GNAT User's Guide for Native Platforms , November 18, 2015AdaCore
Copyright © 2008-2016, Free Software Foundation
`GNAT, The GNU Ada Development Environment'
GCC version 6.4.0
AdaCore
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, with the Front-Cover Texts being "GNAT User's Guide for Native Platforms", and with no Back-Cover Texts. A copy of the license is included in the section entitled GNU Free Documentation License.
--- The Detailed Node Listing ---
About This Guide
Getting Started with GNAT
The GNAT Compilation Model
Foreign Language Representation
File Naming Topics and Utilities
Handling Arbitrary File Naming Conventions with gnatname
File Name Krunching with gnatkr
Renaming Files with gnatchop
Configuration Pragmas
GNAT and Libraries
General Ada Libraries
Stand-alone Ada Libraries
Conditional Compilation
Modeling Conditional Compilation in Ada
Preprocessing with gnatprep
Mixed Language Programming
Building Mixed Ada and C++ Programs
Generating Ada Bindings for C and C++ headers
Generating C Headers for Ada Specifications
GNAT and Other Compilation Models
Using GNAT Files with External Tools
Building Executable Programs with GNAT
Building with gnatmake
Compiling with gcc
Compiler Switches
Binding with gnatbind
Switches for gnatbind
Linking with gnatlink
Using the GNU make Utility
GNAT Project Manager
Building With Projects
Organizing Projects into Subsystems
Library Projects
Project Extension
Aggregate Projects
Aggregate Library Projects
Project File Reference
Attributes
Tools Supporting Project Files
gnatmake and Project Files
GNAT Utility Programs
The File Cleanup Utility gnatclean
The GNAT Library Browser gnatls
The Cross-Referencing Tools gnatxref and gnatfind
Examples of gnatxref Usage
The Ada to HTML Converter gnathtml
GNAT and Program Execution
Running and Debugging Ada Programs
Stack Traceback
Code Coverage and Profiling
Code Coverage of Ada Programs with gcov
Profiling an Ada Program with gprof
Improving Performance
Performance Considerations
Reducing Size of Executables with Unused Subprogram/Data Elimination
Overflow Check Handling in GNAT
Stack Related Facilities
Memory Management Issues
Platform-Specific Information
Run-Time Libraries
Specifying a Run-Time Library
Microsoft Windows Topics
Mixed-Language Programming on Windows
Windows Calling Conventions
Using DLLs with GNAT
Building DLLs with gnatdll
Creating a Spec for Ada DLLs
GNAT and Windows Resources
Debugging a DLL
Windows Specific Add-Ons
Mac OS Topics
Elaboration Order Handling in GNAT
Inline Assembler
Other Asm Functionality
This guide describes the use of GNAT, a compiler and software development toolset for the full Ada programming language. It documents the features of the compiler and tools, and explains how to use them to build Ada applications.
GNAT implements Ada 95, Ada 2005 and Ada 2012, and it may also be invoked in Ada 83 compatibility mode. By default, GNAT assumes Ada 2012, but you can override with a compiler switch (Compiling Different Versions of Ada) to explicitly specify the language version. Throughout this manual, references to 'Ada' without a year suffix apply to all Ada 95/2005/2012 versions of the language.
This guide contains the following chapters:
Appendices cover several additional topics:
This guide assumes a basic familiarity with the Ada 95 language, as described in the International Standard ANSI/ISO/IEC-8652:1995, January 1995. It does not require knowledge of the features introduced by Ada 2005 or Ada 2012. Reference manuals for Ada 95, Ada 2005, and Ada 2012 are included in the GNAT documentation package.
For further information about Ada and related tools, please refer to the following documents:
In early 2015 the GNAT manuals were transitioned to the reStructuredText (rst) / Sphinx documentation generator technology. During that process the GNAT User's Guide was reorganized so that related topics would be described together in the same chapter or appendix. Here's a summary of the major changes realized in the new document structure.
Following are examples of the typographical and graphic conventions used in this guide:
File names
and then shown this way.
$ character followed by a space.
parent-dir/subdir/myfile.adb.
If you are using GNAT on a Windows platform, please note that
the '\' character should be used instead.
This chapter describes how to use GNAT's command line interface to build executable Ada programs. On most platforms a visually oriented Integrated Development Environment is also available, the GNAT Programming Studio (GPS). GPS offers a graphical "look and feel", support for development in other programming languages, comprehensive browsing features, and many other capabilities. For information on GPS please refer to Using the GNAT Programming Studio.
Three steps are needed to create an executable file from an Ada source file:
All three steps are most commonly handled by using the `gnatmake' utility program that, given the name of the main program, automatically performs the necessary compilation, binding and linking steps.
Any text editor may be used to prepare an Ada program. (If Emacs is used, the optional Ada mode may be helpful in laying out the program.) The program text is a normal text file. We will assume in our initial example that you have used your editor to prepare the following standard format text file:
with Ada.Text_IO; use Ada.Text_IO;
procedure Hello is
begin
Put_Line ("Hello WORLD!");
end Hello;
This file should be named hello.adb.
With the normal default file naming conventions, GNAT requires
that each file
contain a single compilation unit whose file name is the
unit name,
with periods replaced by hyphens; the
extension is ads for a
spec and adb for a body.
You can override this default file naming convention by use of the
special pragma Source_File_Name (for further information please
see Using Other File Names).
Alternatively, if you want to rename your files according to this default
convention, which is probably more convenient if you will be using GNAT
for all your compilations, then the gnatchop utility
can be used to generate correctly-named source files
(see Renaming Files with gnatchop).
You can compile the program using the following command ($ is used as the command prompt in the examples in this document):
$ gcc -c hello.adb
`gcc' is the command used to run the compiler. This compiler is
capable of compiling programs in several languages, including Ada and
C. It assumes that you have given it an Ada program if the file extension is
either .ads or .adb, and it will then call
the GNAT compiler to compile the specified file.
The -c switch is required. It tells `gcc' to only do a
compilation. (For C programs, `gcc' can also do linking, but this
capability is not used directly for Ada programs, so the -c
switch must always be present.)
This compile command generates a file
hello.o, which is the object
file corresponding to your Ada program. It also generates
an 'Ada Library Information' file hello.ali,
which contains additional information used to check
that an Ada program is consistent.
To build an executable file,
use gnatbind to bind the program
and `gnatlink' to link it. The
argument to both gnatbind and `gnatlink' is the name of the
ALI file, but the default extension of .ali can
be omitted. This means that in the most common case, the argument
is simply the name of the main program:
$ gnatbind hello
$ gnatlink hello
A simpler method of carrying out these steps is to use `gnatmake', a master program that invokes all the required compilation, binding and linking tools in the correct order. In particular, `gnatmake' automatically recompiles any sources that have been modified since they were last compiled, or sources that depend on such modified sources, so that 'version skew' is avoided.
$ gnatmake hello.adb
The result is an executable program called hello, which can be
run by entering:
$ hello
assuming that the current directory is on the search path for executable programs.
and, if all has gone well, you will see:
Hello WORLD!
appear in response to this command.
Consider a slightly more complicated example that has three files: a main program, and the spec and body of a package:
package Greetings is
procedure Hello;
procedure Goodbye;
end Greetings;
with Ada.Text_IO; use Ada.Text_IO;
package body Greetings is
procedure Hello is
begin
Put_Line ("Hello WORLD!");
end Hello;
procedure Goodbye is
begin
Put_Line ("Goodbye WORLD!");
end Goodbye;
end Greetings;
with Greetings;
procedure Gmain is
begin
Greetings.Hello;
Greetings.Goodbye;
end Gmain;
Following the one-unit-per-file rule, place this program in the following three separate files:
To build an executable version of this program, we could use four separate steps to compile, bind, and link the program, as follows:
$ gcc -c gmain.adb
$ gcc -c greetings.adb
$ gnatbind gmain
$ gnatlink gmain
Note that there is no required order of compilation when using GNAT.
In particular it is perfectly fine to compile the main program first.
Also, it is not necessary to compile package specs in the case where
there is an accompanying body; you only need to compile the body. If you want
to submit these files to the compiler for semantic checking and not code
generation, then use the -gnatc switch:
$ gcc -c greetings.ads -gnatc
Although the compilation can be done in separate steps as in the above example, in practice it is almost always more convenient to use the `gnatmake' tool. All you need to know in this case is the name of the main program's source file. The effect of the above four commands can be achieved with a single one:
$ gnatmake gmain.adb
In the next section we discuss the advantages of using `gnatmake' in more detail.
If you work on a program by compiling single components at a time using
`gcc', you typically keep track of the units you modify. In order to
build a consistent system, you compile not only these units, but also any
units that depend on the units you have modified.
For example, in the preceding case,
if you edit gmain.adb, you only need to recompile that file. But if
you edit greetings.ads, you must recompile both
greetings.adb and gmain.adb, because both files contain
units that depend on greetings.ads.
`gnatbind' will warn you if you forget one of these compilation steps, so that it is impossible to generate an inconsistent program as a result of forgetting to do a compilation. Nevertheless it is tedious and error-prone to keep track of dependencies among units. One approach to handle the dependency-bookkeeping is to use a makefile. However, makefiles present maintenance problems of their own: if the dependencies change as you change the program, you must make sure that the makefile is kept up-to-date manually, which is also an error-prone process.
The `gnatmake' utility takes care of these details automatically. Invoke it using either one of the following forms:
$ gnatmake gmain.adb
$ gnatmake gmain
The argument is the name of the file containing the main program;
you may omit the extension. `gnatmake'
examines the environment, automatically recompiles any files that need
recompiling, and binds and links the resulting set of object files,
generating the executable file, gmain.
In a large program, it
can be extremely helpful to use `gnatmake', because working out by hand
what needs to be recompiled can be difficult.
Note that `gnatmake' takes into account all the Ada rules that establish dependencies among units. These include dependencies that result from inlining subprogram bodies, and from generic instantiation. Unlike some other Ada make tools, `gnatmake' does not rely on the dependencies that were found by the compiler on a previous compilation, which may possibly be wrong when sources change. `gnatmake' determines the exact set of dependencies from scratch each time it is run.
This chapter describes the compilation model used by GNAT. Although similar to that used by other languages such as C and C++, this model is substantially different from the traditional Ada compilation models, which are based on a centralized program library. The chapter covers the following material:
Ada source programs are represented in standard text files, using Latin-1 coding. Latin-1 is an 8-bit code that includes the familiar 7-bit ASCII set, plus additional characters used for representing foreign languages (see Foreign Language Representation for support of non-USA character sets). The format effector characters are represented using their standard ASCII encodings, as follows:
Character Effect
Code
VTVertical tab
16#0B#
HTHorizontal tab
16#09#
CRCarriage return
16#0D#
LFLine feed
16#0A#
FFForm feed
16#0C#
Source files are in standard text file format. In addition, GNAT will recognize a wide variety of stream formats, in which the end of physical lines is marked by any of the following sequences: LF, CR, CR-LF, or LF-CR. This is useful in accommodating files that are imported from other operating systems.
The end of a source file is normally represented by the physical end of
file. However, the control character 16#1A# (SUB) is also
recognized as signalling the end of the source file. Again, this is
provided for compatibility with other operating systems where this
code is used to represent the end of file.
Each file contains a single Ada compilation unit, including any pragmas associated with the unit. For example, this means you must place a package declaration (a package spec) and the corresponding body in separate files. An Ada compilation (which is a sequence of compilation units) is represented using a sequence of files. Similarly, you will place each subunit or child unit in a separate file.
GNAT supports the standard character sets defined in Ada as well as several other non-standard character sets for use in localized versions of the compiler (Character Set Control).
The basic character set is Latin-1. This character set is defined by ISO standard 8859, part 1. The lower half (character codes 16#00# ... 16#7F#) is identical to standard ASCII coding, but the upper half is used to represent additional characters. These include extended letters used by European languages, such as French accents, the vowels with umlauts used in German, and the extra letter A-ring used in Swedish.
For a complete list of Latin-1 codes and their encodings, see the source
file of library unit Ada.Characters.Latin_1 in file
a-chlat1.ads.
You may use any of these extended characters freely in character or
string literals. In addition, the extended characters that represent
letters can be used in identifiers.
GNAT also supports several other 8-bit coding schemes:
For precise data on the encodings permitted, and the uppercase and lowercase
equivalences that are recognized, see the file csets.adb in
the GNAT compiler sources. You will need to obtain a full source release
of GNAT to obtain this file.
GNAT allows wide character codes to appear in character and string literals, and also optionally in identifiers, by means of the following possible encoding schemes:
ESC a b c d
where a, b, c, d are the four hexadecimal
characters (using uppercase letters) of the wide character code. For
example, ESC A345 is used to represent the wide character with code
16#A345#.
This scheme is compatible with use of the full Wide_Character set.
16#0000#-16#007f#: 2#0`xxxxxxx`#
16#0080#-16#07ff#: 2#110`xxxxx`# 2#10`xxxxxx`#
16#0800#-16#ffff#: 2#1110`xxxx`# 2#10`xxxxxx`# 2#10`xxxxxx`#
where the xxx bits correspond to the left-padded bits of the
16-bit character value. Note that all lower half ASCII characters
are represented as ASCII bytes and all upper half characters and
other wide characters are represented as sequences of upper-half
(The full UTF-8 scheme allows for encoding 31-bit characters as
6-byte sequences, and in the following section on wide wide
characters, the use of these sequences is documented).
[ " a b c d " ]
where a, b, c, d are the four hexadecimal characters (using uppercase letters) of the wide character code. For example, ['A345'] is used to represent the wide character with code 16#A345#. It is also possible (though not required) to use the Brackets coding for upper half characters. For example, the code 16#A3# can be represented as ['A3'].
This scheme is compatible with use of the full Wide_Character set, and is also the method used for wide character encoding in some standard ACATS (Ada Conformity Assessment Test Suite) test suite distributions.
Note: Some of these coding schemes do not permit the full use of the Ada character set. For example, neither Shift JIS nor EUC allow the use of the upper half of the Latin-1 set. |
GNAT allows wide wide character codes to appear in character and string literals, and also optionally in identifiers, by means of the following possible encoding schemes:
16#01_0000#-16#10_FFFF#: 11110xxx 10xxxxxx 10xxxxxx
10xxxxxx
16#0020_0000#-16#03FF_FFFF#: 111110xx 10xxxxxx 10xxxxxx
10xxxxxx 10xxxxxx
16#0400_0000#-16#7FFF_FFFF#: 1111110x 10xxxxxx 10xxxxxx
10xxxxxx 10xxxxxx 10xxxxxx
where the xxx bits correspond to the left-padded bits of the
32-bit character value.
[ " a b c d e f " ]
[ " a b c d e f g h " ]
where a-h are the six or eight hexadecimal characters (using uppercase letters) of the wide wide character code. For example, ["1F4567"] is used to represent the wide wide character with code 16#001F_4567#.
This scheme is compatible with use of the full Wide_Wide_Character set, and is also the method used for wide wide character encoding in some standard ACATS (Ada Conformity Assessment Test Suite) test suite distributions.
GNAT has a default file naming scheme and also provides the user with a high degree of control over how the names and extensions of the source files correspond to the Ada compilation units that they contain.
The default file name is determined by the name of the unit that the file contains. The name is formed by taking the full expanded name of the unit and replacing the separating dots with hyphens and using lowercase for all letters.
An exception arises if the file name generated by the above rules starts with one of the characters a, g, i, or s, and the second character is a minus. In this case, the character tilde is used in place of the minus. The reason for this special rule is to avoid clashes with the standard names for child units of the packages System, Ada, Interfaces, and GNAT, which use the prefixes s-, a-, i-, and g-, respectively.
The file extension is .ads for a spec and
.adb for a body. The following table shows some
examples of these rules.
Source File Ada Compilation Unit
main.adsMain (spec)
main.adbMain (body)
arith_functions.adsArith_Functions (package spec)
arith_functions.adbArith_Functions (package body)
func-spec.adsFunc.Spec (child package spec)
func-spec.adbFunc.Spec (child package body)
main-sub.adbSub (subunit of Main)
a~bad.adbA.Bad (child package body)
Following these rules can result in excessively long file names if corresponding unit names are long (for example, if child units or subunits are heavily nested). An option is available to shorten such long file names (called file name 'krunching'). This may be particularly useful when programs being developed with GNAT are to be used on operating systems with limited file name lengths. Using gnatkr.
Of course, no file shortening algorithm can guarantee uniqueness over all possible unit names; if file name krunching is used, it is your responsibility to ensure no name clashes occur. Alternatively you can specify the exact file names that you want used, as described in the next section. Finally, if your Ada programs are migrating from a compiler with a different naming convention, you can use the gnatchop utility to produce source files that follow the GNAT naming conventions. (For details see Renaming Files with gnatchop.)
Note: in the case of Windows or Mac OS operating systems, case is not
significant. So for example on Windows if the canonical name is
main-sub.adb, you can use the file name Main-Sub.adb instead.
However, case is significant for other operating systems, so for example,
if you want to use other than canonically cased file names on a Unix system,
you need to follow the procedures described in the next section.
In the previous section, we have described the default rules used by GNAT to determine the file name in which a given unit resides. It is often convenient to follow these default rules, and if you follow them, the compiler knows without being explicitly told where to find all the files it needs.
However, in some cases, particularly when a program is imported from another Ada compiler environment, it may be more convenient for the programmer to specify which file names contain which units. GNAT allows arbitrary file names to be used by means of the Source_File_Name pragma. The form of this pragma is as shown in the following examples:
pragma Source_File_Name (My_Utilities.Stacks,
Spec_File_Name => "myutilst_a.ada");
pragma Source_File_name (My_Utilities.Stacks,
Body_File_Name => "myutilst.ada");
As shown in this example, the first argument for the pragma is the unit name (in this example a child unit). The second argument has the form of a named association. The identifier indicates whether the file name is for a spec or a body; the file name itself is given by a string literal.
The source file name pragma is a configuration pragma, which means that
normally it will be placed in the gnat.adc
file used to hold configuration
pragmas that apply to a complete compilation environment.
For more details on how the gnat.adc file is created and used
see Handling of Configuration Pragmas.
GNAT allows completely arbitrary file names to be specified using the
source file name pragma. However, if the file name specified has an
extension other than .ads or .adb it is necessary to use
a special syntax when compiling the file. The name in this case must be
preceded by the special sequence `-x' followed by a space and the name
of the language, here ada, as in:
$ gcc -c -x ada peculiar_file_name.sim
gnatmake handles non-standard file names in the usual manner (the non-standard file name for the main program is simply used as the argument to gnatmake). Note that if the extension is also non-standard, then it must be included in the gnatmake command, it may not be omitted.
The previous section described the use of the Source_File_Name
pragma to allow arbitrary names to be assigned to individual source files.
However, this approach requires one pragma for each file, and especially in
large systems can result in very long gnat.adc files, and also create
a maintenance problem.
GNAT also provides a facility for specifying systematic file naming schemes other than the standard default naming scheme previously described. An alternative scheme for naming is specified by the use of Source_File_Name pragmas having the following format:
pragma Source_File_Name (
Spec_File_Name => FILE_NAME_PATTERN
[ , Casing => CASING_SPEC]
[ , Dot_Replacement => STRING_LITERAL ] );
pragma Source_File_Name (
Body_File_Name => FILE_NAME_PATTERN
[ , Casing => CASING_SPEC ]
[ , Dot_Replacement => STRING_LITERAL ] ) ;
pragma Source_File_Name (
Subunit_File_Name => FILE_NAME_PATTERN
[ , Casing => CASING_SPEC ]
[ , Dot_Replacement => STRING_LITERAL ] ) ;
FILE_NAME_PATTERN ::= STRING_LITERAL
CASING_SPEC ::= Lowercase | Uppercase | Mixedcase
The FILE_NAME_PATTERN string shows how the file name is constructed. It contains a single asterisk character, and the unit name is substituted systematically for this asterisk. The optional parameter Casing indicates whether the unit name is to be all upper-case letters, all lower-case letters, or mixed-case. If no Casing parameter is used, then the default is all lower-case.
The optional Dot_Replacement string is used to replace any periods that occur in subunit or child unit names. If no Dot_Replacement argument is used then separating dots appear unchanged in the resulting file name. Although the above syntax indicates that the Casing argument must appear before the Dot_Replacement argument, but it is also permissible to write these arguments in the opposite order.
As indicated, it is possible to specify different naming schemes for bodies, specs, and subunits. Quite often the rule for subunits is the same as the rule for bodies, in which case, there is no need to give a separate Subunit_File_Name rule, and in this case the Body_File_name rule is used for subunits as well.
The separate rule for subunits can also be used to implement the rather unusual case of a compilation environment (e.g., a single directory) which contains a subunit and a child unit with the same unit name. Although both units cannot appear in the same partition, the Ada Reference Manual allows (but does not require) the possibility of the two units coexisting in the same environment.
The file name translation works in the following steps:
As an example of the use of this mechanism, consider a commonly used scheme
in which file names are all lower case, with separating periods copied
unchanged to the resulting file name, and specs end with .1.ada, and
bodies end with .2.ada. GNAT will follow this scheme if the following
two pragmas appear:
pragma Source_File_Name
(Spec_File_Name => ".1.ada");
pragma Source_File_Name
(Body_File_Name => ".2.ada");
The default GNAT scheme is actually implemented by providing the following default pragmas internally:
pragma Source_File_Name
(Spec_File_Name => ".ads", Dot_Replacement => "-");
pragma Source_File_Name
(Body_File_Name => ".adb", Dot_Replacement => "-");
Our final example implements a scheme typically used with one of the
Ada 83 compilers, where the separator character for subunits was '__'
(two underscores), specs were identified by adding _.ADA, bodies
by adding .ADA, and subunits by
adding .SEP. All file names were
upper case. Child units were not present of course since this was an
Ada 83 compiler, but it seems reasonable to extend this scheme to use
the same double underscore separator for child units.
pragma Source_File_Name
(Spec_File_Name => "_.ADA",
Dot_Replacement => "__",
Casing = Uppercase);
pragma Source_File_Name
(Body_File_Name => ".ADA",
Dot_Replacement => "__",
Casing = Uppercase);
pragma Source_File_Name
(Subunit_File_Name => ".SEP",
Dot_Replacement => "__",
Casing = Uppercase);
The GNAT compiler must be able to know the source file name of a compilation unit. When using the standard GNAT default file naming conventions (.ads for specs, .adb for bodies), the GNAT compiler does not need additional information.
When the source file names do not follow the standard GNAT default file naming conventions, the GNAT compiler must be given additional information through a configuration pragmas file (Configuration Pragmas) or a project file. When the non-standard file naming conventions are well-defined, a small number of pragmas Source_File_Name specifying a naming pattern (Alternative File Naming Schemes) may be sufficient. However, if the file naming conventions are irregular or arbitrary, a number of pragma Source_File_Name for individual compilation units must be defined. To help maintain the correspondence between compilation unit names and source file names within the compiler, GNAT provides a tool gnatname to generate the required pragmas for a set of files.
The usual form of the gnatname command is:
$ gnatname [`switches`] `naming_pattern` [`naming_patterns`]
[--and [`switches`] `naming_pattern` [`naming_patterns`]]
All of the arguments are optional. If invoked without any argument, gnatname will display its usage.
When used with at least one naming pattern, gnatname will attempt to find all the compilation units in files that follow at least one of the naming patterns. To find these compilation units, gnatname will use the GNAT compiler in syntax-check-only mode on all regular files.
One or several Naming Patterns may be given as arguments to gnatname. Each Naming Pattern is enclosed between double quotes (or single quotes on Windows). A Naming Pattern is a regular expression similar to the wildcard patterns used in file names by the Unix shells or the DOS prompt.
gnatname may be called with several sections of directories/patterns. Sections are separated by switch –and. In each section, there must be at least one pattern. If no directory is specified in a section, the current directory (or the project directory is -P is used) is implied. The options other that the directory switches and the patterns apply globally even if they are in different sections.
Examples of Naming Patterns are:
"*.[12].ada"
"*.ad[sb]*"
"body_*" "spec_*"
For a more complete description of the syntax of Naming Patterns,
see the second kind of regular expressions described in g-regexp.ads
(the 'Glob' regular expressions).
When invoked with no switch -P, gnatname will create a
configuration pragmas file gnat.adc in the current working directory,
with pragmas Source_File_Name for each file that contains a valid Ada
unit.
Switches for gnatname must precede any specified Naming Pattern.
You may specify any of the following switches to gnatname:
--version--help--subdirs=`dir'--no-backup--and-c`filename'filename (instead of the default
gnat.adc).
There may be zero, one or more space between `-c' and
filename.
filename may include directory information. filename must be
writable. There may be only one switch `-c'.
When a switch `-c' is
specified, no switch `-P' may be specified (see below).
-d`dir'dir. There may be zero, one or more
spaces between `-d' and dir.
dir may end with /**, that is it may be of the form
root_dir/**. In this case, the directory root_dir and all of its
subdirectories, recursively, have to be searched for sources.
When a switch `-d'
is specified, the current working directory will not be searched for source
files, unless it is explicitly specified with a `-d'
or `-D' switch.
Several switches `-d' may be specified.
If dir is a relative path, it is relative to the directory of
the configuration pragmas file specified with switch
`-c',
or to the directory of the project file specified with switch
`-P' or,
if neither switch `-c'
nor switch `-P' are specified, it is relative to the
current working directory. The directory
specified with switch `-d' must exist and be readable.
-D`filename'filename.
There may be zero, one or more spaces between `-D'
and filename.
filename must be an existing, readable text file.
Each nonempty line in filename must be a directory.
Specifying switch `-D' is equivalent to specifying as many
switches `-d' as there are nonempty lines in
file.
-eL-f`pattern'gnatname -Pprj -f"*.c" "*.ada"
will look for Ada units in all files with the .ada extension,
and will add to the list of file for project prj.gpr the C files
with extension .c.
-hstdout.
-P`proj'proj. There may be zero, one or more space
between `-P' and proj. proj may include directory
information. proj must be writable.
There may be only one switch `-P'.
When a switch `-P' is specified,
no switch `-c' may be specified.
On all platforms, except on VMS, when gnatname is invoked for an
existing project file <proj>.gpr, a backup copy of the project file is created
in the project directory with file name <proj>.gpr.saved_x. 'x' is the first
non negative number that makes this backup copy a new file.
-vstdout.
This includes name of the file written, the name of the directories to search
and, for each file in those directories whose name matches at least one of
the Naming Patterns, an indication of whether the file contains a unit,
and if so the name of the unit.
-v -v-x`pattern'gnatname -x "*_nt.ada" "*.ada"
will look for Ada units in all files with the .ada extension,
except those whose names end with _nt.ada.
$ gnatname -c /home/me/names.adc -d sources "[a-z]*.ada*"
In this example, the directory /home/me must already exist
and be writable. In addition, the directory
/home/me/sources (specified by
`-d sources') must exist and be readable.
Note the optional spaces after `-c' and `-d'.
$ gnatname -P/home/me/proj -x "*_nt_body.ada"
-dsources -dsources/plus -Dcommon_dirs.txt "body_*" "spec_*"
Note that several switches `-d' may be used, even in conjunction with one or several switches `-D'. Several Naming Patterns and one excluded pattern are used in this example.
This chapter discusses the method used by the compiler to shorten the default file names chosen for Ada units so that they do not exceed the maximum length permitted. It also describes the gnatkr utility that can be used to determine the result of applying this shortening.
The default file naming rule in GNAT is that the file name must be derived from the unit name. The exact default rule is as follows:
a, g, s, or i,
then replace the dot by the character
~ (tilde)
instead of a minus.
The reason for this exception is to avoid clashes
with the standard names for children of System, Ada, Interfaces,
and GNAT, which use the prefixes
s-, a-, i-, and g-,
respectively.
The -gnatk`nn'
switch of the compiler activates a 'krunching'
circuit that limits file names to nn characters (where nn is a decimal
integer).
The gnatkr utility can be used to determine the krunched name for a given file, when krunched to a specified maximum length.
The gnatkr command has the form:
$ gnatkr `name` [`length`]
name is the uncrunched file name, derived from the name of the unit
in the standard manner described in the previous section (i.e., in particular
all dots are replaced by hyphens). The file name may or may not have an
extension (defined as a suffix of the form period followed by arbitrary
characters other than period). If an extension is present then it will
be preserved in the output. For example, when krunching hellofile.ads
to eight characters, the result will be hellofil.ads.
Note: for compatibility with previous versions of gnatkr dots may
appear in the name instead of hyphens, but the last dot will always be
taken as the start of an extension. So if gnatkr is given an argument
such as Hello.World.adb it will be treated exactly as if the first
period had been a hyphen, and for example krunching to eight characters
gives the result hellworl.adb.
Note that the result is always all lower case. Characters of the other case are folded as required.
length represents the length of the krunched name. The default when no argument is given is 8 characters. A length of zero stands for unlimited, in other words do not chop except for system files where the implied crunching length is always eight characters.
The output is the krunched name. The output has an extension only if the original argument was a file name with an extension.
The initial file name is determined by the name of the unit that the file
contains. The name is formed by taking the full expanded name of the
unit and replacing the separating dots with hyphens and
using lowercase
for all letters, except that a hyphen in the second character position is
replaced by a tilde if the first character is
a, i, g, or s.
The extension is .ads for a
spec and .adb for a body.
Krunching does not affect the extension, but the file name is shortened to
the specified length by following these rules:
As an example, consider the krunching of our-strings-wide_fixed.adb
to fit the name into 8 characters as required by some operating systems:
our-strings-wide_fixed 22
our strings wide fixed 19
our string wide fixed 18
our strin wide fixed 17
our stri wide fixed 16
our stri wide fixe 15
our str wide fixe 14
our str wid fixe 13
our str wid fix 12
ou str wid fix 11
ou st wid fix 10
ou st wi fix 9
ou st wi fi 8
Final file name: oustwifi.adb
| Prefix |
Replacement |
ada-
|
|
gnat-
|
|
interfac es-
|
|
system-
|
|
These system files have a hyphen in the second character position. That is why normal user files replace such a character with a tilde, to avoid confusion with system file names.
As an example of this special rule, consider
ada-strings-wide_fixed.adb, which gets krunched as follows:
ada-strings-wide_fixed 22
a- strings wide fixed 18
a- string wide fixed 17
a- strin wide fixed 16
a- stri wide fixed 15
a- stri wide fixe 14
a- str wide fixe 13
a- str wid fixe 12
a- str wid fix 11
a- st wid fix 10
a- st wi fix 9
a- st wi fi 8
Final file name: a-stwifi.adb
Of course no file shortening algorithm can guarantee uniqueness over all possible unit names, and if file name krunching is used then it is your responsibility to ensure that no name clashes occur. The utility program gnatkr is supplied for conveniently determining the krunched name of a file.
$ gnatkr very_long_unit_name.ads --> velounna.ads
$ gnatkr grandparent-parent-child.ads --> grparchi.ads
$ gnatkr Grandparent.Parent.Child.ads --> grparchi.ads
$ gnatkr grandparent-parent-child --> grparchi
$ gnatkr very_long_unit_name.ads/count=6 --> vlunna.ads
$ gnatkr very_long_unit_name.ads/count=0 --> very_long_unit_name.ads
This chapter discusses how to handle files with multiple units by using the gnatchop utility. This utility is also useful in renaming files to meet the standard GNAT default file naming conventions.
The basic compilation model of GNAT requires that a file submitted to the compiler have only one unit and there be a strict correspondence between the file name and the unit name.
The gnatchop utility allows both of these rules to be relaxed, allowing GNAT to process files which contain multiple compilation units and files with arbitrary file names. gnatchop reads the specified file and generates one or more output files, containing one unit per file. The unit and the file name correspond, as required by GNAT.
If you want to permanently restructure a set of 'foreign' files so that they match the GNAT rules, and do the remaining development using the GNAT structure, you can simply use `gnatchop' once, generate the new set of files and work with them from that point on.
Alternatively, if you want to keep your files in the 'foreign' format, perhaps to maintain compatibility with some other Ada compilation system, you can set up a procedure where you use `gnatchop' each time you compile, regarding the source files that it writes as temporary files that you throw away.
Note that if your file containing multiple units starts with a byte order mark (BOM) specifying UTF-8 encoding, then the files generated by gnatchop will each start with a copy of this BOM, meaning that they can be compiled automatically in UTF-8 mode without needing to specify an explicit encoding.
The basic function of gnatchop is to take a file with multiple units and split it into separate files. The boundary between files is reasonably clear, except for the issue of comments and pragmas. In default mode, the rule is that any pragmas between units belong to the previous unit, except that configuration pragmas always belong to the following unit. Any comments belong to the following unit. These rules almost always result in the right choice of the split point without needing to mark it explicitly and most users will find this default to be what they want. In this default mode it is incorrect to submit a file containing only configuration pragmas, or one that ends in configuration pragmas, to gnatchop.
However, using a special option to activate 'compilation mode', gnatchop can perform another function, which is to provide exactly the semantics required by the RM for handling of configuration pragmas in a compilation. In the absence of configuration pragmas (at the main file level), this option has no effect, but it causes such configuration pragmas to be handled in a quite different manner.
First, in compilation mode, if gnatchop is given a file that consists of
only configuration pragmas, then this file is appended to the
gnat.adc file in the current directory. This behavior provides
the required behavior described in the RM for the actions to be taken
on submitting such a file to the compiler, namely that these pragmas
should apply to all subsequent compilations in the same compilation
environment. Using GNAT, the current directory, possibly containing a
gnat.adc file is the representation
of a compilation environment. For more information on the
gnat.adc file, see Handling of Configuration Pragmas.
Second, in compilation mode, if gnatchop is given a file that starts with configuration pragmas, and contains one or more units, then these configuration pragmas are prepended to each of the chopped files. This behavior provides the required behavior described in the RM for the actions to be taken on compiling such a file, namely that the pragmas apply to all units in the compilation, but not to subsequently compiled units.
Finally, if configuration pragmas appear between units, they are appended to the previous unit. This results in the previous unit being illegal, since the compiler does not accept configuration pragmas that follow a unit. This provides the required RM behavior that forbids configuration pragmas other than those preceding the first compilation unit of a compilation.
For most purposes, gnatchop will be used in default mode. The compilation mode described above is used only if you need exactly accurate behavior with respect to compilations, and you have files that contain multiple units and configuration pragmas. In this circumstance the use of gnatchop with the compilation mode switch provides the required behavior, and is for example the mode in which GNAT processes the ACVC tests.
The gnatchop command has the form:
$ gnatchop switches file_name [file_name ...]
[directory]
The only required argument is the file name of the file to be chopped. There are no restrictions on the form of this file name. The file itself contains one or more Ada units, in normal GNAT format, concatenated together. As shown, more than one file may be presented to be chopped.
When run in default mode, gnatchop generates one output file in the current directory for each unit in each of the files.
directory, if specified, gives the name of the directory to which the output files will be written. If it is not specified, all files are written to the current directory.
For example, given a
file called hellofiles containing
procedure Hello;
with Ada.Text_IO; use Ada.Text_IO;
procedure Hello is
begin
Put_Line ("Hello");
end Hello;
the command
$ gnatchop hellofiles
generates two files in the current directory, one called
hello.ads containing the single line that is the procedure spec,
and the other called hello.adb containing the remaining text. The
original file is not affected. The generated files can be compiled in
the normal manner.
When gnatchop is invoked on a file that is empty or that contains only empty lines and/or comments, gnatchop will not fail, but will not produce any new sources.
For example, given a
file called toto.txt containing
-- Just a comment
the command
$ gnatchop toto.txt
will not produce any new file and will result in the following warnings:
toto.txt:1:01: warning: empty file, contains no compilation units
no compilation units found
no source files written
`gnatchop' recognizes the following switches:
--version--help-c-gnat`xxx'-h-k`mm'-p-q-rIf the original file to be chopped itself contains a Source_Reference pragma referencing a third file, then gnatchop respects this pragma, and the generated Source_Reference pragmas in the chopped file refer to the original file, with appropriate line numbers. This is particularly useful when gnatchop is used in conjunction with gnatprep to compile files that contain preprocessing statements and multiple units.
-v-w--GCC=`xxxx'$ gnatchop -w hello_s.ada prerelease/files
Chops the source file hello_s.ada. The output files will be
placed in the directory prerelease/files,
overwriting any
files with matching names in that directory (no files in the current
directory are modified).
$ gnatchop archive
Chops the source file archive
into the current directory. One
useful application of gnatchop is in sending sets of sources
around, for example in email messages. The required sources are simply
concatenated (for example, using a Unix cat
command), and then
`gnatchop' is used at the other end to reconstitute the original
file names.
$ gnatchop file1 file2 file3 direc
Chops all units in files file1, file2, file3, placing
the resulting files in the directory direc. Note that if any units
occur more than once anywhere within this set of files, an error message
is generated, and no files are written. To override this check, use the
`-w' switch,
in which case the last occurrence in the last file will
be the one that is output, and earlier duplicate occurrences for a given
unit will be skipped.
Configuration pragmas include those pragmas described as such in the Ada Reference Manual, as well as implementation-dependent pragmas that are configuration pragmas. See the Implementation_Defined_Pragmas chapter in the GNAT_Reference_Manual for details on these additional GNAT-specific configuration pragmas. Most notably, the pragma Source_File_Name, which allows specifying non-default names for source files, is a configuration pragma. The following is a complete list of configuration pragmas recognized by GNAT:
Ada_83
Ada_95
Ada_05
Ada_2005
Ada_12
Ada_2012
Allow_Integer_Address
Annotate
Assertion_Policy
Assume_No_Invalid_Values
C_Pass_By_Copy
Check_Name
Check_Policy
Compile_Time_Error
Compile_Time_Warning
Compiler_Unit
Component_Alignment
Convention_Identifier
Debug_Policy
Detect_Blocking
Default_Storage_Pool
Discard_Names
Elaboration_Checks
Eliminate
Extend_System
Extensions_Allowed
External_Name_Casing
Fast_Math
Favor_Top_Level
Float_Representation
Implicit_Packing
Initialize_Scalars
Interrupt_State
License
Locking_Policy
Long_Float
No_Run_Time
No_Strict_Aliasing
Normalize_Scalars
Optimize_Alignment
Persistent_BSS
Polling
Priority_Specific_Dispatching
Profile
Profile_Warnings
Propagate_Exceptions
Queuing_Policy
Ravenscar
Restricted_Run_Time
Restrictions
Restrictions_Warnings
Reviewable
Short_Circuit_And_Or
Source_File_Name
Source_File_Name_Project
SPARK_Mode
Style_Checks
Suppress
Suppress_Exception_Locations
Task_Dispatching_Policy
Universal_Data
Unsuppress
Use_VADS_Size
Validity_Checks
Warnings
Wide_Character_Encoding
Configuration pragmas may either appear at the start of a compilation unit, or they can appear in a configuration pragma file to apply to all compilations performed in a given compilation environment.
GNAT also provides the gnatchop utility to provide an automatic
way to handle configuration pragmas following the semantics for
compilations (that is, files with multiple units), described in the RM.
See Operating gnatchop in Compilation Mode for details.
However, for most purposes, it will be more convenient to edit the
gnat.adc file that contains configuration pragmas directly,
as described in the following section.
In the case of Restrictions pragmas appearing as configuration pragmas in individual compilation units, the exact handling depends on the type of restriction.
Restrictions that require partition-wide consistency (like No_Tasking) are recognized wherever they appear and can be freely inherited, e.g. from a `with'ed unit to the `with'ing unit. This makes sense since the binder will in any case insist on seeing consistent use, so any unit not conforming to any restrictions that are anywhere in the partition will be rejected, and you might as well find that out at compile time rather than at bind time.
For restrictions that do not require partition-wide consistency, e.g. SPARK or No_Implementation_Attributes, in general the restriction applies only to the unit in which the pragma appears, and not to any other units.
The exception is No_Elaboration_Code which always applies to the entire object file from a compilation, i.e. to the body, spec, and all subunits. This restriction can be specified in a configuration pragma file, or it can be on the body and/or the spec (in eithe case it applies to all the relevant units). It can appear on a subunit only if it has previously appeared in the body of spec.
In GNAT a compilation environment is defined by the current
directory at the time that a compile command is given. This current
directory is searched for a file whose name is gnat.adc. If
this file is present, it is expected to contain one or more
configuration pragmas that will be applied to the current compilation.
However, if the switch `-gnatA' is used, gnat.adc is not
considered. When taken into account, gnat.adc is added to the
dependencies, so that if gnat.adc is modified later, an invocation of
`gnatmake' will recompile the source.
Configuration pragmas may be entered into the gnat.adc file
either by running gnatchop on a source file that consists only of
configuration pragmas, or more conveniently by direct editing of the
gnat.adc file, which is a standard format source file.
Besides gnat.adc, additional files containing configuration
pragmas may be applied to the current compilation using the switch
-gnatec=`path' where path must designate an existing file that
contains only configuration pragmas. These configuration pragmas are
in addition to those found in gnat.adc (provided gnat.adc
is present and switch `-gnatA' is not used).
It is allowable to specify several switches `-gnatec=', all of which will be taken into account.
Files containing configuration pragmas specified with switches
`-gnatec=' are added to the dependencies, unless they are
temporary files. A file is considered temporary if its name ends in
.tmp or .TMP. Certain tools follow this naming
convention because they pass information to `gcc' via
temporary files that are immediately deleted; it doesn't make sense to
depend on a file that no longer exists. Such tools include
`gprbuild', `gnatmake', and `gnatcheck'.
If you are using project file, a separate mechanism is provided using project attributes, see Specifying Configuration Pragmas for more details.
An Ada program consists of a set of source files, and the first step in compiling the program is to generate the corresponding object files. These are generated by compiling a subset of these source files. The files you need to compile are the following:
The preceding rules describe the set of files that must be compiled to
generate the object files for a program. Each object file has the same
name as the corresponding source file, except that the extension is
.o as usual.
You may wish to compile other files for the purpose of checking their syntactic and semantic correctness. For example, in the case where a package has a separate spec and body, you would not normally compile the spec. However, it is convenient in practice to compile the spec to make sure it is error-free before compiling clients of this spec, because such compilations will fail if there is an error in the spec.
GNAT provides an option for compiling such files purely for the purposes of checking correctness; such compilations are not required as part of the process of building a program. To compile a file in this checking mode, use the `-gnatc' switch.
A given object file clearly depends on the source file which is compiled to produce it. Here we are using "depends" in the sense of a typical make utility; in other words, an object file depends on a source file if changes to the source file require the object file to be recompiled. In addition to this basic dependency, a given object may depend on additional source files as follows:
The use of `-gnatN' activates inlining optimization that is performed by the front end of the compiler. This inlining does not require that the code generation be optimized. Like `-gnatn', the use of this switch generates additional dependencies.
When using a gcc-based back end (in practice this means using any version of GNAT other than for the JVM, .NET or GNAAMP platforms), then the use of `-gnatN' is deprecated, and the use of `-gnatn' is preferred. Historically front end inlining was more extensive than the gcc back end inlining, but that is no longer the case.
O depends on the proper body of a subunit through
inlining or instantiation, it depends on the parent unit of the subunit.
This means that any modification of the parent unit or one of its subunits
affects the compilation of O.
These rules are applied transitively: if unit A `with's
unit B, whose elaboration calls an inlined procedure in package
C, the object file for unit A will depend on the body of
C, in file c.adb.
The set of dependent files described by these rules includes all the files on which the unit is semantically dependent, as dictated by the Ada language standard. However, it is a superset of what the standard describes, because it includes generic, inline, and subunit dependencies.
An object file must be recreated by recompiling the corresponding source file if any of the source files on which it depends are modified. For example, if the make utility is used to control compilation, the rule for an Ada object file must mention all the source files on which the object file depends, according to the above definition. The determination of the necessary recompilations is done automatically when one uses `gnatmake'.
Each compilation actually generates two output files. The first of these
is the normal object file that has a .o extension. The second is a
text file containing full dependency information. It has the same
name as the source file, but an .ali extension.
This file is known as the Ada Library Information (ALI) file.
The following information is contained in the ALI file.
For a full detailed description of the format of the ALI file,
see the source of the body of unit Lib.Writ, contained in file
lib-writ.adb in the GNAT compiler sources.
When using languages such as C and C++, once the source files have been compiled the only remaining step in building an executable program is linking the object modules together. This means that it is possible to link an inconsistent version of a program, in which two units have included different versions of the same header.
The rules of Ada do not permit such an inconsistent program to be built. For example, if two clients have different versions of the same package, it is illegal to build a program containing these two clients. These rules are enforced by the GNAT binder, which also determines an elaboration order consistent with the Ada rules.
The GNAT binder is run after all the object files for a program have been created. It is given the name of the main program unit, and from this it determines the set of units required by the program, by reading the corresponding ALI files. It generates error messages if the program is inconsistent or if no valid order of elaboration exists.
If no errors are detected, the binder produces a main program, in Ada by
default, that contains calls to the elaboration procedures of those
compilation unit that require them, followed by
a call to the main program. This Ada program is compiled to generate the
object file for the main program. The name of
the Ada file is b~xxx.adb` (with the corresponding spec
b~xxx.ads`) where xxx is the name of the
main program unit.
Finally, the linker is used to build the resulting executable program, using the object from the main program from the bind step as well as the object files for the Ada units of the program.
This chapter describes how to build and use libraries with GNAT, and also shows how to recompile the GNAT run-time library. You should be familiar with the Project Manager facility (GNAT Project Manager) before reading this chapter.
A library is, conceptually, a collection of objects which does not have its own main thread of execution, but rather provides certain services to the applications that use it. A library can be either statically linked with the application, in which case its code is directly included in the application, or, on platforms that support it, be dynamically linked, in which case its code is shared by all applications making use of this library.
GNAT supports both types of libraries. In the static case, the compiled code can be provided in different ways. The simplest approach is to provide directly the set of objects resulting from compilation of the library source files. Alternatively, you can group the objects into an archive using whatever commands are provided by the operating system. For the latter case, the objects are grouped into a shared library.
In the GNAT environment, a library has three types of components:
ALI files (see The Ada Library Information Files), and
A GNAT library may expose all its source files, which is useful for documentation purposes. Alternatively, it may expose only the units needed by an external user to make use of the library. That is to say, the specs reflecting the library services along with all the units needed to compile those specs, which can include generic bodies or any body implementing an inlined routine. In the case of `stand-alone libraries' those exposed units are called `interface units' (Stand-alone Ada Libraries).
All compilation units comprising an application, including those in a library,
need to be elaborated in an order partially defined by Ada's semantics. GNAT
computes the elaboration order from the ALI files and this is why they
constitute a mandatory part of GNAT libraries.
`Stand-alone libraries' are the exception to this rule because a specific
library elaboration routine is produced independently of the application(s)
using the library.
The easiest way to build a library is to use the Project Manager, which supports a special type of project called a `Library Project' (see Library Projects).
A project is considered a library project, when two project-level attributes are defined in it: Library_Name and Library_Dir. In order to control different aspects of library configuration, additional optional project-level attributes can be specified:
The GNAT Project Manager takes full care of the library maintenance task,
including recompilation of the source files for which objects do not exist
or are not up to date, assembly of the library archive, and installation of
the library (i.e., copying associated source, object and ALI files
to the specified location).
Here is a simple library project file:
project My_Lib is
for Source_Dirs use ("src1", "src2");
for Object_Dir use "obj";
for Library_Name use "mylib";
for Library_Dir use "lib";
for Library_Kind use "dynamic";
end My_lib;
and the compilation command to build and install the library:
$ gnatmake -Pmy_lib
It is not entirely trivial to perform manually all the steps required to produce a library. We recommend that you use the GNAT Project Manager for this task. In special cases where this is not desired, the necessary steps are discussed below.
There are various possibilities for compiling the units that make up the library: for example with a Makefile (Using the GNU make Utility) or with a conventional script. For simple libraries, it is also possible to create a dummy main program which depends upon all the packages that comprise the interface of the library. This dummy main program can then be given to `gnatmake', which will ensure that all necessary objects are built.
After this task is accomplished, you should follow the standard procedure of the underlying operating system to produce the static or shared library.
Here is an example of such a dummy program:
with My_Lib.Service1;
with My_Lib.Service2;
with My_Lib.Service3;
procedure My_Lib_Dummy is
begin
null;
end;
Here are the generic commands that will build an archive or a shared library.
# compiling the library
$ gnatmake -c my_lib_dummy.adb
# we don't need the dummy object itself
$ rm my_lib_dummy.o my_lib_dummy.ali
# create an archive with the remaining objects
$ ar rc libmy_lib.a *.o
# some systems may require "ranlib" to be run as well
# or create a shared library
$ gcc -shared -o libmy_lib.so *.o
# some systems may require the code to have been compiled with -fPIC
# remove the object files that are now in the library
$ rm *.o
# Make the ALI files read-only so that gnatmake will not try to
# regenerate the objects that are in the library
$ chmod -w *.ali
Please note that the library must have a name of the form lib`xxx'.a
or lib`xxx'.so (or lib`xxx'.dll on Windows) in order to
be accessed by the directive -l`xxx' at link time.
If you use project files, library installation is part of the library build process (Installing a library with project files).
When project files are not an option, it is also possible, but not recommended,
to install the library so that the sources needed to use the library are on the
Ada source path and the ALI files & libraries be on the Ada Object path (see
Search Paths and the Run-Time Library (RTL). Alternatively, the system
administrator can place general-purpose libraries in the default compiler
paths, by specifying the libraries' location in the configuration files
ada_source_path and ada_object_path. These configuration files
must be located in the GNAT installation tree at the same place as the gcc spec
file. The location of the gcc spec file can be determined as follows:
$ gcc -v
The configuration files mentioned above have a simple format: each line must contain one unique directory name. Those names are added to the corresponding path in their order of appearance in the file. The names can be either absolute or relative; in the latter case, they are relative to where theses files are located.
The files ada_source_path and ada_object_path might not be
present in a
GNAT installation, in which case, GNAT will look for its run-time library in
the directories adainclude (for the sources) and adalib (for the
objects and ALI files). When the files exist, the compiler does not
look in adainclude and adalib, and thus the
ada_source_path file
must contain the location for the GNAT run-time sources (which can simply
be adainclude). In the same way, the ada_object_path file must
contain the location for the GNAT run-time objects (which can simply
be adalib).
You can also specify a new default path to the run-time library at compilation time with the switch `–RTS=rts-path'. You can thus choose / change the run-time library you want your program to be compiled with. This switch is recognized by `gcc', `gnatmake', `gnatbind', `gnatls', `gnatfind' and `gnatxref'.
It is possible to install a library before or after the standard GNAT library, by reordering the lines in the configuration files. In general, a library must be installed before the GNAT library if it redefines any part of it.
Once again, the project facility greatly simplifies the use of libraries. In this context, using a library is just a matter of adding a `with' clause in the user project. For instance, to make use of the library My_Lib shown in examples in earlier sections, you can write:
with "my_lib";
project My_Proj is
...
end My_Proj;
Even if you have a third-party, non-Ada library, you can still use GNAT's
Project Manager facility to provide a wrapper for it. For example, the
following project, when `with'ed by your main project, will link with the
third-party library liba.a:
project Liba is
for Externally_Built use "true";
for Source_Files use ();
for Library_Dir use "lib";
for Library_Name use "a";
for Library_Kind use "static";
end Liba;
This is an alternative to the use of pragma Linker_Options. It is especially interesting in the context of systems with several interdependent static libraries where finding a proper linker order is not easy and best be left to the tools having visibility over project dependence information.
In order to use an Ada library manually, you need to make sure that this library is on both your source and object path (see Search Paths and the Run-Time Library (RTL) and Search Paths for gnatbind). Furthermore, when the objects are grouped in an archive or a shared library, you need to specify the desired library at link time.
For example, you can use the library mylib installed in
/dir/my_lib_src and /dir/my_lib_obj with the following commands:
$ gnatmake -aI/dir/my_lib_src -aO/dir/my_lib_obj my_appl \\
-largs -lmy_lib
This can be expressed more simply:
$ gnatmake my_appl
when the following conditions are met:
/dir/my_lib_src has been added by the user to the environment
variable
ADA_INCLUDE_PATH, or by the administrator to the file
ada_source_path
/dir/my_lib_obj has been added by the user to the environment
variable
ADA_OBJECTS_PATH, or by the administrator to the file
ada_object_path
pragma Linker_Options ("-lmy_lib");
Note that you may also load a library dynamically at
run time given its filename, as illustrated in the GNAT plugins example
in the directory share/examples/gnat/plugins within the GNAT
install area.
A Stand-alone Library (abbreviated 'SAL') is a library that contains the
necessary code to
elaborate the Ada units that are included in the library. In contrast with
an ordinary library, which consists of all sources, objects and ALI
files of the
library, a SAL may specify a restricted subset of compilation units
to serve as a library interface. In this case, the fully
self-sufficient set of files will normally consist of an objects
archive, the sources of interface units' specs, and the ALI
files of interface units.
If an interface spec contains a generic unit or an inlined subprogram,
the body's
source must also be provided; if the units that must be provided in the source
form depend on other units, the source and ALI files of those must
also be provided.
The main purpose of a SAL is to minimize the recompilation overhead of client applications when a new version of the library is installed. Specifically, if the interface sources have not changed, client applications do not need to be recompiled. If, furthermore, a SAL is provided in the shared form and its version, controlled by Library_Version attribute, is not changed, then the clients do not need to be relinked.
SALs also allow the library providers to minimize the amount of library source text exposed to the clients. Such 'information hiding' might be useful or necessary for various reasons.
Stand-alone libraries are also well suited to be used in an executable whose main routine is not written in Ada.
GNAT's Project facility provides a simple way of building and installing stand-alone libraries; see Stand-alone Library Projects. To be a Stand-alone Library Project, in addition to the two attributes that make a project a Library Project (Library_Name and Library_Dir; see Library Projects), the attribute Library_Interface must be defined. For example:
for Library_Dir use "lib_dir";
for Library_Name use "dummy";
for Library_Interface use ("int1", "int1.child");
Attribute Library_Interface has a non-empty string list value, each string in the list designating a unit contained in an immediate source of the project file.
When a Stand-alone Library is built, first the binder is invoked to build
a package whose name depends on the library name
(b~dummy.ads/b in the example above).
This binder-generated package includes initialization and
finalization procedures whose
names depend on the library name (dummyinit and dummyfinal
in the example
above). The object corresponding to this package is included in the library.
You must ensure timely (e.g., prior to any use of interfaces in the SAL) calling of these procedures if a static SAL is built, or if a shared SAL is built with the project-level attribute Library_Auto_Init set to "false".
For a Stand-Alone Library, only the ALI files of the Interface Units
(those that are listed in attribute Library_Interface) are copied to
the Library Directory. As a consequence, only the Interface Units may be
imported from Ada units outside of the library. If other units are imported,
the binding phase will fail.
It is also possible to build an encapsulated library where not only the code to elaborate and finalize the library is embedded but also ensuring that the library is linked only against static libraries. So an encapsulated library only depends on system libraries, all other code, including the GNAT runtime, is embedded. To build an encapsulated library the attribute Library_Standalone must be set to encapsulated:
for Library_Dir use "lib_dir";
for Library_Name use "dummy";
for Library_Kind use "dynamic";
for Library_Interface use ("int1", "int1.child");
for Library_Standalone use "encapsulated";
The default value for this attribute is standard in which case a stand-alone library is built.
The attribute Library_Src_Dir may be specified for a Stand-Alone Library. Library_Src_Dir is a simple attribute that has a single string value. Its value must be the path (absolute or relative to the project directory) of an existing directory. This directory cannot be the object directory or one of the source directories, but it can be the same as the library directory. The sources of the Interface Units of the library that are needed by an Ada client of the library will be copied to the designated directory, called the Interface Copy directory. These sources include the specs of the Interface Units, but they may also include bodies and subunits, when pragmas Inline or Inline_Always are used, or when there is a generic unit in the spec. Before the sources are copied to the Interface Copy directory, an attempt is made to delete all files in the Interface Copy directory.
Building stand-alone libraries by hand is somewhat tedious, but for those occasions when it is necessary here are the steps that you need to perform:
ALI files of the interfaces, and
with the switch `-L' to give specific names to the init
and final procedures. For example:
$ gnatbind -n int1.ali int2.ali -Lsal1
$ gcc -c b~int2.adb
ALI file that starts
with letter 'P') and make the modified copy of the ALI file
read-only.
Using SALs is not different from using other libraries (see Using a library).
It is easy to adapt the SAL build procedure discussed above for use of a SAL in a non-Ada context.
The only extra step required is to ensure that library interface subprograms are compatible with the main program, by means of pragma Export or pragma Convention.
Here is an example of simple library interface for use with C main program:
package My_Package is
procedure Do_Something;
pragma Export (C, Do_Something, "do_something");
procedure Do_Something_Else;
pragma Export (C, Do_Something_Else, "do_something_else");
end My_Package;
On the foreign language side, you must provide a 'foreign' view of the library interface; remember that it should contain elaboration routines in addition to interface subprograms.
The example below shows the content of mylib_interface.h (note that there is no rule for the naming of this file, any name can be used)
/* the library elaboration procedure */
extern void mylibinit (void);
/* the library finalization procedure */
extern void mylibfinal (void);
/* the interface exported by the library */
extern void do_something (void);
extern void do_something_else (void);
Libraries built as explained above can be used from any program, provided that the elaboration procedures (named mylibinit in the previous example) are called before the library services are used. Any number of libraries can be used simultaneously, as long as the elaboration procedure of each library is called.
Below is an example of a C program that uses the mylib library.
#include "mylib_interface.h"
int
main (void)
{
/* First, elaborate the library before using it */
mylibinit ();
/* Main program, using the library exported entities */
do_something ();
do_something_else ();
/* Library finalization at the end of the program */
mylibfinal ();
return 0;
}
Note that invoking any library finalization procedure generated by gnatbind shuts down the Ada run-time environment. Consequently, the finalization of all Ada libraries must be performed at the end of the program. No call to these libraries or to the Ada run-time library should be made after the finalization phase.
Note also that special care must be taken with multi-tasks applications. The initialization and finalization routines are not protected against concurrent access. If such requirement is needed it must be ensured at the application level using a specific operating system services like a mutex or a critical-section.
The pragmas listed below should be used with caution inside libraries, as they can create incompatibilities with other Ada libraries:
When using a library that contains such pragmas, the user must make sure that all libraries use the same pragmas with the same values. Otherwise, Program_Error will be raised during the elaboration of the conflicting libraries. The usage of these pragmas and its consequences for the user should therefore be well documented.
Similarly, the traceback in the exception occurrence mechanism should be enabled or disabled in a consistent manner across all libraries. Otherwise, Program_Error will be raised during the elaboration of the conflicting libraries.
If the Version or Body_Version
attributes are used inside a library, then you need to
perform a gnatbind step that specifies all ALI files in all
libraries, so that version identifiers can be properly computed.
In practice these attributes are rarely used, so this is unlikely
to be a consideration.
It may be useful to recompile the GNAT library in various contexts, the most important one being the use of partition-wide configuration pragmas such as Normalize_Scalars. A special Makefile called Makefile.adalib is provided to that effect and can be found in the directory containing the GNAT library. The location of this directory depends on the way the GNAT environment has been installed and can be determined by means of the command:
$ gnatls -v
The last entry in the object search path usually contains the gnat library. This Makefile contains its own documentation and in particular the set of instructions needed to rebuild a new library and to use it.
This section presents some guidelines for modeling conditional compilation in Ada and describes the gnatprep preprocessor utility.
It is often necessary to arrange for a single source program to serve multiple purposes, where it is compiled in different ways to achieve these different goals. Some examples of the need for this feature are
In C, or C++, the typical approach would be to use the preprocessor that is defined as part of the language. The Ada language does not contain such a feature. This is not an oversight, but rather a very deliberate design decision, based on the experience that overuse of the preprocessing features in C and C++ can result in programs that are extremely difficult to maintain. For example, if we have ten switches that can be on or off, this means that there are a thousand separate programs, any one of which might not even be syntactically correct, and even if syntactically correct, the resulting program might not work correctly. Testing all combinations can quickly become impossible.
Nevertheless, the need to tailor programs certainly exists, and in this section we will discuss how this can be achieved using Ada in general, and GNAT in particular.
In the case where the difference is simply which code sequence is executed, the cleanest solution is to use Boolean constants to control which code is executed.
FP_Initialize_Required : constant Boolean := True;
...
if FP_Initialize_Required then
...
end if;
Not only will the code inside the if statement not be executed if the constant Boolean is False, but it will also be completely deleted from the program. However, the code is only deleted after the if statement has been checked for syntactic and semantic correctness. (In contrast, with preprocessors the code is deleted before the compiler ever gets to see it, so it is not checked until the switch is turned on.)
Typically the Boolean constants will be in a separate package, something like:
package Config is
FP_Initialize_Required : constant Boolean := True;
Reset_Available : constant Boolean := False;
...
end Config;
The Config package exists in multiple forms for the various targets, with an appropriate script selecting the version of Config needed. Then any other unit requiring conditional compilation can do a `with' of Config to make the constants visible.
A common use of conditional code is to execute statements (for example dynamic checks, or output of intermediate results) under control of a debug switch, so that the debugging behavior can be turned on and off. This can be done using a Boolean constant to control whether the code is active:
if Debugging then
Put_Line ("got to the first stage!");
end if;
or
if Debugging and then Temperature > 999.0 then
raise Temperature_Crazy;
end if;
Since this is a common case, there are special features to deal with this in a convenient manner. For the case of tests, Ada 2005 has added a pragma Assert that can be used for such tests. This pragma is modeled on the Assert pragma that has always been available in GNAT, so this feature may be used with GNAT even if you are not using Ada 2005 features. The use of pragma Assert is described in the GNAT_Reference_Manual, but as an example, the last test could be written:
pragma Assert (Temperature <= 999.0, "Temperature Crazy");
or simply
pragma Assert (Temperature <= 999.0);
In both cases, if assertions are active and the temperature is excessive, the exception Assert_Failure will be raised, with the given string in the first case or a string indicating the location of the pragma in the second case used as the exception message.
You can turn assertions on and off by using the Assertion_Policy pragma.
This is an Ada 2005 pragma which is implemented in all modes by GNAT. Alternatively, you can use the `-gnata' switch to enable assertions from the command line, which applies to all versions of Ada.
For the example above with the Put_Line, the GNAT-specific pragma Debug can be used:
pragma Debug (Put_Line ("got to the first stage!"));
If debug pragmas are enabled, the argument, which must be of the form of a procedure call, is executed (in this case, Put_Line will be called). Only one call can be present, but of course a special debugging procedure containing any code you like can be included in the program and then called in a pragma Debug argument as needed.
One advantage of pragma Debug over the if Debugging then construct is that pragma Debug can appear in declarative contexts, such as at the very beginning of a procedure, before local declarations have been elaborated.
Debug pragmas are enabled using either the `-gnata' switch that also controls assertions, or with a separate Debug_Policy pragma.
The latter pragma is new in the Ada 2005 versions of GNAT (but it can be used in Ada 95 and Ada 83 programs as well), and is analogous to pragma Assertion_Policy to control assertions.
Assertion_Policy and Debug_Policy are configuration pragmas,
and thus they can appear in gnat.adc if you are not using a
project file, or in the file designated to contain configuration pragmas
in a project file.
They then apply to all subsequent compilations. In practice the use of
the `-gnata' switch is often the most convenient method of controlling
the status of these pragmas.
Note that a pragma is not a statement, so in contexts where a statement sequence is required, you can't just write a pragma on its own. You have to add a null statement.
if ... then
... -- some statements
else
pragma Assert (Num_Cases < 10);
null;
end if;
In some cases it may be necessary to conditionalize declarations to meet different requirements. For example we might want a bit string whose length is set to meet some hardware message requirement.
This may be possible using declare blocks controlled by conditional constants:
if Small_Machine then
declare
X : Bit_String (1 .. 10);
begin
...
end;
else
declare
X : Large_Bit_String (1 .. 1000);
begin
...
end;
end if;
Note that in this approach, both declarations are analyzed by the compiler so this can only be used where both declarations are legal, even though one of them will not be used.
Another approach is to define integer constants, e.g., Bits_Per_Word, or Boolean constants, e.g., Little_Endian, and then write declarations that are parameterized by these constants. For example
for Rec use
Field1 at 0 range Boolean'Pos (Little_Endian) * 10 .. Bits_Per_Word;
end record;
If Bits_Per_Word is set to 32, this generates either
for Rec use
Field1 at 0 range 0 .. 32;
end record;
for the big endian case, or
for Rec use record
Field1 at 0 range 10 .. 32;
end record;
for the little endian case. Since a powerful subset of Ada expression notation is usable for creating static constants, clever use of this feature can often solve quite difficult problems in conditionalizing compilation (note incidentally that in Ada 95, the little endian constant was introduced as System.Default_Bit_Order, so you do not need to define this one yourself).
In some cases, none of the approaches described above are adequate. This can occur for example if the set of declarations required is radically different for two different configurations.
In this situation, the official Ada way of dealing with conditionalizing such code is to write separate units for the different cases. As long as this does not result in excessive duplication of code, this can be done without creating maintenance problems. The approach is to share common code as far as possible, and then isolate the code and declarations that are different. Subunits are often a convenient method for breaking out a piece of a unit that is to be conditionalized, with separate files for different versions of the subunit for different targets, where the build script selects the right one to give to the compiler.
As an example, consider a situation where a new feature in Ada 2005 allows something to be done in a really nice way. But your code must be able to compile with an Ada 95 compiler. Conceptually you want to say:
if Ada_2005 then
... neat Ada 2005 code
else
... not quite as neat Ada 95 code
end if;
where Ada_2005 is a Boolean constant.
But this won't work when Ada_2005 is set to False, since the then clause will be illegal for an Ada 95 compiler. (Recall that although such unreachable code would eventually be deleted by the compiler, it still needs to be legal. If it uses features introduced in Ada 2005, it will be illegal in Ada 95.)
So instead we write
procedure Insert is separate;
Then we have two files for the subunit Insert, with the two sets of code. If the package containing this is called File_Queries, then we might have two files
file_queries-insert-2005.adb
file_queries-insert-95.adb
and the build script renames the appropriate file to file_queries-insert.adb and then carries out the compilation.
This can also be done with project files' naming schemes. For example:
for body ("File_Queries.Insert") use "file_queries-insert-2005.ada";
Note also that with project files it is desirable to use a different extension
than ads / adb for alternative versions. Otherwise a naming
conflict may arise through another commonly used feature: to declare as part
of the project a set of directories containing all the sources obeying the
default naming scheme.
The use of alternative units is certainly feasible in all situations,
and for example the Ada part of the GNAT run-time is conditionalized
based on the target architecture using this approach. As a specific example,
consider the implementation of the AST feature in VMS. There is one
spec: s-asthan.ads which is the same for all architectures, and three
bodies:
s-asthan.adbs-asthan-vms-alpha.adbs-asthan-vms-ia64.adbThe dummy version s-asthan.adb simply raises exceptions noting that
this operating system feature is not available, and the two remaining
versions interface with the corresponding versions of VMS to provide
VMS-compatible AST handling. The GNAT build script knows the architecture
and operating system, and automatically selects the right version,
renaming it if necessary to s-asthan.adb before the run-time build.
Another style for arranging alternative implementations is through Ada's access-to-subprogram facility. In case some functionality is to be conditionally included, you can declare an access-to-procedure variable Ref that is initialized to designate a 'do nothing' procedure, and then invoke Ref.all when appropriate. In some library package, set Ref to Proc'Access for some procedure Proc that performs the relevant processing. The initialization only occurs if the library package is included in the program. The same idea can also be implemented using tagged types and dispatching calls.
Although it is quite possible to conditionalize code without the use of C-style preprocessing, as described earlier in this section, it is nevertheless convenient in some cases to use the C approach. Moreover, older Ada compilers have often provided some preprocessing capability, so legacy code may depend on this approach, even though it is not standard.
To accommodate such use, GNAT provides a preprocessor (modeled to a large extent on the various preprocessors that have been used with legacy code on other compilers, to enable easier transition).
The preprocessor may be used in two separate modes. It can be used quite separately from the compiler, to generate a separate output source file that is then fed to the compiler as a separate step. This is the gnatprep utility, whose use is fully described in Preprocessing with gnatprep.
The preprocessing language allows such constructs as
#if DEBUG or else (PRIORITY > 4) then
bunch of declarations
#else
completely different bunch of declarations
#end if;
The values of the symbols DEBUG and PRIORITY can be defined either on the command line or in a separate file.
The other way of running the preprocessor is even closer to the C style and often more convenient. In this approach the preprocessing is integrated into the compilation process. The compiler is fed the preprocessor input which includes #if lines etc, and then the compiler carries out the preprocessing internally and processes the resulting output. For more details on this approach, see Integrated Preprocessing.
This section discusses how to use GNAT's gnatprep utility for simple preprocessing. Although designed for use with GNAT, gnatprep does not depend on any special GNAT features. For further discussion of conditional compilation in general, see Conditional Compilation.
Preprocessing symbols are defined in definition files and referred to in sources to be preprocessed. A Preprocessing symbol is an identifier, following normal Ada (case-insensitive) rules for its syntax, with the restriction that all characters need to be in the ASCII set (no accented letters).
To call gnatprep use:
$ gnatprep [`switches`] `infile` `outfile` [`deffile`]
where
-b-c-C-D`symbol'=`value'-rNote that if the file to be preprocessed contains multiple units, then it will be necessary to gnatchop the output file from gnatprep. If a Source_Reference pragma is present in the preprocessed file, it will be respected by gnatchop -r so that the final chopped files will correctly refer to the original input source file for gnatprep.
-s-uNote: if neither `-b' nor `-c' is present, then preprocessor lines and deleted lines are completely removed from the output, unless -r is specified, in which case -b is assumed.
The definitions file contains lines of the form:
symbol := value
where symbol is a preprocessing symbol, and value is one of the following:
Comment lines may also appear in the definitions file, starting with
the usual --,
and comments may be added to the definitions lines.
The input text may contain preprocessor conditional inclusion lines, as well as general symbol substitution sequences.
The preprocessor conditional inclusion commands have the form:
#if <expression> [then]
lines
#elsif <expression> [then]
lines
#elsif <expression> [then]
lines
...
#else
lines
#end if;
In this example, <expression> is defined by the following grammar:
<expression> ::= <symbol>
<expression> ::= <symbol> = "<value>"
<expression> ::= <symbol> = <symbol>
<expression> ::= <symbol> = <integer>
<expression> ::= <symbol> > <integer>
<expression> ::= <symbol> >= <integer>
<expression> ::= <symbol> < <integer>
<expression> ::= <symbol> <= <integer>
<expression> ::= <symbol> 'Defined
<expression> ::= not <expression>
<expression> ::= <expression> and <expression>
<expression> ::= <expression> or <expression>
<expression> ::= <expression> and then <expression>
<expression> ::= <expression> or else <expression>
<expression> ::= ( <expression> )
Note the following restriction: it is not allowed to have "and" or "or" following "not" in the same expression without parentheses. For example, this is not allowed:
not X or Y
This can be expressed instead as one of the following forms:
(not X) or Y
not (X or Y)
For the first test (<expression> ::= <symbol>) the symbol must have either the value true or false, that is to say the right-hand of the symbol definition must be one of the (case-insensitive) literals True or False. If the value is true, then the corresponding lines are included, and if the value is false, they are excluded.
When comparing a symbol to an integer, the integer is any non negative literal integer as defined in the Ada Reference Manual, such as 3, 16#FF# or 2#11#. The symbol value must also be a non negative integer. Integer values in the range 0 .. 2**31-1 are supported.
The test (<expression> ::= <symbol>'Defined) is true only if the symbol has been defined in the definition file or by a `-D' switch on the command line. Otherwise, the test is false.
The equality tests are case insensitive, as are all the preprocessor lines.
If the symbol referenced is not defined in the symbol definitions file, then the effect depends on whether or not switch `-u' is specified. If so, then the symbol is treated as if it had the value false and the test fails. If this switch is not specified, then it is an error to reference an undefined symbol. It is also an error to reference a symbol that is defined with a value other than True or False.
The use of the not operator inverts the sense of this logical test. The not operator cannot be combined with the or or and operators, without parentheses. For example, "if not X or Y then" is not allowed, but "if (not X) or Y then" and "if not (X or Y) then" are.
The then keyword is optional as shown
The # must be the first non-blank character on a line, but otherwise the format is free form. Spaces or tabs may appear between the # and the keyword. The keywords and the symbols are case insensitive as in normal Ada code. Comments may be used on a preprocessor line, but other than that, no other tokens may appear on a preprocessor line. Any number of elsif clauses can be present, including none at all. The else is optional, as in Ada.
The # marking the start of a preprocessor line must be the first non-blank character on the line, i.e., it must be preceded only by spaces or horizontal tabs.
Symbol substitution outside of preprocessor lines is obtained by using the sequence:
$symbol
anywhere within a source line, except in a comment or within a string literal. The identifier following the $ must match one of the symbols defined in the symbol definition file, and the result is to substitute the value of the symbol in place of $symbol in the output file.
Note that although the substitution of strings within a string literal is not possible, it is possible to have a symbol whose defined value is a string literal. So instead of setting XYZ to hello and writing:
Header : String := "$XYZ";
you should set XYZ to "hello" and write:
Header : String := $XYZ;
and then the substitution will occur as desired.
GNAT sources may be preprocessed immediately before compilation.
In this case, the actual
text of the source is not the text of the source file, but is derived from it
through a process called preprocessing. Integrated preprocessing is specified
through switches `-gnatep' and/or `-gnateD'. `-gnatep'
indicates, through a text file, the preprocessing data to be used.
-gnateD specifies or modifies the values of preprocessing symbol.
Note that integrated preprocessing applies only to Ada source files, it is
not available for configuration pragma files.
Note that when integrated preprocessing is used, the output from the preprocessor is not written to any external file. Instead it is passed internally to the compiler. If you need to preserve the result of preprocessing in a file, then you should use `gnatprep' to perform the desired preprocessing in stand-alone mode.
It is recommended that `gnatmake' switch -s should be used when Integrated Preprocessing is used. The reason is that preprocessing with another Preprocessing Data file without changing the sources will not trigger recompilation without this switch.
Note that `gnatmake' switch -m will almost always trigger recompilation for sources that are preprocessed, because `gnatmake' cannot compute the checksum of the source after preprocessing.
The actual preprocessing function is described in detail in section Preprocessing with gnatprep. This section only describes how integrated preprocessing is triggered and parameterized.
-gnatep=`file'A preprocessing data file is a text file with significant lines indicating how should be preprocessed either a specific source or all sources not mentioned in other lines. A significant line is a nonempty, non-comment line. Comments are similar to Ada comments.
Each significant line starts with either a literal string or the character '*'. A literal string is the file name (without directory information) of the source to preprocess. A character '*' indicates the preprocessing for all the sources that are not specified explicitly on other lines (order of the lines is not significant). It is an error to have two lines with the same file name or two lines starting with the character '*'.
After the file name or the character '*', another optional literal string indicating the file name of the definition file to be used for preprocessing (Form of Definitions File). The definition files are found by the compiler in one of the source directories. In some cases, when compiling a source in a directory other than the current directory, if the definition file is in the current directory, it may be necessary to add the current directory as a source directory through switch -I., otherwise the compiler would not find the definition file.
Then, optionally, switches similar to those of gnatprep may be found. Those switches are:
-b-c-Dsymbol=`value'-s-uExamples of valid lines in a preprocessor data file:
"toto.adb" "prep.def" -u
-- preprocess "toto.adb", using definition file "prep.def",
-- undefined symbol are False.
* -c -DVERSION=V101
-- preprocess all other sources without a definition file;
-- suppressed lined are commented; symbol VERSION has the value V101.
"titi.adb" "prep2.def" -s
-- preprocess "titi.adb", using definition file "prep2.def";
-- list all symbols with their values.
-gnateDsymbol[=value]Examples:
-gnateDToto=Titi
-gnateDFoo
-gnateDFoo=\"Foo-Bar\"
A symbol declared with this switch on the command line replaces a symbol with the same name either in a definition file or specified with a switch -D in the preprocessor data file.
This switch is similar to switch `-D' of gnatprep.
-gnateGThis section describes how to develop a mixed-language program, with a focus on combining Ada with C or C++.
Interfacing Ada with a foreign language such as C involves using compiler directives to import and/or export entity definitions in each language – using extern statements in C, for instance, and the Import, Export, and Convention pragmas in Ada. A full treatment of these topics is provided in Appendix B, section 1 of the Ada Reference Manual.
There are two ways to build a program using GNAT that contains some Ada sources and some foreign language sources, depending on whether or not the main subprogram is written in Ada. Here is a source example with the main subprogram in Ada:
/* file1.c */
#include <stdio.h>
void print_num (int num)
{
printf ("num is %d.\\n", num);
return;
}
/* file2.c */
/* num_from_Ada is declared in my_main.adb */
extern int num_from_Ada;
int get_num (void)
{
return num_from_Ada;
}
-- my_main.adb
procedure My_Main is
-- Declare then export an Integer entity called num_from_Ada
My_Num : Integer := 10;
pragma Export (C, My_Num, "num_from_Ada");
-- Declare an Ada function spec for Get_Num, then use
-- C function get_num for the implementation.
function Get_Num return Integer;
pragma Import (C, Get_Num, "get_num");
-- Declare an Ada procedure spec for Print_Num, then use
-- C function print_num for the implementation.
procedure Print_Num (Num : Integer);
pragma Import (C, Print_Num, "print_num";
begin
Print_Num (Get_Num);
end My_Main;
To build this example:
$ gcc -c file1.c
$ gcc -c file2.c
$ gnatmake -c my_main.adb
$ gnatbind my_main.ali
$ gnatlink my_main.ali file1.o file2.o
The last three steps can be grouped in a single command:
$ gnatmake my_main.adb -largs file1.o file2.o
If the main program is in a language other than Ada, then you may have
more than one entry point into the Ada subsystem. You must use a special
binder option to generate callable routines that initialize and
finalize the Ada units (Binding with Non-Ada Main Programs).
Calls to the initialization and finalization routines must be inserted
in the main program, or some other appropriate point in the code. The
call to initialize the Ada units must occur before the first Ada
subprogram is called, and the call to finalize the Ada units must occur
after the last Ada subprogram returns. The binder will place the
initialization and finalization subprograms into the
b~xxx.adb file where they can be accessed by your C
sources. To illustrate, we have the following example:
/* main.c */
extern void adainit (void);
extern void adafinal (void);
extern int add (int, int);
extern int sub (int, int);
int main (int argc, char *argv[])
{
int a = 21, b = 7;
adainit();
/* Should print "21 + 7 = 28" */
printf ("%d + %d = %d\\n", a, b, add (a, b));
/* Should print "21 - 7 = 14" */
printf ("%d - %d = %d\\n", a, b, sub (a, b));
adafinal();
}
-- unit1.ads
package Unit1 is
function Add (A, B : Integer) return Integer;
pragma Export (C, Add, "add");
end Unit1;
-- unit1.adb
package body Unit1 is
function Add (A, B : Integer) return Integer is
begin
return A + B;
end Add;
end Unit1;
-- unit2.ads
package Unit2 is
function Sub (A, B : Integer) return Integer;
pragma Export (C, Sub, "sub");
end Unit2;
-- unit2.adb
package body Unit2 is
function Sub (A, B : Integer) return Integer is
begin
return A - B;
end Sub;
end Unit2;
The build procedure for this application is similar to the last example's:
$ gcc -c main.c
$ gnatmake -c unit1.adb
$ gnatmake -c unit2.adb
-n option to specify a foreign main program:
$ gnatbind -n unit1.ali unit2.ali
$ gnatlink unit2.ali main.o -o exec_file
This procedure yields a binary executable called exec_file.
Depending on the circumstances (for example when your non-Ada main object
does not provide symbol main), you may also need to instruct the
GNAT linker not to include the standard startup objects by passing the
-nostartfiles switch to gnatlink.
GNAT follows standard calling sequence conventions and will thus interface to any other language that also follows these conventions. The following Convention identifiers are recognized by GNAT:
Note that in the case of GNAT running on a platform that supports HP Ada 83, a higher degree of compatibility can be guaranteed, and in particular records are laid out in an identical manner in the two compilers. Note also that if output from two different compilers is mixed, the program is responsible for dealing with elaboration issues. Probably the safest approach is to write the main program in the version of Ada other than GNAT, so that it takes care of its own elaboration requirements, and then call the GNAT-generated adainit procedure to ensure elaboration of the GNAT components. Consult the documentation of the other Ada compiler for further details on elaboration.
However, it is not possible to mix the tasking run time of GNAT and HP Ada 83, All the tasking operations must either be entirely within GNAT compiled sections of the program, or entirely within HP Ada 83 compiled sections of the program.
A note on interfacing to a C 'varargs' function:
In C, varargs allows a function to take a variable number of arguments. There is no direct equivalent in this to Ada. One approach that can be used is to create a C wrapper for each different profile and then interface to this C wrapper. For example, to print an int value using printf, create a C function printfi that takes two arguments, a pointer to a string and an int, and calls printf. Then in the Ada program, use pragma Import to interface to printfi.
It may work on some platforms to directly interface to a varargs function by providing a specific Ada profile for a particular call. However, this does not work on all platforms, since there is no guarantee that the calling sequence for a two argument normal C function is the same as for calling a varargs C function with the same two arguments.
type Distance is new Long_Float;
type Time is new Long_Float;
type Velocity is new Long_Float;
function "/" (D : Distance; T : Time)
return Velocity;
pragma Import (Intrinsic, "/");
This common idiom is often programmed with a generic definition and an
explicit body. The pragma makes it simpler to introduce such declarations.
It incurs no overhead in compilation time or code size, because it is
implemented as a single machine instruction.
function builtin_sqrt (F : Float) return Float;
pragma Import (Intrinsic, builtin_sqrt, "__builtin_sqrtf");
Most of the GCC builtins are accessible this way, and as for other import conventions (e.g. C), it is the user's responsibility to ensure that the Ada subprogram profile matches the underlying builtin expectations.
GNAT additionally provides a useful pragma Convention_Identifier that can be used to parameterize conventions and allow additional synonyms to be specified. For example if you have legacy code in which the convention identifier Fortran77 was used for Fortran, you can use the configuration pragma:
pragma Convention_Identifier (Fortran77, Fortran);
And from now on the identifier Fortran77 may be used as a convention identifier (for example in an Import pragma) with the same meaning as Fortran.
A programmer inexperienced with mixed-language development may find that building an application containing both Ada and C++ code can be a challenge. This section gives a few hints that should make this task easier.
GNAT supports interfacing with the G++ compiler (or any C++ compiler
generating code that is compatible with the G++ Application Binary
Interface —see <http://www.codesourcery.com/archives/cxx-abi>).
Interfacing can be done at 3 levels: simple data, subprograms, and classes. In the first two cases, GNAT offers a specific Convention C_Plus_Plus (or CPP) that behaves exactly like Convention C. Usually, C++ mangles the names of subprograms. To generate proper mangled names automatically, see Generating Ada Bindings for C and C++ headers). This problem can also be addressed manually in two ways:
Interfacing at the class level can be achieved by using the GNAT specific pragmas such as CPP_Constructor. See the GNAT_Reference_Manual for additional information.
Usually the linker of the C++ development system must be used to link mixed applications because most C++ systems will resolve elaboration issues (such as calling constructors on global class instances) transparently during the link phase. GNAT has been adapted to ease the use of a foreign linker for the last phase. Three cases can be considered:
Note that if the C++ code uses inline functions, you will need to compile your C++ code with the -fkeep-inline-functions switch in order to provide an existing function implementation that the Ada code can link with.
$ g++ -c -fkeep-inline-functions file1.C
$ g++ -c -fkeep-inline-functions file2.C
$ gnatmake ada_unit -largs file1.o file2.o --LINK=g++
C_INCLUDE_PATH,
GCC_EXEC_PREFIX,
BINUTILS_ROOT, and
GCC_ROOT will affect both compilers
at the same time and may make one of the two compilers operate
improperly if set during invocation of the wrong compiler. It is also
very important that the linker uses the proper libgcc.a GCC
library – that is, the one from the C++ compiler installation. The
implicit link command as suggested in the gnatmake command
from the former example can be replaced by an explicit link command with
the full-verbosity option in order to verify which library is used:
$ gnatbind ada_unit
$ gnatlink -v -v ada_unit file1.o file2.o --LINK=c++
If there is a problem due to interfering environment variables, it can be worked around by using an intermediate script. The following example shows the proper script to use when GNAT has not been installed at its default location and g++ has been installed at its default location:
$ cat ./my_script
#!/bin/sh
unset BINUTILS_ROOT
unset GCC_ROOT
c++ $*
$ gnatlink -v -v ada_unit file1.o file2.o --LINK=./my_script
If the setjmp/longjmp exception mechanism is used, only the paths to the libgcc libraries are required:
$ cat ./my_script
#!/bin/sh
CC $* `gcc -print-file-name=libgcc.a` `gcc -print-file-name=libgcc_eh.a`
$ gnatlink ada_unit file1.o file2.o --LINK=./my_script
where CC is the name of the non-GNU C++ compiler.
If the zero cost exception mechanism is used, and the platform supports automatic registration of exception tables (e.g., Solaris), paths to more objects are required:
$ cat ./my_script
#!/bin/sh
CC `gcc -print-file-name=crtbegin.o` $* \\
`gcc -print-file-name=libgcc.a` `gcc -print-file-name=libgcc_eh.a` \\
`gcc -print-file-name=crtend.o`
$ gnatlink ada_unit file1.o file2.o --LINK=./my_script
If the "zero cost exception" mechanism is used, and the platform doesn't support automatic registration of exception tables (e.g., HP-UX or AIX), the simple approach described above will not work and a pre-linking phase using GNAT will be necessary.
Another alternative is to use the gprbuild multi-language builder
which has a large knowledge base and knows how to link Ada and C++ code
together automatically in most cases.
The following example, provided as part of the GNAT examples, shows how to achieve procedural interfacing between Ada and C++ in both directions. The C++ class A has two methods. The first method is exported to Ada by the means of an extern C wrapper function. The second method calls an Ada subprogram. On the Ada side, The C++ calls are modelled by a limited record with a layout comparable to the C++ class. The Ada subprogram, in turn, calls the C++ method. So, starting from the C++ main program, the process passes back and forth between the two languages.
Here are the compilation commands:
$ gnatmake -c simple_cpp_interface
$ g++ -c cpp_main.C
$ g++ -c ex7.C
$ gnatbind -n simple_cpp_interface
$ gnatlink simple_cpp_interface -o cpp_main --LINK=g++ -lstdc++ ex7.o cpp_main.o
Here are the corresponding sources:
//cpp_main.C
#include "ex7.h"
extern "C" {
void adainit (void);
void adafinal (void);
void method1 (A *t);
}
void method1 (A *t)
{
t->method1 ();
}
int main ()
{
A obj;
adainit ();
obj.method2 (3030);
adafinal ();
}
//ex7.h
class Origin {
public:
int o_value;
};
class A : public Origin {
public:
void method1 (void);
void method2 (int v);
A();
int a_value;
};
//ex7.C
#include "ex7.h"
#include <stdio.h>
extern "C" { void ada_method2 (A *t, int v);}
void A::method1 (void)
{
a_value = 2020;
printf ("in A::method1, a_value = %d \\n",a_value);
}
void A::method2 (int v)
{
ada_method2 (this, v);
printf ("in A::method2, a_value = %d \\n",a_value);
}
A::A(void)
{
a_value = 1010;
printf ("in A::A, a_value = %d \\n",a_value);
}
-- simple_cpp_interface.ads
with System;
package Simple_Cpp_Interface is
type A is limited
record
Vptr : System.Address;
O_Value : Integer;
A_Value : Integer;
end record;
pragma Convention (C, A);
procedure Method1 (This : in out A);
pragma Import (C, Method1);
procedure Ada_Method2 (This : in out A; V : Integer);
pragma Export (C, Ada_Method2);
end Simple_Cpp_Interface;
-- simple_cpp_interface.adb
package body Simple_Cpp_Interface is
procedure Ada_Method2 (This : in out A; V : Integer) is
begin
Method1 (This);
This.A_Value := V;
end Ada_Method2;
end Simple_Cpp_Interface;
In order to interface with C++ constructors GNAT provides the pragma CPP_Constructor (see the GNAT_Reference_Manual for additional information). In this section we present some common uses of C++ constructors in mixed-languages programs in GNAT.
Let us assume that we need to interface with the following C++ class:
class Root {
public:
int a_value;
int b_value;
virtual int Get_Value ();
Root(); // Default constructor
Root(int v); // 1st non-default constructor
Root(int v, int w); // 2nd non-default constructor
};
For this purpose we can write the following package spec (further information on how to build this spec is available in Interfacing with C++ at the Class Level and Generating Ada Bindings for C and C++ headers).
with Interfaces.C; use Interfaces.C;
package Pkg_Root is
type Root is tagged limited record
A_Value : int;
B_Value : int;
end record;
pragma Import (CPP, Root);
function Get_Value (Obj : Root) return int;
pragma Import (CPP, Get_Value);
function Constructor return Root;
pragma Cpp_Constructor (Constructor, "_ZN4RootC1Ev");
function Constructor (v : Integer) return Root;
pragma Cpp_Constructor (Constructor, "_ZN4RootC1Ei");
function Constructor (v, w : Integer) return Root;
pragma Cpp_Constructor (Constructor, "_ZN4RootC1Eii");
end Pkg_Root;
On the Ada side the constructor is represented by a function (whose name is arbitrary) that returns the classwide type corresponding to the imported C++ class. Although the constructor is described as a function, it is typically a procedure with an extra implicit argument (the object being initialized) at the implementation level. GNAT issues the appropriate call, whatever it is, to get the object properly initialized.
Constructors can only appear in the following contexts:
In a declaration of an object whose type is a class imported from C++, either the default C++ constructor is implicitly called by GNAT, or else the required C++ constructor must be explicitly called in the expression that initializes the object. For example:
Obj1 : Root;
Obj2 : Root := Constructor;
Obj3 : Root := Constructor (v => 10);
Obj4 : Root := Constructor (30, 40);
The first two declarations are equivalent: in both cases the default C++ constructor is invoked (in the former case the call to the constructor is implicit, and in the latter case the call is explicit in the object declaration). Obj3 is initialized by the C++ non-default constructor that takes an integer argument, and Obj4 is initialized by the non-default C++ constructor that takes two integers.
Let us derive the imported C++ class in the Ada side. For example:
type DT is new Root with record
C_Value : Natural := 2009;
end record;
In this case the components DT inherited from the C++ side must be initialized by a C++ constructor, and the additional Ada components of type DT are initialized by GNAT. The initialization of such an object is done either by default, or by means of a function returning an aggregate of type DT, or by means of an extension aggregate.
Obj5 : DT;
Obj6 : DT := Function_Returning_DT (50);
Obj7 : DT := (Constructor (30,40) with C_Value => 50);
The declaration of Obj5 invokes the default constructors: the C++ default constructor of the parent type takes care of the initialization of the components inherited from Root, and GNAT takes care of the default initialization of the additional Ada components of type DT (that is, C_Value is initialized to value 2009). The order of invocation of the constructors is consistent with the order of elaboration required by Ada and C++. That is, the constructor of the parent type is always called before the constructor of the derived type.
Let us now consider a record that has components whose type is imported from C++. For example:
type Rec1 is limited record
Data1 : Root := Constructor (10);
Value : Natural := 1000;
end record;
type Rec2 (D : Integer := 20) is limited record
Rec : Rec1;
Data2 : Root := Constructor (D, 30);
end record;
The initialization of an object of type Rec2 will call the non-default C++ constructors specified for the imported components. For example:
Obj8 : Rec2 (40);
Using Ada 2005 we can use limited aggregates to initialize an object invoking C++ constructors that differ from those specified in the type declarations. For example:
Obj9 : Rec2 := (Rec => (Data1 => Constructor (15, 16),
others => <>),
others => <>);
The above declaration uses an Ada 2005 limited aggregate to initialize Obj9, and the C++ constructor that has two integer arguments is invoked to initialize the Data1 component instead of the constructor specified in the declaration of type Rec1. In Ada 2005 the box in the aggregate indicates that unspecified components are initialized using the expression (if any) available in the component declaration. That is, in this case discriminant D is initialized to value 20, Value is initialized to value 1000, and the non-default C++ constructor that handles two integers takes care of initializing component Data2 with values 20,30.
In Ada 2005 we can use the extended return statement to build the Ada equivalent to C++ non-default constructors. For example:
function Constructor (V : Integer) return Rec2 is
begin
return Obj : Rec2 := (Rec => (Data1 => Constructor (V, 20),
others => <>),
others => <>) do
-- Further actions required for construction of
-- objects of type Rec2
...
end record;
end Constructor;
In this example the extended return statement construct is used to build in place the returned object whose components are initialized by means of a limited aggregate. Any further action associated with the constructor can be placed inside the construct.
In this section we demonstrate the GNAT features for interfacing with C++ by means of an example making use of Ada 2005 abstract interface types. This example consists of a classification of animals; classes have been used to model our main classification of animals, and interfaces provide support for the management of secondary classifications. We first demonstrate a case in which the types and constructors are defined on the C++ side and imported from the Ada side, and latter the reverse case.
The root of our derivation will be the Animal class, with a single private attribute (the Age of the animal), a constructor, and two public primitives to set and get the value of this attribute.
class Animal {
public:
virtual void Set_Age (int New_Age);
virtual int Age ();
Animal() {Age_Count = 0;};
private:
int Age_Count;
};
Abstract interface types are defined in C++ by means of classes with pure virtual functions and no data members. In our example we will use two interfaces that provide support for the common management of Carnivore and Domestic animals:
class Carnivore {
public:
virtual int Number_Of_Teeth () = 0;
};
class Domestic {
public:
virtual void Set_Owner (char* Name) = 0;
};
Using these declarations, we can now say that a Dog is an animal that is both Carnivore and Domestic, that is:
class Dog : Animal, Carnivore, Domestic {
public:
virtual int Number_Of_Teeth ();
virtual void Set_Owner (char* Name);
Dog(); // Constructor
private:
int Tooth_Count;
char *Owner;
};
In the following examples we will assume that the previous declarations are located in a file named animals.h. The following package demonstrates how to import these C++ declarations from the Ada side:
with Interfaces.C.Strings; use Interfaces.C.Strings;
package Animals is
type Carnivore is limited interface;
pragma Convention (C_Plus_Plus, Carnivore);
function Number_Of_Teeth (X : Carnivore)
return Natural is abstract;
type Domestic is limited interface;
pragma Convention (C_Plus_Plus, Domestic);
procedure Set_Owner
(X : in out Domestic;
Name : Chars_Ptr) is abstract;
type Animal is tagged limited record
Age : Natural;
end record;
pragma Import (C_Plus_Plus, Animal);
procedure Set_Age (X : in out Animal; Age : Integer);
pragma Import (C_Plus_Plus, Set_Age);
function Age (X : Animal) return Integer;
pragma Import (C_Plus_Plus, Age);
function New_Animal return Animal;
pragma CPP_Constructor (New_Animal);
pragma Import (CPP, New_Animal, "_ZN6AnimalC1Ev");
type Dog is new Animal and Carnivore and Domestic with record
Tooth_Count : Natural;
Owner : String (1 .. 30);
end record;
pragma Import (C_Plus_Plus, Dog);
function Number_Of_Teeth (A : Dog) return Natural;
pragma Import (C_Plus_Plus, Number_Of_Teeth);
procedure Set_Owner (A : in out Dog; Name : Chars_Ptr);
pragma Import (C_Plus_Plus, Set_Owner);
function New_Dog return Dog;
pragma CPP_Constructor (New_Dog);
pragma Import (CPP, New_Dog, "_ZN3DogC2Ev");
end Animals;
Thanks to the compatibility between GNAT run-time structures and the C++ ABI, interfacing with these C++ classes is easy. The only requirement is that all the primitives and components must be declared exactly in the same order in the two languages.
Regarding the abstract interfaces, we must indicate to the GNAT compiler by means of a pragma Convention (C_Plus_Plus), the convention used to pass the arguments to the called primitives will be the same as for C++. For the imported classes we use pragma Import with convention C_Plus_Plus to indicate that they have been defined on the C++ side; this is required because the dispatch table associated with these tagged types will be built in the C++ side and therefore will not contain the predefined Ada primitives which Ada would otherwise expect.
As the reader can see there is no need to indicate the C++ mangled names associated with each subprogram because it is assumed that all the calls to these primitives will be dispatching calls. The only exception is the constructor, which must be registered with the compiler by means of pragma CPP_Constructor and needs to provide its associated C++ mangled name because the Ada compiler generates direct calls to it.
With the above packages we can now declare objects of type Dog on the Ada side and dispatch calls to the corresponding subprograms on the C++ side. We can also extend the tagged type Dog with further fields and primitives, and override some of its C++ primitives on the Ada side. For example, here we have a type derivation defined on the Ada side that inherits all the dispatching primitives of the ancestor from the C++ side.
with Animals; use Animals;
package Vaccinated_Animals is
type Vaccinated_Dog is new Dog with null record;
function Vaccination_Expired (A : Vaccinated_Dog) return Boolean;
end Vaccinated_Animals;
It is important to note that, because of the ABI compatibility, the programmer does not need to add any further information to indicate either the object layout or the dispatch table entry associated with each dispatching operation.
Now let us define all the types and constructors on the Ada side and export them to C++, using the same hierarchy of our previous example:
with Interfaces.C.Strings;
use Interfaces.C.Strings;
package Animals is
type Carnivore is limited interface;
pragma Convention (C_Plus_Plus, Carnivore);
function Number_Of_Teeth (X : Carnivore)
return Natural is abstract;
type Domestic is limited interface;
pragma Convention (C_Plus_Plus, Domestic);
procedure Set_Owner
(X : in out Domestic;
Name : Chars_Ptr) is abstract;
type Animal is tagged record
Age : Natural;
end record;
pragma Convention (C_Plus_Plus, Animal);
procedure Set_Age (X : in out Animal; Age : Integer);
pragma Export (C_Plus_Plus, Set_Age);
function Age (X : Animal) return Integer;
pragma Export (C_Plus_Plus, Age);
function New_Animal return Animal'Class;
pragma Export (C_Plus_Plus, New_Animal);
type Dog is new Animal and Carnivore and Domestic with record
Tooth_Count : Natural;
Owner : String (1 .. 30);
end record;
pragma Convention (C_Plus_Plus, Dog);
function Number_Of_Teeth (A : Dog) return Natural;
pragma Export (C_Plus_Plus, Number_Of_Teeth);
procedure Set_Owner (A : in out Dog; Name : Chars_Ptr);
pragma Export (C_Plus_Plus, Set_Owner);
function New_Dog return Dog'Class;
pragma Export (C_Plus_Plus, New_Dog);
end Animals;
Compared with our previous example the only differences are the use of pragma Convention (instead of pragma Import), and the use of pragma Export to indicate to the GNAT compiler that the primitives will be available to C++. Thanks to the ABI compatibility, on the C++ side there is nothing else to be done; as explained above, the only requirement is that all the primitives and components are declared in exactly the same order.
For completeness, let us see a brief C++ main program that uses the declarations available in animals.h (presented in our first example) to import and use the declarations from the Ada side, properly initializing and finalizing the Ada run-time system along the way:
#include "animals.h"
#include <iostream>
using namespace std;
void Check_Carnivore (Carnivore *obj) {...}
void Check_Domestic (Domestic *obj) {...}
void Check_Animal (Animal *obj) {...}
void Check_Dog (Dog *obj) {...}
extern "C" {
void adainit (void);
void adafinal (void);
Dog* new_dog ();
}
void test ()
{
Dog *obj = new_dog(); // Ada constructor
Check_Carnivore (obj); // Check secondary DT
Check_Domestic (obj); // Check secondary DT
Check_Animal (obj); // Check primary DT
Check_Dog (obj); // Check primary DT
}
int main ()
{
adainit (); test(); adafinal ();
return 0;
}
GNAT includes a binding generator for C and C++ headers which is intended to do 95% of the tedious work of generating Ada specs from C or C++ header files.
Note that this capability is not intended to generate 100% correct Ada specs, and will is some cases require manual adjustments, although it can often be used out of the box in practice.
Some of the known limitations include:
The code generated is using the Ada 2005 syntax, which makes it easier to interface with other languages than previous versions of Ada.
The binding generator is part of the `gcc' compiler and can be invoked via the `-fdump-ada-spec' switch, which will generate Ada spec files for the header files specified on the command line, and all header files needed by these files transitively. For example:
$ g++ -c -fdump-ada-spec -C /usr/include/time.h
$ gcc -c -gnat05 *.ads
will generate, under GNU/Linux, the following files: time_h.ads,
bits_time_h.ads, stddef_h.ads, bits_types_h.ads which
correspond to the files /usr/include/time.h,
/usr/include/bits/time.h, etc..., and will then compile in Ada 2005
mode these Ada specs.
The -C switch tells `gcc' to extract comments from headers, and will attempt to generate corresponding Ada comments.
If you want to generate a single Ada file and not the transitive closure, you can use instead the `-fdump-ada-spec-slim' switch.
You can optionally specify a parent unit, of which all generated units will be children, using -fada-spec-parent=<unit>.
Note that we recommend when possible to use the `g++' driver to generate bindings, even for most C headers, since this will in general generate better Ada specs. For generating bindings for C++ headers, it is mandatory to use the `g++' command, or `gcc -x c++' which is equivalent in this case. If `g++' cannot work on your C headers because of incompatibilities between C and C++, then you can fallback to `gcc' instead.
For an example of better bindings generated from the C++ front-end, the name of the parameters (when available) are actually ignored by the C front-end. Consider the following C header:
extern void foo (int variable);
with the C front-end, variable is ignored, and the above is handled as:
extern void foo (int);
generating a generic:
procedure foo (param1 : int);
with the C++ front-end, the name is available, and we generate:
procedure foo (variable : int);
In some cases, the generated bindings will be more complete or more meaningful
when defining some macros, which you can do via the `-D' switch. This
is for example the case with Xlib.h under GNU/Linux:
$ g++ -c -fdump-ada-spec -DXLIB_ILLEGAL_ACCESS -C /usr/include/X11/Xlib.h
The above will generate more complete bindings than a straight call without the `-DXLIB_ILLEGAL_ACCESS' switch.
In other cases, it is not possible to parse a header file in a stand-alone
manner, because other include files need to be included first. In this
case, the solution is to create a small header file including the needed
#include and possible #define directives. For example, to
generate Ada bindings for readline/readline.h, you need to first
include stdio.h, so you can create a file with the following two
lines in e.g. readline1.h:
#include <stdio.h>
#include <readline/readline.h>
and then generate Ada bindings from this file:
$ g++ -c -fdump-ada-spec readline1.h
Generating bindings for C++ headers is done using the same options, always with the `g++' compiler. Note that generating Ada spec from C++ headers is a much more complex job and support for C++ headers is much more limited that support for C headers. As a result, you will need to modify the resulting bindings by hand more extensively when using C++ headers.
In this mode, C++ classes will be mapped to Ada tagged types, constructors will be mapped using the CPP_Constructor pragma, and when possible, multiple inheritance of abstract classes will be mapped to Ada interfaces (see the `Interfacing to C++' section in the GNAT Reference Manual for additional information on interfacing to C++).
For example, given the following C++ header file:
class Carnivore {
public:
virtual int Number_Of_Teeth () = 0;
};
class Domestic {
public:
virtual void Set_Owner (char* Name) = 0;
};
class Animal {
public:
int Age_Count;
virtual void Set_Age (int New_Age);
};
class Dog : Animal, Carnivore, Domestic {
public:
int Tooth_Count;
char *Owner;
virtual int Number_Of_Teeth ();
virtual void Set_Owner (char* Name);
Dog();
};
The corresponding Ada code is generated:
package Class_Carnivore is
type Carnivore is limited interface;
pragma Import (CPP, Carnivore);
function Number_Of_Teeth (this : access Carnivore) return int is abstract;
end;
use Class_Carnivore;
package Class_Domestic is
type Domestic is limited interface;
pragma Import (CPP, Domestic);
procedure Set_Owner
(this : access Domestic;
Name : Interfaces.C.Strings.chars_ptr) is abstract;
end;
use Class_Domestic;
package Class_Animal is
type Animal is tagged limited record
Age_Count : aliased int;
end record;
pragma Import (CPP, Animal);
procedure Set_Age (this : access Animal; New_Age : int);
pragma Import (CPP, Set_Age, "_ZN6Animal7Set_AgeEi");
end;
use Class_Animal;
package Class_Dog is
type Dog is new Animal and Carnivore and Domestic with record
Tooth_Count : aliased int;
Owner : Interfaces.C.Strings.chars_ptr;
end record;
pragma Import (CPP, Dog);
function Number_Of_Teeth (this : access Dog) return int;
pragma Import (CPP, Number_Of_Teeth, "_ZN3Dog15Number_Of_TeethEv");
procedure Set_Owner
(this : access Dog; Name : Interfaces.C.Strings.chars_ptr);
pragma Import (CPP, Set_Owner, "_ZN3Dog9Set_OwnerEPc");
function New_Dog return Dog;
pragma CPP_Constructor (New_Dog);
pragma Import (CPP, New_Dog, "_ZN3DogC1Ev");
end;
use Class_Dog;
-fdump-ada-spec-fdump-ada-spec-slim-fada-spec-parent=`unit'-CGNAT includes a C header generator for Ada specifications which supports Ada types that have a direct mapping to C types. This includes in particular support for:
The C header generator is part of the GNAT compiler and can be invoked via
the `-gnatceg' combination of switches, which will generate a .h
file corresponding to the given input file (Ada spec or body). Note that
only spec files are processed in any case, so giving a spec or a body file
as input is equivalent. For example:
$ gcc -c -gnatceg pack1.ads
will generate a self-contained file called pack1.h including
common definitions from the Ada Standard package, followed by the
definitions included in pack1.ads, as well as all the other units
withed by this file.
For instance, given the following Ada files:
package Pack2 is
type Int is range 1 .. 10;
end Pack2;
with Pack2;
package Pack1 is
type Rec is record
Field1, Field2 : Pack2.Int;
end record;
Global : Rec := (1, 2);
procedure Proc1 (R : Rec);
procedure Proc2 (R : in out Rec);
end Pack1;
The above gcc command will generate the following pack1.h file:
/* Standard definitions skipped */
#ifndef PACK2_ADS
#define PACK2_ADS
typedef short_short_integer pack2__TintB;
typedef pack2__TintB pack2__int;
#endif /* PACK2_ADS */
#ifndef PACK1_ADS
#define PACK1_ADS
typedef struct _pack1__rec {
pack2__int field1;
pack2__int field2;
} pack1__rec;
extern pack1__rec pack1__global;
extern void pack1__proc1(const pack1__rec r);
extern void pack1__proc2(pack1__rec *r);
#endif /* PACK1_ADS */
You can then include pack1.h from a C source file and use the types,
call subprograms, reference objects, and constants.
This section compares the GNAT model with the approaches taken in other environents, first the C/C++ model and then the mechanism that has been used in other Ada systems, in particular those traditionally used for Ada 83.
The GNAT model of compilation is close to the C and C++ models. You can think of Ada specs as corresponding to header files in C. As in C, you don't need to compile specs; they are compiled when they are used. The Ada `with' is similar in effect to the #include of a C header.
One notable difference is that, in Ada, you may compile specs separately to check them for semantic and syntactic accuracy. This is not always possible with C headers because they are fragments of programs that have less specific syntactic or semantic rules.
The other major difference is the requirement for running the binder, which performs two important functions. First, it checks for consistency. In C or C++, the only defense against assembling inconsistent programs lies outside the compiler, in a makefile, for example. The binder satisfies the Ada requirement that it be impossible to construct an inconsistent program when the compiler is used in normal mode.
The other important function of the binder is to deal with elaboration issues. There are also elaboration issues in C++ that are handled automatically. This automatic handling has the advantage of being simpler to use, but the C++ programmer has no control over elaboration. Where gnatbind might complain there was no valid order of elaboration, a C++ compiler would simply construct a program that malfunctioned at run time.
This section is intended for Ada programmers who have used an Ada compiler implementing the traditional Ada library model, as described in the Ada Reference Manual.
In GNAT, there is no 'library' in the normal sense. Instead, the set of source files themselves acts as the library. Compiling Ada programs does not generate any centralized information, but rather an object file and a ALI file, which are of interest only to the binder and linker. In a traditional system, the compiler reads information not only from the source file being compiled, but also from the centralized library. This means that the effect of a compilation depends on what has been previously compiled. In particular:
In GNAT, compiling one unit never affects the compilation of any other units because the compiler reads only source files. Only changes to source files can affect the results of a compilation. In particular:
The most important result of these differences is that order of compilation is never significant in GNAT. There is no situation in which one is required to do one compilation before another. What shows up as order of compilation requirements in the traditional Ada library becomes, in GNAT, simple source dependencies; in other words, there is only a set of rules saying what source files must be present when a file is compiled.
This section explains how files that are produced by GNAT may be used with tools designed for other languages.
The object files generated by GNAT are in standard system format and in particular the debugging information uses this format. This means programs generated by GNAT can be used with existing utilities that depend on these formats.
In general, any utility program that works with C will also often work with Ada programs generated by GNAT. This includes software utilities such as gprof (a profiling program), gdb (the FSF debugger), and utilities such as Purify.
In order to interpret the output from GNAT, when using tools that are originally intended for use with other languages, it is useful to understand the conventions used to generate link names from the Ada entity names.
All link names are in all lowercase letters. With the exception of library procedure names, the mechanism used is simply to use the full expanded Ada name with dots replaced by double underscores. For example, suppose we have the following package spec:
package QRS is
MN : Integer;
end QRS;
The variable MN has a full expanded Ada name of QRS.MN, so the corresponding link name is qrs__mn. Of course if a pragma Export is used this may be overridden:
package Exports is
Var1 : Integer;
pragma Export (Var1, C, External_Name => "var1_name");
Var2 : Integer;
pragma Export (Var2, C, Link_Name => "var2_link_name");
end Exports;
In this case, the link name for Var1 is whatever link name the C compiler would assign for the C function var1_name. This typically would be either var1_name or _var1_name, depending on operating system conventions, but other possibilities exist. The link name for Var2 is var2_link_name, and this is not operating system dependent.
One exception occurs for library level procedures. A potential ambiguity arises between the required name _main for the C main program, and the name we would otherwise assign to an Ada library level procedure called Main (which might well not be the main program).
To avoid this ambiguity, we attach the prefix _ada_ to such names. So if we have a library level procedure such as:
procedure Hello (S : String);
the external name of this procedure will be _ada_hello.
This chapter describes first the gnatmake tool (Building with gnatmake), which automatically determines the set of sources needed by an Ada compilation unit and executes the necessary (re)compilations, binding and linking. It also explains how to use each tool individually: the compiler (gcc, see Compiling with gcc), binder (gnatbind, see Binding with gnatbind), and linker (gnatlink, see Linking with gnatlink) to build executable programs. Finally, this chapter provides examples of how to make use of the general GNU make mechanism in a GNAT context (see Using the GNU make Utility).
A typical development cycle when working on an Ada program consists of the following steps:
The third step in particular can be tricky, because not only do the modified files have to be compiled, but any files depending on these files must also be recompiled. The dependency rules in Ada can be quite complex, especially in the presence of overloading, use clauses, generics and inlined subprograms.
`gnatmake' automatically takes care of the third and fourth steps of this process. It determines which sources need to be compiled, compiles them, and binds and links the resulting object files.
Unlike some other Ada make programs, the dependencies are always accurately recomputed from the new sources. The source based approach of the GNAT compilation model makes this possible. This means that if changes to the source program cause corresponding changes in dependencies, they will always be tracked exactly correctly by `gnatmake'.
Note that for advanced description of project structure, we recommend creating a project file as explained in GNAT Project Manager and use the `gprbuild' tool which supports building with project files and works similarly to `gnatmake'.
The usual form of the `gnatmake' command is
$ gnatmake [<switches>] <file_name> [<file_names>] [<mode_switches>]
The only required argument is one file_name, which specifies a compilation unit that is a main program. Several file_names can be specified: this will result in several executables being built. If switches are present, they can be placed before the first file_name, between file_names or after the last file_name. If mode_switches are present, they must always be placed after the last file_name and all switches.
If you are using standard file extensions (.adb and
.ads), then the
extension may be omitted from the file_name arguments. However, if
you are using non-standard extensions, then it is required that the
extension be given. A relative or absolute directory path can be
specified in a file_name, in which case, the input source file will
be searched for in the specified directory only. Otherwise, the input
source file will first be searched in the directory where
`gnatmake' was invoked and if it is not found, it will be search on
the source path of the compiler as described in
Search Paths and the Run-Time Library (RTL).
All `gnatmake' output (except when you specify `-M') is sent to
stderr. The output produced by the
`-M' switch is sent to stdout.
You may specify any of the following switches to `gnatmake':
--version--help--version was not used, display usage, then exit disregarding
all other options.
--GCC=`compiler_name'gcc. You need to use
quotes around compiler_name if compiler_name contains
spaces or other separator characters.
As an example --GCC="foo -x -y"
will instruct `gnatmake' to use foo -x -y as your
compiler. A limitation of this syntax is that the name and path name of
the executable itself must not include any embedded spaces. Note that
switch -c is always inserted after your command name. Thus in the
above example the compiler command that will be used by `gnatmake'
will be foo -c -x -y. If several --GCC=compiler_name are
used, only the last compiler_name is taken into account. However,
all the additional switches are also taken into account. Thus,
--GCC="foo -x -y" --GCC="bar -z -t" is equivalent to
--GCC="bar -x -y -z -t".
--GNATBIND=`binder_name'gnatbind. You need to
use quotes around binder_name if binder_name contains spaces
or other separator characters.
As an example --GNATBIND="bar -x -y"
will instruct `gnatmake' to use bar -x -y as your
binder. Binder switches that are normally appended by `gnatmake'
to gnatbind are now appended to the end of bar -x -y.
A limitation of this syntax is that the name and path name of the executable
itself must not include any embedded spaces.
--GNATLINK=`linker_name'gnatlink. You need to
use quotes around linker_name if linker_name contains spaces
or other separator characters.
As an example --GNATLINK="lan -x -y"
will instruct `gnatmake' to use lan -x -y as your
linker. Linker switches that are normally appended by gnatmake to
gnatlink are now appended to the end of lan -x -y.
A limitation of this syntax is that the name and path name of the executable
itself must not include any embedded spaces.
--create-map-file--create-map-file=`mapfile'--create-missing-dirs-P`project'), automatically create
missing object directories, library directories and exec
directories.
--single-compile-per-obj-dir--subdirs=`subdir'--unchecked-shared-lib-imports--source-info=`source info file'-a-a is also useful
in conjunction with -f
if you need to recompile an entire application,
including run-time files, using special configuration pragmas,
such as a Normalize_Scalars pragma.
By default
gnatmake -a compiles all GNAT
internal files with
gcc -c -gnatpg rather than gcc -c.
-b-c-C-C=`file'-dcompleted x out of y (zz%)
If the file needs to be compiled this is displayed after the invocation of the compiler. These lines are displayed even in quiet output mode.
-D `dir'This switch cannot be used when using a project file.
-eI`nnn'-eLThis also assumes that no directory matches the naming scheme for files (for instance that you do not have a directory called "sources.ads" when using the default GNAT naming scheme).
When you do not have to use this switch (i.e., by default), gnatmake is able to save a lot of system calls (several per source file and object file), which can result in a significant speed up to load and manipulate a project file, especially when using source files from a remote system.
-eS-f-F-g-i-j`n'-kIf `gnatmake' is invoked with several file_names and with this
switch, if there are compilation errors when building an executable,
`gnatmake' will not attempt to build the following executables.
-l-m-Mstdout in a form that can be directly exploited in
a Makefile. By default, each source file is prefixed with its
(relative or absolute) directory name. This name is whatever you
specified in the various `-aI'
and `-I' switches. If you use
gnatmake -M `-q'
(see below), only the source file names,
without relative paths, are output. If you just specify the `-M'
switch, dependencies of the GNAT internal system files are omitted. This
is typically what you want. If you also specify
the `-a' switch,
dependencies of the GNAT internal files are also listed. Note that
dependencies of the objects in external Ada libraries (see
switch -aL`dir' in the following list)
are never reported.
-n-o `exec_name'This switch cannot be used when invoking `gnatmake' with several
file_names.
-p--create-missing-dirs
-P`project'-q-sThis switch is recommended when Integrated Preprocessing is used.
-u-U-v-vl-vm-vh-vP`x'-x-X`name'=`value'-zAny uppercase or multi-character switch that is not a `gnatmake' switch is passed to `gcc' (e.g., `-O', `-gnato,' etc.)
-aI`dir'-aL`dir'.ALI
files have been located in directory dir. This allows you to have
missing bodies for the units in dir and to ignore out of date bodies
for the same units. You still need to specify
the location of the specs for these units by using the switches
-aI`dir' or -I`dir'.
Note: this switch is provided for compatibility with previous versions
of `gnatmake'. The easier method of causing standard libraries
to be excluded from consideration is to write-protect the corresponding
ALI files.
-aO`dir'-I--L`dir'-largs -L`dir'.
Furthermore, under Windows, the sources pointed to by the libraries path
set in the registry are not searched for.
-nostdinc-nostdlib--RTS=`rts-path'adainclude or ada_source_path, and adalib or
ada_object_path present):
The mode switches (referred to as mode_switches) allow the inclusion of switches that are to be passed to the compiler itself, the binder or the linker. The effect of a mode switch is to cause all subsequent switches up to the end of the switch list, or up to the next mode switch, to be interpreted as switches to be passed on to the designated component of GNAT.
-cargs `switches'-bargs `switches'-largs `switches'-margs `switches'This section contains some additional useful notes on the operation of the `gnatmake' command.
gnatmake foo.adb, where foo
is a subunit or body of a generic unit, `gnatmake' recompiles
foo.adb (because it finds no ALI) and stops, issuing a
warning.
$ gnatmake -aI`include-dir` -aL`obj-dir` main
Generally `gnatmake' automatically performs all necessary recompilations and you don't need to worry about how it works. However, it may be useful to have some basic understanding of the `gnatmake' approach and in particular to understand how it uses the results of previous compilations without incorrectly depending on them.
First a definition: an object file is considered `up to date' if the corresponding ALI file exists and if all the source files listed in the dependency section of this ALI file have time stamps matching those in the ALI file. This means that neither the source file itself nor any files that it depends on have been modified, and hence there is no need to recompile this file.
`gnatmake' works by first checking if the specified main unit is up to date. If so, no compilations are required for the main unit. If not, `gnatmake' compiles the main program to build a new ALI file that reflects the latest sources. Then the ALI file of the main unit is examined to find all the source files on which the main program depends, and `gnatmake' recursively applies the above procedure on all these files.
This process ensures that `gnatmake' only trusts the dependencies in an existing ALI file if they are known to be correct. Otherwise it always recompiles to determine a new, guaranteed accurate set of dependencies. As a result the program is compiled 'upside down' from what may be more familiar as the required order of compilation in some other Ada systems. In particular, clients are compiled before the units on which they depend. The ability of GNAT to compile in any order is critical in allowing an order of compilation to be chosen that guarantees that `gnatmake' will recompute a correct set of new dependencies if necessary.
When invoking `gnatmake' with several file_names, if a unit is imported by several of the executables, it will be recompiled at most once.
Note: when using non-standard naming conventions (Using Other File Names), changing through a configuration pragmas file the version of a source and invoking `gnatmake' to recompile may have no effect, if the previous version of the source is still accessible by `gnatmake'. It may be necessary to use the switch -f.
hello.adb (containing unit Hello) and bind and link the
resulting object files to generate an executable file hello.
main1.adb (containing unit Main1), main2.adb
(containing unit Main2) and main3.adb
(containing unit Main3) and bind and link the resulting object files
to generate three executable files main1,
main2 and main3.
main_unit.adb). All compilations will
be done with optimization level 2 and the order of elaboration will be
listed by the binder. `gnatmake' will operate in quiet mode, not
displaying commands it is executing.
This section discusses how to compile Ada programs using the `gcc' command. It also describes the set of switches that can be used to control the behavior of the compiler.
The first step in creating an executable program is to compile the units of the program using the `gcc' command. You must compile the following files:
.adb) for a library level subprogram or generic
subprogram
.ads) for a library level package or generic
package that has no body
.adb) for a library level package
or generic package that has a body
You need `not' compile the following files
because they are compiled as part of compiling related units. GNAT package specs when the corresponding body is compiled, and subunits when the parent is compiled.
If you attempt to compile any of these files, you will get one of the following error messages (where fff is the name of the file you compiled):
cannot generate code for file `fff` (package spec)
to check package spec, use -gnatc
cannot generate code for file `fff` (missing subunits)
to check parent unit, use -gnatc
cannot generate code for file `fff` (subprogram spec)
to check subprogram spec, use -gnatc
cannot generate code for file `fff` (subunit)
to check subunit, use -gnatc
As indicated by the above error messages, if you want to submit one of these files to the compiler to check for correct semantics without generating code, then use the `-gnatc' switch.
The basic command for compiling a file containing an Ada unit is:
$ gcc -c [switches] <file name>
where file name is the name of the Ada file (usually
having an extension .ads for a spec or .adb for a body).
You specify the
-c switch to tell `gcc' to compile, but not link, the file.
The result of a successful compilation is an object file, which has the
same name as the source file but an extension of .o and an Ada
Library Information (ALI) file, which also has the same name as the
source file, but with .ali as the extension. GNAT creates these
two output files in the current directory, but you may specify a source
file in any directory using an absolute or relative path specification
containing the directory information.
`gcc' is actually a driver program that looks at the extensions of
the file arguments and loads the appropriate compiler. For example, the
GNU C compiler is cc1, and the Ada compiler is gnat1.
These programs are in directories known to the driver program (in some
configurations via environment variables you set), but need not be in
your path. The `gcc' driver also calls the assembler and any other
utilities needed to complete the generation of the required object
files.
It is possible to supply several file names on the same `gcc' command. This causes `gcc' to call the appropriate compiler for each file. For example, the following command lists two separate files to be compiled:
$ gcc -c x.adb y.adb
calls gnat1 (the Ada compiler) twice to compile x.adb and
y.adb.
The compiler generates two object files x.o and y.o
and the two ALI files x.ali and y.ali.
Any switches apply to all the files listed, see Compiler Switches for a list of available `gcc' switches.
With the GNAT source-based library system, the compiler must be able to find source files for units that are needed by the unit being compiled. Search paths are used to guide this process.
The compiler compiles one source file whose name must be given explicitly on the command line. In other words, no searching is done for this file. To find all other source files that are needed (the most common being the specs of units), the compiler examines the following directories, in the following order:
ADA_PRJ_INCLUDE_FILE environment variable.
ADA_PRJ_INCLUDE_FILE is normally set by gnatmake or by the gnat
driver when project files are used. It should not normally be set
by other means.
ADA_INCLUDE_PATH environment variable.
Construct this value
exactly as the
PATH environment variable: a list of directory
names separated by colons (semicolons when working with the NT version).
ada_source_path file which is part of the GNAT
installation tree and is used to store standard libraries such as the
GNAT Run Time Library (RTL) source files.
Installing a library
Specifying the switch `-I-' inhibits the use of the directory containing the source file named in the command line. You can still have this directory on your search path, but in this case it must be explicitly requested with a `-I' switch.
Specifying the switch `-nostdinc' inhibits the search of the default location for the GNAT Run Time Library (RTL) source files.
The compiler outputs its object files and ALI files in the current
working directory.
Caution: The object file can be redirected with the `-o' switch;
however, `gcc' and gnat1 have not been coordinated on this
so the ALI file will not go to the right place. Therefore, you should
avoid using the `-o' switch.
The packages Ada, System, and Interfaces and their children make up the GNAT RTL, together with the simple System.IO package used in the "Hello World" example. The sources for these units are needed by the compiler and are kept together in one directory. Not all of the bodies are needed, but all of the sources are kept together anyway. In a normal installation, you need not specify these directory names when compiling or binding. Either the environment variables or the built-in defaults cause these files to be found.
In addition to the language-defined hierarchies (System, Ada and Interfaces), the GNAT distribution provides a fourth hierarchy, consisting of child units of GNAT. This is a collection of generally useful types, subprograms, etc. See the GNAT_Reference_Manual for further details.
Besides simplifying access to the RTL, a major use of search paths is in compiling sources from multiple directories. This can make development environments much more flexible.
If, in our earlier example, there was a spec for the hello
procedure, it would be contained in the file hello.ads; yet this
file would not have to be explicitly compiled. This is the result of the
model we chose to implement library management. Some of the consequences
of this model are as follows:
The following are some typical Ada compilation command line examples:
$ gcc -c xyz.adb
Compile body in file xyz.adb with all default options.
$ gcc -c -O2 -gnata xyz-def.adb
Compile the child unit package in file xyz-def.adb with extensive
optimizations, and pragma Assert/Debug statements
enabled.
$ gcc -c -gnatc abc-def.adb
Compile the subunit in file abc-def.adb in semantic-checking-only
mode.
The `gcc' command accepts switches that control the compilation process. These switches are fully described in this section: first an alphabetical listing of all switches with a brief description, and then functionally grouped sets of switches with more detailed information.
More switches exist for GCC than those documented here, especially for specific targets. However, their use is not recommended as they may change code generation in ways that are incompatible with the Ada run-time library, or can cause inconsistencies between compilation units.
-b `target'-B`dir'-cNote: for some other languages when using `gcc', notably in the case of C and C++, it is possible to use use `gcc' without a `-c' switch to compile and link in one step. In the case of GNAT, you cannot use this approach, because the binder must be run and `gcc' cannot be used to run the GNAT binder.
-fcallgraph-info[=su,da]-fdump-scosSCOs
in the compiler sources for details in files scos.ads and
scos.adb.
-fdump-xref-flto[=`n']-fno-inline-fno-inline-functions-fno-inline-small-functions-fno-inline-functions-called-once-fno-ivopts-fno-strict-aliasing-fno-strict-overflow-fstack-check-fstack-usage-g-gnat05-gnat12-gnat2005-gnat2012-gnat83-gnat95Note: for compatibility with some Ada 95 compilers which support only the overriding keyword of Ada 2005, the `-gnatd.D' switch can be used along with `-gnat95' to achieve a similar effect with GNAT.
`-gnatd.D' instructs GNAT to consider overriding as a keyword and handle its associated semantic checks, even in Ada 95 mode.
-gnata-gnatAgnat.adc. If a gnat.adc file is present,
it will be ignored.
-gnatbstderr even if verbose mode set.
-gnatB-gnatc-gnatC.scil files). This switch is not compatible with
code generation (it will, among other things, disable some switches such
as -gnatn, and enable others such as -gnata).
-gnatddebug.adb for details of the implemented
debug options. Certain debug options are relevant to applications
programmers, and these are documented at appropriate points in this
users guide.
-gnatD-gnateA type Rec_Typ is record
Data : Integer := 0;
end record;
function Self (Val : Rec_Typ) return Rec_Typ is
begin
return Val;
end Self;
procedure Detect_Aliasing (Val_1 : in out Rec_Typ; Val_2 : Rec_Typ) is
begin
null;
end Detect_Aliasing;
Obj : Rec_Typ;
Detect_Aliasing (Obj, Obj);
Detect_Aliasing (Obj, Self (Obj));
In the example above, the first call to Detect_Aliasing fails with a Program_Error at runtime because the actuals for Val_1 and Val_2 denote the same object. The second call executes without raising an exception because Self(Obj) produces an anonymous object which does not share the memory location of Obj.
-gnatec=`path'-gnateC-gnated-gnateDsymbol[=`value']-gnateE-gnatef-gnateFThe -gnatc switch must always be specified before this switch, e.g. -gnatceg. Generate a C header from the Ada input file. See Generating C Headers for Ada Specifications for more information.
-gnateG-gnatei`nnn'-gnateI`nnn'-gnatel-gnateL-gnatem=`path'-gnatep=`file'-gnateP-gnateS-gnatet=`path'-gnateT=`path'The following target dependent values should be defined, where Nat denotes a natural integer value, Pos denotes a positive integer value, and fields marked with a question mark are boolean fields, where a value of 0 is False, and a value of 1 is True:
Bits_BE : Nat; -- Bits stored big-endian?
Bits_Per_Unit : Pos; -- Bits in a storage unit
Bits_Per_Word : Pos; -- Bits in a word
Bytes_BE : Nat; -- Bytes stored big-endian?
Char_Size : Pos; -- Standard.Character'Size
Double_Float_Alignment : Nat; -- Alignment of double float
Double_Scalar_Alignment : Nat; -- Alignment of double length scalar
Double_Size : Pos; -- Standard.Long_Float'Size
Float_Size : Pos; -- Standard.Float'Size
Float_Words_BE : Nat; -- Float words stored big-endian?
Int_Size : Pos; -- Standard.Integer'Size
Long_Double_Size : Pos; -- Standard.Long_Long_Float'Size
Long_Long_Size : Pos; -- Standard.Long_Long_Integer'Size
Long_Size : Pos; -- Standard.Long_Integer'Size
Maximum_Alignment : Pos; -- Maximum permitted alignment
Max_Unaligned_Field : Pos; -- Maximum size for unaligned bit field
Pointer_Size : Pos; -- System.Address'Size
Short_Enums : Nat; -- Short foreign convention enums?
Short_Size : Pos; -- Standard.Short_Integer'Size
Strict_Alignment : Nat; -- Strict alignment?
System_Allocator_Alignment : Nat; -- Alignment for malloc calls
Wchar_T_Size : Pos; -- Interfaces.C.wchar_t'Size
Words_BE : Nat; -- Words stored big-endian?
The format of the input file is as follows. First come the values of the variables defined above, with one line per value:
name value
where name is the name of the parameter, spelled out in full, and cased as in the above list, and value is an unsigned decimal integer. Two or more blanks separates the name from the value.
All the variables must be present, in alphabetical order (i.e. the same order as the list above).
Then there is a blank line to separate the two parts of the file. Then come the lines showing the floating-point types to be registered, with one line per registered mode:
name digs float_rep size alignment
where name is the string name of the type (which can have single spaces embedded in the name (e.g. long double), digs is the number of digits for the floating-point type, float_rep is the float representation (I/V/A for IEEE-754-Binary, Vax_Native, AAMP), size is the size in bits, alignment is the alignment in bits. The name is followed by at least two blanks, fields are separated by at least one blank, and a LF character immediately follows the alignment field.
Here is an example of a target parameterization file:
Bits_BE 0
Bits_Per_Unit 8
Bits_Per_Word 64
Bytes_BE 0
Char_Size 8
Double_Float_Alignment 0
Double_Scalar_Alignment 0
Double_Size 64
Float_Size 32
Float_Words_BE 0
Int_Size 64
Long_Double_Size 128
Long_Long_Size 64
Long_Size 64
Maximum_Alignment 16
Max_Unaligned_Field 64
Pointer_Size 64
Short_Size 16
Strict_Alignment 0
System_Allocator_Alignment 16
Wchar_T_Size 32
Words_BE 0
float 15 I 64 64
double 15 I 64 64
long double 18 I 80 128
TF 33 I 128 128
-gnateu-gnateV-gnateY-gnatE-gnatf-gnatF-gnatg-gnatG=nn-gnathstdout.
-gnati`c'-gnatINote that when -gnatct is used to generate trees for input into ASIS tools, these representation clauses are removed from the tree and ignored. This means that the tool will not see them.
-gnatj`nn'-gnatk=`n'-gnatl-gnatL-gnatm=`n'-gnatn[12]-gnatNWhen using a gcc-based back end (in practice this means using any version of GNAT other than the JGNAT, .NET or GNAAMP versions), then the use of `-gnatN' is deprecated, and the use of `-gnatn' is preferred. Historically front end inlining was more extensive than the gcc back end inlining, but that is no longer the case.
-gnato0-gnato??| Digit |
Interpretation |
| `1' |
All intermediate overflows checked against base type (STRICT) |
| `2' |
Minimize intermediate overflows (MINIMIZED) |
| `3' |
Eliminate intermediate overflows (ELIMINATED)
|
If only one digit appears, then it applies to all cases; if two digits are given, then the first applies outside assertions, pre/postconditions, and type invariants, and the second applies within assertions, pre/postconditions, and type invariants.
If no digits follow the `-gnato', then it is equivalent to `-gnato11', causing all intermediate overflows to be handled in strict mode.
This switch also causes arithmetic overflow checking to be performed (as though pragma Unsuppress (Overflow_Mode) had been specified).
The default if no option `-gnato' is given is that overflow handling is in STRICT mode (computations done using the base type), and that overflow checking is enabled.
Note that division by zero is a separate check that is not controlled by this switch (divide-by-zero checking is on by default).
See also Specifying the Desired Mode.
-gnatp-gnat-p-gnatP-gnatq-gnatQALI and tree files even if illegalities.
Note that code generation is still suppressed in the presence of any
errors, so even with `-gnatQ' no object file is generated.
-gnatr-gnatR[0/1/2/3[s]]-gnatRm[s]-gnats-gnatS-gnatt-gnatT`nnn'-gnatu-gnatU-gnatvstdout.
-gnatV-gnatw`xxx'-gnatW`e'-gnatx-gnatX-gnaty-gnatz`m'-I`dir'-I--o `file'-nostdinc-nostdlib-O[`n']| `n' |
Effect |
| `0' |
No optimization, the default setting if no `-O' appears |
| `1' |
Normal optimization, the default if you specify `-O' without an operand. A good compromise between code quality and compilation time. |
| `2' |
Extensive optimization, may improve execution time, possibly at the cost of substantially increased compilation time. |
| `3' |
Same as `-O2', and also includes inline expansion for small subprograms in the same unit. |
| `s' |
Optimize space usage
|
See also Optimization Levels.
-pass-exit-codes--RTS=`rts-path'-S.s as the extension,
instead of the object file.
This may be useful if you need to examine the generated assembly code.
-fverbose-asm-v-V `ver'-wYou may combine a sequence of GNAT switches into a single switch. For example, the combined switch
-gnatofi3
is equivalent to specifying the following sequence of switches:
-gnato -gnatf -gnati3
The following restrictions apply to the combination of switches in this manner:
The standard default format for error messages is called 'brief format'.
Brief format messages are written to stderr (the standard error
file) and have the following form:
e.adb:3:04: Incorrect spelling of keyword "function"
e.adb:4:20: ";" should be "is"
The first integer after the file name is the line number in the file, and the second integer is the column number within the line. GPS can parse the error messages and point to the referenced character. The following switches provide control over the error message format:
-gnatvstdout (the standard output file.
The same program compiled with the
`-gnatv' switch would generate:
3. funcion X (Q : Integer)
|
>>> Incorrect spelling of keyword "function"
4. return Integer;
|
>>> ";" should be "is"
The vertical bar indicates the location of the error, and the >>>
prefix can be used to search for error messages. When this switch is
used the only source lines output are those with errors.
-gnatlp.adb might look like:
Compiling: p.adb
1. package body p is
2. procedure a;
3. procedure a is separate;
4. begin
5. null
|
>>> missing ";"
6. end;
Compiling: p.ads
1. package p is
2. pragma Elaborate_Body
|
>>> missing ";"
3. end p;
Compiling: p-a.adb
1. separate p
|
>>> missing "("
2. procedure a is
3. begin
4. null
|
>>> missing ";"
5. end;
When you specify the `-gnatv' or `-gnatl' switches and
standard output is redirected, a brief summary is written to
stderr (standard error) giving the number of error messages and
warning messages generated.
-gnatl=`fname'fname does not start with a period, then it is the full name
of the file to be written. If fname is an extension, it is
appended to the name of the file being compiled. For example, if
file xyz.adb is compiled with `-gnatl=.lst',
then the output is written to file xyz.adb.lst.
-gnatU-gnatbstderr (the standard error
file) as well as the verbose
format message or full listing (which as usual is written to
stdout (the standard output file).
-gnatm=`n' e.adb:3:04: Incorrect spelling of keyword "function"
e.adb:5:35: missing ".."
fatal error: maximum number of errors detected
compilation abandoned
The default setting if no switch is given is 9999. If the number of warnings reaches this limit, then a message is output and further warnings are suppressed, but the compilation is continued. If the number of error messages reaches this limit, then a message is output and the compilation is abandoned. A value of zero means that no limit applies.
Note that the equal sign is optional, so the switches `-gnatm2' and `-gnatm=2' are equivalent.
-gnatfe.adb:7:07: "V" is undefined (more references follow)
where the parenthetical comment warns that there are additional references to the variable V. Compiling the same program with the `-gnatf' switch yields
e.adb:7:07: "V" is undefined
e.adb:8:07: "V" is undefined
e.adb:8:12: "V" is undefined
e.adb:8:16: "V" is undefined
e.adb:9:07: "V" is undefined
e.adb:9:12: "V" is undefined
The `-gnatf' switch also generates additional information for some error messages. Some examples are:
-gnatjnnIf the `-gnatjnn' switch is used with a positive value for nn, then messages are output in a different manner. A message and all its continuation lines are treated as a unit, and count as only one warning or message in the statistics totals. Furthermore, the message is reformatted so that no line is longer than nn characters.
-gnatq-gnatQALI file is not generated if any
illegalities are detected in the program. The use of `-gnatQ' forces
generation of the ALI file. This file is marked as being in
error, so it cannot be used for binding purposes, but it does contain
reasonably complete cross-reference information, and thus may be useful
for use by tools (e.g., semantic browsing tools or integrated development
environments) that are driven from the ALI file. This switch
implies `-gnatq', since the semantic phase must be run to get a
meaningful ALI file.
In addition, if `-gnatt' is also specified, then the tree file is generated even if there are illegalities. It may be useful in this case to also specify `-gnatq' to ensure that full semantic processing occurs. The resulting tree file can be processed by ASIS, for the purpose of providing partial information about illegal units, but if the error causes the tree to be badly malformed, then ASIS may crash during the analysis.
When `-gnatQ' is used and the generated ALI file is marked as
being in error, `gnatmake' will attempt to recompile the source when it
finds such an ALI file, including with switch `-gnatc'.
Note that `-gnatQ' has no effect if `-gnats' is specified, since ALI files are never generated if `-gnats' is set.
In addition to error messages, which correspond to illegalities as defined in the Ada Reference Manual, the compiler detects two kinds of warning situations.
First, the compiler considers some constructs suspicious and generates a warning message to alert you to a possible error. Second, if the compiler detects a situation that is sure to raise an exception at run time, it generates a warning message. The following shows an example of warning messages:
e.adb:4:24: warning: creation of object may raise Storage_Error
e.adb:10:17: warning: static value out of range
e.adb:10:17: warning: "Constraint_Error" will be raised at run time
GNAT considers a large number of situations as appropriate for the generation of warning messages. As always, warnings are not definite indications of errors. For example, if you do an out-of-range assignment with the deliberate intention of raising a Constraint_Error exception, then the warning that may be issued does not indicate an error. Some of the situations for which GNAT issues warnings (at least some of the time) are given in the following list. This list is not complete, and new warnings are often added to subsequent versions of GNAT. The list is intended to give a general idea of the kinds of warnings that are generated.
The following section lists compiler switches that are available to control the handling of warning messages. It is also possible to exercise much finer control over what warnings are issued and suppressed using the GNAT pragma Warnings (see the description of the pragma in the GNAT_Reference_manual).
-gnatwaThis switch activates most optional warning messages. See the remaining list in this section for details on optional warning messages that can be individually controlled. The warnings that are not turned on by this switch are:
-gnatwd (implicit dereferencing)
-gnatw.d (tag warnings with -gnatw switch)
-gnatwh (hiding)
-gnatw.h (holes in record layouts)
-gnatw.k (redefinition of names in standard)
-gnatwl (elaboration warnings)
-gnatw.l (inherited aspects)
-gnatw.n (atomic synchronization)
-gnatwo (address clause overlay)
-gnatw.o (values set by out parameters ignored)
-gnatw.s (overridden size clause)
-gnatwt (tracking of deleted conditional code)
-gnatw.u (unordered enumeration)
-gnatw.w (use of Warnings Off)
-gnatw.y (reasons for package needing body)
All other optional warnings are turned on.
-gnatwAThis switch suppresses all optional warning messages, see remaining list in this section for details on optional warning messages that can be individually controlled. Note that unlike switch `-gnatws', the use of switch `-gnatwA' does not suppress warnings that are normally given unconditionally and cannot be individually controlled (for example, the warning about a missing exit path in a function). Also, again unlike switch `-gnatws', warnings suppressed by the use of switch `-gnatwA' can be individually turned back on. For example the use of switch `-gnatwA' followed by switch `-gnatwd' will suppress all optional warnings except the warnings for implicit dereferencing.
-gnatw.aThis switch activates warnings for assertions where the compiler can tell at compile time that the assertion will fail. Note that this warning is given even if assertions are disabled. The default is that such warnings are generated.
-gnatw.AThis switch suppresses warnings for assertions where the compiler can tell at compile time that the assertion will fail.
-gnatwbThis switch activates warnings for static fixed-point expressions whose value is not an exact multiple of Small. Such values are implementation dependent, since an implementation is free to choose either of the multiples that surround the value. GNAT always chooses the closer one, but this is not required behavior, and it is better to specify a value that is an exact multiple, ensuring predictable execution. The default is that such warnings are not generated.
-gnatwBThis switch suppresses warnings for static fixed-point expressions whose value is not an exact multiple of Small.
-gnatw.bThis switch activates warnings when a size clause, value size clause, component clause, or component size clause forces the use of biased representation for an integer type (e.g. representing a range of 10..11 in a single bit by using 0/1 to represent 10/11). The default is that such warnings are generated.
-gnatw.BThis switch suppresses warnings for representation clauses that force the use of biased representation.
-gnatwcThis switch activates warnings for conditional expressions used in tests that are known to be True or False at compile time. The default is that such warnings are not generated. Note that this warning does not get issued for the use of boolean variables or constants whose values are known at compile time, since this is a standard technique for conditional compilation in Ada, and this would generate too many false positive warnings.
This warning option also activates a special test for comparisons using the operators '>=' and' <='. If the compiler can tell that only the equality condition is possible, then it will warn that the '>' or '<' part of the test is useless and that the operator could be replaced by '='. An example would be comparing a Natural variable <= 0.
This warning option also generates warnings if one or both tests is optimized away in a membership test for integer values if the result can be determined at compile time. Range tests on enumeration types are not included, since it is common for such tests to include an end point.
This warning can also be turned on using `-gnatwa'.
-gnatwCThis switch suppresses warnings for conditional expressions used in tests that are known to be True or False at compile time.
-gnatw.cThis switch activates warnings for record components where a record representation clause is present and has component clauses for the majority, but not all, of the components. A warning is given for each component for which no component clause is present.
-gnatw.CThis switch suppresses warnings for record components that are missing a component clause in the situation described above.
-gnatwdIf this switch is set, then the use of a prefix of an access type in an indexed component, slice, or selected component without an explicit .all will generate a warning. With this warning enabled, access checks occur only at points where an explicit .all appears in the source code (assuming no warnings are generated as a result of this switch). The default is that such warnings are not generated.
-gnatwDThis switch suppresses warnings for implicit dereferences in indexed components, slices, and selected components.
-gnatw.dIf this switch is set, then warning messages are tagged, with one of the following strings:
- `[-gnatw?]' Used to tag warnings controlled by the switch `-gnatwx' where x is a letter a-z.
- `[-gnatw.?]' Used to tag warnings controlled by the switch `-gnatw.x' where x is a letter a-z.
- `[-gnatel]' Used to tag elaboration information (info) messages generated when the static model of elaboration is used and the `-gnatel' switch is set.
- `[restriction warning]' Used to tag warning messages for restriction violations, activated by use of the pragma `Restriction_Warnings'.
- `[warning-as-error]' Used to tag warning messages that have been converted to error messages by use of the pragma Warning_As_Error. Note that such warnings are prefixed by the string "error: " rather than "warning: ".
- `[enabled by default]' Used to tag all other warnings that are always given by default, unless warnings are completely suppressed using pragma `Warnings(Off)' or the switch `-gnatws'.
-gnatw.DIf this switch is set, then warning messages return to the default mode in which warnings and info messages are not tagged as described above for -gnatw.d.
-gnatweThis switch causes warning messages and style check messages to be treated as errors. The warning string still appears, but the warning messages are counted as errors, and prevent the generation of an object file. Note that this is the only -gnatw switch that affects the handling of style check messages. Note also that this switch has no effect on info (information) messages, which are not treated as errors if this switch is present.
-gnatw.eThis switch activates all optional warnings, including those which are not activated by -gnatwa. The use of this switch is not recommended for normal use. If you turn this switch on, it is almost certain that you will get large numbers of useless warnings. The warnings that are excluded from -gnatwa are typically highly specialized warnings that are suitable for use only in code that has been specifically designed according to specialized coding rules.
-gnatwfThis switch causes a warning to be generated if a formal parameter is not referenced in the body of the subprogram. This warning can also be turned on using `-gnatwu'. The default is that these warnings are not generated.
-gnatwFThis switch suppresses warnings for unreferenced formal parameters. Note that the combination `-gnatwu' followed by `-gnatwF' has the effect of warning on unreferenced entities other than subprogram formals.
-gnatwgThis switch causes a warning to be generated if an unrecognized pragma is encountered. Apart from issuing this warning, the pragma is ignored and has no effect. The default is that such warnings are issued (satisfying the Ada Reference Manual requirement that such warnings appear).
-gnatwGThis switch suppresses warnings for unrecognized pragmas.
-gnatw.gThis switch sets the warning categories that are used by the standard GNAT style. Currently this is equivalent to `-gnatwAao.sI.C.V.X' but more warnings may be added in the future without advanced notice.
-gnatwhThis switch activates warnings on hiding declarations. A declaration is considered hiding if it is for a non-overloadable entity, and it declares an entity with the same name as some other entity that is directly or use-visible. The default is that such warnings are not generated.
-gnatwHThis switch suppresses warnings on hiding declarations.
-gnatw.hThis switch activates warnings on component clauses in record representation clauses that leave holes (gaps) in the record layout. If this warning option is active, then record representation clauses should specify a contiguous layout, adding unused fill fields if needed.
-gnatw.HThis switch suppresses warnings on component clauses in record representation clauses that leave holes (haps) in the record layout.
-gnatwiThis switch activates warnings for a `with' of an internal GNAT implementation unit, defined as any unit from the Ada, Interfaces, GNAT, or System hierarchies that is not documented in either the Ada Reference Manual or the GNAT Programmer's Reference Manual. Such units are intended only for internal implementation purposes and should not be `with'ed by user programs. The default is that such warnings are generated
-gnatwIThis switch disables warnings for a `with' of an internal GNAT implementation unit.
-gnatw.iThis switch enables a warning on statically detectable overlapping actuals in a subprogram call, when one of the actuals is an in-out parameter, and the types of the actuals are not by-copy types. This warning is off by default.
-gnatw.IThis switch disables warnings on overlapping actuals in a call..
-gnatwjIf this warning option is activated, then warnings are generated for calls to subprograms marked with pragma Obsolescent and for use of features in Annex J of the Ada Reference Manual. In the case of Annex J, not all features are flagged. In particular use of the renamed packages (like Text_IO) and use of package ASCII are not flagged, since these are very common and would generate many annoying positive warnings. The default is that such warnings are not generated.
In addition to the above cases, warnings are also generated for GNAT features that have been provided in past versions but which have been superseded (typically by features in the new Ada standard). For example, pragma Ravenscar will be flagged since its function is replaced by pragma Profile(Ravenscar), and pragma Interface_Name will be flagged since its function is replaced by pragma Import.
Note that this warning option functions differently from the restriction No_Obsolescent_Features in two respects. First, the restriction applies only to annex J features. Second, the restriction does flag uses of package ASCII.
-gnatwJThis switch disables warnings on use of obsolescent features.
-gnatwkThis switch activates warnings for variables that are initialized but never modified, and then could be declared constants. The default is that such warnings are not given.
-gnatwKThis switch disables warnings on variables that could be declared constants.
-gnatw.kThis switch activates warnings for declarations that declare a name that is defined in package Standard. Such declarations can be confusing, especially since the names in package Standard continue to be directly visible, meaning that use visibiliy on such redeclared names does not work as expected. Names of discriminants and components in records are not included in this check.
-gnatw.KThis switch activates warnings for declarations that declare a name that is defined in package Standard.
-gnatwlThis switch activates warnings for possible elaboration problems, including suspicious use of Elaborate pragmas, when using the static elaboration model, and possible situations that may raise Program_Error when using the dynamic elaboration model. See the section in this guide on elaboration checking for further details. The default is that such warnings are not generated.
-gnatwLThis switch suppresses warnings for possible elaboration problems.
-gnatw.lThis switch causes the compiler to list inherited invariants, preconditions, and postconditions from Type_Invariant'Class, Invariant'Class, Pre'Class, and Post'Class aspects. Also list inherited subtype predicates.
-gnatw.LThis switch suppresses listing of inherited aspects.
-gnatwmThis switch activates warnings for variables that are assigned (using an initialization value or with one or more assignment statements) but whose value is never read. The warning is suppressed for volatile variables and also for variables that are renamings of other variables or for which an address clause is given. The default is that these warnings are not given.
-gnatwMThis switch disables warnings for variables that are assigned or initialized, but never read.
-gnatw.mThis switch activates warnings for modulus values that seem suspicious. The cases caught are where the size is the same as the modulus (e.g. a modulus of 7 with a size of 7 bits), and modulus values of 32 or 64 with no size clause. The guess in both cases is that 2**x was intended rather than x. In addition expressions of the form 2*x for small x generate a warning (the almost certainly accurate guess being that 2**x was intended). The default is that these warnings are given.
-gnatw.MThis switch disables warnings for suspicious modulus values.
-gnatwnThis switch sets normal warning mode, in which enabled warnings are issued and treated as warnings rather than errors. This is the default mode. the switch `-gnatwn' can be used to cancel the effect of an explicit `-gnatws' or `-gnatwe'. It also cancels the effect of the implicit `-gnatwe' that is activated by the use of `-gnatg'.
-gnatw.nThis switch actives warnings when an access to an atomic variable requires the generation of atomic synchronization code. These warnings are off by default.
-gnatw.NThis switch suppresses warnings when an access to an atomic variable requires the generation of atomic synchronization code.
-gnatwoThis switch activates warnings for possibly unintended initialization effects of defining address clauses that cause one variable to overlap another. The default is that such warnings are generated.
-gnatwOThis switch suppresses warnings on possibly unintended initialization effects of defining address clauses that cause one variable to overlap another.
-gnatw.oThis switch activates warnings for variables that are modified by using them as actuals for a call to a procedure with an out mode formal, where the resulting assigned value is never read. It is applicable in the case where there is more than one out mode formal. If there is only one out mode formal, the warning is issued by default (controlled by -gnatwu). The warning is suppressed for volatile variables and also for variables that are renamings of other variables or for which an address clause is given. The default is that these warnings are not given.
-gnatw.OThis switch suppresses warnings for variables that are modified by using them as actuals for a call to a procedure with an out mode formal, where the resulting assigned value is never read.
-gnatwpThis switch activates warnings for failure of front end inlining (activated by `-gnatN') to inline a particular call. There are many reasons for not being able to inline a call, including most commonly that the call is too complex to inline. The default is that such warnings are not given. Warnings on ineffective inlining by the gcc back-end can be activated separately, using the gcc switch -Winline.
-gnatwPThis switch suppresses warnings on ineffective pragma Inlines. If the inlining mechanism cannot inline a call, it will simply ignore the request silently.
-gnatw.pThis switch activates warnings for cases of suspicious parameter ordering when the list of arguments are all simple identifiers that match the names of the formals, but are in a different order. The warning is suppressed if any use of named parameter notation is used, so this is the appropriate way to suppress a false positive (and serves to emphasize that the "misordering" is deliberate). The default is that such warnings are not given.
-gnatw.PThis switch suppresses warnings on cases of suspicious parameter ordering.
-gnatwqThis switch activates warnings for cases where parentheses are not used and the result is potential ambiguity from a readers point of view. For example (not a > b) when a and b are modular means ((not a) > b) and very likely the programmer intended (not (a > b)). Similarly (-x mod 5) means (-(x mod 5)) and quite likely ((-x) mod 5) was intended. In such situations it seems best to follow the rule of always parenthesizing to make the association clear, and this warning switch warns if such parentheses are not present. The default is that these warnings are given.
-gnatwQThis switch suppresses warnings for cases where the association is not clear and the use of parentheses is preferred.
-gnatwrThis switch activates warnings for redundant constructs. The following is the current list of constructs regarded as redundant:
The default is that warnings for redundant constructs are not given.
-gnatwRThis switch suppresses warnings for redundant constructs.
-gnatw.rThis switch activates warnings for an object renaming that renames a function call, which is equivalent to a constant declaration (as opposed to renaming the function itself). The default is that these warnings are given.
-gnatw.RThis switch suppresses warnings for object renaming function.
-gnatwsThis switch completely suppresses the output of all warning messages from the GNAT front end, including both warnings that can be controlled by switches described in this section, and those that are normally given unconditionally. The effect of this suppress action can only be cancelled by a subsequent use of the switch `-gnatwn'.
Note that switch `-gnatws' does not suppress warnings from the `gcc' back end. To suppress these back end warnings as well, use the switch `-w' in addition to `-gnatws'. Also this switch has no effect on the handling of style check messages.
-gnatw.sThis switch activates warnings on component clauses in record representation clauses where the length given overrides that specified by an explicit size clause for the component type. A warning is similarly given in the array case if a specified component size overrides an explicit size clause for the array component type.
-gnatw.SThis switch suppresses warnings on component clauses in record representation clauses that override size clauses, and similar warnings when an array component size overrides a size clause.
-gnatwtThis switch activates warnings for tracking of code in conditionals (IF and CASE statements) that is detected to be dead code which cannot be executed, and which is removed by the front end. This warning is off by default. This may be useful for detecting deactivated code in certified applications.
-gnatwTThis switch suppresses warnings for tracking of deleted conditional code.
-gnatw.tThis switch activates warnings on suspicious contracts. This includes warnings on suspicious postconditions (whether a pragma Postcondition or a Post aspect in Ada 2012) and suspicious contract cases (pragma or aspect Contract_Cases). A function postcondition or contract case is suspicious when no postcondition or contract case for this function mentions the result of the function. A procedure postcondition or contract case is suspicious when it only refers to the pre-state of the procedure, because in that case it should rather be expressed as a precondition. This switch also controls warnings on suspicious cases of expressions typically found in contracts like quantified expressions and uses of Update attribute. The default is that such warnings are generated.
-gnatw.TThis switch suppresses warnings on suspicious contracts.
-gnatwuThis switch activates warnings to be generated for entities that are declared but not referenced, and for units that are `with'ed and not referenced. In the case of packages, a warning is also generated if no entities in the package are referenced. This means that if a with'ed package is referenced but the only references are in use clauses or renames declarations, a warning is still generated. A warning is also generated for a generic package that is `with'ed but never instantiated. In the case where a package or subprogram body is compiled, and there is a `with' on the corresponding spec that is only referenced in the body, a warning is also generated, noting that the `with' can be moved to the body. The default is that such warnings are not generated. This switch also activates warnings on unreferenced formals (it includes the effect of `-gnatwf').
-gnatwUThis switch suppresses warnings for unused entities and packages. It also turns off warnings on unreferenced formals (and thus includes the effect of `-gnatwF').
-gnatw.uThis switch causes enumeration types to be considered as conceptually unordered, unless an explicit pragma Ordered is given for the type. The effect is to generate warnings in clients that use explicit comparisons or subranges, since these constructs both treat objects of the type as ordered. (A `client' is defined as a unit that is other than the unit in which the type is declared, or its body or subunits.) Please refer to the description of pragma Ordered in the GNAT Reference Manual for further details. The default is that such warnings are not generated.
-gnatw.UThis switch causes all enumeration types to be considered as ordered, so that no warnings are given for comparisons or subranges for any type.
-gnatwvThis switch activates warnings for access to variables which may not be properly initialized. The default is that such warnings are generated.
-gnatwVThis switch suppresses warnings for access to variables which may not be properly initialized. For variables of a composite type, the warning can also be suppressed in Ada 2005 by using a default initialization with a box. For example, if Table is an array of records whose components are only partially uninitialized, then the following code:
Tab : Table := (others => <>);
will suppress warnings on subsequent statements that access components of variable Tab.
-gnatw.vThis switch activates messages (labeled "info", they are not warnings, just informational messages) about the effects of non-default bit-order on records to which a component clause is applied. The effect of specifying non-default bit ordering is a bit subtle (and changed with Ada 2005), so these messages, which are given by default, are useful in understanding the exact consequences of using this feature.
-gnatw.VThis switch suppresses information messages for the effects of specifying non-default bit order on record components with component clauses.
-gnatwwThis switch activates warnings for indexing an unconstrained string parameter with a literal or S'Length. This is a case where the code is assuming that the low bound is one, which is in general not true (for example when a slice is passed). The default is that such warnings are generated.
-gnatwWThis switch suppresses warnings for indexing an unconstrained string parameter with a literal or S'Length. Note that this warning can also be suppressed in a particular case by adding an assertion that the lower bound is 1, as shown in the following example:
procedure K (S : String) is
pragma Assert (S'First = 1);
...
-gnatw.wThis switch activates warnings for use of pragma Warnings (Off, entity) where either the pragma is entirely useless (because it suppresses no warnings), or it could be replaced by pragma Unreferenced or pragma Unmodified. Also activates warnings for the case of Warnings (Off, String), where either there is no matching Warnings (On, String), or the Warnings (Off) did not suppress any warning. The default is that these warnings are not given.
-gnatw.WThis switch suppresses warnings for use of pragma Warnings (Off, ...).
-gnatwxThis switch activates warnings on Export/Import pragmas when the compiler detects a possible conflict between the Ada and foreign language calling sequences. For example, the use of default parameters in a convention C procedure is dubious because the C compiler cannot supply the proper default, so a warning is issued. The default is that such warnings are generated.
-gnatwXThis switch suppresses warnings on Export/Import pragmas. The sense of this is that you are telling the compiler that you know what you are doing in writing the pragma, and it should not complain at you.
-gnatw.xThis switch activates warnings for exception usage when pragma Restrictions
(No_Exception_Propagation) is in effect. Warnings are given for implicit or
explicit exception raises which are not covered by a local handler, and for
exception handlers which do not cover a local raise. The default is that these
warnings are not given.
-gnatw.XThis switch disables warnings for exception usage when pragma Restrictions (No_Exception_Propagation) is in effect.
-gnatwyFor the most part, newer versions of Ada are upwards compatible with older versions. For example, Ada 2005 programs will almost always work when compiled as Ada 2012. However there are some exceptions (for example the fact that some is now a reserved word in Ada 2012). This switch activates several warnings to help in identifying and correcting such incompatibilities. The default is that these warnings are generated. Note that at one point Ada 2005 was called Ada 0Y, hence the choice of character.
-gnatwYThis switch suppresses the warnings intended to help in identifying incompatibilities between Ada language versions.
-gnatw.yThere are a number of cases in which a package spec needs a body. For example, the use of pragma Elaborate_Body, or the declaration of a procedure specification requiring a completion. This switch causes information messages to be output showing why a package specification requires a body. This can be useful in the case of a large package specification which is unexpectedly requiring a body. The default is that such information messages are not output.
-gnatw.YThis switch suppresses the output of information messages showing why a package specification needs a body.
-gnatwzThis switch activates warnings for unchecked conversions where the types are known at compile time to have different sizes. The default is that such warnings are generated. Warnings are also generated for subprogram pointers with different conventions.
-gnatwZThis switch suppresses warnings for unchecked conversions where the types are known at compile time to have different sizes or conventions.
-gnatw.zThis switch activates warnings for cases of record types with specified Size and Alignment attributes where the size is not a multiple of the alignment, resulting in an object size that is greater than the specified size. The default is that such warnings are generated.
-gnatw.ZThis switch suppresses warnings for cases of record types with specified Size and Alignment attributes where the size is not a multiple of the alignment, resulting in an object size that is greater than the specified size. The warning can also be suppressed by giving an explicit Object_Size value.
-Wunused-Wuninitialized-Wstack-usage=`len'-Wall-w-WerrorA string of warning parameters can be used in the same parameter. For example:
-gnatwaGe
will turn on all optional warnings except for unrecognized pragma warnings, and also specify that warnings should be treated as errors.
When no switch `-gnatw' is used, this is equivalent to:
-gnatw.a-gnatwB-gnatw.b-gnatwC-gnatw.C-gnatwD-gnatwF-gnatwg-gnatwH-gnatwi-gnatw.I-gnatwJ-gnatwK-gnatwL-gnatw.L-gnatwM-gnatw.m-gnatwn-gnatwo-gnatw.O-gnatwP-gnatw.P-gnatwq-gnatwR-gnatw.R-gnatw.S-gnatwT-gnatw.T-gnatwU-gnatwv-gnatww-gnatw.W-gnatwx-gnatw.X-gnatwy-gnatwz
-gnatapragma Assertion_Policy (Check);
Which is a shorthand for:
pragma Assertion_Policy
(Assert => Check,
Static_Predicate => Check,
Dynamic_Predicate => Check,
Pre => Check,
Pre'Class => Check,
Post => Check,
Post'Class => Check,
Type_Invariant => Check,
Type_Invariant'Class => Check);
The pragmas Assert and Debug normally have no effect and
are ignored. This switch, where a stands for assert, causes
pragmas Assert and Debug to be activated. This switch also
causes preconditions, postconditions, subtype predicates, and
type invariants to be activated.
The pragmas have the form:
pragma Assert (<Boolean-expression> [, <static-string-expression>])
pragma Debug (<procedure call>)
pragma Type_Invariant (<type-local-name>, <Boolean-expression>)
pragma Predicate (<type-local-name>, <Boolean-expression>)
pragma Precondition (<Boolean-expression>, <string-expression>)
pragma Postcondition (<Boolean-expression>, <string-expression>)
The aspects have the form:
with [Pre|Post|Type_Invariant|Dynamic_Predicate|Static_Predicate]
=> <Boolean-expression>;
The Assert pragma causes Boolean-expression to be tested. If the result is True, the pragma has no effect (other than possible side effects from evaluating the expression). If the result is False, the exception Assert_Failure declared in the package System.Assertions is raised (passing static-string-expression, if present, as the message associated with the exception). If no string expression is given, the default is a string containing the file name and line number of the pragma.
The Debug pragma causes procedure to be called. Note that pragma Debug may appear within a declaration sequence, allowing debugging procedures to be called between declarations.
For the aspect specification, the <Boolean-expression> is evaluated. If the result is True, the aspect has no effect. If the result is False, the exception Assert_Failure is raised.
The Ada Reference Manual defines the concept of invalid values (see RM 13.9.1). The primary source of invalid values is uninitialized variables. A scalar variable that is left uninitialized may contain an invalid value; the concept of invalid does not apply to access or composite types.
It is an error to read an invalid value, but the RM does not require run-time checks to detect such errors, except for some minimal checking to prevent erroneous execution (i.e. unpredictable behavior). This corresponds to the `-gnatVd' switch below, which is the default. For example, by default, if the expression of a case statement is invalid, it will raise Constraint_Error rather than causing a wild jump, and if an array index on the left-hand side of an assignment is invalid, it will raise Constraint_Error rather than overwriting an arbitrary memory location.
The `-gnatVa' may be used to enable additional validity checks, which are not required by the RM. These checks are often very expensive (which is why the RM does not require them). These checks are useful in tracking down uninitialized variables, but they are not usually recommended for production builds, and in particular we do not recommend using these extra validity checking options in combination with optimization, since this can confuse the optimizer. If performance is a consideration, leading to the need to optimize, then the validity checking options should not be used.
The other `-gnatV'x switches below allow finer-grained
control; you can enable whichever validity checks you desire. However,
for most debugging purposes, `-gnatVa' is sufficient, and the
default `-gnatVd' (i.e. standard Ada behavior) is usually
sufficient for non-debugging use.
The `-gnatB' switch tells the compiler to assume that all values are valid (that is, within their declared subtype range) except in the context of a use of the Valid attribute. This means the compiler can generate more efficient code, since the range of values is better known at compile time. However, an uninitialized variable can cause wild jumps and memory corruption in this mode.
The `-gnatV'x switch allows control over the validity
checking mode as described below.
The x argument is a string of letters that
indicate validity checks that are performed or not performed in addition
to the default checks required by Ada as described above.
-gnatVaAll validity checks are turned on. That is, `-gnatVa' is equivalent to `gnatVcdfimorst'.
-gnatVcThe right hand side of assignments, and the initializing values of object declarations are validity checked.
-gnatVdSome validity checks are done by default following normal Ada semantics (RM 13.9.1 (9-11)). A check is done in case statements that the expression is within the range of the subtype. If it is not, Constraint_Error is raised. For assignments to array components, a check is done that the expression used as index is within the range. If it is not, Constraint_Error is raised. Both these validity checks may be turned off using switch `-gnatVD'. They are turned on by default. If `-gnatVD' is specified, a subsequent switch `-gnatVd' will leave the checks turned on. Switch `-gnatVD' should be used only if you are sure that all such expressions have valid values. If you use this switch and invalid values are present, then the program is erroneous, and wild jumps or memory overwriting may occur.
-gnatVeIn the absence of this switch, assignments to record or array components are not validity checked, even if validity checks for assignments generally (`-gnatVc') are turned on. In Ada, assignment of composite values do not require valid data, but assignment of individual components does. So for example, there is a difference between copying the elements of an array with a slice assignment, compared to assigning element by element in a loop. This switch allows you to turn off validity checking for components, even when they are assigned component by component.
-gnatVfIn the absence of this switch, validity checking occurs only for discrete values. If `-gnatVf' is specified, then validity checking also applies for floating-point values, and NaNs and infinities are considered invalid, as well as out of range values for constrained types. Note that this means that standard IEEE infinity mode is not allowed. The exact contexts in which floating-point values are checked depends on the setting of other options. For example, `-gnatVif' or `-gnatVfi' (the order does not matter) specifies that floating-point parameters of mode in should be validity checked.
-gnatViArguments for parameters of mode in are validity checked in function and procedure calls at the point of call.
-gnatVmArguments for parameters of mode in out are validity checked in procedure calls at the point of call. The 'm' here stands for modify, since this concerns parameters that can be modified by the call. Note that there is no specific option to test out parameters, but any reference within the subprogram will be tested in the usual manner, and if an invalid value is copied back, any reference to it will be subject to validity checking.
-gnatVnThis switch turns off all validity checking, including the default checking for case statements and left hand side subscripts. Note that the use of the switch `-gnatp' suppresses all run-time checks, including validity checks, and thus implies `-gnatVn'. When this switch is used, it cancels any other `-gnatV' previously issued.
-gnatVoArguments for predefined operators and attributes are validity checked.
This includes all operators in package Standard,
the shift operators defined as intrinsic in package Interfaces
and operands for attributes such as Pos. Checks are also made
on individual component values for composite comparisons, and on the
expressions in type conversions and qualified expressions. Checks are
also made on explicit ranges using .. (e.g., slices, loops etc).
-gnatVpThis controls the treatment of parameters within a subprogram (as opposed to `-gnatVi' and `-gnatVm' which control validity testing of parameters on a call. If either of these call options is used, then normally an assumption is made within a subprogram that the input arguments have been validity checking at the point of call, and do not need checking again within a subprogram). If `-gnatVp' is set, then this assumption is not made, and parameters are not assumed to be valid, so their validity will be checked (or rechecked) within the subprogram.
-gnatVrThe expression in return statements in functions is validity checked.
-gnatVsAll subscripts expressions are checked for validity, whether they appear on the right side or left side (in default mode only left side subscripts are validity checked).
-gnatVtExpressions used as conditions in if, while or exit statements are checked, as well as guard expressions in entry calls.
The `-gnatV' switch may be followed by a string of letters
to turn on a series of validity checking options.
For example, -gnatVcr
specifies that in addition to the default validity checking, copies and
function return expressions are to be validity checked.
In order to make it easier to specify the desired combination of effects,
the upper case letters CDFIMORST may
be used to turn off the corresponding lower case option.
Thus -gnatVaM turns on all validity checking options except for
checking of **in out** procedure arguments.
The specification of additional validity checking generates extra code (and in the case of `-gnatVa' the code expansion can be substantial). However, these additional checks can be very useful in detecting uninitialized variables, incorrect use of unchecked conversion, and other errors leading to invalid values. The use of pragma Initialize_Scalars is useful in conjunction with the extra validity checking, since this ensures that wherever possible uninitialized variables have invalid values.
See also the pragma Validity_Checks which allows modification of the validity checking mode at the program source level, and also allows for temporary disabling of validity checks.
The `-gnatyx' switch causes the compiler to enforce specified style rules. A limited set of style rules has been used in writing the GNAT sources themselves. This switch allows user programs to activate all or some of these checks. If the source program fails a specified style check, an appropriate message is given, preceded by the character sequence '(style)'. This message does not prevent successful compilation (unless the `-gnatwe' switch is used).
Note that this is by no means intended to be a general facility for checking arbitrary coding standards. It is simply an embedding of the style rules we have chosen for the GNAT sources. If you are starting a project which does not have established style standards, you may find it useful to adopt the entire set of GNAT coding standards, or some subset of them.
The string x is a sequence of letters or digits indicating the particular style checks to be performed. The following checks are defined:
-gnaty0If a digit from 1-9 appears in the string after `-gnaty' then proper indentation is checked, with the digit indicating the indentation level required. A value of zero turns off this style check. The general style of required indentation is as specified by the examples in the Ada Reference Manual. Full line comments must be aligned with the – starting on a column that is a multiple of the alignment level, or they may be aligned the same way as the following non-blank line (this is useful when full line comments appear in the middle of a statement, or they may be aligned with the source line on the previous non-blank line.
-gnatyaAttribute names, including the case of keywords such as digits used as attributes names, must be written in mixed case, that is, the initial letter and any letter following an underscore must be uppercase. All other letters must be lowercase.
-gnatyAWhen using the array attributes First, Last, Range, or Length, the index number must be omitted for one-dimensional arrays and is required for multi-dimensional arrays.
-gnatybTrailing blanks are not allowed at the end of statements. The purpose of this rule, together with h (no horizontal tabs), is to enforce a canonical format for the use of blanks to separate source tokens.
-gnatyBThe use of AND/OR operators is not permitted except in the cases of modular operands, array operands, and simple stand-alone boolean variables or boolean constants. In all other cases and then/or else are required.
-gnatycComments must meet the following set of rules:
---------------------------
-- This is a box comment --
-- with two text lines. --
---------------------------
-gnatyCThis is identical to c except that only one space is required following the – of a comment instead of two.
-gnatydAll lines must be terminated by a single ASCII.LF character (in particular the DOS line terminator sequence CR/LF is not allowed).
-gnatyeOptional labels on end statements ending subprograms and on exit statements exiting named loops, are required to be present.
-gnatyfNeither form feeds nor vertical tab characters are permitted in the source text.
-gnatygThe set of style check switches is set to match that used by the GNAT sources. This may be useful when developing code that is eventually intended to be incorporated into GNAT. Currently this is equivalent to `-gnatwydISux') but additional style switches may be added to this set in the future without advance notice.
-gnatyhHorizontal tab characters are not permitted in the source text. Together with the b (no blanks at end of line) check, this enforces a canonical form for the use of blanks to separate source tokens.
-gnatyiThe keyword then must appear either on the same line as corresponding if, or on a line on its own, lined up under the if.
-gnatyIMode in (the default mode) is not allowed to be given explicitly. in out is fine, but not in on its own.
-gnatykAll keywords must be in lower case (with the exception of keywords such as digits used as attribute names to which this check does not apply).
-gnatylLayout of statement and declaration constructs must follow the recommendations in the Ada Reference Manual, as indicated by the form of the syntax rules. For example an else keyword must be lined up with the corresponding if keyword.
There are two respects in which the style rule enforced by this check option are more liberal than those in the Ada Reference Manual. First in the case of record declarations, it is permissible to put the record keyword on the same line as the type keyword, and then the end in end record must line up under type. This is also permitted when the type declaration is split on two lines. For example, any of the following three layouts is acceptable:
type q is record
a : integer;
b : integer;
end record;
type q is
record
a : integer;
b : integer;
end record;
type q is
record
a : integer;
b : integer;
end record;
Second, in the case of a block statement, a permitted alternative is to put the block label on the same line as the declare or begin keyword, and then line the end keyword up under the block label. For example both the following are permitted:
Block : declare
A : Integer := 3;
begin
Proc (A, A);
end Block;
Block :
declare
A : Integer := 3;
begin
Proc (A, A);
end Block;
The same alternative format is allowed for loops. For example, both of the following are permitted:
Clear : while J < 10 loop
A (J) := 0;
end loop Clear;
Clear :
while J < 10 loop
A (J) := 0;
end loop Clear;
-gnatyLThe maximum level of nesting of constructs (including subprograms, loops, blocks, packages, and conditionals) may not exceed the given value `nnn'. A value of zero disconnects this style check.
-gnatymThe length of source lines must not exceed 79 characters, including any trailing blanks. The value of 79 allows convenient display on an 80 character wide device or window, allowing for possible special treatment of 80 character lines. Note that this count is of characters in the source text. This means that a tab character counts as one character in this count and a wide character sequence counts as a single character (however many bytes are needed in the encoding).
-gnatyMThe length of lines must not exceed the given value `nnn'. The maximum value that can be specified is 32767. If neither style option for setting the line length is used, then the default is 255. This also controls the maximum length of lexical elements, where the only restriction is that they must fit on a single line.
-gnatynAny identifier from Standard must be cased to match the presentation in the Ada Reference Manual (for example, Integer and ASCII.NUL).
-gnatyNAll style check options are turned off.
-gnatyoAll subprogram bodies in a given scope (e.g., a package body) must be in alphabetical order. The ordering rule uses normal Ada rules for comparing strings, ignoring casing of letters, except that if there is a trailing numeric suffix, then the value of this suffix is used in the ordering (e.g., Junk2 comes before Junk10).
-gnatyOThis applies to all subprograms of a derived type that override a primitive operation of the type, for both tagged and untagged types. In particular, the declaration of a primitive operation of a type extension that overrides an inherited operation must carry an overriding indicator. Another case is the declaration of a function that overrides a predefined operator (such as an equality operator).
-gnatypPragma names must be written in mixed case, that is, the initial letter and any letter following an underscore must be uppercase. All other letters must be lowercase. An exception is that SPARK_Mode is allowed as an alternative for Spark_Mode.
-gnatyrAll identifier references must be cased in the same way as the corresponding declaration. No specific casing style is imposed on identifiers. The only requirement is for consistency of references with declarations.
-gnatysSeparate declarations ('specs') are required for subprograms (a body is not allowed to serve as its own declaration). The only exception is that parameterless library level procedures are not required to have a separate declaration. This exception covers the most frequent form of main program procedures.
-gnatySNo statements are allowed on the same line as a then or else keyword following the keyword in an if statement. or else and and then are not affected, and a special exception allows a pragma to appear after else.
-gnatytThe following token spacing rules are enforced:
Exactly one blank (and no other white space) must appear between a not token and a following in token.
-gnatyuUnnecessary blank lines are not allowed. A blank line is considered unnecessary if it appears at the end of the file, or if more than one blank line occurs in sequence.
-gnatyxUnnecessary extra level of parentheses (C-style) are not allowed around conditions in if statements, while statements and exit statements.
-gnatyyThis is equivalent to gnaty3aAbcefhiklmnprst, that is all checking options enabled with the exception of `-gnatyB', `-gnatyd', `-gnatyI', `-gnatyLnnn', `-gnatyo', `-gnatyO', `-gnatyS', `-gnatyu', and `-gnatyx'.
-gnaty-This causes any subsequent options in the string to act as canceling the corresponding style check option. To cancel maximum nesting level control, use `L' parameter witout any integer value after that, because any digit following `-' in the parameter string of the `-gnaty' option will be threated as canceling indentation check. The same is true for `M' parameter. `y' and `N' parameters are not allowed after `-'.
-gnaty+This causes any subsequent options in the string to enable the corresponding style check option. That is, it cancels the effect of a previous -, if any.
In the above rules, appearing in column one is always permitted, that is, counts as meeting either a requirement for a required preceding space, or as meeting a requirement for no preceding space.
Appearing at the end of a line is also always permitted, that is, counts as meeting either a requirement for a following space, or as meeting a requirement for no following space.
If any of these style rules is violated, a message is generated giving details on the violation. The initial characters of such messages are always '(style)'. Note that these messages are treated as warning messages, so they normally do not prevent the generation of an object file. The `-gnatwe' switch can be used to treat warning messages, including style messages, as fatal errors.
The switch -gnaty on its own (that is not
followed by any letters or digits) is equivalent
to the use of `-gnatyy' as described above, that is all
built-in standard style check options are enabled.
The switch -gnatyN clears any previously set style checks.
By default, the following checks are suppressed: stack overflow checks, and checks for access before elaboration on subprogram calls. All other checks, including overflow checks, range checks and array bounds checks, are turned on by default. The following `gcc' switches refine this default behavior.
-gnatpNote that when checks are suppressed, the compiler is allowed, but not required, to omit the checking code. If the run-time cost of the checking code is zero or near-zero, the compiler will generate it even if checks are suppressed. In particular, if the compiler can prove that a certain check will necessarily fail, it will generate code to do an unconditional 'raise', even if checks are suppressed. The compiler warns in this case. Another case in which checks may not be eliminated is when they are embedded in certain run time routines such as math library routines.
Of course, run-time checks are omitted whenever the compiler can prove that they will not fail, whether or not checks are suppressed.
Note that if you suppress a check that would have failed, program execution is erroneous, which means the behavior is totally unpredictable. The program might crash, or print wrong answers, or do anything else. It might even do exactly what you wanted it to do (and then it might start failing mysteriously next week or next year). The compiler will generate code based on the assumption that the condition being checked is true, which can result in erroneous execution if that assumption is wrong.
The checks subject to suppression include all the checks defined by the Ada standard, the additional implementation defined checks Alignment_Check, Duplicated_Tag_Check, Predicate_Check, Container_Checks, Tampering_Check, and Validity_Check, as well as any checks introduced using pragma Check_Name. Note that Atomic_Synchronization is not automatically suppressed by use of this option.
If the code depends on certain checks being active, you can use pragma Unsuppress either as a configuration pragma or as a local pragma to make sure that a specified check is performed even if `gnatp' is specified.
The `-gnatp' switch has no effect if a subsequent `-gnat-p' switch appears.
-gnat-p-gnato??Overflow checks are always enabled by this switch. The argument controls the mode, using the codes
If two digits are present after `-gnato' then the first digit sets the mode for expressions outside assertions, and the second digit sets the mode for expressions within assertions. Here assertions is used in the technical sense (which includes for example precondition and postcondition expressions).
If one digit is present, the corresponding mode is applicable to both expressions within and outside assertion expressions.
If no digits are present, the default is to enable overflow checks and set STRICT mode for both kinds of expressions. This is compatible with the use of `-gnato' in previous versions of GNAT.
Note that the `-gnato??' switch does not affect the code generated for any floating-point operations; it applies only to integer semantics. For floating-point, GNAT has the Machine_Overflows attribute set to False and the normal mode of operation is to generate IEEE NaN and infinite values on overflow or invalid operations (such as dividing 0.0 by 0.0).
The reason that we distinguish overflow checking from other kinds of range constraint checking is that a failure of an overflow check, unlike for example the failure of a range check, can result in an incorrect value, but cannot cause random memory destruction (like an out of range subscript), or a wild jump (from an out of range case value). Overflow checking is also quite expensive in time and space, since in general it requires the use of double length arithmetic.
Note again that the default is `-gnato11' (equivalent to `-gnato1'), so overflow checking is performed in STRICT mode by default.
-gnatE-fstack-checkThe setting of these switches only controls the default setting of the checks. You may modify them using either Suppress (to remove checks) or Unsuppress (to add back suppressed checks) pragmas in the program source.
-gnatsRun GNAT in syntax checking only mode. For example, the command
$ gcc -c -gnats x.adb
compiles file x.adb in syntax-check-only mode. You can check a
series of files in a single command
, and can use wild cards to specify such a group of files.
Note that you must specify the `-c' (compile
only) flag in addition to the `-gnats' flag.
You may use other switches in conjunction with `-gnats'. In particular, `-gnatl' and `-gnatv' are useful to control the format of any generated error messages.
When the source file is empty or contains only empty lines and/or comments, the output is a warning:
$ gcc -c -gnats -x ada toto.txt
toto.txt:1:01: warning: empty file, contains no compilation units
$
Otherwise, the output is simply the error messages, if any. No object file or ALI file is generated by a syntax-only compilation. Also, no units other than the one specified are accessed. For example, if a unit X `with's a unit Y, compiling unit X in syntax check only mode does not access the source file containing unit Y.
Normally, GNAT allows only a single unit in a source file. However, this restriction does not apply in syntax-check-only mode, and it is possible to check a file containing multiple compilation units concatenated together. This is primarily used by the gnatchop utility (Renaming Files with gnatchop).
-gnatcBecause dependent files must be accessed, you must follow the GNAT semantic restrictions on file structuring to operate in this mode:
The output consists of error messages as appropriate. No object file is
generated. An ALI file is generated for use in the context of
cross-reference tools, but this file is marked as not being suitable
for binding (since no object file is generated).
The checking corresponds exactly to the notion of
legality in the Ada Reference Manual.
Any unit can be compiled in semantics-checking-only mode, including units that would not normally be compiled (subunits, and specifications where a separate body is present).
The switches described in this section allow you to explicitly specify the version of the Ada language that your programs are written in. The default mode is Ada 2012, but you can also specify Ada 95, Ada 2005 mode, or indicate Ada 83 compatibility mode.
-gnat83 (Ada 83 Compatibility Mode)With few exceptions (most notably the need to use <> on unconstrained generic formal parameters, the use of the new Ada 95 / Ada 2005 reserved words, and the use of packages with optional bodies), it is not necessary to specify the `-gnat83' switch when compiling Ada 83 programs, because, with rare exceptions, Ada 95 and Ada 2005 are upwardly compatible with Ada 83. Thus a correct Ada 83 program is usually also a correct program in these later versions of the language standard. For further information please refer to the Compatibility_and_Porting_Guide chapter in the GNAT Reference Manual.
-gnat95 (Ada 95 mode)This switch also can be used to cancel the effect of a previous `-gnat83', `-gnat05/2005', or `-gnat12/2012' switch earlier in the command line.
-gnat05 or -gnat2005 (Ada 2005 mode)-gnat12 or -gnat2012 (Ada 2012 mode)-gnatX (Enable GNAT Extensions)-gnati`c'| `1' |
ISO 8859-1 (Latin-1) identifiers |
| `2' |
ISO 8859-2 (Latin-2) letters allowed in identifiers |
| `3' |
ISO 8859-3 (Latin-3) letters allowed in identifiers |
| `4' |
ISO 8859-4 (Latin-4) letters allowed in identifiers |
| `5' |
ISO 8859-5 (Cyrillic) letters allowed in identifiers |
| `9' |
ISO 8859-15 (Latin-9) letters allowed in identifiers |
| `p' |
IBM PC letters (code page 437) allowed in identifiers |
| `8' |
IBM PC letters (code page 850) allowed in identifiers |
| `f' |
Full upper-half codes allowed in identifiers |
| `n' |
No upper-half codes allowed in identifiers |
| `w' |
Wide-character codes (that is, codes greater than 255)
allowed in identifiers
|
See Foreign Language Representation for full details on the implementation of these character sets.
-gnatW`e'| `h' |
Hex encoding (brackets coding also recognized) |
| `u' |
Upper half encoding (brackets encoding also recognized) |
| `s' |
Shift/JIS encoding (brackets encoding also recognized) |
| `e' |
EUC encoding (brackets encoding also recognized) |
| `8' |
UTF-8 encoding (brackets encoding also recognized) |
| `b' |
Brackets encoding only (default value)
|
For full details on these encoding methods see Wide_Character Encodings. Note that brackets coding is always accepted, even if one of the other options is specified, so for example `-gnatW8' specifies that both brackets and UTF-8 encodings will be recognized. The units that are with'ed directly or indirectly will be scanned using the specified representation scheme, and so if one of the non-brackets scheme is used, it must be used consistently throughout the program. However, since brackets encoding is always recognized, it may be conveniently used in standard libraries, allowing these libraries to be used with any of the available coding schemes.
Note that brackets encoding only applies to program text. Within comments, brackets are considered to be normal graphic characters, and bracket sequences are never recognized as wide characters.
If no `-gnatW?' parameter is present, then the default representation is normally Brackets encoding only. However, if the first three characters of the file are 16#EF# 16#BB# 16#BF# (the standard byte order mark or BOM for UTF-8), then these three characters are skipped and the default representation for the file is set to UTF-8.
Note that the wide character representation that is specified (explicitly or by default) for the main program also acts as the default encoding used for Wide_Text_IO files if not specifically overridden by a WCEM form parameter.
When no `-gnatW?' is specified, then characters (other than wide characters represented using brackets notation) are treated as 8-bit Latin-1 codes. The codes recognized are the Latin-1 graphic characters, and ASCII format effectors (CR, LF, HT, VT). Other lower half control characters in the range 16#00#..16#1F# are not accepted in program text or in comments. Upper half control characters (16#80#..16#9F#) are rejected in program text, but allowed and ignored in comments. Note in particular that the Next Line (NEL) character whose encoding is 16#85# is not recognized as an end of line in this default mode. If your source program contains instances of the NEL character used as a line terminator, you must use UTF-8 encoding for the whole source program. In default mode, all lines must be ended by a standard end of line sequence (CR, CR/LF, or LF).
Note that the convention of simply accepting all upper half characters in comments means that programs that use standard ASCII for program text, but UTF-8 encoding for comments are accepted in default mode, providing that the comments are ended by an appropriate (CR, or CR/LF, or LF) line terminator. This is a common mode for many programs with foreign language comments.
-gnatk`n'.ads or .adb extension). The default is not
to enable file name krunching.
For the source file naming rules, File Naming Rules.
-gnatn[12]You can optionally specify the inlining level: 1 for moderate inlining across modules, which is a good compromise between compilation times and performances at run time, or 2 for full inlining across modules, which may bring about longer compilation times. If no inlining level is specified, the compiler will pick it based on the optimization level: 1 for `-O1', `-O2' or `-Os' and 2 for `-O3'.
If you specify this switch the compiler will access these bodies, creating an extra source dependency for the resulting object file, and where possible, the call will be inlined. For further details on when inlining is possible see Inlining of Subprograms.
-gnatNWhen using a gcc-based back end (in practice this means using any version of GNAT other than the JGNAT, .NET or GNAAMP versions), then the use of `-gnatN' is deprecated, and the use of `-gnatn' is preferred. Historically front end inlining was more extensive than the gcc back end inlining, but that is no longer the case.
-gnatt.adt.
This not normally required, but is used by separate analysis tools.
Typically
these tools do the necessary compilations automatically, so you should
not have to specify this switch in normal operation.
Note that the combination of switches `-gnatct'
generates a tree in the form required by ASIS applications.
-gnatustdout.
The listing includes all units on which the unit being compiled depends
either directly or indirectly.
-pass-exit-codesWhen `-pass-exit-codes' is used, `gcc' exits with an extended exit status and allows an integrated development environment to better react to a compilation failure. Those exit status are:
| `5' |
There was an error in at least one source file. |
| `3' |
At least one source file did not generate an object file. |
| `2' |
The compiler died unexpectedly (internal error for example). |
| `0' |
An object file has been generated for every source file.
|
-gnatd`x'debug.adb.
-gnatG[=`nn']The optional parameter nn if present after -gnatG specifies an alternative maximum line length that overrides the normal default of 72. This value is in the range 40-999999, values less than 40 being silently reset to 40. The equal sign is optional.
The format of the output is very similar to standard Ada source, and is
easily understood by an Ada programmer. The following special syntactic
additions correspond to low level features used in the generated code that
do not have any exact analogies in pure Ada source form. The following
is a partial list of these special constructions. See the spec
of package Sprint in file sprint.ads for a full list.
If the switch `-gnatL' is used in conjunction with `-gnatG', then the original source lines are interspersed in the expanded source (as comment lines with the original line number).
new `xxx' [storage_pool = `yyy']at end `procedure-name';(if `expr' then `expr' else `expr')`target'^(`source')`target'?(`source')`target'?^(`source')`x' #/ `y'
`x' #mod `y'
`x' # `y'
`x' #rem `y'free `expr' [storage_pool = `xxx'][subtype or type declaration]freeze `type-name' [`actions']reference `itype'`function-name'! (`arg', `arg', `arg')`label-name' : label#$ `subprogram-name'`expr' && `expr' && `expr' ... && `expr'[constraint_error]`expression''reference`target-type'!(`source-expression')[`numerator'/`denominator']-gnatD[=nn]xxx.dg, where xxx is the normal file name,
instead of to the standard output file. For
example, if the source file name is hello.adb, then a file
hello.adb.dg will be written. The debugging
information generated by the `gcc' `-g' switch
will refer to the generated xxx.dg file. This allows
you to do source level debugging using the generated code which is
sometimes useful for complex code, for example to find out exactly
which part of a complex construction raised an exception. This switch
also suppress generation of cross-reference information (see
`-gnatx') since otherwise the cross-reference information
would refer to the .dg file, which would cause
confusion since this is not the original source file.
Note that `-gnatD' actually implies `-gnatG' automatically, so it is not necessary to give both options. In other words `-gnatD' is equivalent to `-gnatDG').
If the switch `-gnatL' is used in conjunction with `-gnatDG', then the original source lines are interspersed in the expanded source (as comment lines with the original line number).
The optional parameter nn if present after -gnatD specifies an alternative maximum line length that overrides the normal default of 72. This value is in the range 40-999999, values less than 40 being silently reset to 40. The equal sign is optional.
-gnatr-gnatR[0|1|2|3[s]]Finally `-gnatR3' includes symbolic
expressions for values that are computed at run time for
variant records. These symbolic expressions have a mostly obvious
format with #n being used to represent the value of the n'th
discriminant. See source files repinfo.ads/adb in the
GNAT sources for full details on the format of `-gnatR3'
output. If the switch is followed by an s (e.g., `-gnatR2s'), then
the output is to a file with the name file.rep where
file is the name of the corresponding source file.
-gnatRm[s]Note that it is possible for record components to have zero size. In this case, the component clause uses an obvious extension of permitted Ada syntax, for example at 0 range 0 .. -1.
Representation information requires that code be generated (since it is the code generator that lays out complex data structures). If an attempt is made to output representation information when no code is generated, for example when a subunit is compiled on its own, then no information can be generated and the compiler outputs a message to this effect.
-gnatS-gnatxALI file. This information is used by a number of tools,
including gnatfind and gnatxref. The `-gnatx' switch
suppresses this information. This saves some space and may slightly
speed up compilation, but means that these tools cannot be used.
GNAT uses two methods for handling exceptions at run-time. The setjmp/longjmp method saves the context when entering a frame with an exception handler. Then when an exception is raised, the context can be restored immediately, without the need for tracing stack frames. This method provides very fast exception propagation, but introduces significant overhead for the use of exception handlers, even if no exception is raised.
The other approach is called 'zero cost' exception handling. With this method, the compiler builds static tables to describe the exception ranges. No dynamic code is required when entering a frame containing an exception handler. When an exception is raised, the tables are used to control a back trace of the subprogram invocation stack to locate the required exception handler. This method has considerably poorer performance for the propagation of exceptions, but there is no overhead for exception handlers if no exception is raised. Note that in this mode and in the context of mixed Ada and C/C++ programming, to propagate an exception through a C/C++ code, the C/C++ code must be compiled with the `-funwind-tables' GCC's option.
The following switches may be used to control which of the two exception handling methods is used.
--RTS=sjlj--RTS=zcxThe same option `–RTS' must be used both for `gcc' and `gnatbind'. Passing this option to `gnatmake' (Switches for gnatmake) will ensure the required consistency through the compilation and binding steps.
-gnatem=`path'The use of mapping files is not required for correct operation of the compiler, but mapping files can improve efficiency, particularly when sources are read over a slow network connection. In normal operation, you need not be concerned with the format or use of mapping files, and the `-gnatem' switch is not a switch that you would use explicitly. It is intended primarily for use by automatic tools such as `gnatmake' running under the project file facility. The description here of the format of mapping files is provided for completeness and for possible use by other tools.
A mapping file is a sequence of sets of three lines. In each set, the first line is the unit name, in lower case, with %s appended for specs and %b appended for bodies; the second line is the file name; and the third line is the path name.
Example:
main%b
main.2.ada
/gnat/project1/sources/main.2.ada
When the switch `-gnatem' is specified, the compiler will create in memory the two mappings from the specified file. If there is any problem (nonexistent file, truncated file or duplicate entries), no mapping will be created.
Several `-gnatem' switches may be specified; however, only the last one on the command line will be taken into account.
When using a project file, `gnatmake' creates a temporary mapping file and communicates it to the compiler using this switch.
The GCC technology provides a wide range of target dependent
-m switches for controlling
details of code generation with respect to different versions of
architectures. This includes variations in instruction sets (e.g.,
different members of the power pc family), and different requirements
for optimal arrangement of instructions (e.g., different members of
the x86 family). The list of available `-m' switches may be
found in the GCC documentation.
Use of these `-m' switches may in some cases result in improved code performance.
The GNAT technology is tested and qualified without any
-m switches,
so generally the most reliable approach is to avoid the use of these
switches. However, we generally expect most of these switches to work
successfully with GNAT, and many customers have reported successful
use of these options.
Our general advice is to avoid the use of `-m' switches unless special needs lead to requirements in this area. In particular, there is no point in using `-m' switches to improve performance unless you actually see a performance improvement.
This chapter describes the GNAT binder, gnatbind, which is used to bind compiled GNAT objects.
Note: to invoke gnatbind with a project file, use the gnat driver (see The GNAT Driver and Project Files).
The gnatbind program performs four separate functions:
The form of the gnatbind command is
$ gnatbind [`switches`] `mainprog`[.ali] [`switches`]
where mainprog.adb is the Ada file containing the main program
unit body. gnatbind constructs an Ada
package in two files whose names are
b~mainprog.ads, and b~mainprog.adb.
For example, if given the
parameter hello.ali, for a main program contained in file
hello.adb, the binder output files would be b~hello.ads
and b~hello.adb.
When doing consistency checking, the binder takes into consideration
any source files it can locate. For example, if the binder determines
that the given main program requires the package Pack, whose
.ALI
file is pack.ali and whose corresponding source spec file is
pack.ads, it attempts to locate the source file pack.ads
(using the same search path conventions as previously described for the
`gcc' command). If it can locate this source file, it checks that
the time stamps
or source checksums of the source and its references to in ALI files
match. In other words, any ALI files that mentions this spec must have
resulted from compiling this version of the source file (or in the case
where the source checksums match, a version close enough that the
difference does not matter).
The effect of this consistency checking, which includes source files, is that the binder ensures that the program is consistent with the latest version of the source files that can be located at bind time. Editing a source file without compiling files that depend on the source file cause error messages to be generated by the binder.
For example, suppose you have a main program hello.adb and a
package P, from file p.ads and you perform the following
steps:
p.ads.
At this point, the file p.ali contains an out-of-date time stamp
because the file p.ads has been edited. The attempt at binding
fails, and the binder generates the following error messages:
error: "hello.adb" must be recompiled ("p.ads" has been modified)
error: "p.ads" has been modified and must be recompiled
Now both files must be recompiled as indicated, and then the bind can succeed, generating a main program. You need not normally be concerned with the contents of this file, but for reference purposes a sample binder output file is given in Example of Binder Output File.
In most normal usage, the default mode of `gnatbind' which is to generate the main package in Ada, as described in the previous section. In particular, this means that any Ada programmer can read and understand the generated main program. It can also be debugged just like any other Ada code provided the `-g' switch is used for `gnatbind' and `gnatlink'.
The following switches are available with gnatbind; details will be presented in subsequent sections.
--version--help-a-aO-aI-A[=`filename']-bstderr even if verbose mode set.
-c-d`nn'[k|m][k|m] suffix, this switch is equivalent,
in effect, to completing all task specs with
pragma Storage_Size (nn);
When they do not already have such a pragma.
-D`nn'[k|m]The secondary stack is used to deal with functions that return a variable sized result, for example a function returning an unconstrained String. There are two ways in which this secondary stack is allocated.
For most targets, the secondary stack is growing on demand and is allocated as a chain of blocks in the heap. The -D option is not very relevant. It only give some control over the size of the allocated blocks (whose size is the minimum of the default secondary stack size value, and the actual size needed for the current allocation request).
For certain targets, notably VxWorks 653, the secondary stack is allocated by carving off a fixed ratio chunk of the primary task stack. The -D option is used to define the size of the environment task's secondary stack.
-e-EaSee also the packages GNAT.Traceback and GNAT.Traceback.Symbolic for more information. Note that on x86 ports, you must not use `-fomit-frame-pointer' `gcc' option.
-Es-E-F-h-H32-H64-I-I--l-L`xxx'-M`xyz'-m`n'-n-nostdinc-nostdlib--RTS=`rts-path'-o `file'b~`xxx.adb`).
Note that if this option is used, then linking must be done manually,
gnatlink cannot be used.
-O[=`filename']-p-P-R-Ra-s-S`xxx'in for an invalid value.
If zero is invalid for the discrete type in question, then the scalar value is set to all zero bits. For signed discrete types, the largest possible negative value of the underlying scalar is set (i.e. a one bit followed by all zero bits). For unsigned discrete types, the underlying scalar value is set to all one bits. For floating-point types, a NaN value is set (see body of package System.Scalar_Values for exact values).
lo for low value.
If zero is invalid for the discrete type in question, then the scalar value is set to all zero bits. For signed discrete types, the largest possible negative value of the underlying scalar is set (i.e. a one bit followed by all zero bits). For unsigned discrete types, the underlying scalar value is set to all zero bits. For floating-point, a small value is set (see body of package System.Scalar_Values for exact values).
hi for high value.
If zero is invalid for the discrete type in question, then the scalar value is set to all one bits. For signed discrete types, the largest possible positive value of the underlying scalar is set (i.e. a zero bit followed by all one bits). For unsigned discrete types, the underlying scalar value is set to all one bits. For floating-point, a large value is set (see body of package System.Scalar_Values for exact values).
The underlying scalar is set to a value consisting of repeated bytes, whose
value corresponds to the given value. For example if BF is given,
then a 32-bit scalar value will be set to the bit patterm 16#BFBFBFBF#.
In addition, you can specify `-Sev' to indicate that the value is
to be set at run time. In this case, the program will look for an environment
variable of the form GNAT_INIT_SCALARS=`yy', where yy is one
of `in/lo/hi/`xx*` with the same meanings as above.
If no environment variable is found, or if it does not have a valid value,
then the default is *in' (invalid values).
-static-shared-t-T`n'A value of zero is treated specially. It turns off time slicing, and in addition, indicates to the tasking run time that the semantics should match as closely as possible the Annex D requirements of the Ada RM, and in particular sets the default scheduling policy to FIFO_Within_Priorities.
-u`n'-vstdout.
-V`key'=`value'-w`x'-Wx`e'-x-X`nnn'-y-zYou may obtain this listing of switches by running gnatbind with no arguments.
As described earlier, by default gnatbind checks that object files are consistent with one another and are consistent with any source files it can locate. The following switches control binder access to sources.
-s-Wx`e'-xThe following switches provide control over the generation of error messages from the binder:
-vstderr. If this switch is present, a header is written
to stdout and any error messages are directed to stdout.
All that is written to stderr is a brief summary message.
-bstderr even if verbose mode is
specified. This is relevant only when used with the
`-v' switch.
-m`n'-M`xxx'-ws-we-tNormally failure of such checks, in accordance with the consistency requirements of the Ada Reference Manual, causes error messages to be generated which abort the binder and prevent the output of a binder file and subsequent link to obtain an executable.
The `-t' switch converts these error messages into warnings, so that binding and linking can continue to completion even in the presence of such errors. The result may be a failed link (due to missing symbols), or a non-functional executable which has undefined semantics.
Note: This means that `-t' should be used only in unusual situations, with extreme care. |
The following switches provide additional control over the elaboration order. For full details see Elaboration Order Handling in GNAT.
-pNormally it only makes sense to use the `-p' switch if dynamic elaboration checking is used (`-gnatE' switch used for compilation). This is because in the default static elaboration mode, all necessary Elaborate and Elaborate_All pragmas are implicitly inserted. These implicit pragmas are still respected by the binder in `-p' mode, so a safe elaboration order is assured.
Note that `-p' is not intended for production use; it is more for debugging/experimental use.
The following switches allow additional control over the output generated by the binder.
-c-estdout.
-hstdout.
-Kstdout. Includes library search paths,
contents of pragmas Ident and Linker_Options, and libraries added
by gnatbind.
-lstdout.
-Ostdout.
This list includes the files explicitly supplied and referenced by the user
as well as implicitly referenced run-time unit files. The latter are
omitted if the corresponding units reside in shared libraries. The
directory names for the run-time units depend on the system configuration.
-o `file'b~`mainprog.adb` default. Note that file denote the Ada
binder generated body filename.
Note that if this option is used, then linking must be done manually.
It is not possible to use gnatlink in this case, since it cannot locate
the binder file.
-rThe heap control switches – `-H32' and `-H64' – determine whether dynamic allocation uses 32-bit or 64-bit memory. They only affect compiler-generated allocations via __gnat_malloc; explicit calls to malloc and related functions from the C run-time library are unaffected.
-H32-H64These switches are only effective on VMS platforms.
The description so far has assumed that the main program is in Ada, and that the task of the binder is to generate a corresponding function main that invokes this Ada main program. GNAT also supports the building of executable programs where the main program is not in Ada, but some of the called routines are written in Ada and compiled using GNAT (Mixed Language Programming). The following switch is used in this situation:
-nIn this case, most of the functions of the binder are still required, but instead of generating a main program, the binder generates a file containing the following callable routines:
- `adainit'
- You must call this routine to initialize the Ada part of the program by calling the necessary elaboration routines. A call to adainit is required before the first call to an Ada subprogram.
Note that it is assumed that the basic execution environment must be setup to be appropriate for Ada execution at the point where the first Ada subprogram is called. In particular, if the Ada code will do any floating-point operations, then the FPU must be setup in an appropriate manner. For the case of the x86, for example, full precision mode is required. The procedure GNAT.Float_Control.Reset may be used to ensure that the FPU is in the right state.
- `adafinal'
- You must call this routine to perform any library-level finalization required by the Ada subprograms. A call to adafinal is required after the last call to an Ada subprogram, and before the program terminates.
If the `-n' switch is given, more than one ALI file may appear on the command line for gnatbind. The normal `closure' calculation is performed for each of the specified units. Calculating the closure means finding out the set of units involved by tracing `with' references. The reason it is necessary to be able to specify more than one ALI file is that a given program may invoke two or more quite separate groups of Ada units.
The binder takes the name of its output file from the last specified ALI file, unless overridden by the use of the `-o file'.
The output is an Ada unit in source form that can be compiled with GNAT. This compilation occurs automatically as part of the `gnatlink' processing.
Currently the GNAT run time requires a FPU using 80 bits mode precision. Under targets where this is not the default it is required to call GNAT.Float_Control.Reset before using floating point numbers (this include float computation, float input and output) in the Ada code. A side effect is that this could be the wrong mode for the foreign code where floating point computation could be broken after this call.
It is possible to have an Ada program which does not have a main subprogram. This program will call the elaboration routines of all the packages, then the finalization routines.
The following switch is used to bind programs organized in this manner:
-zThe package Ada.Command_Line provides access to the command-line arguments and program name. In order for this interface to operate correctly, the two variables
int gnat_argc;
char **gnat_argv;
are declared in one of the GNAT library routines. These variables must be set from the actual argc and argv values passed to the main program. With no `n' present, gnatbind generates the C main program to automatically set these variables. If the `n' switch is used, there is no automatic way to set these variables. If they are not set, the procedures in Ada.Command_Line will not be available, and any attempt to use them will raise Constraint_Error. If command line access is required, your main program must set gnat_argc and gnat_argv from the argc and argv values passed to it.
The binder takes the name of an ALI file as its argument and needs to locate source files as well as other ALI files to verify object consistency.
For source files, it follows exactly the same search rules as `gcc' (see Search Paths and the Run-Time Library (RTL)). For ALI files the directories searched are:
ADA_PRJ_OBJECTS_FILE environment variable.
ADA_PRJ_OBJECTS_FILE is normally set by gnatmake or by the gnat
driver when project files are used. It should not normally be set
by other means.
ADA_OBJECTS_PATH environment variable.
Construct this value
exactly as the
PATH environment variable: a list of directory
names separated by colons (semicolons when working with the NT version
of GNAT).
ada_object_path file which is part of the GNAT
installation tree and is used to store standard libraries such as the
GNAT Run Time Library (RTL) unless the switch `-nostdlib' is
specified. See Installing a library
In the binder the switch `-I'
is used to specify both source and
library file paths. Use `-aI'
instead if you want to specify
source paths only, and `-aO'
if you want to specify library paths
only. This means that for the binder
-I`dir' is equivalent to
-aI`dir'
-aO``dir'.
The binder generates the bind file (a C language source file) in the
current working directory.
The packages Ada, System, and Interfaces and their children make up the GNAT Run-Time Library, together with the package GNAT and its children, which contain a set of useful additional library functions provided by GNAT. The sources for these units are needed by the compiler and are kept together in one directory. The ALI files and object files generated by compiling the RTL are needed by the binder and the linker and are kept together in one directory, typically different from the directory containing the sources. In a normal installation, you need not specify these directory names when compiling or binding. Either the environment variables or the built-in defaults cause these files to be found.
Besides simplifying access to the RTL, a major use of search paths is in compiling sources from multiple directories. This can make development environments much more flexible.
Here are some examples of gnatbind invovations:
gnatbind helloThe main program Hello (source program in
hello.adb) is bound using the standard switch settings. The generated main program isb~hello.adb. This is the normal, default use of the binder.gnatbind hello -o mainprog.adbThe main program Hello (source program in
hello.adb) is bound using the standard switch settings. The generated main program ismainprog.adbwith the associated spec inmainprog.ads. Note that you must specify the body here not the spec. Note that if this option is used, then linking must be done manually, since gnatlink will not be able to find the generated file.
This chapter discusses `gnatlink', a tool that links an Ada program and builds an executable file. This utility invokes the system linker (via the `gcc' command) with a correct list of object files and library references. `gnatlink' automatically determines the list of files and references for the Ada part of a program. It uses the binder file generated by the `gnatbind' to determine this list.
Note: to invoke gnatlink with a project file, use the gnat driver (see The GNAT Driver and Project Files).
The form of the `gnatlink' command is
$ gnatlink [`switches`] `mainprog`[.ali]
[`non-Ada objects`] [`linker options`]
The arguments of `gnatlink' (switches, main ALI file,
non-Ada objects
or linker options) may be in any order, provided that no non-Ada object may
be mistaken for a main ALI file.
Any file name F without the .ali
extension will be taken as the main ALI file if a file exists
whose name is the concatenation of F and .ali.
mainprog.ali references the ALI file of the main program.
The .ali extension of this file can be omitted. From this
reference, `gnatlink' locates the corresponding binder file
b~mainprog.adb and, using the information in this file along
with the list of non-Ada objects and linker options, constructs a
linker command file to create the executable.
The arguments other than the `gnatlink' switches and the main
ALI file are passed to the linker uninterpreted.
They typically include the names of
object files for units written in other languages than Ada and any library
references required to resolve references in any of these foreign language
units, or in Import pragmas in any Ada units.
linker options is an optional list of linker specific switches. The default linker called by gnatlink is `gcc' which in turn calls the appropriate system linker.
One useful option for the linker is `-s': it reduces the size of the executable by removing all symbol table and relocation information from the executable.
Standard options for the linker such as `-lmy_lib' or `-Ldir' can be added as is. For options that are not recognized by `gcc' as linker options, use the `gcc' switches `-Xlinker' or `-Wl,'.
Refer to the GCC documentation for details.
Here is an example showing how to generate a linker map:
$ gnatlink my_prog -Wl,-Map,MAPFILE
Using linker options it is possible to set the program stack and heap size. See Setting Stack Size from gnatlink and Setting Heap Size from gnatlink.
`gnatlink' determines the list of objects required by the Ada program and prepends them to the list of objects passed to the linker. `gnatlink' also gathers any arguments set by the use of pragma Linker_Options and adds them to the list of arguments presented to the linker.
The following switches are available with the `gnatlink' utility:
--version--help-f-gb~mainprog.adb) to be compiled with `-g'.
In addition, the binder does not delete the b~mainprog.adb,
b~mainprog.o and b~mainprog.ali files.
Without `-g', the binder removes these files by default.
-n-v-v -v-o `exec-name'try.
-b `target'-B`dir'-M-M=`mapfile'--GCC=`compiler_name'gcc. You need to use quotes around compiler_name if
compiler_name contains spaces or other separator characters.
As an example --GCC="foo -x -y" will instruct `gnatlink' to
use foo -x -y as your compiler. Note that switch -c is always
inserted after your command name. Thus in the above example the compiler
command that will be used by `gnatlink' will be foo -c -x -y.
A limitation of this syntax is that the name and path name of the executable
itself must not include any embedded spaces. If the compiler executable is
different from the default one (gcc or <prefix>-gcc), then the back-end
switches in the ALI file are not used to compile the binder generated source.
For example, this is the case with --GCC="foo -x -y". But the back end
switches will be used for --GCC="gcc -gnatv". If several
--GCC=compiler_name are used, only the last compiler_name
is taken into account. However, all the additional switches are also taken
into account. Thus,
--GCC="foo -x -y" --GCC="bar -z -t" is equivalent to
--GCC="bar -x -y -z -t".
--LINK=`name'This chapter offers some examples of makefiles that solve specific problems. It does not explain how to write a makefile, nor does it try to replace the `gnatmake' utility (Building with gnatmake).
All the examples in this section are specific to the GNU version of make. Although `make' is a standard utility, and the basic language is the same, these examples use some advanced features found only in GNU make.
Complex project organizations can be handled in a very powerful way by using GNU make combined with gnatmake. For instance, here is a Makefile which allows you to build each subsystem of a big project into a separate shared library. Such a makefile allows you to significantly reduce the link time of very big applications while maintaining full coherence at each step of the build process.
The list of dependencies are handled automatically by `gnatmake'. The Makefile is simply used to call gnatmake in each of the appropriate directories.
Note that you should also read the example on how to automatically create the list of directories (Automatically Creating a List of Directories) which might help you in case your project has a lot of subdirectories.
## This Makefile is intended to be used with the following directory
## configuration:
## - The sources are split into a series of csc (computer software components)
## Each of these csc is put in its own directory.
## Their name are referenced by the directory names.
## They will be compiled into shared library (although this would also work
## with static libraries
## - The main program (and possibly other packages that do not belong to any
## csc is put in the top level directory (where the Makefile is).
## toplevel_dir __ first_csc (sources) __ lib (will contain the library)
## \\_ second_csc (sources) __ lib (will contain the library)
## \\_ ...
## Although this Makefile is build for shared library, it is easy to modify
## to build partial link objects instead (modify the lines with -shared and
## gnatlink below)
##
## With this makefile, you can change any file in the system or add any new
## file, and everything will be recompiled correctly (only the relevant shared
## objects will be recompiled, and the main program will be re-linked).
# The list of computer software component for your project. This might be
# generated automatically.
CSC_LIST=aa bb cc
# Name of the main program (no extension)
MAIN=main
# If we need to build objects with -fPIC, uncomment the following line
#NEED_FPIC=-fPIC
# The following variable should give the directory containing libgnat.so
# You can get this directory through 'gnatls -v'. This is usually the last
# directory in the Object_Path.
GLIB=...
# The directories for the libraries
# (This macro expands the list of CSC to the list of shared libraries, you
# could simply use the expanded form:
# LIB_DIR=aa/lib/libaa.so bb/lib/libbb.so cc/lib/libcc.so
LIB_DIR=${foreach dir,${CSC_LIST},${dir}/lib/lib${dir}.so}
${MAIN}: objects ${LIB_DIR}
gnatbind ${MAIN} ${CSC_LIST:%=-aO%/lib} -shared
gnatlink ${MAIN} ${CSC_LIST:%=-l%}
objects::
# recompile the sources
gnatmake -c -i ${MAIN}.adb ${NEED_FPIC} ${CSC_LIST:%=-I%}
# Note: In a future version of GNAT, the following commands will be simplified
# by a new tool, gnatmlib
${LIB_DIR}:
mkdir -p ${dir $@ }
cd ${dir $@ } && gcc -shared -o ${notdir $@ } ../*.o -L${GLIB} -lgnat
cd ${dir $@ } && cp -f ../*.ali .
# The dependencies for the modules
# Note that we have to force the expansion of *.o, since in some cases
# make won't be able to do it itself.
aa/lib/libaa.so: ${wildcard aa/*.o}
bb/lib/libbb.so: ${wildcard bb/*.o}
cc/lib/libcc.so: ${wildcard cc/*.o}
# Make sure all of the shared libraries are in the path before starting the
# program
run::
LD_LIBRARY_PATH=`pwd`/aa/lib:`pwd`/bb/lib:`pwd`/cc/lib ./${MAIN}
clean::
${RM} -rf ${CSC_LIST:%=%/lib}
${RM} ${CSC_LIST:%=%/*.ali}
${RM} ${CSC_LIST:%=%/*.o}
${RM} *.o *.ali ${MAIN}
In most makefiles, you will have to specify a list of directories, and store it in a variable. For small projects, it is often easier to specify each of them by hand, since you then have full control over what is the proper order for these directories, which ones should be included.
However, in larger projects, which might involve hundreds of subdirectories, it might be more convenient to generate this list automatically.
The example below presents two methods. The first one, although less general, gives you more control over the list. It involves wildcard characters, that are automatically expanded by `make'. Its shortcoming is that you need to explicitly specify some of the organization of your project, such as for instance the directory tree depth, whether some directories are found in a separate tree, etc.
The second method is the most general one. It requires an external program, called `find', which is standard on all Unix systems. All the directories found under a given root directory will be added to the list.
# The examples below are based on the following directory hierarchy:
# All the directories can contain any number of files
# ROOT_DIRECTORY -> a -> aa -> aaa
# -> ab
# -> ac
# -> b -> ba -> baa
# -> bb
# -> bc
# This Makefile creates a variable called DIRS, that can be reused any time
# you need this list (see the other examples in this section)
# The root of your project's directory hierarchy
ROOT_DIRECTORY=.
####
# First method: specify explicitly the list of directories
# This allows you to specify any subset of all the directories you need.
####
DIRS := a/aa/ a/ab/ b/ba/
####
# Second method: use wildcards
# Note that the argument(s) to wildcard below should end with a '/'.
# Since wildcards also return file names, we have to filter them out
# to avoid duplicate directory names.
# We thus use make's `dir` and `sort` functions.
# It sets DIRs to the following value (note that the directories aaa and baa
# are not given, unless you change the arguments to wildcard).
# DIRS= ./a/a/ ./b/ ./a/aa/ ./a/ab/ ./a/ac/ ./b/ba/ ./b/bb/ ./b/bc/
####
DIRS := ${sort ${dir ${wildcard ${ROOT_DIRECTORY}/*/
${ROOT_DIRECTORY}/*/*/}}}
####
# Third method: use an external program
# This command is much faster if run on local disks, avoiding NFS slowdowns.
# This is the most complete command: it sets DIRs to the following value:
# DIRS= ./a ./a/aa ./a/aa/aaa ./a/ab ./a/ac ./b ./b/ba ./b/ba/baa ./b/bb ./b/bc
####
DIRS := ${shell find ${ROOT_DIRECTORY} -type d -print}
Once you have created the list of directories as explained in the previous section (Automatically Creating a List of Directories), you can easily generate the command line arguments to pass to gnatmake.
For the sake of completeness, this example assumes that the source path is not the same as the object path, and that you have two separate lists of directories.
# see "Automatically creating a list of directories" to create
# these variables
SOURCE_DIRS=
OBJECT_DIRS=
GNATMAKE_SWITCHES := ${patsubst %,-aI%,${SOURCE_DIRS}}
GNATMAKE_SWITCHES += ${patsubst %,-aO%,${OBJECT_DIRS}}
all:
gnatmake ${GNATMAKE_SWITCHES} main_unit
One problem that might be encountered on big projects is that many operating systems limit the length of the command line. It is thus hard to give gnatmake the list of source and object directories.
This example shows how you can set up environment variables, which will make `gnatmake' behave exactly as if the directories had been specified on the command line, but have a much higher length limit (or even none on most systems).
It assumes that you have created a list of directories in your Makefile, using one of the methods presented in Automatically Creating a List of Directories. For the sake of completeness, we assume that the object path (where the ALI files are found) is different from the sources patch.
Note a small trick in the Makefile below: for efficiency reasons, we create two temporary variables (SOURCE_LIST and OBJECT_LIST), that are expanded immediately by make. This way we overcome the standard make behavior which is to expand the variables only when they are actually used.
On Windows, if you are using the standard Windows command shell, you must replace colons with semicolons in the assignments to these variables.
# In this example, we create both ADA_INCLUDE_PATH and ADA_OBJECTS_PATH.
# This is the same thing as putting the -I arguments on the command line.
# (the equivalent of using -aI on the command line would be to define
# only ADA_INCLUDE_PATH, the equivalent of -aO is ADA_OBJECTS_PATH).
# You can of course have different values for these variables.
#
# Note also that we need to keep the previous values of these variables, since
# they might have been set before running 'make' to specify where the GNAT
# library is installed.
# see "Automatically creating a list of directories" to create these
# variables
SOURCE_DIRS=
OBJECT_DIRS=
empty:=
space:=${empty} ${empty}
SOURCE_LIST := ${subst ${space},:,${SOURCE_DIRS}}
OBJECT_LIST := ${subst ${space},:,${OBJECT_DIRS}}
ADA_INCLUDE_PATH += ${SOURCE_LIST}
ADA_OBJECTS_PATH += ${OBJECT_LIST}
export ADA_INCLUDE_PATH
export ADA_OBJECTS_PATH
all:
gnatmake main_unit
This chapter describes GNAT's `Project Manager', a facility that allows you to manage complex builds involving a number of source files, directories, and options for different system configurations. In particular, project files allow you to specify:
ALI files, object files, tree files, etc.) is to be placed
Project files are written in a syntax close to that of Ada, using familiar notions such as packages, context clauses, declarations, default values, assignments, and inheritance (see Project File Reference).
Project files can be built hierarchically from other project files, simplifying complex system integration and project reuse (see Organizing Projects into Subsystems).
Several tools support project files, generally in addition to specifying
the information on the command line itself). They share common switches
to control the loading of the project (in particular
-P`projectfile' and
-X`vbl'=`value').
The Project Manager supports a wide range of development strategies, for systems of all sizes. Here are some typical practices that are easily handled:
Project files can be used to achieve some of the effects of a source versioning system (for example, defining separate projects for the different sets of sources that comprise different releases) but the Project Manager is independent of any source configuration management tool that might be used by the developers.
The various sections below introduce the different concepts related to projects. Each section starts with examples and use cases, and then goes into the details of related project file capabilities.
In its simplest form, a unique project is used to build a single executable. This section concentrates on such a simple setup. Later sections will extend this basic model to more complex setups.
The following concepts are the foundation of project files, and will be further detailed later in this documentation. They are summarized here as a reference.
.gpr
extension. It defines build-related characteristics of an application.
The characteristics include the list of sources, the location of those
sources, the location for the generated object files, the name of
the main program, and the options for the various tools involved in the
build process.
The following subsections introduce gradually all the attributes of interest for simple build needs. Here is the simple setup that will be used in the following examples.
The Ada source files pack.ads, pack.adb, and proc.adb are in
the common/ directory. The file proc.adb contains an Ada main
subprogram Proc that `with's package Pack. We want to compile
these source files with the switch
`-O2', and put the resulting files in
the directory obj/.
common/
pack.ads
pack.adb
proc.adb
common/obj/
proc.ali, proc.o pack.ali, pack.o
Our project is to be called `Build'. The name of the
file is the name of the project (case-insensitive) with the
.gpr extension, therefore the project file name is build.gpr. This
is not mandatory, but a warning is issued when this convention is not followed.
This is a very simple example, and as stated above, a single project file is enough for it. We will thus create a new file, that for now should contain the following code:
project Build is
end Build;
When you create a new project, the first thing to describe is how to find the corresponding source files. These are the only settings that are needed by all the tools that will use this project (builder, compiler, binder and linker for the compilation, IDEs to edit the source files,...).
The first step is to declare the source directories, which are the directories
to be searched to find source files. In the case of the example,
the common directory is the only source directory.
There are several ways of defining source directories:
build.gpr is placed in the common
directory, the project has the needed implicit source directory.
"/usr/local/common/" on UNIX), or relative to the
directory in which the project file resides (for instance "." if
build.gpr is inside common/, or "common" if it is one level up).
Each of the source directories must exist and be readable.
The syntax for directories is platform specific. For portability, however, the project manager will always properly translate UNIX-like path names to the native format of the specific platform. For instance, when the same project file is to be used both on Unix and Windows, "/" should be used as the directory separator rather than "\".
**", then that path and all its subdirectories
(recursively) are included in the list of source directories. For instance,
** and ./** represent the complete directory tree rooted at
the directory in which the project file resides.
When using that construct, it can sometimes be convenient to also use the attribute `Excluded_Source_Dirs', which is also a list of paths. Each entry specifies a directory whose immediate content, not including subdirs, is to be excluded. It is also possible to exclude a complete directory subtree using the "**" notation.
It is often desirable to remove, from the source directories, directory subtrees rooted at some subdirectories. An example is the subdirectories created by a Version Control System such as Subversion that creates directory subtrees rooted at subdirectories ".svn". To do that, attribute `Ignore_Source_Sub_Dirs' can be used. It specifies the list of simple file names for the roots of these undesirable directory subtrees.
for Source_Dirs use ("./**");
for Ignore_Source_Sub_Dirs use (".svn");
When applied to the simple example, and because we generally prefer to have the project file at the toplevel directory rather than mixed with the sources, we will create the following file
build.gpr
project Build is
for Source_Dirs use ("common"); -- <<<<
end Build;
Once source directories have been specified, one may need to indicate source files of interest. By default, all source files present in the source directories are considered by the project manager. When this is not desired, it is possible to specify the list of sources to consider explicitly. In such a case, only source file base names are indicated and not their absolute or relative path names. The project manager is in charge of locating the specified source files in the specified source directories.
Since the project manager was initially developed for Ada environments, the default language is usually Ada and the above project file is complete: it defines without ambiguity the sources composing the project: that is to say, all the sources in subdirectory "common" for the default language (Ada) using the default naming convention.
However, when compiling a multi-language application, or a pure C application, the project manager must be told which languages are of interest, which is done by setting the `Languages' attribute to a list of strings, each of which is the name of a language.
Even when using only Ada, the default naming might not be suitable. Indeed,
how does the project manager recognizes an "Ada file" from any other
file? Project files can describe the naming scheme used for source files,
and override the default (see Naming Schemes). The default is the
standard GNAT extension (.adb for bodies and .ads for
specs), which is what is used in our example, explaining why no naming scheme
is explicitly specified.
See Naming Schemes.
A warning is issued if both attributes Source_Files and Source_List_File are given explicit values. In this case, the attribute Source_Files prevails.
In most simple cases, such as the above example, the default source file search
behavior provides the expected result, and we do not need to add anything after
setting Source_Dirs. The project manager automatically finds
pack.ads, pack.adb, and proc.adb as source files of the
project.
Note that by default a warning is issued when a project has no sources attached to it and this is not explicitly indicated in the project file.
If the order of the source directories is known statically, that is if "/**" is not used in the string list Source_Dirs, then there may be several files with the same name sitting in different directories of the project. In this case, only the file in the first directory is considered as a source of the project and the others are hidden. If "/**" is used in the string list Source_Dirs, it is an error to have several files with the same name in the same directory "/**" subtree, since there would be an ambiguity as to which one should be used. However, two files with the same name may exist in two single directories or directory subtrees. In this case, the one in the first directory or directory subtree is a source of the project.
If there are two sources in different directories of the same "/**" subtree, one way to resolve the problem is to exclude the directory of the file that should not be used as a source of the project.
The next step when writing a project is to indicate where the compiler should put the object files. In fact, the compiler and other tools might create several different kind of files (for GNAT, there is the object file and the ALI file for instance). One of the important concepts in projects is that most tools may consider source directories as read-only and do not attempt to create new or temporary files there. Instead, all files are created in the object directory. It is of course not true for project-aware IDEs, whose purpose it is to create the source files.
The object directory is specified through the `Object_Dir' attribute. Its value is the path to the object directory, either absolute or relative to the directory containing the project file. This directory must already exist and be readable and writable, although some tools have a switch to create the directory if needed (See the switch -p for `gprbuild').
If the attribute Object_Dir is not specified, it defaults to the project directory, that is the directory containing the project file.
For our example, we can specify the object dir in this way:
project Build is
for Source_Dirs use ("common");
for Object_Dir use "obj"; -- <<<<
end Build;
As mentioned earlier, there is a single object directory per project. As a result, if you have an existing system where the object files are spread across several directories, you can either move all of them into the same directory if you want to build it with a single project file, or study the section on subsystems (see Organizing Projects into Subsystems) to see how each separate object directory can be associated with one of the subsystems constituting the application.
When the `linker' is called, it usually creates an executable. By default, this executable is placed in the object directory of the project. It might be convenient to store it in its own directory.
This can be done through the Exec_Dir attribute, which, like `Object_Dir' contains a single absolute or relative path and must point to an existing and writable directory, unless you ask the tool to create it on your behalf. When not specified, It defaults to the object directory and therefore to the project file's directory if neither `Object_Dir' nor `Exec_Dir' was specified.
In the case of the example, let's place the executable in the root
of the hierarchy, ie the same directory as build.gpr. Hence
the project file is now
project Build is
for Source_Dirs use ("common");
for Object_Dir use "obj";
for Exec_Dir use "."; -- <<<<
end Build;
In the previous section, executables were mentioned. The project manager needs to be taught what they are. In a project file, an executable is indicated by pointing to the source file of a main subprogram. In C this is the file that contains the main function, and in Ada the file that contains the main unit.
There can be any number of such main files within a given project, and thus several executables can be built in the context of a single project file. Of course, one given executable might not (and in fact will not) need all the source files referenced by the project. As opposed to other build environments such as `makefile', one does not need to specify the list of dependencies of each executable, the project-aware builder knows enough of the semantics of the languages to build and link only the necessary elements.
The list of main files is specified via the `Main' attribute. It contains a list of file names (no directories). If a project defines this attribute, it is not necessary to identify main files on the command line when invoking a builder, and editors like `GPS' will be able to create extra menus to spawn or debug the corresponding executables.
project Build is
for Source_Dirs use ("common");
for Object_Dir use "obj";
for Exec_Dir use ".";
for Main use ("proc.adb"); -- <<<<
end Build;
If this attribute is defined in the project, then spawning the builder with a command such as
gprbuild -Pbuild
automatically builds all the executables corresponding to the files listed in the `Main' attribute. It is possible to specify one or more executables on the command line to build a subset of them.
We now have a project file that fully describes our environment, and can be
used to build the application with a simple `gprbuild' command as seen
in the previous section. In fact, the empty project we showed immediately at
the beginning (with no attribute at all) could already fulfill that need if it
was put in the common directory.
Of course, we might want more control. This section shows you how to specify the compilation switches that the various tools involved in the building of the executable should use.
Since source names and locations are described in the project file, it is not necessary to use switches on the command line for this purpose (switches such as -I for gcc). This removes a major source of command line length overflow. Clearly, the builders will have to communicate this information one way or another to the underlying compilers and tools they call but they usually use response files for this and thus are not subject to command line overflows.
Several tools participate to the creation of an executable: the compiler produces object files from the source files; the binder (in the Ada case) creates a "source" file that takes care, among other things, of elaboration issues and global variable initialization; and the linker gathers everything into a single executable that users can execute. All these tools are known to the project manager and will be called with user defined switches from the project files. However, we need to introduce a new project file concept to express the switches to be used for any of the tools involved in the build.
A project file is subdivided into zero or more `packages', each of which contains the attributes specific to one tool (or one set of tools). Project files use an Ada-like syntax for packages. Package names permitted in project files are restricted to a predefined set (see Packages), and the contents of packages are limited to a small set of constructs and attributes (see Attributes).
Our example project file can be extended with the following empty packages. At this stage, they could all be omitted since they are empty, but they show which packages would be involved in the build process.
project Build is
for Source_Dirs use ("common");
for Object_Dir use "obj";
for Exec_Dir use ".";
for Main use ("proc.adb");
package Builder is --<<< for gprbuild
end Builder;
package Compiler is --<<< for the compiler
end Compiler;
package Binder is --<<< for the binder
end Binder;
package Linker is --<<< for the linker
end Linker;
end Build;
Let's first examine the compiler switches. As stated in the initial description of the example, we want to compile all files with `-O2'. This is a compiler switch, although it is usual, on the command line, to pass it to the builder which then passes it to the compiler. It is recommended to use directly the right package, which will make the setup easier to understand for other people.
Several attributes can be used to specify the switches:
This is the first mention in this manual of an `indexed attribute'. When this attribute is defined, one must supply an `index' in the form of a literal string. In the case of `Default_Switches', the index is the name of the language to which the switches apply (since a different compiler will likely be used for each language, and each compiler has its own set of switches). The value of the attribute is a list of switches.
In this example, we want to compile all Ada source files with the switch `-O2', and the resulting project file is as follows (only the Compiler package is shown):
package Compiler is for Default_Switches ("Ada") use ("-O2"); end Compiler;
In some cases, we might want to use specific switches for one or more files. For instance, compiling
proc.adbmight not be possible at high level of optimization because of a compiler issue. In such a case, the `Switches' attribute (indexed on the file name) can be used and will override the switches defined by `Default_Switches'. Our project file would become:package Compiler is for Default_Switches ("Ada") use ("-O2"); for Switches ("proc.adb") use ("-O0"); end Compiler;Switches may take a pattern as an index, such as in:
package Compiler is for Default_Switches ("Ada") use ("-O2"); for Switches ("pkg*") use ("-O0"); end Compiler;Sources
pkg.adbandpkg-child.adbwould be compiled with -O0, not -O2.Switches can also be given a language name as index instead of a file name in which case it has the same semantics as `Default_Switches'. However, indexes with wild cards are never valid for language name.
`Local_Configuration_Pragmas':
This attribute may specify the path of a file containing configuration pragmas for use by the Ada compiler, such as pragma Restrictions (No_Tasking). These pragmas will be used for all the sources of the project.
The switches for the other tools are defined in a similar manner through the `Default_Switches' and `Switches' attributes, respectively in the `Builder' package (for `gprbuild'), the `Binder' package (binding Ada executables) and the `Linker' package (for linking executables).
Now that our project files are written, let's build our executable. Here is the command we would use from the command line:
gprbuild -Pbuild
This will automatically build the executables specified through the `Main' attribute: for each, it will compile or recompile the sources for which the object file does not exist or is not up-to-date; it will then run the binder; and finally run the linker to create the executable itself.
The `gprbuild' builder, can automatically manage C files the
same way: create the file utils.c in the common directory,
set the attribute `Languages' to "(Ada, C)", and re-run
gprbuild -Pbuild
Gprbuild knows how to recompile the C files and will recompile them only if one of their dependencies has changed. No direct indication on how to build the various elements is given in the project file, which describes the project properties rather than a set of actions to be executed. Here is the invocation of `gprbuild' when building a multi-language program:
$ gprbuild -Pbuild
gcc -c proc.adb
gcc -c pack.adb
gcc -c utils.c
gprbind proc
...
gcc proc.o -o proc
Notice the three steps described earlier:
The default output of GPRbuild's execution is kept reasonably simple and easy to understand. In particular, some of the less frequently used commands are not shown, and some parameters are abbreviated. So it is not possible to rerun the effect of the `gprbuild' command by cut-and-pasting its output. GPRbuild's option -v provides a much more verbose output which includes, among other information, more complete compilation, post-compilation and link commands.
By default, the executable name corresponding to a main file is computed from the main source file name. Through the attribute `Builder.Executable', it is possible to change this default.
For instance, instead of building `proc' (or `proc.exe' on Windows), we could configure our project file to build "proc1" (resp proc1.exe) with the following addition:
project Build is
... -- same as before
package Builder is
for Executable ("proc.adb") use "proc1";
end Builder
end Build;
Attribute `Executable_Suffix', when specified, may change the suffix of the executable files, when no attribute Executable applies: its value replaces the platform-specific executable suffix. The default executable suffix is empty on UNIX and ".exe" on Windows.
It is also possible to change the name of the produced executable by using the command line switch `-o'. When several mains are defined in the project, it is not possible to use the `-o' switch and the only way to change the names of the executable is provided by Attributes Executable and Executable_Suffix.
To illustrate some other project capabilities, here is a slightly more complex project using similar sources and a main program in C:
project C_Main is
for Languages use ("Ada", "C");
for Source_Dirs use ("common");
for Object_Dir use "obj";
for Main use ("main.c");
package Compiler is
C_Switches := ("-pedantic");
for Default_Switches ("C") use C_Switches;
for Default_Switches ("Ada") use ("-gnaty");
for Switches ("main.c") use C_Switches & ("-g");
end Compiler;
end C_Main;
This project has many similarities with the previous one.
As expected, its Main attribute now refers to a C source.
The attribute `Exec_Dir' is now omitted, thus the resulting
executable will be put in the directory obj.
The most noticeable difference is the use of a variable in the `Compiler' package to store settings used in several attributes. This avoids text duplication, and eases maintenance (a single place to modify if we want to add new switches for C files). We will revisit the use of variables in the context of scenarios (see Scenarios in Projects).
In this example, we see how the file main.c can be compiled with
the switches used for all the other C files, plus `-g'.
In this specific situation the use of a variable could have been
replaced by a reference to the Default_Switches attribute:
for Switches ("c_main.c") use Compiler'Default_Switches ("C") & ("-g");
Note the tick (`'') used to refer to attributes defined in a package.
Here is the output of the GPRbuild command using this project:
$ gprbuild -Pc_main
gcc -c -pedantic -g main.c
gcc -c -gnaty proc.adb
gcc -c -gnaty pack.adb
gcc -c -pedantic utils.c
gprbind main.bexch
...
gcc main.o -o main
The default switches for Ada sources,
the default switches for C sources (in the compilation of lib.c),
and the specific switches for main.c have all been taken into
account.
Sometimes an Ada software system is ported from one compilation environment to another (say GNAT), and the file are not named using the default GNAT conventions. Instead of changing all the file names, which for a variety of reasons might not be possible, you can define the relevant file naming scheme in the `Naming' package of your project file.
The naming scheme has two distinct goals for the project manager: it allows finding of source files when searching in the source directories, and given a source file name it makes it possible to guess the associated language, and thus the compiler to use.
Note that the use by the Ada compiler of pragmas Source_File_Name is not supported when using project files. You must use the features described in this paragraph. You can however specify other configuration pragmas.
The following attributes can be defined in package Naming:
Its value must be one of "lowercase" (the default if unspecified), "uppercase" or "mixedcase". It describes the casing of file names with regards to the Ada unit name. Given an Ada unit My_Unit, the file name will respectively be
my_unit.adb(lowercase),MY_UNIT.ADB(uppercase) orMy_Unit.adb(mixedcase). On Windows, file names are case insensitive, so this attribute is irrelevant.
This attribute specifies the string that should replace the "." in unit names. Its default value is "-" so that a unit Parent.Child is expected to be found in the file
parent-child.adb. The replacement string must satisfy the following requirements to avoid ambiguities in the naming scheme:
- It must not be empty
- It cannot start or end with an alphanumeric character
- It cannot be a single underscore
- It cannot start with an underscore followed by an alphanumeric
- It cannot contain a dot '.' except if the entire string is "."
`Spec_Suffix' and `Specification_Suffix':
For Ada, these attributes give the suffix used in file names that contain specifications. For other languages, they give the extension for files that contain declaration (header files in C for instance). The attribute is indexed on the language. The two attributes are equivalent, but the latter is obsolescent.
If the value of the attribute is the empty string, it indicates to the Project Manager that the only specifications/header files for the language are those specified with attributes Spec or Specification_Exceptions.
If Spec_Suffix ("Ada") is not specified, then the default is ".ads".
A non empty value must satisfy the following requirements:
- It must include at least one dot
- If Dot_Replacement is a single dot, then it cannot include more than one dot.
`Body_Suffix' and `Implementation_Suffix':
These attributes give the extension used for file names that contain code (bodies in Ada). They are indexed on the language. The second version is obsolescent and fully replaced by the first attribute.
For each language of a project, one of these two attributes need to be specified, either in the project itself or in the configuration project file.
If the value of the attribute is the empty string, it indicates to the Project Manager that the only source files for the language are those specified with attributes Body or Implementation_Exceptions.
These attributes must satisfy the same requirements as Spec_Suffix. In addition, they must be different from any of the values in Spec_Suffix. If Body_Suffix ("Ada") is not specified, then the default is ".adb".
If Body_Suffix ("Ada") and Spec_Suffix ("Ada") end with the same string, then a file name that ends with the longest of these two suffixes will be a body if the longest suffix is Body_Suffix ("Ada") or a spec if the longest suffix is Spec_Suffix ("Ada").
If the suffix does not start with a '.', a file with a name exactly equal to the suffix will also be part of the project (for instance if you define the suffix as Makefile.in, a file called
Makefile.inwill be part of the project. This capability is usually not interesting when building. However, it might become useful when a project is also used to find the list of source files in an editor, like the GNAT Programming System (GPS).
This attribute is specific to Ada. It denotes the suffix used in file names that contain separate bodies. If it is not specified, then it defaults to same value as Body_Suffix ("Ada").
The value of this attribute cannot be the empty string.
Otherwise, the same rules apply as for the Body_Suffix attribute. The only accepted index is "Ada".
`Spec' or `Specification':
This attribute Spec can be used to define the source file name for a given Ada compilation unit's spec. The index is the literal name of the Ada unit (case insensitive). The value is the literal base name of the file that contains this unit's spec (case sensitive or insensitive depending on the operating system). This attribute allows the definition of exceptions to the general naming scheme, in case some files do not follow the usual convention.
When a source file contains several units, the relative position of the unit can be indicated. The first unit in the file is at position 1
for Spec ("MyPack.MyChild") use "mypack.mychild.spec"; for Spec ("top") use "foo.a" at 1; for Spec ("foo") use "foo.a" at 2;
These attribute play the same role as `Spec' for Ada bodies.
`Specification_Exceptions' and `Implementation_Exceptions':
These attributes define exceptions to the naming scheme for languages other than Ada. They are indexed on the language name, and contain a list of file names respectively for headers and source code.
For example, the following package models the Apex file naming rules:
package Naming is
for Casing use "lowercase";
for Dot_Replacement use ".";
for Spec_Suffix ("Ada") use ".1.ada";
for Body_Suffix ("Ada") use ".2.ada";
end Naming;
After building an application or a library it is often required to install it into the development environment. For instance this step is required if the library is to be used by another application. The `gprinstall' tool provides an easy way to install libraries, executable or object code generated during the build. The `Install' package can be used to change the default locations.
The following attributes can be defined in package Install:
An array attribute to declare a set of files not part of the sources to be installed. The array discriminant is the directory where the file is to be installed. If a relative directory then Prefix (see below) is prepended. Note also that if the same file name occurs multiple time in the attribute list, the last one will be the one installed.
Root directory for the installation.
`Exec_Subdir'
Subdirectory of `Prefix' where executables are to be installed. Default is `bin'.
`Lib_Subdir'
Subdirectory of `Prefix' where directory with the library or object files is to be installed. Default is `lib'.
`Sources_Subdir'
Subdirectory of `Prefix' where directory with sources is to be installed. Default is `include'.
`Project_Subdir'
Subdirectory of `Prefix' where the generated project file is to be installed. Default is `share/gpr'.
`Mode'
The installation mode, it is either `dev' (default) or `usage'. See `gprbuild' user's guide for details.
`Install_Name'
Specify the name to use for recording the installation. The default is the project name without the extension.
For large projects the compilation time can become a limitation in the development cycle. To cope with that, GPRbuild supports distributed compilation.
The following attributes can be defined in package Remote:
Root directory of the project's sources. The default value is the project's directory.
A `subsystem' is a coherent part of the complete system to be built. It is represented by a set of sources and one single object directory. A system can be composed of a single subsystem when it is simple as we have seen in the first section. Complex systems are usually composed of several interdependent subsystems. A subsystem is dependent on another subsystem if knowledge of the other one is required to build it, and in particular if visibility on some of the sources of this other subsystem is required. Each subsystem is usually represented by its own project file.
In this section, the previous example is being extended. Let's assume some sources of our Build project depend on other sources. For instance, when building a graphical interface, it is usual to depend upon a graphical library toolkit such as GtkAda. Furthermore, we also need sources from a logging module we had previously written.
GtkAda comes with its own project file (appropriately called
gtkada.gpr), and we will assume we have already built a project
called logging.gpr for the logging module. With the information provided
so far in build.gpr, building the application would fail with an error
indicating that the gtkada and logging units that are relied upon by the sources
of this project cannot be found.
This is solved by adding the following `with' clauses at the beginning of our project:
with "gtkada.gpr";
with "a/b/logging.gpr";
project Build is
... -- as before
end Build;
When such a project is compiled, `gprbuild' will automatically check the other projects and recompile their sources when needed. It will also recompile the sources from Build when needed, and finally create the executable. In some cases, the implementation units needed to recompile a project are not available, or come from some third party and you do not want to recompile it yourself. In this case, set the attribute `Externally_Built' to "true", indicating to the builder that this project can be assumed to be up-to-date, and should not be considered for recompilation. In Ada, if the sources of this externally built project were compiled with another version of the compiler or with incompatible options, the binder will issue an error.
The project's `with' clause has several effects. It provides source visibility between projects during the compilation process. It also guarantees that the necessary object files from Logging and GtkAda are available when linking Build.
As can be seen in this example, the syntax for importing projects is similar to the syntax for importing compilation units in Ada. However, project files use literal strings instead of names, and the `with' clause identifies project files rather than packages.
Each literal string after `with' is the path
(absolute or relative) to a project file. The .gpr extension is
optional, although we recommend adding it. If no extension is specified,
and no project file with the .gpr extension is found, then
the file is searched for exactly as written in the `with' clause,
that is with no extension.
As mentioned above, the path after a `with' has to be a literal string, and you cannot use concatenation, or lookup the value of external variables to change the directories from which a project is loaded. A solution if you need something like this is to use aggregate projects (see Aggregate Projects).
When a relative path or a base name is used, the project files are searched relative to each of the directories in the `project path'. This path includes all the directories found with the following algorithm, in this order; the first matching file is used:
<prefix>/<target>/lib/gnat if option `–target' is specified
<prefix>/<target>/share/gpr if option `–target' is specified
<prefix>/share/gpr/
<prefix>/lib/gnat/
In our example, gtkada.gpr is found in the predefined directory if
it was installed at the same root as GNAT.
Some tools also support extending the project path from the command line, generally through the `-aP'. You can see the value of the project path by using the `gnatls -v' command.
Any symbolic link will be fully resolved in the directory of the importing project file before the imported project file is examined.
Any source file in the imported project can be used by the sources of the importing project, transitively. Thus if A imports B, which imports C, the sources of A may depend on the sources of C, even if A does not import C explicitly. However, this is not recommended, because if and when B ceases to import C, some sources in A will no longer compile. `gprbuild' has a switch `–no-indirect-imports' that will report such indirect dependencies.
Note: One very important aspect of a project hierarchy is that `a given source can only belong to one project' (otherwise the project manager would not know which settings apply to it and when to recompile it). It means that different project files do not usually share source directories or when they do, they need to specify precisely which project owns which sources using attribute Source_Files or equivalent. By contrast, 2 projects can each own a source with the same base file name as long as they live in different directories. The latter is not true for Ada Sources because of the correlation between source files and Ada units. |
Cyclic dependencies are mostly forbidden: if A imports B (directly or indirectly) then B is not allowed to import A. However, there are cases when cyclic dependencies would be beneficial. For these cases, another form of import between projects exists: the `limited with'. A project A that imports a project B with a straight `with' may also be imported, directly or indirectly, by B through a limited with.
The difference between straight `with' and limited with is that the name of a project imported with a limited with cannot be used in the project importing it. In particular, its packages cannot be renamed and its variables cannot be referred to.
with "b.gpr";
with "c.gpr";
project A is
for Exec_Dir use B'Exec_Dir; -- ok
end A;
limited with "a.gpr"; -- Cyclic dependency: A -> B -> A
project B is
for Exec_Dir use A'Exec_Dir; -- not ok
end B;
with "d.gpr";
project C is
end C;
limited with "a.gpr"; -- Cyclic dependency: A -> C -> D -> A
project D is
for Exec_Dir use A'Exec_Dir; -- not ok
end D;
When building an application, it is common to have similar needs in several of the projects corresponding to the subsystems under construction. For instance, they will all have the same compilation switches.
As seen before (see Tools Options in Project Files), setting compilation switches for all sources of a subsystem is simple: it is just a matter of adding a Compiler.Default_Switches attribute to each project files with the same value. Of course, that means duplication of data, and both places need to be changed in order to recompile the whole application with different switches. It can become a real problem if there are many subsystems and thus many project files to edit.
There are two main approaches to avoiding this duplication:
build.gpr imports logging.gpr, we could change it
to reference the attribute in Logging, either through a package renaming,
or by referencing the attribute. The following example shows both cases:
project Logging is
package Compiler is
for Switches ("Ada")
use ("-O2");
end Compiler;
package Binder is
for Switches ("Ada")
use ("-E");
end Binder;
end Logging;
with "logging.gpr";
project Build is
package Compiler renames Logging.Compiler;
package Binder is
for Switches ("Ada") use Logging.Binder'Switches ("Ada");
end Binder;
end Build;
The solution used for Compiler gets the same value for all attributes of the package, but you cannot modify anything from the package (adding extra switches or some exceptions). The second version is more flexible, but more verbose.
If you need to refer to the value of a variable in an imported project, rather than an attribute, the syntax is similar but uses a "." rather than an apostrophe. For instance:
with "imported";
project Main is
Var1 := Imported.Var;
end Main;
shared.gpr.
abstract project Shared is
for Source_Files use (); -- no sources
package Compiler is
for Switches ("Ada")
use ("-O2");
end Compiler;
end Shared;
with "shared.gpr";
project Logging is
package Compiler renames Shared.Compiler;
end Logging;
with "shared.gpr";
project Build is
package Compiler renames Shared.Compiler;
end Build;
As for the first example, we could have chosen to set the attributes one by one rather than to rename a package. The reason we explicitly indicate that Shared has no sources is so that it can be created in any directory and we are sure it shares no sources with Build or Logging, which of course would be invalid.
Note the additional use of the `abstract' qualifier in shared.gpr.
This qualifier is optional, but helps convey the message that we do not
intend this project to have sources (see Qualified Projects for
more qualifiers).
We have already seen many examples of attributes used to specify a special option of one of the tools involved in the build process. Most of those attributes are project specific. That it to say, they only affect the invocation of tools on the sources of the project where they are defined.
There are a few additional attributes that apply to all projects in a hierarchy as long as they are defined on the "main" project. The main project is the project explicitly mentioned on the command-line. The project hierarchy is the "with"-closure of the main project.
Here is a list of commonly used global attributes:
`Builder.Global_Configuration_Pragmas':
This attribute points to a file that contains configuration pragmas to use when building executables. These pragmas apply for all executables built from this project hierarchy. As we have seen before, additional pragmas can be specified on a per-project basis by setting the Compiler.Local_Configuration_Pragmas attribute.
`Builder.Global_Compilation_Switches':
This attribute is a list of compiler switches to use when compiling any source file in the project hierarchy. These switches are used in addition to the ones defined in the Compiler package, which only apply to the sources of the corresponding project. This attribute is indexed on the name of the language.
Using such global capabilities is convenient. It can also lead to unexpected behavior. Especially when several subsystems are shared among different main projects and the different global attributes are not compatible. Note that using aggregate projects can be a safer and more powerful replacement to global attributes.
Various aspects of the projects can be modified based on `scenarios'. These are user-defined modes that change the behavior of a project. Typical examples are the setup of platform-specific compiler options, or the use of a debug and a release mode (the former would activate the generation of debug information, while the second will focus on improving code optimization).
Let's enhance our example to support debug and release modes. The issue is to let the user choose what kind of system he is building: use `-g' as compiler switches in debug mode and `-O2' in release mode. We will also set up the projects so that we do not share the same object directory in both modes; otherwise switching from one to the other might trigger more recompilations than needed or mix objects from the two modes.
One naive approach is to create two different project files, say
build_debug.gpr and build_release.gpr, that set the appropriate
attributes as explained in previous sections. This solution does not scale
well, because in the presence of multiple projects depending on each other, you
will also have to duplicate the complete hierarchy and adapt the project files
to point to the right copies.
Instead, project files support the notion of scenarios controlled by external values. Such values can come from several sources (in decreasing order of priority):
gprbuild -Pbuild.gpr -Xmode=release
`External function second parameter'.
We now need to get that value in the project. The general form is to use
the predefined function `external' which returns the current value of
the external. For instance, we could set up the object directory to point to
either obj/debug or obj/release by changing our project to
project Build is
for Object_Dir use "obj/" & external ("mode", "debug");
... -- as before
end Build;
The second parameter to external is optional, and is the default value to use if "mode" is not set from the command line or the environment.
In order to set the switches according to the different scenarios, other constructs have to be introduced such as typed variables and case constructions.
A `typed variable' is a variable that can take only a limited number of values, similar to an enumeration in Ada. Such a variable can then be used in a `case construction' and create conditional sections in the project. The following example shows how this can be done:
project Build is
type Mode_Type is ("debug", "release"); -- all possible values
Mode : Mode_Type := external ("mode", "debug"); -- a typed variable
package Compiler is
case Mode is
when "debug" =>
for Switches ("Ada")
use ("-g");
when "release" =>
for Switches ("Ada")
use ("-O2");
end case;
end Compiler;
end Build;
The project has suddenly grown in size, but has become much more flexible. Mode_Type defines the only valid values for the mode variable. If any other value is read from the environment, an error is reported and the project is considered as invalid.
The Mode variable is initialized with an external value defaulting to "debug". This default could be omitted and that would force the user to define the value. Finally, we can use a case construction to set the switches depending on the scenario the user has chosen.
Most aspects of the projects can depend on scenarios. The notable exception are project dependencies (`with' clauses), which cannot depend on a scenario.
Scenarios work the same way with `project hierarchies': you can either
duplicate a variable similar to Mode in each of the project (as long
as the first argument to external is always the same and the type is
the same), or simply set the variable in the shared.gpr project
(see Sharing Between Projects).
So far, we have seen examples of projects that create executables. However, it is also possible to create libraries instead. A `library' is a specific type of subsystem where, for convenience, objects are grouped together using system-specific means such as archives or windows DLLs.
Library projects provide a system- and language-independent way of building both `static' and `dynamic' libraries. They also support the concept of `standalone libraries' (SAL) which offer two significant properties: the elaboration (e.g. initialization) of the library is either automatic or very simple; a change in the implementation part of the library implies minimal post-compilation actions on the complete system and potentially no action at all for the rest of the system in the case of dynamic SALs.
There is a restriction on shared library projects: by default, they are only allowed to import other shared library projects. They are not allowed to import non library projects or static library projects.
The GNAT Project Manager takes complete care of the library build, rebuild and
installation tasks, including recompilation of the source files for which
objects do not exist or are not up to date, assembly of the library archive, and
installation of the library (i.e., copying associated source, object and
ALI files to the specified location).
Let's enhance our example and transform the logging subsystem into a
library. In order to do so, a few changes need to be made to
logging.gpr. Some attributes need to be defined: at least
Library_Name and Library_Dir; in addition, some other attributes
can be used to specify specific aspects of the library. For readability, it is
also recommended (although not mandatory), to use the qualifier library
in front of the project keyword.
This attribute is the name of the library to be built. There is no restriction on the name of a library imposed by the project manager, except for stand-alone libraries whose names must follow the syntax of Ada identifiers; however, there may be system-specific restrictions on the name. In general, it is recommended to stick to alphanumeric characters (and possibly single underscores) to help portability.
This attribute is the path (absolute or relative) of the directory where the library is to be installed. In the process of building a library, the sources are compiled, the object files end up in the explicit or implicit Object_Dir directory. When all sources of a library are compiled, some of the compilation artifacts, including the library itself, are copied to the library_dir directory. This directory must exist and be writable. It must also be different from the object directory so that cleanup activities in the Library_Dir do not affect recompilation needs.
Here is the new version of logging.gpr that makes it a library:
library project Logging is -- "library" is optional
for Library_Name use "logging"; -- will create "liblogging.a" on Unix
for Object_Dir use "obj";
for Library_Dir use "lib"; -- different from object_dir
end Logging;
Once the above two attributes are defined, the library project is valid and is enough for building a library with default characteristics. Other library-related attributes can be used to change the defaults:
The value of this attribute must be either "static", "dynamic" or "relocatable" (the latter is a synonym for dynamic). It indicates which kind of library should be built (the default is to build a static library, that is an archive of object files that can potentially be linked into a static executable). When the library is set to be dynamic, a separate image is created that will be loaded independently, usually at the start of the main program execution. Support for dynamic libraries is very platform specific, for instance on Windows it takes the form of a DLL while on GNU/Linux, it is a dynamic elf image whose suffix is usually
.so. Library project files, on the other hand, can be written in a platform independent way so that the same project file can be used to build a library on different operating systems.If you need to build both a static and a dynamic library, it is recommended to use two different object directories, since in some cases some extra code needs to be generated for the latter. For such cases, one can either define two different project files, or a single one that uses scenarios to indicate the various kinds of library to be built and their corresponding object_dir.
This attribute may be specified to indicate the directory where the ALI files of the library are installed. By default, they are copied into the Library_Dir directory, but as for the executables where we have a separate Exec_Dir attribute, you might want to put them in a separate directory since there can be hundreds of them. The same restrictions as for the Library_Dir attribute apply.
This attribute is platform dependent, and has no effect on Windows. On Unix, it is used only for dynamic libraries as the internal name of the library (the "soname"). If the library file name (built from the Library_Name) is different from the Library_Version, then the library file will be a symbolic link to the actual file whose name will be Library_Version. This follows the usual installation schemes for dynamic libraries on many Unix systems.
project Logging is Version := "1"; for Library_Dir use "lib"; for Library_Name use "logging"; for Library_Kind use "dynamic"; for Library_Version use "liblogging.so." & Version; end Logging;After the compilation, the directory
libwill contain both alibdummy.so.1library and a symbolic link to it calledlibdummy.so.
This attribute is the name of the tool to use instead of "gcc" to link shared libraries. A common use of this attribute is to define a wrapper script that accomplishes specific actions before calling gcc (which itself calls the linker to build the library image).
This attribute may be used to specify additional switches (last switches) when linking a shared library.
It may also be used to add foreign object files to a static library. Each string in Library_Options is an absolute or relative path of an object file. When a relative path, it is relative to the object directory.
This attribute, that is taken into account only by `gprbuild', may be used to specified leading options (first switches) when linking a shared library.
This attribute specifies additional switches to be given to the linker when linking an executable. It is ignored when defined in the main project and taken into account in all other projects that are imported directly or indirectly. These switches complement the Linker.Switches defined in the main project. This is useful when a particular subsystem depends on an external library: adding this dependency as a Linker_Options in the project of the subsystem is more convenient than adding it to all the Linker.Switches of the main projects that depend upon this subsystem.
When the builder detects that a project file is a library project file, it recompiles all sources of the project that need recompilation and rebuild the library if any of the sources have been recompiled. It then groups all object files into a single file, which is a shared or a static library. This library can later on be linked with multiple executables. Note that the use of shard libraries reduces the size of the final executable and can also reduce the memory footprint at execution time when the library is shared among several executables.
`gprbuild also allows to build **multi-language libraries*' when specifying sources from multiple languages.
A non-library project can import a library project. When the builder is invoked
on the former, the library of the latter is only rebuilt when absolutely
necessary. For instance, if a unit of the library is not up-to-date but none of
the executables need this unit, then the unit is not recompiled and the library
is not reassembled. For instance, let's assume in our example that logging has
the following sources: log1.ads, log1.adb, log2.ads and
log2.adb. If log1.adb has been modified, then the library
liblogging will be rebuilt when compiling all the sources of
Build only if proc.ads, pack.ads or pack.adb
include a "with Log1".
To ensure that all the sources in the Logging library are up to date, and that all the sources of Build are also up to date, the following two commands need to be used:
gprbuild -Plogging.gpr
gprbuild -Pbuild.gpr
All ALI files will also be copied from the object directory to the
library directory. To build executables, `gprbuild' will use the
library rather than the individual object files.
Library projects can also be useful to describe a library that needs to be used but, for some reason, cannot be rebuilt. For instance, it is the case when some of the library sources are not available. Such library projects need to use the Externally_Built attribute as in the example below:
library project Extern_Lib is
for Languages use ("Ada", "C");
for Source_Dirs use ("lib_src");
for Library_Dir use "lib2";
for Library_Kind use "dynamic";
for Library_Name use "l2";
for Externally_Built use "true"; -- <<<<
end Extern_Lib;
In the case of externally built libraries, the Object_Dir attribute does not need to be specified because it will never be used.
The main effect of using such an externally built library project is mostly to affect the linker command in order to reference the desired library. It can also be achieved by using Linker.Linker_Options or Linker.Switches in the project corresponding to the subsystem needing this external library. This latter method is more straightforward in simple cases but when several subsystems depend upon the same external library, finding the proper place for the Linker.Linker_Options might not be easy and if it is not placed properly, the final link command is likely to present ordering issues. In such a situation, it is better to use the externally built library project so that all other subsystems depending on it can declare this dependency thanks to a project `with' clause, which in turn will trigger the builder to find the proper order of libraries in the final link command.
A `stand-alone library' is a library that contains the necessary code to elaborate the Ada units that are included in the library. A stand-alone library is a convenient way to add an Ada subsystem to a more global system whose main is not in Ada since it makes the elaboration of the Ada part mostly transparent. However, stand-alone libraries are also useful when the main is in Ada: they provide a means for minimizing relinking & redeployment of complex systems when localized changes are made.
The name of a stand-alone library, specified with attribute Library_Name, must have the syntax of an Ada identifier.
The most prominent characteristic of a stand-alone library is that it offers a distinction between interface units and implementation units. Only the former are visible to units outside the library. A stand-alone library project is thus characterised by a third attribute, usually `Library_Interface', in addition to the two attributes that make a project a Library Project (Library_Name and Library_Dir). This third attribute may also be `Interfaces'. `Library_Interface' only works when the interface is in Ada and takes a list of units as parameter. `Interfaces' works for any supported language and takes a list of sources as parameter.
This attribute defines an explicit subset of the units of the project. Units from projects importing this library project may only "with" units whose sources are listed in the Library_Interface. Other sources are considered implementation units.
for Library_Dir use "lib"; for Library_Name use "logging"; for Library_Interface use ("lib1", "lib2"); -- unit names
`Interfaces'
This attribute defines an explicit subset of the source files of a project. Sources from projects importing this project, can only depend on sources from this subset. This attribute can be used on non library projects. It can also be used as a replacement for attribute Library_Interface, in which case, units have to be replaced by source files. For multi-language library projects, it is the only way to make the project a Stand-Alone Library project whose interface is not purely Ada.
This attribute defines the kind of standalone library to build. Values are either standard (the default), no or encapsulated. When standard is used the code to elaborate and finalize the library is embedded, when encapsulated is used the library can furthermore depend only on static libraries (including the GNAT runtime). This attribute can be set to no to make it clear that the library should not be standalone in which case the Library_Interface should not defined. Note that this attribute only applies to shared libraries, so Library_Kind must be set to dynamic.
for Library_Dir use "lib"; for Library_Name use "logging"; for Library_Kind use "dynamic"; for Library_Interface use ("lib1", "lib2"); -- unit names for Library_Standalone use "encapsulated";
In order to include the elaboration code in the stand-alone library, the binder is invoked on the closure of the library units creating a package whose name depends on the library name (b~logging.ads/b in the example). This binder-generated package includes `initialization' and `finalization' procedures whose names depend on the library name (logginginit and loggingfinal in the example). The object corresponding to this package is included in the library.
A dynamic stand-alone Library is automatically initialized if automatic initialization of Stand-alone Libraries is supported on the platform and if attribute `Library_Auto_Init' is not specified or is specified with the value "true". A static Stand-alone Library is never automatically initialized. Specifying "false" for this attribute prevents automatic initialization.
When a non-automatically initialized stand-alone library is used in an executable, its initialization procedure must be called before any service of the library is used. When the main subprogram is in Ada, it may mean that the initialization procedure has to be called during elaboration of another package.
For a stand-alone library, only the
ALIfiles of the interface units (those that are listed in attribute Library_Interface) are copied to the library directory. As a consequence, only the interface units may be imported from Ada units outside of the library. If other units are imported, the binding phase will fail.
`Binder.Default_Switches':
When a stand-alone library is bound, the switches that are specified in the attribute `Binder.Default_Switches ("Ada")' are used in the call to `gnatbind'.
This attribute defines the location (absolute or relative to the project directory) where the sources of the interface units are copied at installation time. These sources includes the specs of the interface units along with the closure of sources necessary to compile them successfully. That may include bodies and subunits, when pragmas Inline are used, or when there are generic units in specs. This directory cannot point to the object directory or one of the source directories, but it can point to the library directory, which is the default value for this attribute.
This attribute controls the export of symbols and, on some platforms (like VMS) that have the notions of major and minor IDs built in the library files, it controls the setting of these IDs. It is not supported on all platforms (where it will just have no effect). It may have one of the following values:
- "autonomous" or "default": exported symbols are not controlled
- "compliant": if attribute `Library_Reference_Symbol_File' is not defined, then it is equivalent to policy "autonomous". If there are exported symbols in the reference symbol file that are not in the object files of the interfaces, the major ID of the library is increased. If there are symbols in the object files of the interfaces that are not in the reference symbol file, these symbols are put at the end of the list in the newly created symbol file and the minor ID is increased.
- "controlled": the attribute `Library_Reference_Symbol_File' must be defined. The library will fail to build if the exported symbols in the object files of the interfaces do not match exactly the symbol in the symbol file.
- "restricted": The attribute `Library_Symbol_File' must be defined. The library will fail to build if there are symbols in the symbol file that are not in the exported symbols of the object files of the interfaces. Additional symbols in the object files are not added to the symbol file.
- "direct": The attribute `Library_Symbol_File' must be defined and must designate an existing file in the object directory. This symbol file is passed directly to the underlying linker without any symbol processing.
`Library_Reference_Symbol_File'
This attribute may define the path name of a reference symbol file that is read when the symbol policy is either "compliant" or "controlled", on platforms that support symbol control, such as VMS, when building a stand-alone library. The path may be an absolute path or a path relative to the project directory.
This attribute may define the name of the symbol file to be created when building a stand-alone library when the symbol policy is either "compliant", "controlled" or "restricted", on platforms that support symbol control, such as VMS. When symbol policy is "direct", then a file with this name must exist in the object directory.
When using project files, a usable version of the library is created in the directory specified by the Library_Dir attribute of the library project file. Thus no further action is needed in order to make use of the libraries that are built as part of the general application build.
You may want to install a library in a context different from where the library is built. This situation arises with third party suppliers, who may want to distribute a library in binary form where the user is not expected to be able to recompile the library. The simplest option in this case is to provide a project file slightly different from the one used to build the library, by using the externally_built attribute. See Using Library Projects
Another option is to use `gprinstall' to install the library in a different context than the build location. `gprinstall' automatically generates a project to use this library, and also copies the minimum set of sources needed to use the library to the install location. Installation
During development of a large system, it is sometimes necessary to use modified versions of some of the source files, without changing the original sources. This can be achieved through the `project extension' facility.
Suppose for instance that our example Build project is built every night for the whole team, in some shared directory. A developer usually needs to work on a small part of the system, and might not want to have a copy of all the sources and all the object files (mostly because that would require too much disk space, time to recompile everything). He prefers to be able to override some of the source files in his directory, while taking advantage of all the object files generated at night.
Another example can be taken from large software systems, where it is common to have multiple implementations of a common interface; in Ada terms, multiple versions of a package body for the same spec. For example, one implementation might be safe for use in tasking programs, while another might be used only in sequential applications. This can be modeled in GNAT using the concept of `project extension'. If one project (the 'child') `extends' another project (the 'parent') then by default all source files of the parent project are inherited by the child, but the child project can override any of the parent's source files with new versions, and can also add new files or remove unnecessary ones. This facility is the project analog of a type extension in object-oriented programming. Project hierarchies are permitted (an extending project may itself be extended), and a project that extends a project can also import other projects.
A third example is that of using project extensions to provide different versions of the same system. For instance, assume that a Common project is used by two development branches. One of the branches has now been frozen, and no further change can be done to it or to Common. However, the other development branch still needs evolution of Common. Project extensions provide a flexible solution to create a new version of a subsystem while sharing and reusing as much as possible from the original one.
A project extension implicitly inherits all the sources and objects from the project it extends. It is possible to create a new version of some of the sources in one of the additional source directories of the extending project. Those new versions hide the original versions. Adding new sources or removing existing ones is also possible. Here is an example on how to extend the project Build from previous examples:
project Work extends "../bld/build.gpr" is
end Work;
The project after `extends' is the one being extended. As usual, it can be specified using an absolute path, or a path relative to any of the directories in the project path (see Project Dependencies). This project does not specify source or object directories, so the default values for these attributes will be used that is to say the current directory (where project Work is placed). We can compile that project with
gprbuild -Pwork
If no sources have been placed in the current directory, this command won't do anything, since this project does not change the sources it inherited from Build, therefore all the object files in Build and its dependencies are still valid and are reused automatically.
Suppose we now want to supply an alternate version of pack.adb but use
the existing versions of pack.ads and proc.adb. We can create
the new file in Work's current directory (likely by copying the one from the
Build project and making changes to it. If new packages are needed at
the same time, we simply create new files in the source directory of the
extending project.
When we recompile, `gprbuild' will now automatically recompile
this file (thus creating pack.o in the current directory) and
any file that depends on it (thus creating proc.o). Finally, the
executable is also linked locally.
Note that we could have obtained the desired behavior using project import
rather than project inheritance. A base project would contain the
sources for pack.ads and proc.adb, and Work would
import base and add pack.adb. In this scenario, base
cannot contain the original version of pack.adb otherwise there would be
2 versions of the same unit in the closure of the project and this is not
allowed. Generally speaking, it is not recommended to put the spec and the
body of a unit in different projects since this affects their autonomy and
reusability.
In a project file that extends another project, it is possible to indicate that an inherited source is `not part' of the sources of the extending project. This is necessary sometimes when a package spec has been overridden and no longer requires a body: in this case, it is necessary to indicate that the inherited body is not part of the sources of the project, otherwise there will be a compilation error when compiling the spec.
For that purpose, the attribute `Excluded_Source_Files' is used. Its value is a list of file names. It is also possible to use attribute Excluded_Source_List_File. Its value is the path of a text file containing one file name per line.
project Work extends "../bld/build.gpr" is
for Source_Files use ("pack.ads");
-- New spec of Pkg does not need a completion
for Excluded_Source_Files use ("pack.adb");
end Work;
All packages that are not declared in the extending project are inherited from the project being extended, with their attributes, with the exception of Linker'Linker_Options which is never inherited. In particular, an extending project retains all the switches specified in the project being extended.
At the project level, if they are not declared in the extending project, some attributes are inherited from the project being extended. They are: Languages, Main (for a root non library project) and Library_Name (for a project extending a library project).
One of the fundamental restrictions in project extension is the following: `A project is not allowed to import directly or indirectly at the same time an extending project and one of its ancestors'.
For example, consider the following hierarchy of projects.
a.gpr contains package A1
b.gpr, imports a.gpr and contains B1, which depends on A1
c.gpr, imports b.gpr and contains C1, which depends on B1
If we want to locally extend the packages A1 and C1, we need to create several extending projects:
a_ext.gpr which extends a.gpr, and overrides A1
b_ext.gpr which extends b.gpr and imports a_ext.gpr
c_ext.gpr which extends c.gpr, imports b_ext.gpr and overrides C1
project A_Ext extends "a.gpr" is
for Source_Files use ("a1.adb", "a1.ads");
end A_Ext;
with "a_ext.gpr";
project B_Ext extends "b.gpr" is
end B_Ext;
with "b_ext.gpr";
project C_Ext extends "c.gpr" is
for Source_Files use ("c1.adb");
end C_Ext;
The extension b_ext.gpr is required, even though we are not overriding
any of the sources of b.gpr because otherwise c_expr.gpr would
import b.gpr which itself knows nothing about a_ext.gpr.
When extending a large system spanning multiple projects, it is often inconvenient to extend every project in the hierarchy that is impacted by a small change introduced in a low layer. In such cases, it is possible to create an `implicit extension' of an entire hierarchy using `extends all' relationship.
When the project is extended using extends all inheritance, all projects that are imported by it, both directly and indirectly, are considered virtually extended. That is, the project manager creates implicit projects that extend every project in the hierarchy; all these implicit projects do not control sources on their own and use the object directory of the "extending all" project.
It is possible to explicitly extend one or more projects in the hierarchy in order to modify the sources. These extending projects must be imported by the "extending all" project, which will replace the corresponding virtual projects with the explicit ones.
When building such a project hierarchy extension, the project manager will ensure that both modified sources and sources in implicit extending projects that depend on them are recompiled.
Thus, in our example we could create the following projects instead:
a_ext.gpr, extends a.gpr and overrides A1
c_ext.gpr, "extends all" c.gpr, imports a_ext.gpr and overrides C1
project A_Ext extends "a.gpr" is
for Source_Files use ("a1.adb", "a1.ads");
end A_Ext;
with "a_ext.gpr";
project C_Ext extends all "c.gpr" is
for Source_Files use ("c1.adb");
end C_Ext;
When building project c_ext.gpr, the entire modified project space is
considered for recompilation, including the sources of b.gpr that are
impacted by the changes in A1 and C1.
Aggregate projects are an extension of the project paradigm, and are meant to solve a few specific use cases that cannot be solved directly using standard projects. This section will go over a few of these use cases to try to explain what you can use aggregate projects for.
Most often, an application is organized into modules and submodules, which are very conveniently represented as a project tree or graph (the root project A `with's the projects for each modules (say B and C), which in turn `with' projects for submodules.
Very often, modules will build their own executables (for testing purposes for instance), or libraries (for easier reuse in various contexts).
However, if you build your project through `gprbuild', using a syntax similar to
gprbuild -PA.gpr
this will only rebuild the main programs of project A, not those of the imported projects B and C. Therefore you have to spawn several `gprbuild' commands, one per project, to build all executables. This is a little inconvenient, but more importantly is inefficient because `gprbuild' needs to do duplicate work to ensure that sources are up-to-date, and cannot easily compile things in parallel when using the -j switch.
Also libraries are always rebuilt when building a project.
You could therefore define an aggregate project Agg that groups A, B and C. Then, when you build with
gprbuild -PAgg.gpr
this will build all mains from A, B and C.
aggregate project Agg is
for Project_Files use ("a.gpr", "b.gpr", "c.gpr");
end Agg;
If B or C do not define any main program (through their Main attribute), all their sources are built. When you do not group them in the aggregate project, only those sources that are needed by A will be built.
If you add a main to a project P not already explicitly referenced in the aggregate project, you will need to add "p.gpr" in the list of project files for the aggregate project, or the main will not be built when building the aggregate project.
One other case is when you have multiple applications and libraries that are built independently from each other (but can be built in parallel). For instance, you have a project tree rooted at A, and another one (which might share some subprojects) rooted at B.
Using only `gprbuild', you could do
gprbuild -PA.gpr
gprbuild -PB.gpr
to build both. But again, `gprbuild' has to do some duplicate work for those files that are shared between the two, and cannot truly build things in parallel efficiently.
If the two projects are really independent, share no sources other than through a common subproject, and have no source files with a common basename, you could create a project C that imports A and B. But these restrictions are often too strong, and one has to build them independently. An aggregate project does not have these limitations and can aggregate two project trees that have common sources.
This scenario is particularly useful in environments like VxWorks 653 where the applications running in the multiple partitions can be built in parallel through a single `gprbuild' command. This also works nicely with Annex E.
The environment variables at the time you launch `gprbuild' will influence the view these tools have of the project (PATH to find the compiler, ADA_PROJECT_PATH or GPR_PROJECT_PATH to find the projects, environment variables that are referenced in project files through the "external" built-in function, ...). Several command line switches can be used to override those (-X or -aP), but on some systems and with some projects, this might make the command line too long, and on all systems often make it hard to read.
An aggregate project can be used to set the environment for all projects built through that aggregate. One of the nice aspects is that you can put the aggregate project under configuration management, and make sure all your user have a consistent environment when building. The syntax looks like
aggregate project Agg is
for Project_Files use ("A.gpr", "B.gpr");
for Project_Path use ("../dir1", "../dir1/dir2");
for External ("BUILD") use "PRODUCTION";
package Builder is
for Global_Compilation_Switches ("Ada") use ("-g");
end Builder;
end Agg;
One of the often requested features in projects is to be able to reference external variables in `with' declarations, as in
with external("SETUP") & "path/prj.gpr"; -- ILLEGAL
project MyProject is
...
end MyProject;
For various reasons, this is not allowed. But using aggregate projects provide an elegant solution. For instance, you could use a project file like:
aggregate project Agg is
for Project_Path use (external("SETUP") & "path");
for Project_Files use ("myproject.gpr");
end Agg;
with "prj.gpr"; -- searched on Agg'Project_Path
project MyProject is
...
end MyProject;
The loading of aggregate projects is optimized in `gprbuild', so that all files are searched for only once on the disk (thus reducing the number of system calls and contributing to faster compilation times, especially on systems with sources on remote servers). As part of the loading, `gprbuild' computes how and where a source file should be compiled, and even if it is found several times in the aggregated projects it will be compiled only once.
Since there is no ambiguity as to which switches should be used, files can be compiled in parallel (through the usual -j switch) and this can be done while maximizing the use of CPUs (compared to launching multiple `gprbuild' commands in parallel).
An aggregate project follows the general syntax of project files. The
recommended extension is still .gpr. However, a special
aggregate qualifier must be put before the keyword
project.
An aggregate project cannot `with' any other project (standard or aggregate), except an abstract project which can be used to share attribute values. Also, aggregate projects cannot be extended or imported though a `with' clause by any other project. Building other aggregate projects from an aggregate project is done through the Project_Files attribute (see below).
An aggregate project does not have any source files directly (only through other standard projects). Therefore a number of the standard attributes and packages are forbidden in an aggregate project. Here is the (non exhaustive) list:
The only package that is authorized (albeit optional) is Builder. Other packages (in particular Compiler, Binder and Linker) are forbidden.
The following three attributes can be used only in an aggregate project:
This attribute is compulsory (or else we are not aggregating any project, and thus not doing anything). It specifies a list of
.gprfiles that are grouped in the aggregate. The list may be empty. The project files can be either other aggregate projects, or standard projects. When grouping standard projects, you can have both the root of a project tree (and you do not need to specify all its imported projects), and any project within the tree.Basically, the idea is to specify all those projects that have main programs you want to build and link, or libraries you want to build. You can even specify projects that do not use the Main attribute nor the Library_* attributes, and the result will be to build all their source files (not just the ones needed by other projects).
The file can include paths (absolute or relative). Paths are relative to the location of the aggregate project file itself (if you use a base name, we expect to find the .gpr file in the same directory as the aggregate project file). The environment variables ADA_PROJECT_PATH, GPR_PROJECT_PATH and GPR_PROJECT_PATH_FILE are not used to find the project files. The extension
.gpris mandatory, since this attribute contains file names, not project names.Paths can also include the "*" and "**" globbing patterns. The latter indicates that any subdirectory (recursively) will be searched for matching files. The latter ("**") can only occur at the last position in the directory part (ie "a/**/*.gpr" is supported, but not "**/a/*.gpr"). Starting the pattern with "**" is equivalent to starting with "./**".
For now, the pattern "*" is only allowed in the filename part, not in the directory part. This is mostly for efficiency reasons to limit the number of system calls that are needed.
Here are a few valid examples:
for Project_Files use ("a.gpr", "subdir/b.gpr"); -- two specific projects relative to the directory of agg.gpr for Project_Files use ("/.gpr"); -- all projects recursively
This attribute can be used to specify a list of directories in which to look for project files in `with' declarations.
When you specify a project in Project_Files (say x/y/a.gpr), and a.gpr imports a project b.gpr, only b.gpr is searched in the project path. a.gpr must be exactly at <dir of the aggregate>/x/y/a.gpr.
This attribute, however, does not affect the search for the aggregated project files specified with Project_Files.
Each aggregate project has its own Project_Path (that is if agg1.gpr includes agg2.gpr, they can potentially both have a different Project_Path).
This project path is defined as the concatenation, in that order, of:
- the current directory;
- followed by the command line -aP switches;
- then the directories from the GPR_PROJECT_PATH and ADA_PROJECT_PATH environment variables;
- then the directories from the Project_Path attribute;
- and finally the predefined directories.
In the example above, agg2.gpr's project path is not influenced by the attribute agg1'Project_Path, nor is agg1 influenced by agg2'Project_Path.
This can potentially lead to errors. Consider the following example:
-- -- +---------------+ +----------------+ -- | Agg1.gpr |-=--includes--=-->| Agg2.gpr | -- | 'project_path| | 'project_path | -- | | | | -- +---------------+ +----------------+ -- : : -- includes includes -- : : -- v v -- +-------+ +---------+ -- | P.gpr |<---------- withs --------| Q.gpr | -- +-------+---------\ +---------+ -- | | -- withs | -- | | -- v v -- +-------+ +---------+ -- | R.gpr | | R'.gpr | -- +-------+ +---------+When looking for p.gpr, both aggregates find the same physical file on the disk. However, it might happen that with their different project paths, both aggregate projects would in fact find a different r.gpr. Since we have a common project (p.gpr) "with"ing two different r.gpr, this will be reported as an error by the builder.
Directories are relative to the location of the aggregate project file.
Example:
for Project_Path use ("/usr/local/gpr", "gpr/");
This attribute can be used to set the value of environment variables as retrieved through the external function in projects. It does not affect the environment variables themselves (so for instance you cannot use it to change the value of your PATH as seen from the spawned compiler).
This attribute affects the external values as seen in the rest of the aggregate project, and in the aggregated projects.
The exact value of external a variable comes from one of three sources (each level overrides the previous levels):
- An External attribute in aggregate project, for instance for External ("BUILD_MODE") use "DEBUG";
- Environment variables. These override the value given by the attribute, so that users can override the value set in the (presumably shared with others team members) aggregate project.
- The -X command line switch to `gprbuild'. This always takes precedence.
This attribute is only taken into account in the main aggregate project (i.e. the one specified on the command line to `gprbuild'), and ignored in other aggregate projects. It is invalid in standard projects. The goal is to have a consistent value in all projects that are built through the aggregate, which would not be the case in the diamond case: A groups the aggregate projects B and C, which both (either directly or indirectly) build the project P. If B and C could set different values for the environment variables, we would have two different views of P, which in particular might impact the list of source files in P.
As mentioned above, only the package Builder can be specified in an aggregate project. In this package, only the following attributes are valid:
This attribute gives the list of switches to use for `gprbuild'. Because no mains can be specified for aggregate projects, the only possible index for attribute Switches is others. All other indexes will be ignored.
Example:
for Switches (others) use ("-v", "-k", "-j8");These switches are only read from the main aggregate project (the one passed on the command line), and ignored in all other aggregate projects or projects.
It can only contain builder switches, not compiler switches.
This attribute gives the list of compiler switches for the various languages. For instance,
for Global_Compilation_Switches ("Ada") use ("O1", "-g"); for Global_Compilation_Switches ("C") use ("-O2");This attribute is only taken into account in the aggregate project specified on the command line, not in other aggregate projects.
In the projects grouped by that aggregate, the attribute Builder.Global_Compilation_Switches is also ignored. However, the attribute Compiler.Default_Switches will be taken into account (but that of the aggregate have higher priority). The attribute Compiler.Switches is also taken into account and can be used to override the switches for a specific file. As a result, it always has priority.
The rules are meant to avoid ambiguities when compiling. For instance, aggregate project Agg groups the projects A and B, that both depend on C. Here is an extra for all of these projects:
aggregate project Agg is for Project_Files use ("a.gpr", "b.gpr"); package Builder is for Global_Compilation_Switches ("Ada") use ("-O2"); end Builder; end Agg; with "c.gpr"; project A is package Builder is for Global_Compilation_Switches ("Ada") use ("-O1"); -- ignored end Builder; package Compiler is for Default_Switches ("Ada") use ("-O1", "-g"); for Switches ("a_file1.adb") use ("-O0"); end Compiler; end A; with "c.gpr"; project B is package Compiler is for Default_Switches ("Ada") use ("-O0"); end Compiler; end B; project C is package Compiler is for Default_Switches ("Ada") use ("-O3", "-gnatn"); for Switches ("c_file1.adb") use ("-O0", "-g"); end Compiler; end C;then the following switches are used:
- all files from project A except a_file1.adb are compiled with "-O2 -g", since the aggregate project has priority.
- the file a_file1.adb is compiled with "-O0", since the Compiler.Switches has priority
- all files from project B are compiled with "-O2", since the aggregate project has priority
- all files from C are compiled with "-O2 -gnatn", except for c_file1.adb which is compiled with "-O0 -g"
Even though C is seen through two paths (through A and through B), the switches used by the compiler are unambiguous.
`Global_Configuration_Pragmas'
This attribute can be used to specify a file containing configuration pragmas, to be passed to the Ada compiler. Since we ignore the package Builder in other aggregate projects and projects, only those pragmas defined in the main aggregate project will be taken into account.
Projects can locally add to those by using the Compiler.Local_Configuration_Pragmas attribute if they need.
This attribute, indexed with a language name, can be used to specify a config when compiling sources of the language. For Ada, these files are configuration pragmas files.
For projects that are built through the aggregate, the package Builder is ignored, except for the Executable attribute which specifies the name of the executables resulting from the link of the main programs, and for the Executable_Suffix.
Aggregate library projects make it possible to build a single library using object files built using other standard or library projects. This gives the flexibility to describe an application as having multiple modules (a GUI, database access, ...) using different project files (so possibly built with different compiler options) and yet create a single library (static or relocatable) out of the corresponding object files.
For example, we can define an aggregate project Agg that groups A, B and C:
aggregate library project Agg is
for Project_Files use ("a.gpr", "b.gpr", "c.gpr");
for Library_Name use ("agg");
for Library_Dir use ("lagg");
end Agg;
Then, when you build with:
gprbuild agg.gpr
This will build all units from projects A, B and C and will create a
static library named libagg.a in the lagg
directory. An aggregate library project has the same set of
restriction as a standard library project.
Note that a shared aggregate library project cannot aggregate a static library project. In platforms where a compiler option is required to create relocatable object files, a Builder package in the aggregate library project may be used:
aggregate library project Agg is
for Project_Files use ("a.gpr", "b.gpr", "c.gpr");
for Library_Name use ("agg");
for Library_Dir use ("lagg");
for Library_Kind use "relocatable";
package Builder is
for Global_Compilation_Switches ("Ada") use ("-fPIC");
end Builder;
end Agg;
With the above aggregate library Builder package, the -fPIC
option will be passed to the compiler when building any source code
from projects a.gpr, b.gpr and c.gpr.
An aggregate library project follows the general syntax of project
files. The recommended extension is still .gpr. However, a special
aggregate library qualifier must be put before the keyword
project.
An aggregate library project cannot `with' any other project (standard or aggregate), except an abstract project which can be used to share attribute values.
An aggregate library project does not have any source files directly (only through other standard projects). Therefore a number of the standard attributes and packages are forbidden in an aggregate library project. Here is the (non exhaustive) list:
The only package that is authorized (albeit optional) is Builder.
The Project_Files attribute (See Aggregate Projects) is used to described the aggregated projects whose object files have to be included into the aggregate library. The environment variables ADA_PROJECT_PATH, GPR_PROJECT_PATH and GPR_PROJECT_PATH_FILE are not used to find the project files.
This section describes the syntactic structure of project files, the various constructs that can be used. Finally, it ends with a summary of all available attributes.
Project files have an Ada-like syntax. The minimal project file is:
project Empty is
end Empty;
The identifier Empty is the name of the project. This project name must be present after the reserved word end at the end of the project file, followed by a semi-colon.
`Identifiers' (i.e., the user-defined names such as project or variable names) have the same syntax as Ada identifiers: they must start with a letter, and be followed by zero or more letters, digits or underscore characters; it is also illegal to have two underscores next to each other. Identifiers are always case-insensitive ("Name" is the same as "name").
simple_name ::= identifier
name ::= simple_name { . simple_name }
`Strings' are used for values of attributes or as indexes for these attributes. They are in general case sensitive, except when noted otherwise (in particular, strings representing file names will be case insensitive on some systems, so that "file.adb" and "File.adb" both represent the same file).
`Reserved words' are the same as for standard Ada 95, and cannot
be used for identifiers. In particular, the following words are currently
used in project files, but others could be added later on. In bold are the
extra reserved words in project files:
all, at, case, end, for, is, limited,
null, others, package, renames, type, use, when,
with, `extends', `external', `project'.
`Comments' in project files have the same syntax as in Ada, two consecutive hyphens through the end of the line.
A project may be an `independent project', entirely defined by a single project file. Any source file in an independent project depends only on the predefined library and other source files in the same project. But a project may also depend on other projects, either by importing them through `with clauses', or by `extending' at most one other project. Both types of dependency can be used in the same project.
A path name denotes a project file. It can be absolute or relative. An absolute path name includes a sequence of directories, in the syntax of the host operating system, that identifies uniquely the project file in the file system. A relative path name identifies the project file, relative to the directory that contains the current project, or relative to a directory listed in the environment variables ADA_PROJECT_PATH and GPR_PROJECT_PATH. Path names are case sensitive if file names in the host operating system are case sensitive. As a special case, the directory separator can always be "/" even on Windows systems, so that project files can be made portable across architectures. The syntax of the environment variables ADA_PROJECT_PATH and GPR_PROJECT_PATH is a list of directory names separated by colons on UNIX and semicolons on Windows.
A given project name can appear only once in a context clause.
It is illegal for a project imported by a context clause to refer, directly or indirectly, to the project in which this context clause appears (the dependency graph cannot contain cycles), except when one of the with clauses in the cycle is a `limited with'.
with "other_project.gpr";
project My_Project extends "extended.gpr" is
end My_Project;
These dependencies form a `directed graph', potentially cyclic when using `limited with'. The subgraph reflecting the `extends' relations is a tree.
A project's `immediate sources' are the source files directly defined by that project, either implicitly by residing in the project source directories, or explicitly through any of the source-related attributes. More generally, a project's `sources' are the immediate sources of the project together with the immediate sources (unless overridden) of any project on which it depends directly or indirectly.
A `project hierarchy' can be created, where projects are children of other projects. The name of such a child project must be Parent.Child, where Parent is the name of the parent project. In particular, this makes all `with' clauses of the parent project automatically visible in the child project.
project ::= context_clause project_declaration
context_clause ::= {with_clause}
with_clause ::= *with* path_name { , path_name } ;
path_name ::= string_literal
project_declaration ::= simple_project_declaration | project_extension
simple_project_declaration ::=
project <project_>name is
{declarative_item}
end <project_>simple_name;
Before the reserved project, there may be one or two `qualifiers', that is identifiers or reserved words, to qualify the project. The current list of qualifiers is:
Declarations introduce new entities that denote types, variables, attributes, and packages. Some declarations can only appear immediately within a project declaration. Others can appear within a project or within a package.
declarative_item ::= simple_declarative_item
| typed_string_declaration
| package_declaration
simple_declarative_item ::= variable_declaration
| typed_variable_declaration
| attribute_declaration
| case_construction
| empty_declaration
empty_declaration ::= *null* ;
An empty declaration is allowed anywhere a declaration is allowed. It has no effect.
A project file may contain `packages', that group attributes (typically all the attributes that are used by one of the GNAT tools).
A package with a given name may only appear once in a project file. The following packages are currently supported in project files (See Attributes for the list of attributes that each can contain).
This package specifies the naming conventions that apply to the source files in a project. In particular, these conventions are used to automatically find all source files in the source directories, or given a file name to find out its language for proper processing. See Naming Schemes.
In its simplest form, a package may be empty:
project Simple is
package Builder is
end Builder;
end Simple;
A package may contain `attribute declarations', `variable declarations' and `case constructions', as will be described below.
When there is ambiguity between a project name and a package name, the name always designates the project. To avoid possible confusion, it is always a good idea to avoid naming a project with one of the names allowed for packages or any name that starts with gnat.
A package can also be defined by a `renaming declaration'. The new package renames a package declared in a different project file, and has the same attributes as the package it renames. The name of the renamed package must be the same as the name of the renaming package. The project must contain a package declaration with this name, and the project must appear in the context clause of the current project, or be its parent project. It is not possible to add or override attributes to the renaming project. If you need to do so, you should use an `extending declaration' (see below).
Packages that are renamed in other project files often come from project files that have no sources: they are just used as templates. Any modification in the template will be reflected automatically in all the project files that rename a package from the template. This is a very common way to share settings between projects.
Finally, a package can also be defined by an `extending declaration'. This is similar to a `renaming declaration', except that it is possible to add or override attributes.
package_declaration ::= package_spec | package_renaming | package_extension
package_spec ::=
package <package_>simple_name is
{simple_declarative_item}
end package_identifier ;
package_renaming ::==
package <package_>simple_name renames <project_>simple_name.package_identifier ;
package_extension ::==
package <package_>simple_name extends <project_>simple_name.package_identifier is
{simple_declarative_item}
end package_identifier ;
An expression is any value that can be assigned to an attribute or a variable. It is either a literal value, or a construct requiring runtime computation by the project manager. In a project file, the computed value of an expression is either a string or a list of strings.
A string value is one of:
A list of strings is one of the following:
The following is the grammar for expressions
string_literal ::= "{string_element}" -- Same as Ada
string_expression ::= string_literal
| *variable_*name
| external_value
| attribute_reference
| ( string_expression { & string_expression } )
string_list ::= ( string_expression { , string_expression } )
| *string_variable*_name
| *string_*attribute_reference
term ::= string_expression | string_list
expression ::= term { & term } -- Concatenation
Concatenation involves strings and list of strings. As soon as a list of strings is involved, the result of the concatenation is a list of strings. The following Ada declarations show the existing operators:
function "&" (X : String; Y : String) return String;
function "&" (X : String_List; Y : String) return String_List;
function "&" (X : String_List; Y : String_List) return String_List;
Here are some specific examples:
List := () & File_Name; -- One string in this list
List2 := List & (File_Name & ".orig"); -- Two strings
Big_List := List & Lists2; -- Three strings
Illegal := "gnat.adc" & List2; -- Illegal, must start with list
An external value is an expression whose value is obtained from the command that invoked the processing of the current project file (typically a `gprbuild' command).
There are two kinds of external values, one that returns a single string, and one that returns a string list.
The syntax of a single string external value is:
external_value ::= *external* ( string_literal [, string_literal] )
The first string_literal is the string to be used on the command line or in the environment to specify the external value. The second string_literal, if present, is the default to use if there is no specification for this external value either on the command line or in the environment.
Typically, the external value will either exist in the
environment variables
or be specified on the command line through the
-X`vbl'=`value' switch. If both
are specified, then the command line value is used, so that a user can more
easily override the value.
The function external always returns a string. It is an error if the value was not found in the environment and no default was specified in the call to external.
An external reference may be part of a string expression or of a string list expression, and can therefore appear in a variable declaration or an attribute declaration.
Most of the time, this construct is used to initialize typed variables, which are then used in `case' constructions to control the value assigned to attributes in various scenarios. Thus such variables are often called `scenario variables'.
The syntax for a string list external value is:
external_value ::= *external_as_list* ( string_literal , string_literal )
The first string_literal is the string to be used on the command line or in the environment to specify the external value. The second string_literal is the separator between each component of the string list.
If the external value does not exist in the environment or on the command line, the result is an empty list. This is also the case, if the separator is an empty string or if the external value is only one separator.
Any separator at the beginning or at the end of the external value is discarded. Then, if there is no separator in the external value, the result is a string list with only one string. Otherwise, any string between the beginning and the first separator, between two consecutive separators and between the last separator and the end are components of the string list.
*external_as_list* ("SWITCHES", ",")
If the external value is "-O2,-g", the result is ("-O2", "-g").
If the external value is ",-O2,-g,", the result is also ("-O2", "-g").
if the external value is "-gnatv", the result is ("-gnatv").
If the external value is ",,", the result is ("").
If the external value is ",", the result is (), the empty string list.
A `type declaration' introduces a discrete set of string literals. If a string variable is declared to have this type, its value is restricted to the given set of literals. These are the only named types in project files. A string type may only be declared at the project level, not inside a package.
typed_string_declaration ::=
*type* *<typed_string_>*_simple_name *is* ( string_literal {, string_literal} );
The string literals in the list are case sensitive and must all be different. They may include any graphic characters allowed in Ada, including spaces. Here is an example of a string type declaration:
type OS is ("NT", "nt", "Unix", "GNU/Linux", "other OS");
Variables of a string type are called `typed variables'; all other variables are called `untyped variables'. Typed variables are particularly useful in case constructions, to support conditional attribute declarations. (See Case Constructions).
A string type may be referenced by its name if it has been declared in the same project file, or by an expanded name whose prefix is the name of the project in which it is declared.
`Variables' store values (strings or list of strings) and can appear as part of an expression. The declaration of a variable creates the variable and assigns the value of the expression to it. The name of the variable is available immediately after the assignment symbol, if you need to reuse its old value to compute the new value. Before the completion of its first declaration, the value of a variable defaults to the empty string ("").
A `typed' variable can be used as part of a `case' expression to compute the value, but it can only be declared once in the project file, so that all case constructions see the same value for the variable. This provides more consistency and makes the project easier to understand. The syntax for its declaration is identical to the Ada syntax for an object declaration. In effect, a typed variable acts as a constant.
An `untyped' variable can be declared and overridden multiple times within the same project. It is declared implicitly through an Ada assignment. The first declaration establishes the kind of the variable (string or list of strings) and successive declarations must respect the initial kind. Assignments are executed in the order in which they appear, so the new value replaces the old one and any subsequent reference to the variable uses the new value.
A variable may be declared at the project file level, or within a package.
typed_variable_declaration ::=
*<typed_variable_>*simple_name : *<typed_string_>*name := string_expression;
variable_declaration ::= *<variable_>*simple_name := expression;
Here are some examples of variable declarations:
This_OS : OS := external ("OS"); -- a typed variable declaration
That_OS := "GNU/Linux"; -- an untyped variable declaration
Name := "readme.txt";
Save_Name := Name & ".saved";
Empty_List := ();
List_With_One_Element := ("-gnaty");
List_With_Two_Elements := List_With_One_Element & "-gnatg";
Long_List := ("main.ada", "pack1_.ada", "pack1.ada", "pack2_.ada");
A `variable reference' may take several forms:
A `context' may be one of the following:
A `case' construction is used in a project file to effect conditional behavior. Through this construction, you can set the value of attributes and variables depending on the value previously assigned to a typed variable.
All choices in a choice list must be distinct. Unlike Ada, the choice lists of all alternatives do not need to include all values of the type. An others choice must appear last in the list of alternatives.
The syntax of a case construction is based on the Ada case construction (although the null declaration for empty alternatives is optional).
The case expression must be a string variable, either typed or not, whose value is often given by an external reference (see External Values).
Each alternative starts with the reserved word when, either a list of literal strings separated by the "|" character or the reserved word others, and the "=>" token. When the case expression is a typed string variable, each literal string must belong to the string type that is the type of the case variable. After each =>, there are zero or more declarations. The only declarations allowed in a case construction are other case constructions, attribute declarations and variable declarations. String type declarations and package declarations are not allowed. Variable declarations are restricted to variables that have already been declared before the case construction.
case_construction ::=
*case* *<variable_>*name *is* {case_item} *end case* ;
case_item ::=
*when* discrete_choice_list =>
{case_declaration
| attribute_declaration
| variable_declaration
| empty_declaration}
discrete_choice_list ::= string_literal {| string_literal} | *others*
Here is a typical example, with a typed string variable:
project MyProj is
type OS_Type is ("GNU/Linux", "Unix", "NT", "VMS");
OS : OS_Type := external ("OS", "GNU/Linux");
package Compiler is
case OS is
when "GNU/Linux" | "Unix" =>
for Switches ("Ada")
use ("-gnath");
when "NT" =>
for Switches ("Ada")
use ("-gnatP");
when others =>
null;
end case;
end Compiler;
end MyProj;
A project (and its packages) may have `attributes' that define the project's properties. Some attributes have values that are strings; others have values that are string lists.
attribute_declaration ::=
simple_attribute_declaration | indexed_attribute_declaration
simple_attribute_declaration ::= *for* attribute_designator *use* expression ;
indexed_attribute_declaration ::=
*for* *<indexed_attribute_>*simple_name ( string_literal) *use* expression ;
attribute_designator ::=
*<simple_attribute_>*simple_name
| *<indexed_attribute_>*simple_name ( string_literal )
There are two categories of attributes: `simple attributes' and `indexed attributes'. Each simple attribute has a default value: the empty string (for string attributes) and the empty list (for string list attributes). An attribute declaration defines a new value for an attribute, and overrides the previous value. The syntax of a simple attribute declaration is similar to that of an attribute definition clause in Ada.
Some attributes are indexed. These attributes are mappings whose domain is a set of strings. They are declared one association at a time, by specifying a point in the domain and the corresponding image of the attribute. Like untyped variables and simple attributes, indexed attributes may be declared several times. Each declaration supplies a new value for the attribute, and replaces the previous setting.
Here are some examples of attribute declarations:
-- simple attributes
for Object_Dir use "objects";
for Source_Dirs use ("units", "test/drivers");
-- indexed attributes
for Body ("main") use "Main.ada";
for Switches ("main.ada")
use ("-v", "-gnatv");
for Switches ("main.ada") use Builder'Switches ("main.ada") & "-g";
-- indexed attributes copy (from package Builder in project Default)
-- The package name must always be specified, even if it is the current
-- package.
for Default_Switches use Default.Builder'Default_Switches;
Attributes references may appear anywhere in expressions, and are used to retrieve the value previously assigned to the attribute. If an attribute has not been set in a given package or project, its value defaults to the empty string or the empty list, with some exceptions.
attribute_reference ::=
attribute_prefix ' *<simple_attribute>_*simple_name [ (string_literal) ]
attribute_prefix ::= *project*
| *<project_>*simple_name
| package_identifier
| *<project_>*simple_name . package_identifier
Examples are:
<project>'Object_Dir
Naming'Dot_Replacement
Imported_Project'Source_Dirs
Imported_Project.Naming'Casing
Builder'Default_Switches ("Ada")
The exceptions to the empty defaults are:
The prefix of an attribute may be:
In the following sections, all predefined attributes are succinctly described, first the project level attributes, that is those attributes that are not in a package, then the attributes in the different packages.
It is possible for different tools to dynamically create new packages with attributes, or new attributes in predefined packages. These attributes are not documented here.
The attributes under Configuration headings are usually found only in configuration project files.
The characteristics of each attribute are indicated as follows:
The value of an attribute may be a single string, indicated by the word "single", or a string list, indicated by the word "list".
When the attribute is read-only, that is when it is not allowed to declare the attribute, this is indicated by the words "read-only".
If it is allowed in the value of the attribute (both single and list) to have an optional index, this is indicated by the words "optional index".
When it is an indexed attribute, this is indicated by the word "indexed".
For an indexed attribute, if the index is case-insensitive, this is indicated by the words "case-insensitive index".
For an indexed attribute, when the index is a file name, this is indicated by the words "file name index". The index may or may not be case-sensitive, depending on the platform.
For an indexed attribute, if it is allowed to use `others' as the index, this is indicated by the words "others allowed".
When `others' is used as the index of an indexed attribute, the value of the attribute indexed by `others' is used when no other index would apply.
The name of the project.
The path name of the project directory.
The list of main sources for the executables.
The list of languages of the sources of the project.
The index is the file name of an executable source. Indicates the list of units from the main project that need to be bound and linked with their closures with the executable. The index is either a file name, a language name or "*". The roots for an executable source are those in `Roots' with an index that is the executable source file name, if declared. Otherwise, they are those in `Roots' with an index that is the language name of the executable source, if present. Otherwise, they are those in `Roots ("*")', if declared. If none of these three possibilities are declared, then there are no roots for the executable source.
Indicates if the project is externally built. Only case-insensitive values allowed are "true" and "false", the default.
Indicates the object directory for the project.
Indicates the exec directory for the project, that is the directory where the executables are.
The list of source directories of the project.
Index is a language name. Value is a list of language names. Indicates that in the source search path of the index language the source directories of the languages in the list should be included.
Example:
for Inherit_Source_Path ("C++") use ("C");
The list of directories that are included in Source_Dirs but are not source directories of the project.
Value is a list of simple names for subdirectories that are removed from the list of source directories, including theur subdirectories.
Value is a list of source file simple names.
Obsolescent. Equivalent to Excluded_Source_Files.
Value is a list of simple file names that are not sources of the project. Allows to remove sources that are inherited or found in the source directories and that match the naming scheme.
Value is a text file name that contains a list of source file simple names, one on each line.
Value is a text file name that contains a list of file simple names that are not sources of the project.
Value is a list of file names that constitutes the interfaces of the project.
Value is the list of aggregated projects.
Value is a list of directories that are added to the project search path when looking for the aggregated projects.
Index is the name of an external reference. Value is the value of the external reference to be used when parsing the aggregated projects.
Value is the name of the library directory. This attribute needs to be declared for each library project.
Value is the name of the library. This attribute needs to be declared or inherited for each library project.
Specifies the kind of library: static library (archive) or shared library. Case-insensitive values must be one of "static" for archives (the default) or "dynamic" or "relocatable" for shared libraries.
Value is the name of the library file.
Value is the list of unit names that constitutes the interfaces of a Stand-Alone Library project.
Specifies if a Stand-Alone Library (SAL) is encapsulated or not. Only authorized case-insensitive values are "standard" for non encapsulated SALs, "encapsulated" for encapsulated SALs or "no" for non SAL library project.
Value is a list of options that need to be used when linking an encapsulated Stand-Alone Library.
Indicates if encapsulated Stand-Alone Libraries are supported. Only authorized case-insensitive values are "true" and "false" (the default).
Indicates if a Stand-Alone Library is auto-initialized. Only authorized case-insentive values are "true" and "false".
Value is a list of options that are to be used at the beginning of the command line when linking a shared library.
Value is a list of options that are to be used when linking a shared library.
Index is a language name. Value is a list of options for an invocation of the compiler of the language. This invocation is done for a shared library project with sources of the language. The output of the invocation is the path name of a shared library file. The directory name is to be put in the run path option switch when linking the shared library for the project.
Value is the name of the directory where copies of the sources of the interfaces of a Stand-Alone Library are to be copied.
Value is the name of the directory where the ALI files of the interfaces of a Stand-Alone Library are to be copied. When this attribute is not declared, the directory is the library directory.
Obsolescent attribute. Specify the linker driver used to link a shared library. Use instead attribute Linker'Driver.
Value is the name of the library symbol file.
Indicates the symbol policy kind. Only authorized case-insensitive values are "autonomous", "default", "compliant", "controlled" or "direct".
Value is the name of the reference symbol file.
Value is the case-insensitive name of the language of a project when attribute Languages is not specified.
Value is the list of switches to be used when specifying the run path option in an executable.
Value is the the string that may replace the path name of the executable directory in the run path options.
Indicates if there may be several run path options specified when linking an executable. Only authorized case-insensitive values are "true" or "false" (the default).
Index is a language name. Specify the version of a toolchain for a language.
Obsolescent. No longer used.
Index is a language name. Indicates if invoking the compiler for a language produces an object file. Only authorized case-insensitive values are "false" and "true" (the default).
Index is a language name. Indicates if the object files created by the compiler for a language need to be linked in the executable. Only authorized case-insensitive values are "false" and "true" (the default).
Value is the name of the target platform. Taken into account only in the main project.
Note that when the target is specified on the command line (usually with a switch –target=), the value of attribute reference 'Target is the one specified on the command line.
Index is a language name. Indicates the runtime directory that is to be used when using the compiler of the language. Taken into account only in the main project.
Note that when the runtime is specified for a language on the command line (usually with a switch –RTS), the value of attribute reference 'Runtime for this language is the one specified on the command line.
Value is the path name of the application that is to be used to build libraries. Usually the path name of "gprlib".
Indicates the level of support of libraries. Only authorized case-insensitive values are "static_only", "full" or "none" (the default).
Value is the name of the application to be used to create a static library (archive), followed by the options to be used.
Value is the list of options to be used when invoking the archive builder to add project files into an archive.
Value is the name of the archive indexer, followed by the required options.
Value is the extension of archives. When not declared, the extension is ".a".
Value is the name of the partial linker executable, followed by the required options.
Value is the prefix in the name of shared library files. When not declared, the prefix is "lib".
Value is the the extension of the name of shared library files. When not declared, the extension is ".so".
Indicates if symbolic links are supported on the platform. Only authorized case-insensitive values are "true" and "false" (the default).
Indicates if major and minor ids for shared library names are supported on the platform. Only authorized case-insensitive values are "true" and "false" (the default).
Indicates if auto-initialization of Stand-Alone Libraries is supported. Only authorized case-insensitive values are "true" and "false" (the default).
Value is the list of required switches when linking a shared library.
Value is the list of switches to specify a internal name for a shared library.
Value is the name of the option that needs to be used, concatenated with the path name of the library file, when linking a shared library.
Index is a language name. Value is the path name of the directory where the runtime libraries are located.
Index is a language name. Value is the path name of the directory where the sources of runtime libraries are located.
Index is a language name. Value is the list of switches to be used when binding code of the language, if there is no applicable attribute Switches.
Index is either a language name or a source file name. Value is the list of switches to be used when binding code. Index is either the source file name of the executable to be bound or the language name of the code to be bound.
Index is a language name. Value is the name of the application to be used when binding code of the language.
Index is a language name. Value is the list of the required switches to be used when binding code of the language.
Index is a language name. Value is a prefix to be used for the binder exchange file name for the language. Used to have different binder exchange file names when binding different languages.
Index is a language name. Value is the name of the environment variable that contains the path for the object directories.
Index is a language name. Value is the name of the environment variable. The value of the environment variable is the path name of a text file that contains the list of object directories.
Index is a language name. Value is the list of builder switches to be used when building an executable of the language, if there is no applicable attribute Switches.
Index is either a language name or a source file name. Value is the list of builder switches to be used when building an executable. Index is either the source file name of the executable to be built or its language name.
Index is either a language name or a source file name. Value is the list of compilation switches to be used when building an executable. Index is either the source file name of the executable to be built or its language name.
Index is an executable source file name. Value is the simple file name of the executable to be built.
Value is the extension of the file names of executable. When not specified, the extension is the default extension of executables on the platform.
Value is the file name of a configuration pragmas file that is specified to the Ada compiler when compiling any Ada source in the project tree.
Index is a language name. Value is the file name of a configuration file that is specified to the compiler when compiling any source of the language in the project tree.
Value is a list of switches to be used by the cleaning application.
Index is a language names. Value is the list of extensions for file names derived from object file names that need to be cleaned in the object directory of the project.
Index is a language names. Value is the list of extensions for file names derived from source file names that need to be cleaned in the object directory of the project.
Value is a list of file names expressed as regular expressions that are to be deleted by gprclean in the object directory of the project.
Value is list of file names expressed as regular expressions that are to be deleted by gprclean in the exec directory of the main project.
Index is a language name. Value is a list of switches to be used when invoking the compiler for the language for a source of the project, if there is no applicable attribute Switches.
Index is a source file name or a language name. Value is the list of switches to be used when invoking the compiler for the source or for its language.
Value is the file name of a configuration pragmas file that is specified to the Ada compiler when compiling any Ada source in the project.
Index is a language name. Value is the file name of a configuration file that is specified to the compiler when compiling any source of the language in the project.
Index is a language name. Value is the name of the executable for the compiler of the language.
Index is a language name. Indicates the kind of the language, either file based or unit based. Only authorized case-insensitive values are "unit_based" and "file_based" (the default).
Index is a language name. Indicates how the dependencies are handled for the language. Only authorized case-insensitive values are "makefile", "ali_file", "ali_closure" or "none" (the default).
Equivalent to attribute Leading_Required_Switches.
Index is a language name. Value is the list of the minimum switches to be used at the beginning of the command line when invoking the compiler for the language.
Index is a language name. Value is the list of the minimum switches to be used at the end of the command line when invoking the compiler for the language.
Index is a language name. Value is the list of switches to be used when compiling a source of the language when the project is a shared library project.
Index is a language name. Value is the kind of path syntax to be used when invoking the compiler for the language. Only authorized case-insensitive values are "canonical" and "host" (the default).
Index is a language name. Value is a list of switches to be used just before the path name of the source to compile when invoking the compiler for a source of the language.
Index is a language name. Value is the extension of the object files created by the compiler of the language. When not specified, the extension is the default one for the platform.
Index is a language name. Value is the list of switches to be used by the compiler of the language to specify the path name of the object file. When not specified, the switch used is "-o".
Index is a language name. Value is the list of switches to be used to compile a unit in a multi unit source of the language. The index of the unit in the source is concatenated with the last switches in the list.
Index is a language name. Value is the string to be used in the object file name before the index of the unit, when compiling a unit in a multi unit source of the language.
Index is a language name. Value is the list of switches to be used to specify a mapping file when invoking the compiler for a source of the language.
Index is a language name. Value is the suffix to be used in a mapping file to indicate that the source is a spec.
Index is a language name. Value is the suffix to be used in a mapping file to indicate that the source is a body.
Index is a language name. Value is the list of switches to specify to the compiler of the language a configuration file.
Index is a language name. Value is the template to be used to indicate a configuration specific to a body of the language in a configuration file.
Index is a language name. Value is the template to be used to indicate a configuration specific to the body a unit in a multi unit source of the language in a configuration file.
Index is a language name. Value is the template to be used to indicate a configuration for all bodies of the languages in a configuration file.
Index is a language name. Value is the template to be used to indicate a configuration specific to a spec of the language in a configuration file.
Index is a language name. Value is the template to be used to indicate a configuration specific to the spec a unit in a multi unit source of the language in a configuration file.
Index is a language name. Value is the template to be used to indicate a configuration for all specs of the languages in a configuration file.
Index is a language name. Indicates if there should be only one configuration file specified to the compiler of the language. Only authorized case-insensitive values are "true" and "false" (the default).
Index is a language name. Value is the list of switches to be used to specify to the compiler the dependency file when the dependency kind of the language is file based, and when Dependency_Driver is not specified for the language.
Index is a language name. Value is the name of the executable to be used to create the dependency file for a source of the language, followed by the required switches.
Index is a language name. Value is the list of switches to specify to the compiler of the language to indicate a directory to look for sources.
Index is a language name. Value is the name of an environment variable that contains the path of all the directories that the compiler of the language may search for sources.
Index is a language name. Value is the name of an environment variable the value of which is the path name of a text file that contains the directories that the compiler of the language may search for sources.
Index is a language name. Value is the list of switches to specify to the compiler of the language the name of a text file that contains the list of object directories. When this attribute is not declared, the text file is not created.
Index is a language name. Value is a list of switches to be used when invoking gnatxref for a source of the language, if there is no applicable attribute Switches.
Index is a source file name. Value is the list of switches to be used when invoking gnatxref for the source.
Index is a language name. Value is a list of switches to be used when invoking gnatfind for a source of the language, if there is no applicable attribute Switches.
Index is a source file name. Value is the list of switches to be used when invoking gnatfind for the source.
Value is a list of switches to be used when invoking gnatls.
Index is the name of an external tool that the GNAT Programming System (GPS) is supporting. Value is a list of switches to use when invoking that tool.
Value is a string that designates the remote host in a cross-compilation environment, to be used for remote compilation and debugging. This attribute should not be specified when running on the local machine.
Value is a string that specifies the name of IP address of the embedded target in a cross-compilation environment, on which the program should execute.
Value is the name of the protocol to use to communicate with the target in a cross-compilation environment, for example "wtx" or "vxworks".
Index is a language Name. Value is a string that denotes the command to be used to invoke the compiler. For historical reasons, the value of Compiler_Command ("Ada") is expected to be a reference to `gnatmake' or `cross-gnatmake'.
Value is a string that specifies the name of the debugger to be used, such as gdb, powerpc-wrs-vxworks-gdb or gdb-4.
Value is a string that specifies the name of the `gnatls' utility to be used to retrieve information about the predefined path; for example, "gnatls", "powerpc-wrs-vxworks-gnatls".
Value is a string used to specify the Version Control System (VCS) to be used for this project, for example "Subversion", "ClearCase". If the value is set to "Auto", the IDE will try to detect the actual VCS used on the list of supported ones.
Value is a string that specifies the command used by the VCS to check the validity of a file, either when the user explicitly asks for a check, or as a sanity check before doing the check-in.
Value is a string that specifies the command used by the VCS to check the validity of a log file.
Value is the directory used to generate the documentation of source code.
An array attribute to declare a set of files not part of the sources to be installed. The array discriminant is the directory where the file is to be installed. If a relative directory then Prefix (see below) is prepended. Note also that if the same file name occurs multiple time in the attribute list, the last one will be the one installed.
Value is the install destination directory.
Value is the sources directory or subdirectory of Prefix.
Value is the executables directory or subdirectory of Prefix.
Value is library directory or subdirectory of Prefix.
Value is the project directory or subdirectory of Prefix.
Indicates that the project is to be installed or not. Case-insensitive value "false" means that the project is not to be installed, all other values mean that the project is to be installed.
Value is the installation mode, it is either `dev' (default) or `usage'.
Specify the name to use for recording the installation. The default is the project name without the extension.
Value is a list of switches that are required when invoking the linker to link an executable.
Index is a language name. Value is a list of switches for the linker when linking an executable for a main source of the language, when there is no applicable Switches.
Index is a source file name or a language name. Value is the list of switches to be used at the beginning of the command line when invoking the linker to build an executable for the source or for its language.
Index is a source file name or a language name. Value is the list of switches to be used when invoking the linker to build an executable for the source or for its language.
Index is a source file name or a language name. Value is the list of switches to be used at the end of the command line when invoking the linker to build an executable for the source or for its language. These switches may override the Required_Switches.
Value is a list of switches/options that are to be added when linking an executable from a project importing the current project directly or indirectly. Linker_Options are not used when linking an executable from the current project.
Value is the switch to specify the map file name that the linker needs to create.
Value is the name of the linker executable.
Value is the maximum number of character in the command line when invoking the linker to link an executable.
Indicates the kind of response file to create when the length of the linking command line is too large. Only authorized case-insensitive values are "none", "gnu", "object_list", "gcc_gnu", "gcc_option_list" and "gcc_object_list".
Value is the list of switches to specify a response file to the linker.
Equivalent to attribute Spec_Suffix.
Index is a language name. Value is the extension of file names for specs of the language.
Equivalent to attribute Body_Suffix.
Index is a language name. Value is the extension of file names for bodies of the language.
Value is the extension of file names for subunits of Ada.
Indicates the casing of sources of the Ada language. Only authorized case-insensitive values are "lowercase", "uppercase" and "mixedcase".
Value is the string that replace the dot of unit names in the source file names of the Ada language.
Equivalent to attribute Spec.
Index is a unit name. Value is the file name of the spec of the unit.
Equivalent to attribute Body.
Index is a unit name. Value is the file name of the body of the unit.
Index is a language name. Value is a list of specs for the language that do not necessarily follow the naming scheme for the language and that may or may not be found in the source directories of the project.
Index is a language name. Value is a list of bodies for the language that do not necessarily follow the naming scheme for the language and that may or may not be found in the source directories of the project.
If this attribute is defined it sets the patterns to synchronized from the master to the slaves. It is exclusive with Excluded_Patterns, that is it is an error to define both.
If this attribute is defined it sets the patterns of compilation artifacts to synchronized from the slaves to the build master. This attribute replace the default hard-coded patterns.
Set of patterns to ignore when synchronizing sources from the build master to the slaves. A set of predefined patterns are supported (e.g. *.o, *.ali, *.exe, etc.), this attributes make it possible to add some more patterns.
Value is the root directory used by the slave machines.
Value is the list of switches to be used when invoking gnatstack.
Index is a language name. Value is a list of switches to be used when invoking gnatsync for a source of the language, if there is no applicable attribute Switches.
Index is a source file name. Value is the list of switches to be used when invoking gnatsync for the source.
This section describes how project files can be used in conjunction with a number of GNAT tools.
This section covers several topics related to `gnatmake' and project files: defining switches for `gnatmake' and for the tools that it invokes; specifying configuration pragmas; the use of the Main attribute; building and rebuilding library project files.
The following switches are used by GNAT tools that support project files:
-P`project'There must be only one `-P' switch on the command line.
Since the Project Manager parses the project file only after all the switches on the command line are checked, the order of the switches `-P', `-vP*x*' or `-X' is not significant.
-X`name'=`value'If name or value includes a space, then name=value should be put between quotes.
-XOS=NT
-X"user=John Doe"
Several `-X' switches can be used simultaneously. If several `-X' switches specify the same name, only the last one is used.
An external variable specified with a `-X' switch takes precedence over the value of the same name in the environment.
-vP`x'`-vP0' means Default; `-vP1' means Medium; `-vP2' means High.
The default is Default: no output for syntactically correct project files. If several `-vP*x*' switches are present, only the last one is used.
-aP`dir'-eL--subdirs=`subdir'For each of the packages Builder, Compiler, Binder, and Linker, you can specify a Default_Switches attribute, a Switches attribute, or both; as their names imply, these switch-related attributes affect the switches that are used for each of these GNAT components when `gnatmake' is invoked. As will be explained below, these component-specific switches precede the switches provided on the `gnatmake' command line.
The Default_Switches attribute is an attribute indexed by language name (case insensitive) whose value is a string list. For example:
package Compiler is
for Default_Switches ("Ada")
use ("-gnaty",
"-v");
end Compiler;
The Switches attribute is indexed on a file name (which may or may not be case sensitive, depending on the operating system) whose value is a string list. For example:
package Builder is
for Switches ("main1.adb")
use ("-O2");
for Switches ("main2.adb")
use ("-g");
end Builder;
For the Builder package, the file names must designate source files
for main subprograms. For the Binder and Linker packages, the
file names must designate ALI or source files for main subprograms.
In each case just the file name without an explicit extension is acceptable.
For each tool used in a program build (`gnatmake', the compiler, the binder, and the linker), the corresponding package @dfn{contributes} a set of switches for each file on which the tool is invoked, based on the switch-related attributes defined in the package. In particular, the switches that each of these packages contributes for a given file f comprise:
If neither of these attributes is defined in the package, then the package does not contribute any switches for the given file.
When `gnatmake' is invoked on a file, the switches comprise two sets, in the following order: those contributed for the file by the Builder package; and the switches passed on the command line.
When `gnatmake' invokes a tool (compiler, binder, linker) on a file, the switches passed to the tool comprise three sets, in the following order:
The term `applicable switches' reflects the fact that `gnatmake' switches may or may not be passed to individual tools, depending on the individual switch.
`gnatmake' may invoke the compiler on source files from different projects. The Project Manager will use the appropriate project file to determine the Compiler package for each source file being compiled. Likewise for the Binder and Linker packages.
As an example, consider the following package in a project file:
project Proj1 is
package Compiler is
for Default_Switches ("Ada")
use ("-g");
for Switches ("a.adb")
use ("-O1");
for Switches ("b.adb")
use ("-O2",
"-gnaty");
end Compiler;
end Proj1;
If `gnatmake' is invoked with this project file, and it needs to
compile, say, the files a.adb, b.adb, and c.adb, then
a.adb will be compiled with the switch `-O1',
b.adb with switches `-O2' and `-gnaty',
and c.adb with `-g'.
The following example illustrates the ordering of the switches contributed by different packages:
project Proj2 is
package Builder is
for Switches ("main.adb")
use ("-g",
"-O1",
"-f");
end Builder;
package Compiler is
for Switches ("main.adb")
use ("-O2");
end Compiler;
end Proj2;
If you issue the command:
$ gnatmake -Pproj2 -O0 main
then the compiler will be invoked on main.adb with the following
sequence of switches
-g -O1 -O2 -O0
with the last `-O' switch having precedence over the earlier ones; several other switches (such as `-c') are added implicitly.
The switches `-g' and `-O1' are contributed by package Builder, `-O2' is contributed by the package Compiler and `-O0' comes from the command line.
The `-g' switch will also be passed in the invocation of `Gnatlink.'
A final example illustrates switch contributions from packages in different project files:
project Proj3 is for Source_Files use ("pack.ads", "pack.adb"); package Compiler is for Default_Switches ("Ada") use ("-gnata"); end Compiler; end Proj3; with "Proj3"; project Proj4 is for Source_Files use ("foo_main.adb", "bar_main.adb"); package Builder is for Switches ("foo_main.adb") use ("-s", "-g"); end Builder; end Proj4;-- Ada source file: with Pack; procedure Foo_Main is ... end Foo_Main;
If the command is
$ gnatmake -PProj4 foo_main.adb -cargs -gnato
then the switches passed to the compiler for foo_main.adb are
`-g' (contributed by the package Proj4.Builder) and
`-gnato' (passed on the command line).
When the imported package Pack is compiled, the switches used
are `-g' from Proj4.Builder,
`-gnata' (contributed from package Proj3.Compiler,
and `-gnato' from the command line.
When using `gnatmake' with project files, some switches or arguments may be expressed as relative paths. As the working directory where compilation occurs may change, these relative paths are converted to absolute paths. For the switches found in a project file, the relative paths are relative to the project file directory, for the switches on the command line, they are relative to the directory where `gnatmake' is invoked. The switches for which this occurs are: -I, -A, -L, -aO, -aL, -aI, as well as all arguments that are not switches (arguments to switch -o, object files specified in package Linker or after -largs on the command line). The exception to this rule is the switch –RTS= for which a relative path argument is never converted.
When using `gnatmake' with project files, if there exists a file
gnat.adc that contains configuration pragmas, this file will be
ignored.
Configuration pragmas can be defined by means of the following attributes in project files: Global_Configuration_Pragmas in package Builder and Local_Configuration_Pragmas in package Compiler.
Both these attributes are single string attributes. Their values is the path name of a file containing configuration pragmas. If a path name is relative, then it is relative to the project directory of the project file where the attribute is defined.
When compiling a source, the configuration pragmas used are, in order, those listed in the file designated by attribute Global_Configuration_Pragmas in package Builder of the main project file, if it is specified, and those listed in the file designated by attribute Local_Configuration_Pragmas in package Compiler of the project file of the source, if it exists.
When using a project file, you can invoke `gnatmake' with one or several main subprograms, by specifying their source files on the command line.
$ gnatmake -Pprj main1.adb main2.adb main3.adb
Each of these needs to be a source file of the same project, except when the switch -u is used.
When -u is not used, all the mains need to be sources of the same project, one of the project in the tree rooted at the project specified on the command line. The package Builder of this common project, the "main project" is the one that is considered by `gnatmake'.
When -u is used, the specified source files may be in projects imported directly or indirectly by the project specified on the command line. Note that if such a source file is not part of the project specified on the command line, the switches found in package Builder of the project specified on the command line, if any, that are transmitted to the compiler will still be used, not those found in the project file of the source file.
When using a project file, you can also invoke `gnatmake' without explicitly specifying any main, and the effect depends on whether you have defined the Main attribute. This attribute has a string list value, where each element in the list is the name of a source file (the file extension is optional) that contains a unit that can be a main subprogram.
If the Main attribute is defined in a project file as a non-empty string list and the switch `-u' is not used on the command line, then invoking `gnatmake' with this project file but without any main on the command line is equivalent to invoking `gnatmake' with all the file names in the Main attribute on the command line.
Example:
project Prj is
for Main use ("main1.adb", "main2.adb", "main3.adb");
end Prj;
With this project file, "gnatmake -Pprj" is equivalent to "gnatmake -Pprj main1.adb main2.adb main3.adb".
When the project attribute Main is not specified, or is specified as an empty string list, or when the switch `-u' is used on the command line, then invoking `gnatmake' with no main on the command line will result in all immediate sources of the project file being checked, and potentially recompiled. Depending on the presence of the switch `-u', sources from other project files on which the immediate sources of the main project file depend are also checked and potentially recompiled. In other words, the `-u' switch is applied to all of the immediate sources of the main project file.
When no main is specified on the command line and attribute Main exists and includes several mains, or when several mains are specified on the command line, the default switches in package Builder will be used for all mains, even if there are specific switches specified for one or several mains.
But the switches from package Binder or Linker will be the specific switches for each main, if they are specified.
When `gnatmake' is invoked with a main project file that is a library project file, it is not allowed to specify one or more mains on the command line.
When a library project file is specified, switches -b and -l have special meanings.
A number of GNAT tools beyond `gnatmake' can benefit from project files:
However, none of these tools can be invoked directly with a project file switch (`-P'). They must be invoked through the `gnat' driver.
The `gnat' driver is a wrapper that accepts a number of commands and calls the corresponding tool. It was designed initially for VMS platforms (to convert VMS qualifiers to Unix-style switches), but it is now available on all GNAT platforms.
On non-VMS platforms, the `gnat' driver accepts the following commands (case insensitive):
Note that the command `gnatmake -c -f -u' is used to invoke the compiler.
On non-VMS platforms, between `gnat' and the command, two special switches may be used:
The command may be followed by switches and arguments for the invoked tool.
$ gnat bind -C main.ali
$ gnat ls -a main
$ gnat chop foo.txt
Switches may also be put in text files, one switch per line, and the text files may be specified with their path name preceded by '@'.
$ gnat bind @args.txt main.ali
In addition, for the following commands the project file related switches (`-P', `-X' and `-vPx') may be used in addition to the switches of the invoking tool:
For each of the following commands, there is optionally a corresponding package in the main project.
Package Gnatls has a unique attribute Switches, a simple variable with a string list value. It contains switches for the invocation of gnatls.
project Proj1 is
package gnatls is
for Switches
use ("-a",
"-v");
end gnatls;
end Proj1;
All other packages have two attribute Switches and Default_Switches.
Switches is an indexed attribute, indexed by the source file name, that has a string list value: the switches to be used when the tool corresponding to the package is invoked for the specific source file.
Default_Switches is an attribute, indexed by the programming language that has a string list value. Default_Switches ("Ada") contains the switches for the invocation of the tool corresponding to the package, except if a specific Switches attribute is specified for the source file.
project Proj is
for Source_Dirs use ("");
package gnatls is
for Switches use
("-a",
"-v");
end gnatls;
package Compiler is
for Default_Switches ("Ada")
use ("-gnatv",
"-gnatwa");
end Binder;
package Binder is
for Default_Switches ("Ada")
use ("-C",
"-e");
end Binder;
package Linker is
for Default_Switches ("Ada")
use ("-C");
for Switches ("main.adb")
use ("-C",
"-v",
"-v");
end Linker;
package Finder is
for Default_Switches ("Ada")
use ("-a",
"-f");
end Finder;
package Cross_Reference is
for Default_Switches ("Ada")
use ("-a",
"-f",
"-d",
"-u");
end Cross_Reference;
end Proj;
With the above project file, commands such as
$ gnat comp -Pproj main
$ gnat ls -Pproj main
$ gnat xref -Pproj main
$ gnat bind -Pproj main.ali
$ gnat link -Pproj main.ali
will set up the environment properly and invoke the tool with the switches found in the package corresponding to the tool: Default_Switches ("Ada") for all tools, except Switches ("main.adb") for gnatlink.
This chapter describes a number of utility programs:
Other GNAT utilities are described elsewhere in this manual:
gnatclean is a tool that allows the deletion of files produced by the compiler, binder and linker, including ALI files, object files, tree files, expanded source files, library files, interface copy source files, binder generated files and executable files.
The gnatclean command has the form:
$ gnatclean switches `names`
where names is a list of source file names. Suffixes .ads and
adb may be omitted. If a project file is specified using switch
-P, then names may be completely omitted.
In normal mode, gnatclean delete the files produced by the compiler and, if switch -c is not specified, by the binder and the linker. In informative-only mode, specified by switch -n, the list of files that would have been deleted in normal mode is listed, but no file is actually deleted.
gnatclean recognizes the following switches:
--version--help--subdirs=`subdir'--unchecked-shared-lib-imports-c-D `dir'-F-h-n-P`project'-q-r-v-vP`x'-X`name'=`value'-aO`dir'-I`dir'-aO`dir'.
-I-gnatls is a tool that outputs information about compiled units. It gives the relationship between objects, unit names and source files. It can also be used to check the source dependencies of a unit as well as various characteristics.
Note: to invoke gnatls with a project file, use the gnat driver (see The GNAT Driver and Project Files).
The gnatls command has the form
$ gnatls switches `object_or_ali_file`
The main argument is the list of object or ali files
(see The Ada Library Information Files)
for which information is requested.
In normal mode, without additional option, gnatls produces a four-column listing. Each line represents information for a specific object. The first column gives the full path of the object, the second column gives the name of the principal unit in this object, the third column gives the status of the source and the fourth column gives the full path of the source representing this unit. Here is a simple example of use:
$ gnatls *.o
./demo1.o demo1 DIF demo1.adb
./demo2.o demo2 OK demo2.adb
./hello.o h1 OK hello.adb
./instr-child.o instr.child MOK instr-child.adb
./instr.o instr OK instr.adb
./tef.o tef DIF tef.adb
./text_io_example.o text_io_example OK text_io_example.adb
./tgef.o tgef DIF tgef.adb
The first line can be interpreted as follows: the main unit which is
contained in
object file demo1.o is demo1, whose main source is in
demo1.adb. Furthermore, the version of the source used for the
compilation of demo1 has been modified (DIF). Each source file has a status
qualifier which can be:
gnatls recognizes the following switches:
--version*--help-a-d-h-o-s-u-files=`file'-aO`dir', -aI`dir', -I`dir', -I-, -nostdinc-aP`dir'--RTS=`rts-path'`-vExample of using the verbose switch. Note how the source and object paths are affected by the -I switch.
$ gnatls -v -I.. demo1.o
GNATLS 5.03w (20041123-34)
Copyright 1997-2004 Free Software Foundation, Inc.
Source Search Path:
<Current_Directory>
../
/home/comar/local/adainclude/
Object Search Path:
<Current_Directory>
../
/home/comar/local/lib/gcc-lib/x86-linux/3.4.3/adalib/
Project Search Path:
<Current_Directory>
/home/comar/local/lib/gnat/
./demo1.o
Unit =>
Name => demo1
Kind => subprogram body
Flags => No_Elab_Code
Source => demo1.adb modified
The following is an example of use of the dependency list. Note the use of the -s switch which gives a straight list of source files. This can be useful for building specialized scripts.
$ gnatls -d demo2.o
./demo2.o demo2 OK demo2.adb
OK gen_list.ads
OK gen_list.adb
OK instr.ads
OK instr-child.ads
$ gnatls -d -s -a demo1.o
demo1.adb
/home/comar/local/adainclude/ada.ads
/home/comar/local/adainclude/a-finali.ads
/home/comar/local/adainclude/a-filico.ads
/home/comar/local/adainclude/a-stream.ads
/home/comar/local/adainclude/a-tags.ads
gen_list.ads
gen_list.adb
/home/comar/local/adainclude/gnat.ads
/home/comar/local/adainclude/g-io.ads
instr.ads
/home/comar/local/adainclude/system.ads
/home/comar/local/adainclude/s-exctab.ads
/home/comar/local/adainclude/s-finimp.ads
/home/comar/local/adainclude/s-finroo.ads
/home/comar/local/adainclude/s-secsta.ads
/home/comar/local/adainclude/s-stalib.ads
/home/comar/local/adainclude/s-stoele.ads
/home/comar/local/adainclude/s-stratt.ads
/home/comar/local/adainclude/s-tasoli.ads
/home/comar/local/adainclude/s-unstyp.ads
/home/comar/local/adainclude/unchconv.ads
The compiler generates cross-referencing information (unless
you set the -gnatx switch), which are saved in the .ali files.
This information indicates where in the source each entity is declared and
referenced. Note that entities in package Standard are not included, but
entities in all other predefined units are included in the output.
Before using any of these two tools, you need to compile successfully your application, so that GNAT gets a chance to generate the cross-referencing information.
The two tools gnatxref and gnatfind take advantage of this information to provide the user with the capability to easily locate the declaration and references to an entity. These tools are quite similar, the difference being that gnatfind is intended for locating definitions and/or references to a specified entity or entities, whereas gnatxref is oriented to generating a full report of all cross-references.
To use these tools, you must not compile your application using the `-gnatx' switch on the `gnatmake' command line (see Building with gnatmake). Otherwise, cross-referencing information will not be generated.
Note: to invoke gnatxref or gnatfind with a project file, use the gnat driver (see The GNAT Driver and Project Files).
The command invocation for gnatxref is:
$ gnatxref [`switches`] `sourcefile1` [`sourcefile2` ...]
where
These file names are considered to be regular expressions, so for instance
specifying source*.adb is the same as giving every file in the current
directory whose name starts with source and whose extension is
adb.
You shouldn't specify any directory name, just base names. `gnatxref' and `gnatfind' will be able to locate these files by themselves using the source path. If you specify directories, no result is produced.
The following switches are available for `gnatxref':
-version-helpaaI`DIR'aO`DIR'nostdincnostdlib-ext=`extension'-RTS=`rts-path'dfgI`DIR'-aODIR -aIDIR.
p`FILE'.gpr
project files, you should use gnatxref through the GNAT driver
(`gnat xref -Pproject').
By default, gnatxref and gnatfind will try to locate a project file in the current directory.
If a project file is either specified or found by the tools, then the content
of the source directory and object directory lines are added as if they
had been specified respectively by -aI
and -aO.
uvtags file that can be used by vi. For examples how to use this
feature, see Examples of gnatxref Usage. The tags file is output
to the standard output, thus you will have to redirect it to a file.
All these switches may be in any order on the command line, and may even
appear after the file names. They need not be separated by spaces, thus
you can say gnatxref -ag instead of gnatxref -a -g.
The command invocation for gnatfind is:
$ gnatfind [`switches`] `pattern`[:`sourcefile`[:`line`[:`column`]]]
[`file1` `file2` ...]
with the following iterpretation of the command arguments:
Omitting the pattern is equivalent to specifying *, which
will match any entity. Note that if you do not provide a pattern, you
have to provide both a sourcefile and a line.
Entity names are given in Latin-1, with uppercase/lowercase equivalence
for matching purposes. At the current time there is no support for
8-bit codes other than Latin-1, or for wide characters in identifiers.
sourcefile, at line line
and column column. See Examples of gnatfind Usage
for syntax examples.
These file names are considered to be regular expressions, so for instance
specifying source*.adb is the same as giving every file in the current
directory whose name starts with source and whose extension is
adb.
The location of the spec of the entity will always be displayed, even if it
isn't in one of file1, file2, ... The
occurrences of the entity in the separate units of the ones given on the
command line will also be displayed.
Note that if you specify at least one file in this part, gnatfind may sometimes not be able to find the body of the subprograms.
At least one of 'sourcefile' or 'pattern' has to be present on the command line.
The following switches are available:
--version-helpaaI`DIR'aO`DIR'nostdincnostdlib-ext=`extension'-RTS=`rts-path'defgI`DIR'-aODIR -aIDIR.
p`FILE'If a project file is either specified or found by the tools, then the content
of the source directory and object directory lines are added as if they
had been specified respectively by -aI and
-aO.
rstAll these switches may be in any order on the command line, and may even
appear after the file names. They need not be separated by spaces, thus
you can say gnatxref -ag instead of
gnatxref -a -g.
As stated previously, gnatfind will search in every directory in the search path. You can force it to look only in the current directory if you specify * at the end of the command line.
Project files allow a programmer to specify how to compile its application, where to find sources, etc. These files are used primarily by GPS, but they can also be used by the two tools gnatxref and gnatfind.
A project file name must end with .gpr. If a single one is
present in the current directory, then gnatxref and gnatfind will
extract the information from it. If multiple project files are found, none of
them is read, and you have to use the -p switch to specify the one
you want to use.
The following lines can be included, even though most of them have default
values which can be used in most cases.
The lines can be entered in any order in the file.
Except for src_dir and obj_dir, you can only have one instance of
each line. If you have multiple instances, only the last one is taken into
account.
$`bind_opt' notation. This is intended to store the default
switches given to `gnatbind'.
$`link_opt' notation. This is intended to store the default
switches given to `gnatlink'.
`${main' notation.
`gnatxref' and `gnatfind' only take into account the src_dir and obj_dir lines, and ignore the others.
As specified in the section about `gnatfind', the pattern can be a regular expression. Two kinds of regular expressions are recognized:
Here is a more formal grammar:
regexp ::= term
term ::= elmt -- matches elmt
term ::= elmt elmt -- concatenation (elmt then elmt)
term ::= * -- any string of 0 or more characters
term ::= ? -- matches any character
term ::= [char {char}] -- matches any character listed
term ::= [char - char] -- matches any character in range
grep.
The following is the form of a regular expression, expressed in same BNF style as is found in the Ada Reference Manual:
regexp ::= term {| term} -- alternation (term or term ...)
term ::= item {item} -- concatenation (item then item)
item ::= elmt -- match elmt
item ::= elmt * -- zero or more elmt's
item ::= elmt + -- one or more elmt's
item ::= elmt ? -- matches elmt or nothing
elmt ::= nschar -- matches given character
elmt ::= [nschar {nschar}] -- matches any character listed
elmt ::= [^ nschar {nschar}] -- matches any character not listed
elmt ::= [char - char] -- matches chars in given range
elmt ::= \\ char -- matches given character
elmt ::= . -- matches any single character
elmt ::= ( regexp ) -- parens used for grouping
char ::= any character, including special characters
nschar ::= any character except ()[].*+?^
Here are a few examples:
abcde|fghi- will match any of the two strings
abcdeandfghi,abc*d- will match any string like
abd,abcd,abccd,abcccd, and so on,[a-z]+- will match any string which has only lowercase characters in it (and at least one character.
For the following examples, we will consider the following units:
main.ads:
1: with Bar;
2: package Main is
3: procedure Foo (B : in Integer);
4: C : Integer;
5: private
6: D : Integer;
7: end Main;
main.adb:
1: package body Main is
2: procedure Foo (B : in Integer) is
3: begin
4: C := B;
5: D := B;
6: Bar.Print (B);
7: Bar.Print (C);
8: end Foo;
9: end Main;
bar.ads:
1: package Bar is
2: procedure Print (B : Integer);
3: end bar;
The first thing to do is to recompile your application (for instance, in
that case just by doing a gnatmake main, so that GNAT generates
the cross-referencing information.
You can then issue any of the following commands:
gnatxref main.adbgnatxref generates cross-reference information for main.adb and every unit 'with'ed by main.adb.The output would be:
B Type: Integer Decl: bar.ads 2:22 B Type: Integer Decl: main.ads 3:20 Body: main.adb 2:20 Ref: main.adb 4:13 5:13 6:19 Bar Type: Unit Decl: bar.ads 1:9 Ref: main.adb 6:8 7:8 main.ads 1:6 C Type: Integer Decl: main.ads 4:5 Modi: main.adb 4:8 Ref: main.adb 7:19 D Type: Integer Decl: main.ads 6:5 Modi: main.adb 5:8 Foo Type: Unit Decl: main.ads 3:15 Body: main.adb 2:15 Main Type: Unit Decl: main.ads 2:9 Body: main.adb 1:14 Print Type: Unit Decl: bar.ads 2:15 Ref: main.adb 6:12 7:12This shows that the entity Main is declared in main.ads, line 2, column 9, its body is in main.adb, line 1, column 14 and is not referenced any where.
The entity Print is declared in bar.ads, line 2, column 15 and it is referenced in main.adb, line 6 column 12 and line 7 column 12.
gnatxref package1.adb package2.adsgnatxref will generates cross-reference information for package1.adb, package2.ads and any other package 'with'ed by any of these.
gnatxref can generate a tags file output, which can be used directly from `vi'. Note that the standard version of `vi' will not work properly with overloaded symbols. Consider using another free implementation of `vi', such as `vim'.
$ gnatxref -v gnatfind.adb > tags
The following command will generate the tags file for gnatfind itself (if the sources are in the search path!):
$ gnatxref -v gnatfind.adb > tags
From `vi', you can then use the command :tag `entity'
(replacing entity by whatever you are looking for), and vi will
display a new file with the corresponding declaration of entity.
gnatfind -f xyz:main.adb
Find declarations for all entities xyz referenced at least once in
main.adb. The references are search in every library file in the search
path.
The directories will be printed as well (as the -f
switch is set)
The output will look like:
directory/main.ads:106:14: xyz <= declaration
directory/main.adb:24:10: xyz <= body
directory/foo.ads:45:23: xyz <= declaration
I.e., one of the entities xyz found in main.adb is declared at line 12 of main.ads (and its body is in main.adb), and another one is declared at line 45 of foo.ads
gnatfind -fs xyz:main.adb
This is the same command as the previous one, but gnatfind will
display the content of the Ada source file lines.
The output will look like:
directory/main.ads:106:14: xyz <= declaration
procedure xyz;
directory/main.adb:24:10: xyz <= body
procedure xyz is
directory/foo.ads:45:23: xyz <= declaration
xyz : Integer;
This can make it easier to find exactly the location your are looking for.
gnatfind -r "*x*":main.ads:123 foo.adb
Find references to all entities containing an x that are
referenced on line 123 of main.ads.
The references will be searched only in main.ads and foo.adb.
gnatfind main.ads:123
Find declarations and bodies for all entities that are referenced on
line 123 of main.ads.
This is the same as gnatfind "*":main.adb:123`
gnatfind mydir/main.adb:123:45
Find the declaration for the entity referenced at column 45 in
line 123 of file main.adb in directory mydir. Note that it
is usual to omit the identifier name when the column is given,
since the column position identifies a unique reference.
The column has to be the beginning of the identifier, and should not point to any character in the middle of the identifier.
`gnathtml' is a Perl script that allows Ada source files to be browsed using standard Web browsers. For installation information, see Installing gnathtml.
Ada reserved keywords are highlighted in a bold font and Ada comments in a blue font. Unless your program was compiled with the gcc `-gnatx' switch to suppress the generation of cross-referencing information, user defined variables and types will appear in a different color; you will be able to click on any identifier and go to its declaration.
The command line is as follows:
$ perl gnathtml.pl [`switches`] `ada-files`
You can specify as many Ada files as you want. gnathtml will generate
an html file for every ada file, and a global file called index.htm.
This file is an index of every identifier defined in the files.
The following switches are available:
83cc `color'dDext `extension'htm.
fl `number'I `dir'.ALI files) and
source files. You can provide several -I switches on the command line,
and the directories will be parsed in the order of the command line.
o `dir'html/.
p `file'.gpr files
to give the directories where Emacs can find sources and object files.
Using this switch, you can tell gnathtml to use these files. This allows you to get an html version of your application, even if it is spread over multiple directories.
sc `color't `file'Perl needs to be installed on your machine to run this script. Perl is freely available for almost every architecture and operating system via the Internet.
On Unix systems, you may want to modify the first line of the script gnathtml, to explicitly specify where Perl is located. The syntax of this line is:
#!full_path_name_to_perl
Alternatively, you may run the script using the following command line:
$ perl gnathtml.pl [`switches`] `files`
This chapter covers several topics:
This section discusses how to debug Ada programs.
An incorrect Ada program may be handled in three ways by the GNAT compiler:
GDB is a general purpose, platform-independent debugger that can be used to debug mixed-language programs compiled with `gcc', and in particular is capable of debugging Ada programs compiled with GNAT. The latest versions of GDB are Ada-aware and can handle complex Ada data structures.
See Debugging with GDB, for full details on the usage of GDB, including a section on its usage on programs. This manual should be consulted for full details. The section that follows is a brief introduction to the philosophy and use of GDB.
When GNAT programs are compiled, the compiler optionally writes debugging information into the generated object file, including information on line numbers, and on declared types and variables. This information is separate from the generated code. It makes the object files considerably larger, but it does not add to the size of the actual executable that will be loaded into memory, and has no impact on run-time performance. The generation of debug information is triggered by the use of the -g switch in the `gcc' or `gnatmake' command used to carry out the compilations. It is important to emphasize that the use of these options does not change the generated code.
The debugging information is written in standard system formats that are used by many tools, including debuggers and profilers. The format of the information is typically designed to describe C types and semantics, but GNAT implements a translation scheme which allows full details about Ada types and variables to be encoded into these standard C formats. Details of this encoding scheme may be found in the file exp_dbug.ads in the GNAT source distribution. However, the details of this encoding are, in general, of no interest to a user, since GDB automatically performs the necessary decoding.
When a program is bound and linked, the debugging information is collected from the object files, and stored in the executable image of the program. Again, this process significantly increases the size of the generated executable file, but it does not increase the size of the executable program itself. Furthermore, if this program is run in the normal manner, it runs exactly as if the debug information were not present, and takes no more actual memory.
However, if the program is run under control of GDB, the debugger is activated. The image of the program is loaded, at which point it is ready to run. If a run command is given, then the program will run exactly as it would have if GDB were not present. This is a crucial part of the GDB design philosophy. GDB is entirely non-intrusive until a breakpoint is encountered. If no breakpoint is ever hit, the program will run exactly as it would if no debugger were present. When a breakpoint is hit, GDB accesses the debugging information and can respond to user commands to inspect variables, and more generally to report on the state of execution.
This section describes how to initiate the debugger.
The debugger can be launched from a GPS menu or directly from the command line. The description below covers the latter use. All the commands shown can be used in the GPS debug console window, but there are usually more GUI-based ways to achieve the same effect.
The command to run GDB is
$ gdb program
where program is the name of the executable file. This activates the debugger and results in a prompt for debugger commands. The simplest command is simply run, which causes the program to run exactly as if the debugger were not present. The following section describes some of the additional commands that can be given to GDB.
GDB contains a large repertoire of commands. See Debugging with GDB for extensive documentation on the use of these commands, together with examples of their use. Furthermore, the command `help' invoked from within GDB activates a simple help facility which summarizes the available commands and their options. In this section we summarize a few of the most commonly used commands to give an idea of what GDB is about. You should create a simple program with debugging information and experiment with the use of these GDB commands on the program as you read through the following section.
The above list is a very short introduction to the commands that GDB provides. Important additional capabilities, including conditional breakpoints, the ability to execute command sequences on a breakpoint, the ability to debug at the machine instruction level and many other features are described in detail in Debugging with GDB. Note that most commands can be abbreviated (for example, c for continue, bt for backtrace).
GDB supports a fairly large subset of Ada expression syntax, with some extensions. The philosophy behind the design of this subset is
- That GDB should provide basic literals and access to operations for arithmetic, dereferencing, field selection, indexing, and subprogram calls, leaving more sophisticated computations to subprograms written into the program (which therefore may be called from GDB).
- That type safety and strict adherence to Ada language restrictions are not particularly relevant in a debugging context.
- That brevity is important to the GDB user.
Thus, for brevity, the debugger acts as if there were implicit with and use clauses in effect for all user-written packages, thus making it unnecessary to fully qualify most names with their packages, regardless of context. Where this causes ambiguity, GDB asks the user's intent.
For details on the supported Ada syntax, see Debugging with GDB.
An important capability of GDB is the ability to call user-defined subprograms while debugging. This is achieved simply by entering a subprogram call statement in the form:
call subprogram-name (parameters)
The keyword call can be omitted in the normal case where the subprogram-name does not coincide with any of the predefined GDB commands.
The effect is to invoke the given subprogram, passing it the list of parameters that is supplied. The parameters can be expressions and can include variables from the program being debugged. The subprogram must be defined at the library level within your program, and GDB will call the subprogram within the environment of your program execution (which means that the subprogram is free to access or even modify variables within your program).
The most important use of this facility is in allowing the inclusion of debugging routines that are tailored to particular data structures in your program. Such debugging routines can be written to provide a suitably high-level description of an abstract type, rather than a low-level dump of its physical layout. After all, the standard GDB print command only knows the physical layout of your types, not their abstract meaning. Debugging routines can provide information at the desired semantic level and are thus enormously useful.
For example, when debugging GNAT itself, it is crucial to have access to the contents of the tree nodes used to represent the program internally. But tree nodes are represented simply by an integer value (which in turn is an index into a table of nodes). Using the print command on a tree node would simply print this integer value, which is not very useful. But the PN routine (defined in file treepr.adb in the GNAT sources) takes a tree node as input, and displays a useful high level representation of the tree node, which includes the syntactic category of the node, its position in the source, the integers that denote descendant nodes and parent node, as well as varied semantic information. To study this example in more detail, you might want to look at the body of the PN procedure in the stated file.
Another useful application of this capability is to deal with situations of complex data which are not handled suitably by GDB. For example, if you specify Convention Fortran for a multi-dimensional array, GDB does not know that the ordering of array elements has been switched and will not properly address the array elements. In such a case, instead of trying to print the elements directly from GDB, you can write a callable procedure that prints the elements in the desired format.
When you use the next command in a function, the current source location will advance to the next statement as usual. A special case arises in the case of a return statement.
Part of the code for a return statement is the 'epilogue' of the function. This is the code that returns to the caller. There is only one copy of this epilogue code, and it is typically associated with the last return statement in the function if there is more than one return. In some implementations, this epilogue is associated with the first statement of the function.
The result is that if you use the next command from a return statement that is not the last return statement of the function you may see a strange apparent jump to the last return statement or to the start of the function. You should simply ignore this odd jump. The value returned is always that from the first return statement that was stepped through.
You can set catchpoints that stop the program execution when your program raises selected exceptions.
GDB allows the following task-related commands:
(gdb) info tasks
ID TID P-ID Thread Pri State Name
1 8088000 0 807e000 15 Child Activation Wait main_task
2 80a4000 1 80ae000 15 Accept/Select Wait b
3 809a800 1 80a4800 15 Child Activation Wait a
* 4 80ae800 3 80b8000 15 Running c
In this listing, the asterisk before the first task indicates it to be the currently running task. The first column lists the task ID that is used to refer to tasks in the following commands.
These commands are like the break ... thread .... linespec specifies source lines.
Use the qualifier
task `taskid'with a breakpoint command to specify that you only want GDB to stop the program when a particular Ada task reaches this breakpoint. taskid is one of the numeric task identifiers assigned by GDB, shown in the first column of theinfo tasksdisplay.If you do not specify
task `taskid'when you set a breakpoint, the breakpoint applies to `all' tasks of your program.You can use the task qualifier on conditional breakpoints as well; in this case, place
task `taskid'before the breakpoint condition (before the if).
This command allows switching to the task referred by taskno. In particular, this allows browsing of the backtrace of the specified task. It is advisable to switch back to the original task before continuing execution otherwise the scheduling of the program may be perturbed.
For more detailed information on the tasking support, see Debugging with GDB.
GNAT always uses code expansion for generic instantiation. This means that each time an instantiation occurs, a complete copy of the original code is made, with appropriate substitutions of formals by actuals.
It is not possible to refer to the original generic entities in GDB, but it is always possible to debug a particular instance of a generic, by using the appropriate expanded names. For example, if we have
procedure g is
generic package k is
procedure kp (v1 : in out integer);
end k;
package body k is
procedure kp (v1 : in out integer) is
begin
v1 := v1 + 1;
end kp;
end k;
package k1 is new k;
package k2 is new k;
var : integer := 1;
begin
k1.kp (var);
k2.kp (var);
k1.kp (var);
k2.kp (var);
end;
Then to break on a call to procedure kp in the k2 instance, simply use the command:
(gdb) break g.k2.kp
When the breakpoint occurs, you can step through the code of the instance in the normal manner and examine the values of local variables, as for other units.
On platforms where gdbserver is supported, it is possible to use this tool to debug your application remotely. This can be useful in situations where the program needs to be run on a target host that is different from the host used for development, particularly when the target has a limited amount of resources (either CPU and/or memory).
To do so, start your program using gdbserver on the target machine. gdbserver then automatically suspends the execution of your program at its entry point, waiting for a debugger to connect to it. The following commands starts an application and tells gdbserver to wait for a connection with the debugger on localhost port 4444.
$ gdbserver localhost:4444 program
Process program created; pid = 5685
Listening on port 4444
Once gdbserver has started listening, we can tell the debugger to establish a connection with this gdbserver, and then start the same debugging session as if the program was being debugged on the same host, directly under the control of GDB.
$ gdb program
(gdb) target remote targethost:4444
Remote debugging using targethost:4444
0x00007f29936d0af0 in ?? () from /lib64/ld-linux-x86-64.so.
(gdb) b foo.adb:3
Breakpoint 1 at 0x401f0c: file foo.adb, line 3.
(gdb) continue
Continuing.
Breakpoint 1, foo () at foo.adb:4
4 end foo;
It is also possible to use gdbserver to attach to an already running program, in which case the execution of that program is simply suspended until the connection between the debugger and gdbserver is established.
For more information on how to use gdbserver, see the `Using the gdbserver Program' section in Debugging with GDB. GNAT provides support for gdbserver on x86-linux, x86-windows and x86_64-linux.
When presented with programs that contain serious errors in syntax or semantics, GNAT may on rare occasions experience problems in operation, such as aborting with a segmentation fault or illegal memory access, raising an internal exception, terminating abnormally, or failing to terminate at all. In such cases, you can activate various features of GNAT that can help you pinpoint the construct in your program that is the likely source of the problem.
The following strategies are presented in increasing order of difficulty, corresponding to your experience in using GNAT and your familiarity with compiler internals.
The `-gnatdO' switch causes errors to be displayed as soon as they are encountered, rather than after compilation is terminated. If GNAT terminates prematurely or goes into an infinite loop, the last error message displayed may help to pinpoint the culprit.
In order to examine the workings of the GNAT system, the following brief description of its organization may be helpful:
sc contain the lexical scanner.
par are components of the parser. The
numbers correspond to chapters of the Ada Reference Manual. For example,
parsing of select statements can be found in par-ch9.adb.
sem perform semantic analysis. The
numbers correspond to chapters of the Ada standard. For example, all
issues involving context clauses can be found in sem_ch10.adb. In
addition, some features of the language require sufficient special processing
to justify their own semantic files: sem_aggr for aggregates, sem_disp for
dynamic dispatching, etc.
exp perform normalization and
expansion of the intermediate representation (abstract syntax tree, or AST).
these files use the same numbering scheme as the parser and semantics files.
For example, the construction of record initialization procedures is done in
exp_ch3.adb.
bind implement the binder, which
verifies the consistency of the compilation, determines an order of
elaboration, and generates the bind file.
atree.ads and atree.adb detail the low-level
data structures used by the front-end.
sinfo.ads and sinfo.adb detail the structure of
the abstract syntax tree as produced by the parser.
einfo.ads and einfo.adb detail the attributes of
all entities, computed during semantic analysis.
lib.
a- are children of Ada, as
defined in Annex A.
i- are children of Interfaces, as
defined in Annex B.
s- are children of System. This includes
both language-defined children and GNAT run-time routines.
g- are children of GNAT. These are useful
general-purpose packages, fully documented in their specs. All
the other .c files are modifications of common `gcc' files.
Most compilers have internal debugging switches and modes. GNAT
does also, except GNAT internal debugging switches and modes are not
secret. A summary and full description of all the compiler and binder
debug flags are in the file debug.adb. You must obtain the
sources of the compiler to see the full detailed effects of these flags.
The switches that print the source of the program (reconstructed from the internal tree) are of general interest for user programs, as are the options to print the full internal tree, and the entity table (the symbol table information). The reconstructed source provides a readable version of the program after the front-end has completed analysis and expansion, and is useful when studying the performance of specific constructs. For example, constraint checks are indicated, complex aggregates are replaced with loops and assignments, and tasking primitives are replaced with run-time calls.
Traceback is a mechanism to display the sequence of subprogram calls that leads to a specified execution point in a program. Often (but not always) the execution point is an instruction at which an exception has been raised. This mechanism is also known as `stack unwinding' because it obtains its information by scanning the run-time stack and recovering the activation records of all active subprograms. Stack unwinding is one of the most important tools for program debugging.
The first entry stored in traceback corresponds to the deepest calling level, that is to say the subprogram currently executing the instruction from which we want to obtain the traceback.
Note that there is no runtime performance penalty when stack traceback is enabled, and no exception is raised during program execution.
Note: this feature is not supported on all platforms. See
GNAT.Traceback spec in g-traceb.ads
for a complete list of supported platforms.
A runtime non-symbolic traceback is a list of addresses of call instructions. To enable this feature you must use the `-E' gnatbind's option. With this option a stack traceback is stored as part of exception information. You can retrieve this information using the addr2line tool.
Here is a simple example:
procedure STB is procedure P1 is begin raise Constraint_Error; end P1; procedure P2 is begin P1; end P2; begin P2; end STB;$ gnatmake stb -bargs -E $ stb Execution terminated by unhandled exception Exception name: CONSTRAINT_ERROR Message: stb.adb:5 Call stack traceback locations: 0x401373 0x40138b 0x40139c 0x401335 0x4011c4 0x4011f1 0x77e892a4
As we see the traceback lists a sequence of addresses for the unhandled exception CONSTRAINT_ERROR raised in procedure P1. It is easy to guess that this exception come from procedure P1. To translate these addresses into the source lines where the calls appear, the addr2line tool, described below, is invaluable. The use of this tool requires the program to be compiled with debug information.
$ gnatmake -g stb -bargs -E
$ stb
Execution terminated by unhandled exception
Exception name: CONSTRAINT_ERROR
Message: stb.adb:5
Call stack traceback locations:
0x401373 0x40138b 0x40139c 0x401335 0x4011c4 0x4011f1 0x77e892a4
$ addr2line --exe=stb 0x401373 0x40138b 0x40139c 0x401335 0x4011c4
0x4011f1 0x77e892a4
00401373 at d:/stb/stb.adb:5
0040138B at d:/stb/stb.adb:10
0040139C at d:/stb/stb.adb:14
00401335 at d:/stb/b~stb.adb:104
004011C4 at /build/.../crt1.c:200
004011F1 at /build/.../crt1.c:222
77E892A4 in ?? at ??:0
The addr2line tool has several other useful options:
--functionsto get the function name corresponding to any location
--demangle=gnatto use the gnat decoding mode for the function names. Note that for binutils version 2.9.x the option is simply
--demangle.$ addr2line --exe=stb --functions --demangle=gnat 0x401373 0x40138b 0x40139c 0x401335 0x4011c4 0x4011f1 00401373 in stb.p1 at d:/stb/stb.adb:5 0040138B in stb.p2 at d:/stb/stb.adb:10 0040139C in stb at d:/stb/stb.adb:14 00401335 in main at d:/stb/b~stb.adb:104 004011C4 in <__mingw_CRTStartup> at /build/.../crt1.c:200 004011F1 in <mainCRTStartup> at /build/.../crt1.c:222
From this traceback we can see that the exception was raised in
stb.adb at line 5, which was reached from a procedure call in
stb.adb at line 10, and so on. The b~std.adb is the binder file,
which contains the call to the main program.
Running gnatbind. The remaining entries are assorted runtime routines,
and the output will vary from platform to platform.
It is also possible to use GDB with these traceback addresses to debug the program. For example, we can break at a given code location, as reported in the stack traceback:
$ gdb -nw stb
Furthermore, this feature is not implemented inside Windows DLL. Only the non-symbolic traceback is reported in this case.
(gdb) break *0x401373
Breakpoint 1 at 0x401373: file stb.adb, line 5.
It is important to note that the stack traceback addresses do not change when debug information is included. This is particularly useful because it makes it possible to release software without debug information (to minimize object size), get a field report that includes a stack traceback whenever an internal bug occurs, and then be able to retrieve the sequence of calls with the same program compiled with debug information.
Non-symbolic tracebacks are obtained by using the `-E' binder argument. The stack traceback is attached to the exception information string, and can be retrieved in an exception handler within the Ada program, by means of the Ada facilities defined in Ada.Exceptions. Here is a simple example:
with Ada.Text_IO;
with Ada.Exceptions;
procedure STB is
use Ada;
use Ada.Exceptions;
procedure P1 is
K : Positive := 1;
begin
K := K - 1;
exception
when E : others =>
Text_IO.Put_Line (Exception_Information (E));
end P1;
procedure P2 is
begin
P1;
end P2;
begin
P2;
end STB;
This program will output:
$ stb
Exception name: CONSTRAINT_ERROR
Message: stb.adb:12
Call stack traceback locations:
0x4015e4 0x401633 0x401644 0x401461 0x4011c4 0x4011f1 0x77e892a4
It is also possible to retrieve a stack traceback from anywhere in a program. For this you need to use the GNAT.Traceback API. This package includes a procedure called Call_Chain that computes a complete stack traceback, as well as useful display procedures described below. It is not necessary to use the `-E gnatbind' option in this case, because the stack traceback mechanism is invoked explicitly.
In the following example we compute a traceback at a specific location in the program, and we display it using GNAT.Debug_Utilities.Image to convert addresses to strings:
with Ada.Text_IO; with GNAT.Traceback; with GNAT.Debug_Utilities; procedure STB is use Ada; use GNAT; use GNAT.Traceback; procedure P1 is TB : Tracebacks_Array (1 .. 10); -- We are asking for a maximum of 10 stack frames. Len : Natural; -- Len will receive the actual number of stack frames returned. begin Call_Chain (TB, Len); Text_IO.Put ("In STB.P1 : "); for K in 1 .. Len loop Text_IO.Put (Debug_Utilities.Image (TB (K))); Text_IO.Put (' '); end loop; Text_IO.New_Line; end P1; procedure P2 is begin P1; end P2; begin P2; end STB;$ gnatmake -g stb $ stb In STB.P1 : 16#0040_F1E4# 16#0040_14F2# 16#0040_170B# 16#0040_171C# 16#0040_1461# 16#0040_11C4# 16#0040_11F1# 16#77E8_92A4#
You can then get further information by invoking the addr2line tool as described earlier (note that the hexadecimal addresses need to be specified in C format, with a leading '0x').
A symbolic traceback is a stack traceback in which procedure names are associated with each code location.
Note that this feature is not supported on all platforms. See
GNAT.Traceback.Symbolic spec in g-trasym.ads for a complete
list of currently supported platforms.
Note that the symbolic traceback requires that the program be compiled with debug information. If it is not compiled with debug information only the non-symbolic information will be valid.
Here is an example:
with Ada.Text_IO; with GNAT.Traceback.Symbolic; procedure STB is procedure P1 is begin raise Constraint_Error; end P1; procedure P2 is begin P1; end P2; procedure P3 is begin P2; end P3; begin P3; exception when E : others => Ada.Text_IO.Put_Line (GNAT.Traceback.Symbolic.Symbolic_Traceback (E)); end STB;$ gnatmake -g .\stb -bargs -E $ stb 0040149F in stb.p1 at stb.adb:8 004014B7 in stb.p2 at stb.adb:13 004014CF in stb.p3 at stb.adb:18 004015DD in ada.stb at stb.adb:22 00401461 in main at b~stb.adb:168 004011C4 in __mingw_CRTStartup at crt1.c:200 004011F1 in mainCRTStartup at crt1.c:222 77E892A4 in ?? at ??:0
In the above example the .\ syntax in the `gnatmake' command
is currently required by `addr2line' for files that are in
the current working directory.
Moreover, the exact sequence of linker options may vary from platform
to platform.
The above `-largs' section is for Windows platforms. By contrast,
under Unix there is no need for the `-largs' section.
Differences across platforms are due to details of linker implementation.
It is possible to get a symbolic stack traceback from anywhere in a program, just as for non-symbolic tracebacks. The first step is to obtain a non-symbolic traceback, and then call Symbolic_Traceback to compute the symbolic information. Here is an example:
with Ada.Text_IO;
with GNAT.Traceback;
with GNAT.Traceback.Symbolic;
procedure STB is
use Ada;
use GNAT.Traceback;
use GNAT.Traceback.Symbolic;
procedure P1 is
TB : Tracebacks_Array (1 .. 10);
-- We are asking for a maximum of 10 stack frames.
Len : Natural;
-- Len will receive the actual number of stack frames returned.
begin
Call_Chain (TB, Len);
Text_IO.Put_Line (Symbolic_Traceback (TB (1 .. Len)));
end P1;
procedure P2 is
begin
P1;
end P2;
begin
P2;
end STB;
Symbolic tracebacks may also be enabled by using the -Es switch to gnatbind (as in gprbuild -g ... -bargs -Es). This will cause the Exception_Information to contain a symbolic traceback, which will also be printed if an unhandled exception terminates the program.
This section describes how to use the gcov coverage testing tool and the gprof profiler tool on Ada programs.
gcov is a test coverage program: it analyzes the execution of a given program on selected tests, to help you determine the portions of the program that are still untested.
gcov is part of the GCC suite, and is described in detail in the GCC User's Guide. You can refer to this documentation for a more complete description.
This chapter provides a quick startup guide, and details some GNAT-specific features.
In order to perform coverage analysis of a program using gcov, several steps are needed:
The code instrumentation needed by gcov is created at the object level. The source code is not modified in any way, because the instrumentation code is inserted by gcc during the compilation process. To compile your code with code coverage activated, you need to recompile your whole project using the switches -fprofile-arcs and -ftest-coverage, and link it using -fprofile-arcs.
$ gnatmake -P my_project.gpr -f -cargs -fprofile-arcs -ftest-coverage \\
-largs -fprofile-arcs
This compilation process will create .gcno files together with
the usual object files.
Once the program is compiled with coverage instrumentation, you can
run it as many times as needed – on portions of a test suite for
example. The first execution will produce .gcda files at the
same location as the .gcno files. Subsequent executions
will update those files, so that a cumulative result of the covered
portions of the program is generated.
Finally, you need to call the gcov tool. The different options of gcov are described in the GCC User's Guide, section 'Invoking gcov'.
This will create annotated source files with a .gcov extension:
my_main.adb file will be analyzed in my_main.adb.gcov.
Because of Ada semantics, portions of the source code may be shared among several object files. This is the case for example when generics are involved, when inlining is active or when declarations generate initialisation calls. In order to take into account this shared code, you need to call gcov on all source files of the tested program at once.
The list of source files might exceed the system's maximum command line
length. In order to bypass this limitation, a new mechanism has been
implemented in gcov: you can now list all your project's files into a
text file, and provide this file to gcov as a parameter, preceded by a @
(e.g. gcov @mysrclist.txt).
Note that on AIX compiling a static library with -fprofile-arcs is not supported as there can be unresolved symbols during the final link.
This section is not meant to be an exhaustive documentation of gprof. Full documentation for it can be found in the GNU Profiler User's Guide documentation that is part of this GNAT distribution.
Profiling a program helps determine the parts of a program that are executed most often, and are therefore the most time-consuming.
gprof is the standard GNU profiling tool; it has been enhanced to better handle Ada programs and multitasking. It is currently supported on the following platforms
In order to profile a program using gprof, several steps are needed:
The following sections detail the different steps, and indicate how to interpret the results.
In order to profile a program the first step is to tell the compiler
to generate the necessary profiling information. The compiler switch to be used
is -pg, which must be added to other compilation switches. This
switch needs to be specified both during compilation and link stages, and can
be specified once when using gnatmake:
$ gnatmake -f -pg -P my_project
Note that only the objects that were compiled with the -pg switch will
be profiled; if you need to profile your whole project, use the -f
gnatmake switch to force full recompilation.
Once the program has been compiled for profiling, you can run it as usual.
The only constraint imposed by profiling is that the program must terminate normally. An interrupted program (via a Ctrl-C, kill, etc.) will not be properly analyzed.
Once the program completes execution, a data file called gmon.out is
generated in the directory where the program was launched from. If this file
already exists, it will be overwritten.
The gprof tool is called as follow:
$ gprof my_prog gmon.out
or simply:
$ gprof my_prog
The complete form of the gprof command line is the following:
$ gprof [switches] [executable [data-file]]
gprof supports numerous switches. The order of these switch does not matter. The full list of options can be found in the GNU Profiler User's Guide documentation that comes with this documentation.
The following is the subset of those switches that is most relevant:
--demangle[=`style'], --no-demangle--no-demangle option may be used to turn off demangling. Different
compilers have different mangling styles. The optional demangling style
argument can be used to choose an appropriate demangling style for your
compiler, in particular Ada symbols generated by GNAT can be demangled using
--demangle=gnat.
-e `function_name'-e `function' option tells gprof not to print
information about the function function_name (and its
children...) in the call graph. The function will still be listed
as a child of any functions that call it, but its index number will be
shown as [not printed]. More than one -e option may be
given; only one function_name may be indicated with each -e
option.
-E `function_name'-E `function' option works like the -e option, but
execution time spent in the function (and children who were not called from
anywhere else), will not be used to compute the percentages-of-time for
the call graph. More than one -E option may be given; only one
function_name may be indicated with each -E option.
-f `function_name'-f `function' option causes gprof to limit the
call graph to the function function_name and its children (and
their children...). More than one -f option may be given;
only one function_name may be indicated with each -f
option.
-F `function_name'-F `function' option works like the -f option, but
only time spent in the function and its children (and their
children...) will be used to determine total-time and
percentages-of-time for the call graph. More than one -F option
may be given; only one function_name may be indicated with each
-F option. The -F option overrides the -E option.
The results of the profiling analysis are represented by two arrays: the 'flat profile' and the 'call graph'. Full documentation of those outputs can be found in the GNU Profiler User's Guide.
The flat profile shows the time spent in each function of the program, and how many time it has been called. This allows you to locate easily the most time-consuming functions.
The call graph shows, for each subprogram, the subprograms that call it, and the subprograms that it calls. It also provides an estimate of the time spent in each of those callers/called subprograms.
This section presents several topics related to program performance. It first describes some of the tradeoffs that need to be considered and some of the techniques for making your program run faster.
It then documents the unused subprogram/data elimination feature, which can reduce the size of program executables.
The GNAT system provides a number of options that allow a trade-off between
The defaults (if no options are selected) aim at improving the speed of compilation and minimizing dependences, at the expense of performance of the generated code:
These options are suitable for most program development purposes. This section describes how you can modify these choices, and also provides some guidelines on debugging optimized code.
By default, GNAT generates all run-time checks, except stack overflow checks, and checks for access before elaboration on subprogram calls. The latter are not required in default mode, because all necessary checking is done at compile time.
The gnat switch, `-gnatp' allows this default to be modified. See Run-Time Checks.
Our experience is that the default is suitable for most development purposes.
Elaboration checks are off by default, and also not needed by default, since GNAT uses a static elaboration analysis approach that avoids the need for run-time checking. This manual contains a full chapter discussing the issue of elaboration checks, and if the default is not satisfactory for your use, you should read this chapter.
For validity checks, the minimal checks required by the Ada Reference Manual (for case statements and assignments to array elements) are on by default. These can be suppressed by use of the `-gnatVn' switch. Note that in Ada 83, there were no validity checks, so if the Ada 83 mode is acceptable (or when comparing GNAT performance with an Ada 83 compiler), it may be reasonable to routinely use `-gnatVn'. Validity checks are also suppressed entirely if `-gnatp' is used.
Note that the setting of the switches controls the default setting of the checks. They may be modified using either pragma Suppress (to remove checks) or pragma Unsuppress (to add back suppressed checks) in the program source.
The use of pragma Restrictions allows you to control which features are permitted in your program. Apart from the obvious point that if you avoid relatively expensive features like finalization (enforceable by the use of pragma Restrictions (No_Finalization), the use of this pragma does not affect the generated code in most cases.
One notable exception to this rule is that the possibility of task abort results in some distributed overhead, particularly if finalization or exception handlers are used. The reason is that certain sections of code have to be marked as non-abortable.
If you use neither the abort statement, nor asynchronous transfer of control (select ... then abort), then this distributed overhead is removed, which may have a general positive effect in improving overall performance. Especially code involving frequent use of tasking constructs and controlled types will show much improved performance. The relevant restrictions pragmas are
pragma Restrictions (No_Abort_Statements);
pragma Restrictions (Max_Asynchronous_Select_Nesting => 0);
It is recommended that these restriction pragmas be used if possible. Note that this also means that you can write code without worrying about the possibility of an immediate abort at any point.
Without any optimization option, the compiler's goal is to reduce the cost of compilation and to make debugging produce the expected results. Statements are independent: if you stop the program with a breakpoint between statements, you can then assign a new value to any variable or change the program counter to any other statement in the subprogram and get exactly the results you would expect from the source code.
Turning on optimization makes the compiler attempt to improve the performance and/or code size at the expense of compilation time and possibly the ability to debug the program.
If you use multiple -O options, with or without level numbers, the last such option is the one that is effective.
The default is optimization off. This results in the fastest compile times, but GNAT makes absolutely no attempt to optimize, and the generated programs are considerably larger and slower than when optimization is enabled. You can use the `-O' switch (the permitted forms are `-O0', `-O1' `-O2', `-O3', and `-Os') to `gcc' to control the optimization level:
Note that many other compilers do fairly extensive optimization even if 'no optimization' is specified. With gcc, it is very unusual to use -O0 for production if execution time is of any concern, since -O0 really does mean no optimization at all. This difference between gcc and other compilers should be kept in mind when doing performance comparisons.
Higher optimization levels perform more global transformations on the program and apply more expensive analysis algorithms in order to generate faster and more compact code. The price in compilation time, and the resulting improvement in execution time, both depend on the particular application and the hardware environment. You should experiment to find the best level for your application.
Since the precise set of optimizations done at each level will vary from release to release (and sometime from target to target), it is best to think of the optimization settings in general terms. See the `Options That Control Optimization' section in Using the GNU Compiler Collection (GCC) for details about the `-O' settings and a number of `-f' options that individually enable or disable specific optimizations.
Unlike some other compilation systems, `gcc' has been tested extensively at all optimization levels. There are some bugs which appear only with optimization turned on, but there have also been bugs which show up only in `unoptimized' code. Selecting a lower level of optimization does not improve the reliability of the code generator, which in practice is highly reliable at all optimization levels.
Note regarding the use of `-O3': The use of this optimization level is generally discouraged with GNAT, since it often results in larger executables which may run more slowly. See further discussion of this point in Inlining of Subprograms.
Although it is possible to do a reasonable amount of debugging at nonzero optimization levels, the higher the level the more likely that source-level constructs will have been eliminated by optimization. For example, if a loop is strength-reduced, the loop control variable may be completely eliminated and thus cannot be displayed in the debugger. This can only happen at `-O2' or `-O3'. Explicit temporary variables that you code might be eliminated at level `-O1' or higher.
The use of the `-g' switch, which is needed for source-level debugging, affects the size of the program executable on disk, and indeed the debugging information can be quite large. However, it has no effect on the generated code (and thus does not degrade performance)
Since the compiler generates debugging tables for a compilation unit before it performs optimizations, the optimizing transformations may invalidate some of the debugging data. You therefore need to anticipate certain anomalous situations that may arise while debugging optimized code. These are the most common cases:
In general, when an unexpected value appears for a local variable or parameter you should first ascertain if that value was actually computed by your program, as opposed to being incorrectly reported by the debugger. Record fields or array elements in an object designated by an access value are generally less of a problem, once you have ascertained that the access value is sensible. Typically, this means checking variables in the preceding code and in the calling subprogram to verify that the value observed is explainable from other values (one must apply the procedure recursively to those other values); or re-running the code and stopping a little earlier (perhaps before the call) and stepping to better see how the variable obtained the value in question; or continuing to step `from' the point of the strange value to see if code motion had simply moved the variable's assignments later.
In light of such anomalies, a recommended technique is to use `-O0' early in the software development cycle, when extensive debugging capabilities are most needed, and then move to `-O1' and later `-O2' as the debugger becomes less critical. Whether to use the `-g' switch in the release version is a release management issue. Note that if you use `-g' you can then use the `strip' program on the resulting executable, which removes both debugging information and global symbols.
A call to a subprogram in the current unit is inlined if all the following conditions are met:
Calls to subprograms in `with'ed units are normally not inlined. To achieve actual inlining (that is, replacement of the call by the code in the body of the subprogram), the following conditions must all be true:
Even if all these conditions are met, it may not be possible for the compiler to inline the call, due to the length of the body, or features in the body that make it impossible for the compiler to do the inlining.
Note that specifying the `-gnatn' switch causes additional compilation dependencies. Consider the following:
package R is
procedure Q;
pragma Inline (Q);
end R;
package body R is
...
end R;
with R;
procedure Main is
begin
...
R.Q;
end Main;
With the default behavior (no `-gnatn' switch specified), the
compilation of the Main procedure depends only on its own source,
main.adb, and the spec of the package in file r.ads. This
means that editing the body of R does not require recompiling
Main.
On the other hand, the call R.Q is not inlined under these
circumstances. If the `-gnatn' switch is present when Main
is compiled, the call will be inlined if the body of Q is small
enough, but now Main depends on the body of R in
r.adb as well as on the spec. This means that if this body is edited,
the main program must be recompiled. Note that this extra dependency
occurs whether or not the call is in fact inlined by `gcc'.
The use of front end inlining with `-gnatN' generates similar additional dependencies.
Note: The `-fno-inline' switch overrides all other conditions and ensures that no inlining occurs, unless requested with pragma Inline_Always for gcc back-ends. The extra dependences resulting from `-gnatn' will still be active, even if this switch is used to suppress the resulting inlining actions.
Note: The `-fno-inline-functions' switch can be used to prevent automatic inlining of subprograms if `-O3' is used.
Note: The `-fno-inline-small-functions' switch can be used to prevent automatic inlining of small subprograms if `-O2' is used.
Note: The `-fno-inline-functions-called-once' switch can be used to prevent inlining of subprograms local to the unit and called once from within it if `-O1' is used.
Note regarding the use of `-O3': `-gnatn' is made up of two sub-switches `-gnatn1' and `-gnatn2' that can be directly specified in lieu of it, `-gnatn' being translated into one of them based on the optimization level. With `-O2' or below, `-gnatn' is equivalent to `-gnatn1' which activates pragma Inline with moderate inlining across modules. With `-O3', `-gnatn' is equivalent to `-gnatn2' which activates pragma Inline with full inlining across modules. If you have used pragma Inline in appropriate cases, then it is usually much better to use `-O2' and `-gnatn' and avoid the use of `-O3' which has the additional effect of inlining subprograms you did not think should be inlined. We have found that the use of `-O3' may slow down the compilation and increase the code size by performing excessive inlining, leading to increased instruction cache pressure from the increased code size and thus minor performance improvements. So the bottom line here is that you should not automatically assume that `-O3' is better than `-O2', and indeed you should use `-O3' only if tests show that it actually improves performance for your program.
On almost all targets, GNAT maps Float and Long_Float to the 32-bit and 64-bit standard IEEE floating-point representations, and operations will use standard IEEE arithmetic as provided by the processor. On most, but not all, architectures, the attribute Machine_Overflows is False for these types, meaning that the semantics of overflow is implementation-defined. In the case of GNAT, these semantics correspond to the normal IEEE treatment of infinities and NaN (not a number) values. For example, 1.0 / 0.0 yields plus infinitiy and 0.0 / 0.0 yields a NaN. By avoiding explicit overflow checks, the performance is greatly improved on many targets. However, if required, floating-point overflow can be enabled by the use of the pragma Check_Float_Overflow.
Another consideration that applies specifically to x86 32-bit architectures is which form of floating-point arithmetic is used. By default the operations use the old style x86 floating-point, which implements an 80-bit extended precision form (on these architectures the type Long_Long_Float corresponds to that form). In addition, generation of efficient code in this mode means that the extended precision form will be used for intermediate results. This may be helpful in improving the final precision of a complex expression. However it means that the results obtained on the x86 will be different from those on other architectures, and for some algorithms, the extra intermediate precision can be detrimental.
In addition to this old-style floating-point, all modern x86 chips implement an alternative floating-point operation model referred to as SSE2. In this model there is no extended form, and furthermore execution performance is significantly enhanced. To force GNAT to use this more modern form, use both of the switches:
-msse2 -mfpmath=sse
A unit compiled with these switches will automatically use the more efficient SSE2 instruction set for Float and Long_Float operations. Note that the ABI has the same form for both floating-point models, so it is permissible to mix units compiled with and without these switches.
You can take advantage of the auto-vectorizer present in the `gcc' back end to vectorize loops with GNAT. The corresponding command line switch is `-ftree-vectorize' but, as it is enabled by default at `-O3' and other aggressive optimizations helpful for vectorization also are enabled by default at this level, using `-O3' directly is recommended.
You also need to make sure that the target architecture features a supported SIMD instruction set. For example, for the x86 architecture, you should at least specify `-msse2' to get significant vectorization (but you don't need to specify it for x86-64 as it is part of the base 64-bit architecture). Similarly, for the PowerPC architecture, you should specify `-maltivec'.
The preferred loop form for vectorization is the for iteration scheme. Loops with a while iteration scheme can also be vectorized if they are very simple, but the vectorizer will quickly give up otherwise. With either iteration scheme, the flow of control must be straight, in particular no exit statement may appear in the loop body. The loop may however contain a single nested loop, if it can be vectorized when considered alone:
A : array (1..4, 1..4) of Long_Float;
S : array (1..4) of Long_Float;
procedure Sum is
begin
for I in A'Range(1) loop
for J in A'Range(2) loop
S (I) := S (I) + A (I, J);
end loop;
end loop;
end Sum;
The vectorizable operations depend on the targeted SIMD instruction set, but the adding and some of the multiplying operators are generally supported, as well as the logical operators for modular types. Note that compiling with `-gnatp' might well reveal cases where some checks do thwart vectorization.
Type conversions may also prevent vectorization if they involve semantics that are not directly supported by the code generator or the SIMD instruction set. A typical example is direct conversion from floating-point to integer types. The solution in this case is to use the following idiom:
Integer (S'Truncation (F))
if S is the subtype of floating-point object F.
In most cases, the vectorizable loops are loops that iterate over arrays. All kinds of array types are supported, i.e. constrained array types with static bounds:
type Array_Type is array (1 .. 4) of Long_Float;
constrained array types with dynamic bounds:
type Array_Type is array (1 .. Q.N) of Long_Float;
type Array_Type is array (Q.K .. 4) of Long_Float;
type Array_Type is array (Q.K .. Q.N) of Long_Float;
or unconstrained array types:
type Array_Type is array (Positive range <>) of Long_Float;
The quality of the generated code decreases when the dynamic aspect of the array type increases, the worst code being generated for unconstrained array types. This is so because, the less information the compiler has about the bounds of the array, the more fallback code it needs to generate in order to fix things up at run time.
It is possible to specify that a given loop should be subject to vectorization preferably to other optimizations by means of pragma Loop_Optimize:
pragma Loop_Optimize (Vector);
placed immediately within the loop will convey the appropriate hint to the compiler for this loop.
It is also possible to help the compiler generate better vectorized code for a given loop by asserting that there are no loop-carried dependencies in the loop. Consider for example the procedure:
type Arr is array (1 .. 4) of Long_Float;
procedure Add (X, Y : not null access Arr; R : not null access Arr) is
begin
for I in Arr'Range loop
R(I) := X(I) + Y(I);
end loop;
end;
By default, the compiler cannot unconditionally vectorize the loop because assigning to a component of the array designated by R in one iteration could change the value read from the components of the array designated by X or Y in a later iteration. As a result, the compiler will generate two versions of the loop in the object code, one vectorized and the other not vectorized, as well as a test to select the appropriate version at run time. This can be overcome by another hint:
pragma Loop_Optimize (Ivdep);
placed immediately within the loop will tell the compiler that it can safely omit the non-vectorized version of the loop as well as the run-time test.
Since GNAT uses the `gcc' back end, all the specialized `gcc' optimization switches are potentially usable. These switches have not been extensively tested with GNAT but can generally be expected to work. Examples of switches in this category are `-funroll-loops' and the various target-specific `-m' options (in particular, it has been observed that `-march=xxx' can significantly improve performance on appropriate machines). For full details of these switches, see the Submodel Options section in the Hardware Models and Configurations chapter of Using the GNU Compiler Collection (GCC).
The strong typing capabilities of Ada allow an optimizer to generate efficient code in situations where other languages would be forced to make worst case assumptions preventing such optimizations. Consider the following example:
procedure R is
type Int1 is new Integer;
type Int2 is new Integer;
type Int1A is access Int1;
type Int2A is access Int2;
Int1V : Int1A;
Int2V : Int2A;
...
begin
...
for J in Data'Range loop
if Data (J) = Int1V.all then
Int2V.all := Int2V.all + 1;
end if;
end loop;
...
end R;
In this example, since the variable Int1V can only access objects of type Int1, and Int2V can only access objects of type Int2, there is no possibility that the assignment to Int2V.all affects the value of Int1V.all. This means that the compiler optimizer can "know" that the value Int1V.all is constant for all iterations of the loop and avoid the extra memory reference required to dereference it each time through the loop.
This kind of optimization, called strict aliasing analysis, is triggered by specifying an optimization level of `-O2' or higher or `-Os' and allows GNAT to generate more efficient code when access values are involved.
However, although this optimization is always correct in terms of the formal semantics of the Ada Reference Manual, difficulties can arise if features like Unchecked_Conversion are used to break the typing system. Consider the following complete program example:
package p1 is
type int1 is new integer;
type int2 is new integer;
type a1 is access int1;
type a2 is access int2;
end p1;
with p1; use p1;
package p2 is
function to_a2 (Input : a1) return a2;
end p2;
with Unchecked_Conversion;
package body p2 is
function to_a2 (Input : a1) return a2 is
function to_a2u is
new Unchecked_Conversion (a1, a2);
begin
return to_a2u (Input);
end to_a2;
end p2;
with p2; use p2;
with p1; use p1;
with Text_IO; use Text_IO;
procedure m is
v1 : a1 := new int1;
v2 : a2 := to_a2 (v1);
begin
v1.all := 1;
v2.all := 0;
put_line (int1'image (v1.all));
end;
This program prints out 0 in `-O0' or `-O1' mode, but it prints out 1 in `-O2' mode. That's because in strict aliasing mode, the compiler can and does assume that the assignment to v2.all could not affect the value of v1.all, since different types are involved.
This behavior is not a case of non-conformance with the standard, since the Ada RM specifies that an unchecked conversion where the resulting bit pattern is not a correct value of the target type can result in an abnormal value and attempting to reference an abnormal value makes the execution of a program erroneous. That's the case here since the result does not point to an object of type int2. This means that the effect is entirely unpredictable.
However, although that explanation may satisfy a language lawyer, in practice an applications programmer expects an unchecked conversion involving pointers to create true aliases and the behavior of printing 1 seems plain wrong. In this case, the strict aliasing optimization is unwelcome.
Indeed the compiler recognizes this possibility, and the unchecked conversion generates a warning:
p2.adb:5:07: warning: possible aliasing problem with type "a2"
p2.adb:5:07: warning: use -fno-strict-aliasing switch for references
p2.adb:5:07: warning: or use "pragma No_Strict_Aliasing (a2);"
Unfortunately the problem is recognized when compiling the body of package p2, but the actual "bad" code is generated while compiling the body of m and this latter compilation does not see the suspicious Unchecked_Conversion.
As implied by the warning message, there are approaches you can use to avoid the unwanted strict aliasing optimization in a case like this.
One possibility is to simply avoid the use of `-O2', but that is a bit drastic, since it throws away a number of useful optimizations that do not involve strict aliasing assumptions.
A less drastic approach is to compile the program using the option `-fno-strict-aliasing'. Actually it is only the unit containing the dereferencing of the suspicious pointer that needs to be compiled. So in this case, if we compile unit m with this switch, then we get the expected value of zero printed. Analyzing which units might need the switch can be painful, so a more reasonable approach is to compile the entire program with options `-O2' and `-fno-strict-aliasing'. If the performance is satisfactory with this combination of options, then the advantage is that the entire issue of possible "wrong" optimization due to strict aliasing is avoided.
To avoid the use of compiler switches, the configuration pragma No_Strict_Aliasing with no parameters may be used to specify that for all access types, the strict aliasing optimization should be suppressed.
However, these approaches are still overkill, in that they causes all manipulations of all access values to be deoptimized. A more refined approach is to concentrate attention on the specific access type identified as problematic.
First, if a careful analysis of uses of the pointer shows that there are no possible problematic references, then the warning can be suppressed by bracketing the instantiation of Unchecked_Conversion to turn the warning off:
pragma Warnings (Off);
function to_a2u is
new Unchecked_Conversion (a1, a2);
pragma Warnings (On);
Of course that approach is not appropriate for this particular example, since indeed there is a problematic reference. In this case we can take one of two other approaches.
The first possibility is to move the instantiation of unchecked conversion to the unit in which the type is declared. In this example, we would move the instantiation of Unchecked_Conversion from the body of package p2 to the spec of package p1. Now the warning disappears. That's because any use of the access type knows there is a suspicious unchecked conversion, and the strict aliasing optimization is automatically suppressed for the type.
If it is not practical to move the unchecked conversion to the same unit in which the destination access type is declared (perhaps because the source type is not visible in that unit), you may use pragma No_Strict_Aliasing for the type. This pragma must occur in the same declarative sequence as the declaration of the access type:
type a2 is access int2;
pragma No_Strict_Aliasing (a2);
Here again, the compiler now knows that the strict aliasing optimization should be suppressed for any reference to type a2 and the expected behavior is obtained.
Finally, note that although the compiler can generate warnings for simple cases of unchecked conversions, there are tricker and more indirect ways of creating type incorrect aliases which the compiler cannot detect. Examples are the use of address overlays and unchecked conversions involving composite types containing access types as components. In such cases, no warnings are generated, but there can still be aliasing problems. One safe coding practice is to forbid the use of address clauses for type overlaying, and to allow unchecked conversion only for primitive types. This is not really a significant restriction since any possible desired effect can be achieved by unchecked conversion of access values.
The aliasing analysis done in strict aliasing mode can certainly have significant benefits. We have seen cases of large scale application code where the time is increased by up to 5% by turning this optimization off. If you have code that includes significant usage of unchecked conversion, you might want to just stick with `-O1' and avoid the entire issue. If you get adequate performance at this level of optimization level, that's probably the safest approach. If tests show that you really need higher levels of optimization, then you can experiment with `-O2' and `-O2 -fno-strict-aliasing' to see how much effect this has on size and speed of the code. If you really need to use `-O2' with strict aliasing in effect, then you should review any uses of unchecked conversion of access types, particularly if you are getting the warnings described above.
There are scenarios in which programs may use low level techniques to modify variables that otherwise might be considered to be unassigned. For example, a variable can be passed to a procedure by reference, which takes the address of the parameter and uses the address to modify the variable's value, even though it is passed as an IN parameter. Consider the following example:
procedure P is
Max_Length : constant Natural := 16;
type Char_Ptr is access all Character;
procedure Get_String(Buffer: Char_Ptr; Size : Integer);
pragma Import (C, Get_String, "get_string");
Name : aliased String (1 .. Max_Length) := (others => ' ');
Temp : Char_Ptr;
function Addr (S : String) return Char_Ptr is
function To_Char_Ptr is
new Ada.Unchecked_Conversion (System.Address, Char_Ptr);
begin
return To_Char_Ptr (S (S'First)'Address);
end;
begin
Temp := Addr (Name);
Get_String (Temp, Max_Length);
end;
where Get_String is a C function that uses the address in Temp to modify the variable Name. This code is dubious, and arguably erroneous, and the compiler would be entitled to assume that Name is never modified, and generate code accordingly.
However, in practice, this would cause some existing code that seems to work with no optimization to start failing at high levels of optimzization.
What the compiler does for such cases is to assume that marking a variable as aliased indicates that some "funny business" may be going on. The optimizer recognizes the aliased keyword and inhibits optimizations that assume the value cannot be assigned. This means that the above example will in fact "work" reliably, that is, it will produce the expected results.
There are two considerations with regard to performance when atomic variables are used.
First, the RM only guarantees that access to atomic variables be atomic, it has nothing to say about how this is achieved, though there is a strong implication that this should not be achieved by explicit locking code. Indeed GNAT will never generate any locking code for atomic variable access (it will simply reject any attempt to make a variable or type atomic if the atomic access cannot be achieved without such locking code).
That being said, it is important to understand that you cannot assume that the entire variable will always be accessed. Consider this example:
type R is record
A,B,C,D : Character;
end record;
for R'Size use 32;
for R'Alignment use 4;
RV : R;
pragma Atomic (RV);
X : Character;
...
X := RV.B;
You cannot assume that the reference to RV.B will read the entire 32-bit variable with a single load instruction. It is perfectly legitimate if the hardware allows it to do a byte read of just the B field. This read is still atomic, which is all the RM requires. GNAT can and does take advantage of this, depending on the architecture and optimization level. Any assumption to the contrary is non-portable and risky. Even if you examine the assembly language and see a full 32-bit load, this might change in a future version of the compiler.
If your application requires that all accesses to RV in this example be full 32-bit loads, you need to make a copy for the access as in:
declare
RV_Copy : constant R := RV;
begin
X := RV_Copy.B;
end;
Now the reference to RV must read the whole variable. Actually one can imagine some compiler which figures out that the whole copy is not required (because only the B field is actually accessed), but GNAT certainly won't do that, and we don't know of any compiler that would not handle this right, and the above code will in practice work portably across all architectures (that permit the Atomic declaration).
The second issue with atomic variables has to do with the possible requirement of generating synchronization code. For more details on this, consult the sections on the pragmas Enable/Disable_Atomic_Synchronization in the GNAT Reference Manual. If performance is critical, and such synchronization code is not required, it may be useful to disable it.
A passive task is one which is sufficiently simple that in theory a compiler could recognize it an implement it efficiently without creating a new thread. The original design of Ada 83 had in mind this kind of passive task optimization, but only a few Ada 83 compilers attempted it. The problem was that it was difficult to determine the exact conditions under which the optimization was possible. The result is a very fragile optimization where a very minor change in the program can suddenly silently make a task non-optimizable.
With the revisiting of this issue in Ada 95, there was general agreement that this approach was fundamentally flawed, and the notion of protected types was introduced. When using protected types, the restrictions are well defined, and you KNOW that the operations will be optimized, and furthermore this optimized performance is fully portable.
Although it would theoretically be possible for GNAT to attempt to do this optimization, but it really doesn't make sense in the context of Ada 95, and none of the Ada 95 compilers implement this optimization as far as we know. In particular GNAT never attempts to perform this optimization.
In any new Ada 95 code that is written, you should always use protected types in place of tasks that might be able to be optimized in this manner. Of course this does not help if you have legacy Ada 83 code that depends on this optimization, but it is unusual to encounter a case where the performance gains from this optimization are significant.
Your program should work correctly without this optimization. If you have performance problems, then the most practical approach is to figure out exactly where these performance problems arise, and update those particular tasks to be protected types. Note that typically clients of the tasks who call entries, will not have to be modified, only the task definition itself.
The Ada.Text_IO package has fairly high overheads due in part to the requirement of maintaining page and line counts. If performance is critical, a recommendation is to use Stream_IO instead of Text_IO for volume output, since this package has less overhead.
If Text_IO must be used, note that by default output to the standard output and standard error files is unbuffered (this provides better behavior when output statements are used for debugging, or if the progress of a program is observed by tracking the output, e.g. by using the Unix `tail -f' command to watch redirected output.
If you are generating large volumes of output with Text_IO and performance is an important factor, use a designated file instead of the standard output file, or change the standard output file to be buffered using Interfaces.C_Streams.setvbuf.
This section describes how you can eliminate unused subprograms and data from your executable just by setting options at compilation time.
By default, an executable contains all code and data of its composing objects (directly linked or coming from statically linked libraries), even data or code never used by this executable.
This feature will allow you to eliminate such unused code from your executable, making it smaller (in disk and in memory).
This functionality is available on all Linux platforms except for the IA-64 architecture and on all cross platforms using the ELF binary file format. In both cases GNU binutils version 2.16 or later are required to enable it.
The operation of eliminating the unused code and data from the final executable is directly performed by the linker.
In order to do this, it has to work with objects compiled with the following options: `-ffunction-sections' `-fdata-sections'.
These options are usable with C and Ada files. They will place respectively each function or data in a separate section in the resulting object file.
Once the objects and static libraries are created with these options, the linker can perform the dead code elimination. You can do this by setting the `-Wl,–gc-sections' option to gcc command or in the `-largs' section of `gnatmake'. This will perform a garbage collection of code and data never referenced.
If the linker performs a partial link (`-r' linker option), then you will need to provide the entry point using the `-e' / `–entry' linker option.
Note that objects compiled without the `-ffunction-sections' and `-fdata-sections' options can still be linked with the executable. However, no dead code elimination will be performed on those objects (they will be linked as is).
The GNAT static library is now compiled with -ffunction-sections and -fdata-sections on some platforms. This allows you to eliminate the unused code and data of the GNAT library from your executable.
Here is a simple example:
with Aux;
procedure Test is
begin
Aux.Used (10);
end Test;
package Aux is
Used_Data : Integer;
Unused_Data : Integer;
procedure Used (Data : Integer);
procedure Unused (Data : Integer);
end Aux;
package body Aux is
procedure Used (Data : Integer) is
begin
Used_Data := Data;
end Used;
procedure Unused (Data : Integer) is
begin
Unused_Data := Data;
end Unused;
end Aux;
Unused and Unused_Data are never referenced in this code excerpt, and hence they may be safely removed from the final executable.
$ gnatmake test
$ nm test | grep used
020015f0 T aux__unused
02005d88 B aux__unused_data
020015cc T aux__used
02005d84 B aux__used_data
$ gnatmake test -cargs -fdata-sections -ffunction-sections \\
-largs -Wl,--gc-sections
$ nm test | grep used
02005350 T aux__used
0201ffe0 B aux__used_data
It can be observed that the procedure Unused and the object Unused_Data are removed by the linker when using the appropriate options.
This section explains how to control the handling of overflow checks.
Overflow checks are checks that the compiler may make to ensure that intermediate results are not out of range. For example:
A : Integer;
...
A := A + 1;
If A has the value Integer'Last, then the addition may cause overflow since the result is out of range of the type Integer. In this case Constraint_Error will be raised if checks are enabled.
A trickier situation arises in examples like the following:
A, C : Integer;
...
A := (A + 1) + C;
where A is Integer'Last and C is -1. Now the final result of the expression on the right hand side is Integer'Last which is in range, but the question arises whether the intermediate addition of (A + 1) raises an overflow error.
The (perhaps surprising) answer is that the Ada language definition does not answer this question. Instead it leaves it up to the implementation to do one of two things if overflow checks are enabled.
If the compiler chooses the first approach, then the assignment of this example will indeed raise Constraint_Error if overflow checking is enabled, or result in erroneous execution if overflow checks are suppressed.
But if the compiler chooses the second approach, then it can perform both additions yielding the correct mathematical result, which is in range, so no exception will be raised, and the right result is obtained, regardless of whether overflow checks are suppressed.
Note that in the first example an exception will be raised in either case, since if the compiler gives the correct mathematical result for the addition, it will be out of range of the target type of the assignment, and thus fails the range check.
This lack of specified behavior in the handling of overflow for intermediate results is a source of non-portability, and can thus be problematic when programs are ported. Most typically this arises in a situation where the original compiler did not raise an exception, and then the application is moved to a compiler where the check is performed on the intermediate result and an unexpected exception is raised.
Furthermore, when using Ada 2012's preconditions and other assertion forms, another issue arises. Consider:
procedure P (A, B : Integer) with
Pre => A + B <= Integer'Last;
One often wants to regard arithmetic in a context like this from a mathematical point of view. So for example, if the two actual parameters for a call to P are both Integer'Last, then the precondition should be regarded as False. If we are executing in a mode with run-time checks enabled for preconditions, then we would like this precondition to fail, rather than raising an exception because of the intermediate overflow.
However, the language definition leaves the specification of whether the above condition fails (raising Assert_Error) or causes an intermediate overflow (raising Constraint_Error) up to the implementation.
The situation is worse in a case such as the following:
procedure Q (A, B, C : Integer) with
Pre => A + B + C <= Integer'Last;
Consider the call
Q (A => Integer'Last, B => 1, C => -1);
From a mathematical point of view the precondition is True, but at run time we may (but are not guaranteed to) get an exception raised because of the intermediate overflow (and we really would prefer this precondition to be considered True at run time).
To deal with the portability issue, and with the problem of mathematical versus run-time interpretation of the expressions in assertions, GNAT provides comprehensive control over the handling of intermediate overflow. GNAT can operate in three modes, and furthemore, permits separate selection of operating modes for the expressions within assertions (here the term 'assertions' is used in the technical sense, which includes preconditions and so forth) and for expressions appearing outside assertions.
The three modes are:
In this mode, all intermediate results for predefined arithmetic operators are computed using the base type, and the result must be in range of the base type. If this is not the case then either an exception is raised (if overflow checks are enabled) or the execution is erroneous (if overflow checks are suppressed). This is the normal default mode.
In this mode, the compiler attempts to avoid intermediate overflows by using a larger integer type, typically Long_Long_Integer, as the type in which arithmetic is performed for predefined arithmetic operators. This may be slightly more expensive at run time (compared to suppressing intermediate overflow checks), though the cost is negligible on modern 64-bit machines. For the examples given earlier, no intermediate overflows would have resulted in exceptions, since the intermediate results are all in the range of Long_Long_Integer (typically 64-bits on nearly all implementations of GNAT). In addition, if checks are enabled, this reduces the number of checks that must be made, so this choice may actually result in an improvement in space and time behavior.
However, there are cases where Long_Long_Integer is not large enough, consider the following example:
procedure R (A, B, C, D : Integer) with
Pre => (A**2 * B**2) / (C**2 * D**2) <= 10;
where A = B = C = D = Integer'Last. Now the intermediate results are out of the range of Long_Long_Integer even though the final result is in range and the precondition is True (from a mathematical point of view). In such a case, operating in this mode, an overflow occurs for the intermediate computation (which is why this mode says `most' intermediate overflows are avoided). In this case, an exception is raised if overflow checks are enabled, and the execution is erroneous if overflow checks are suppressed.
In this mode, the compiler avoids all intermediate overflows by using arbitrary precision arithmetic as required. In this mode, the above example with A**2 * B**2 would not cause intermediate overflow, because the intermediate result would be evaluated using sufficient precision, and the result of evaluating the precondition would be True.
This mode has the advantage of avoiding any intermediate overflows, but at the expense of significant run-time overhead, including the use of a library (included automatically in this mode) for multiple-precision arithmetic.
This mode provides cleaner semantics for assertions, since now the run-time behavior emulates true arithmetic behavior for the predefined arithmetic operators, meaning that there is never a conflict between the mathematical view of the assertion, and its run-time behavior.
Note that in this mode, the behavior is unaffected by whether or not overflow checks are suppressed, since overflow does not occur. It is possible for gigantic intermediate expressions to raise Storage_Error as a result of attempting to compute the results of such expressions (e.g. Integer'Last ** Integer'Last) but overflow is impossible.
Note that these modes apply only to the evaluation of predefined arithmetic, membership, and comparison operators for signed integer aritmetic.
For fixed-point arithmetic, checks can be suppressed. But if checks are enabled then fixed-point values are always checked for overflow against the base type for intermediate expressions (that is such checks always operate in the equivalent of STRICT mode).
For floating-point, on nearly all architectures, Machine_Overflows is False, and IEEE infinities are generated, so overflow exceptions are never raised. If you want to avoid infinities, and check that final results of expressions are in range, then you can declare a constrained floating-point type, and range checks will be carried out in the normal manner (with infinite values always failing all range checks).
The desired mode of for handling intermediate overflow can be specified using either the Overflow_Mode pragma or an equivalent compiler switch. The pragma has the form
pragma Overflow_Mode ([General =>] MODE [, [Assertions =>] MODE]);
where MODE is one of
The case is ignored, so MINIMIZED, Minimized and minimized all have the same effect.
If only the General parameter is present, then the given MODE applies to expressions both within and outside assertions. If both arguments are present, then General applies to expressions outside assertions, and Assertions applies to expressions within assertions. For example:
pragma Overflow_Mode
(General => Minimized, Assertions => Eliminated);
specifies that general expressions outside assertions be evaluated in 'minimize intermediate overflows' mode, and expressions within assertions be evaluated in 'eliminate intermediate overflows' mode. This is often a reasonable choice, avoiding excessive overhead outside assertions, but assuring a high degree of portability when importing code from another compiler, while incurring the extra overhead for assertion expressions to ensure that the behavior at run time matches the expected mathematical behavior.
The Overflow_Mode pragma has the same scoping and placement rules as pragma Suppress, so it can occur either as a configuration pragma, specifying a default for the whole program, or in a declarative scope, where it applies to the remaining declarations and statements in that scope.
Note that pragma Overflow_Mode does not affect whether overflow checks are enabled or suppressed. It only controls the method used to compute intermediate values. To control whether overflow checking is enabled or suppressed, use pragma Suppress or Unsuppress in the usual manner
Additionally, a compiler switch `-gnato?' or `-gnato??' can be used to control the checking mode default (which can be subsequently overridden using pragmas).
Here ? is one of the digits 1 through 3:
1use base type for intermediate operations (STRICT)
2minimize intermediate overflows (MINIMIZED)
3eliminate intermediate overflows (ELIMINATED)
As with the pragma, if only one digit appears then it applies to all cases; if two digits are given, then the first applies outside assertions, and the second within assertions. Thus the equivalent of the example pragma above would be `-gnato23'.
If no digits follow the `-gnato', then it is equivalent to `-gnato11', causing all intermediate operations to be computed using the base type (STRICT mode).
The default mode for overflow checks is
General => Strict
which causes all computations both inside and outside assertions to use the base type.
This retains compatibility with previous versions of GNAT which suppressed overflow checks by default and always used the base type for computation of intermediate results.
The switch `-gnato' (with no digits following) is equivalent to
General => Strict
which causes overflow checking of all intermediate overflows both inside and outside assertions against the base type.
The pragma Suppress (Overflow_Check) disables overflow checking, but it has no effect on the method used for computing intermediate results.
The pragma Unsuppress (Overflow_Check) enables overflow checking, but it has no effect on the method used for computing intermediate results.
In practice on typical 64-bit machines, the MINIMIZED mode is reasonably efficient, and can be generally used. It also helps to ensure compatibility with code imported from some other compiler to GNAT.
Setting all intermediate overflows checking (CHECKED mode) makes sense if you want to make sure that your code is compatible with any other possible Ada implementation. This may be useful in ensuring portability for code that is to be exported to some other compiler than GNAT.
The Ada standard allows the reassociation of expressions at the same precedence level if no parentheses are present. For example, A+B+C parses as though it were (A+B)+C, but the compiler can reintepret this as A+(B+C), possibly introducing or eliminating an overflow exception. The GNAT compiler never takes advantage of this freedom, and the expression A+B+C will be evaluated as (A+B)+C. If you need the other order, you can write the parentheses explicitly A+(B+C) and GNAT will respect this order.
The use of ELIMINATED mode will cause the compiler to automatically include an appropriate arbitrary precision integer arithmetic package. The compiler will make calls to this package, though only in cases where it cannot be sure that Long_Long_Integer is sufficient to guard against intermediate overflows. This package does not use dynamic alllocation, but it does use the secondary stack, so an appropriate secondary stack package must be present (this is always true for standard full Ada, but may require specific steps for restricted run times such as ZFP).
Although ELIMINATED mode causes expressions to use arbitrary precision arithmetic, avoiding overflow, the final result must be in an appropriate range. This is true even if the final result is of type [Long_[Long_]]Integer'Base, which still has the same bounds as its associated constrained type at run-time.
Currently, the ELIMINATED mode is only available on target platforms for which Long_Long_Integer is 64-bits (nearly all GNAT platforms).
The GNAT compiler supports dimensionality checking. The user can specify physical units for objects, and the compiler will verify that uses of these objects are compatible with their dimensions, in a fashion that is familiar to engineering practice. The dimensions of algebraic expressions (including powers with static exponents) are computed from their constituents.
This feature depends on Ada 2012 aspect specifications, and is available from version 7.0.1 of GNAT onwards. The GNAT-specific aspect Dimension_System allows you to define a system of units; the aspect Dimension then allows the user to declare dimensioned quantities within a given system. (These aspects are described in the `Implementation Defined Aspects' chapter of the `GNAT Reference Manual').
The major advantage of this model is that it does not require the declaration of multiple operators for all possible combinations of types: it is only necessary to use the proper subtypes in object declarations.
The simplest way to impose dimensionality checking on a computation is to make
use of the package System.Dim.Mks,
which is part of the GNAT library. This
package defines a floating-point type MKS_Type,
for which a sequence of
dimension names are specified, together with their conventional abbreviations.
The following should be read together with the full specification of the
package, in file s-dimmks.ads.
type Mks_Type is new Long_Long_Float
with
Dimension_System => (
(Unit_Name => Meter, Unit_Symbol => 'm', Dim_Symbol => 'L'),
(Unit_Name => Kilogram, Unit_Symbol => "kg", Dim_Symbol => 'M'),
(Unit_Name => Second, Unit_Symbol => 's', Dim_Symbol => 'T'),
(Unit_Name => Ampere, Unit_Symbol => 'A', Dim_Symbol => 'I'),
(Unit_Name => Kelvin, Unit_Symbol => 'K', Dim_Symbol => "Theta"),
(Unit_Name => Mole, Unit_Symbol => "mol", Dim_Symbol => 'N'),
(Unit_Name => Candela, Unit_Symbol => "cd", Dim_Symbol => 'J'));
The package then defines a series of subtypes that correspond to these conventional units. For example:
subtype Length is Mks_Type
with
Dimension => (Symbol => 'm', Meter => 1, others => 0);
and similarly for Mass, Time, Electric_Current, Thermodynamic_Temperature, Amount_Of_Substance, and Luminous_Intensity (the standard set of units of the SI system).
The package also defines conventional names for values of each unit, for example:
as well as useful multiples of these units:
cm : constant Length := 1.0E-02;
g : constant Mass := 1.0E-03;
min : constant Time := 60.0;
day : constant Time := 60.0 * 24.0 * min;
...
Using this package, you can then define a derived unit by providing the aspect that specifies its dimensions within the MKS system, as well as the string to be used for output of a value of that unit:
subtype Acceleration is Mks_Type
with Dimension => ("m/sec^2",
Meter => 1,
Second => -2,
others => 0);
Here is a complete example of use:
with System.Dim.MKS; use System.Dim.Mks;
with System.Dim.Mks_IO; use System.Dim.Mks_IO;
with Text_IO; use Text_IO;
procedure Free_Fall is
subtype Acceleration is Mks_Type
with Dimension => ("m/sec^2", 1, 0, -2, others => 0);
G : constant acceleration := 9.81 * m / (s ** 2);
T : Time := 10.0*s;
Distance : Length;
begin
Put ("Gravitational constant: ");
Put (G, Aft => 2, Exp => 0); Put_Line ("");
Distance := 0.5 * G * T ** 2;
Put ("distance travelled in 10 seconds of free fall ");
Put (Distance, Aft => 2, Exp => 0);
Put_Line ("");
end Free_Fall;
Execution of this program yields:
Gravitational constant: 9.81 m/sec^2
distance travelled in 10 seconds of free fall 490.50 m
However, incorrect assignments such as:
Distance := 5.0;
Distance := 5.0 * kg:
are rejected with the following diagnoses:
Distance := 5.0;
>>> dimensions mismatch in assignment
>>> left-hand side has dimension [L]
>>> right-hand side is dimensionless
Distance := 5.0 * kg:
>>> dimensions mismatch in assignment
>>> left-hand side has dimension [L]
>>> right-hand side has dimension [M]
The dimensions of an expression are properly displayed, even if there is no explicit subtype for it. If we add to the program:
Put ("Final velocity: ");
Put (G * T, Aft =>2, Exp =>0);
Put_Line ("");
then the output includes:
Final velocity: 98.10 m.s**(-1)
This section describes some useful tools associated with stack checking and analysis. In particular, it deals with dynamic and static stack usage measurements.
For most operating systems, `gcc' does not perform stack overflow checking by default. This means that if the main environment task or some other task exceeds the available stack space, then unpredictable behavior will occur. Most native systems offer some level of protection by adding a guard page at the end of each task stack. This mechanism is usually not enough for dealing properly with stack overflow situations because a large local variable could "jump" above the guard page. Furthermore, when the guard page is hit, there may not be any space left on the stack for executing the exception propagation code. Enabling stack checking avoids such situations.
To activate stack checking, compile all units with the gcc option -fstack-check. For example:
$ gcc -c -fstack-check package1.adb
Units compiled with this option will generate extra instructions to check that any use of the stack (for procedure calls or for declaring local variables in declare blocks) does not exceed the available stack space. If the space is exceeded, then a Storage_Error exception is raised.
For declared tasks, the stack size is controlled by the size
given in an applicable Storage_Size pragma or by the value specified
at bind time with -d (Switches for gnatbind) or is set to
the default size as defined in the GNAT runtime otherwise.
For the environment task, the stack size depends on
system defaults and is unknown to the compiler. Stack checking
may still work correctly if a fixed
size stack is allocated, but this cannot be guaranteed.
To ensure that a clean exception is signalled for stack
overflow, set the environment variable
GNAT_STACK_LIMIT to indicate the maximum
stack area that can be used, as in:
$ SET GNAT_STACK_LIMIT 1600
The limit is given in kilobytes, so the above declaration would set the stack limit of the environment task to 1.6 megabytes. Note that the only purpose of this usage is to limit the amount of stack used by the environment task. If it is necessary to increase the amount of stack for the environment task, then this is an operating systems issue, and must be addressed with the appropriate operating systems commands.
A unit compiled with -fstack-usage will generate an extra file
that specifies
the maximum amount of stack used, on a per-function basis.
The file has the same
basename as the target object file with a .su extension.
Each line of this file is made up of three fields:
The second field corresponds to the size of the known part of the function frame.
The qualifier static means that the function frame size is purely static. It usually means that all local variables have a static size. In this case, the second field is a reliable measure of the function stack utilization.
The qualifier dynamic means that the function frame size is not static. It happens mainly when some local variables have a dynamic size. When this qualifier appears alone, the second field is not a reliable measure of the function stack analysis. When it is qualified with bounded, it means that the second field is a reliable maximum of the function stack utilization.
A unit compiled with -Wstack-usage will issue a warning for each
subprogram whose stack usage might be larger than the specified amount of
bytes. The wording is in keeping with the qualifier documented above.
It is possible to measure the maximum amount of stack used by a task, by adding a switch to `gnatbind', as:
$ gnatbind -u0 file
With this option, at each task termination, its stack usage is output on
stderr.
It is not always convenient to output the stack usage when the program
is still running. Hence, it is possible to delay this output until program
termination. for a given number of tasks specified as the argument of the
-u option. For instance:
$ gnatbind -u100 file
will buffer the stack usage information of the first 100 tasks to terminate and output this info at program termination. Results are displayed in four columns:
Index | Task Name | Stack Size | Stack Usage
where:
The environment task stack, e.g., the stack that contains the main unit, is only processed when the environment variable GNAT_STACK_LIMIT is set.
The package GNAT.Task_Stack_Usage provides facilities to get stack usage reports at run-time. See its body for the details.
This section describes some useful memory pools provided in the GNAT library and in particular the GNAT Debug Pool facility, which can be used to detect incorrect uses of access values (including 'dangling references').
The System.Pool_Global package offers the Unbounded_No_Reclaim_Pool storage pool. Allocations use the standard system call malloc while deallocations use the standard system call free. No reclamation is performed when the pool goes out of scope. For performance reasons, the standard default Ada allocators/deallocators do not use any explicit storage pools but if they did, they could use this storage pool without any change in behavior. That is why this storage pool is used when the user manages to make the default implicit allocator explicit as in this example:
type T1 is access Something;
-- no Storage pool is defined for T2
type T2 is access Something_Else;
for T2'Storage_Pool use T1'Storage_Pool;
-- the above is equivalent to
for T2'Storage_Pool use System.Pool_Global.Global_Pool_Object;
The System.Pool_Local package offers the Unbounded_Reclaim_Pool storage pool. The allocation strategy is similar to Pool_Local's except that the all storage allocated with this pool is reclaimed when the pool object goes out of scope. This pool provides a explicit mechanism similar to the implicit one provided by several Ada 83 compilers for allocations performed through a local access type and whose purpose was to reclaim memory when exiting the scope of a given local access. As an example, the following program does not leak memory even though it does not perform explicit deallocation:
with System.Pool_Local;
procedure Pooloc1 is
procedure Internal is
type A is access Integer;
X : System.Pool_Local.Unbounded_Reclaim_Pool;
for A'Storage_Pool use X;
v : A;
begin
for I in 1 .. 50 loop
v := new Integer;
end loop;
end Internal;
begin
for I in 1 .. 100 loop
Internal;
end loop;
end Pooloc1;
The System.Pool_Size package implements the Stack_Bounded_Pool used when Storage_Size is specified for an access type. The whole storage for the pool is allocated at once, usually on the stack at the point where the access type is elaborated. It is automatically reclaimed when exiting the scope where the access type is defined. This package is not intended to be used directly by the user and it is implicitly used for each such declaration:
type T1 is access Something;
for T1'Storage_Size use 10_000;
The use of unchecked deallocation and unchecked conversion can easily lead to incorrect memory references. The problems generated by such references are usually difficult to tackle because the symptoms can be very remote from the origin of the problem. In such cases, it is very helpful to detect the problem as early as possible. This is the purpose of the Storage Pool provided by GNAT.Debug_Pools.
In order to use the GNAT specific debugging pool, the user must associate a debug pool object with each of the access types that may be related to suspected memory problems. See Ada Reference Manual 13.11.
type Ptr is access Some_Type;
Pool : GNAT.Debug_Pools.Debug_Pool;
for Ptr'Storage_Pool use Pool;
GNAT.Debug_Pools is derived from a GNAT-specific kind of pool: the Checked_Pool. Such pools, like standard Ada storage pools, allow the user to redefine allocation and deallocation strategies. They also provide a checkpoint for each dereference, through the use of the primitive operation Dereference which is implicitly called at each dereference of an access value.
Once an access type has been associated with a debug pool, operations on values of the type may raise four distinct exceptions, which correspond to four potential kinds of memory corruption:
For types associated with a Debug_Pool, dynamic allocation is performed using the standard GNAT allocation routine. References to all allocated chunks of memory are kept in an internal dictionary. Several deallocation strategies are provided, whereupon the user can choose to release the memory to the system, keep it allocated for further invalid access checks, or fill it with an easily recognizable pattern for debug sessions. The memory pattern is the old IBM hexadecimal convention: 16#DEADBEEF#.
See the documentation in the file g-debpoo.ads for more information on the various strategies.
Upon each dereference, a check is made that the access value denotes a properly allocated memory location. Here is a complete example of use of Debug_Pools, that includes typical instances of memory corruption:
with Gnat.Io; use Gnat.Io;
with Unchecked_Deallocation;
with Unchecked_Conversion;
with GNAT.Debug_Pools;
with System.Storage_Elements;
with Ada.Exceptions; use Ada.Exceptions;
procedure Debug_Pool_Test is
type T is access Integer;
type U is access all T;
P : GNAT.Debug_Pools.Debug_Pool;
for T'Storage_Pool use P;
procedure Free is new Unchecked_Deallocation (Integer, T);
function UC is new Unchecked_Conversion (U, T);
A, B : aliased T;
procedure Info is new GNAT.Debug_Pools.Print_Info(Put_Line);
begin
Info (P);
A := new Integer;
B := new Integer;
B := A;
Info (P);
Free (A);
begin
Put_Line (Integer'Image(B.all));
exception
when E : others => Put_Line ("raised: " & Exception_Name (E));
end;
begin
Free (B);
exception
when E : others => Put_Line ("raised: " & Exception_Name (E));
end;
B := UC(A'Access);
begin
Put_Line (Integer'Image(B.all));
exception
when E : others => Put_Line ("raised: " & Exception_Name (E));
end;
begin
Free (B);
exception
when E : others => Put_Line ("raised: " & Exception_Name (E));
end;
Info (P);
end Debug_Pool_Test;
The debug pool mechanism provides the following precise diagnostics on the execution of this erroneous program:
Debug Pool info:
Total allocated bytes : 0
Total deallocated bytes : 0
Current Water Mark: 0
High Water Mark: 0
Debug Pool info:
Total allocated bytes : 8
Total deallocated bytes : 0
Current Water Mark: 8
High Water Mark: 8
raised: GNAT.DEBUG_POOLS.ACCESSING_DEALLOCATED_STORAGE
raised: GNAT.DEBUG_POOLS.FREEING_DEALLOCATED_STORAGE
raised: GNAT.DEBUG_POOLS.ACCESSING_NOT_ALLOCATED_STORAGE
raised: GNAT.DEBUG_POOLS.FREEING_NOT_ALLOCATED_STORAGE
Debug Pool info:
Total allocated bytes : 8
Total deallocated bytes : 4
Current Water Mark: 4
High Water Mark: 8
This appendix contains information relating to the implementation of run-time libraries on various platforms and also covers topics related to the GNAT implementation on Windows and Mac OS.
The GNAT run-time implementation may vary with respect to both the underlying threads library and the exception handling scheme. For threads support, one or more of the following are supplied:
For exception handling, either or both of two models are supplied:
Most programs should experience a substantial speed improvement by being compiled with a ZCX run-time. This is especially true for tasking applications or applications with many exception handlers.}
This section summarizes which combinations of threads and exception support are supplied on various GNAT platforms. It then shows how to select a particular library either permanently or temporarily, explains the properties of (and tradeoffs among) the various threads libraries, and provides some additional information about several specific platforms.
| Platform |
Run-Time |
Tasking |
Exceptions |
|---|---|---|---|
| ppc-aix |
rts-native (default) |
native AIX threads |
ZCX |
| rts-sjlj |
native AIX threads |
SJLJ | |
| sparc-solaris |
rts-native (default) |
native Solaris threads library |
ZCX |
| rts-pthread |
pthread library |
ZCX | |
| rts-sjlj |
native Solaris threads library |
SJLJ | |
| sparc64-solaris |
rts-native (default) |
native Solaris threads library |
ZCX |
| x86-linux |
rts-native (default) |
pthread library |
ZCX |
| rts-sjlj |
pthread library |
SJLJ | |
| x86-lynx |
rts-native (default) |
native LynxOS threads |
SJLJ |
| x86-solaris |
rts-native (default) |
native Solaris threads library |
ZCX |
| rts-sjlj |
native Solaris threads library |
SJLJ | |
| x86-windows |
rts-native (default) |
native Win32 threads |
ZCX |
| rts-sjlj |
native Win32 threads |
SJLJ | |
| x86_64-linux |
rts-native (default) |
pthread library |
ZCX |
| rts-sjlj |
pthread library |
SJLJ
|
The adainclude subdirectory containing the sources of the GNAT
run-time library, and the adalib subdirectory containing the
ALI files and the static and/or shared GNAT library, are located
in the gcc target-dependent area:
target=$prefix/lib/gcc/gcc-*dumpmachine*/gcc-*dumpversion*/
As indicated above, on some platforms several run-time libraries are supplied. These libraries are installed in the target dependent area and contain a complete source and binary subdirectory. The detailed description below explains the differences between the different libraries in terms of their thread support.
The default run-time library (when GNAT is installed) is `rts-native'. This default run time is selected by the means of soft links. For example on x86-linux:
--
-- $(target-dir)
-- |
-- +--- adainclude----------+
-- | |
-- +--- adalib-----------+ |
-- | | |
-- +--- rts-native | |
-- | | | |
-- | +--- adainclude <---+
-- | | |
-- | +--- adalib <----+
-- |
-- +--- rts-sjlj
-- |
-- +--- adainclude
-- |
-- +--- adalib
If the `rts-sjlj' library is to be selected on a permanent basis, these soft links can be modified with the following commands:
$ cd $target
$ rm -f adainclude adalib
$ ln -s rts-sjlj/adainclude adainclude
$ ln -s rts-sjlj/adalib adalib
Alternatively, you can specify rts-sjlj/adainclude in the file
$target/ada_source_path and rts-sjlj/adalib in
$target/ada_object_path.
Selecting another run-time library temporarily can be achieved by using the `–RTS' switch, e.g., `–RTS=sjlj'
When using a POSIX threads implementation, you have a choice of several scheduling policies: SCHED_FIFO, SCHED_RR and SCHED_OTHER.
Typically, the default is SCHED_OTHER, while using SCHED_FIFO or SCHED_RR requires special (e.g., root) privileges.
By default, GNAT uses the SCHED_OTHER policy. To specify SCHED_FIFO, you can use one of the following:
To specify SCHED_RR, you should use pragma Time_Slice with a value greater than 0.0, or else use the corresponding `-T' binder option.
This section addresses some topics related to the various threads libraries on Sparc Solaris.
GNAT under Solaris/Sparc 32 bits comes with an alternate tasking run-time library based on POSIX threads — `rts-pthread'.
This run-time library has the advantage of being mostly shared across all POSIX-compliant thread implementations, and it also provides under Solaris 8 the PTHREAD_PRIO_INHERIT and PTHREAD_PRIO_PROTECT semantics that can be selected using the predefined pragma Locking_Policy with respectively Inheritance_Locking and Ceiling_Locking as the policy.
As explained above, the native run-time library is based on the Solaris thread library (libthread) and is the default library.
When the Solaris threads library is used (this is the default), programs
compiled with GNAT can automatically take advantage of
and can thus execute on multiple processors.
The user can alternatively specify a processor on which the program should run
to emulate a single-processor system. The multiprocessor / uniprocessor choice
is made by
setting the environment variable
GNAT_PROCESSOR
to one of the following:
GNAT_PROCESSORValueEffect
`-2' Use the default configuration (run the program on all available processors) - this is the same as having GNAT_PROCESSOR unset
`-1' Let the run-time implementation choose one processor and run the program on that processor
`0 .. Last_Proc' Run the program on the specified processor. Last_Proc is equal to _SC_NPROCESSORS_CONF - 1 (where _SC_NPROCESSORS_CONF is a system variable).
On AIX, the resolver library initializes some internal structure on the first call to get*by* functions, which are used to implement GNAT.Sockets.Get_Host_By_Name and GNAT.Sockets.Get_Host_By_Address. If such initialization occurs within an Ada task, and the stack size for the task is the default size, a stack overflow may occur.
To avoid this overflow, the user should either ensure that the first call to GNAT.Sockets.Get_Host_By_Name or GNAT.Sockets.Get_Host_By_Addrss occurs in the environment task, or use pragma Storage_Size to specify a sufficiently large size for the stack of the task that contains this call.
This section describes topics that are specific to the Microsoft Windows platforms.
One of the strengths of the GNAT technology is that its tool set (`gcc', `gnatbind', `gnatlink', `gnatmake', the gdb debugger, etc.) is used in the same way regardless of the platform.
On Windows this tool set is complemented by a number of Microsoft-specific tools that have been provided to facilitate interoperability with Windows when this is required. With these tools:
Immediately below are listed all known general GNAT-for-Windows restrictions. Other restrictions about specific features like Windows Resources and DLLs are listed in separate sections below.
Make sure the system on which GNAT is installed is accessible from the current machine, i.e., the install location is shared over the network. Shared resources are accessed on Windows by means of UNC paths, which have the format \\server\sharename\path
In order to use such a network installation, simply add the UNC path of the
bin directory of your GNAT installation in front of your PATH. For
example, if GNAT is installed in \GNAT directory of a share location
called c-drive on a machine LOKI, the following command will
make it available:
$ path \\loki\c-drive\gnat\bin;%path%`
Be aware that every compilation using the network installation results in the transfer of large amounts of data across the network and will likely cause serious performance penalty.
There are two main subsystems under Windows. The CONSOLE subsystem (which is the default subsystem) will always create a console when launching the application. This is not something desirable when the application has a Windows GUI. To get rid of this console the application must be using the WINDOWS subsystem. To do so the `-mwindows' linker option must be specified.
$ gnatmake winprog -largs -mwindows
It is possible to control where temporary files gets created by setting
the
TMP environment variable. The file will be created:
TMP environment variable if
this directory exists.
c:\temp, if the
TMP environment variable is not
set (or not pointing to a directory) and if this directory exists.
This allows you to determine exactly where the temporary file will be created. This is particularly useful in networked environments where you may not have write access to some directories.
Developing pure Ada applications on Windows is no different than on other GNAT-supported platforms. However, when developing or porting an application that contains a mix of Ada and C/C++, the choice of your Windows C/C++ development environment conditions your overall interoperability strategy.
If you use `gcc' or Microsoft C to compile the non-Ada part of your application, there are no Windows-specific restrictions that affect the overall interoperability with your Ada code. If you do want to use the Microsoft tools for your C++ code, you have two choices:
In addition to the description about C main in Mixed Language Programming section, if the C main uses a stand-alone library it is required on x86-windows to setup the SEH context. For this the C main must looks like this:
/* main.c */
extern void adainit (void);
extern void adafinal (void);
extern void __gnat_initialize(void*);
extern void call_to_ada (void);
int main (int argc, char *argv[])
{
int SEH [2];
/* Initialize the SEH context */
__gnat_initialize (&SEH);
adainit();
/* Then call Ada services in the stand-alone library */
call_to_ada();
adafinal();
}
Note that this is not needed on x86_64-windows where the Windows native SEH support is used.
This section pertain only to Win32. On Win64 there is a single native calling convention. All convention specifiers are ignored on this platform.
When a subprogram F (caller) calls a subprogram G (callee), there are several ways to push G's parameters on the stack and there are several possible scenarios to clean up the stack upon G's return. A calling convention is an agreed upon software protocol whereby the responsibilities between the caller (F) and the callee (G) are clearly defined. Several calling conventions are available for Windows:
This is the default calling convention used when interfacing to C/C++ routines compiled with either `gcc' or Microsoft Visual C++.
In the C calling convention subprogram parameters are pushed on the stack by the caller from right to left. The caller itself is in charge of cleaning up the stack after the call. In addition, the name of a routine with C calling convention is mangled by adding a leading underscore.
The name to use on the Ada side when importing (or exporting) a routine with C calling convention is the name of the routine. For instance the C function:
int get_val (long);
should be imported from Ada as follows:
function Get_Val (V : Interfaces.C.long) return Interfaces.C.int;
pragma Import (C, Get_Val, External_Name => "get_val");
Note that in this particular case the External_Name parameter could have been omitted since, when missing, this parameter is taken to be the name of the Ada entity in lower case. When the Link_Name parameter is missing, as in the above example, this parameter is set to be the External_Name with a leading underscore.
When importing a variable defined in C, you should always use the C calling convention unless the object containing the variable is part of a DLL (in which case you should use the Stdcall calling convention, Stdcall Calling Convention).
This convention, which was the calling convention used for Pascal programs, is used by Microsoft for all the routines in the Win32 API for efficiency reasons. It must be used to import any routine for which this convention was specified.
In the Stdcall calling convention subprogram parameters are pushed
on the stack by the caller from right to left. The callee (and not the
caller) is in charge of cleaning the stack on routine exit. In addition,
the name of a routine with Stdcall calling convention is mangled by
adding a leading underscore (as for the C calling convention) and a
trailing @`nn', where nn is the overall size (in
bytes) of the parameters passed to the routine.
The name to use on the Ada side when importing a C routine with a
Stdcall calling convention is the name of the C routine. The leading
underscore and trailing @`nn' are added automatically by
the compiler. For instance the Win32 function:
APIENTRY int get_val (long);
should be imported from Ada as follows:
function Get_Val (V : Interfaces.C.long) return Interfaces.C.int;
pragma Import (Stdcall, Get_Val);
-- On the x86 a long is 4 bytes, so the Link_Name is "_get_val@4"
As for the C calling convention, when the External_Name parameter is missing, it is taken to be the name of the Ada entity in lower case. If instead of writing the above import pragma you write:
function Get_Val (V : Interfaces.C.long) return Interfaces.C.int;
pragma Import (Stdcall, Get_Val, External_Name => "retrieve_val");
then the imported routine is _retrieve_val@4. However, if instead of specifying the External_Name parameter you specify the Link_Name as in the following example:
function Get_Val (V : Interfaces.C.long) return Interfaces.C.int;
pragma Import (Stdcall, Get_Val, Link_Name => "retrieve_val");
then the imported routine is retrieve_val, that is, there is no
decoration at all. No leading underscore and no Stdcall suffix
@`nn'.
This is especially important as in some special cases a DLL's entry
point name lacks a trailing @`nn' while the exported
name generated for a call has it.
It is also possible to import variables defined in a DLL by using an import pragma for a variable. As an example, if a DLL contains a variable defined as:
int my_var;
then, to access this variable from Ada you should write:
My_Var : Interfaces.C.int;
pragma Import (Stdcall, My_Var);
Note that to ease building cross-platform bindings this convention will be handled as a C calling convention on non-Windows platforms.
This convention, which is GNAT-specific is fully equivalent to the Stdcall calling convention described above.
This convention, which is GNAT-specific is fully equivalent to the Stdcall calling convention described above.
A Dynamically Linked Library (DLL) is a library that can be shared by several applications running under Windows. A DLL can contain any number of routines and variables.
One advantage of DLLs is that you can change and enhance them without forcing all the applications that depend on them to be relinked or recompiled. However, you should be aware than all calls to DLL routines are slower since, as you will understand below, such calls are indirect.
To illustrate the remainder of this section, suppose that an application
wants to use the services of a DLL API.dll. To use the services
provided by API.dll you must statically link against the DLL or
an import library which contains a jump table with an entry for each
routine and variable exported by the DLL. In the Microsoft world this
import library is called API.lib. When using GNAT this import
library is called either libAPI.dll.a, libapi.dll.a,
libAPI.a or libapi.a (names are case insensitive).
After you have linked your application with the DLL or the import library and you run your application, here is what happens:
API.dll is mapped into the address space of your
application. This means that:
libAPI.dll.a
or API.lib or automatically created when linking against a DLL)
which is part of your application are initialized with the addresses
of the routines and variables in API.dll.
API.dll, routines DllMain or
DllMainCRTStartup are invoked. These routines typically contain
the initialization code needed for the well-being of the routines and
variables exported by the DLL.
There is an additional point which is worth mentioning. In the Windows world there are two kind of DLLs: relocatable and non-relocatable DLLs. Non-relocatable DLLs can only be loaded at a very specific address in the target application address space. If the addresses of two non-relocatable DLLs overlap and these happen to be used by the same application, a conflict will occur and the application will run incorrectly. Hence, when possible, it is always preferable to use and build relocatable DLLs. Both relocatable and non-relocatable DLLs are supported by GNAT. Note that the `-s' linker option (see GNU Linker User's Guide) removes the debugging symbols from the DLL but the DLL can still be relocated.
As a side note, an interesting difference between Microsoft DLLs and Unix shared libraries, is the fact that on most Unix systems all public routines are exported by default in a Unix shared library, while under Windows it is possible (but not required) to list exported routines in a definition file (see The Definition File).
To use the services of a DLL, say API.dll, in your Ada application
you must have:
API.dll. If not available this Ada spec must be built from the C/C++
header files provided with the DLL.
libAPI.dll.a or API.lib). As previously
mentioned an import library is a statically linked library containing the
import table which will be filled at load time to point to the actual
API.dll routines. Sometimes you don't have an import library for the
DLL you want to use. The following sections will explain how to build
one. Note that this is optional.
API.dll.
Once you have all the above, to compile an Ada application that uses the
services of API.dll and whose main subprogram is My_Ada_App,
you simply issue the command
$ gnatmake my_ada_app -largs -lAPI
The argument `-largs -lAPI' at the end of the `gnatmake' command tells the GNAT linker to look for an import library. The linker will look for a library name in this specific order:
libAPI.dll.a
API.dll.a
libAPI.a
API.lib
libAPI.dll
API.dll
The first three are the GNU style import libraries. The third is the Microsoft style import libraries. The last two are the actual DLL names.
Note that if the Ada package spec for API.dll contains the
following pragma
pragma Linker_Options ("-lAPI");
you do not have to add `-largs -lAPI' at the end of the `gnatmake' command.
If any one of the items above is missing you will have to create it
yourself. The following sections explain how to do so using as an
example a fictitious DLL called API.dll.
A DLL typically comes with a C/C++ header file which provides the
definitions of the routines and variables exported by the DLL. The Ada
equivalent of this header file is a package spec that contains definitions
for the imported entities. If the DLL you intend to use does not come with
an Ada spec you have to generate one such spec yourself. For example if
the header file of API.dll is a file api.h containing the
following two definitions:
int some_var;
int get (char *);
then the equivalent Ada spec could be:
with Interfaces.C.Strings;
package API is
use Interfaces;
Some_Var : C.int;
function Get (Str : C.Strings.Chars_Ptr) return C.int;
private
pragma Import (C, Get);
pragma Import (DLL, Some_Var);
end API;
If a Microsoft-style import library API.lib or a GNAT-style
import library libAPI.dll.a or libAPI.a is available
with API.dll you can skip this section. You can also skip this
section if API.dll or libAPI.dll is built with GNU tools
as in this case it is possible to link directly against the
DLL. Otherwise read on.
As previously mentioned, and unlike Unix systems, the list of symbols that are exported from a DLL must be provided explicitly in Windows. The main goal of a definition file is precisely that: list the symbols exported by a DLL. A definition file (usually a file with a .def suffix) has the following structure:
[LIBRARY `name`]
[DESCRIPTION `string`]
EXPORTS
`symbol1`
`symbol2`
...
API.dll the EXPORTS
section of API.def looks like:
EXPORTS
some_var
get
Note that you must specify the correct suffix (@`nn')
(see Windows Calling Conventions) for a Stdcall
calling convention function in the exported symbols list.
There can actually be other sections in a definition file, but these sections are not relevant to the discussion at hand.
You can automatically create the definition file API.def
(see The Definition File) from a DLL.
For that use the dlltool program as follows:
$ dlltool API.dll -z API.def --export-all-symbolsNote that if some routines in the DLL have the Stdcall convention (Windows Calling Conventions) with stripped
@`nn'suffix then you'll have to editapi.defto add it, and specify `-k' to `gnatdll' when creating the import library.Here are some hints to find the right
@`nn'suffix.
- If you have the Microsoft import library (.lib), it is possible to get the right symbols by using Microsoft dumpbin tool (see the corresponding Microsoft documentation for further details).
$ dumpbin /exports api.lib- If you have a message about a missing symbol at link time the compiler tells you what symbol is expected. You just have to go back to the definition file and add the right suffix.
To create a static import library from API.dll with the GNAT tools
you should create the .def file, then use gnatdll tool
(see Using gnatdll) as follows:
$ gnatdll -e API.def -d API.dllgnatdll takes as input a definition file
API.defand the name of the DLL containing the services listed in the definition fileAPI.dll. The name of the static import library generated is computed from the name of the definition file as follows: if the definition file name is xyz`.def`, the import library name will be lib``xyz`.a`. Note that in the previous example option `-e' could have been removed because the name of the definition file (before the '.def' suffix) is the same as the name of the DLL (Using gnatdll for more information about gnatdll).
A Microsoft import library is needed only if you plan to make an Ada DLL available to applications developed with Microsoft tools (Mixed-Language Programming on Windows).
To create a Microsoft-style import library for API.dll you
should create the .def file, then build the actual import library using
Microsoft's lib utility:
$ lib -machine:IX86 -def:API.def -out:API.libIf you use the above command the definition file
API.defmust contain a line giving the name of the DLL:LIBRARY "API"See the Microsoft documentation for further details about the usage of lib.
There is nothing specific to Windows in the build process. Library Projects.
Due to a system limitation, it is not possible under Windows to create threads when inside the DllMain routine which is used for auto-initialization of shared libraries, so it is not possible to have library level tasks in SALs.
This section explain how to build DLLs using the GNAT built-in DLL support. With the following procedure it is straight forward to build and use DLLs with GNAT.
$ gcc -shared -shared-libgcc -o api.dll obj1.o obj2.o ...
It is important to note that in this case all symbols found in the object files are automatically exported. It is possible to restrict the set of symbols to export by passing to `gcc' a definition file (see The Definition File). For example:
$ gcc -shared -shared-libgcc -o api.dll api.def obj1.o obj2.o ...
If you use a definition file you must export the elaboration procedures for every package that required one. Elaboration procedures are named using the package name followed by "_E".
$ mkdir apilib
$ copy *.ads *.ali api.dll apilib
$ attrib +R apilib\\*.ali
At this point it is possible to use the DLL by directly linking against it. Note that you must use the GNAT shared runtime when using GNAT shared libraries. This is achieved by using `-shared' binder's option.
$ gnatmake main -Iapilib -bargs -shared -largs -Lapilib -lAPI
Note that it is preferred to use GNAT Project files (Building DLLs with GNAT Project files) or the built-in GNAT DLL support (Building DLLs with GNAT) or to build DLLs.
This section explains how to build DLLs containing Ada code using gnatdll. These DLLs will be referred to as Ada DLLs in the remainder of this section.
The steps required to build an Ada DLL that is to be used by Ada as well as non-Ada applications are as follows:
Note that a relocatable DLL stripped using the strip binutils tool will not be relocatable anymore. To build a DLL without debug information pass -largs -s to gnatdll. This restriction does not apply to a DLL built using a Library Project. See Library Projects.
When using Ada DLLs from Ada applications there is a limitation users should be aware of. Because on Windows the GNAT run time is not in a DLL of its own, each Ada DLL includes a part of the GNAT run time. Specifically, each Ada DLL includes the services of the GNAT run time that are necessary to the Ada code inside the DLL. As a result, when an Ada program uses an Ada DLL there are two independent GNAT run times: one in the Ada DLL and one in the main program.
It is therefore not possible to exchange GNAT run-time objects between the Ada DLL and the main Ada program. Example of GNAT run-time objects are file handles (e.g., Text_IO.File_Type), tasks types, protected objects types, etc.
It is completely safe to exchange plain elementary, array or record types, Windows object handles, etc.
Building a DLL is a way to encapsulate a set of services usable from any application. As a result, the Ada entities exported by a DLL should be exported with the C or Stdcall calling conventions to avoid any Ada name mangling. As an example here is an Ada package API, spec and body, exporting two procedures, a function, and a variable:
with Interfaces.C; use Interfaces; package API is Count : C.int := 0; function Factorial (Val : C.int) return C.int; procedure Initialize_API; procedure Finalize_API; -- Initialization & Finalization routines. More in the next section. private pragma Export (C, Initialize_API); pragma Export (C, Finalize_API); pragma Export (C, Count); pragma Export (C, Factorial); end API;package body API is function Factorial (Val : C.int) return C.int is Fact : C.int := 1; begin Count := Count + 1; for K in 1 .. Val loop Fact := Fact * K; end loop; return Fact; end Factorial; procedure Initialize_API is procedure Adainit; pragma Import (C, Adainit); begin Adainit; end Initialize_API; procedure Finalize_API is procedure Adafinal; pragma Import (C, Adafinal); begin Adafinal; end Finalize_API; end API;
If the Ada DLL you are building will only be used by Ada applications you do not have to export Ada entities with a C or Stdcall convention. As an example, the previous package could be written as follows:
package API is Count : Integer := 0; function Factorial (Val : Integer) return Integer; procedure Initialize_API; procedure Finalize_API; -- Initialization and Finalization routines. end API;package body API is function Factorial (Val : Integer) return Integer is Fact : Integer := 1; begin Count := Count + 1; for K in 1 .. Val loop Fact := Fact * K; end loop; return Fact; end Factorial; ... -- The remainder of this package body is unchanged. end API;
Note that if you do not export the Ada entities with a C or Stdcall convention you will have to provide the mangled Ada names in the definition file of the Ada DLL (Creating the Definition File).
The DLL that you are building contains your Ada code as well as all the routines in the Ada library that are needed by it. The first thing a user of your DLL must do is elaborate the Ada code (Elaboration Order Handling in GNAT).
To achieve this you must export an initialization routine (Initialize_API in the previous example), which must be invoked before using any of the DLL services. This elaboration routine must call the Ada elaboration routine adainit generated by the GNAT binder (Binding with Non-Ada Main Programs). See the body of Initialize_Api for an example. Note that the GNAT binder is automatically invoked during the DLL build process by the gnatdll tool (Using gnatdll).
When a DLL is loaded, Windows systematically invokes a routine called DllMain. It would therefore be possible to call adainit directly from DllMain without having to provide an explicit initialization routine. Unfortunately, it is not possible to call adainit from the DllMain if your program has library level tasks because access to the DllMain entry point is serialized by the system (that is, only a single thread can execute 'through' it at a time), which means that the GNAT run time will deadlock waiting for the newly created task to complete its initialization.
When the services of an Ada DLL are no longer needed, the client code should invoke the DLL finalization routine, if available. The DLL finalization routine is in charge of releasing all resources acquired by the DLL. In the case of the Ada code contained in the DLL, this is achieved by calling routine adafinal generated by the GNAT binder (Binding with Non-Ada Main Programs). See the body of Finalize_Api for an example. As already pointed out the GNAT binder is automatically invoked during the DLL build process by the gnatdll tool (Using gnatdll).
To use the services exported by the Ada DLL from another programming language (e.g., C), you have to translate the specs of the exported Ada entities in that language. For instance in the case of API.dll, the corresponding C header file could look like:
extern int *_imp__count;
#define count (*_imp__count)
int factorial (int);
It is important to understand that when building an Ada DLL to be used by other Ada applications, you need two different specs for the packages contained in the DLL: one for building the DLL and the other for using the DLL. This is because the DLL calling convention is needed to use a variable defined in a DLL, but when building the DLL, the variable must have either the Ada or C calling convention. As an example consider a DLL comprising the following package API:
package API is
Count : Integer := 0;
...
-- Remainder of the package omitted.
end API;
After producing a DLL containing package API, the spec that must be used to import API.Count from Ada code outside of the DLL is:
package API is
Count : Integer;
pragma Import (DLL, Count);
end API;
The definition file is the last file needed to build the DLL. It lists the exported symbols. As an example, the definition file for a DLL containing only package API (where all the entities are exported with a C calling convention) is:
EXPORTS
count
factorial
finalize_api
initialize_api
If the C calling convention is missing from package API, then the definition file contains the mangled Ada names of the above entities, which in this case are:
EXPORTS
api__count
api__factorial
api__finalize_api
api__initialize_api
gnatdll is a tool to automate the DLL build process once all the Ada and non-Ada sources that make up your DLL have been compiled. gnatdll is actually in charge of two distinct tasks: build the static import library for the DLL and the actual DLL. The form of the gnatdll command is
$ gnatdll [`switches`] `list-of-files` [-largs `opts`]
where list-of-files is a list of ALI and object files. The object file list must be the exact list of objects corresponding to the non-Ada sources whose services are to be included in the DLL. The ALI file list must be the exact list of ALI files for the corresponding Ada sources whose services are to be included in the DLL. If list-of-files is missing, only the static import library is generated.
You may specify any of the following switches to gnatdll:
-a[`address']-b `address'-bargs `opts'-d `dllfile'-e `deffile'-g-h-I`dir'-k@`nn' suffix from the import library's exported
names, but keeps them for the link names. You must specify this
option if you want to use a Stdcall function in a DLL for which
the @`nn' suffix has been removed. This is the case for most
of the Windows NT DLL for example. This option has no effect when
`-n' option is specified.
-l `file'-n-q-v-largs `opts'As an example the command to build a relocatable DLL from api.adb
once api.adb has been compiled and api.def created is
$ gnatdll -d api.dll api.ali
The above command creates two files: libapi.dll.a (the import
library) and api.dll (the actual DLL). If you want to create
only the DLL, just type:
$ gnatdll -d api.dll -n api.ali
Alternatively if you want to create just the import library, type:
$ gnatdll -d api.dll
This section details the steps involved in creating a DLL. gnatdll does these steps for you. Unless you are interested in understanding what goes on behind the scenes, you should skip this section.
We use the previous example of a DLL containing the Ada package API,
to illustrate the steps necessary to build a DLL. The starting point is a
set of objects that will make up the DLL and the corresponding ALI
files. In the case of this example this means that api.o and
api.ali are available. To build a relocatable DLL, gnatdll does
the following:
api.base). A base file gives
the information necessary to generate relocation information for the
DLL.
$ gnatbind -n api
$ gnatlink api -o api.jnk -mdll -Wl,--base-file,api.base
In addition to the base file, the `gnatlink' command generates an
output file api.jnk which can be discarded. The `-mdll' switch
asks `gnatlink' to generate the routines DllMain and
DllMainCRTStartup that are called by the Windows loader when the DLL
is loaded into memory.
api.exp). The export table contains the relocation
information in a form which can be used during the final link to ensure
that the Windows loader is able to place the DLL anywhere in memory.
$ dlltool --dllname api.dll --def api.def --base-file api.base \\
--output-exp api.exp
$ gnatbind -n api
$ gnatlink api -o api.jnk api.exp -mdll
-Wl,--base-file,api.base
libAPI.dll.a.
$ dlltool --dllname api.dll --def api.def --base-file api.base \\
--output-exp api.exp --output-lib libAPI.a
$ gnatbind -n api
$ gnatlink api api.exp -o api.dll -mdll
dlltool is the low-level tool used by gnatdll to build DLLs and static import libraries. This section summarizes the most common dlltool switches. The form of the dlltool command is
$ dlltool [`switches`]
dlltool switches include:
--base-file `basefile'--def `deffile'--dllname `name'-k@`nn' from exported names
(Windows Calling Conventions
for a discussion about Stdcall-style symbols.
--help--output-exp `exportfile'--output-lib `libfile'-v--as `assembler-name'Resources are an easy way to add Windows specific objects to your application. The objects that can be added as resources include:
For example, a version information resource can be defined as follow and embedded into an executable or DLL:
A version information resource can be used to embed information into an executable or a DLL. These information can be viewed using the file properties from the Windows Explorer. Here is an example of a version information resource:
1 VERSIONINFO
FILEVERSION 1,0,0,0
PRODUCTVERSION 1,0,0,0
BEGIN
BLOCK "StringFileInfo"
BEGIN
BLOCK "080904E4"
BEGIN
VALUE "CompanyName", "My Company Name"
VALUE "FileDescription", "My application"
VALUE "FileVersion", "1.0"
VALUE "InternalName", "my_app"
VALUE "LegalCopyright", "My Name"
VALUE "OriginalFilename", "my_app.exe"
VALUE "ProductName", "My App"
VALUE "ProductVersion", "1.0"
END
END
BLOCK "VarFileInfo"
BEGIN
VALUE "Translation", 0x809, 1252
END
END
The value 0809 (langID) is for the U.K English language and 04E4 (charsetID), which is equal to 1252 decimal, for multilingual.
This section explains how to build, compile and use resources. Note that this section does not cover all resource objects, for a complete description see the corresponding Microsoft documentation.
A resource file is an ASCII file. By convention resource files have an
.rc extension.
The easiest way to build a resource file is to use Microsoft tools
such as imagedit.exe to build bitmaps, icons and cursors and
dlgedit.exe to build dialogs.
It is always possible to build an .rc file yourself by writing a
resource script.
It is not our objective to explain how to write a resource file. A complete description of the resource script language can be found in the Microsoft documentation.
This section describes how to build a GNAT-compatible (COFF) object file containing the resources. This is done using the Resource Compiler windres as follows:
$ windres -i myres.rc -o myres.o
By default windres will run `gcc' to preprocess the .rc
file. You can specify an alternate preprocessor (usually named
cpp.exe) using the windres `–preprocessor'
parameter. A list of all possible options may be obtained by entering
the command windres `–help'.
It is also possible to use the Microsoft resource compiler rc.exe
to produce a .res file (binary resource file). See the
corresponding Microsoft documentation for further details. In this case
you need to use windres to translate the .res file to a
GNAT-compatible object file as follows:
$ windres -i myres.res -o myres.o
To include the resource file in your program just add the GNAT-compatible object file for the resource(s) to the linker arguments. With `gnatmake' this is done by using the `-largs' option:
$ gnatmake myprog -largs myres.o
This section describes a common case of mixed GNAT/Microsoft Visual Studio application development, where the main program is developed using MSVS, and is linked with a DLL developed using GNAT. Such a mixed application should be developed following the general guidelines outlined above; below is the cookbook-style sequence of steps to follow:
$ gprbuild -p mylib.gpr
$ dlltool libmylib.dll -z libmylib.def --export-all-symbols
$ lib -machine:IX86 -def:libmylib.def -out:libmylib.lib
If you are using a 64-bit toolchain, the above becomes...
$ lib -machine:X64 -def:libmylib.def -out:libmylib.lib
$ cl /O2 /MD main.c libmylib.lib
Debugging a DLL is similar to debugging a standard program. But we have to deal with two different executable parts: the DLL and the program that uses it. We have the following four possibilities:
In this section we address only cases one and two above. There is no point in trying to debug a DLL with GNU/GDB, if there is no GDB-compatible debugging information in it. To do so you must use a debugger compatible with the tools suite used to build the DLL.
This is the simplest case. Both the DLL and the program have GDB compatible debugging information. It is then possible to break anywhere in the process. Let's suppose here that the main procedure is named ada_main and that in the DLL there is an entry point named ada_dll.
The DLL (Introduction to Dynamic Link Libraries (DLLs)) and program must have been built with the debugging information (see GNAT -g switch). Here are the step-by-step instructions for debugging it:
$ gdb -nw ada_main
(gdb) start
This step is required to be able to set a breakpoint inside the DLL. As long as the program is not run, the DLL is not loaded. This has the consequence that the DLL debugging information is also not loaded, so it is not possible to set a breakpoint in the DLL.
(gdb) break ada_dll
(gdb) cont
At this stage a breakpoint is set inside the DLL. From there on you can use the standard approach to debug the whole program (Running and Debugging Ada Programs).
In this case things are slightly more complex because it is not possible to start the main program and then break at the beginning to load the DLL and the associated DLL debugging information. It is not possible to break at the beginning of the program because there is no GDB debugging information, and therefore there is no direct way of getting initial control. This section addresses this issue by describing some methods that can be used to break somewhere in the DLL to debug it.
First suppose that the main procedure is named main (this is for example some C code built with Microsoft Visual C) and that there is a DLL named test.dll containing an Ada entry point named ada_dll.
The DLL (see Introduction to Dynamic Link Libraries (DLLs)) must have been built with debugging information (see GNAT -g option).
$ objdump --file-header main.exe
The starting address is reported on the last line. For example:
main.exe: file format pei-i386
architecture: i386, flags 0x0000010a:
EXEC_P, HAS_DEBUG, D_PAGED
start address 0x00401010
$ gdb main.exe
$ (gdb) break *0x00401010
$ (gdb) run
The program will stop at the given address.
(gdb) break ada_dll.adb:45
Or if you want to break using a symbol on the DLL, you need first to select the Ada language (language used by the DLL).
(gdb) set language ada
(gdb) break ada_dll
(gdb) cont
This will run the program until it reaches the breakpoint that has been set. From that point you can use the standard way to debug a program as described in (Running and Debugging Ada Programs).
It is also possible to debug the DLL by attaching to a running process.
With GDB it is always possible to debug a running process by attaching to it. It is possible to debug a DLL this way. The limitation of this approach is that the DLL must run long enough to perform the attach operation. It may be useful for instance to insert a time wasting loop in the code of the DLL to meet this criterion.
main.exe.
$ main
main.exe is 208.
$ gdb
(gdb) attach 208
(gdb) symbol-file main.exe
(gdb) break ada_dll
(gdb) cont
This last step will resume the process execution, and stop at the breakpoint we have set. From there you can use the standard approach to debug a program as described in Running and Debugging Ada Programs.
It is possible to specify the program stack size at link time. On modern versions of Windows, starting with XP, this is mostly useful to set the size of the main stack (environment task). The other task stacks are set with pragma Storage_Size or with the `gnatbind -d' command.
Since older versions of Windows (2000, NT4, etc.) do not allow setting the reserve size of individual tasks, the link-time stack size applies to all tasks, and pragma Storage_Size has no effect. In particular, Stack Overflow checks are made against this link-time specified size.
This setting can be done with `gnatlink' using either of the following:
$ gnatlink hello -Xlinker --stack=0x10000,0x1000
This sets the stack reserve size to 0x10000 bytes and the stack commit size to 0x1000 bytes.
$ gnatlink hello -Wl,--stack=0x1000000
This sets the stack reserve size to 0x1000000 bytes. Note that with `-Wl' option it is not possible to set the stack commit size because the coma is a separator for this option.
Under Windows systems, it is possible to specify the program heap size from `gnatlink' using either of the following:
$ gnatlink hello -Xlinker --heap=0x10000,0x1000
This sets the heap reserve size to 0x10000 bytes and the heap commit size to 0x1000 bytes.
$ gnatlink hello -Wl,--heap=0x1000000
This sets the heap reserve size to 0x1000000 bytes. Note that with `-Wl' option it is not possible to set the heap commit size because the coma is a separator for this option.
This section describes the Windows specific add-ons.
Win32Ada is a binding for the Microsoft Win32 API. This binding can be easily installed from the provided installer. To use the Win32Ada binding you need to use a project file, and adding a single with_clause will give you full access to the Win32Ada binding sources and ensure that the proper libraries are passed to the linker.
with "win32ada";
project P is
for Sources use ...;
end P;
To build the application you just need to call gprbuild for the application's project, here p.gpr:
gprbuild p.gpr
wPOSIX is a minimal POSIX binding whose goal is to help with building cross-platforms applications. This binding is not complete though, as the Win32 API does not provide the necessary support for all POSIX APIs.
To use the wPOSIX binding you need to use a project file, and adding a single with_clause will give you full access to the wPOSIX binding sources and ensure that the proper libraries are passed to the linker.
with "wposix";
project P is
for Sources use ...;
end P;
To build the application you just need to call gprbuild for the application's project, here p.gpr:
gprbuild p.gpr
This section describes topics that are specific to Apple's OS X platform.
The Darwin Kernel requires the debugger to have special permissions before it is allowed to control other processes. These permissions are granted by codesigning the GDB executable. Without these permissions, the debugger will report error messages such as:
Starting program: /x/y/foo
Unable to find Mach task port for process-id 28885: (os/kern) failure (0x5).
(please check gdb is codesigned - see taskgated(8))
Codesigning requires a certificate. The following procedure explains how to create one:
Once a certificate has been created, the debugger can be codesigned as follow. In a Terminal, run the following command:
$ codesign -f -s "gdb-cert" <gnat_install_prefix>/bin/gdb
where "gdb-cert" should be replaced by the actual certificate
name chosen above, and <gnat_install_prefix> should be replaced by
the location where you installed GNAT. Also, be sure that users are
in the Unix group _developer.
This Appendix displays the source code for the output file generated by `gnatbind' for a simple 'Hello World' program. Comments have been added for clarification purposes.
-- The package is called Ada_Main unless this name is actually used
-- as a unit name in the partition, in which case some other unique
-- name is used.
pragma Ada_95;
with System;
package ada_main is
pragma Warnings (Off);
-- The main program saves the parameters (argument count,
-- argument values, environment pointer) in global variables
-- for later access by other units including
-- Ada.Command_Line.
gnat_argc : Integer;
gnat_argv : System.Address;
gnat_envp : System.Address;
-- The actual variables are stored in a library routine. This
-- is useful for some shared library situations, where there
-- are problems if variables are not in the library.
pragma Import (C, gnat_argc);
pragma Import (C, gnat_argv);
pragma Import (C, gnat_envp);
-- The exit status is similarly an external location
gnat_exit_status : Integer;
pragma Import (C, gnat_exit_status);
GNAT_Version : constant String :=
"GNAT Version: Pro 7.4.0w (20141119-49)" & ASCII.NUL;
pragma Export (C, GNAT_Version, "__gnat_version");
Ada_Main_Program_Name : constant String := "_ada_hello" & ASCII.NUL;
pragma Export (C, Ada_Main_Program_Name, "__gnat_ada_main_program_name");
-- This is the generated adainit routine that performs
-- initialization at the start of execution. In the case
-- where Ada is the main program, this main program makes
-- a call to adainit at program startup.
procedure adainit;
pragma Export (C, adainit, "adainit");
-- This is the generated adafinal routine that performs
-- finalization at the end of execution. In the case where
-- Ada is the main program, this main program makes a call
-- to adafinal at program termination.
procedure adafinal;
pragma Export (C, adafinal, "adafinal");
-- This routine is called at the start of execution. It is
-- a dummy routine that is used by the debugger to breakpoint
-- at the start of execution.
-- This is the actual generated main program (it would be
-- suppressed if the no main program switch were used). As
-- required by standard system conventions, this program has
-- the external name main.
function main
(argc : Integer;
argv : System.Address;
envp : System.Address)
return Integer;
pragma Export (C, main, "main");
-- The following set of constants give the version
-- identification values for every unit in the bound
-- partition. This identification is computed from all
-- dependent semantic units, and corresponds to the
-- string that would be returned by use of the
-- Body_Version or Version attributes.
-- The following Export pragmas export the version numbers
-- with symbolic names ending in B (for body) or S
-- (for spec) so that they can be located in a link. The
-- information provided here is sufficient to track down
-- the exact versions of units used in a given build.
type Version_32 is mod 2 ** 32;
u00001 : constant Version_32 := 16#8ad6e54a#;
pragma Export (C, u00001, "helloB");
u00002 : constant Version_32 := 16#fbff4c67#;
pragma Export (C, u00002, "system__standard_libraryB");
u00003 : constant Version_32 := 16#1ec6fd90#;
pragma Export (C, u00003, "system__standard_libraryS");
u00004 : constant Version_32 := 16#3ffc8e18#;
pragma Export (C, u00004, "adaS");
u00005 : constant Version_32 := 16#28f088c2#;
pragma Export (C, u00005, "ada__text_ioB");
u00006 : constant Version_32 := 16#f372c8ac#;
pragma Export (C, u00006, "ada__text_ioS");
u00007 : constant Version_32 := 16#2c143749#;
pragma Export (C, u00007, "ada__exceptionsB");
u00008 : constant Version_32 := 16#f4f0cce8#;
pragma Export (C, u00008, "ada__exceptionsS");
u00009 : constant Version_32 := 16#a46739c0#;
pragma Export (C, u00009, "ada__exceptions__last_chance_handlerB");
u00010 : constant Version_32 := 16#3aac8c92#;
pragma Export (C, u00010, "ada__exceptions__last_chance_handlerS");
u00011 : constant Version_32 := 16#1d274481#;
pragma Export (C, u00011, "systemS");
u00012 : constant Version_32 := 16#a207fefe#;
pragma Export (C, u00012, "system__soft_linksB");
u00013 : constant Version_32 := 16#467d9556#;
pragma Export (C, u00013, "system__soft_linksS");
u00014 : constant Version_32 := 16#b01dad17#;
pragma Export (C, u00014, "system__parametersB");
u00015 : constant Version_32 := 16#630d49fe#;
pragma Export (C, u00015, "system__parametersS");
u00016 : constant Version_32 := 16#b19b6653#;
pragma Export (C, u00016, "system__secondary_stackB");
u00017 : constant Version_32 := 16#b6468be8#;
pragma Export (C, u00017, "system__secondary_stackS");
u00018 : constant Version_32 := 16#39a03df9#;
pragma Export (C, u00018, "system__storage_elementsB");
u00019 : constant Version_32 := 16#30e40e85#;
pragma Export (C, u00019, "system__storage_elementsS");
u00020 : constant Version_32 := 16#41837d1e#;
pragma Export (C, u00020, "system__stack_checkingB");
u00021 : constant Version_32 := 16#93982f69#;
pragma Export (C, u00021, "system__stack_checkingS");
u00022 : constant Version_32 := 16#393398c1#;
pragma Export (C, u00022, "system__exception_tableB");
u00023 : constant Version_32 := 16#b33e2294#;
pragma Export (C, u00023, "system__exception_tableS");
u00024 : constant Version_32 := 16#ce4af020#;
pragma Export (C, u00024, "system__exceptionsB");
u00025 : constant Version_32 := 16#75442977#;
pragma Export (C, u00025, "system__exceptionsS");
u00026 : constant Version_32 := 16#37d758f1#;
pragma Export (C, u00026, "system__exceptions__machineS");
u00027 : constant Version_32 := 16#b895431d#;
pragma Export (C, u00027, "system__exceptions_debugB");
u00028 : constant Version_32 := 16#aec55d3f#;
pragma Export (C, u00028, "system__exceptions_debugS");
u00029 : constant Version_32 := 16#570325c8#;
pragma Export (C, u00029, "system__img_intB");
u00030 : constant Version_32 := 16#1ffca443#;
pragma Export (C, u00030, "system__img_intS");
u00031 : constant Version_32 := 16#b98c3e16#;
pragma Export (C, u00031, "system__tracebackB");
u00032 : constant Version_32 := 16#831a9d5a#;
pragma Export (C, u00032, "system__tracebackS");
u00033 : constant Version_32 := 16#9ed49525#;
pragma Export (C, u00033, "system__traceback_entriesB");
u00034 : constant Version_32 := 16#1d7cb2f1#;
pragma Export (C, u00034, "system__traceback_entriesS");
u00035 : constant Version_32 := 16#8c33a517#;
pragma Export (C, u00035, "system__wch_conB");
u00036 : constant Version_32 := 16#065a6653#;
pragma Export (C, u00036, "system__wch_conS");
u00037 : constant Version_32 := 16#9721e840#;
pragma Export (C, u00037, "system__wch_stwB");
u00038 : constant Version_32 := 16#2b4b4a52#;
pragma Export (C, u00038, "system__wch_stwS");
u00039 : constant Version_32 := 16#92b797cb#;
pragma Export (C, u00039, "system__wch_cnvB");
u00040 : constant Version_32 := 16#09eddca0#;
pragma Export (C, u00040, "system__wch_cnvS");
u00041 : constant Version_32 := 16#6033a23f#;
pragma Export (C, u00041, "interfacesS");
u00042 : constant Version_32 := 16#ece6fdb6#;
pragma Export (C, u00042, "system__wch_jisB");
u00043 : constant Version_32 := 16#899dc581#;
pragma Export (C, u00043, "system__wch_jisS");
u00044 : constant Version_32 := 16#10558b11#;
pragma Export (C, u00044, "ada__streamsB");
u00045 : constant Version_32 := 16#2e6701ab#;
pragma Export (C, u00045, "ada__streamsS");
u00046 : constant Version_32 := 16#db5c917c#;
pragma Export (C, u00046, "ada__io_exceptionsS");
u00047 : constant Version_32 := 16#12c8cd7d#;
pragma Export (C, u00047, "ada__tagsB");
u00048 : constant Version_32 := 16#ce72c228#;
pragma Export (C, u00048, "ada__tagsS");
u00049 : constant Version_32 := 16#c3335bfd#;
pragma Export (C, u00049, "system__htableB");
u00050 : constant Version_32 := 16#99e5f76b#;
pragma Export (C, u00050, "system__htableS");
u00051 : constant Version_32 := 16#089f5cd0#;
pragma Export (C, u00051, "system__string_hashB");
u00052 : constant Version_32 := 16#3bbb9c15#;
pragma Export (C, u00052, "system__string_hashS");
u00053 : constant Version_32 := 16#807fe041#;
pragma Export (C, u00053, "system__unsigned_typesS");
u00054 : constant Version_32 := 16#d27be59e#;
pragma Export (C, u00054, "system__val_lluB");
u00055 : constant Version_32 := 16#fa8db733#;
pragma Export (C, u00055, "system__val_lluS");
u00056 : constant Version_32 := 16#27b600b2#;
pragma Export (C, u00056, "system__val_utilB");
u00057 : constant Version_32 := 16#b187f27f#;
pragma Export (C, u00057, "system__val_utilS");
u00058 : constant Version_32 := 16#d1060688#;
pragma Export (C, u00058, "system__case_utilB");
u00059 : constant Version_32 := 16#392e2d56#;
pragma Export (C, u00059, "system__case_utilS");
u00060 : constant Version_32 := 16#84a27f0d#;
pragma Export (C, u00060, "interfaces__c_streamsB");
u00061 : constant Version_32 := 16#8bb5f2c0#;
pragma Export (C, u00061, "interfaces__c_streamsS");
u00062 : constant Version_32 := 16#6db6928f#;
pragma Export (C, u00062, "system__crtlS");
u00063 : constant Version_32 := 16#4e6a342b#;
pragma Export (C, u00063, "system__file_ioB");
u00064 : constant Version_32 := 16#ba56a5e4#;
pragma Export (C, u00064, "system__file_ioS");
u00065 : constant Version_32 := 16#b7ab275c#;
pragma Export (C, u00065, "ada__finalizationB");
u00066 : constant Version_32 := 16#19f764ca#;
pragma Export (C, u00066, "ada__finalizationS");
u00067 : constant Version_32 := 16#95817ed8#;
pragma Export (C, u00067, "system__finalization_rootB");
u00068 : constant Version_32 := 16#52d53711#;
pragma Export (C, u00068, "system__finalization_rootS");
u00069 : constant Version_32 := 16#769e25e6#;
pragma Export (C, u00069, "interfaces__cB");
u00070 : constant Version_32 := 16#4a38bedb#;
pragma Export (C, u00070, "interfaces__cS");
u00071 : constant Version_32 := 16#07e6ee66#;
pragma Export (C, u00071, "system__os_libB");
u00072 : constant Version_32 := 16#d7b69782#;
pragma Export (C, u00072, "system__os_libS");
u00073 : constant Version_32 := 16#1a817b8e#;
pragma Export (C, u00073, "system__stringsB");
u00074 : constant Version_32 := 16#639855e7#;
pragma Export (C, u00074, "system__stringsS");
u00075 : constant Version_32 := 16#e0b8de29#;
pragma Export (C, u00075, "system__file_control_blockS");
u00076 : constant Version_32 := 16#b5b2aca1#;
pragma Export (C, u00076, "system__finalization_mastersB");
u00077 : constant Version_32 := 16#69316dc1#;
pragma Export (C, u00077, "system__finalization_mastersS");
u00078 : constant Version_32 := 16#57a37a42#;
pragma Export (C, u00078, "system__address_imageB");
u00079 : constant Version_32 := 16#bccbd9bb#;
pragma Export (C, u00079, "system__address_imageS");
u00080 : constant Version_32 := 16#7268f812#;
pragma Export (C, u00080, "system__img_boolB");
u00081 : constant Version_32 := 16#e8fe356a#;
pragma Export (C, u00081, "system__img_boolS");
u00082 : constant Version_32 := 16#d7aac20c#;
pragma Export (C, u00082, "system__ioB");
u00083 : constant Version_32 := 16#8365b3ce#;
pragma Export (C, u00083, "system__ioS");
u00084 : constant Version_32 := 16#6d4d969a#;
pragma Export (C, u00084, "system__storage_poolsB");
u00085 : constant Version_32 := 16#e87cc305#;
pragma Export (C, u00085, "system__storage_poolsS");
u00086 : constant Version_32 := 16#e34550ca#;
pragma Export (C, u00086, "system__pool_globalB");
u00087 : constant Version_32 := 16#c88d2d16#;
pragma Export (C, u00087, "system__pool_globalS");
u00088 : constant Version_32 := 16#9d39c675#;
pragma Export (C, u00088, "system__memoryB");
u00089 : constant Version_32 := 16#445a22b5#;
pragma Export (C, u00089, "system__memoryS");
u00090 : constant Version_32 := 16#6a859064#;
pragma Export (C, u00090, "system__storage_pools__subpoolsB");
u00091 : constant Version_32 := 16#e3b008dc#;
pragma Export (C, u00091, "system__storage_pools__subpoolsS");
u00092 : constant Version_32 := 16#63f11652#;
pragma Export (C, u00092, "system__storage_pools__subpools__finalizationB");
u00093 : constant Version_32 := 16#fe2f4b3a#;
pragma Export (C, u00093, "system__storage_pools__subpools__finalizationS");
-- BEGIN ELABORATION ORDER
-- ada%s
-- interfaces%s
-- system%s
-- system.case_util%s
-- system.case_util%b
-- system.htable%s
-- system.img_bool%s
-- system.img_bool%b
-- system.img_int%s
-- system.img_int%b
-- system.io%s
-- system.io%b
-- system.parameters%s
-- system.parameters%b
-- system.crtl%s
-- interfaces.c_streams%s
-- interfaces.c_streams%b
-- system.standard_library%s
-- system.exceptions_debug%s
-- system.exceptions_debug%b
-- system.storage_elements%s
-- system.storage_elements%b
-- system.stack_checking%s
-- system.stack_checking%b
-- system.string_hash%s
-- system.string_hash%b
-- system.htable%b
-- system.strings%s
-- system.strings%b
-- system.os_lib%s
-- system.traceback_entries%s
-- system.traceback_entries%b
-- ada.exceptions%s
-- system.soft_links%s
-- system.unsigned_types%s
-- system.val_llu%s
-- system.val_util%s
-- system.val_util%b
-- system.val_llu%b
-- system.wch_con%s
-- system.wch_con%b
-- system.wch_cnv%s
-- system.wch_jis%s
-- system.wch_jis%b
-- system.wch_cnv%b
-- system.wch_stw%s
-- system.wch_stw%b
-- ada.exceptions.last_chance_handler%s
-- ada.exceptions.last_chance_handler%b
-- system.address_image%s
-- system.exception_table%s
-- system.exception_table%b
-- ada.io_exceptions%s
-- ada.tags%s
-- ada.streams%s
-- ada.streams%b
-- interfaces.c%s
-- system.exceptions%s
-- system.exceptions%b
-- system.exceptions.machine%s
-- system.finalization_root%s
-- system.finalization_root%b
-- ada.finalization%s
-- ada.finalization%b
-- system.storage_pools%s
-- system.storage_pools%b
-- system.finalization_masters%s
-- system.storage_pools.subpools%s
-- system.storage_pools.subpools.finalization%s
-- system.storage_pools.subpools.finalization%b
-- system.memory%s
-- system.memory%b
-- system.standard_library%b
-- system.pool_global%s
-- system.pool_global%b
-- system.file_control_block%s
-- system.file_io%s
-- system.secondary_stack%s
-- system.file_io%b
-- system.storage_pools.subpools%b
-- system.finalization_masters%b
-- interfaces.c%b
-- ada.tags%b
-- system.soft_links%b
-- system.os_lib%b
-- system.secondary_stack%b
-- system.address_image%b
-- system.traceback%s
-- ada.exceptions%b
-- system.traceback%b
-- ada.text_io%s
-- ada.text_io%b
-- hello%b
-- END ELABORATION ORDER
end ada_main;
pragma Ada_95;
-- The following source file name pragmas allow the generated file
-- names to be unique for different main programs. They are needed
-- since the package name will always be Ada_Main.
pragma Source_File_Name (ada_main, Spec_File_Name => "b~hello.ads");
pragma Source_File_Name (ada_main, Body_File_Name => "b~hello.adb");
pragma Suppress (Overflow_Check);
with Ada.Exceptions;
-- Generated package body for Ada_Main starts here
package body ada_main is
pragma Warnings (Off);
-- These values are reference counter associated to units which have
-- been elaborated. It is also used to avoid elaborating the
-- same unit twice.
E72 : Short_Integer; pragma Import (Ada, E72, "system__os_lib_E");
E13 : Short_Integer; pragma Import (Ada, E13, "system__soft_links_E");
E23 : Short_Integer; pragma Import (Ada, E23, "system__exception_table_E");
E46 : Short_Integer; pragma Import (Ada, E46, "ada__io_exceptions_E");
E48 : Short_Integer; pragma Import (Ada, E48, "ada__tags_E");
E45 : Short_Integer; pragma Import (Ada, E45, "ada__streams_E");
E70 : Short_Integer; pragma Import (Ada, E70, "interfaces__c_E");
E25 : Short_Integer; pragma Import (Ada, E25, "system__exceptions_E");
E68 : Short_Integer; pragma Import (Ada, E68, "system__finalization_root_E");
E66 : Short_Integer; pragma Import (Ada, E66, "ada__finalization_E");
E85 : Short_Integer; pragma Import (Ada, E85, "system__storage_pools_E");
E77 : Short_Integer; pragma Import (Ada, E77, "system__finalization_masters_E");
E91 : Short_Integer; pragma Import (Ada, E91, "system__storage_pools__subpools_E");
E87 : Short_Integer; pragma Import (Ada, E87, "system__pool_global_E");
E75 : Short_Integer; pragma Import (Ada, E75, "system__file_control_block_E");
E64 : Short_Integer; pragma Import (Ada, E64, "system__file_io_E");
E17 : Short_Integer; pragma Import (Ada, E17, "system__secondary_stack_E");
E06 : Short_Integer; pragma Import (Ada, E06, "ada__text_io_E");
Local_Priority_Specific_Dispatching : constant String := "";
Local_Interrupt_States : constant String := "";
Is_Elaborated : Boolean := False;
procedure finalize_library is
begin
E06 := E06 - 1;
declare
procedure F1;
pragma Import (Ada, F1, "ada__text_io__finalize_spec");
begin
F1;
end;
E77 := E77 - 1;
E91 := E91 - 1;
declare
procedure F2;
pragma Import (Ada, F2, "system__file_io__finalize_body");
begin
E64 := E64 - 1;
F2;
end;
declare
procedure F3;
pragma Import (Ada, F3, "system__file_control_block__finalize_spec");
begin
E75 := E75 - 1;
F3;
end;
E87 := E87 - 1;
declare
procedure F4;
pragma Import (Ada, F4, "system__pool_global__finalize_spec");
begin
F4;
end;
declare
procedure F5;
pragma Import (Ada, F5, "system__storage_pools__subpools__finalize_spec");
begin
F5;
end;
declare
procedure F6;
pragma Import (Ada, F6, "system__finalization_masters__finalize_spec");
begin
F6;
end;
declare
procedure Reraise_Library_Exception_If_Any;
pragma Import (Ada, Reraise_Library_Exception_If_Any, "__gnat_reraise_library_exception_if_any");
begin
Reraise_Library_Exception_If_Any;
end;
end finalize_library;
-------------
-- adainit --
-------------
procedure adainit is
Main_Priority : Integer;
pragma Import (C, Main_Priority, "__gl_main_priority");
Time_Slice_Value : Integer;
pragma Import (C, Time_Slice_Value, "__gl_time_slice_val");
WC_Encoding : Character;
pragma Import (C, WC_Encoding, "__gl_wc_encoding");
Locking_Policy : Character;
pragma Import (C, Locking_Policy, "__gl_locking_policy");
Queuing_Policy : Character;
pragma Import (C, Queuing_Policy, "__gl_queuing_policy");
Task_Dispatching_Policy : Character;
pragma Import (C, Task_Dispatching_Policy, "__gl_task_dispatching_policy");
Priority_Specific_Dispatching : System.Address;
pragma Import (C, Priority_Specific_Dispatching, "__gl_priority_specific_dispatching");
Num_Specific_Dispatching : Integer;
pragma Import (C, Num_Specific_Dispatching, "__gl_num_specific_dispatching");
Main_CPU : Integer;
pragma Import (C, Main_CPU, "__gl_main_cpu");
Interrupt_States : System.Address;
pragma Import (C, Interrupt_States, "__gl_interrupt_states");
Num_Interrupt_States : Integer;
pragma Import (C, Num_Interrupt_States, "__gl_num_interrupt_states");
Unreserve_All_Interrupts : Integer;
pragma Import (C, Unreserve_All_Interrupts, "__gl_unreserve_all_interrupts");
Detect_Blocking : Integer;
pragma Import (C, Detect_Blocking, "__gl_detect_blocking");
Default_Stack_Size : Integer;
pragma Import (C, Default_Stack_Size, "__gl_default_stack_size");
Leap_Seconds_Support : Integer;
pragma Import (C, Leap_Seconds_Support, "__gl_leap_seconds_support");
procedure Runtime_Initialize;
pragma Import (C, Runtime_Initialize, "__gnat_runtime_initialize");
Finalize_Library_Objects : No_Param_Proc;
pragma Import (C, Finalize_Library_Objects, "__gnat_finalize_library_objects");
-- Start of processing for adainit
begin
-- Record various information for this partition. The values
-- are derived by the binder from information stored in the ali
-- files by the compiler.
if Is_Elaborated then
return;
end if;
Is_Elaborated := True;
Main_Priority := -1;
Time_Slice_Value := -1;
WC_Encoding := 'b';
Locking_Policy := ' ';
Queuing_Policy := ' ';
Task_Dispatching_Policy := ' ';
Priority_Specific_Dispatching :=
Local_Priority_Specific_Dispatching'Address;
Num_Specific_Dispatching := 0;
Main_CPU := -1;
Interrupt_States := Local_Interrupt_States'Address;
Num_Interrupt_States := 0;
Unreserve_All_Interrupts := 0;
Detect_Blocking := 0;
Default_Stack_Size := -1;
Leap_Seconds_Support := 0;
Runtime_Initialize;
Finalize_Library_Objects := finalize_library'access;
-- Now we have the elaboration calls for all units in the partition.
-- The Elab_Spec and Elab_Body attributes generate references to the
-- implicit elaboration procedures generated by the compiler for
-- each unit that requires elaboration. Increment a counter of
-- reference for each unit.
System.Soft_Links'Elab_Spec;
System.Exception_Table'Elab_Body;
E23 := E23 + 1;
Ada.Io_Exceptions'Elab_Spec;
E46 := E46 + 1;
Ada.Tags'Elab_Spec;
Ada.Streams'Elab_Spec;
E45 := E45 + 1;
Interfaces.C'Elab_Spec;
System.Exceptions'Elab_Spec;
E25 := E25 + 1;
System.Finalization_Root'Elab_Spec;
E68 := E68 + 1;
Ada.Finalization'Elab_Spec;
E66 := E66 + 1;
System.Storage_Pools'Elab_Spec;
E85 := E85 + 1;
System.Finalization_Masters'Elab_Spec;
System.Storage_Pools.Subpools'Elab_Spec;
System.Pool_Global'Elab_Spec;
E87 := E87 + 1;
System.File_Control_Block'Elab_Spec;
E75 := E75 + 1;
System.File_Io'Elab_Body;
E64 := E64 + 1;
E91 := E91 + 1;
System.Finalization_Masters'Elab_Body;
E77 := E77 + 1;
E70 := E70 + 1;
Ada.Tags'Elab_Body;
E48 := E48 + 1;
System.Soft_Links'Elab_Body;
E13 := E13 + 1;
System.Os_Lib'Elab_Body;
E72 := E72 + 1;
System.Secondary_Stack'Elab_Body;
E17 := E17 + 1;
Ada.Text_Io'Elab_Spec;
Ada.Text_Io'Elab_Body;
E06 := E06 + 1;
end adainit;
--------------
-- adafinal --
--------------
procedure adafinal is
procedure s_stalib_adafinal;
pragma Import (C, s_stalib_adafinal, "system__standard_library__adafinal");
procedure Runtime_Finalize;
pragma Import (C, Runtime_Finalize, "__gnat_runtime_finalize");
begin
if not Is_Elaborated then
return;
end if;
Is_Elaborated := False;
Runtime_Finalize;
s_stalib_adafinal;
end adafinal;
-- We get to the main program of the partition by using
-- pragma Import because if we try to with the unit and
-- call it Ada style, then not only do we waste time
-- recompiling it, but also, we don't really know the right
-- switches (e.g.@: identifier character set) to be used
-- to compile it.
procedure Ada_Main_Program;
pragma Import (Ada, Ada_Main_Program, "_ada_hello");
----------
-- main --
----------
-- main is actually a function, as in the ANSI C standard,
-- defined to return the exit status. The three parameters
-- are the argument count, argument values and environment
-- pointer.
function main
(argc : Integer;
argv : System.Address;
envp : System.Address)
return Integer
is
-- The initialize routine performs low level system
-- initialization using a standard library routine which
-- sets up signal handling and performs any other
-- required setup. The routine can be found in file
-- a-init.c.
procedure initialize;
pragma Import (C, initialize, "__gnat_initialize");
-- The finalize routine performs low level system
-- finalization using a standard library routine. The
-- routine is found in file a-final.c and in the standard
-- distribution is a dummy routine that does nothing, so
-- really this is a hook for special user finalization.
procedure finalize;
pragma Import (C, finalize, "__gnat_finalize");
-- The following is to initialize the SEH exceptions
SEH : aliased array (1 .. 2) of Integer;
Ensure_Reference : aliased System.Address := Ada_Main_Program_Name'Address;
pragma Volatile (Ensure_Reference);
-- Start of processing for main
begin
-- Save global variables
gnat_argc := argc;
gnat_argv := argv;
gnat_envp := envp;
-- Call low level system initialization
Initialize (SEH'Address);
-- Call our generated Ada initialization routine
adainit;
-- Now we call the main program of the partition
Ada_Main_Program;
-- Perform Ada finalization
adafinal;
-- Perform low level system finalization
Finalize;
-- Return the proper exit status
return (gnat_exit_status);
end;
-- This section is entirely comments, so it has no effect on the
-- compilation of the Ada_Main package. It provides the list of
-- object files and linker options, as well as some standard
-- libraries needed for the link. The gnatlink utility parses
-- this b~hello.adb file to read these comment lines to generate
-- the appropriate command line arguments for the call to the
-- system linker. The BEGIN/END lines are used for sentinels for
-- this parsing operation.
-- The exact file names will of course depend on the environment,
-- host/target and location of files on the host system.
-- BEGIN Object file/option list
-- ./hello.o
-- -L./
-- -L/usr/local/gnat/lib/gcc-lib/i686-pc-linux-gnu/2.8.1/adalib/
-- /usr/local/gnat/lib/gcc-lib/i686-pc-linux-gnu/2.8.1/adalib/libgnat.a
-- END Object file/option list
end ada_main;
The Ada code in the above example is exactly what is generated by the binder. We have added comments to more clearly indicate the function of each part of the generated Ada_Main package.
The code is standard Ada in all respects, and can be processed by any tools that handle Ada. In particular, it is possible to use the debugger in Ada mode to debug the generated Ada_Main package. For example, suppose that for reasons that you do not understand, your program is crashing during elaboration of the body of Ada.Text_IO. To locate this bug, you can place a breakpoint on the call:
Ada.Text_Io'Elab_Body;
and trace the elaboration routine for this package to find out where the problem might be (more usually of course you would be debugging elaboration code in your own application).
This appendix describes the handling of elaboration code in Ada and in GNAT, and discusses how the order of elaboration of program units can be controlled in GNAT, either automatically or with explicit programming features.
Ada provides rather general mechanisms for executing code at elaboration time, that is to say before the main program starts executing. Such code arises in three contexts:
Variables declared at the library level, in package specs or bodies, can require initialization that is performed at elaboration time, as in:
Sqrt_Half : Float := Sqrt (0.5);
Code in a BEGIN-END section at the outer level of a package body is executed as part of the package body elaboration code.
Tasks that are declared using task allocators at the library level start executing immediately and hence can execute at elaboration time.
Subprogram calls are possible in any of these contexts, which means that any arbitrary part of the program may be executed as part of the elaboration code. It is even possible to write a program which does all its work at elaboration time, with a null main program, although stylistically this would usually be considered an inappropriate way to structure a program.
An important concern arises in the context of elaboration code: we have to be sure that it is executed in an appropriate order. What we have is a series of elaboration code sections, potentially one section for each unit in the program. It is important that these execute in the correct order. Correctness here means that, taking the above example of the declaration of Sqrt_Half, if some other piece of elaboration code references Sqrt_Half, then it must run after the section of elaboration code that contains the declaration of Sqrt_Half.
There would never be any order of elaboration problem if we made a rule that whenever you `with' a unit, you must elaborate both the spec and body of that unit before elaborating the unit doing the `with'ing:
with Unit_1;
package Unit_2 is ...
would require that both the body and spec of Unit_1 be elaborated before the spec of Unit_2. However, a rule like that would be far too restrictive. In particular, it would make it impossible to have routines in separate packages that were mutually recursive.
You might think that a clever enough compiler could look at the actual elaboration code and determine an appropriate correct order of elaboration, but in the general case, this is not possible. Consider the following example.
In the body of Unit_1, we have a procedure Func_1 that references the variable Sqrt_1, which is declared in the elaboration code of the body of Unit_1:
Sqrt_1 : Float := Sqrt (0.1);
The elaboration code of the body of Unit_1 also contains:
if expression_1 = 1 then
Q := Unit_2.Func_2;
end if;
Unit_2 is exactly parallel, it has a procedure Func_2 that references the variable Sqrt_2, which is declared in the elaboration code of the body Unit_2:
Sqrt_2 : Float := Sqrt (0.1);
The elaboration code of the body of Unit_2 also contains:
if expression_2 = 2 then
Q := Unit_1.Func_1;
end if;
Now the question is, which of the following orders of elaboration is acceptable:
Spec of Unit_1
Spec of Unit_2
Body of Unit_1
Body of Unit_2
or
Spec of Unit_2
Spec of Unit_1
Body of Unit_2
Body of Unit_1
If you carefully analyze the flow here, you will see that you cannot tell at compile time the answer to this question. If expression_1 is not equal to 1, and expression_2 is not equal to 2, then either order is acceptable, because neither of the function calls is executed. If both tests evaluate to true, then neither order is acceptable and in fact there is no correct order.
If one of the two expressions is true, and the other is false, then one of the above orders is correct, and the other is incorrect. For example, if expression_1 /= 1 and expression_2 = 2, then the call to Func_1 will occur, but not the call to Func_2. This means that it is essential to elaborate the body of Unit_1 before the body of Unit_2, so the first order of elaboration is correct and the second is wrong.
By making expression_1 and expression_2 depend on input data, or perhaps the time of day, we can make it impossible for the compiler or binder to figure out which of these expressions will be true, and hence it is impossible to guarantee a safe order of elaboration at run time.
In some languages that involve the same kind of elaboration problems, e.g., Java and C++, the programmer needs to take these ordering problems into account, and it is common to write a program in which an incorrect elaboration order gives surprising results, because it references variables before they are initialized. Ada is designed to be a safe language, and a programmer-beware approach is clearly not sufficient. Consequently, the language provides three lines of defense:
Some standard rules restrict the possible choice of elaboration order. In particular, if you `with' a unit, then its spec is always elaborated before the unit doing the `with'. Similarly, a parent spec is always elaborated before the child spec, and finally a spec is always elaborated before its corresponding body.
Dynamic checks are made at run time, so that if some entity is accessed before it is elaborated (typically by means of a subprogram call) then the exception (Program_Error) is raised.
Facilities are provided for the programmer to specify the desired order of elaboration.
Let's look at these facilities in more detail. First, the rules for dynamic checking. One possible rule would be simply to say that the exception is raised if you access a variable which has not yet been elaborated. The trouble with this approach is that it could require expensive checks on every variable reference. Instead Ada has two rules which are a little more restrictive, but easier to check, and easier to state:
A subprogram can only be called at elaboration time if its body has been elaborated. The rules for elaboration given above guarantee that the spec of the subprogram has been elaborated before the call, but not the body. If this rule is violated, then the exception Program_Error is raised.
A generic unit can only be instantiated if the body of the generic unit has been elaborated. Again, the rules for elaboration given above guarantee that the spec of the generic unit has been elaborated before the instantiation, but not the body. If this rule is violated, then the exception Program_Error is raised.
The idea is that if the body has been elaborated, then any variables it references must have been elaborated; by checking for the body being elaborated we guarantee that none of its references causes any trouble. As we noted above, this is a little too restrictive, because a subprogram that has no non-local references in its body may in fact be safe to call. However, it really would be unsafe to rely on this, because it would mean that the caller was aware of details of the implementation in the body. This goes against the basic tenets of Ada.
A plausible implementation can be described as follows. A Boolean variable is associated with each subprogram and each generic unit. This variable is initialized to False, and is set to True at the point body is elaborated. Every call or instantiation checks the variable, and raises Program_Error if the variable is False.
Note that one might think that it would be good enough to have one Boolean variable for each package, but that would not deal with cases of trying to call a body in the same package as the call that has not been elaborated yet. Of course a compiler may be able to do enough analysis to optimize away some of the Boolean variables as unnecessary, and GNAT indeed does such optimizations, but still the easiest conceptual model is to think of there being one variable per subprogram.
In the previous section we discussed the rules in Ada which ensure that Program_Error is raised if an incorrect elaboration order is chosen. This prevents erroneous executions, but we need mechanisms to specify a correct execution and avoid the exception altogether. To achieve this, Ada provides a number of features for controlling the order of elaboration. We discuss these features in this section.
First, there are several ways of indicating to the compiler that a given unit has no elaboration problems:
A library package that does not require a body does not permit a body (this rule was introduced in Ada 95). Thus if we have a such a package, as in:
package Definitions is
generic
type m is new integer;
package Subp is
type a is array (1 .. 10) of m;
type b is array (1 .. 20) of m;
end Subp;
end Definitions;
A package that `with's Definitions may safely instantiate Definitions.Subp because the compiler can determine that there definitely is no package body to worry about in this case
This pragma places sufficient restrictions on a unit to guarantee that no call to any subprogram in the unit can result in an elaboration problem. This means that the compiler does not need to worry about the point of elaboration of such units, and in particular, does not need to check any calls to any subprograms in this unit.
This pragma places slightly less stringent restrictions on a unit than does pragma Pure, but these restrictions are still sufficient to ensure that there are no elaboration problems with any calls to the unit.
This pragma requires that the body of a unit be elaborated immediately after its spec. Suppose a unit A has such a pragma, and unit B does a `with' of unit A. Recall that the standard rules require the spec of unit A to be elaborated before the `with'ing unit; given the pragma in A, we also know that the body of A will be elaborated before B, so that calls to A are safe and do not need a check.
Note that, unlike pragma Pure and pragma Preelaborate, the use of Elaborate_Body does not guarantee that the program is free of elaboration problems, because it may not be possible to satisfy the requested elaboration order. Let's go back to the example with Unit_1 and Unit_2. If a programmer marks Unit_1 as Elaborate_Body, and not Unit_2, then the order of elaboration will be:
Spec of Unit_2
Spec of Unit_1
Body of Unit_1
Body of Unit_2
Now that means that the call to Func_1 in Unit_2 need not be checked, it must be safe. But the call to Func_2 in Unit_1 may still fail if Expression_1 is equal to 1, and the programmer must still take responsibility for this not being the case.
If all units carry a pragma Elaborate_Body, then all problems are eliminated, except for calls entirely within a body, which are in any case fully under programmer control. However, using the pragma everywhere is not always possible. In particular, for our Unit_1/Unit_2 example, if we marked both of them as having pragma Elaborate_Body, then clearly there would be no possible elaboration order.
The above pragmas allow a server to guarantee safe use by clients, and clearly this is the preferable approach. Consequently a good rule is to mark units as Pure or Preelaborate if possible, and if this is not possible, mark them as Elaborate_Body if possible. As we have seen, there are situations where neither of these three pragmas can be used. So we also provide methods for clients to control the order of elaboration of the servers on which they depend:
This pragma is placed in the context clause, after a `with' clause, and it requires that the body of the named unit be elaborated before the unit in which the pragma occurs. The idea is to use this pragma if the current unit calls at elaboration time, directly or indirectly, some subprogram in the named unit.
This is a stronger version of the Elaborate pragma. Consider the following example:
Unit A |withs| unit B and calls B.Func in elab code
Unit B |withs| unit C, and B.Func calls C.Func
Now if we put a pragma Elaborate (B) in unit A, this ensures that the body of B is elaborated before the call, but not the body of C, so the call to C.Func could still cause Program_Error to be raised.
The effect of a pragma Elaborate_All is stronger, it requires not only that the body of the named unit be elaborated before the unit doing the `with', but also the bodies of all units that the named unit uses, following `with' links transitively. For example, if we put a pragma Elaborate_All (B) in unit A, then it requires not only that the body of B be elaborated before A, but also the body of C, because B `with's C.
We are now in a position to give a usage rule in Ada for avoiding elaboration problems, at least if dynamic dispatching and access to subprogram values are not used. We will handle these cases separately later.
The rule is simple:
`If a unit has elaboration code that can directly or indirectly make a call to a subprogram in a |withed| unit, or instantiate a generic package in a |withed| unit, then if the |withed| unit does not have pragma `Pure` or `Preelaborate`, then the client should have a pragma `Elaborate_All`for the |withed| unit.*'
By following this rule a client is assured that calls can be made without risk of an exception.
For generic subprogram instantiations, the rule can be relaxed to require only a pragma Elaborate since elaborating the body of a subprogram cannot cause any transitive elaboration (we are not calling the subprogram in this case, just elaborating its declaration).
If this rule is not followed, then a program may be in one of four states:
No order of elaboration exists which follows the rules, taking into account any Elaborate, Elaborate_All, or Elaborate_Body pragmas. In this case, an Ada compiler must diagnose the situation at bind time, and refuse to build an executable program.
One or more acceptable elaboration orders exist, and all of them generate an elaboration order problem. In this case, the binder can build an executable program, but Program_Error will be raised when the program is run.
One or more acceptable elaboration orders exists, and some of them work, and some do not. The programmer has not controlled the order of elaboration, so the binder may or may not pick one of the correct orders, and the program may or may not raise an exception when it is run. This is the worst case, because it means that the program may fail when moved to another compiler, or even another version of the same compiler.
One ore more acceptable elaboration orders exist, and all of them work. In this case the program runs successfully. This state of affairs can be guaranteed by following the rule we gave above, but may be true even if the rule is not followed.
Note that one additional advantage of following our rules on the use of Elaborate and Elaborate_All is that the program continues to stay in the ideal (all orders OK) state even if maintenance changes some bodies of some units. Conversely, if a program that does not follow this rule happens to be safe at some point, this state of affairs may deteriorate silently as a result of maintenance changes.
You may have noticed that the above discussion did not mention the use of Elaborate_Body. This was a deliberate omission. If you `with' an Elaborate_Body unit, it still may be the case that code in the body makes calls to some other unit, so it is still necessary to use Elaborate_All on such units.
In the case of internal calls, i.e., calls within a single package, the programmer has full control over the order of elaboration, and it is up to the programmer to elaborate declarations in an appropriate order. For example writing:
function One return Float;
Q : Float := One;
function One return Float is
begin
return 1.0;
end One;
will obviously raise Program_Error at run time, because function One will be called before its body is elaborated. In this case GNAT will generate a warning that the call will raise Program_Error:
1. procedure y is
2. function One return Float;
3.
4. Q : Float := One;
|
>>> warning: cannot call "One" before body is elaborated
>>> warning: Program_Error will be raised at run time
5.
6. function One return Float is
7. begin
8. return 1.0;
9. end One;
10.
11. begin
12. null;
13. end;
Note that in this particular case, it is likely that the call is safe, because the function One does not access any global variables. Nevertheless in Ada, we do not want the validity of the check to depend on the contents of the body (think about the separate compilation case), so this is still wrong, as we discussed in the previous sections.
The error is easily corrected by rearranging the declarations so that the body of One appears before the declaration containing the call (note that in Ada 95 as well as later versions of the Ada standard, declarations can appear in any order, so there is no restriction that would prevent this reordering, and if we write:
function One return Float;
function One return Float is
begin
return 1.0;
end One;
Q : Float := One;
then all is well, no warning is generated, and no Program_Error exception will be raised. Things are more complicated when a chain of subprograms is executed:
function A return Integer;
function B return Integer;
function C return Integer;
function B return Integer is begin return A; end;
function C return Integer is begin return B; end;
X : Integer := C;
function A return Integer is begin return 1; end;
Now the call to C at elaboration time in the declaration of X is correct, because the body of C is already elaborated, and the call to B within the body of C is correct, but the call to A within the body of B is incorrect, because the body of A has not been elaborated, so Program_Error will be raised on the call to A. In this case GNAT will generate a warning that Program_Error may be raised at the point of the call. Let's look at the warning:
1. procedure x is
2. function A return Integer;
3. function B return Integer;
4. function C return Integer;
5.
6. function B return Integer is begin return A; end;
|
>>> warning: call to "A" before body is elaborated may
raise Program_Error
>>> warning: "B" called at line 7
>>> warning: "C" called at line 9
7. function C return Integer is begin return B; end;
8.
9. X : Integer := C;
10.
11. function A return Integer is begin return 1; end;
12.
13. begin
14. null;
15. end;
Note that the message here says 'may raise', instead of the direct case, where the message says 'will be raised'. That's because whether A is actually called depends in general on run-time flow of control. For example, if the body of B said
function B return Integer is
begin
if some-condition-depending-on-input-data then
return A;
else
return 1;
end if;
end B;
then we could not know until run time whether the incorrect call to A would actually occur, so Program_Error might or might not be raised. It is possible for a compiler to do a better job of analyzing bodies, to determine whether or not Program_Error might be raised, but it certainly couldn't do a perfect job (that would require solving the halting problem and is provably impossible), and because this is a warning anyway, it does not seem worth the effort to do the analysis. Cases in which it would be relevant are rare.
In practice, warnings of either of the forms given above will usually correspond to real errors, and should be examined carefully and eliminated. In the rare case where a warning is bogus, it can be suppressed by any of the following methods:
For the internal elaboration check case, GNAT by default generates the necessary run-time checks to ensure that Program_Error is raised if any call fails an elaboration check. Of course this can only happen if a warning has been issued as described above. The use of pragma Suppress (Elaboration_Check) may (but is not guaranteed to) suppress some of these checks, meaning that it may be possible (but is not guaranteed) for a program to be able to call a subprogram whose body is not yet elaborated, without raising a Program_Error exception.
The previous section discussed the case in which the execution of a particular thread of elaboration code occurred entirely within a single unit. This is the easy case to handle, because a programmer has direct and total control over the order of elaboration, and furthermore, checks need only be generated in cases which are rare and which the compiler can easily detect. The situation is more complex when separate compilation is taken into account. Consider the following:
package Math is
function Sqrt (Arg : Float) return Float;
end Math;
package body Math is
function Sqrt (Arg : Float) return Float is
begin
...
end Sqrt;
end Math;
with Math;
package Stuff is
X : Float := Math.Sqrt (0.5);
end Stuff;
with Stuff;
procedure Main is
begin
...
end Main;
where Main is the main program. When this program is executed, the elaboration code must first be executed, and one of the jobs of the binder is to determine the order in which the units of a program are to be elaborated. In this case we have four units: the spec and body of Math, the spec of Stuff and the body of Main). In what order should the four separate sections of elaboration code be executed?
There are some restrictions in the order of elaboration that the binder can choose. In particular, if unit U has a `with' for a package X, then you are assured that the spec of X is elaborated before U , but you are not assured that the body of X is elaborated before U. This means that in the above case, the binder is allowed to choose the order:
spec of Math
spec of Stuff
body of Math
body of Main
but that's not good, because now the call to Math.Sqrt that happens during the elaboration of the Stuff spec happens before the body of Math.Sqrt is elaborated, and hence causes Program_Error exception to be raised. At first glance, one might say that the binder is misbehaving, because obviously you want to elaborate the body of something you `with' first, but that is not a general rule that can be followed in all cases. Consider
package X is ...
package Y is ...
with X;
package body Y is ...
with Y;
package body X is ...
This is a common arrangement, and, apart from the order of elaboration problems that might arise in connection with elaboration code, this works fine. A rule that says that you must first elaborate the body of anything you `with' cannot work in this case: the body of X `with's Y, which means you would have to elaborate the body of Y first, but that `with's X, which means you have to elaborate the body of X first, but ... and we have a loop that cannot be broken.
It is true that the binder can in many cases guess an order of elaboration that is unlikely to cause a Program_Error exception to be raised, and it tries to do so (in the above example of Math/Stuff/Spec, the GNAT binder will by default elaborate the body of Math right after its spec, so all will be well).
However, a program that blindly relies on the binder to be helpful can get into trouble, as we discussed in the previous sections, so GNAT provides a number of facilities for assisting the programmer in developing programs that are robust with respect to elaboration order.
The default behavior in GNAT ensures elaboration safety. In its default mode GNAT implements the rule we previously described as the right approach. Let's restate it:
`If a unit has elaboration code that can directly or indirectly make a call to a subprogram in a |withed| unit, or instantiate a generic package in a |withed| unit, then if the |withed| unit does not have pragma `Pure` or `Preelaborate`, then the client should have an `Elaborate_All` pragma for the |withed| unit.'
`In the case of instantiating a generic subprogram, it is always sufficient to have only an `Elaborate` pragma for the |withed| unit.'
By following this rule a client is assured that calls and instantiations can be made without risk of an exception.
In this mode GNAT traces all calls that are potentially made from elaboration code, and puts in any missing implicit Elaborate and Elaborate_All pragmas. The advantage of this approach is that no elaboration problems are possible if the binder can find an elaboration order that is consistent with these implicit Elaborate and Elaborate_All pragmas. The disadvantage of this approach is that no such order may exist.
If the binder does not generate any diagnostics, then it means that it has found an elaboration order that is guaranteed to be safe. However, the binder may still be relying on implicitly generated Elaborate and Elaborate_All pragmas so portability to other compilers than GNAT is not guaranteed.
If it is important to guarantee portability, then the compilations should use the `-gnatel' (info messages for elaboration pragmas) switch. This will cause info messages to be generated indicating the missing Elaborate and Elaborate_All pragmas. Consider the following source program:
with k;
package j is
m : integer := k.r;
end;
where it is clear that there should be a pragma Elaborate_All for unit k. An implicit pragma will be generated, and it is likely that the binder will be able to honor it. However, if you want to port this program to some other Ada compiler than GNAT. it is safer to include the pragma explicitly in the source. If this unit is compiled with the `-gnatel' switch, then the compiler outputs an information message:
1. with k;
2. package j is
3. m : integer := k.r;
|
>>> info: call to "r" may raise Program_Error
>>> info: missing pragma Elaborate_All for "k"
4. end;
and these messages can be used as a guide for supplying manually the missing pragmas. It is usually a bad idea to use this option during development. That's because it will tell you when you need to put in a pragma, but cannot tell you when it is time to take it out. So the use of pragma Elaborate_All may lead to unnecessary dependencies and even false circularities.
This default mode is more restrictive than the Ada Reference Manual, and it is possible to construct programs which will compile using the dynamic model described there, but will run into a circularity using the safer static model we have described.
Of course any Ada compiler must be able to operate in a mode consistent with the requirements of the Ada Reference Manual, and in particular must have the capability of implementing the standard dynamic model of elaboration with run-time checks.
In GNAT, this standard mode can be achieved either by the use of the `-gnatE' switch on the compiler (`gcc' or `gnatmake') command, or by the use of the configuration pragma:
pragma Elaboration_Checks (DYNAMIC);
Either approach will cause the unit affected to be compiled using the standard dynamic run-time elaboration checks described in the Ada Reference Manual. The static model is generally preferable, since it is clearly safer to rely on compile and link time checks rather than run-time checks. However, in the case of legacy code, it may be difficult to meet the requirements of the static model. This issue is further discussed in What to Do If the Default Elaboration Behavior Fails.
Note that the static model provides a strict subset of the allowed behavior and programs of the Ada Reference Manual, so if you do adhere to the static model and no circularities exist, then you are assured that your program will work using the dynamic model, providing that you remove any pragma Elaborate statements from the source.
The use of pragma Elaborate should generally be avoided in Ada 95 and Ada 2005 programs, since there is no guarantee that transitive calls will be properly handled. Indeed at one point, this pragma was placed in Annex J (Obsolescent Features), on the grounds that it is never useful.
Now that's a bit restrictive. In practice, the case in which pragma Elaborate is useful is when the caller knows that there are no transitive calls, or that the called unit contains all necessary transitive pragma Elaborate statements, and legacy code often contains such uses.
Strictly speaking the static mode in GNAT should ignore such pragmas, since there is no assurance at compile time that the necessary safety conditions are met. In practice, this would cause GNAT to be incompatible with correctly written Ada 83 code that had all necessary pragma Elaborate statements in place. Consequently, we made the decision that GNAT in its default mode will believe that if it encounters a pragma Elaborate then the programmer knows what they are doing, and it will trust that no elaboration errors can occur.
The result of this decision is two-fold. First to be safe using the static mode, you should remove all pragma Elaborate statements. Second, when fixing circularities in existing code, you can selectively use pragma Elaborate statements to convince the static mode of GNAT that it need not generate an implicit pragma Elaborate_All statement.
When using the static mode with `-gnatwl', any use of pragma Elaborate will generate a warning about possible problems.
In this section we examine special elaboration issues that arise for programs that declare library level tasks.
Generally the model of execution of an Ada program is that all units are elaborated, and then execution of the program starts. However, the declaration of library tasks definitely does not fit this model. The reason for this is that library tasks start as soon as they are declared (more precisely, as soon as the statement part of the enclosing package body is reached), that is to say before elaboration of the program is complete. This means that if such a task calls a subprogram, or an entry in another task, the callee may or may not be elaborated yet, and in the standard Reference Manual model of dynamic elaboration checks, you can even get timing dependent Program_Error exceptions, since there can be a race between the elaboration code and the task code.
The static model of elaboration in GNAT seeks to avoid all such dynamic behavior, by being conservative, and the conservative approach in this particular case is to assume that all the code in a task body is potentially executed at elaboration time if a task is declared at the library level.
This can definitely result in unexpected circularities. Consider the following example
package Decls is
task Lib_Task is
entry Start;
end Lib_Task;
type My_Int is new Integer;
function Ident (M : My_Int) return My_Int;
end Decls;
with Utils;
package body Decls is
task body Lib_Task is
begin
accept Start;
Utils.Put_Val (2);
end Lib_Task;
function Ident (M : My_Int) return My_Int is
begin
return M;
end Ident;
end Decls;
with Decls;
package Utils is
procedure Put_Val (Arg : Decls.My_Int);
end Utils;
with Text_IO;
package body Utils is
procedure Put_Val (Arg : Decls.My_Int) is
begin
Text_IO.Put_Line (Decls.My_Int'Image (Decls.Ident (Arg)));
end Put_Val;
end Utils;
with Decls;
procedure Main is
begin
Decls.Lib_Task.Start;
end;
If the above example is compiled in the default static elaboration mode, then a circularity occurs. The circularity comes from the call Utils.Put_Val in the task body of Decls.Lib_Task. Since this call occurs in elaboration code, we need an implicit pragma Elaborate_All for Utils. This means that not only must the spec and body of Utils be elaborated before the body of Decls, but also the spec and body of any unit that is `with'ed by the body of Utils must also be elaborated before the body of Decls. This is the transitive implication of pragma Elaborate_All and it makes sense, because in general the body of Put_Val might have a call to something in a `with'ed unit.
In this case, the body of Utils (actually its spec) `with's Decls. Unfortunately this means that the body of Decls must be elaborated before itself, in case there is a call from the body of Utils.
Here is the exact chain of events we are worrying about:
Indeed, if you add an explicit pragma Elaborate_All for Utils in the body of Decls you will get a true Ada Reference Manual circularity that makes the program illegal.
In practice, we have found that problems with the static model of elaboration in existing code often arise from library tasks, so we must address this particular situation.
Note that if we compile and run the program above, using the dynamic model of elaboration (that is to say use the `-gnatE' switch), then it compiles, binds, links, and runs, printing the expected result of 2. Therefore in some sense the circularity here is only apparent, and we need to capture the properties of this program that distinguish it from other library-level tasks that have real elaboration problems.
We have four possible answers to this question:
If we use the `-gnatE' switch, then as noted above, the program works. Why is this? If we examine the task body, it is apparent that the task cannot proceed past the accept statement until after elaboration has been completed, because the corresponding entry call comes from the main program, not earlier. This is why the dynamic model works here. But that's really giving up on a precise analysis, and we prefer to take this approach only if we cannot solve the problem in any other manner. So let us examine two ways to reorganize the program to avoid the potential elaboration problem.
Write separate packages, so that library tasks are isolated from other declarations as much as possible. Let us look at a variation on the above program.
package Decls1 is
task Lib_Task is
entry Start;
end Lib_Task;
end Decls1;
with Utils;
package body Decls1 is
task body Lib_Task is
begin
accept Start;
Utils.Put_Val (2);
end Lib_Task;
end Decls1;
package Decls2 is
type My_Int is new Integer;
function Ident (M : My_Int) return My_Int;
end Decls2;
with Utils;
package body Decls2 is
function Ident (M : My_Int) return My_Int is
begin
return M;
end Ident;
end Decls2;
with Decls2;
package Utils is
procedure Put_Val (Arg : Decls2.My_Int);
end Utils;
with Text_IO;
package body Utils is
procedure Put_Val (Arg : Decls2.My_Int) is
begin
Text_IO.Put_Line (Decls2.My_Int'Image (Decls2.Ident (Arg)));
end Put_Val;
end Utils;
with Decls1;
procedure Main is
begin
Decls1.Lib_Task.Start;
end;
All we have done is to split Decls into two packages, one containing the library task, and one containing everything else. Now there is no cycle, and the program compiles, binds, links and executes using the default static model of elaboration.
A significant part of the problem arises because of the use of the single task declaration form. This means that the elaboration of the task type, and the elaboration of the task itself (i.e., the creation of the task) happen at the same time. A good rule of style in Ada is to always create explicit task types. By following the additional step of placing task objects in separate packages from the task type declaration, many elaboration problems are avoided. Here is another modified example of the example program:
package Decls is
task type Lib_Task_Type is
entry Start;
end Lib_Task_Type;
type My_Int is new Integer;
function Ident (M : My_Int) return My_Int;
end Decls;
with Utils;
package body Decls is
task body Lib_Task_Type is
begin
accept Start;
Utils.Put_Val (2);
end Lib_Task_Type;
function Ident (M : My_Int) return My_Int is
begin
return M;
end Ident;
end Decls;
with Decls;
package Utils is
procedure Put_Val (Arg : Decls.My_Int);
end Utils;
with Text_IO;
package body Utils is
procedure Put_Val (Arg : Decls.My_Int) is
begin
Text_IO.Put_Line (Decls.My_Int'Image (Decls.Ident (Arg)));
end Put_Val;
end Utils;
with Decls;
package Declst is
Lib_Task : Decls.Lib_Task_Type;
end Declst;
with Declst;
procedure Main is
begin
Declst.Lib_Task.Start;
end;
What we have done here is to replace the task declaration in package Decls with a task type declaration. Then we introduce a separate package Declst to contain the actual task object. This separates the elaboration issues for the task type declaration, which causes no trouble, from the elaboration issues of the task object, which is also unproblematic, since it is now independent of the elaboration of Utils. This separation of concerns also corresponds to a generally sound engineering principle of separating declarations from instances. This version of the program also compiles, binds, links, and executes, generating the expected output.
The previous two approaches described how a program can be restructured to avoid the special problems caused by library task bodies. in practice, however, such restructuring may be difficult to apply to existing legacy code, so we must consider solutions that do not require massive rewriting.
Let us consider more carefully why our original sample program works under the dynamic model of elaboration. The reason is that the code in the task body blocks immediately on the accept statement. Now of course there is nothing to prohibit elaboration code from making entry calls (for example from another library level task), so we cannot tell in isolation that the task will not execute the accept statement during elaboration.
However, in practice it is very unusual to see elaboration code make any entry calls, and the pattern of tasks starting at elaboration time and then immediately blocking on accept or select statements is very common. What this means is that the compiler is being too pessimistic when it analyzes the whole package body as though it might be executed at elaboration time.
If we know that the elaboration code contains no entry calls, (a very safe assumption most of the time, that could almost be made the default behavior), then we can compile all units of the program under control of the following configuration pragma:
pragma Restrictions (No_Entry_Calls_In_Elaboration_Code);
This pragma can be placed in the gnat.adc file in the usual
manner. If we take our original unmodified program and compile it
in the presence of a gnat.adc containing the above pragma,
then once again, we can compile, bind, link, and execute, obtaining
the expected result. In the presence of this pragma, the compiler does
not trace calls in a task body, that appear after the first accept
or select statement, and therefore does not report a potential
circularity in the original program.
The compiler will check to the extent it can that the above restriction is not violated, but it is not always possible to do a complete check at compile time, so it is important to use this pragma only if the stated restriction is in fact met, that is to say no task receives an entry call before elaboration of all units is completed.
So far, we have assumed that the entire program is either compiled using the dynamic model or static model, ensuring consistency. It is possible to mix the two models, but rules have to be followed if this mixing is done to ensure that elaboration checks are not omitted.
The basic rule is that `a unit compiled with the static model cannot be |withed| by a unit compiled with the dynamic model'. The reason for this is that in the static model, a unit assumes that its clients guarantee to use (the equivalent of) pragma Elaborate_All so that no elaboration checks are required in inner subprograms, and this assumption is violated if the client is compiled with dynamic checks.
The precise rule is as follows. A unit that is compiled with dynamic checks can only `with' a unit that meets at least one of the following criteria:
If this rule is violated, that is if a unit with dynamic elaboration checks `with's a unit that does not meet one of the above four criteria, then the binder (gnatbind) will issue a warning similar to that in the following example:
warning: "x.ads" has dynamic elaboration checks and with's
warning: "y.ads" which has static elaboration checks
These warnings indicate that the rule has been violated, and that as a result elaboration checks may be missed in the resulting executable file. This warning may be suppressed using the `-ws' binder switch in the usual manner.
One useful application of this mixing rule is in the case of a subsystem which does not itself `with' units from the remainder of the application. In this case, the entire subsystem can be compiled with dynamic checks to resolve a circularity in the subsystem, while allowing the main application that uses this subsystem to be compiled using the more reliable default static model.
If the binder cannot find an acceptable order, it outputs detailed diagnostics. For example:
error: elaboration circularity detected
info: "proc (body)" must be elaborated before "pack (body)"
info: reason: Elaborate_All probably needed in unit "pack (body)"
info: recompile "pack (body)" with -gnatel
info: for full details
info: "proc (body)"
info: is needed by its spec:
info: "proc (spec)"
info: which is withed by:
info: "pack (body)"
info: "pack (body)" must be elaborated before "proc (body)"
info: reason: pragma Elaborate in unit "proc (body)"
In this case we have a cycle that the binder cannot break. On the one hand, there is an explicit pragma Elaborate in proc for pack. This means that the body of pack must be elaborated before the body of proc. On the other hand, there is elaboration code in pack that calls a subprogram in proc. This means that for maximum safety, there should really be a pragma Elaborate_All in pack for proc which would require that the body of proc be elaborated before the body of pack. Clearly both requirements cannot be satisfied. Faced with a circularity of this kind, you have three different options.
The most desirable option from the point of view of long-term maintenance is to rearrange the program so that the elaboration problems are avoided. One useful technique is to place the elaboration code into separate child packages. Another is to move some of the initialization code to explicitly called subprograms, where the program controls the order of initialization explicitly. Although this is the most desirable option, it may be impractical and involve too much modification, especially in the case of complex legacy code.
If the compilations are done using the `-gnatE' (dynamic elaboration check) switch, then GNAT behaves in a quite different manner. Dynamic checks are generated for all calls that could possibly result in raising an exception. With this switch, the compiler does not generate implicit Elaborate or Elaborate_All pragmas. The behavior then is exactly as specified in the Ada Reference Manual. The binder will generate an executable program that may or may not raise Program_Error, and then it is the programmer's job to ensure that it does not raise an exception. Note that it is important to compile all units with the switch, it cannot be used selectively.
The drawback of dynamic checks is that they generate a
significant overhead at run time, both in space and time. If you
are absolutely sure that your program cannot raise any elaboration
exceptions, and you still want to use the dynamic elaboration model,
then you can use the configuration pragma
Suppress (Elaboration_Check) to suppress all such checks. For
example this pragma could be placed in the gnat.adc file.
When you know that certain calls or instantiations in elaboration code cannot possibly lead to an elaboration error, and the binder nevertheless complains about implicit Elaborate and Elaborate_All pragmas that lead to elaboration circularities, it is possible to remove those warnings locally and obtain a program that will bind. Clearly this can be unsafe, and it is the responsibility of the programmer to make sure that the resulting program has no elaboration anomalies. The pragma Suppress (Elaboration_Check) can be used with different granularity to suppress warnings and break elaboration circularities:
As previously described in section Treatment of Pragma Elaborate, GNAT in static mode assumes that a pragma Elaborate indicates correctly that no elaboration checks are required on calls to the designated unit. There may be cases in which the caller knows that no transitive calls can occur, so that a pragma Elaborate will be sufficient in a case where pragma Elaborate_All would cause a circularity.
These five cases are listed in order of decreasing safety, and therefore require increasing programmer care in their application. Consider the following program:
package Pack1 is
function F1 return Integer;
X1 : Integer;
end Pack1;
package Pack2 is
function F2 return Integer;
function Pure (x : integer) return integer;
-- pragma Suppress (Elaboration_Check, On => Pure); -- (3)
-- pragma Suppress (Elaboration_Check); -- (4)
end Pack2;
with Pack2;
package body Pack1 is
function F1 return Integer is
begin
return 100;
end F1;
Val : integer := Pack2.Pure (11); -- Elab. call (1)
begin
declare
-- pragma Suppress(Elaboration_Check, Pack2.F2); -- (1)
-- pragma Suppress(Elaboration_Check); -- (2)
begin
X1 := Pack2.F2 + 1; -- Elab. call (2)
end;
end Pack1;
with Pack1;
package body Pack2 is
function F2 return Integer is
begin
return Pack1.F1;
end F2;
function Pure (x : integer) return integer is
begin
return x ** 3 - 3 * x;
end;
end Pack2;
with Pack1, Ada.Text_IO;
procedure Proc3 is
begin
Ada.Text_IO.Put_Line(Pack1.X1'Img); -- 101
end Proc3;
In the absence of any pragmas, an attempt to bind this program produces the following diagnostics:
error: elaboration circularity detected
info: "pack1 (body)" must be elaborated before "pack1 (body)"
info: reason: Elaborate_All probably needed in unit "pack1 (body)"
info: recompile "pack1 (body)" with -gnatel for full details
info: "pack1 (body)"
info: must be elaborated along with its spec:
info: "pack1 (spec)"
info: which is withed by:
info: "pack2 (body)"
info: which must be elaborated along with its spec:
info: "pack2 (spec)"
info: which is withed by:
info: "pack1 (body)"
The sources of the circularity are the two calls to Pack2.Pure and Pack2.F2 in the body of Pack1. We can see that the call to F2 is safe, even though F2 calls F1, because the call appears after the elaboration of the body of F1. Therefore the pragma (1) is safe, and will remove the warning on the call. It is also possible to use pragma (2) because there are no other potentially unsafe calls in the block.
The call to Pure is safe because this function does not depend on the state of Pack2. Therefore any call to this function is safe, and it is correct to place pragma (3) in the corresponding package spec.
Finally, we could place pragma (4) in the spec of Pack2 to disable warnings on all calls to functions declared therein. Note that this is not necessarily safe, and requires more detailed examination of the subprogram bodies involved. In particular, a call to F2 requires that F1 be already elaborated.
It is hard to generalize on which of these four approaches should be taken. Obviously if it is possible to fix the program so that the default treatment works, this is preferable, but this may not always be practical. It is certainly simple enough to use `-gnatE' but the danger in this case is that, even if the GNAT binder finds a correct elaboration order, it may not always do so, and certainly a binder from another Ada compiler might not. A combination of testing and analysis (for which the information messages generated with the `-gnatel' switch can be useful) must be used to ensure that the program is free of errors. One switch that is useful in this testing is the `-p (pessimistic elaboration order)' switch for gnatbind. Normally the binder tries to find an order that has the best chance of avoiding elaboration problems. However, if this switch is used, the binder plays a devil's advocate role, and tries to choose the order that has the best chance of failing. If your program works even with this switch, then it has a better chance of being error free, but this is still not a guarantee.
For an example of this approach in action, consider the C-tests (executable tests) from the ACATS suite. If these are compiled and run with the default treatment, then all but one of them succeed without generating any error diagnostics from the binder. However, there is one test that fails, and this is not surprising, because the whole point of this test is to ensure that the compiler can handle cases where it is impossible to determine a correct order statically, and it checks that an exception is indeed raised at run time.
This one test must be compiled and run using the `-gnatE' switch, and then it passes. Alternatively, the entire suite can be run using this switch. It is never wrong to run with the dynamic elaboration switch if your code is correct, and we assume that the C-tests are indeed correct (it is less efficient, but efficiency is not a factor in running the ACATS tests.)
In rare cases, the static elaboration model fails to prevent dispatching calls to not-yet-elaborated subprograms. In such cases, we fall back to run-time checks; premature calls to any primitive operation of a tagged type before the body of the operation has been elaborated will raise Program_Error.
Access-to-subprogram types, however, are handled conservatively, and do not require run-time checks. This was not true in earlier versions of the compiler; you can use the `-gnatd.U' debug switch to revert to the old behavior if the new conservative behavior causes elaboration cycles. Here, 'conservative' means that if you do P'Access during elaboration, the compiler will assume that you might call P indirectly during elaboration, so it adds an implicit pragma Elaborate_All on the library unit containing P. The `-gnatd.U' switch is safe if you know there are no such calls. If the program worked before, it will continue to work with `-gnatd.U'. But beware that code modifications such as adding an indirect call can cause erroneous behavior in the presence of `-gnatd.U'.
First, compile your program with the default options, using none of the special elaboration control switches. If the binder successfully binds your program, then you can be confident that, apart from issues raised by the use of access-to-subprogram types and dynamic dispatching, the program is free of elaboration errors. If it is important that the program be portable to other compilers than GNAT, then use the `-gnatel' switch to generate messages about missing Elaborate or Elaborate_All pragmas, and supply the missing pragmas.
If the program fails to bind using the default static elaboration handling, then you can fix the program to eliminate the binder message, or recompile the entire program with the `-gnatE' switch to generate dynamic elaboration checks, and, if you are sure there really are no elaboration problems, use a global pragma Suppress (Elaboration_Check).
This section has been entirely concerned with the issue of finding a valid elaboration order, as defined by the Ada Reference Manual. In a case where several elaboration orders are valid, the task is to find one of the possible valid elaboration orders (and the static model in GNAT will ensure that this is achieved).
The purpose of the elaboration rules in the Ada Reference Manual is to make sure that no entity is accessed before it has been elaborated. For a subprogram, this means that the spec and body must have been elaborated before the subprogram is called. For an object, this means that the object must have been elaborated before its value is read or written. A violation of either of these two requirements is an access before elaboration order, and this section has been all about avoiding such errors.
In the case where more than one order of elaboration is possible, in the sense that access before elaboration errors are avoided, then any one of the orders is 'correct' in the sense that it meets the requirements of the Ada Reference Manual, and no such error occurs.
However, it may be the case for a given program, that there are constraints on the order of elaboration that come not from consideration of avoiding elaboration errors, but rather from extra-lingual logic requirements. Consider this example:
with Init_Constants;
package Constants is
X : Integer := 0;
Y : Integer := 0;
end Constants;
package Init_Constants is
procedure P; --* require a body*
end Init_Constants;
with Constants;
package body Init_Constants is
procedure P is begin null; end;
begin
Constants.X := 3;
Constants.Y := 4;
end Init_Constants;
with Constants;
package Calc is
Z : Integer := Constants.X + Constants.Y;
end Calc;
with Calc;
with Text_IO; use Text_IO;
procedure Main is
begin
Put_Line (Calc.Z'Img);
end Main;
In this example, there is more than one valid order of elaboration. For example both the following are correct orders:
Init_Constants spec
Constants spec
Calc spec
Init_Constants body
Main body
and
Init_Constants spec
Init_Constants body
Constants spec
Calc spec
Main body
There is no language rule to prefer one or the other, both are correct from an order of elaboration point of view. But the programmatic effects of the two orders are very different. In the first, the elaboration routine of Calc initializes Z to zero, and then the main program runs with this value of zero. But in the second order, the elaboration routine of Calc runs after the body of Init_Constants has set X and Y and thus Z is set to 7 before Main runs.
One could perhaps by applying pretty clever non-artificial intelligence to the situation guess that it is more likely that the second order of elaboration is the one desired, but there is no formal linguistic reason to prefer one over the other. In fact in this particular case, GNAT will prefer the second order, because of the rule that bodies are elaborated as soon as possible, but it's just luck that this is what was wanted (if indeed the second order was preferred).
If the program cares about the order of elaboration routines in a case like this, it is important to specify the order required. In this particular case, that could have been achieved by adding to the spec of Calc:
pragma Elaborate_All (Constants);
which requires that the body (if any) and spec of Constants, as well as the body and spec of any unit `with'ed by Constants be elaborated before Calc is elaborated.
Clearly no automatic method can always guess which alternative you require, and if you are working with legacy code that had constraints of this kind which were not properly specified by adding Elaborate or Elaborate_All pragmas, then indeed it is possible that two different compilers can choose different orders.
However, GNAT does attempt to diagnose the common situation where there are uninitialized variables in the visible part of a package spec, and the corresponding package body has an elaboration block that directly or indirectly initialized one or more of these variables. This is the situation in which a pragma Elaborate_Body is usually desirable, and GNAT will generate a warning that suggests this addition if it detects this situation.
The gnatbind `-p' switch may be useful in smoking out problems. This switch causes bodies to be elaborated as late as possible instead of as early as possible. In the example above, it would have forced the choice of the first elaboration order. If you get different results when using this switch, and particularly if one set of results is right, and one is wrong as far as you are concerned, it shows that you have some missing Elaborate pragmas. For the example above, we have the following output:
$ gnatmake -f -q main
$ main
7
$ gnatmake -f -q main -bargs -p
$ main
0
It is of course quite unlikely that both these results are correct, so it is up to you in a case like this to investigate the source of the difference, by looking at the two elaboration orders that are chosen, and figuring out which is correct, and then adding the necessary Elaborate or Elaborate_All pragmas to ensure the desired order.
To see the elaboration order that the binder chooses, you can look at the last part of the file:b~xxx.adb binder output file. Here is an example:
System.Soft_Links'Elab_Body;
E14 := True;
System.Secondary_Stack'Elab_Body;
E18 := True;
System.Exception_Table'Elab_Body;
E24 := True;
Ada.Io_Exceptions'Elab_Spec;
E67 := True;
Ada.Tags'Elab_Spec;
Ada.Streams'Elab_Spec;
E43 := True;
Interfaces.C'Elab_Spec;
E69 := True;
System.Finalization_Root'Elab_Spec;
E60 := True;
System.Os_Lib'Elab_Body;
E71 := True;
System.Finalization_Implementation'Elab_Spec;
System.Finalization_Implementation'Elab_Body;
E62 := True;
Ada.Finalization'Elab_Spec;
E58 := True;
Ada.Finalization.List_Controller'Elab_Spec;
E76 := True;
System.File_Control_Block'Elab_Spec;
E74 := True;
System.File_Io'Elab_Body;
E56 := True;
Ada.Tags'Elab_Body;
E45 := True;
Ada.Text_Io'Elab_Spec;
Ada.Text_Io'Elab_Body;
E07 := True;
Here Elab_Spec elaborates the spec
and Elab_Body elaborates the body. The assignments to the E`xx' flags
flag that the corresponding body is now elaborated.
You can also ask the binder to generate a more readable list of the elaboration order using the -l switch when invoking the binder. Here is an example of the output generated by this switch:
ada (spec)
interfaces (spec)
system (spec)
system.case_util (spec)
system.case_util (body)
system.concat_2 (spec)
system.concat_2 (body)
system.concat_3 (spec)
system.concat_3 (body)
system.htable (spec)
system.parameters (spec)
system.parameters (body)
system.crtl (spec)
interfaces.c_streams (spec)
interfaces.c_streams (body)
system.restrictions (spec)
system.restrictions (body)
system.standard_library (spec)
system.exceptions (spec)
system.exceptions (body)
system.storage_elements (spec)
system.storage_elements (body)
system.secondary_stack (spec)
system.stack_checking (spec)
system.stack_checking (body)
system.string_hash (spec)
system.string_hash (body)
system.htable (body)
system.strings (spec)
system.strings (body)
system.traceback (spec)
system.traceback (body)
system.traceback_entries (spec)
system.traceback_entries (body)
ada.exceptions (spec)
ada.exceptions.last_chance_handler (spec)
system.soft_links (spec)
system.soft_links (body)
ada.exceptions.last_chance_handler (body)
system.secondary_stack (body)
system.exception_table (spec)
system.exception_table (body)
ada.io_exceptions (spec)
ada.tags (spec)
ada.streams (spec)
interfaces.c (spec)
interfaces.c (body)
system.finalization_root (spec)
system.finalization_root (body)
system.memory (spec)
system.memory (body)
system.standard_library (body)
system.os_lib (spec)
system.os_lib (body)
system.unsigned_types (spec)
system.stream_attributes (spec)
system.stream_attributes (body)
system.finalization_implementation (spec)
system.finalization_implementation (body)
ada.finalization (spec)
ada.finalization (body)
ada.finalization.list_controller (spec)
ada.finalization.list_controller (body)
system.file_control_block (spec)
system.file_io (spec)
system.file_io (body)
system.val_uns (spec)
system.val_util (spec)
system.val_util (body)
system.val_uns (body)
system.wch_con (spec)
system.wch_con (body)
system.wch_cnv (spec)
system.wch_jis (spec)
system.wch_jis (body)
system.wch_cnv (body)
system.wch_stw (spec)
system.wch_stw (body)
ada.tags (body)
ada.exceptions (body)
ada.text_io (spec)
ada.text_io (body)
text_io (spec)
gdbstr (body)
If you need to write low-level software that interacts directly with the hardware, Ada provides two ways to incorporate assembly language code into your program. First, you can import and invoke external routines written in assembly language, an Ada feature fully supported by GNAT. However, for small sections of code it may be simpler or more efficient to include assembly language statements directly in your Ada source program, using the facilities of the implementation-defined package System.Machine_Code, which incorporates the gcc Inline Assembler. The Inline Assembler approach offers a number of advantages, including the following:
This appendix presents a series of examples to show you how to use the Inline Assembler. Although it focuses on the Intel x86, the general approach applies also to other processors. It is assumed that you are familiar with Ada and with assembly language programming.
The assembler used by GNAT and gcc is based not on the Intel assembly language, but rather on a language that descends from the AT&T Unix assembler `as' (and which is often referred to as 'AT&T syntax'). The following table summarizes the main features of `as' syntax and points out the differences from the Intel conventions. See the gcc `as' and `gas' (an `as' macro pre-processor) documentation for further information.
`Register names'
gcc / `as': Prefix with '%'; for example %eax
Intel: No extra punctuation; for example eax
`Immediate operand'
gcc / `as': Prefix with '$'; for example $4
Intel: No extra punctuation; for example 4
`Address'
gcc / `as': Prefix with '$'; for example $loc
Intel: No extra punctuation; for example loc
`Memory contents'
gcc / `as': No extra punctuation; for example loc
Intel: Square brackets; for example [loc]
`Register contents'
gcc / `as': Parentheses; for example (%eax)
Intel: Square brackets; for example [eax]
`Hexadecimal numbers'
gcc / `as': Leading '0x' (C language syntax); for example 0xA0
Intel: Trailing 'h'; for example A0h
`Operand size'
gcc / `as': Explicit in op code; for example movw to move a 16-bit word
Intel: Implicit, deduced by assembler; for example mov
`Instruction repetition'
gcc / `as': Split into two lines; for example
rep
stosl
Intel: Keep on one line; for example rep stosl
`Order of operands'
gcc / `as': Source first; for example movw $4, %eax
Intel: Destination first; for example mov eax, 4
The following example will generate a single assembly language statement, nop, which does nothing. Despite its lack of run-time effect, the example will be useful in illustrating the basics of the Inline Assembler facility.
with System.Machine_Code; use System.Machine_Code;
procedure Nothing is
begin
Asm ("nop");
end Nothing;
Asm is a procedure declared in package System.Machine_Code; here it takes one parameter, a `template string' that must be a static expression and that will form the generated instruction. Asm may be regarded as a compile-time procedure that parses the template string and additional parameters (none here), from which it generates a sequence of assembly language instructions.
The examples in this chapter will illustrate several of the forms for invoking Asm; a complete specification of the syntax is found in the Machine_Code_Insertions section of the GNAT Reference Manual.
Under the standard GNAT conventions, the Nothing procedure
should be in a file named nothing.adb.
You can build the executable in the usual way:
$ gnatmake nothing
However, the interesting aspect of this example is not its run-time behavior but rather the generated assembly code. To see this output, invoke the compiler as follows:
$ gcc -c -S -fomit-frame-pointer -gnatp nothing.adb
where the options are:
-c-S-fomit-frame-pointer-gnatpThis gives a human-readable assembler version of the code. The resulting
file will have the same name as the Ada source file, but with a .s
extension. In our example, the file nothing.s has the following
contents:
.file "nothing.adb"
gcc2_compiled.:
___gnu_compiled_ada:
.text
.align 4
.globl __ada_nothing
__ada_nothing:
#APP
nop
#NO_APP
jmp L1
.align 2,0x90
L1:
ret
The assembly code you included is clearly indicated by the compiler, between the #APP and #NO_APP delimiters. The character before the 'APP' and 'NOAPP' can differ on different targets. For example, GNU/Linux uses '#APP' while on NT you will see '/APP'.
If you make a mistake in your assembler code (such as using the wrong size modifier, or using a wrong operand for the instruction) GNAT will report this error in a temporary file, which will be deleted when the compilation is finished. Generating an assembler file will help in such cases, since you can assemble this file separately using the `as' assembler that comes with gcc.
Assembling the file using the command
$ as nothing.s
will give you error messages whose lines correspond to the assembler
input file, so you can easily find and correct any mistakes you made.
If there are no errors, `as' will generate an object file
nothing.out.
The examples in this section, showing how to access the processor flags, illustrate how to specify the destination operands for assembly language statements.
with Interfaces; use Interfaces;
with Ada.Text_IO; use Ada.Text_IO;
with System.Machine_Code; use System.Machine_Code;
procedure Get_Flags is
Flags : Unsigned_32;
use ASCII;
begin
Asm ("pushfl" & LF & HT & -- push flags on stack
"popl %%eax" & LF & HT & -- load eax with flags
"movl %%eax, %0", -- store flags in variable
Outputs => Unsigned_32'Asm_Output ("=g", Flags));
Put_Line ("Flags register:" & Flags'Img);
end Get_Flags;
In order to have a nicely aligned assembly listing, we have separated multiple assembler statements in the Asm template string with linefeed (ASCII.LF) and horizontal tab (ASCII.HT) characters. The resulting section of the assembly output file is:
#APP
pushfl
popl %eax
movl %eax, -40(%ebp)
#NO_APP
It would have been legal to write the Asm invocation as:
Asm ("pushfl popl %%eax movl %%eax, %0")
but in the generated assembler file, this would come out as:
#APP
pushfl popl %eax movl %eax, -40(%ebp)
#NO_APP
which is not so convenient for the human reader.
We use Ada comments at the end of each line to explain what the assembler instructions actually do. This is a useful convention.
When writing Inline Assembler instructions, you need to precede each register and variable name with a percent sign. Since the assembler already requires a percent sign at the beginning of a register name, you need two consecutive percent signs for such names in the Asm template string, thus %%eax. In the generated assembly code, one of the percent signs will be stripped off.
Names such as %0, %1, %2, etc., denote input or output variables: operands you later define using Input or Output parameters to Asm. An output variable is illustrated in the third statement in the Asm template string:
movl %%eax, %0
The intent is to store the contents of the eax register in a variable that can be accessed in Ada. Simply writing movl %%eax, Flags would not necessarily work, since the compiler might optimize by using a register to hold Flags, and the expansion of the movl instruction would not be aware of this optimization. The solution is not to store the result directly but rather to advise the compiler to choose the correct operand form; that is the purpose of the %0 output variable.
Information about the output variable is supplied in the Outputs parameter to Asm:
Outputs => Unsigned_32'Asm_Output ("=g", Flags));
The output is defined by the Asm_Output attribute of the target type; the general format is
Type'Asm_Output (constraint_string, variable_name)
The constraint string directs the compiler how to store/access the associated variable. In the example
Unsigned_32'Asm_Output ("=m", Flags);
the "m" (memory) constraint tells the compiler that the variable Flags should be stored in a memory variable, thus preventing the optimizer from keeping it in a register. In contrast,
Unsigned_32'Asm_Output ("=r", Flags);
uses the "r" (register) constraint, telling the compiler to store the variable in a register.
If the constraint is preceded by the equal character '=', it tells the compiler that the variable will be used to store data into it.
In the Get_Flags example, we used the "g" (global) constraint, allowing the optimizer to choose whatever it deems best.
There are a fairly large number of constraints, but the ones that are most useful (for the Intel x86 processor) are the following:
`=' output constraint
`g' global (i.e., can be stored anywhere)
`m' in memory
`I' a constant
`a' use eax
`b' use ebx
`c' use ecx
`d' use edx
`S' use esi
`D' use edi
`r' use one of eax, ebx, ecx or edx
`q' use one of eax, ebx, ecx, edx, esi or edi
The full set of constraints is described in the gcc and `as' documentation; note that it is possible to combine certain constraints in one constraint string.
You specify the association of an output variable with an assembler operand
through the %`n' notation, where `n' is a non-negative
integer. Thus in
Asm ("pushfl" & LF & HT & -- push flags on stack
"popl %%eax" & LF & HT & -- load eax with flags
"movl %%eax, %0", -- store flags in variable
Outputs => Unsigned_32'Asm_Output ("=g", Flags));
%0 will be replaced in the expanded code by the appropriate operand, whatever the compiler decided for the Flags variable.
In general, you may have any number of output variables:
For example:
Asm ("movl %%eax, %0" & LF & HT &
"movl %%ebx, %1" & LF & HT &
"movl %%ecx, %2",
Outputs => (Unsigned_32'Asm_Output ("=g", Var_A), -- %0 = Var_A
Unsigned_32'Asm_Output ("=g", Var_B), -- %1 = Var_B
Unsigned_32'Asm_Output ("=g", Var_C))); -- %2 = Var_C
where Var_A, Var_B, and Var_C are variables in the Ada program.
As a variation on the Get_Flags example, we can use the constraints string to direct the compiler to store the eax register into the Flags variable, instead of including the store instruction explicitly in the Asm template string:
with Interfaces; use Interfaces;
with Ada.Text_IO; use Ada.Text_IO;
with System.Machine_Code; use System.Machine_Code;
procedure Get_Flags_2 is
Flags : Unsigned_32;
use ASCII;
begin
Asm ("pushfl" & LF & HT & -- push flags on stack
"popl %%eax", -- save flags in eax
Outputs => Unsigned_32'Asm_Output ("=a", Flags));
Put_Line ("Flags register:" & Flags'Img);
end Get_Flags_2;
The "a" constraint tells the compiler that the Flags variable will come from the eax register. Here is the resulting code:
#APP
pushfl
popl %eax
#NO_APP
movl %eax,-40(%ebp)
The compiler generated the store of eax into Flags after expanding the assembler code.
Actually, there was no need to pop the flags into the eax register; more simply, we could just pop the flags directly into the program variable:
with Interfaces; use Interfaces;
with Ada.Text_IO; use Ada.Text_IO;
with System.Machine_Code; use System.Machine_Code;
procedure Get_Flags_3 is
Flags : Unsigned_32;
use ASCII;
begin
Asm ("pushfl" & LF & HT & -- push flags on stack
"pop %0", -- save flags in Flags
Outputs => Unsigned_32'Asm_Output ("=g", Flags));
Put_Line ("Flags register:" & Flags'Img);
end Get_Flags_3;
The example in this section illustrates how to specify the source operands for assembly language statements. The program simply increments its input value by 1:
with Interfaces; use Interfaces;
with Ada.Text_IO; use Ada.Text_IO;
with System.Machine_Code; use System.Machine_Code;
procedure Increment is
function Incr (Value : Unsigned_32) return Unsigned_32 is
Result : Unsigned_32;
begin
Asm ("incl %0",
Outputs => Unsigned_32'Asm_Output ("=a", Result),
Inputs => Unsigned_32'Asm_Input ("a", Value));
return Result;
end Incr;
Value : Unsigned_32;
begin
Value := 5;
Put_Line ("Value before is" & Value'Img);
Value := Incr (Value);
Put_Line ("Value after is" & Value'Img);
end Increment;
The Outputs parameter to Asm specifies that the result will be in the eax register and that it is to be stored in the Result variable.
The Inputs parameter looks much like the Outputs parameter, but with an Asm_Input attribute. The "=" constraint, indicating an output value, is not present.
You can have multiple input variables, in the same way that you can have more than one output variable.
The parameter count (%0, %1) etc, still starts at the first output statement, and continues with the input statements.
Just as the Outputs parameter causes the register to be stored into the target variable after execution of the assembler statements, so does the Inputs parameter cause its variable to be loaded into the register before execution of the assembler statements.
Thus the effect of the Asm invocation is:
The resulting assembler file (with `-O2' optimization) contains:
_increment__incr.1:
subl $4,%esp
movl 8(%esp),%eax
#APP
incl %eax
#NO_APP
movl %eax,%edx
movl %ecx,(%esp)
addl $4,%esp
ret
For a short subprogram such as the Incr function in the previous section, the overhead of the call and return (creating / deleting the stack frame) can be significant, compared to the amount of code in the subprogram body. A solution is to apply Ada's Inline pragma to the subprogram, which directs the compiler to expand invocations of the subprogram at the point(s) of call, instead of setting up a stack frame for out-of-line calls. Here is the resulting program:
with Interfaces; use Interfaces;
with Ada.Text_IO; use Ada.Text_IO;
with System.Machine_Code; use System.Machine_Code;
procedure Increment_2 is
function Incr (Value : Unsigned_32) return Unsigned_32 is
Result : Unsigned_32;
begin
Asm ("incl %0",
Outputs => Unsigned_32'Asm_Output ("=a", Result),
Inputs => Unsigned_32'Asm_Input ("a", Value));
return Result;
end Incr;
pragma Inline (Increment);
Value : Unsigned_32;
begin
Value := 5;
Put_Line ("Value before is" & Value'Img);
Value := Increment (Value);
Put_Line ("Value after is" & Value'Img);
end Increment_2;
Compile the program with both optimization (`-O2') and inlining (`-gnatn') enabled.
The Incr function is still compiled as usual, but at the point in Increment where our function used to be called:
pushl %edi
call _increment__incr.1
the code for the function body directly appears:
movl %esi,%eax
#APP
incl %eax
#NO_APP
movl %eax,%edx
thus saving the overhead of stack frame setup and an out-of-line call.
This section describes two important parameters to the Asm procedure: Clobber, which identifies register usage; and Volatile, which inhibits unwanted optimizations.
One of the dangers of intermixing assembly language and a compiled language such as Ada is that the compiler needs to be aware of which registers are being used by the assembly code. In some cases, such as the earlier examples, the constraint string is sufficient to indicate register usage (e.g., "a" for the eax register). But more generally, the compiler needs an explicit identification of the registers that are used by the Inline Assembly statements.
Using a register that the compiler doesn't know about could be a side effect of an instruction (like mull storing its result in both eax and edx). It can also arise from explicit register usage in your assembly code; for example:
Asm ("movl %0, %%ebx" & LF & HT &
"movl %%ebx, %1",
Outputs => Unsigned_32'Asm_Output ("=g", Var_Out),
Inputs => Unsigned_32'Asm_Input ("g", Var_In));
where the compiler (since it does not analyze the Asm template string) does not know you are using the ebx register.
In such cases you need to supply the Clobber parameter to Asm, to identify the registers that will be used by your assembly code:
Asm ("movl %0, %%ebx" & LF & HT &
"movl %%ebx, %1",
Outputs => Unsigned_32'Asm_Output ("=g", Var_Out),
Inputs => Unsigned_32'Asm_Input ("g", Var_In),
Clobber => "ebx");
The Clobber parameter is a static string expression specifying the register(s) you are using. Note that register names are `not' prefixed by a percent sign. Also, if more than one register is used then their names are separated by commas; e.g., "eax, ebx"
The Clobber parameter has several additional uses:
Compiler optimizations in the presence of Inline Assembler may sometimes have unwanted effects. For example, when an Asm invocation with an input variable is inside a loop, the compiler might move the loading of the input variable outside the loop, regarding it as a one-time initialization.
If this effect is not desired, you can disable such optimizations by setting the Volatile parameter to True; for example:
Asm ("movl %0, %%ebx" & LF & HT &
"movl %%ebx, %1",
Outputs => Unsigned_32'Asm_Output ("=g", Var_Out),
Inputs => Unsigned_32'Asm_Input ("g", Var_In),
Clobber => "ebx",
Volatile => True);
By default, Volatile is set to False unless there is no Outputs parameter.
Although setting Volatile to True prevents unwanted optimizations, it will also disable other optimizations that might be important for efficiency. In general, you should set Volatile to True only if the compiler's optimizations have created problems.
Version 1.3, 3 November 2008
Copyright 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc
<http://fsf.org/>
Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed.
`Preamble'
The purpose of this License is to make a manual, textbook, or other functional and useful document "free" in the sense of freedom: to assure everyone the effective freedom to copy and redistribute it, with or without modifying it, either commercially or noncommercially. Secondarily, this License preserves for the author and publisher a way to get credit for their work, while not being considered responsible for modifications made by others.
This License is a kind of "copyleft", which means that derivative works of the document must themselves be free in the same sense. It complements the GNU General Public License, which is a copyleft license designed for free software.
We have designed this License in order to use it for manuals for free software, because free software needs free documentation: a free program should come with manuals providing the same freedoms that the software does. But this License is not limited to software manuals; it can be used for any textual work, regardless of subject matter or whether it is published as a printed book. We recommend this License principally for works whose purpose is instruction or reference.
`1. APPLICABILITY AND DEFINITIONS'
This License applies to any manual or other work, in any medium, that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. Such a notice grants a world-wide, royalty-free license, unlimited in duration, to use that work under the conditions stated herein. The `Document', below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as "`you'". You accept the license if you copy, modify or distribute the work in a way requiring permission under copyright law.
A "`Modified Version'" of the Document means any work containing the Document or a portion of it, either copied verbatim, or with modifications and/or translated into another language.
A "`Secondary Section'" is a named appendix or a front-matter section of the Document that deals exclusively with the relationship of the publishers or authors of the Document to the Document's overall subject (or to related matters) and contains nothing that could fall directly within that overall subject. (Thus, if the Document is in part a textbook of mathematics, a Secondary Section may not explain any mathematics.) The relationship could be a matter of historical connection with the subject or with related matters, or of legal, commercial, philosophical, ethical or political position regarding them.
The "`Invariant Sections'" are certain Secondary Sections whose titles are designated, as being those of Invariant Sections, in the notice that says that the Document is released under this License. If a section does not fit the above definition of Secondary then it is not allowed to be designated as Invariant. The Document may contain zero Invariant Sections. If the Document does not identify any Invariant Sections then there are none.
The "`Cover Texts'" are certain short passages of text that are listed, as Front-Cover Texts or Back-Cover Texts, in the notice that says that the Document is released under this License. A Front-Cover Text may be at most 5 words, and a Back-Cover Text may be at most 25 words.
A "`Transparent'" copy of the Document means a machine-readable copy, represented in a format whose specification is available to the general public, that is suitable for revising the document straightforwardly with generic text editors or (for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor, and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters. A copy made in an otherwise Transparent file format whose markup, or absence of markup, has been arranged to thwart or discourage subsequent modification by readers is not Transparent. An image format is not Transparent if used for any substantial amount of text. A copy that is not "Transparent" is called `Opaque'.
Examples of suitable formats for Transparent copies include plain ASCII without markup, Texinfo input format, LaTeX input format, SGML or XML using a publicly available DTD, and standard-conforming simple HTML, PostScript or PDF designed for human modification. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats include proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML, PostScript or PDF produced by some word processors for output purposes only.
The "`Title Page'" means, for a printed book, the title page itself, plus such following pages as are needed to hold, legibly, the material this License requires to appear in the title page. For works in formats which do not have any title page as such, "Title Page" means the text near the most prominent appearance of the work's title, preceding the beginning of the body of the text.
The "`publisher'" means any person or entity that distributes copies of the Document to the public.
A section "`Entitled XYZ'" means a named subunit of the Document whose title either is precisely XYZ or contains XYZ in parentheses following text that translates XYZ in another language. (Here XYZ stands for a specific section name mentioned below, such as "`Acknowledgements'", "`Dedications'", "`Endorsements'", or "`History'".) To "`Preserve the Title'" of such a section when you modify the Document means that it remains a section "Entitled XYZ" according to this definition.
The Document may include Warranty Disclaimers next to the notice which states that this License applies to the Document. These Warranty Disclaimers are considered to be included by reference in this License, but only as regards disclaiming warranties: any other implication that these Warranty Disclaimers may have is void and has no effect on the meaning of this License.
`2. VERBATIM COPYING'
You may copy and distribute the Document in any medium, either commercially or noncommercially, provided that this License, the copyright notices, and the license notice saying this License applies to the Document are reproduced in all copies, and that you add no other conditions whatsoever to those of this License. You may not use technical measures to obstruct or control the reading or further copying of the copies you make or distribute. However, you may accept compensation in exchange for copies. If you distribute a large enough number of copies you must also follow the conditions in section 3.
You may also lend copies, under the same conditions stated above, and you may publicly display copies.
`3. COPYING IN QUANTITY'
If you publish printed copies (or copies in media that commonly have printed covers) of the Document, numbering more than 100, and the Document's license notice requires Cover Texts, you must enclose the copies in covers that carry, clearly and legibly, all these Cover Texts: Front-Cover Texts on the front cover, and Back-Cover Texts on the back cover. Both covers must also clearly and legibly identify you as the publisher of these copies. The front cover must present the full title with all words of the title equally prominent and visible. You may add other material on the covers in addition. Copying with changes limited to the covers, as long as they preserve the title of the Document and satisfy these conditions, can be treated as verbatim copying in other respects.
If the required texts for either cover are too voluminous to fit legibly, you should put the first ones listed (as many as fit reasonably) on the actual cover, and continue the rest onto adjacent pages.
If you publish or distribute Opaque copies of the Document numbering more than 100, you must either include a machine-readable Transparent copy along with each Opaque copy, or state in or with each Opaque copy a computer-network location from which the general network-using public has access to download using public-standard network protocols a complete Transparent copy of the Document, free of added material. If you use the latter option, you must take reasonably prudent steps, when you begin distribution of Opaque copies in quantity, to ensure that this Transparent copy will remain thus accessible at the stated location until at least one year after the last time you distribute an Opaque copy (directly or through your agents or retailers) of that edition to the public.
It is requested, but not required, that you contact the authors of the Document well before redistributing any large number of copies, to give them a chance to provide you with an updated version of the Document.
`4. MODIFICATIONS'
You may copy and distribute a Modified Version of the Document under the conditions of sections 2 and 3 above, provided that you release the Modified Version under precisely this License, with the Modified Version filling the role of the Document, thus licensing distribution and modification of the Modified Version to whoever possesses a copy of it. In addition, you must do these things in the Modified Version:
If the Modified Version includes new front-matter sections or appendices that qualify as Secondary Sections and contain no material copied from the Document, you may at your option designate some or all of these sections as invariant. To do this, add their titles to the list of Invariant Sections in the Modified Version's license notice. These titles must be distinct from any other section titles.
You may add a section Entitled "Endorsements", provided it contains nothing but endorsements of your Modified Version by various parties—for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard.
You may add a passage of up to five words as a Front-Cover Text, and a passage of up to 25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified Version. Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or through arrangements made by) any one entity. If the Document already includes a cover text for the same cover, previously added by you or by arrangement made by the same entity you are acting on behalf of, you may not add another; but you may replace the old one, on explicit permission from the previous publisher that added the old one.
The author(s) and publisher(s) of the Document do not by this License give permission to use their names for publicity for or to assert or imply endorsement of any Modified Version.
`5. COMBINING DOCUMENTS'
You may combine the Document with other documents released under this License, under the terms defined in section 4 above for modified versions, provided that you include in the combination all of the Invariant Sections of all of the original documents, unmodified, and list them all as Invariant Sections of your combined work in its license notice, and that you preserve all their Warranty Disclaimers.
The combined work need only contain one copy of this License, and multiple identical Invariant Sections may be replaced with a single copy. If there are multiple Invariant Sections with the same name but different contents, make the title of each such section unique by adding at the end of it, in parentheses, the name of the original author or publisher of that section if known, or else a unique number. Make the same adjustment to the section titles in the list of Invariant Sections in the license notice of the combined work.
In the combination, you must combine any sections Entitled "History" in the various original documents, forming one section Entitled "History"; likewise combine any sections Entitled "Acknowledgements", and any sections Entitled "Dedications". You must delete all sections Entitled "Endorsements".
`6. COLLECTIONS OF DOCUMENTS'
You may make a collection consisting of the Document and other documents released under this License, and replace the individual copies of this License in the various documents with a single copy that is included in the collection, provided that you follow the rules of this License for verbatim copying of each of the documents in all other respects.
You may extract a single document from such a collection, and distribute it individually under this License, provided you insert a copy of this License into the extracted document, and follow this License in all other respects regarding verbatim copying of that document.
`7. AGGREGATION WITH INDEPENDENT WORKS'
A compilation of the Document or its derivatives with other separate and independent documents or works, in or on a volume of a storage or distribution medium, is called an "aggregate" if the copyright resulting from the compilation is not used to limit the legal rights of the compilation's users beyond what the individual works permit. When the Document is included in an aggregate, this License does not apply to the other works in the aggregate which are not themselves derivative works of the Document.
If the Cover Text requirement of section 3 is applicable to these copies of the Document, then if the Document is less than one half of the entire aggregate, the Document's Cover Texts may be placed on covers that bracket the Document within the aggregate, or the electronic equivalent of covers if the Document is in electronic form. Otherwise they must appear on printed covers that bracket the whole aggregate.
`8. TRANSLATION'
Translation is considered a kind of modification, so you may distribute translations of the Document under the terms of section 4. Replacing Invariant Sections with translations requires special permission from their copyright holders, but you may include translations of some or all Invariant Sections in addition to the original versions of these Invariant Sections. You may include a translation of this License, and all the license notices in the Document, and any Warranty Disclaimers, provided that you also include the original English version of this License and the original versions of those notices and disclaimers. In case of a disagreement between the translation and the original version of this License or a notice or disclaimer, the original version will prevail.
If a section in the Document is Entitled "Acknowledgements", "Dedications", or "History", the requirement (section 4) to Preserve its Title (section 1) will typically require changing the actual title.
`9. TERMINATION'
You may not copy, modify, sublicense, or distribute the Document except as expressly provided under this License. Any attempt otherwise to copy, modify, sublicense, or distribute it is void, and will automatically terminate your rights under this License.
However, if you cease all violation of this License, then your license from a particular copyright holder is reinstated (a) provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the copyright holder fails to notify you of the violation by some reasonable means prior to 60 days after the cessation.
Moreover, your license from a particular copyright holder is reinstated permanently if the copyright holder notifies you of the violation by some reasonable means, this is the first time you have received notice of violation of this License (for any work) from that copyright holder, and you cure the violation prior to 30 days after your receipt of the notice.
Termination of your rights under this section does not terminate the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, receipt of a copy of some or all of the same material does not give you any rights to use it.
`10. FUTURE REVISIONS OF THIS LICENSE'
The Free Software Foundation may publish new, revised versions
of the GNU Free Documentation License from time to time. Such new
versions will be similar in spirit to the present version, but may
differ in detail to address new problems or concerns. See
<http://www.gnu.org/copyleft/>.
Each version of the License is given a distinguishing version number. If the Document specifies that a particular numbered version of this License "or any later version" applies to it, you have the option of following the terms and conditions either of that specified version or of any later version that has been published (not as a draft) by the Free Software Foundation. If the Document does not specify a version number of this License, you may choose any version ever published (not as a draft) by the Free Software Foundation. If the Document specifies that a proxy can decide which future versions of this License can be used, that proxy's public statement of acceptance of a version permanently authorizes you to choose that version for the Document.
`11. RELICENSING'
"Massive Multiauthor Collaboration Site" (or "MMC Site") means any World Wide Web server that publishes copyrightable works and also provides prominent facilities for anybody to edit those works. A public wiki that anybody can edit is an example of such a server. A "Massive Multiauthor Collaboration" (or "MMC") contained in the site means any set of copyrightable works thus published on the MMC site.
"CC-BY-SA" means the Creative Commons Attribution-Share Alike 3.0 license published by Creative Commons Corporation, a not-for-profit corporation with a principal place of business in San Francisco, California, as well as future copyleft versions of that license published by that same organization.
"Incorporate" means to publish or republish a Document, in whole or in part, as part of another Document.
An MMC is "eligible for relicensing" if it is licensed under this License, and if all works that were first published under this License somewhere other than this MMC, and subsequently incorporated in whole or in part into the MMC, (1) had no cover texts or invariant sections, and (2) were thus incorporated prior to November 1, 2008.
The operator of an MMC Site may republish an MMC contained in the site under CC-BY-SA on the same site at any time before August 1, 2009, provided the MMC is eligible for relicensing.
`ADDENDUM: How to use this License for your documents'
To use this License in a document you have written, include a copy of the License in the document and put the following copyright and license notices just after the title page:
Copyright © YEAR YOUR NAME. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled "GNU Free Documentation License".
If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, replace the "with ... Texts." line with this:
with the Invariant Sections being LIST THEIR TITLES, with the Front-Cover Texts being LIST, and with the Back-Cover Texts being LIST.
If you have Invariant Sections without Cover Texts, or some other combination of the three, merge those two alternatives to suit the situation.
If your document contains nontrivial examples of program code, we recommend releasing these examples in parallel under your choice of free software license, such as the GNU General Public License, to permit their use in free software.