GNAT User's Guide
*****************

   GNAT User's Guide

   GNAT, The GNU Ada 95 Compiler

   GNAT Version 3.14p

   Date: 2001/05/10 16:08:26

   Ada Core Technologies, Inc.

   (C) Copyright 1995-2000, Ada Core Technologies, Inc.

   Permission is granted to make and distribute verbatim copies of this
manual provided the copyright notice and this permission notice are
preserved on all copies.

   Silicon Graphics and IRIS are registered trademarks and IRIX is a
trademark of Silicon Graphics, Inc.

   IBM PC is a trademark of International Business Machines Corporation.

   UNIX is a registered trademark of AT&T Bell Laboratories.

   The following are trademarks of Compaq Computers: DEC, DEC Ada,
DECthreads, DIGITAL, DECset, OpenVMS, and VAX.

   The following are trademarks of Microsoft Corporation: Windows NT,
Windows 95, Windows 98.

   The following are trademarks of Wind River Systems: VxWorks, Tornado.

About This Guide
****************

This guide describes the use of GNAT, a compiler and software
development toolset for the full Ada 95 programming language.  It
describes the features of the compiler and tools, and details how to
use them to build Ada 95 applications.

What This Guide Contains
========================

This guide contains the following chapters:
   * *Note Getting Started with GNAT::, describes how to get started
     compiling and running Ada programs with the GNAT Ada programming
     environment.

   * *Note The GNAT Compilation Model::, describes the compilation
     model used by GNAT.

   * *Note Compiling Using gcc::, describes how to compile Ada programs
     with `gcc', the Ada compiler.

   * *Note Binding Using gnatbind::, describes how to perform binding
     of Ada programs with `gnatbind', the GNAT binding utility.

   * *Note Linking Using gnatlink::, describes `gnatlink', a program
     that provides for linking using the GNAT run-time library to
     construct a program. `gnatlink' can also incorporate foreign
     language object units into the executable.

   * *Note The GNAT Make Program gnatmake::, describes `gnatmake', a
     utility that automatically determines the set of sources needed by
     an Ada compilation unit, and executes the necessary compilations
     binding and link.

   * *Note Renaming Files Using gnatchop::, describes `gnatchop', a
     utility that allows you to preprocess a file that contains Ada
     source code, and split it into one or more new files, one for each
     compilation unit.

   * *Note The Cross-Referencing Tools gnatxref and gnatfind::,
     discusses `gnatxref' and `gnatfind', two tools that provide an easy
     way to navigate through sources.

   * *Note File Name Krunching Using gnatkr::, describes the `gnatkr'
     file name krunching utility, used to handle shortened file names
     on operating systems with a limit on the length of names.

   * *Note Preprocessing Using gnatprep::, describes `gnatprep', a
     preprocessor utility that allows a single source file to be used to
     generate multiple or parameterized source files, by means of macro
     substitution.

   * *Note The GNAT Library Browser gnatls::, describes `gnatls', a
     utility that displays information about compiled units, including
     dependences on the corresponding sources files, and consistency of
     compilations.

   * *Note GNAT and Libraries::, describes the process of creating and
     using Libraries with GNAT. It also describes how to recompile the
     GNAT run-time library.

   * *Note Using the GNU make Utility::, describes some techniques for
     using the GNAT toolset in Makefiles.

   * *Note Finding Memory Problems with gnatmem::, describes `gnatmem',
     a utility that monitors dynamic allocation and deallocation
     activity in a program, and displays information about incorrect
     deallocations and sources of possible memory leaks.

   * *Note Finding Memory Problems with GNAT Debug Pool::, describes
     how to use the GNAT-specific Debug Pool in order to detect as
     early as possible the use of incorrect memory references.

   * *Note Creating Sample Bodies Using gnatstub::, discusses
     `gnatstub', a utility that generates empty but compilable bodies
     for library units.

   * *Note Reducing the Size of Ada Executables with gnatelim::,
     describes `gnatelim', a tool which detects unused subprograms and
     helps the compiler to create a smaller executable for the program.

   * *Note Other Utility Programs::, discusses several other GNAT
     utilities, including `gnatpsta' and `gnatpsys'.

   * *Note Running and Debugging Ada Programs::, describes how to run
     and debug Ada programs.

   * *Note Building Mixed Ada & C++ Programs::, gives hints on how to
     interface with c++.

   * *Note Performance Considerations::, reviews the trade offs between
     using defaults or options in program development.

What You Should Know before Reading This Guide
==============================================

This user's guide assumes that you are familiar with Ada 95 language, as
described in the International Standard ANSI/ISO/IEC-8652:1995, Jan
1995.

Related Information
===================

For further information about related tools, refer to the following
documents:

   * `GNAT Reference Manual', which contains all reference material for
     the GNAT implementation of Ada 95.

   * `Ada 95 Language Reference Manual', which contains all reference
     material for the Ada 95 programming language.

   * `Debugging with GDB' contains all details on the use of the GNU
     source-level debugger.

   * `GNU Emacs Manual' contains full information on the extensible
     editor and programming environment Emacs.

Conventions
===========

Following are examples of the typographical and graphic conventions used
in this guide:

   * `Functions', `utility program names', `standard names', and
     `classes'.

   * `Option flags'

   * `File Names', `button names', and `field names'.

   * VARIABLES.

   * *Emphasis*.

   * [optional information or parameters]

   * Examples are described by text
          and then shown this way.

Commands that are entered by the user are preceded in this manual by the
characters "`$ '" (dollar sign followed by space). If your system uses
this sequence as a prompt, then the commands will appear exactly as you
see them in the manual. If your system uses some other prompt, then the
command will appear with the `$' replaced by whatever prompt character
you are using.

Getting Started with GNAT
*************************

This chapter describes some simple ways of using GNAT to build
executable Ada programs.

Running GNAT
============

Three steps are needed to create an executable file from an Ada source
file:

  1. The source file(s) must be compiled.

  2. The file(s) must be bound using the GNAT binder.

  3. All appropriate object files must be linked to produce an
     executable.

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.

Running a Simple Ada Program
============================

Any text editor may be used to prepare an Ada program. If `glide' is
used, the optional Ada mode may be helpful in laying out the program.
The program text is a normal text file. We will suppose 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, and whose 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' (*note 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 (*note Renaming Files
Using 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 95 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 a 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', which
is a master program that invokes all of 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 a consistent compilation is ensured.

     $ 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.

Running a Program with Multiple Units
=====================================

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:

`greetings.ads'
     spec of package `Greetings'

`greetings.adb'
     body of package `Greetings'

`gmain.adb'
     body of main program

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.

Using the `gnatmake' Utility
============================

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 95 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.

The GNAT Compilation Model
**************************

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 library. The model is initially described without
reference to the library-based model. If you have not previously used an
Ada compiler, you need only read the first part of this chapter. The
last section describes and discusses the differences between the GNAT
model and the traditional Ada compiler models. If you have used other
Ada compilers, this section will help you to understand those
differences, and the advantages of the GNAT model.

Source Representation
=====================

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 (*note Foreign Language Representation::.  for
support of non-USA character sets). The format effector characters are
represented using their standard ASCII encodings, as follows:

`VT'
     Vertical tab, `16#0B#'

`HT'
     Horizontal tab, `16#09#'

`CR'
     Carriage return, `16#0D#'

`LF'
     Line feed, `16#0A#'

`FF'
     Form 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
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.

Foreign Language Representation
===============================

GNAT supports the standard character sets defined in Ada 95 as well as
several other non-standard character sets for use in localized versions
of the compiler (*note Character Set Control::.).

Latin-1
-------

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.

Other 8-Bit Codes
-----------------

GNAT also supports several other 8-bit coding schemes:

Latin-2
     Latin-2 letters allowed in identifiers, with uppercase and
     lowercase equivalence.

Latin-3
     Latin-3 letters allowed in identifiers, with uppercase and
     lowercase equivalence.

Latin-4
     Latin-4 letters allowed in identifiers, with uppercase and
     lowercase equivalence.

IBM PC (code page 437)
     This code page is the normal default for PCs in the U.S. It
     corresponds to the original IBM PC character set. This set has
     some, but not all, of the extended Latin-1 letters, but these
     letters do not have the same encoding as Latin-1. In this mode,
     these letters are allowed in identifiers with uppercase and
     lowercase equivalence.

IBM PC (code page 850)
     This code page is a modification of 437 extended to include all the
     Latin-1 letters, but still not with the usual Latin-1 encoding. In
     this mode, all these letters are allowed in identifiers with
     uppercase and lowercase equivalence.

Full Upper 8-bit
     Any character in the range 80-FF allowed in identifiers, and all
     are considered distinct. In other words, there are no uppercase
     and lowercase equivalences in this range. This is useful in
     conjunction with certain encoding schemes used for some foreign
     character sets (e.g.  the typical method of representing Chinese
     characters on the PC).

No Upper-Half
     No upper-half characters in the range 80-FF are allowed in
     identifiers.  This gives Ada 83 compatibility for identifier names.

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.

Wide Character Encodings
------------------------

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:

Hex Coding
     In this encoding, a wide character is represented by the following
     five character sequence:

          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.

Upper-Half Coding
     The wide character with encoding `16#abcd#' where the upper bit is
     on (in other words, "a" is in the range 8-F) is represented as two
     bytes, `16#ab#' and `16#cd#'. The second byte cannot be a format
     control character, but is not required to be in the upper half.
     This method can be also used for shift-JIS or EUC, where the
     internal coding matches the external coding.

Shift JIS Coding
     A wide character is represented by a two-character sequence,
     `16#ab#' and `16#cd#', with the restrictions described for
     upper-half encoding as described above. The internal character
     code is the corresponding JIS character according to the standard
     algorithm for Shift-JIS conversion. Only characters defined in the
     JIS code set table can be used with this encoding method.

EUC Coding
     A wide character is represented by a two-character sequence
     `16#ab#' and `16#cd#', with both characters being in the upper
     half. The internal character code is the corresponding JIS
     character according to the EUC encoding algorithm. Only characters
     defined in the JIS code set table can be used with this encoding
     method.

UTF-8 Coding
     A wide character is represented using UCS Transformation Format 8
     (UTF-8) as defined in Annex R of ISO 10646-1/Am.2. Depending on
     the character value, the representation is a one, two, or three
     byte sequence:
          16#0000#-16#007f#: 2#0xxxxxxx#
          16#0080#-16#07ff#: 2#110xxxxx# 2#10xxxxxx#
          16#0800#-16#ffff#: 2#1110xxxx# 2#10xxxxxx# 2#10xxxxxx#

     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, but in this implementation, all UTF-8 sequences
     of four or more bytes length will be treated as illegal).

Brackets Coding
     In this encoding, a wide character is represented by the following
     eight character sequence:

          [ " 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 the
     standard ACVC (Ada Compiler Validation Capability) test suite
     distributions.

Note: Some of these coding schemes do not permit the full use of the
Ada 95 character set. For example, neither Shift JIS, nor EUC allow the
use of the upper half of the Latin-1 set.

File Naming Rules
=================

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 list shows some examples of these rules.

`main.ads'
     Main (spec)

`main.adb'
     Main (body)

`arith_functions.ads'
     Arith_Functions (package spec)

`arith_functions.adb'
     Arith_Functions (package body)

`func-spec.ads'
     Func.Spec (child package spec)

`func-spec.adb'
     Func.Spec (child package body)

`main-sub.adb'
     Sub (subunit of Main)

`a~bad.adb'
     A.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. *Note 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 *note Renaming Files Using gnatchop::..)

Using Other File Names
======================

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 *note
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.

Alternative File Naming Schemes
===============================

   In the previous section, we 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:

   * If there is a specific `Source_File_Name' pragma for the given
     unit, then this is always used, and any general pattern rules are
     ignored.

   * If there is a pattern type `Source_File_Name' pragma that applies
     to the unit, then the resulting file name will be used if the file
     exists. If more than one pattern matches, the latest one will be
     tried first, and the first attempt resulting in a reference to a
     file that exists will be used.

   * If no pattern type `Source_File_Name' pragma that applies to the
     unit for which the corresponding file exists, then the standard
     GNAT default naming rules are used.

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);

Generating Object Files
=======================

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:

   * If a package spec has no body, compile the package spec to produce
     the object file for the package.

   * If a package has both a spec and a body, compile the body to
     produce the object file for the package. The source file for the
     package spec need not be compiled in this case because there is
     only one object file, which contains the code for both the spec
     and body of the package.

   * For a subprogram, compile the subprogram body to produce the
     object file for the subprogram. The spec, if one is present, is as
     usual in a separate file, and need not be compiled.

   * In the case of subunits, only compile the parent unit. A single
     object file is generated for the entire subunit tree, which
     includes all the subunits.

   * Compile child units independently of their parent units (though,
     of course, the spec of all the ancestor unit must be present in
     order to compile a child unit).

   * Compile generic units in the same manner as any other units. The
     object files in this case are small dummy files that contain at
     most the flag used for elaboration checking. This is because GNAT
     always handles generic instantiation by means of macro expansion.
     However, it is still necessary to compile generic units, for
     dependency checking and elaboration purposes.

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.

Source Dependencies
===================

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:

   * If a file being compiled `with''s a unit X, the object file
     depends on the file containing the spec of unit X. This includes
     files that are `with''ed implicitly either because they are parents
     of `with''ed child units or they are run-time units required by the
     language constructs used in a particular unit.

   * If a file being compiled instantiates a library level generic
     unit, the object file depends on both the spec and body files for
     this generic unit.

   * If a file being compiled instantiates a generic unit defined
     within a package, the object file depends on the body file for the
     package as well as the spec file.

   * If a file being compiled contains a call to a subprogram for which
     pragma `Inline' applies and inlining is activated with the
     `-gnatn' switch, the object file depends on the file containing the
     body of this subprogram as well as on the file containing the
     spec. Note that for inlining to actually occur as a result of the
     use of this switch, it is necessary to compile in optimizing mode.

     The use of `-gnatN' activates a more extensive 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.

   * If an object file 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.

   * The object file for a parent unit depends on all its subunit body
     files.

   * The previous two rules meant that for purposes of computing
     dependencies and recompilation, a body and all its subunits are
     treated as an indivisible whole.

     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
     described in the Ada 95 Language Reference Manual. However, it is
     a superset of what the ARM 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'.

The Ada Library Information Files
=================================

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.

   * Version information (indicates which version of GNAT was used to
     compile the unit(s) in question)

   * Main program information (including priority and time slice
     settings, as well as the wide character encoding used during
     compilation).

   * List of arguments used in the `gcc' command for the compilation

   * Attributes of the unit, including configuration pragmas used, an
     indication of whether the compilation was successful, exception
     model used etc.

   * A list of relevant restrictions applying to the unit (used for
     consistency) checking.

   * Categorization information (e.g. use of pragma `Pure').

   * Information on all `with''ed units, including presence of
     `Elaborate' or `Elaborate_All' pragmas.

   * Information from any `Linker_Options' pragmas used in the unit

   * Information on the use of `Body_Version' or `Version' attributes
     in the unit.

   * Dependency information. This is a list of files, together with
     time stamp and checksum information. These are files on which the
     unit depends in the sense that recompilation is required if any of
     these units are modified.

   * Cross-reference data. Contains information on all entities
     referenced in the unit. Used by tools like `gnatxref' and
     `gnatfind' to provide cross-reference information.

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.

Binding an Ada Program
======================

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.

Mixed Language Programming
==========================

Interfacing to C
----------------

There are two ways to build a program that contains some Ada files and
some other language files depending on whether the main program is in
Ada or not.  If the main program is in Ada, you should proceed as
follows:

  1. Compile the other language files to generate object files. For
     instance:
          gcc -c file1.c
          gcc -c file2.c

  2. Compile the Ada units to produce a set of object files and ALI
     files. For instance:
          gnatmake -c my_main.adb

  3. Run the Ada binder on the Ada main program. For instance:
          gnatbind my_main.ali

  4. Link the Ada main program, the Ada objects and the other language
     objects. For instance:
          gnatlink my_main.ali file1.o file2.o

   The three last steps can be grouped in a single command:
     gnatmake my_main.adb -largs file1.o file2.o

If the main program is in some language other than Ada, Then you may
have more than one entry point in the Ada subsystem. You must use a
special option of the binder to generate callable routines to initialize
and finalize the Ada units (*note 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. You use the same
procedure for building the program as described previously. In this
case, however, the binder only places the initialization and
finalization subprograms into file `b~XXX.adb' instead of the main
program.  So, if the main program is not in Ada, you should proceed as
follows:

  1. Compile the other language files to generate object files. For
     instance:
          gcc -c file1.c
          gcc -c file2.c

  2. Compile the Ada units to produce a set of object files and ALI
     files. For instance:
          gnatmake -c entry_point1.adb
          gnatmake -c entry_point2.adb

  3. Run the Ada binder on the Ada main program. For instance:
          gnatbind -n entry_point1.ali entry_point2.ali

  4. Link the Ada main program, the Ada objects and the other language
     objects. You only need to give the last entry point here. For
     instance:
          gnatlink entry_point2.ali file1.o file2.o

Calling Conventions
-------------------

   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:

   * Ada. This indicates that the standard Ada calling sequence will be
     used and all Ada data items may be passed without any limitations
     in the case where GNAT is used to generate both the caller and
     callee. It is also possible to mix GNAT generated code and code
     generated by another Ada compiler. In this case, the data types
     should be restricted to simple cases, including primitive types.
     Whether complex data types can be passed depends on the situation.
     Probably it is safe to pass simple arrays, such as arrays of
     integers or floats. Records may or may not work, depending on
     whether both compilers lay them out identically. Complex structures
     involving variant records, access parameters, tasks, or protected
     types, are unlikely to be able to be passed.

     Note that in the case of GNAT running on a platform that supports
     DEC Ada 83, a higher degree of compatibility can be guaranteed,
     and in particular records are layed 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
     DEC Ada 83, All the tasking operations must either be entirely
     within GNAT compiled sections of the program, or entirely within
     DEC Ada 83 compiled sections of the program.

   * Asm. Equivalent to Ada.

   * Assembler. Equivalent to Ada.

   * COBOL. Data will be passed according to the conventions described
     in section B.4 of the Ada 95 Reference Manual.

   * C. Data will be passed according to the conventions described in
     section B.3 of the Ada 95 Reference Manual.

   * CPP. This stands for C++. For most purposes this is identical to C.
     See the separate description of the specialized GNAT pragmas
     relating to C++ interfacing for further details.

   * Fortran. Data will be passed according to the conventions described
     in section B.5 of the Ada 95 Reference Manual.

   * Intrinsic. This applies to an intrinsic operation, as defined in
     the Ada 95 Reference Manual. If a a pragma Import (Intrinsic)
     applies to a subprogram, this means that the body of the
     subprogram is provided by the compiler itself, usually by means of
     an efficient code sequence, and that the user does not supply an
     explicit body for it. In an application program, the pragma can
     only be applied to the following two sets of names, which the GNAT
     compiler recognizes.
        * Rotate_Left, Rotate_Right, Shift_Left, Shift_Right,
          Shift_Right_- Arithmetic.  The corresponding subprogram
          declaration must have two formal parameters. The first one
          must be a signed integer type or a modular type with a binary
          modulus, and the second parameter must be of type Natural.
          The return type must be the same as the type of the first
          argument. The size of this type can only be 8, 16, 32, or 64.

        * binary arithmetic operators: "+", "-", "*", "/" The
          corresponding operator declaration must have parameters and
          result type that have the same root numeric type (for
          example, all three are long_float types). This simplifies the
          definition of operations that use type checking to perform
          dimensional checks:
               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.

   * Stdcall. This is relevant only to NT/Win95 implementations of GNAT,
     and specifies that the Stdcall calling sequence will be used, as
     defined by the NT API.

   * Stubbed. This is a special convention that indicates that the
     compiler should provide a stub body that raises `Program_Error'.

Building Mixed Ada & C++ Programs
=================================

Building a mixed application containing both Ada and C++ code may be a
challenge for the unaware programmer. As a matter of fact, this
interfacing has not been standardized in the Ada 95 reference manual due
to the immaturity and lack of standard of C++ at the time. This section
gives a few hints that should make this task easier. In particular the
first section addresses the differences with interfacing with C. The
second section looks into the delicate problem of linking the complete
application from its Ada and C++ parts. The last section give some
hints on how the GNAT run time can be adapted in order to allow
inter-language dispatching with a new C++ compiler.

Interfacing to C++
------------------

GNAT supports interfacing with C++ compilers generating code that is
compatible with the standard Application Binary Interface of the given
platform.

Interfacing can be done at 3 levels: simple data, subprograms and
classes. In the first 2 cases, GNAT offer a specific CONVENTION CPP
that behaves exactly like CONVENTION C. Usually C++ mangle names of
subprograms and currently GNAT does not provide any help to solve the
demangling problem. This problem can be addressed in 2 ways:
   * by modifying the C++ code in order to force a C convention using
     the EXTERN "C" syntax.

   * by figuring out the mangled name and use it as the Link_Name
     argument of the pragma import.

Interfacing at the class level can be achieved by using the GNAT
specific pragmas such as `CPP_Class' and ` CPP_Virtual'. See the GNAT
Reference Manual for additional information.

Linking a Mixed C++ & Ada Program
---------------------------------

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:
  1. Using GNAT and G++ (GNU C++ compiler) from the same GCC
     installation. The c++ linker can simply be called by using the c++
     specific driver called `c++'. Note that this setup is not very
     common because it may request recompiling the whole GCC tree from
     sources and it does not allow to upgrade easily to a new version
     of one compiler for one of the two languages without taking the
     risk of destabilizing the other.

          $ c++ -c file1.C
          $ c++ -c file2.C
          $ gnatmake ada_unit -largs file1.o file2.o --LINK=c++

  2. Using GNAT and G++ from 2 different GCC installations. If both
     compilers are on the PATH, the same method can be used. It is
     important to be aware that environment variables such as
     C_INCLUDE_PATH, GCC_EXEC_PREFIX, BINUTILS_ROOT or GCC_ROOT will
     affect both compilers at the same time and thus may make one of
     the 2 compilers operate improperly if they are set for the other.
     In particular it is important that the link command has access to
     the proper gcc library `libgcc.a', that is to say the one that is
     part of 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 full verbosity 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 workaround 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:

          $ gnatlink -v -v ada_unit file1.o file2.o --LINK=./my_script
          $ cat ./my_script
          #!/bin/sh
          unset BINUTILS_ROOT
          unset GCC_ROOT
          c++ $*

  3. Using a non GNU C++ compiler. The same set of command as previously
     described can be used to insure that the c++ linker is used.
     Nonetheless, the Ada code may implicitly depend on the gcc
     library. The latter can be located thanks to gnatls: it is to be
     found on the last directory of the object path. It must then be
     explicitly mentioned in the link command :
          $ gnatls -v
          $ Gdir=<the last directory on the object path>
          $ gnatlink ada_unit file1.o file2.o -L$Gdir -lgcc
                --LINK=<cpp_linker>

A Simple Example
----------------

The following example, provided as part of the GNAT examples, show how
to achieve procedural interfacing between Ada and C++ in both
directions. The C++ class A has 2 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++ calss is modelized by
a limited record with a layout comparable to the C++ class. The Ada
subprogram, in turn, calls the c++ method. So from the C++ main program
the code goes back and forth between the 2 languages.

Here are the compilation commands for native configurations:
     $ gnatmake -c simple_cpp_interface
     $ c++ -c cpp_main.C
     $ c++ -c ex7.C
     $ gnatbind -n simple_cpp_interface
     $ gnatlink simple_cpp_interface -o cpp_main --LINK=$(CPLUSPLUS)
           -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);
       virtual 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);
     }
     
     -- Ada sources
     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;
     
     package Simple_Cpp_Interface is
        type A is limited
           record
              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;

Adapting the Run Time to a New C++ Compiler
-------------------------------------------

GNAT offers the capability to derive Ada 95 tagged types directly from
preexisting C++ classes and . See "Interfacing with C++" in the GNAT
reference manual. The mechanism used by GNAT for achieving such a goal
has been made user configurable through a GNAT library unit
`Interfaces.CPP'. The default version of this file is adapted to the
GNU c++ compiler. Internal knowledge of the virtual table layout used
by the new C++ compiler is needed to configure properly this unit. The
Interface of this unit is known by the compiler and cannot be changed
except for the value of the constants defining the characteristics of
the virtual table: CPP_DT_Prologue_Size, CPP_DT_Entry_Size,
CPP_TSD_Prologue_Size, CPP_TSD_Entry_Size. Read comments in the source
of this unit for more details.

Comparison between GNAT and C/C++ Compilation Models
====================================================

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.

Comparison between GNAT and Conventional Ada Library Models
===========================================================

This section is intended to be useful to Ada programmers who have
previously used an Ada compiler implementing the traditional Ada library
model, as described in the Ada 95 Language Reference Manual. If you
have not used such a system, please go on to the next section.

   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:

   * When a unit is `with''ed, the unit seen by the compiler corresponds
     to the version of the unit most recently compiled into the library.

   * Inlining is effective only if the necessary body has already been
     compiled into the library.

   * Compiling a unit may obsolete other units in the library.

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:

   * When a unit is `with''ed, the unit seen by the compiler corresponds
     to the source version of the unit that is currently accessible to
     the compiler.

   * Inlining requires the appropriate source files for the package or
     subprogram bodies to be available to the compiler. Inlining is
     always effective, independent of the order in which units are
     complied.

   * Compiling a unit never affects any other compilations. The editing
     of sources may cause previous compilations to be out of date if
     they depended on the source file being modified.

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.

Compiling Using `gcc'
*********************

This chapter 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.

Compiling Programs
==================

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:

   * the body file (`.adb') for a library level subprogram or generic
     subprogram

   * the spec file (`.ads') for a library level package or generic
     package that has no body

   * the body file (`.adb') for a library level package or generic
     package that has a body

You need *not* compile the following files

   * the spec of a library unit which has a body

   * subunits

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):

     No code generated for file FFF (PACKAGE SPEC)
     No code generated for file FFF (SUBUNIT)

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 three separate files to
be compiled:

     $ gcc -c x.adb y.adb z.c

calls `gnat1' (the Ada compiler) twice to compile `x.adb' and `y.adb',
and `cc1' (the C compiler) once to compile `z.c'.  The compiler
generates three object files `x.o', `y.o' and `z.o' and the two ALI
files `x.ali' and `y.ali' from the Ada compilations. Any switches apply
to all the files listed, except for `-gnatX' switches, which apply only
to Ada compilations.

Switches for `gcc'
==================

The `gcc' command accepts switches that control the compilation
process. These switches are fully described in this section.  First we
briefly list all the switches, in alphabetical order, then we describe
the switches in more detail in functionally grouped sections.

`-b TARGET'
     Compile your program to run on TARGET, which is the name of a
     system configuration. You must have a GNAT cross-compiler built if
     TARGET is not the same as your host system.

`-BDIR'
     Load compiler executables (for example, `gnat1', the Ada compiler)
     from DIR instead of the default location. Only use this switch
     when multiple versions of the GNAT compiler are available. See the
     `gcc' manual page for further details. You would normally use the
     `-b' or `-V' switch instead.

`-c'
     Compile. Always use this switch when compiling Ada programs.

     Note: 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.

`-g'
     Generate debugging information. This information is stored in the
     object file and copied from there to the final executable file by
     the linker, where it can be read by the debugger. You must use the
     `-g' switch if you plan on using the debugger.

`-IDIR'
     Direct GNAT to search the DIR directory for source files needed by
     the current compilation (*note Search Paths and the Run-Time
     Library (RTL)::.).

`-I-'
     Except for the source file named in the command line, do not look
     for source files in the directory containing the source file named
     in the command line (*note Search Paths and the Run-Time Library
     (RTL)::.).

`-o FILE'
     This switch is used in `gcc' to redirect the generated object file
     and its associated ALI file. Beware of this switch with GNAT,
     because it may cause the object file and ALI file to have
     different names which in turn may confuse the binder and the
     linker.

`-O[N]'
     N controls the optimization level.

    n = 0
          No optimization, the default setting if no `-O' appears

    n = 1
          Normal optimization, the default if you specify `-O' without
          an operand.

    n = 2
          Extensive optimization

    n = 3
          Extensive optimization with automatic inlining. This applies
          only to inlining within a unit. For details on control of
          inter-unit inlining see *Note Subprogram Inlining Control::.

`-S'
     Used in place of `-c' to cause the assembler source file to be
     generated, using `.s' as the extension, instead of the object file.
     This may be useful if you need to examine the generated assembly
     code.

`-v'
     Show commands generated by the `gcc' driver. Normally used only for
     debugging purposes or if you need to be sure what version of the
     compiler you are executing.

`-V VER'
     Execute VER version of the compiler. This is the `gcc' version,
     not the GNAT version.

`-funwind-tables'
     This switch causes the object files to be generated with unwind
     table information. This is required for use of zero cost exception
     handling, of for use of the trace capabilities in the GNAT library.

`-gnata'
     Assertions enabled. `Pragma Assert' and `pragma Debug' to be
     activated.

`-gnatb'
     Generate brief messages to `stderr' even if verbose mode set.

`-gnatc'
     Check syntax and semantics only (no code generation attempted).

`-gnatC'
     Compress debug information and external symbol name table entries.

`-gnatD'
     Output expanded source files for source level debugging. This
     switch also suppress generation of cross-reference information
     (see -gnatx).

`-gnatE'
     Full dynamic elaboration checks.

`-gnatf'
     Full errors. Multiple errors per line, all undefined references.

`-gnatF'
     Externals names are folded to all uppercase.

`-gnatg'
     Internal GNAT implementation mode. This should not be used for
     applications programs, it is intended only for use by the compiler
     and its run-time library. For documentation, see the GNAT sources.

`-gnatG'
     List generated expanded code in source form.

`-gnatiC'
     Identifier character set (C=1/2/3/4/8/p/f/n/w).

`-gnath'
     Output usage information. The output is written to `stdout'.

`-gnatkN'
     Limit file names to N (1-999) characters (`k' = krunch).

`-gnatl'
     Output full source listing with embedded error messages.

`-gnatmN'
     Limit number of detected errors to N (1-999).

`-gnatn'
     Activate inlining across unit boundaries for subprograms for which
     pragma `inline' is specified.

`-gnatN'
     Activate front end inlining.

`-fno-inline'
     Suppresses all inlining, even if other optimization or inlining
     switches are set.

`-fstack-check'
     Activates stack checking. See separate section on stack checking
     for details of the use of this option.

`-gnato'
     Enable numeric overflow checking (which is not normally enabled by
     default).

`-gnatp'
     Suppress all checks.

`-gnatq'
     Don't quit; try semantics, even if parse errors.

`-gnatQ'
     Don't quit; generate `ali' and tree files even if illegalities.

`-gnatP'
     Enable polling. This is required on some systems (notably Windows
     NT) to obtain asynchronous abort and asynchronous transfer of
     control capability.  See the description of pragma Polling in the
     GNAT Reference Manual for full details.

`-gnatR0/1/2'
     Output representation information for declared types and objects.

`-gnats'
     Syntax check only.

`-gnatt'
     Tree output file to be generated.

`-gnatT nnn'
     Set time slice to specified number of microseconds

`-gnatu'
     List units for this compilation.

`-gnatU'
     Tag all error messages with the unique string "error:"

`-gnatv'
     Verbose mode. Full error output with source lines to `stdout'.

`-gnatwM'
     Warning mode (M=`s,e,l' for suppress, treat as error, elaboration
     warnings).

`-gnatWE'
     Wide character encoding method (E=n/h/u/s/e/8).

`-gnatx'
     Suppress generation of cross-reference information.

`-gnatwM'
     Warning mode

`-gnaty'
     Enable built-in style checks. See separate section describing this
     feature.

`-gnatzM'
     Distribution stub generation and compilation (M=r/c for
     receiver/caller stubs).

`-gnat83'
     Enforce Ada 83 restrictions.

`-pass-exit-codes'
     Catch exit codes from the compiler and use the most meaningful as
     exit status.

   You may combine a sequence of GNAT switches into a single switch. For
example, the combined switch

     -gnatcfi3

is equivalent to specifying the following sequence of switches:

     -gnatc -gnatf -gnati3

Output and Error Message Control
--------------------------------

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.  `glide'
can parse the error messages and point to the referenced character.
The following switches provide control over the error message format:

`-gnatv'
     The v stands for verbose.  The effect of this setting is to write
     long-format error messages to `stdout' (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.

`-gnatl'
     The `l' stands for list.  This switch causes a full listing of the
     file to be generated. The output might look as follows:

          1. procedure E is
           2.    V : Integer;
           3.    funcion X (Q : Integer)
                 |
              >>> Incorrect spelling of keyword "function"
           4.     return Integer;
                                |
              >>> ";" should be "is"
           5.    begin
           6.       return Q + Q;
           7.    end;
           8. begin
           9.    V := X + X;
          10.end E;

     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.

`-gnatU'
     This switch forces all error messages to be preceded by the unique
     string "error:". This means that error messages take a few more
     characters in space, but allows easy searching for and
     identification of error messages.

`-gnatb'
     The `b' stands for brief.  This switch causes GNAT to generate the
     brief format error messages to `stderr' (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).

`-gnatmN'
     The `m' stands for maximum.  N is a decimal integer in the range
     of 1 to 999 and limits the number of error messages to be
     generated. For example, using `-gnatm2' might yield

          e.adb:3:04: Incorrect spelling of keyword "function"
          e.adb:5:35: missing ".."
          fatal error: maximum errors reached
          compilation abandoned

`-gnatf'
     The `f' stands for full.  Normally, the compiler suppresses error
     messages that are likely to be redundant. This switch causes all
     error messages to be generated. In particular, in the case of
     references to undefined variables. If a given variable is
     referenced several times, the normal format of messages is
          e.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

`-gnatq'
     The `q' stands for quit (really "don't quit").  In normal
     operation mode, the compiler first parses the program and
     determines if there are any syntax errors. If there are,
     appropriate error messages are generated and compilation is
     immediately terminated.  This switch tells GNAT to continue with
     semantic analysis even if syntax errors have been found. This may
     enable the detection of more errors in a single run. On the other
     hand, the semantic analyzer is more likely to encounter some
     internal fatal error when given a syntactically invalid tree.

`-gnatQ'
     In normal operation mode, the `ali' 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 browing tools or
     integrated development environments) that are driven from the
     `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.

     In addition to error messages, which correspond to illegalities as
     defined in the Ada 95 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:

        * Possible infinitely recursive calls

        * Out-of-range values being assigned

        * Possible order of elaboration problems

        * Unreachable code

        * Variables that are never assigned a value

        * Variables that are referenced before being initialized

        * Task entries with no corresponding Accept statement

        * Duplicate Accepts for the same task entry in a select

        * Objects that take too much storage

        * Unchecked conversion between types of differing sizes

        * Missing return statements along some execution paths in a
          function

        * Incorrect pragmas

        * Incorrect external names

        * Allocation from empty storage pool

        * Potentially blocking operations in protected types

        * Suspicious parenthesization of expressions

        * Mismatching bounds in an aggregate

        * Attempt to return local value by reference

        * Unrecognized pragmas

        * Premature instantiation of a generic body

        * Attempt to pack aliased components

        * Out of bounds array subscripts

        * Wrong length on string assignment

        * Violations of style rules if style checking is enabled

     The following switches are available to control the handling of
     warning messages:

    `-gnatwa (activate all optional errors)'
          This swich activates all optional warning messages, see
          remaining list in this section for details on optional
          warning messages that can be individually controlled. The one
          exception is that `-gnatwa' does not activate warnings for
          hiding variables (`-gnatwh'), so if this warning is required,
          it must be explicitly set.

    `-gnatwA (suppress all optional errors)'
          This swich suppresses all optional warning messages, see
          remaining list in this section for details on optional
          warning messages that can be individually controlled.

    `-gnatwb (activate warnings on biased rounding)'
          If a static floating-point expression has a value that is
          exactly half way between two adjacent machine numbers, then
          the rules of Ada (Ada Reference Manual, section 4.9(38))
          require that this rounding be done away from zero, even if
          the normal unbiased rounding rules at run time would require
          rounding towards zero. This warning message alerts you to
          such instances where compile-time rounding and run-time
          rounding are not equivalent. If it is important to get proper
          run-time rounding, then you can force this by making one of
          the operands into a variable. The default is that such
          warnings are not generated.  Note that `-gnatwa' does not
          affect the setting of this warning option.

    `-gnatwB (suppress warnings on biased rounding)'
          This switch disables warnings on biased rounding.

    `-gnatwc (activate warnings on conditionals)'
          This 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.
          This warning can also be turned on using `-gnatwa'.

    `-gnatwC (suppress warnings on conditionals)'
          This switch suppresses warnings for conditional expressions
          used in tests that are known to be True or False at compile
          time.

    `-gnatwe (treat warnings as errors)'
          This switch causes warning 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.

    `-gnatwh (activate warnings on hiding)'
          This 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.  Note that `-gnatwa' does not affect the setting
          of this warning option.

    `-gnatwH (suppress warnings on hiding)'
          This switch suppresses warnings on hiding declarations.

    `-gnatwi (activate warnings on implementation units).'
          This 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 This warning can also be turned on using
          `-gnatwa'.

    `-gnatwI (disable warnings on implementation units).'
          This switch disables warnings for a `with' of an internal GNAT
          implementation unit.

    `-gnatwl (activate warnings on elaboration pragmas)'
          This swich activates warnings on missing pragma Elaborate_All
          statements.  See the section in this guide on elaboration
          checking for details on when such pragma should be used. The
          default is that such warnings are not generated.  This
          warning can also be turned on using `-gnatwa'.

    `-gnatwL (suppress warnings on elaboration pragmas)'
          This swich suppresses warnings on missing pragma
          Elaborate_All statements.  See the section in this guide on
          elaboration checking for details on when such pragma should
          be used.

    `-gnatwo (activate warnings on address clause overlays)'
          This swich 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.  This warning can also be turned on
          using `-gnatwa'.

    `-gnatwO (suppress warnings on address clause overlays)'
          This swich suppresses warnings on possibly unintended
          initialization effects of defining address clauses that cause
          one variable to overlap another.

    `-gnatwp (activate warnings on ineffective pragma Inlines)'
          This swich 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.  This warning can also be turned on using `-gnatwa'.

    `-gnatwP (suppress warnings on ineffective pragma Inlines)'
          This swich suppresses warnings on ineffective pragma Inlines.
          If the inlining mechanism cannot inline a call, it will
          simply ignore the request silently.

    `-gnatwr (activate warnings on redundant constructs)'
          This switch activates warnings for redundant constructs. In
          particular assignments of a variable to itself, and a type
          conversion that converts an object to its own type. The
          default is that such warnings are not generated.

    `-gnatwR (suppress warnings on redundant constructs)'
          This switch suppresses warnings for redundant constructs.

    `-gnatws (suppress all warnings)'
          This switch completely suppresses the output of all warning
          messages from the GNAT front end.  Note that it 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'.

    `-gnatwu (activate warnings on unused entities)'
          This switch activates warnings to be generated for entities
          that are defined 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 the 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
          warning can also be turned on using `-gnatwa'.

    `-gnatwU (suppress warnings on unused entities)'
          This switch suppresses warnings for unused entities and
          packages.

          A string of warning parameters can be used in the same
          parameter. For example:

               -gnatwaLe

          Would turn on all optional warnings except for elaboration
          pragma warnings, and also specify that warnings should be
          treated as errors.

    `-w'
          This switch suppresses warnings from the `gcc' backend. It
          may be used in conjunction with `-gnatws' to ensure that all
          warnings are suppressed during the entire compilation process.

Debugging and Assertion Control
-------------------------------

`-gnata'
     The pragmas `Assert' and `Debug' normally have no effect and are
     ignored. This switch, where `a' stands for assert, causes `Assert'
     and `Debug' pragmas to be activated.

     The pragmas have the form:

          pragma Assert (BOOLEAN-EXPRESSION [,
                                STATIC-STRING-EXPRESSION])
             pragma Debug (PROCEDURE CALL)

     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_Error' 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 giving 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.

Style Checking
--------------

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 warning message is given, preceded by the
character sequence "(style)".  The string X is a sequence of letters or
digits indicating the particular style checks to be performed. The
following checks are defined:

`1-9 (specify indentation level)'
     If a digit from 1-9 appears in the string after `-gnaty' then
     proper indentation is checked, with the digit indicating the
     indentation level required. 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.

`a (check attribute casing)'
     If the letter a appears in the string after `-gnaty' then
     attribute 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.

`b (blanks not allowed at statement end)'
     If the letter b appears in the string after `-gnaty' then trailing
     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.

`c (check comments)'
     If the letter c appears in the string after `-gnaty' then comments
     must meet the following set of rules:

        * The "-" that starts the column must either start in column
          one, or else at least one blank must precede this sequence.

        * Comments that follow other tokens on a line must have at
          least one blank following the "-" at the start of the comment.

        * Full line comments must have two blanks following the "-"
          that starts the comment, with the following exceptions.

        * A line consisting only of the "-" characters, possibly
          preceded by blanks is permitted.

        * A comment starting with "-x" where x is a special character
          is permitted.  This alows proper processing of the output
          generated by specialized tools including `gnatprep' (where -!
          is used) and the SPARK annnotation language (where -# is
          used). For the purposes of this rule, a special character is
          defined as being in one of the ASCII ranges 16#21#..16#2F# or
          16#3A#..16#3F#.

        * A line consisting entirely of minus signs, possibly preceded
          by blanks, is permitted. This allows the construction of box
          comments where lines of minus signs are used to form the top
          and bottom of the box.

        * If a comment starts and ends with "-" is permitted as long as
          at least one blank follows the initial "-". Together with the
          preceding rule, this allows the construction of box comments,
          as shown in the following example:
               ---------------------------
               -- This is a box comment --
               -- with two text lines.  --
               ---------------------------

`e (check end/exit labels)'
     If the letter e appears in the string after `-gnaty' then optional
     labels on `end' statements ending subprograms and on `exit'
     statements exiting named loops, are required to be present.

`f (no form feeds or vertical tabs)'
     If the letter f appears in the string after `-gnaty' then neither
     form feeds nor vertical tab characters are not permitted in the
     source text.

`h (no horizontal tabs)'
     If the letter h appears in the string after `-gnaty' then
     horizontal 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.

`i (check if-then layout)'
     If the letter i appears in the string after `-gnaty', then the
     keyword `then' must appear either on the same line as
     corresponding `if', or on a line on its own, lined up under the
     `if' with at least one non-blank line in between containing all or
     part of the condition to be tested.

`k (check keyword casing)'
     If the letter k appears in the string after `-gnaty' then all
     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).

`l (check layout)'
     If the letter l appears in the string after `-gnaty' then layout
     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'.  For example, either of the following two layouts is
     acceptable:

          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;

`m (check maximum line length)'
     If the letter m appears in the string after `-gnaty' then the
     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.

`Mnnn (set maximum line length)'
     If the sequence Mnnn, where nnn is a decimal number, appears in
     the string after `-gnaty' then the length of lines must not exceed
     the given value.

`n (check casing of entities in Standard)'
     If the letter n appears in the string after `-gnaty' then any
     identifier from Standard must be cased to match the presentation
     in the Ada Reference Manual (for example, `Integer' and
     `ASCII.NUL').

`o (check order of subprogram bodies)'
     If the letter o appears in the string after `-gnaty' then all
     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).

`p (check pragma casing)'
     If the letter p appears in the string after `-gnaty' then pragma
     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.

`r (check references)'
     If the letter r appears in the string after `-gnaty' then all
     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.

`s (check separate specs)'
     If the letter s appears in the string after `-gnaty' then separate
     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.

`t (check token spacing)'
     If the letter t appears in the string after `-gnaty' then the
     following token spacing rules are enforced:

        * The keywords `abs' and `not' must be followed by a space.

        * The token `=>' must be surrounded by spaces.

        * The token `<>' must be preceded by a space or a left
          parenthesis.

        * Binary operators other than `**' must be surrounded by spaces.
          There is no restriction on the layout of the `**' binary
          operator.

        * Colon must be surrounded by spaces.

        * Colon-equal (assignment) must be surrounded by spaces.

        * Comma must be the first non-blank character on the line, or be
          immediately preceded by a non-blank character, and must be
          followed by a space.

        * Left parenthesis must be preceded by a space, and must not be
          followed by a space (it can be at the end of a line).

        * A right parenthesis must either be the first non-blank
          character on a line, or it must be preceded by a non-blank
          character.

        * A semicolon must not be preceded by a space, and must not be
          followed by a non-blank character.

        * A unary plus or minus may not be followed by a space.

        * A vertical bar must be surrounded by spaces.

     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 `gnaty3abcefhiklmprst', that is all checking
options are enabled with the exception of -gnatyo, with an indentation
level of 3. This is the standard checking option that is used for the
GNAT sources.

Run-Time Checks
---------------

If you compile with the default options, GNAT will insert many run-time
checks into the compiled code, including code that performs range
checking against constraints, but not arithmetic overflow checking for
integer operations (including division by zero) or checks for access
before elaboration on subprogram calls. All other run-time checks, as
required by the Ada 95 Reference Manual, are generated by default.  The
following `gcc' switches refine this default behavior:

`-gnatp'
     Suppress all run-time checks as though `pragma Suppress
     (all_checks') had been present in the source. Use this switch to
     improve the performance of the code at the expense of safety in
     the presence of invalid data or program bugs.

`-gnato'
     Enables overflow checking for integer operations.  This causes
     GNAT to generate slower and larger executable programs by adding
     code to check for both overflow and division by zero (resulting in
     raising `Constraint_Error' as required by standard Ada semantics).
     These overflow checks correspond to situations in which the true
     value of the result of an operation may be outside the base range
     of the result type. The following example shows the distinction:

          X1 : Integer := Integer'Last;
          X2 : Integer range 1 .. 5 := 5;
          ...
          X1 := X1 + 1;   -- `-gnato' required to catch the Constraint_Error
          X2 := X2 + 1;   -- range check, `-gnato' has no effect here

     Here the first addition results in a value that is outside the
     base range of Integer, and hence requires an overflow check for
     detection of the constraint error. The second increment operation
     results in a violation of the explicit range constraint, and such
     range checks are always performed. Basically the compiler can
     assume that in the absence of the `-gnato' switch that any value
     of type `xxx' is in range of the base type of `xxx'.

     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 can generate an incorrect value, but cannot cause erroneous
     behavior. This is unlike the situation with a constraint check on
     an array subscript, where failure to perform the check can result
     in random memory description, or the range check on a case
     statement, where failure to perform the check can cause a wild
     jump.

     Note again that `-gnato' is off by default, so overflow checking is
     not performed in default mode. This means that out of the box,
     with the default settings, GNAT does not do all the checks
     expected from the language description in the Ada Reference
     Manual. If you want all constraint checks to be performed, as
     described in this Manual, then you must explicitly use the -gnato
     switch either on the `gnatmake' or `gcc' command.

`-gnatE'
     Enables dynamic checks for access-before-elaboration on subprogram
     calls and generic instantiations.  For full details of the effect
     and use of this switch, *Note Compiling Using gcc::.

The 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.

Stack Overflow Checking
-----------------------

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.

   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) do 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 always controlled by the size
given in an applicable `Storage_Size' pragma (or is set to the default
size if no pragma is used.

   For the environment task, the stack size depends on system defaults
and is unknown to the compiler. The stack may even dynamically grow on
some systems, precluding the normal Ada semantics for stack overflow.
In the worst case, unbounded stack usage, causes unbounded stack
expansion resulting in the system running out of virtual memory.

   The 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.

Run-Time Control
----------------

`-gnatT nnn'
     The `gnatT' switch can be used to specify the time-slicing value
     to be used for task switching between equal priority tasks. The
     value `nnn' is given in microseconds as a decimal integer.

     Setting the time-slicing value is only effective if the underlying
     thread control system can accomodate time slicing. Check the
     documentation of your operating system for details. Note that the
     time-slicing value can also be set by use of pragma `Time_Slice'
     or by use of the `t' switch in the gnatbind step. The pragma
     overrides a command line argument if both are present, and the `t'
     switch for gnatbind overrides both the pragma and the `gcc'
     command line switch.

Using `gcc' for Syntax Checking
-------------------------------

`-gnats'
     The `s' stands for syntax.

     Run 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.

     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 (*note Renaming Files Using gnatchop::.).

Using `gcc' for Semantic Checking
---------------------------------

`-gnatc'
     The `c' stands for check.  Causes the compiler to operate in
     semantic check mode, with full checking for all illegalities
     specified in the Ada 95 Reference Manual, but without generation
     of any object code (no object file is generated).

     Because dependent files must be accessed, you must follow the GNAT
     semantic restrictions on file structuring to operate in this mode:

        * The needed source files must be accessible (*note Search
          Paths and the Run-Time Library (RTL)::.).

        * Each file must contain only one compilation unit.

        * The file name and unit name must match (*note File Naming
          Rules::.).

     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 95 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).

Compiling Ada 83 Programs
-------------------------

`-gnat83'
     Although GNAT is primarily an Ada 95 compiler, it accepts this
     switch to specify that an Ada 83 program is to be compiled in
     Ada83 mode. If you specify this switch, GNAT rejects most Ada 95
     extensions and applies Ada 83 semantics where this can be done
     easily.  It is not possible to guarantee this switch does a perfect
     job; for example, some subtle tests, such as are found in earlier
     ACVC tests (that have been removed from the ACVC suite for Ada
     95), may not compile correctly. However, for most purposes, using
     this switch should help to ensure that programs that compile
     correctly under the `-gnat83' switch can be ported easily to an
     Ada 83 compiler. This is the main use of the switch.

     With few exceptions (most notably the need to use `<>' on
     unconstrained generic formal parameters, the use of the new Ada 95
     keywords, and the use of packages with optional bodies), it is not
     necessary to use the `-gnat83' switch when compiling Ada 83
     programs, because, with rare exceptions, Ada 95 is upwardly
     compatible with Ada 83. This means that a correct Ada 83 program
     is usually also a correct Ada 95 program.

Character Set Control
---------------------

`-gnatiC'
     Normally GNAT recognizes the Latin-1 character set in source
     program identifiers, as described in the Ada 95 Reference Manual.
     This switch causes GNAT to recognize alternate character sets in
     identifiers. C is a single character  indicating the character
     set, as follows:

    `1'
          Latin-1 identifiers

    `2'
          Latin-2 letters allowed in identifiers

    `3'
          Latin-3 letters allowed in identifiers

    `4'
          Latin-4 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 allowed in identifiers

     *Note Foreign Language Representation::, for full details on the
     implementation of these character sets.

`-gnatWE'
     Specify the method of encoding for wide characters.  E is one of
     the following:

    `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
     the these encoding methods see *Note 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.  scheme. If no `-gnatW?' parameter is present,
     then the default representation is Brackets encoding only.

     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.

File Naming Control
-------------------

`-gnatkN'
     Activates file name "krunching". N, a decimal integer in the range
     1-999, indicates the maximum allowable length of a file name (not
     including the `.ads' or `.adb' extension). The default is not to
     enable file name krunching.

     For the source file naming rules, *Note File Naming Rules::.

Subprogram Inlining Control
---------------------------

`-gnatn'
     The `n' here is intended to suggest the first syllable of the word
     "inline".  GNAT recognizes and processes `Inline' pragmas.
     However, for the inlining to actually occur, optimization must be
     enabled. To enable inlining across unit boundaries, this is,
     inlining a call in one unit of a subprogram declared in a
     `with''ed unit, you must also specify this switch.  In the absence
     of this switch, GNAT does not attempt inlining across units and
     does not need to access the bodies of subprograms for which
     `pragma Inline' is specified if they are not in the current unit.

     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 *Note Inlining of Subprograms::.

`-gnatN'
     The front end inlining activated by this switch is generally more
     extensive, and quite often more effective than the standard
     `-gnatn' inlining mode.  It will also generate additional
     dependencies.

Auxiliary Output Control
------------------------

`-gnatt'
     Cause GNAT to write the internal tree for a unit to a file (with
     the extension `.adt'.  This not normally required, but is used by
     separate analysis tools.  Typically these tools do the necessary
     compilations automatically, so you should never have to specify
     this switch in normal operation.

`-gnatu'
     Print a list of units required by this compilation on `stdout'.
     The listing includes all units on which the unit being compiled
     depends either directly or indirectly.

`-pass-exit-codes'
     If this switch is not used, the exit code returned by `gcc' when
     compiling multiple files indicates whether all source files have
     been successfully used to generate object files or not.

     When `-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.

Debugging Control
-----------------

`-gnatdX'
     Activate internal debugging switches. X is a letter or digit, or
     string of letters or digits, which specifies the type of debugging
     outputs desired. Normally these are used only for internal
     development or system debugging purposes. You can find full
     documentation for these switches in the body of the `Debug' unit
     in the compiler source file `debug.adb'.

`-gnatG'
     This switch causes the compiler to generate auxiliary output
     containing a pseudo-source listing of the generated expanded code.
     Like most Ada compilers, GNAT works by first transforming the high
     level Ada code into lower level constructs. For example, tasking
     operations are transformed into calls to the tasking run-time
     routines. A unique capability of GNAT is to list this expanded
     code in a form very close to normal Ada source.  This is very
     useful in understanding the implications of various Ada usage on
     the efficiency of the generated code. There are many cases in Ada
     (e.g. the use of controlled types), where simple Ada statements can
     generate a lot of run-time code. By using `-gnatG' you can identify
     these cases, and consider whether it may be desirable to modify
     the coding approach to improve efficiency.

     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 specification of package `Sprint' in file
     `sprint.ads' for a full list.

    `new XXX [storage_pool = YYY]'
          Shows the storage pool being used for an allocator.

    `at end PROCEDURE-NAME;'
          Shows the finalization (cleanup) procedure for a scope.

    `(if EXPR then EXPR else EXPR)'
          Conditional expression equivalent to the `x?y:z' construction
          in C.

    `TARGET^(SOURCE)'
          A conversion with floating-point truncation instead of
          rounding.

    `TARGET?(SOURCE)'
          A conversion that bypasses normal Ada semantic checking. In
          particular enumeration types and fixed-point types are
          treated simply as integers.

    `TARGET?^(SOURCE)'
          Combines the above two cases.

    `X #/ Y'
    `X #mod Y'
    `X #* Y'
    `X #rem Y'
          A division or multiplication of fixed-point values which are
          treated as integers without any kind of scaling.

    `free EXPR [storage_pool = XXX]'
          Shows the storage pool associated with a `free' statement.

    `freeze TYPENAME [ACTIONS]'
          Shows the point at which TYPENAME is frozen, with possible
          associated actions to be performed at the freeze point.

    `reference ITYPE'
          Reference (and hence definition) to internal type ITYPE.

    `FUNCTION-NAME! (ARG, ARG, ARG)'
          Intrinsic function call.

    `LABELNAME : label'
          Declaration of label LABELNAME.

    `EXPR && EXPR && EXPR ... && EXPR'
          A multiple concatenation (same effect as EXPR & EXPR & EXPR,
          but handled more efficiently).

    `[constraint_error]'
          Raise the `Constraint_Error' exception.

    `EXPRESSION'reference'
          A pointer to the result of evaluating EXPRESSION.

    `TARGET-TYPE!(SOURCE-EXPRESSION)'
          An unchecked conversion of SOURCE-EXPRESSION to TARGET-TYPE.

    `[NUMERATOR/DENOMINATOR]'
          Used to represent internal real literals (that) have no exact
          representation in base 2-16 (for example, the result of
          compile time evaluation of the expression 1.0/27.0).

    `-gnatD'
          This switch is used in conjunction with `-gnatG' to cause the
          expanded source, as described above to be written to files
          with names `xxx.dg', where `xxx' is the normal file name, 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).

    `-gnatC'
          In the generated debugging information, and also in the case
          of long external names, the compiler uses a compression
          mechanism if the name is very long.  This compression method
          uses a checksum, and avoids trouble on some operating systems
          which have difficulty with very long names. The `-gnatC'
          switch forces this compression approach to be used on all
          external names and names in the debugging information tables.
          This reduces the size of the generated executable, at the
          expense of making the naming scheme more complex. The
          compression only affects the qualification of the name. Thus
          a name in the source:

               Very_Long_Package.Very_Long_Inner_Package.Var

          would normally appear in these tables as:

               very_long_package__very_long_inner_package__var

          but if the `-gnatC' switch is used, then the name appears as

               XCb7e0c705__var

          Here b7e0c705 is a compressed encoding of the qualification
          prefix.  The GNAT Ada aware version of GDB understands these
          encoded prefixes, so if this debugger is used, the encoding
          is largely hidden from the user of the compiler.

`-gnatRx'
     This switch controls output from the compiler of a listing showing
     representation information for declared types and objects. For
     `-gnatR0', no information is output (equivalent to omitting the
     `-gnatR' switch). For `-gnatR1' (which is the default, so `-gnatR'
     with no parameter has the same effect), size and alignment
     information is listed for declared array and record types. For
     `-gnatR2', size and alignment information is listed for all
     declared types and objects.

`-gnatx'
     Normally the compiler generates full cross-referencing information
     in the `ALI' 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.

Search Paths and the Run-Time Library (RTL)
===========================================

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:

  1. The directory containing the source file of the main unit being
     compiled (the file name on the command line).

  2. Each directory named by an `-I' switch given on the `gcc' command
     line, in the order given.

  3. Each of the directories listed in the value of the
     `ADA_INCLUDE_PATH' environment variable.  Construct this value
     exactly as the `PATH' environment variable: a list of directory
     names separated by colons.

  4. The content of the "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.  *Note Installing an
     Ada 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 routines. 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.

Order of Compilation Issues
===========================

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:

   * There is no point in compiling specs (except for package specs
     with no bodies) because these are compiled as needed by clients. If
     you attempt a useless compilation, you will receive an error
     message.  It is also useless to compile subunits because they are
     compiled as needed by the parent.

   * There are no order of compilation requirements: performing a
     compilation never obsoletes anything. The only way you can obsolete
     something and require recompilations is to modify one of the
     source files on which it depends.

   * There is no library as such, apart from the ALI files (*note The
     Ada Library Information Files::., for information on the format of
     these files). For now we find it convenient to create separate ALI
     files, but eventually the information therein may be incorporated
     into the object file directly.

   * When you compile a unit, the source files for the specs of all
     units that it `with''s, all its subunits, and the bodies of any
     generics it instantiates must be available (reachable by the
     search-paths mechanism described above), or you will receive a
     fatal error message.

Examples
========

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.

Binding Using `gnatbind'
************************

This chapter describes the GNAT binder, `gnatbind', which is used to
bind compiled GNAT objects. The `gnatbind' program performs four
separate functions:

  1. Checks that a program is consistent, in accordance with the rules
     in Chapter 10 of the Ada 95 Reference Manual. In particular, error
     messages are generated if a program uses inconsistent versions of a
     given unit.

  2. Checks that an acceptable order of elaboration exists for the
     program and issues an error message if it cannot find an order of
     elaboration that satisfies the rules in Chapter 10 of the Ada 95
     Language Manual.

  3. Generates a main program incorporating the given elaboration order.
     This program is a small Ada package (body and spec) that must be
     subsequently compiled using the GNAT compiler. The necessary
     compilation step is usually performed automatically by `gnatlink'.
     The two most important functions of this program are to call the
     elaboration routines of units in an appropriate order and to call
     the main program.

  4. Determines the set of object files required by the given main
     program.  This information is output in the forms of comments in
     the generated C program, to be read by the `gnatlink' utility used
     to link the Ada application.

Running `gnatbind'
==================

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. If no switches are specified, `gnatbind' constructs an Ada
package in two files whose names are `b~ADA_MAIN.ads', and
`b~ADA_MAIN.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 any source files it
can locate into consideration. 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:

  1. Enter `gcc -c hello.adb' to compile the main program.

  2. Enter `gcc -c p.ads' to compile package `P'.

  3. Edit file `p.ads'.

  4. Enter `gnatbind hello'.

   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 it is similar to the following which
is the binder file generated for a simple "hello world" program.

     --  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.
     
     with System;
     package Ada_Main is
     
        --  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);
     
        --  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);
     
        --  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);
     
        --  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.
     
        procedure Break_Start;
        pragma Import (C, Break_Start, "__gnat_break_start");
     
        --  This is the actual generated main program (it would be
        --  suppressed if the no main program swich 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.
     
        u00001 : constant Integer := 16#425FD0AF#;
        u00002 : constant Integer := 16#077A2651#;
        u00003 : constant Integer := 16#08ADDC9E#;
        u00004 : constant Integer := 16#1D370323#;
        u00005 : constant Integer := 16#3043D77B#;
        u00006 : constant Integer := 16#2359F9ED#;
        u00007 : constant Integer := 16#0CA940CF#;
        u00008 : constant Integer := 16#69BA6A59#;
        u00009 : constant Integer := 16#156A40CF#;
        u00010 : constant Integer := 16#033DABE0#;
        u00011 : constant Integer := 16#6AB38FEA#;
        u00012 : constant Integer := 16#7AAA368C#;
        u00013 : constant Integer := 16#7D13B305#;
        u00014 : constant Integer := 16#62D2B79D#;
        u00015 : constant Integer := 16#2E865F1E#;
        u00016 : constant Integer := 16#6379D875#;
        u00017 : constant Integer := 16#72D6A51D#;
        u00018 : constant Integer := 16#6E88E3D7#;
        u00019 : constant Integer := 16#45C8383C#;
        u00020 : constant Integer := 16#385E7AC2#;
        u00021 : constant Integer := 16#08FE4C1F#;
        u00022 : constant Integer := 16#23B87757#;
        u00023 : constant Integer := 16#3A4BFD9A#;
        u00024 : constant Integer := 16#4C9F3930#;
        u00025 : constant Integer := 16#2F1EB794#;
        u00026 : constant Integer := 16#0E2A461A#;
        u00027 : constant Integer := 16#5570D114#;
        u00028 : constant Integer := 16#501FA6BF#;
        u00029 : constant Integer := 16#57692181#;
        u00030 : constant Integer := 16#7C25DE96#;
        u00031 : constant Integer := 16#521B9399#;
        u00032 : constant Integer := 16#689CC1B9#;
        u00033 : constant Integer := 16#0357E00A#;
        u00034 : constant Integer := 16#1345CFE9#;
        u00035 : constant Integer := 16#343244DE#;
        u00036 : constant Integer := 16#6725DC79#;
        u00037 : constant Integer := 16#2DAF477E#;
        u00038 : constant Integer := 16#4F0184F2#;
        u00039 : constant Integer := 16#0A0669D8#;
        u00040 : constant Integer := 16#26610831#;
        u00041 : constant Integer := 16#0B5A4DF9#;
        u00042 : constant Integer := 16#1D4F93E8#;
        u00043 : constant Integer := 16#30B2EC3D#;
        u00044 : constant Integer := 16#34054F96#;
        u00045 : constant Integer := 16#6598BA3E#;
        u00046 : constant Integer := 16#2C9C021D#;
        u00047 : constant Integer := 16#177A51F6#;
        u00048 : constant Integer := 16#1CBC39CD#;
        u00049 : constant Integer := 16#5461BB3E#;
        u00050 : constant Integer := 16#03F36D98#;
        u00051 : constant Integer := 16#208D3EF6#;
        u00052 : constant Integer := 16#33AF4230#;
        u00053 : constant Integer := 16#0B97C6BF#;
        u00054 : constant Integer := 16#34B32999#;
     
        --  The following Export pragms 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.
     
        pragma Export (C, u00001, "helloB");
        pragma Export (C, u00002, "system__standard_libraryB");
        pragma Export (C, u00003, "system__standard_libraryS");
        pragma Export (C, u00004, "systemS");
        pragma Export (C, u00005, "system__exceptionsS");
        pragma Export (C, u00006, "adaS");
        pragma Export (C, u00007, "ada__exceptionsB");
        pragma Export (C, u00008, "ada__exceptionsS");
        pragma Export (C, u00009, "gnatS");
        pragma Export (C, u00010, "gnat__heap_sort_aB");
        pragma Export (C, u00011, "gnat__heap_sort_aS");
        pragma Export (C, u00012, "system__exception_tableB");
        pragma Export (C, u00013, "system__exception_tableS");
        pragma Export (C, u00014, "gnat__htableB");
        pragma Export (C, u00015, "gnat__htableS");
        pragma Export (C, u00016, "system__machine_codeS");
        pragma Export (C, u00017, "system__secondary_stackB");
        pragma Export (C, u00018, "system__secondary_stackS");
        pragma Export (C, u00019, "system__parametersB");
        pragma Export (C, u00020, "system__parametersS");
        pragma Export (C, u00021, "system__soft_linksB");
        pragma Export (C, u00022, "system__soft_linksS");
        pragma Export (C, u00023, "system__stack_checkingB");
        pragma Export (C, u00024, "system__stack_checkingS");
        pragma Export (C, u00025, "system__storage_elementsB");
        pragma Export (C, u00026, "system__storage_elementsS");
        pragma Export (C, u00027, "text_ioS");
        pragma Export (C, u00028, "ada__text_ioB");
        pragma Export (C, u00029, "ada__text_ioS");
        pragma Export (C, u00030, "ada__streamsS");
        pragma Export (C, u00031, "ada__tagsB");
        pragma Export (C, u00032, "ada__tagsS");
        pragma Export (C, u00033, "interfacesS");
        pragma Export (C, u00034, "interfaces__c_streamsB");
        pragma Export (C, u00035, "interfaces__c_streamsS");
        pragma Export (C, u00036, "system__file_ioB");
        pragma Export (C, u00037, "system__file_ioS");
        pragma Export (C, u00038, "ada__finalizationB");
        pragma Export (C, u00039, "ada__finalizationS");
        pragma Export (C, u00040, "system__finalization_rootB");
        pragma Export (C, u00041, "system__finalization_rootS");
        pragma Export (C, u00042, "system__stream_attributesB");
        pragma Export (C, u00043, "system__stream_attributesS");
        pragma Export (C, u00044, "ada__io_exceptionsS");
        pragma Export (C, u00045, "system__unsigned_typesS");
        pragma Export (C, u00046, "system__finalization_implementationB");
        pragma Export (C, u00047, "system__finalization_implementationS");
        pragma Export (C, u00048, "system__string_ops_concat_3B");
        pragma Export (C, u00049, "system__string_ops_concat_3S");
        pragma Export (C, u00050, "system__string_opsB");
        pragma Export (C, u00051, "system__string_opsS");
        pragma Export (C, u00052, "system__file_control_blockS");
        pragma Export (C, u00053, "ada__finalization__list_controllerB");
        pragma Export (C, u00054, "ada__finalization__list_controllerS");
     
     end Ada_Main;
     
     --  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");
     
     package body Ada_Main is
     
     --  Generated package body for Ada_Main starts here
     
        -------------
        -- adainit --
        -------------
     
        procedure adainit is
     
           --  Set_Globals is a library routine that stores away the
           --  value of the indicated set of global values in global
           --  variables within the library.
     
           procedure Set_Globals
             (Main_Priority            : Integer;
              Time_Slice_Value         : Integer;
              WC_Encoding              : Character;
              Locking_Policy           : Character;
              Queuing_Policy           : Character;
              Task_Dispatching_Policy  : Character;
              Adafinal                 : System.Address;
              Unreserve_All_Interrupts : Boolean;
              Exception_Tracebacks     : Boolean);
           pragma Import (C, Set_Globals, "__gnat_set_globals");
     
           --  SDP_Table_Build is a library routine used to build the
           --  exception tables. See unit Ada.Exceptions in files
           --  a-except.ads/adb for full details of how zero cost
           --  exception handling works. This procedure, the call to
           --  it, and the two following tables are all omitted if the
           --  build is in longjmp/setjump exception mode.
     
           procedure SDP_Table_Build
             (SDP_Addresses   : System.Address;
              SDP_Count       : Natural;
              Elab_Addresses  : System.Address;
              Elab_Addr_Count : Natural);
           pragma Import (C, SDP_Table_Build, "__gnat_SDP_Table_Build");
     
           --  Table of Unit_Exception_Table addresses. Used for zero
           --  cost exception handling to build the top level table.
     
           ST : aliased constant array (1 .. 21) of System.Address := (
             Hello'UET_Address,
             Ada.Exceptions'UET_Address,
             Gnat.Heap_Sort_A'UET_Address,
             System.Exception_Table'UET_Address,
             System.Secondary_Stack'UET_Address,
             System.Parameters'UET_Address,
             System.Soft_Links'UET_Address,
             System.Stack_Checking'UET_Address,
             Ada.Text_Io'UET_Address,
             Ada.Streams'UET_Address,
             Ada.Tags'UET_Address,
             Interfaces.C_Streams'UET_Address,
             System.File_Io'UET_Address,
             Ada.Finalization'UET_Address,
             System.Finalization_Root'UET_Address,
             System.Stream_Attributes'UET_Address,
             System.Finalization_Implementation'UET_Address,
             System.String_Ops_Concat_3'UET_Address,
             System.String_Ops'UET_Address,
             System.File_Control_Block'UET_Address,
             Ada.Finalization.List_Controller'UET_Address);
     
           --  Table of addresses of elaboration routines. Used for
           --  zero cost exception handling to make sure these
           --  addresses are included in the top level procedure
           --  address table.
     
           EA : aliased constant array (1 .. 22) of System.Address := (
             adainit'Code_Address,
             adafinal'Code_Address,
             Ada.Exceptions'Elab_Spec'Address,
             System.Exceptions'Elab_Spec'Address,
             Interfaces.C_Streams'Elab_Spec'Address,
             System.Exception_Table'Elab_Body'Address,
             Ada.Io_Exceptions'Elab_Spec'Address,
             Ada.Tags'Elab_Spec'Address,
             Ada.Tags'Elab_Body'Address,
             Ada.Streams'Elab_Spec'Address,
             System.Stack_Checking'Elab_Spec'Address,
             System.Soft_Links'Elab_Body'Address,
             System.Secondary_Stack'Elab_Body'Address,
             Ada.Exceptions'Elab_Body'Address,
             System.Finalization_Root'Elab_Spec'Address,
             System.Finalization_Implementation'Elab_Spec'Address,
             Ada.Finalization'Elab_Spec'Address,
             Ada.Finalization.List_Controller'Elab_Spec'Address,
             System.File_Control_Block'Elab_Spec'Address,
             System.File_Io'Elab_Body'Address,
             Ada.Text_Io'Elab_Spec'Address,
             Ada.Text_Io'Elab_Body'Address);
     
        --  Start of processing for adainit
     
        begin
           --  Call SDP_Table_Build to build the top level procedure
           --  table for zero cost exception handling (omitted in
           --  longjmp/setjump mode).
     
           SDP_Table_Build (ST'Address, 21, EA'Address, 22);
     
           --  Call Set_Globals to record various information for
           --  this partition.  The values are derived by the binder
           --  from information stored in the ali files by the compiler.
     
           Set_Globals
             (Main_Priority            => -1,
              --  Priority of main program, -1 if no pragma Priority used
     
              Time_Slice_Value         => -1,
              --  Time slice from Time_Slice pragma, -1 if none used
     
              WC_Encoding              => 'b',
              --  Wide_Character encoding used, default is brackets
     
              Locking_Policy           => ' ',
              --  Locking_Policy used, default of space means not
              --  specified, otherwise it is the first character of
              --  the policy name.
     
              Queuing_Policy           => ' ',
              --  Queuing_Policy used, default of space means not
              --  specified, otherwise it is the first character of
              --  the policy name.
     
              Task_Dispatching_Policy  => ' ',
              --  Task_Dispatching_Policy used, default of space means
              --  not specified, otherwise first character of the
              --  policy name.
     
              Adafinal                 => adafinal'Address,
              --  Address of generated ada final routine
     
              Unreserve_All_Interrupts => False,
              --  Set true if pragma Unreserve_All_Interrupts was used
     
              Exception_Tracebacks     => False);
              --  Indicates if exception tracebacks are enabled
     
           --  Now we have the elaboration calls for all units in the
           --  partition. The all is commented out if the given unit
           --  has no elaboration code. We retain the commented out call
           --  to indicate the full order chosen. The Elab_Spec and
           --  Elab_Body attributes generate references to the implicit
           --  elaboration procedures generated by the compiler for each
           --  unit that requires elaboration.
     
           -- System'Elab_Spec;
           -- Ada'Elab_Spec;
           -- Gnat'Elab_Spec;
           -- Gnat.Heap_Sort_A'Elab_Spec;
           -- Gnat.Htable'Elab_Spec;
           -- Gnat.Htable'Elab_Body;
           -- Interfaces'Elab_Spec;
           -- System.Machine_Code'Elab_Spec;
           -- System.Parameters'Elab_Spec;
           -- System.Standard_Library'Elab_Spec;
              Ada.Exceptions'Elab_Spec;
              System.Exceptions'Elab_Spec;
           -- System.Parameters'Elab_Body;
           -- Gnat.Heap_Sort_A'Elab_Body;
              Interfaces.C_Streams'Elab_Spec;
           -- Interfaces.C_Streams'Elab_Body;
           -- System.Exception_Table'Elab_Spec;
              System.Exception_Table'Elab_Body;
              Ada.Io_Exceptions'Elab_Spec;
           -- System.Storage_Elements'Elab_Spec;
           -- System.Storage_Elements'Elab_Body;
           -- System.Secondary_Stack'Elab_Spec;
              Ada.Tags'Elab_Spec;
              Ada.Tags'Elab_Body;
              Ada.Streams'Elab_Spec;
              System.Stack_Checking'Elab_Spec;
           -- System.Soft_Links'Elab_Spec;
              System.Soft_Links'Elab_Body;
           -- System.Stack_Checking'Elab_Body;
              System.Secondary_Stack'Elab_Body;
              Ada.Exceptions'Elab_Body;
           -- System.Standard_Library'Elab_Body;
           -- System.String_Ops'Elab_Spec;
           -- System.String_Ops'Elab_Body;
           -- System.String_Ops_Concat_3'Elab_Spec;
           -- System.String_Ops_Concat_3'Elab_Body;
           -- System.Unsigned_Types'Elab_Spec;
           -- System.Stream_Attributes'Elab_Spec;
           -- System.Stream_Attributes'Elab_Body;
              System.Finalization_Root'Elab_Spec;
           -- System.Finalization_Root'Elab_Body;
              System.Finalization_Implementation'Elab_Spec;
           -- System.Finalization_Implementation'Elab_Body;
              Ada.Finalization'Elab_Spec;
           -- Ada.Finalization'Elab_Body;
              Ada.Finalization.List_Controller'Elab_Spec;
           -- Ada.Finalization.List_Controller'Elab_Body;
              System.File_Control_Block'Elab_Spec;
           -- System.File_Io'Elab_Spec;
              System.File_Io'Elab_Body;
              Ada.Text_Io'Elab_Spec;
              Ada.Text_Io'Elab_Body;
           -- Text_Io'Elab_Spec;
           -- hello'elab_body;
           null;
        end adainit;
     
        --------------
        -- adafinal --
        --------------
     
        procedure adafinal is
     
           --  The actual finalization is performed by calling the
           --  library routine in System.Finalization_Implementation.
     
           procedure do_finalize;
           pragma Import
             (C, do_finalize,
              "system__finalization_implementation__finalize_global_list");
     
        begin
           do_finalize;
        end adafinal;
     
        ----------
        -- 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");
     
           --  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");
     
        --  Start of processing for main
     
        begin
           --  Save global variables
     
           gnat_argc := argc;
           gnat_argv := argv;
           gnat_envp := envp;
     
           --  Call low level system initialization
     
           Initialize;
     
           --  Call our generated Ada initialization routine
     
           adainit;
     
           --  This is the point at which we want the debugger to get
           --  control
     
           Break_Start;
     
           --  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
        --   ./system.o
        --   ./ada.o
        --   ./gnat.o
        --   ./g-htable.o
        --   ./interfac.o
        --   ./s-maccod.o
        --   ./s-except.o
        --   ./s-parame.o
        --   ./g-hesora.o
        --   ./i-cstrea.o
        --   ./s-exctab.o
        --   ./a-ioexce.o
        --   ./s-stoele.o
        --   ./a-tags.o
        --   ./a-stream.o
        --   ./s-soflin.o
        --   ./s-stache.o
        --   ./s-secsta.o
        --   ./a-except.o
        --   ./s-stalib.o
        --   ./s-strops.o
        --   ./s-sopco3.o
        --   ./s-unstyp.o
        --   ./s-stratt.o
        --   ./s-finroo.o
        --   ./s-finimp.o
        --   ./a-finali.o
        --   ./a-filico.o
        --   ./s-ficobl.o
        --   ./s-fileio.o
        --   ./a-textio.o
        --   ./text_io.o
        --   ./hello.o
        --   -L./
     ...  Target/System specific library search directives
        --   -lgnat
     -- 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
blowing up during elaboration of the body of `Ada.Text_IO'. To chase
this bug down, 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).

Generating the Binder Program in C
==================================

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'.

   However for some purposes it may be convenient to generate the main
program in C rather than Ada. This may for example be helpful when you
are generating a mixed language program with the main program in C. The
GNAT compiler itself is an example. The use of the `-C' switch for both
`gnatbind' and `gnatlink' will cause the program to be generated in C
(and compiled using the gnu C compiler). The following shows the C code
generated for the same "Hello World" program:

     extern int gnat_argc;
     extern char **gnat_argv;
     extern char **gnat_envp;
     extern int gnat_exit_status;
     void adafinal ();
     
     void adainit ()
     {
        extern void *__gnat_hello__SDP;
        extern void *__gnat_ada__exceptions__SDP;
        extern void *__gnat_gnat__heap_sort_a__SDP;
        extern void *__gnat_system__exception_table__SDP;
        extern void *__gnat_system__secondary_stack__SDP;
        extern void *__gnat_system__parameters__SDP;
        extern void *__gnat_system__soft_links__SDP;
        extern void *__gnat_system__stack_checking__SDP;
        extern void *__gnat_ada__text_io__SDP;
        extern void *__gnat_ada__streams__SDP;
        extern void *__gnat_ada__tags__SDP;
        extern void *__gnat_interfaces__c_streams__SDP;
        extern void *__gnat_system__file_io__SDP;
        extern void *__gnat_ada__finalization__SDP;
        extern void *__gnat_system__finalization_root__SDP;
        extern void *__gnat_system__stream_attributes__SDP;
        extern void *__gnat_system__finalization_implementation__SDP;
        extern void *__gnat_system__string_ops_concat_3__SDP;
        extern void *__gnat_system__string_ops__SDP;
        extern void *__gnat_system__file_control_block__SDP;
        extern void *__gnat_ada__finalization__list_controller__SDP;
     
        void **st[21] = {
          &__gnat_hello__SDP,
          &__gnat_ada__exceptions__SDP,
          &__gnat_gnat__heap_sort_a__SDP,
          &__gnat_system__exception_table__SDP,
          &__gnat_system__secondary_stack__SDP,
          &__gnat_system__parameters__SDP,
          &__gnat_system__soft_links__SDP,
          &__gnat_system__stack_checking__SDP,
          &__gnat_ada__text_io__SDP,
          &__gnat_ada__streams__SDP,
          &__gnat_ada__tags__SDP,
          &__gnat_interfaces__c_streams__SDP,
          &__gnat_system__file_io__SDP,
          &__gnat_ada__finalization__SDP,
          &__gnat_system__finalization_root__SDP,
          &__gnat_system__stream_attributes__SDP,
          &__gnat_system__finalization_implementation__SDP,
          &__gnat_system__string_ops_concat_3__SDP,
          &__gnat_system__string_ops__SDP,
          &__gnat_system__file_control_block__SDP,
          &__gnat_ada__finalization__list_controller__SDP};
     
        extern void ada__exceptions___elabs ();
        extern void system__exceptions___elabs ();
        extern void interfaces__c_streams___elabs ();
        extern void system__exception_table___elabb ();
        extern void ada__io_exceptions___elabs ();
        extern void ada__tags___elabs ();
        extern void ada__tags___elabb ();
        extern void ada__streams___elabs ();
        extern void system__stack_checking___elabs ();
        extern void system__soft_links___elabb ();
        extern void system__secondary_stack___elabb ();
        extern void ada__exceptions___elabb ();
        extern void system__finalization_root___elabs ();
        extern void system__finalization_implementation___elabs ();
        extern void ada__finalization___elabs ();
        extern void ada__finalization__list_controller___elabs ();
        extern void system__file_control_block___elabs ();
        extern void system__file_io___elabb ();
        extern void ada__text_io___elabs ();
        extern void ada__text_io___elabb ();
     
        void (*ea[22]) () = {
          adainit,
          adafinal,
          ada__exceptions___elabs,
          system__exceptions___elabs,
          interfaces__c_streams___elabs,
          system__exception_table___elabb,
          ada__io_exceptions___elabs,
          ada__tags___elabs,
          ada__tags___elabb,
          ada__streams___elabs,
          system__stack_checking___elabs,
          system__soft_links___elabb,
          system__secondary_stack___elabb,
          ada__exceptions___elabb,
          system__finalization_root___elabs,
          system__finalization_implementation___elabs,
          ada__finalization___elabs,
          ada__finalization__list_controller___elabs,
          system__file_control_block___elabs,
          system__file_io___elabb,
          ada__text_io___elabs,
          ada__text_io___elabb};
     
        __gnat_SDP_Table_Build (&st, 21, ea, 22);
        __gnat_set_globals (
           -1,      /* Main_Priority              */
           -1,      /* Time_Slice_Value           */
           'b',     /* WC_Encoding                */
           ' ',     /* Locking_Policy             */
           ' ',     /* Queuing_Policy             */
           ' ',     /* Tasking_Dispatching_Policy */
           adafinal,/* Finalization routine address */
           0,       /* Unreserve_All_Interrupts */
           0);      /* Exception_Tracebacks */
     
     /* system___elabs (); */
     /* ada___elabs (); */
     /* gnat___elabs (); */
     /* gnat__heap_sort_a___elabs (); */
     /* gnat__htable___elabs (); */
     /* gnat__htable___elabb (); */
     /* interfaces___elabs (); */
     /* system__machine_code___elabs (); */
     /* system__parameters___elabs (); */
     /* system__standard_library___elabs (); */
        ada__exceptions___elabs ();
        system__exceptions___elabs ();
     /* system__parameters___elabb (); */
     /* gnat__heap_sort_a___elabb (); */
        interfaces__c_streams___elabs ();
     /* interfaces__c_streams___elabb (); */
     /* system__exception_table___elabs (); */
        system__exception_table___elabb ();
        ada__io_exceptions___elabs ();
     /* system__storage_elements___elabs (); */
     /* system__storage_elements___elabb (); */
     /* system__secondary_stack___elabs (); */
        ada__tags___elabs ();
        ada__tags___elabb ();
        ada__streams___elabs ();
        system__stack_checking___elabs ();
     /* system__soft_links___elabs (); */
        system__soft_links___elabb ();
     /* system__stack_checking___elabb (); */
        system__secondary_stack___elabb ();
        ada__exceptions___elabb ();
     /* system__standard_library___elabb (); */
     /* system__string_ops___elabs (); */
     /* system__string_ops___elabb (); */
     /* system__string_ops_concat_3___elabs (); */
     /* system__string_ops_concat_3___elabb (); */
     /* system__unsigned_types___elabs (); */
     /* system__stream_attributes___elabs (); */
     /* system__stream_attributes___elabb (); */
        system__finalization_root___elabs ();
     /* system__finalization_root___elabb (); */
        system__finalization_implementation___elabs ();
     /* system__finalization_implementation___elabb (); */
        ada__finalization___elabs ();
     /* ada__finalization___elabb (); */
        ada__finalization__list_controller___elabs ();
     /* ada__finalization__list_controller___elabb (); */
        system__file_control_block___elabs ();
     /* system__file_io___elabs (); */
        system__file_io___elabb ();
        ada__text_io___elabs ();
        ada__text_io___elabb ();
     /* text_io___elabs (); */
     /* hello___elabb (); */
     }
     void adafinal () {
        system__finalization_implementation__finalize_global_list ();
     }
     int main (argc, argv, envp)
         int argc;
         char **argv;
         char **envp;
     {
        gnat_argc = argc;
        gnat_argv = argv;
        gnat_envp = envp;
     
        __gnat_initialize();
        adainit();
        __gnat_break_start();
     
        _ada_hello ();
     
        adafinal();
        __gnat_finalize();
        exit (gnat_exit_status);
     }
     unsigned helloB = 0x425FD0AF;
     unsigned system__standard_libraryB = 0x077A2651;
     unsigned system__standard_libraryS = 0x08ADDC9E;
     unsigned systemS = 0x1D370323;
     unsigned system__exceptionsS = 0x3043D77B;
     unsigned adaS = 0x2359F9ED;
     unsigned ada__exceptionsB = 0x0CA940CF;
     unsigned ada__exceptionsS = 0x69BA6A59;
     unsigned gnatS = 0x156A40CF;
     unsigned gnat__heap_sort_aB = 0x033DABE0;
     unsigned gnat__heap_sort_aS = 0x6AB38FEA;
     unsigned system__exception_tableB = 0x7AAA368C;
     unsigned system__exception_tableS = 0x7D13B305;
     unsigned gnat__htableB = 0x62D2B79D;
     unsigned gnat__htableS = 0x2E865F1E;
     unsigned system__machine_codeS = 0x6379D875;
     unsigned system__secondary_stackB = 0x72D6A51D;
     unsigned system__secondary_stackS = 0x6E88E3D7;
     unsigned system__parametersB = 0x45C8383C;
     unsigned system__parametersS = 0x385E7AC2;
     unsigned system__soft_linksB = 0x08FE4C1F;
     unsigned system__soft_linksS = 0x23B87757;
     unsigned system__stack_checkingB = 0x3A4BFD9A;
     unsigned system__stack_checkingS = 0x4C9F3930;
     unsigned system__storage_elementsB = 0x2F1EB794;
     unsigned system__storage_elementsS = 0x0E2A461A;
     unsigned text_ioS = 0x5570D114;
     unsigned ada__text_ioB = 0x501FA6BF;
     unsigned ada__text_ioS = 0x57692181;
     unsigned ada__streamsS = 0x7C25DE96;
     unsigned ada__tagsB = 0x521B9399;
     unsigned ada__tagsS = 0x689CC1B9;
     unsigned interfacesS = 0x0357E00A;
     unsigned interfaces__c_streamsB = 0x1345CFE9;
     unsigned interfaces__c_streamsS = 0x343244DE;
     unsigned system__file_ioB = 0x6725DC79;
     unsigned system__file_ioS = 0x2DAF477E;
     unsigned ada__finalizationB = 0x4F0184F2;
     unsigned ada__finalizationS = 0x0A0669D8;
     unsigned system__finalization_rootB = 0x26610831;
     unsigned system__finalization_rootS = 0x0B5A4DF9;
     unsigned system__stream_attributesB = 0x1D4F93E8;
     unsigned system__stream_attributesS = 0x30B2EC3D;
     unsigned ada__io_exceptionsS = 0x34054F96;
     unsigned system__unsigned_typesS = 0x6598BA3E;
     unsigned system__finalization_implementationB = 0x2C9C021D;
     unsigned system__finalization_implementationS = 0x177A51F6;
     unsigned system__string_ops_concat_3B = 0x1CBC39CD;
     unsigned system__string_ops_concat_3S = 0x5461BB3E;
     unsigned system__string_opsB = 0x03F36D98;
     unsigned system__string_opsS = 0x208D3EF6;
     unsigned system__file_control_blockS = 0x33AF4230;
     unsigned ada__finalization__list_controllerB = 0x0B97C6BF;
     unsigned ada__finalization__list_controllerS = 0x34B32999;
     
     /* BEGIN Object file/option list
     ./system.o
     ./ada.o
     ./gnat.o
     ./g-htable.o
     ./interfac.o
     ./s-maccod.o
     ./s-except.o
     ./s-parame.o
     ./g-hesora.o
     ./i-cstrea.o
     ./s-exctab.o
     ./a-ioexce.o
     ./s-stoele.o
     ./a-tags.o
     ./a-stream.o
     ./s-soflin.o
     ./s-stache.o
     ./s-secsta.o
     ./a-except.o
     ./s-stalib.o
     ./s-strops.o
     ./s-sopco3.o
     ./s-unstyp.o
     ./s-stratt.o
     ./s-finroo.o
     ./s-finimp.o
     ./a-finali.o
     ./a-filico.o
     ./s-ficobl.o
     ./s-fileio.o
     ./a-textio.o
     ./text_io.o
     ./hello.o
     -L./
     ...  Target/System specific library search directives
     -lgnat
        END Object file/option list */

Here again, the C code is exactly what is generated by the binder. The
functions of the various parts of this code correspond in an obvious
manner with the commented Ada code shown in the example in the previous
section.

Consistency-Checking Modes
==========================

As described in the previous section, 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'
     Require source files to be present. In this mode, the binder must
     be able to locate all source files that are referenced, in order
     to check their consistency. In normal mode, if a source file
     cannot be located it is simply ignored. If you specify this
     switch, a missing source file is an error.

`-x'
     Exclude source files. In this mode, the binder only checks that ALI
     files are consistent with one another. Source files are not
     accessed.  The binder runs faster in this mode, and there is still
     a guarantee that the resulting program is self-consistent.  If a
     source file has been edited since it was last compiled, and you
     specify this switch, the binder will not detect that the object
     file is out of date with respect to the source file. Note that
     this is the mode that is automatically used by `gnatmake' because
     in this case the checking against sources has already been
     performed by `gnatmake' in the course of compilation (i.e. before
     binding).

Binder Error Message Control
============================

The following switches provide control over the generation of error
messages from the binder:

`-v'
     Verbose mode. In the normal mode, brief error messages are
     generated to `stderr'. 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.

`-b'
     Generate brief error messages to `stderr' even if verbose mode is
     specified. This is relevant only when used with the `-v' switch.

`-mN'
     Limits the number of error messages to N, a decimal integer in the
     range 1-999. The binder terminates immediately if this limit is
     reached.

`-MXXX'
     Renames the generated main program from `main' to `xxx'.  This is
     useful in the case of some cross-building environments, where the
     actual main program is separate from the one generated by
     `gnatbind'.

`-ws'
     Suppress all warning messages.

`-we'
     Treat any warning messages as fatal errors.

`-t'
     The binder performs a number of consistency checks including:

        * Check that time stamps of a given source unit are consistent

        * Check that checksums of a given source unit are consistent

        * Check that consistent versions of `GNAT' were used for
          compilation

        * Check consistency of configuration pragmas as required

     Normally 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.  *This means that `-t' should be used only in
     unusual situations, with extreme care.*

Elaboration Control
===================

The following switches provide additional control over the elaboration
order. For full details see *Note Elaboration Order Handling in GNAT::.

`-f'
     Instructs the binder to ignore directives from the compiler about
     implied `Elaborate_All' pragmas, and to use full Ada 95 Reference
     Manual semantics in an attempt to find a legal elaboration order,
     even if it seems likely that this order will cause an elaboration
     exception.

`-p'
     Normally the binder attempts to choose an elaboration order that is
     likely to minimize the likelihood of an elaboration order error
     resulting in raising a `Program_Error' exception. This switch
     reverses the action of the binder, and requests that it
     deliberately choose an order that is likely to maximize the
     likelihood of an elaboration error.  This is useful in ensuring
     portability and avoiding dependence on accidental fortuitous
     elaboration ordering.

     Normally 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_All' pragmas are implicitly
     inserted. These implicit pragmas are still respected by the binder
     in `-p' mode, so a safe elaboration order is assured.

Output Control
==============

The following switches allow additional control over the output
generated by the binder.

`-A'
     Generate binder program in Ada (default). The binder program is
     named `b~MAINPROG.adb' by default. This can be changed with `-o'
     `gnatbind' option.

`-C'
     Generate binder program in C. The binder program is named
     `b_MAINPROG.c'. This can be changed with `-o' `gnatbind' option.

`-e'
     Output complete list of elaboration-order dependencies, showing the
     reason for each dependency. This output can be rather extensive
     but may be useful in diagnosing problems with elaboration order.
     The output is written to `stdout'.

`-h'
     Output usage information. The output is written to `stdout'.

`-l'
     Output chosen elaboration order. The output is written to `stdout'.

`-O'
     Output full names of all the object files that must be linked to
     provide the Ada component of the program. The output is written to
     `stdout'.  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'
     Set name of output file to FILE instead of the normal
     `b~MAINPROG.adb' default. Note that FILE denote the Ada binder
     generated body filename. In C mode you would normally give FILE an
     extension of `.c' because it will be a C source program.  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.

`-c'
     Check only. Do not generate the binder output file. In this mode
     the binder performs all error checks but does not generate an
     output file.

Binding with Non-Ada Main Programs
==================================

In our description so far we have 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 (*note Mixed Language Programming::.).  The following switch
is used in this situation:

`-n'
     No main program. The main program is not in Ada.

In 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\/OUTPUT=file\'.
The output is an Ada unit in source form that can be compiled with GNAT
unless the -C switch is used in which case the output is a C source
file, which must be compiled using the C compiler.  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.

Binding Programs with No Main Subprogram
========================================

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:

`-z'
     Normally the binder checks that the unit name given on the command
     line corresponds to a suitable main subprogram. When this switch
     is used, a list of ALI files can be given, and the execution of
     the program consists of elaboration of these units in an
     appropriate order.

Summary of Binder Switches
==========================

The following are the switches available with `gnatbind':

`-aO'
     Specify directory to be searched for ALI files.

`-aI'
     Specify directory to be searched for source file.

`-A'
     Generate binder program in Ada (default)

`-b'
     Generate brief messages to `stderr' even if verbose mode set.

`-c'
     Check only, no generation of binder output file.

`-C'
     Generate binder program in C

`-e'
     Output complete list of elaboration-order dependencies.

`-E'
     Store tracebacks in exception occurrences when the target supports
     it.  This is the default with the zero cost exception mechanism.
     This option is currently only supported on Solaris, Linux and
     Windows ix86. Under Solaris and Linux you need to use explicitly
     the `gcc' flag `-funwind-tables' when compiling every file in your
     application. See also the packages `GNAT.Traceback' and
     `GNAT.Traceback.Symbolic'. Under Windows there is no specific
     option to use to enable this feature but you must not use
     `-fomit-frame-pointer' `gcc' option.

`-f'
     On some targets, the command line length is limited, and `gnatlink'
     will generate a separate file for the linker if the list of object
     files is too long. The `-f' flag forces this file to be generated
     even if the limit is not exceeded. This is useful in some cases to
     deal with special situations where the command line length is
     exceeded.

`-h'
     Output usage (help) information

`-I'
     Specify directory to be searched for source and ALI files.

`-I-'
     Do not look for sources in the current directory where `gnatbind'
     was invoked, and do not look for ALI files in the directory
     containing the ALI file named in the `gnatbind' command line.

`-l'
     Output chosen elaboration order.

`-Mxyz'
     Rename generated main program from main to xyz

`-mN'
     Limit number of detected errors to N (1-999).

`-n'
     No main program.

`-nostdinc'
     Do not look for sources in the system default directory.

`-nostdlib'
     Do not look for library files in the system default directory.

`-o FILE'
     Name the output file FILE (default is `b~XXX.adb').  Note that if
     this option is used, then linking must be done manually, gnatlink
     cannot be used.

`-O'
     Output object list.

`-p'
     Pessimistic (worst-case) elaboration order

`-s'
     Require all source files to be present.

`-static'
     Link against a static GNAT run time.

`-shared'
     Link against a shared GNAT run time when available.

`-t'
     Tolerate time stamp and other consistency errors

`-TN'
     Set the time slice value to n microseconds. A value of zero means
     no time slicing and also indicates to the tasking run time to
     match as close as possible to the annex D requirements of the RM.

`-v'
     Verbose mode. Write error messages, header, summary output to
     `stdout'.

`-wX'
     Warning mode (X=s/e for suppress/treat as error)

`-x'
     Exclude source files (check object consistency only).

`-z'
     No main subprogram.

   You may obtain this listing by running the program `gnatbind' with
no arguments.

Command-Line Access
===================

The 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.

Search Paths for `gnatbind'
===========================

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'
(*note Search Paths and the Run-Time Library (RTL)::.). For ALI files
the directories searched are:

  1. The directory containing the ALI file named in the command line,
     unless the switch `-I-' is specified.

  2. All directories specified by `-I' switches on the `gnatbind'
     command line, in the order given.

  3. Each of the directories listed in the value of the
     `ADA_OBJECTS_PATH' environment variable.  Construct this value
     exactly as the `PATH' environment variable: a list of directory
     names separated by colons.

  4. The content of the "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.  *Note Installing an Ada 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.

Examples of `gnatbind' Usage
============================

This section contains a number of examples of using the GNAT binding
utility `gnatbind'.

`gnatbind hello'
     The main program `Hello' (source program in `hello.adb') is bound
     using the standard switch settings. The generated main program is
     `b~hello.adb'. This is the normal, default use of the binder.

`gnatbind hello -o mainprog.adb'
     The main program `Hello' (source program in `hello.adb') is bound
     using the standard switch settings. The generated main program is
     `mainprog.adb' with the associated spec in `mainprog.ads'. Note
     that you must specify the body here not the spec, in the case
     where the output is in Ada. Note that if this option is used, then
     linking must be done manually, since gnatlink will not be able to
     find the generated file.

`gnatbind main -C -o mainprog.c -x'
     The main program `Main' (source program in `main.adb') is bound,
     excluding source files from the consistency checking, generating
     the file `mainprog.c'.

`gnatbind -x main_program -C -o mainprog.c'
     This command is exactly the same as the previous example. Switches
     may appear anywhere in the command line, and single letter
     switches may be combined into a single switch.

`gnatbind -n math dbase -C -o ada-control.c'
     The main program is in a language other than Ada, but calls to
     subprograms in packages `Math' and `Dbase' appear. This call to
     `gnatbind' generates the file `ada-control.c' containing the
     `adainit' and `adafinal' routines to be called before and after
     accessing the Ada units.

Linking Using `gnatlink'
************************

This chapter discusses `gnatlink', a utility program used to link Ada
programs and build an executable file. This is a simple program that
invokes the UNIX 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 binder to determine
this list.

Running `gnatlink'
==================

The form of the `gnatlink' command is

     $ gnatlink [SWITCHES] MAINPROG[.ali] [NON-ADA OBJECTS]
           [LINKER OPTIONS]

`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 UNIX linker command file to create the
executable.

   The arguments following `MAINPROG.ali' 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 `pragma Import' statements 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 usually called LD. 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, the GCC switches
`-Xlinker' or `-Wl,' shall be used. Refer to the GCC documentation for
details. Here is an example showing how to generate a linker map
assuming that the underlying linker is GNU ld:

     $ gnatlink my_prog -Wl,-Map,MAPFILE

   Using LINKER OPTIONS it is possible to set the program stack and
heap size. See *note Setting Stack Size from gnatlink::. and *note
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.

Switches for `gnatlink'
=======================

The following switches are available with the `gnatlink' utility:

`-A'
     The binder has generated code in Ada. This is the default.

`-C'
     If instead of generating a file in Ada, the binder has generated
     one in C, then the linker needs to know about it. Use this switch
     to signal to `gnatlink' that the binder has generated C code
     rather than Ada code.

`-g'
     The option to include debugging information causes the Ada bind
     file (in other words, `b~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. The same procedure apply if
     a C bind file was generated using `-C' `gnatbind' option, in this
     case the filenames are `b_MAINPROG.c' and `b_MAINPROG.o'.

`-n'
     Do not compile the file generated by the binder. This may be used
     when a link is rerun with different options, but there is no need
     to recompile the binder file.

`-v'
     Causes additional information to be output, including a full list
     of the included object files. This switch option is most useful
     when you want to see what set of object files are being used in
     the link step.

`-v -v'
     Very verbose mode. Requests that the compiler operate in verbose
     mode when it compiles the binder file, and that the system linker
     run in verbose mode.

`-o EXEC-NAME'
     EXEC-NAME specifies an alternate name for the generated executable
     program. If this switch is omitted, the executable has the same
     name as the main unit. For example, `gnatlink try.ali' creates an
     executable called `try'.

`-b TARGET'
     Compile your program to run on TARGET, which is the name of a
     system configuration. You must have a GNAT cross-compiler built if
     TARGET is not the same as your host system.

`-BDIR'
     Load compiler executables (for example, `gnat1', the Ada compiler)
     from DIR instead of the default location. Only use this switch
     when multiple versions of the GNAT compiler are available. See the
     `gcc' manual page for further details. You would normally use the
     `-b' or `-V' switch instead.

`--GCC=COMPILER_NAME'
     Program used for compiling the binder file. The default is `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'.  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'
     NAME is the name of the linker to be invoked. This is especially
     useful in mixed language programs since languages such as c++
     require their own linker to be used. When this switch is omitted,
     the default name for the linker is (`gcc'). When this switch is
     used, the specified linker is called instead of (`gcc') with
     exactly the same parameters that would have been passed to (`gcc')
     so if the desired linker requires different parameters it is
     necessary to use a wrapper script that massages the parameters
     before invoking the real linker. It may be useful to control the
     exact invocation by using the verbose switch.

Setting Stack Size from `gnatlink'
==================================

It is possible to specify the program stack size from `gnatlink'.
Assuming that the underlying linker is GNU ld there is two ways to do
so:

   * using `-Xlinker' linker option

          $ gnatlink hello -Xlinker --stack=0x10000,0x1000

     This set the stack reserve size to 0x10000 bytes and the stack
     commit size to 0x1000 bytes.

   * using `-Wl' linker option

          $ gnatlink hello -Wl,--stack=0x1000000

     This set 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.

Setting Heap Size from `gnatlink'
=================================

It is possible to specify the program heap size from `gnatlink'.
Assuming that the underlying linker is GNU ld there is two ways to do
so:

   * using `-Xlinker' linker option

          $ gnatlink hello -Xlinker --heap=0x10000,0x1000

     This set the heap reserve size to 0x10000 bytes and the heap commit
     size to 0x1000 bytes.

   * using `-Wl' linker option

          $ gnatlink hello -Wl,--heap=0x1000000

     This set 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.

The GNAT Make Program `gnatmake'
********************************

A typical development cycle when working on an Ada program consists of
the following steps:

  1. Edit some sources to fix bugs.

  2. Add enhancements.

  3. Compile all sources affected.

  4. Rebind and relink.

  5. Test.

The third step 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'.

Running `gnatmake'
==================

The form of the `gnatmake' command is

     $ gnatmake [SWITCHES] FILE_NAME [MODE_SWITCHES]

The only required argument is FILE_NAME, which specifies the
compilation unit that is the main program. If `switches' are present,
they can be placed before of after FILE_NAME.  If MODE_SWITCHES are
present, they must always be placed after 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 argument. 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 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 *Note Search Paths and the
Run-Time Library (RTL)::.

   All `gnatmake' output (except when you specify `-M') is to `stderr'.
The output produced by the `-M' switch is send to `stdout'.

Switches for `gnatmake'
=======================

You may specify any of the following switches to `gnatmake':

`--GCC=COMPILER_NAME'
     Program used for compiling. The default is `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. 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'
     Program used for binding. The default is `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'.

`--GNATLINK=LINKER_NAME'
     Program used for linking. The default is `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'
     Consider all files in the make process, even the GNAT internal
     system files (for example, the predefined Ada library files), as
     well as any locked files. Locked files are files whose ALI file is
     write-protected.  By default, `gnatmake' does not check these
     files, because the assumption is that the GNAT internal files are
     properly up to date, and also that any write protected ALI files
     have been properly installed. Note that if there is an
     installation problem, such that one of these files is not up to
     date, it will be properly caught by the binder.  You may have to
     specify this switch if you are working on GNAT itself. `-a' is
     also useful in conjunction with `-f' if you need to recompile an
     entire application, including run-time files, using special
     configuration pragma settings, such as a non-standard
     `Float_Representation' pragma.  By default `gnatmake -a' compiles
     all GNAT internal files with `gcc -c -gnatg' rather than `gcc -c'.

`-c'
     Compile only. Do not perform binding and linking. If the root unit
     specified by FILE_NAME is not a main unit, this is the default.
     Otherwise `gnatmake' will attempt binding and linking unless all
     objects are up to date and the executable is more recent than the
     objects.

`-f'
     Force recompilations. Recompile all sources, even though some
     object files may be up to date, but don't recompile predefined or
     GNAT internal files or locked files (files with a write-protected
     ALI file), unless the `-a' switch is also specified.

`-i'
     In normal mode, `gnatmake' compiles all object files and ALI files
     into the current directory. If the `-i' switch is used, then
     instead object files and ALI files that already exist are
     overwritten in place. This means that once a large project is
     organized into separate directories in the desired manner, then
     `gnatmake' will automatically maintain and update this
     organization. If no ALI files are found on the Ada object path
     (*Note Search Paths and the Run-Time Library (RTL)::), the new
     object and ALI files are created in the directory containing the
     source being compiled. If another organization is desired, where
     objects and sources are kept in different directories, a useful
     technique is to create dummy ALI files in the desired directories.
     When detecting such a dummy file, `gnatmake' will be forced to
     recompile the corresponding source file, and it will be put the
     resulting object and ALI files in the directory where it found the
     dummy file.

`-jN'
     Use N processes to carry out the (re)compilations. On a
     multiprocessor machine compilations will occur in parallel. In the
     event of compilation errors, messages from various compilations
     might get interspersed (but `gnatmake' will give you the full
     ordered list of failing compiles at the end). If this is
     problematic, rerun the make process with n set to 1 to get a clean
     list of messages.

`-k'
     Keep going. Continue as much as possible after a compilation
     error. To ease the programmer's task in case of compilation
     errors, the list of sources for which the compile fails is given
     when `gnatmake' terminates.

`-m'
     Specifies that the minimum necessary amount of recompilations be
     performed. In this mode `gnatmake' ignores time stamp differences
     when the only modifications to a source file consist in
     adding/removing comments, empty lines, spaces or tabs. This means
     that if you have changed the comments in a source file or have
     simply reformatted it, using this switch will tell gnatmake not to
     recompile files that depend on it (provided other sources on which
     these files depend have undergone no semantic modifications).

`-M'
     Check if all objects are up to date. If they are, output the object
     dependences to `stdout' 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'
     Don't compile, bind, or link. Checks if all objects are up to date.
     If they are not, the full name of the first file that needs to be
     recompiled is printed.  Repeated use of this option, followed by
     compiling the indicated source file, will eventually result in
     recompiling all required units.

`-o EXEC_NAME'
     Output executable name. The name of the final executable program
     will be EXEC_NAME. If the `-o' switch is omitted the default name
     for the executable will be the name of the input file in
     appropriate form for an executable file on the host system.

`-q'
     Quiet. When this flag is not set, the commands carried out by
     `gnatmake' are displayed.

`-s'
     Recompile if compiler switches have changed since last compilation.
     All compiler switches but -I and -o are taken into account in the
     following way: orders between different first letter" switches are
     ignored, but orders between same switches are taken into account.
     For example, `-O -O2' is different than `-O2 -O', but `-g -O' is
     equivalent to `-O -g'.

`-u'
     Unique. Recompile at most the main file. It implies -c. Combined
     with -f, it is equivalent to calling the compiler directly.

`-v'
     Verbose. Displays the reason for all recompilations `gnatmake'
     decides are necessary.

`-z'
     No main subprogram. Bind and link the program even if the unit name
     given on the command line is a package name. The resulting
     executable will execute the elaboration routines of the package
     and its closure, then the finalization routines.

``gcc' switches'
     The switch `-g' or any uppercase switch (other than `-A', or `-L')
     or any switch that is more than one character is passed to `gcc'
     (e.g. `-O', `-gnato,' etc.)

Source and library search path switches:

`-aIDIR'
     When looking for source files also look in directory DIR.  The
     order in which source files search is undertaken is described in
     *Note Search Paths and the Run-Time Library (RTL)::.

`-aLDIR'
     Consider DIR as being an externally provided Ada library.
     Instructs `gnatmake' to skip compilation units whose `.ali' files
     have been located in directory DIR. This allows you to have
     missing bodies for the units in DIR. You still need to specify the
     location of the specs for these units by using the switches
     `-aIDIR' or `-IDIR'.  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.

`-aODIR'
     When searching for library and object files, look in directory
     DIR. The order in which library files are searched is described in
     *Note Search Paths for gnatbind::.

`-ADIR'
     Equivalent to `-aLDIR -aIDIR'.

`-IDIR'
     Equivalent to `-aODIR -aIDIR'.

`-I-'
     Do not look for source files in the directory containing the source
     file named in the command line.  Do not look for ALI or object
     files in the directory where `gnatmake' was invoked.

`-LDIR'
     Add directory DIR to the list of directories in which the linker
     will search for libraries. This is equivalent to `-largs -L'DIR.

`-nostdinc'
     Do not look for source files in the system default directory.

`-nostdlib'
     Do not look for library files in the system default directory.

Mode Switches for `gnatmake'
============================

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'
     Compiler switches. Here SWITCHES is a list of switches that are
     valid switches for `gcc'. They will be passed on to all compile
     steps performed by `gnatmake'.

`-bargs SWITCHES'
     Binder switches. Here SWITCHES is a list of switches that are
     valid switches for `gcc'. They will be passed on to all bind steps
     performed by `gnatmake'.

`-largs SWITCHES'
     Linker switches. Here SWITCHES is a list of switches that are
     valid switches for `gcc'. They will be passed on to all link steps
     performed by `gnatmake'.

Notes on the Command Line
=========================

This section contains some additional useful notes on the operation of
the `gnatmake' command.

   * If `gnatmake' finds no ALI files, it recompiles the main program
     and all other units required by the main program.  This means that
     `gnatmake' can be used for the initial compile, as well as during
     subsequent steps of the development cycle.

   * If you enter `gnatmake FILE.adb', where `FILE.adb' is a subunit or
     body of a generic unit, `gnatmake' recompiles `FILE.adb' (because
     it finds no ALI) and stops, issuing a warning.

   * In `gnatmake' the switch `-I' is used to specify both source and
     library file paths. Use `-aI' instead if you just want to specify
     source paths only and `-aO' if you want to specify library paths
     only.

   * `gnatmake' examines both an ALI file and its corresponding object
     file for consistency. If an ALI is more recent than its
     corresponding object, or if the object file is missing, the
     corresponding source will be recompiled.  Note that `gnatmake'
     expects an ALI and the corresponding object file to be in the same
     directory.

   * `gnatmake' will ignore any files whose ALI file is write-protected.
     This may conveniently be used to exclude standard libraries from
     consideration and in particular it means that the use of the `-f'
     switch will not recompile these files unless `-a' is also
     specified.

   * `gnatmake' has been designed to make the use of Ada libraries
     particularly convenient. Assume you have an Ada library organized
     as follows: OBJ-DIR contains the objects and ALI files for of your
     Ada compilation units, whereas INCLUDE-DIR contains the specs of
     these units, but no bodies. Then to compile a unit stored in
     `main.adb', which uses this Ada library you would just type

          $ gnatmake -aIINCLUDE-DIR  -aLOBJ-DIR  main

   * Using `gnatmake' along with the `-m (minimal recompilation)'
     switch provides an extremely powerful tool: you can freely update
     the comments/format of your source files without having to
     recompile everything. Note, however, that adding or deleting lines
     in a source files may render its debugging info obsolete. If the
     file in question is a spec, the impact is rather limited, as that
     debugging info will only be useful during the elaboration phase of
     your program. For bodies the impact can be more significant. In
     all events, your debugger will warn you if a source file is more
     recent than the corresponding object, and therefore obsolescence of
     debugging information will go unnoticed.

How `gnatmake' Works
====================

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 its time stamp predates that of the
object file 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.

Examples of `gnatmake' Usage
============================

`gnatmake hello.adb'
     Compile all files necessary to bind and link the main program
     `hello.adb' (containing unit `Hello') and bind and link the
     resulting object files to generate an executable file `hello'.

`gnatmake -q Main_Unit -cargs -O2 -bargs -l'
     Compile all files necessary to bind and link the main program unit
     `Main_Unit' (from file `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.

Renaming Files Using `gnatchop'
*******************************

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.

Handling Files with Multiple Units
==================================

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.

Operating gnatchop in Compilation Mode
======================================

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 the section on handling of
configuration pragmas *note 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.

Command Line for `gnatchop'
===========================

The `gnatchop' command has the form:

     $ gnatchop switches FILE NAME [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 Text_IO; use 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.

Switches for `gnatchop'
=======================

`gnatchop' recognizes the following switches:

`-c'
     Causes `gnatchop' to operate in compilation mode, in which
     configuration pragmas are handled according to strict RM rules. See
     previous section for a full description of this mode.

`-gnatxxx'
     This passes the given `-gnatxxx' switch to `gnat' which is used to
     parse the given file. Not all `xxx' options make sense, but for
     example, the use of `-gnati2' allows `gnatchop' to process a
     source file that uses Latin-2 coding for identifiers.

`-h'
     Causes `gnatchop' to generate a brief help summary to the standard
     output file showing usage information.

`-kMM'
     Limit generated file names to the specified number `mm' of
     characters.  This is useful if the resulting set of files is
     required to be interoperable with systems which limit the length
     of file names.  No space is allowed between the `-k' and the
     numeric value. The numeric value may be omitted in which case a
     default of `-k8', suitable for use with DOS-like file systems, is
     used. If no `-k' switch is present then there is no limit on the
     length of file names.

`-q'
     Causes output of informational messages indicating the set of
     generated files to be suppressed. Warnings and error messages are
     unaffected.

`-r'
     Generate `Source_Reference' pragmas. Use this switch if the output
     files are regarded as temporary and development is to be done in
     terms of the original unchopped file. This switch causes
     `Source_Reference' pragmas to be inserted into each of the
     generated files to refers back to the original file name and line
     number.  The result is that all error messages refer back to the
     original unchopped file.  In addition, the debugging information
     placed into the object file (when the `-g' switch of `gcc' or
     `gnatmake' is specified) also refers back to this original file so
     that tools like profilers and debuggers will give information in
     terms of the original unchopped file.

     If 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'
     Causes `gnatchop' to operate in verbose mode. The version number
     and copyright notice are output, as well as exact copies of the
     gnat1 commands spawned to obtain the chop control information.

`-w'
     Overwrite existing file names. Normally `gnatchop' regards it as a
     fatal error if there is already a file with the same name as a
     file it would otherwise output, in other words if the files to be
     chopped contain duplicated units. This switch bypasses this check,
     and causes all but the last instance of such duplicated units to
     be skipped.

Examples of `gnatchop' Usage
============================

`gnatchop -w hello_s.ada ichbiah/files'
     Chops the source file `hello_s.ada'. The output files will be
     placed in the directory `ichbiah/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
*********************

In Ada 95, configuration pragmas include those pragmas described as
such in the Ada 95 Reference Manual, as well as
implementation-dependent pragmas that are configuration pragmas. See the
individual descriptions of pragmas 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.

Handling of Configuration Pragmas
=================================

   Configuration pragmas may either appear at the start of a compilation
unit, in which case they apply only to that unit, or they may 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 section *note 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.

The Configuration Pragmas File
==============================

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.

   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.

Elaboration Order Handling in GNAT
**********************************

This chapter describes the handling of elaboration code in Ada 95 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.

Elaboration Code in Ada 95
==========================

Ada 95 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:

Initializers for variables.
     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);

Package initialization code
     Code in a `BEGIN-END' section at the outer level of a package body
     is executed as part of the package body elaboration code.

Library level task allocators
     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_2' will occur, but not the call to `Func_1.'  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.

Checking the Elaboration Order in Ada 95
========================================

In some languages that involve the same kind of elaboration problems,
e.g. Java and C++, the programmer is expected to worry about these
ordering problems himself, 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 95 is designed
to be a safe language, and a programmer-beware approach is clearly not
sufficient. Consequently, the language provides three lines of defense:

Standard rules
     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 elaboration checks
     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.

Elaboration control
     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 95 has two
rules which are a little more restrictive, but easier to check, and
easier to state:

Restrictions on calls
     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.

Restrictions on instantiations
     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.

Controlling the Elaboration Order in Ada 95
===========================================

In the previous section we discussed the rules in Ada 95 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 95 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:

packages that do not require a body
     In Ada 95, a library package that does not require a body does not
     permit a body. This means that 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

pragma Pure
     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.

pragma Preelaborate
     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.

pragma Elaborate_Body
     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 in
Ada 95 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:

pragma Elaborate (unit)
     This pragma is placed in the context clause, after a `with'
     statement, 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.

pragma Elaborate_All (unit)
     This is a stronger version of the Elaborate pragma. Consider the
     following example:

          Unit A `with''s unit B and calls B.Func in elab code
          Unit B `with''s 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 95 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 `with''ed unit, or
instantiate a generic unit in a `with''ed unit, then if the `with''ed
unit does not have pragma Pure, Preelaborate, or Elaborate_Body, then
the client should have an Elaborate_All for the `with''ed unit. By
following this rule a client is assured that calls can be made without
risk of an exception.  If this rule is not followed, then a program may
be in one of four states:

No order exists
     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 95 compiler must diagnose the situation at bind
     time, and refuse to build an executable program.

One or more orders exist, all incorrect
     One or more acceptable elaboration orders exists, 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.

Several orders exist, some right, some incorrect
     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 or more orders exists, all correct
     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 Elaborate_All rule
is that the program continues to stay in the ideal (all orders OK) state
even if maintenance changes some bodies of some subprograms.
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.

Controlling Elaboration in GNAT - Internal Calls
================================================

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 95, 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, 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:

   * Compile with the `-gnatws' switch set

   * Suppress `Elaboration_Checks' for the called subprogram

   * Use pragma `Warnings_Off' to turn warnings off for the call

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_Checks)' 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.

Controlling Elaboration in GNAT - External Calls
================================================

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 in fact always 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.

Default Behavior in GNAT - Ensuring Safety
==========================================

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 `with''ed unit, or instantiate a generic unit
in a `with''ed unit, then if the `with''ed unit does not have pragma
`Pure', `Preelaborate', or `Elaborate_Body', then the client should
have an `Elaborate_All' for the `with''ed 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_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_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_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 `-gnatwl' (warn on elaboration problems) switch. This
will cause warning messages to be generated indicating the missing
`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, it is safer to include
the pragma explicitly in the source. If this unit is compiled with the
`-gnatwl' switch, then the compiler outputs a warning:

     1. with k;
     2. package j is
     3.   m : integer := k.r;
                          |
        >>> warning: call to "r" may raise Program_Error
        >>> warning: missing pragma Elaborate_All for "k"
     
     4. end;

and these warnings can be used as a guide for supplying manually the
missing pragmas.

   This default mode is not strictly compatible with 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 (RM);

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 is issue
is further discussed in the section "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, you are assured that your program will work using
the dynamic model.

Elaboration Issues for Library Tasks
====================================

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:

  1. In the body of `Decls' a call is made from within the body of a
     library task to a subprogram in the package `Utils'. Since this
     call may occur at elaboration time (given that the task is
     activated at elaboration time), we have to assume the worst, i.e.
     that the call does happen at elaboration time.

  2. This means that the body and spec of `Util' must be elaborated
     before the body of `Decls' so that this call does not cause an
     access before elaboration.

  3. Within the body of `Util', specifically within the body of
     `Util.Put_Val' there may be calls to any unit `with''ed by this
     package.

  4. One such `with''ed package is package `Decls', so there might be a
     call to a subprogram in `Decls' in `Put_Val'.  In fact there is
     such a call in this example, but we would have to assume that
     there was such a call even if it were not there, since we are not
     supposed to write the body of `Decls' knowing what is in the body
     of `Utils'; certainly in the case of the static elaboration model,
     the compiler does not know what is in other bodies and must assume
     the worst.

  5. This means that the spec and body of `Decls' must also be
     elaborated before we elaborate the unit containing the call, but
     that unit is `Decls'! This means that the body of `Decls' must be
     elaborated before itself, and that's a circularity.

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:

   * Use the dynamic model of elaboration.

     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.

   * Split library tasks into separate packages.

     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.

   * Declare separate task types.

     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 95 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.

   * Use No_Entry_Calls_In_Elaboration_Code restriction.

     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.

Mixing Elaboration Models
=========================

So far, we have assumed that the entire program is either compiled
using the dynamic model or static model, ensuring consistencty. 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 `with'ed' 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:

   * The `with'ed' unit is itself compiled with dynamic elaboration
     checks (that is with the `-gnatE' switch.

   * The `with'ed' unit is an internal GNAT implementation unit from
     the System, Interfaces, Ada, or GNAT hierarchies.

   * The `with'ed' unit has pragma Preelaborate or pragma Pure.

   * The `with'ing' unit (that is the client) has an explicit pragma
     `Elaborate_All' for the `with'ed' unit.

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, without
allowing the main application that uses this subsystem to be compiled
using the more reliable default static model.

What to Do If the Default Elaboration Behavior Fails
====================================================

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 -gnatwl
     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.

Fix the program
     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.

Perform dynamic checks
     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_All' pragmas.  The
     behavior then is exactly as specified in the Ada 95 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.

Suppress checks
     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, then you can use the `-f' switch for the `gnatbind'
     step, or `-bargs -f' if you are using `gnatmake'.  This switch
     tells the binder to ignore any implicit `Elaborate_All' pragmas
     that were generated by the compiler, and suppresses any
     circularity messages that they cause. The resulting executable
     will work properly if there are no elaboration problems, but if
     there are some order of elaboration problems they will not be
     detected, and unexpected results may occur.

It is hard to generalize on which of these three 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' or `-f' but
the danger in either 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 warnings generated with the `-gnatwl'
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 `-h
(horrible elaboration order)' switch for `gnatbind'.  Normally the
binder tries to find an order that has the best chance of of avoiding
elaboration problems. With this switch, 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 ACVC 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 ACVC tests.)

Elaboration for Access-to-Subprogram Values
===========================================

The introduction of access-to-subprogram types in Ada 95 complicates
the handling of elaboration. The trouble is that it becomes impossible
to tell at compile time which procedure is being called. This means
that it is not possible for the binder to analyze the elaboration
requirements in this case.

   If at the point at which  the access value is created, the body of
the subprogram is known to have been elaborated, then the access value
is safe, and its use does not require a check. This may be achieved by
appropriate arrangement of the order of declarations if the subprogram
is in the current unit, or, if the subprogram is in another unit, by
using pragma `Pure', `Preelaborate', or `Elaborate_Body' on the
referenced unit.

   If the referenced body is not known to have been elaborated at the
point the access value is created, then any use of the access value
must do a dynamic check, and this dynamic check will fail and raise a
`Program_Error' exception if the body has not been elaborated yet.
GNAT will generate the necessary checks, and in addition, if the
`-gnatwl' switch is set, will generate warnings that such checks are
required.

   The use of dynamic dispatching for tagged types similarly generates
a requirement for dynamic checks, and premature calls to any primitive
operation of a tagged type before the body of the operation has been
elaborated, will result in the raising of `Program_Error'.

Summary of Procedures for Elaboration Control
=============================================

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, then use the `-gnatwl' switch to generate warnings
about missing `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, or, if you are sure there really are no
elaboration problems, use the `-f' switch for the binder to cause it to
ignore implicit `Elaborate_All' pragmas generated by the compiler.

Other Elaboration Order Considerations
======================================

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 Calc;
     end Init_Constants;
     
     with Constants;
     package body Init_Constants is
        procedure Calc 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
     Main body
     Init_Constants 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.

   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_All' pragmas to ensure the desired order.

The Cross-Referencing Tools `gnatxref' and `gnatfind'
*****************************************************

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.

   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 (*note (gnat_ug)The GNAT
Make Program gnatmake::). Otherwise, cross-referencing information will
not be generated.

`gnatxref' Switches
===================

The command lines for `gnatxref' is:
     $ gnatxref [switches] sourcefile1 [sourcefile2 ...]

where

`sourcefile1, sourcefile2'
     identifies the source files for which a report is to be generated.
     The 'with'ed units will be processed too. You must provide at
     least one file.

     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 switches can be :
`-a'
     If this switch is present, `gnatfind' and `gnatxref' will parse
     the read-only files found in the library search path. Otherwise,
     these files will be ignored. This option can be used to protect
     Gnat sources or your own libraries from being parsed, thus making
     `gnatfind' and `gnatxref' much faster, and their output much
     smaller.

`-aIDIR'
     When looking for source files also look in directory DIR. The
     order in which source file search is undertaken is the same as for
     `gnatmake'.

`-aODIR'
     When searching for library and object files, look in directory
     DIR. The order in which library files are searched is the same as
     for `gnatmake'.

`-d'
     If this switch is set `gnatxref' will output the parent type
     reference for each matching derived types.

`-f'
     If this switch is set, the output file names will be preceded by
     their directory (if the file was found in the search path). If
     this switch is not set, the directory will not be printed.

`-g'
     If this switch is set, information is output only for library-level
     entities, ignoring local entities. The use of this switch may
     accelerate `gnatfind' and `gnatxref'.

`-IDIR'
     Equivalent to `-aODIR -aIDIR'.

`-pFILE'
     Specify a project file to use *Note Project Files::.  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'.

`-u'
     Output only unused symbols. This may be really useful if you give
     your main compilation unit on the command line, as `gnatxref' will
     then display every unused entity and 'with'ed package.

`-v'
     Instead of producing the default output, `gnatxref' will generate a
     `tags' file that can be used by vi. For examples how to use this
     feature, see *Note 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'.

`gnatfind' Switches
===================

The command line for `gnatfind' is:

     $ gnatfind [switches] pattern[:sourcefile[:line[:column]]]
           [file1 file2 ...]

where

`pattern'
     An entity will be output only if it matches the regular expression
     found in `pattern', see *Note Regular Expressions in gnatfind and
     gnatxref::.

     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'
     `gnatfind' will look for references, bodies or declarations of
     symbols referenced in `sourcefile', at line `line' and column
     `column'. See *note Examples of gnatfind Usage::.  for syntax
     examples.

`line'
     is a decimal integer identifying the line number containing the
     reference to the entity (or entities) to be located.

`column'
     is a decimal integer identifying the exact location on the line of
     the first character of the identifier for the entity reference.
     Columns are numbered from 1.

`file1 file2 ...'
     The search will be restricted to these files. If none are given,
     then the search will be done for every library file in the search
     path.  These file must appear only after the pattern or sourcefile.

     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'.

     Not 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:
`-a'
     If this switch is present, `gnatfind' and `gnatxref' will parse
     the read-only files found in the library search path. Otherwise,
     these files will be ignored. This option can be used to protect
     Gnat sources or your own libraries from being parsed, thus making
     `gnatfind' and `gnatxref' much faster, and their output much
     smaller.

`-aIDIR'
     When looking for source files also look in directory DIR. The
     order in which source file search is undertaken is the same as for
     `gnatmake'.

`-aODIR'
     When searching for library and object files, look in directory
     DIR. The order in which library files are searched is the same as
     for `gnatmake'.

`-d'
     If this switch is set, then `gnatfind' will output the parent type
     reference for each matching derived types.

`-e'
     By default, `gnatfind' accept the simple regular expression set for
     `pattern'. If this switch is set, then the pattern will be
     considered as full Unix-style regular expression.

`-f'
     If this switch is set, the output file names will be preceded by
     their directory (if the file was found in the search path). If
     this switch is not set, the directory will not be printed.

`-g'
     If this switch is set, information is output only for library-level
     entities, ignoring local entities. The use of this switch may
     accelerate `gnatfind' and `gnatxref'.

`-IDIR'
     Equivalent to `-aODIR -aIDIR'.

`-pFILE'
     Specify a project file (*note Project Files::.) to use.  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'.

`-r'
     By default, `gnatfind' will output only the information about the
     declaration, body or type completion of the entities. If this
     switch is set, the `gnatfind' will locate every reference to the
     entities in the files specified on the command line (or in every
     file in the search path if no file is given on the command line).

`-s'
     If this switch is set, then `gnatfind' will output the content of
     the Ada source file lines were the entity was found.

`-t'
     If this switch is set, then `gnatfind' will output the type
     hierarchy for the specified type. It act like -d option but
     recursively from parent type to parent type. When this switch is
     set it is not possible to specify more than one 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'.

   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
=============

The project files allows a programmer to specify how to compile its
application, where to find sources,... These files are used primarily by
the Glide Ada mode, but they can also be used by the two tools
`gnatxref' and `gnatfind'.

   A project file name must end with `.adp'. 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.

`src_dir=DIR         [default: "./"]'
     specifies a directory where to look for source files. Multiple
     src_dir lines can be specified and they will be searched in the
     order they are specified.

`obj_dir=DIR         [default: "./"]'
     specifies a directory where to look for object and library files.
     Multiple obj_dir lines can be specified and they will be searched
     in the order they are specified

`comp_opt=SWITCHES   [default: ""]'
     creates a variable which can be referred to subsequently by using
     the `${comp_opt}' notation. This is intended to store the default
     switches given to `gnatmake' and `gcc'.

`bind_opt=SWITCHES   [default: ""]'
     creates a variable which can be referred to subsequently by using
     the `${bind_opt}' notation. This is intended to store the default
     switches given to `gnatbind'.

`link_opt=SWITCHES   [default: ""]'
     creates a variable which can be referred to subsequently by using
     the `${link_opt}' notation. This is intended to store the default
     switches given to `gnatlink'.

`main=EXECUTABLE     [default: ""]'
     specifies the name of the executable for the application. This
     variable can be referred to in the following lines by using the
     `${main}' notation.

`comp_cmd=COMMAND    [default: "gcc -c -I${src_dir} -g -gnatq"]'
     specifies the command used to compile a single file in the
     application.

`make_cmd=COMMAND    [default: "gnatmake ${main} -aI${src_dir} -aO${obj_dir} -g -gnatq -cargs ${comp_opt} -bargs ${bind_opt} -largs ${link_opt}"]'
     specifies the command used to recompile the whole application.

`run_cmd=COMMAND     [default: "${main}"]'
     specifies the command used to run the application.

`debug_cmd=COMMAND   [default: "gdb ${main}"]'
     specifies the command used to debug the application

   `gnatxref' and `gnatfind' only take into account the `src_dir' and
`obj_dir' lines, and ignore the others.

Regular Expressions in `gnatfind' and `gnatxref'
================================================

As specified in the section about `gnatfind', the pattern can be a
regular expression. Actually, there are to set of regular expressions
which are recognized by the program :

`globbing patterns'
     These are the most usual regular expression. They are the same
     that you generally used in a Unix shell command line, or in a DOS
     session.

     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

`full regular expression'
     The second set of regular expressions is much more powerful. This
     is the type of regular expressions recognized by utilities such a
     `grep'.

     The following is the form of a regular expression, expressed in Ada
     reference manual style BNF is as follows

          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 ()[].*+?^

     Following are a few examples :

    `abcde|fghi'
          will match any of the two strings 'abcde' and 'fghi'.

    `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

Examples of `gnatxref' Usage
============================

General Usage
-------------

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.adb'
     `gnatxref' 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     7:12

     that is 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
     it referenced in main.adb, line 6 column 12 and line 7 column 12.

`gnatxref package1.adb package2.ads'
     `gnatxref' will generates cross-reference information for
     package1.adb, package2.ads and any other package 'with'ed by any
     of these.

Using gnatxref with vi
----------------------

   `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

will generate the tags file for `gnatfind' itself (if the sources are
in the search path!).

   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.

Examples of `gnatfind' Usage
============================

`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

     that is to say, 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, instead `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.adb 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.

File Name Krunching Using `gnatkr'
**********************************

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.

About `gnatkr'
==============

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:
   * Take the unit name and replace all dots by hyphens.

   * If such a replacement occurs in the second character position of a
     name, and the first character is 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 `-gnatkNN' switch of the compiler activates a "krunching"
circuit that limits file names to nn characters (where nn is a decimal
integer). For example, using OpenVMS, where the maximum file name
length is 39, the value of nn is usually set to 39, but if you want to
generate a set of files that would be usable if ported to a system with
some different maximum file length, then a different value can be
specified.  The default value of 39 for OpenVMS need not be specified.

   The `gnatkr' utility can be used to determine the krunched name for
a given file, when krunched to a specified maximum length.

Using `gnatkr'
==============

The `gnatkr' command has the form

     $ gnatkr NAME [LENGTH]

NAME can be an Ada name with dots or the GNAT name of the unit, where
the dots representing child units or subunit are replaced by hyphens.
The only confusion arises if a name ends in `.ads' or `.adb'. `gnatkr'
takes this to be an extension if there are no other dots in the name
and the whole name is in lowercase.

   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 which are
always 8.

The output is the krunched name. The output has an extension only if the
original argument was a file name with an extension.

Krunching Method
================

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 specification 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:

   * The name is divided into segments separated by hyphens, tildes or
     underscores and all hyphens, tildes, and underscores are
     eliminated. If this leaves the name short enough, we are done.

   * If the name is too long, the longest segment is located (left-most
     if there are two of equal length), and shortened by dropping its
     last character. This is repeated until the name is short enough.

     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

   * The file names for all predefined units are always krunched to
     eight characters. The krunching of these predefined units uses the
     following special prefix replacements:

    `ada-'
          replaced by `a-'

    `gnat-'
          replaced by `g-'

    `interfaces-'
          replaced by `i-'

    `system-'
          replaced by `s-'

     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.

Examples of `gnatkr' Usage
==========================

     $ gnatkr very_long_unit_name.ads      --> velounna.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

Preprocessing Using `gnatprep'
******************************

The `gnatprep' utility provides a simple preprocessing capability for
Ada programs.  It is designed for use with GNAT, but is not dependent
on any special features of GNAT.

Using `gnatprep'
================

To call `gnatprep' use

     $ gnatprep [-bcrsu] [-Dsymbol=value] infile outfile [deffile]

where
`infile'
     is the full name of the input file, which is an Ada source file
     containing preprocessor directives.

`outfile'
     is the full name of the output file, which is an Ada source in
     standard Ada form. When used with GNAT, this file name will
     normally have an ads or adb suffix.

`deffile'
     is the full name of a text file containing definitions of symbols
     to be referenced by the preprocessor. This argument is optional,
     and can be replaced by the use of the `-D' switch.

`switches'
     is an optional sequence of switches as described in the next
     section.

Switches for `gnatprep'
=======================

`-b'
     Causes both preprocessor lines and the lines deleted by
     preprocessing to be replaced by blank lines in the output source
     file, preserving line numbers in the output file.

`-c'
     Causes both preprocessor lines and the lines deleted by
     preprocessing to be retained in the output source as comments
     marked with the special string "-! ". This option will result in
     line numbers being preserved in the output file.

`-Dsymbol=value'
     Defines a new symbol, associated with value. If no value is given
     on the command line, then symbol is considered to be `True'. This
     switch can be used in place of a definition file.

`-r'
     Causes a `Source_Reference' pragma to be generated that references
     the original input file, so that error messages will use the file
     name of this original file. The use of this switch implies that
     preprocessor lines are not to be removed from the file, so its use
     will force `-b' mode if `-c' has not been specified explicitly.

     Note 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'
     Causes a sorted list of symbol names and values to be listed on
     the standard output file.

`-u'
     Causes undefined symbols to be treated as having the value FALSE
     in the context of a preprocessor test. In the absence of this
     option, an undefined symbol in a `#if' or `#elsif' test will be
     treated as an error.

Note: 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.

Form of Definitions File
========================

The definitions file contains lines of the form

     symbol := value

where symbol is an identifier, following normal Ada (case-insensitive)
rules for its syntax, and value is one of the following:

   * Empty, corresponding to a null substitution

   * A string literal using normal Ada syntax

   * Any sequence of characters from the set (letters, digits, period,
     underline).

Comment lines may also appear in the definitions file, starting with
the usual `--', and comments may be added to the definitions lines.

Form of Input Text for `gnatprep'
=================================

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> 'Defined
     expression ::=  not expression
     expression ::=  expression and expression
     expression ::=  expression or expression
     expression ::=  expression and then expression
     expression ::=  expression or else expression
     expression ::=  ( expression )

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.

   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, so that the lines are included only if the symbol is not defined.
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. 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.

The GNAT Library Browser `gnatls'
*********************************

`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.

Running `gnatls'
================

The `gnatls' command has the form

     $ gnatls switches OBJECT_OR_ALI_FILE

The main argument is the list of object or `ali' files (*note 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:

`OK (unchanged)'
     The version of the source file used for the compilation of the
     specified unit corresponds exactly to the actual source file.

`MOK (slightly modified)'
     The version of the source file used for the compilation of the
     specified unit differs from the actual source file but not enough
     to require recompilation. If you use gnatmake with the qualifier
     `-m (minimal recompilation)', a file marked MOK will not be
     recompiled.

`DIF (modified)'
     No version of the source found on the path corresponds to the
     source used to build this object.

`??? (file not found)'
     No source file was found for this unit.

`HID (hidden,  unchanged version not first on PATH)'
     The version of the source that corresponds exactly to the source
     used for compilation has been found on the path but it is hidden
     by another version of the same source that has been modified.

Switches for `gnatls'
=====================

`gnatls' recognizes the following switches:

`-a'
     Consider all units, including those of the predefined Ada library.
     Especially useful with `-d'.

`-d'
     List sources from which specified units depend on.

`-h'
     Output the list of options.

`-o'
     Only output information about object files.

`-s'
     Only output information about source files.

`-u'
     Only output information about compilation units.

`-aODIR'
`-aIDIR'
`-IDIR'
`-I-'
`-nostdinc'
     Source and Object path manipulation. Same meaning as the equivalent
     `gnatmake' flags (see *Note Switches for gnatmake::).

`-v'
     Verbose mode. Output the complete source and object paths. Do not
     use the default column layout but instead use long format giving
     as much as information possible on each requested units, including
     special characteristics such as:

    `Preelaborable'
          The unit is preelaborable in the Ada 95 sense.

    `No_Elab_Code'
          No elaboration code has been produced by the compiler for
          this unit.

    `Pure'
          The unit is pure in the Ada 95 sense.

    `Elaborate_Body'
          The unit contains a pragma Elaborate_Body.

    `Remote_Types'
          The unit contains a pragma Remote_Types.

    `Shared_Passive'
          The unit contains a pragma Shared_Passive.

    `Predefined'
          This unit is part of the predefined environment and cannot be
          modified by the user.

    `Remote_Call_Interface'
          The unit contains a pragma Remote_Call_Interface.

Example of `gnatls' Usage
=========================

Example of using the verbose switch. Note how the source and object
paths are affected by the -I switch.

     $ gnatls -v -I.. demo1.o
     
     GNATLS 3.10w (970212) Copyright 1999 Free Software Foundation, Inc.
     
     Source Search Path:
        <Current_Directory>
        ../
        /home/comar/local/adainclude/
     
     Object Search Path:
        <Current_Directory>
        ../
        /home/comar/local/lib/gcc-lib/mips-sni-sysv4/2.7.2/adalib/
     
     ./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

GNAT and Libraries
******************

This chapter addresses some of the issues related to building and using
a library with GNAT. It also shows how the GNAT runtime library can be
recompiled.

Creating an Ada Library
=======================

In the GNAT environment, a library has two components:
   * Source files.

   * Compiled code and Ali files. See *Note The Ada Library Information
     Files::.

In order to use other packages *Note The GNAT Compilation Model::
requires a certain number of sources to be available to the compiler.
The minimal set of sources required includes the specs of all the
packages that make up the visible part of the library as well as all
the sources upon which they depend. The bodies of all visible generic
units must also be provided.

Although it is not strictly mandatory, it is recommended that all
sources needed to recompile the library be provided, so that the user
can make full use of interunit inlining and source-level debugging.
This can also make the situation easier for users that need to upgrade
their compilation toolchain and thus need to recompile the library from
sources.

The compiled code can be provided in different ways. The simplest way is
to provide directly the set of objects produced by the compiler during
the compilation of the library. It is also possible to group the objects
into an archive using whatever commands are provided by the operating
system. Finally, it is also possible to create a shared library (see
option -shared in the GCC manual).

There are various possibilities for compiling the units that make up the
library: for example with a Makefile *Note 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, in order to build all the
necessary objects. Here is an example of such a dummy program and the
generic commands used to build an archive or a shared library.

     with My_Lib.Service1;
     with My_Lib.Service2;
     with My_Lib.Service3;
     procedure My_Lib_Dummy is
     begin
        null;
     end;
     
     # 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

When the objects are grouped in an archive or a shared library, the user
needs to specify the desired library at link time, unless a pragma
linker_options has been used in one of the sources:
     pragma Linker_Options ("-lmy_lib");

Installing an Ada Library
=========================

In the GNAT model, installing a library consists in copying into a
specific location the files that make up this library. It is possible
to install the sources in a different directory from the other files
(ALI, objects, archives) since the source path and the object path can
easily be specified separately.

For general purpose libraries, it is possible for the system
administrator to put those libraries in the default compiler paths. To
achieve this, he must specify their location in the configuration files
"ada_source_path" and "ada_object_path" that 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 simple format: each line
in them 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.

"ada_source_path" and "ada_object_path" might actually 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" at all, 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").

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.

Using an Ada Library
====================

In order to use a Ada library, you need to make sure that this library
is on both your source and object path *Note Search Paths and the
Run-Time Library (RTL):: and *Note Search Paths for gnatbind::. For
instance, 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 simplified down to the following:
     $ 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"

   * a pragma linker_options, as mentioned in *Note Creating an Ada
     Library:: as been added to the sources.

Rebuilding the GNAT Run-Time Library
====================================

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_Scalar. 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.

Using the GNU `make' Utility
****************************

This chapter offers some examples of makefiles that solve specific
problems. It does not explain how to write a makefile (see the GNU make
documentation), nor does it try to replace the `gnatmake' utility
(*note The GNAT Make Program 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'.

Using gnatmake in a Makefile
============================

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 (*note 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}

Automatically Creating a List of Directories
============================================

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, altough 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,...

   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}

Generating the Command Line Switches
====================================

Once you have created the list of directories as explained in the
previous section (*note Automatically Creating a List of
Directories::.), you can easily generate the command line arguments to
pass to gnatmake.

   For the sake of completness, 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

Overcoming Command Line Length Limits
=====================================

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 *Note Automatically
Creating a List of Directories::.  For the sake of completness, 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 immediatly by `make'. This way we overcome the standard make
behavior which is to expand the variables only when they are actually
used.

     # In this example, we create both ADA_INCLUDE_PATH and ADA_OBJECT_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_OBJECT_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_OBJECT_PATH += ${OBJECT_LIST}
     export ADA_INCLUDE_PATH
     export ADA_OBJECT_PATH
     
     all:
             gnatmake main_unit

Finding Memory Problems with `gnatmem'
**************************************

`gnatmem', is a tool that monitors dynamic allocation and deallocation
activity in a program, and displays information about incorrect
deallocations and possible sources of memory leaks. Gnatmem provides
three type of information:
   * General information concerning memory management, such as the total
     number of allocations and deallocations, the amount of allocated
     memory and the high water mark, i.e. the largest amount of
     allocated memory in the course of program execution.

   * Backtraces for all incorrect deallocations, that is to say
     deallocations which do not correspond to a valid allocation.

   * Information on each allocation that is potentially the origin of a
     memory leak.

   The `gnatmem' command has two modes. It can be used with `gdb' or
with instrumented allocation and deallocation routines. The later mode
is called the `GMEM' mode. Both modes produce the very same output.

Running `gnatmem' (GDB Mode)
============================

The `gnatmem' command has the form

        $ gnatmem [-q] [n] [-o file] user_program [program_arg]*
     or
        $ gnatmem [-q] [n] -i file

Gnatmem must be supplied with the executable to examine, followed by its
run-time inputs. For example, if a program is executed with the command:
     $ my_program arg1 arg2
   then it can be run under `gnatmem' control using the command:
     $ gnatmem my_program arg1 arg2

   The program is transparently executed under the control of the
debugger *Note The GNAT Debugger GDB::. This does not affect the
behavior of the program, except for sensitive real-time programs. When
the program has completed execution, `gnatmem' outputs a report
containing general allocation/deallocation information and potential
memory leak.  For better results, the user program should be compiled
with debugging options *Note Switches for gcc::.

   Here is a simple example of use:

   *************** debut cc
     $ gnatmem test_gm
     
     Global information
     ------------------
        Total number of allocations        :  45
        Total number of deallocations      :   6
        Final Water Mark (non freed mem)   :  11.29 Kilobytes
        High Water Mark                    :  11.40 Kilobytes
     
     .
     .
     .
     Allocation Root # 2
     -------------------
      Number of non freed allocations    :  11
      Final Water Mark (non freed mem)   :   1.16 Kilobytes
      High Water Mark                    :   1.27 Kilobytes
      Backtrace                          :
        test_gm.adb:23 test_gm.alloc
     .
     .
     .

   The first block of output give general information. In this case, the
Ada construct "new" was executed 45 times, and only 6 calls to an
unchecked deallocation routine occurred.

   Subsequent paragraphs display  information on all allocation roots.
An allocation root is a specific point in the execution of the program
that generates some dynamic allocation, such as a "new" construct. This
root is represented by an execution backtrace (or subprogram call
stack). By default the backtrace depth for allocations roots is 1, so
that a root corresponds exactly to a source location. The backtrace can
be made deeper, to make the root more specific.

Running `gnatmem' (GMEM Mode)
=============================

The `gnatmem' command has the form

        $ gnatmem [-q] [n] -i gmem.out user_program [program_arg]*

   The program must have been linked with the instrumented version of
the allocation and deallocation routines. This is done with linking
with the `libgmem.a' library. For better results, the user program
should be compiled with debugging options *Note Switches for gcc::. For
example to build `my_program':

     $ gnatmake -g my_program -largs -lgmem

Under Solaris it is required to build the application with
`-funwind-tables' options *Note Switches for gcc::.

When running `my_program' the file `gmem.out' is produced. This file
contains information about all allocations and deallocations done by the
program. It is produced by the instrumented allocations and
deallocations routines and will be used by `gnatmem'.

Gnatmem must be supplied with the `gmem.out' file and the executable to
examine followed by its run-time inputs. For example, if a program is
executed with the command:
     $ my_program arg1 arg2
   then `gmem.out' can be analysed by `gnatmem' using the command:
     $ gnatmem -i gmem.out my_program arg1 arg2

Switches for `gnatmem'
======================

`gnatmem' recognizes the following switches:

``-q''
     Quiet. Gives the minimum output needed to identify the origin of
     the memory leaks. Omit statistical information.

``n''
     N is an integer literal (usually between 1 and 10) which controls
     the depth of the backtraces defining allocation root. The default
     value for N is 1. The deeper the backtrace, the more precise the
     localization of the root. Note that the total number of roots can
     depend on this parameter.

``-o file''
     Direct the gdb output to the specified file. The `gdb' script used
     to generate this output is also saved in the file `gnatmem.tmp'.

``-i file''
     Do the `gnatmem' processing starting from `file' which has been
     generated by a previous call to `gnatmem' with the -o switch or
     `gmem.out' produced by `GMEM' mode. This is useful for post mortem
     processing.

Example of `gnatmem' Usage
==========================

This section is based on the `GDB' mode of `gnatmem'. The same results
can be achieved using `GMEM' mode. See section *Note Running gnatmem
(GMEM Mode)::.

The first example shows the use of `gnatmem' on a simple leaking
program.  Suppose that we have the following Ada program:

     with Unchecked_Deallocation;
     procedure Test_Gm is
     
        type T is array (1..1000) of Integer;
        type Ptr is access T;
        procedure Free is new Unchecked_Deallocation (T, Ptr);
        A : Ptr;
     
        procedure My_Alloc is
        begin
           A := new T;
        end My_Alloc;
     
        procedure My_DeAlloc is
           B : Ptr := A;
        begin
           Free (B);
        end My_DeAlloc;
     
     begin
        My_Alloc;
        for I in 1 .. 5 loop
           for J in I .. 5 loop
              My_Alloc;
           end loop;
           My_Dealloc;
        end loop;
     end;

The program needs to be compiled with debugging option:

     $ gnatmake -g test_gm

   `gnatmem' is invoked simply with
     $ gnatmem test_gm

which produces the following output:

     Global information
     ------------------
        Total number of allocations        :  18
        Total number of deallocations      :   5
        Final Water Mark (non freed mem)   :  53.00 Kilobytes
        High Water Mark                    :  56.90 Kilobytes
     
     Allocation Root # 1
     -------------------
      Number of non freed allocations    :  11
      Final Water Mark (non freed mem)   :  42.97 Kilobytes
      High Water Mark                    :  46.88 Kilobytes
      Backtrace                          :
        test_gm.adb:11 test_gm.my_alloc
     
     Allocation Root # 2
     -------------------
      Number of non freed allocations    :   1
      Final Water Mark (non freed mem)   :  10.02 Kilobytes
      High Water Mark                    :  10.02 Kilobytes
      Backtrace                          :
        s-secsta.adb:81 system.secondary_stack.ss_init
     
     Allocation Root # 3
     -------------------
      Number of non freed allocations    :   1
      Final Water Mark (non freed mem)   :  12 Bytes
      High Water Mark                    :  12 Bytes
      Backtrace                          :
        s-secsta.adb:181 system.secondary_stack.ss_init

Note that the GNAT run time contains itself a certain number of
allocations that have no  corresponding deallocation, as shown here for
root #2 and root #1. This is a normal behavior when the number of non
freed allocations is one, it locates dynamic data structures that the
run time needs for the complete lifetime of the program. Note also that
there is only one allocation root in the user program with a single
line back trace: test_gm.adb:11 test_gm.my_alloc, whereas a careful
analysis of the program shows that 'My_Alloc' is called at 2 different
points in the source (line 21 and line 24). If those two allocation
roots need to be distinguished, the backtrace depth parameter can be
used:

     $ gnatmem 3 test_gm

which will give the following output:

     Global information
     ------------------
        Total number of allocations        :  18
        Total number of deallocations      :   5
        Final Water Mark (non freed mem)   :  53.00 Kilobytes
        High Water Mark                    :  56.90 Kilobytes
     
     Allocation Root # 1
     -------------------
      Number of non freed allocations    :  10
      Final Water Mark (non freed mem)   :  39.06 Kilobytes
      High Water Mark                    :  42.97 Kilobytes
      Backtrace                          :
        test_gm.adb:11 test_gm.my_alloc
        test_gm.adb:24 test_gm
        b_test_gm.c:52 main
     
     Allocation Root # 2
     -------------------
      Number of non freed allocations    :   1
      Final Water Mark (non freed mem)   :  10.02 Kilobytes
      High Water Mark                    :  10.02 Kilobytes
      Backtrace                          :
        s-secsta.adb:81  system.secondary_stack.ss_init
        s-secsta.adb:283 <system__secondary_stack___elabb>
        b_test_gm.c:33   adainit
     
     Allocation Root # 3
     -------------------
      Number of non freed allocations    :   1
      Final Water Mark (non freed mem)   :   3.91 Kilobytes
      High Water Mark                    :   3.91 Kilobytes
      Backtrace                          :
        test_gm.adb:11 test_gm.my_alloc
        test_gm.adb:21 test_gm
        b_test_gm.c:52 main
     
     Allocation Root # 4
     -------------------
      Number of non freed allocations    :   1
      Final Water Mark (non freed mem)   :  12 Bytes
      High Water Mark                    :  12 Bytes
      Backtrace                          :
        s-secsta.adb:181 system.secondary_stack.ss_init
        s-secsta.adb:283 <system__secondary_stack___elabb>
        b_test_gm.c:33   adainit

The allocation root #1 of the first example has been split in 2 roots #1
and #3 thanks to the more precise associated backtrace.

GDB and GMEM Modes
==================

The main advantage of the `GMEM' mode is that it is a lot faster than
the `GDB' mode where the application must be monitored by a `GDB'
script.  But the `GMEM' mode is available only for DEC Unix, SGI Irix,
Linux x86, Solaris (sparc and x86) and Windows 95/98/NT/2000 (x86).

The main advantage of the `GDB' mode is that it is available on all
supported platforms. But it can be very slow if the application does a
lot of allocations and deallocations.

Implementation Note
===================

`gnatmem' Using `GDB' Mode
--------------------------

`gnatmem' executes the user program under the control of `GDB' using a
script that sets breakpoints and gathers information on each dynamic
allocation and deallocation. The output of the script is then analyzed
by `gnatmem' in order to locate memory leaks and their origin in the
program. Gnatmem works by recording each address returned by the
allocation procedure (`__gnat_malloc') along with the backtrace at the
allocation point. On each deallocation, the deallocated address is
matched with the corresponding allocation. At the end of the processing,
the unmatched allocations are considered potential leaks. All the
allocations associated with the same backtrace are grouped together and
form an allocation root. The allocation roots are then sorted so that
those with the biggest number of unmatched allocation are printed
first. A delicate aspect of this technique is to distinguish between the
data produced by the user program and the data produced by the gdb
script. Currently, on systems that allow probing the terminal, the gdb
command "tty" is used to force the program output to be redirected to
the current terminal while the `gdb' output is directed to a file or to
a pipe in order to be processed subsequently by `gnatmem'.

`gnatmem' Using `GMEM' Mode
---------------------------

This mode use the same algorithm to detect memory leak as the `GDB'
mode of `gnatmem', the only difference is in the way data are gathered.
In `GMEM' mode the program is linked with instrumented version of
`__gnat_malloc' and `__gnat_free' routines. Information needed to find
memory leak are recorded by these routines in file `gmem.out'. This
mode also require that the stack traceback be available, this is only
implemented on some platforms *Note GDB and GMEM Modes::.

Finding Memory Problems with GNAT Debug Pool
********************************************

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 specifc 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 of 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:
   * `GNAT.Debug_Pools.Accessing_Not_Allocated_Storage'

   * `GNAT.Debug_Pools.Accessing_Deallocated_Storage'

   * `GNAT.Debug_Pools.Freeing_Not_Allocated_Storage'

   * `GNAT.Debug_Pools.Freeing_Deallocated_Storage '

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. The deallocation
strategy consists in not releasing the memory to the underlying system
but rather to fill it with a memory pattern easily recognizable during
debugging sessions: The memory pattern is the old IBM hexadecimal
convention: 16#DEADBEEF#.  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

Creating Sample Bodies Using `gnatstub'
***************************************

`gnatstub' creates body stubs, that is, empty but compilable bodies for
library unit declarations.

   To create a body stub, `gnatstub' has to compile the library unit
declaration. Therefore, bodies can be created only for legal library
units. Moreover, if a library unit depends semantically upon units
located outside the current directory, you have to provide the source
search path when calling `gnatstub', see the description of `gnatstub'
switches below.

Running `gnatstub'
==================

`gnatstub' has the command-line interface of the form

     $ gnatstub [switches] filename [directory]

where
`filename'
     is the name of the source file that contains a library unit
     declaration for which a body must be created. This name should
     follow the GNAT file name conventions. No crunching is allowed for
     this file name. The file name may contain the path information.

`directory'
     indicates the directory to place a body stub (default is the
     current directory)

`switches'
     is an optional sequence of switches as described in the next
     section

Switches for `gnatstub'
=======================

`-f'
     If the destination directory already contains a file with a name
     of the body file for the argument spec file, replace it with the
     generated body stub.

`-hs'
     Put the comment header (i.e. all the comments preceding the
     compilation unit) from the source of the library unit declaration
     into the body stub.

`-hg'
     Put a sample comment header into the body stub.

`-IDIR'
`-I-'
     These switches have the same meaning as in calls to gcc.  They
     define the source search path in the call to gcc issued by
     `gnatstub' to compile an argument source file.

`-iN'
     (N is a decimal natural number). Set the indentation level in the
     generated body sample to n, '-i0' means "no indentation", the
     default indentation is 3.

`-k'
     Do not remove the tree file (i.e. the snapshot of the compiler
     internal structures used by `gnatstub') after creating the body
     stub.

`-lN'
     (N is a decimal positive number) Set the maximum line length in the
     body stub to n, the default is 78.

`-q'
     Quiet mode: do not generate a confirmation when a body is
     successfully created or a message when a body is not required for
     an argument unit.

`-r'
     Reuse the tree file (if it exists) instead of creating it: instead
     of creating the tree file for the library unit declaration,
     gnatstub tries to find it in the current directory and use it for
     creating a body. If the tree file is not found, no body is
     created. `-r' also implies `-k', whether or not `-k' is set
     explicitly.

`-t'
     Overwrite the existing tree file: if the current directory already
     contains the file which, according to the GNAT file name rules
     should be considered as a tree file for the argument source file,
     gnatstub will refuse to create the tree file needed to create a
     body sampler, unless `-t' option is set

`-v'
     Verbose mode: generate version information.

Reducing the Size of Ada Executables with `gnatelim'
****************************************************

About `gnatelim'
================

When a program shares a set of Ada packages with other programs, it may
happen that this program uses only a fraction of the subprograms
defined in these packages. The code created for these unused
subprograms increases the size of the executable.

   `gnatelim' tracks unused subprograms in an Ada program and outputs a
list of GNAT-specific `Eliminate' pragmas (see next section) marking
all the subprograms that are declared but never called.  By placing the
list of `Eliminate' pragmas in the GNAT configuration file `gnat.adc'
and recompiling your program, you may decrease the size of its
executable, because the compiler will not generate the code for
'eliminated' subprograms.

   `gnatelim' needs as its input data a set of tree files (see *Note
Tree Files::) representing all the components of a program to process
and a bind file for a main subprogram (see *Note Preparing Tree and
Bind Files for gnatelim::).

`Eliminate' Pragma
==================

The simplified syntax of the Eliminate pragma used by `gnatelim' is:

     pragma Eliminate (Library_Unit_Name, Subprogram_Name);

where
`Library_Unit_Name'
     full expanded Ada name of a library unit

`Subprogram_Name'
     a simple or expanded name of a subprogram declared within this
     compilation unit

The effect of an `Eliminate' pragma placed in the GNAT configuration
file `gnat.adc' is:

   * If the subprogram `Subprogram_Name' is declared within the library
     unit `Library_Unit_Name', the compiler will not generate code for
     this subprogram. This applies to all overloaded subprograms denoted
     by `Subprogram_Name'.

   * If a subprogram marked by the pragma `Eliminate' is used (called)
     in a program, the compiler will produce an error message in the
     place where it is called.

Tree Files
==========

A tree file stores a snapshot of the compiler internal data structures
at the very end of a successful compilation. It contains all the
syntactic and semantic information for the copmiled unit and all the
units upon which it depends semantically.  To use tools that make use
of tree files, you need to first produce the right set of tree files.

   GNAT produces correct tree files when -gnatt -gnatc options are set
in a gcc call. The tree files have an .adt extension.  Therefore, to
produce a tree file for the compilation unit contained in a file named
`foo.adb', you must use the command

     $ gcc -c -gnatc -gnatt foo.adb

and you will get the tree file `foo.adt'.  compilation.

Preparing Tree and Bind Files for `gnatelim'
============================================

A set of tree files covering the program to be analyzed with `gnatelim'
and the bind file for the main subprogram does not have to be in the
current directory.  '-T' gnatelim option may be used to provide the
search path for tree files, and '-b' option may be used to point to the
bind file to process (see *Note Running gnatelim::)

   If you do not have the appropriate set of tree files and the right
bind file, you may create them in the current directory using the
following procedure.

   Let `Main_Prog' be the name of a main subprogram, and suppose this
subprogram is in a file named `main_prog.adb'.

   To create a bind file for `gnatelim', run `gnatbind' for the main
subprogram. `gnatelim' can work with both Ada and C bind files; when
both are present, it uses the Ada bind file.  The following commands
will build the program and create the bind file:

     $ gnatmake -c Main_Prog
     $ gnatbind main_prog

To create a minimal set of tree files covering the whole program, call
`gnatmake' for this program as follows:

     $ gnatmake -f -c -gnatc -gnatt Main_Prog

The `-c' gnatmake option turns off the bind and link steps, that are
useless anyway because the sources are compiled with `-gnatc' option
which turns off code generation.

   The `-f' gnatmake option forces recompilation of all the needed
sources.

   This sequence of actions will create all the data needed by
`gnatelim' from scratch and therefore guarantee its consistency. If you
would like to use some existing set of files as `gnatelim' output, you
must make sure that the set of files is complete and consistent. You
can use the `-m' switch to check if there are missed tree files

   Note, that `gnatelim' needs neither object nor ALI files.

Running `gnatelim'
==================

`gnatelim' has the following command-line interface:

     $ gnatelim [options] name

`name' should be a full expanded Ada name of a main subprogram of a
program (partition).

   `gnatelim' options:

`-q'
     Quiet mode: by default `gnatelim' generates to the standard error
     stream a trace of the source file names of the compilation units
     being processed. This option turns this trace off.

`-v'
     Verbose mode: `gnatelim' version information is printed as Ada
     comments to the standard output stream.

`-a'
     Also look for subprograms from the GNAT run time that can be
     eliminated.

`-m'
     Check if any tree files are missing for an accurate result.

`\-T\DIR'
     When looking for tree files also look in directory DIR

`\-b\BIND_FILE'
     Specifices BIND_FILE as the bind file to process. If not set, the
     name of the bind file is computed from the full expanded Ada name
     of a main subprogram.

`-dX'
     Activate internal debugging switches. X is a letter or digit, or
     string of letters or digits, which specifies the type of debugging
     mode desired.  Normally these are used only for internal
     development or system debugging purposes. You can find full
     documentation for these switches in the body of the
     `Gnatelim.Options' unit in the compiler source file
     `gnatelim-options.adb'.

`gnatelim' sends its output to the standard output stream, and all the
tracing and debug information is sent to the standard error stream.  In
order to produce a proper GNAT configuration file `gnat.adc',
redirection must be used:

     $ gnatelim Main_Prog > gnat.adc

or

     $ gnatelim Main_Prog >> gnat.adc

In order to append the `gnatelim' output to the existing contents of
`gnat.adc'.

Correcting the List of Eliminate Pragmas
========================================

In some rare cases it may happen that `gnatelim' will try to eliminate
subprograms which are actually called in the program. In this case, the
compiler will generate an error message of the form:

     file.adb:106:07: cannot call eliminated subprogram "My_Prog"

You will need to manually remove the wrong `Eliminate' pragmas from the
`gnat.adc' file. It is advised that you recompile your program from
scratch after that because you need a consistent `gnat.adc' file during
the entire compilation.

Making Your Executables Smaller
===============================

In order to get a smaller executable for your program you now have to
recompile the program completely with the new `gnat.adc' file created
by `gnatelim' in your current directory:

     $ gnatmake -f Main_Prog

(you will need `-f' option for gnatmake to recompile everything with
the set of pragmas `Eliminate' you have obtained with `gnatelim').

   Be aware that the set of `Eliminate' pragmas is specific to each
program. It is not recommended to merge sets of `Eliminate' pragmas
created for different programs in one `gnat.adc' file.

Summary of the gnatelim Usage Cycle
===================================

Here is a quick summary of the steps to be taken in order to reduce the
size of your executables with `gnatelim'. You may use other GNAT
options to control the optimization level, to produce the debugging
information, to set search path, etc.

  1. Produce a bind file and a set of tree files

          $ gnatmake -c Main_Prog
          $ gnatbind main_prog
          $ gnatmake -f -c -gnatc -gnatt Main_Prog

  2. Generate a list of `Eliminate' pragmas
          $ gnatelim Main_Prog >[>] gnat.adc

  3. Recompile the application

          $ gnatmake -f Main_Prog


Other Utility Programs
**********************

This chapter discusses some other utility programs available in the Ada
environment.

Using Other Utility Programs with GNAT
======================================

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.

The `gnatpsys' Utility Program
==============================

Many of the definitions in package System are implementation-dependent.
Furthermore, although the source of the package System is available for
inspection, it uses special attributes for parameterizing many of the
critical values, so the source is not informative for the casual user.

   The `gnatpsys' utility is designed to deal with this situation.  It
is an Ada program that dynamically determines the values of all the
relevant parameters in System, and prints them out in the form of an
Ada source listing for System, displaying all the values of interest.
This output is generated to `stdout'.

   To determine the value of any parameter in package System, simply
run `gnatpsys' with no qualifiers or arguments, and examine the output.
This is preferable to consulting documentation, because you know that
the values you are getting are the actual ones provided by the
executing system.

The `gnatpsta' Utility Program
==============================

Many of the definitions in package Standard are
implementation-dependent.  However, the source of this package does not
exist as an Ada source file, so these values cannot be determined by
inspecting the source.  They can be determined by examining in detail
the coding of `cstand.adb' which creates the image of Standard in the
compiler, but this is awkward and requires a great deal of internal
knowledge about the system.

   The `gnatpsta' utility is designed to deal with this situation.  It
is an Ada program that dynamically determines the values of all the
relevant parameters in Standard, and prints them out in the form of an
Ada source listing for Standard, displaying all the values of interest.
This output is generated to `stdout'.

   To determine the value of any parameter in package Standard, simply
run `gnatpsta' with no qualifiers or arguments, and examine the output.
This is preferable to consulting documentation, because you know that
the values you are getting are the actual ones provided by the
executing system.

The External Symbol Naming Scheme of GNAT
=========================================

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.

Ada Mode for `Glide'
====================

The Glide mode for programming in Ada (both, Ada83 and Ada95) helps the
user in understanding existing code and facilitates writing new code. It
furthermore provides some utility functions for easier integration of
standard Emacs features when programming in Ada.

General Features:
-----------------

   * Full Integrated Development Environment :

        * support of 'project files' for the configuration (directories,
          compilation options,...)

        * compiling and stepping through error messages.

        * running and debugging your applications within Glide.

   * easy to use for beginners by pull-down menus,

   * user configurable by many user-option variables.

Ada Mode Features That Help Understanding Code:
-----------------------------------------------

   * functions for easy and quick stepping through Ada code,

   * getting cross reference information for identifiers (e.g. find the
     defining place by a keystroke),

   * displaying an index menu of types and subprograms and move point to
     the chosen one,

   * automatic color highlighting of the various entities in Ada code.

Glide Support for Writing Ada Code:
-----------------------------------

   * switching between spec and body files with possible autogeneration
     of body files,

   * automatic formating of subprograms parameter lists.

   * automatic smart indentation according to Ada syntax,

   * automatic completion of identifiers,

   * automatic casing of identifiers, keywords, and attributes,

   * insertion of statement templates,

   * filling comment paragraphs like filling normal text,

   For more information, please see *Note Ada Mode for Glide::.

Converting Ada Files to html with `gnathtml'
============================================

This `Perl' script allows Ada source files to be browsed using standard
Web browsers. For installation procedure, see the section *Note
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 follow:
     $ perl gnathtml.pl [switches] ada-files

   You can pass it 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 available switches are the following ones :

`-83'
     Only the subset on the Ada 83 keywords will be highlighted, not
     the full Ada 95 keywords set.

`-cc COLOR'
     This option allows you to change the color used for comments. The
     default value is green. The color argument can be any name
     accepted by html.

`-d'
     If the ada files depend on some other files (using for instance the
     `with' command, the latter will also be converted to html.  Only
     the files in the user project will be converted to html, not the
     files in the run-time library itself.

`-D'
     This command is the same as -d above, but `gnathtml' will also look
     for files in the run-time library, and generate html files for
     them.

`-f'
     By default, gnathtml will generate html links only for global
     entities ('with'ed units, global variables and types,...). If you
     specify the `-f' on the command line, then links will be generated
     for local entities too.

`-l NUMBER'
     If this switch is provided and NUMBER is not 0, then `gnathtml'
     will number the html files every NUMBER line.

`-I DIR'
     Specify a directory to search for library files (`.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'
     Specify the output directory for html files. By default, gnathtml
     will saved the generated html files in a subdirectory named
     `html/'.

`-p FILE'
     If you are using Emacs and the most recent Emacs Ada mode, which
     provides a full Integrated Development Environment for compiling,
     checking, running and debugging applications, you may be using
     `.adp' 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'
     This option allows you to change the color used for symbol
     definitions.  The default value is red. The color argument can be
     any name accepted by html.

`-t FILE'
     This switch provides the name of a file. This file contains a list
     of file names to be converted, and the effect is exactly as though
     they had appeared explicitly on the command line. This is the
     recommended way to work around the command line length limit on
     some systems.

Installing `gnathtml'
=====================

`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  tell  the Operating  system  where
Perl is. 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

Running and Debugging Ada Programs
**********************************

This chapter discusses how to debug Ada programs. An incorrect Ada
program may be handled in three ways by the GNAT compiler:

  1. The illegality may be a violation of the static semantics of Ada.
     In that case GNAT diagnoses the constructs in the program that are
     illegal.  It is then a straightforward matter for the user to
     modify those parts of the program.

  2. The illegality may be a violation of the dynamic semantics of Ada.
     In that case the program compiles and executes, but may generate
     incorrect results, or may terminate abnormally with some exception.

  3. When presented with a program that contains convoluted errors, GNAT
     itself may terminate abnormally without providing full diagnostics
     on the incorrect user program.

The GNAT Debugger GDB
=====================

`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.  The manual `Debugging with GDB' contains 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.

Running GDB
===========

The debugger can be launched directly and simply from emacs which allows
to browse and modify directly the source code during the debugging
session, *Note Ada Mode for Glide::. Here is described the basic use of
`GDB' is text mode.

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'.

Introduction to GDB Commands
============================

`GDB' contains a large repertoire of commands. The manual `Debugging
with GDB' includes 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.

`set args ARGUMENTS'
     The ARGUMENTS list above is a list of arguments to be passed to
     the program on a subsequent run command, just as though the
     arguments had been entered on a normal invocation of the program.
     The `set args' command is not needed if the program does not
     require arguments.

`run'
     The `run' command causes execution of the program to start from the
     beginning. If the program is already running, that is to say if you
     are currently positioned  at a breakpoint, then a prompt will ask
     for confirmation that you want to abandon the current execution
     and restart.

`breakpoint LOCATION'
     The breakpoint command sets a breakpoint, that is to say a point
     at which execution will halt and `GDB' will await further
     commands. LOCATION is either a line number within a file, given in
     the format `file:linenumber', or it is the name of a subprogram.
     If you request that a breakpoint be set on a subprogram that is
     overloaded, a prompt will ask you to specify on which of those
     subprograms you want to breakpoint. You can also specify that all
     of them should be breakpointed. If the program is run and
     execution encounters the breakpoint, then the program stops and
     `GDB' signals that the breakpoint was encountered by printing the
     line of code before which the program is halted.

`breakpoint exception NAME'
     A special form of the breakpoint command which breakpoints whenever
     exception NAME is raised.  If NAME is omitted, then a breakpoint
     will occur when any exception is raised.

`print EXPRESSION'
     This will print the value of the given expression. Most simple Ada
     expression formats are properly handled by `GDB', so the expression
     can contain function calls, variables, operators, and attribute
     references.

`continue'
     Continues execution following a breakpoint, until the next
     breakpoint or the termination of the program.

`step'
     Executes a single line after a breakpoint. If the next statement
     is a subprogram call, execution continues into (the first
     statement of) the called subprogram.

`next'
     Executes a single line. If this line is a subprogram call,
     executes and returns from the call.

`list'
     Lists a few lines around the current source location. In practice,
     it is usually more convenient to have a separate edit window open
     with the relevant source file displayed. Successive applications
     of this command print subsequent lines. The command can be given
     an argument which is a line number, in which case it displays a
     few lines around the specified one.

`backtrace'
     Displays a backtrace of the call chain. This command is typically
     used after a breakpoint has occurred, to examine the sequence of
     calls that leads to the current breakpoint. The display includes
     one line for each activation record (frame) corresponding to an
     active subprogram.

`up'
     At a breakpoint, `GDB' can display the values of variables local
     to the current frame. The command `up' can be used to examine the
     contents of other active frames, by moving the focus up the stack,
     that is to say from callee to caller, one frame at a time.

`down'
     Moves the focus of `GDB' down from the frame currently being
     examined to the frame of its callee (the reverse of the previous
     command),

`frame N'
     Inspect the frame with the given number. The value 0 denotes the
     frame of the current breakpoint, that is to say the top of the
     call stack.

   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).

Using Ada Expressions
=====================

`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 important to the `GDB' user.

   * 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'.

Calling User-Defined Subprograms
================================

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.

Breaking on Ada Exceptions
==========================

You can set breakpoints that trip when your program raises selected
exceptions.

`break exception'
     Set a breakpoint that trips whenever (any task in the) program
     raises any exception.

`break exception NAME'
     Set a breakpoint that trips whenever (any task in the) program
     raises the exception NAME.

`break exception unhandled'
     Set a breakpoint that trips whenever (any task in the) program
     raises an exception for which there is no handler.

`info exceptions'
`info exceptions REGEXP'
     The `info exceptions' command permits the user to examine all
     defined exceptions within Ada programs. With a regular expression,
     REGEXP, as argument, prints out only those exceptions whose name
     matches REGEXP.

Ada Tasks
=========

`GDB' allows the following task-related commands:

`info tasks'
     This command shows a list of current Ada tasks, as in the
     following example:

          (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.

`break LINESPEC task TASKID'
`break LINESPEC task TASKID if ...'
     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 the `info tasks' display.

     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').

`task TASKNO'
     This command allows to switch to the task referred by TASKNO. In
     particular, This allows to browse the backtrace of the specified
     task. It is advised to switch back to the original task before
     continuing execution otherwise the scheduling of the program may be
     perturbated.

For more detailed information on the tasking support, see `Debugging
with GDB'.

Debugging Generic Units
=======================

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.

GNAT Abnormal Termination or Failure to Terminate
=================================================

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.

  1. Run `gcc' with the `-gnatf'. This first switch causes all errors
     on a given line to be reported. In its absence, only the first
     error on a line is displayed.

     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.

  2. Run `gcc' with the `-v (verbose)' switch. In this mode, `gcc'
     produces ongoing information about the progress of the compilation
     and provides the name of each procedure as code is generated. This
     switch allows you to find which Ada procedure was being compiled
     when it encountered a code generation problem.

  3. Run `gcc' with the `-gnatdc' switch. This is a GNAT specific
     switch that does for the front-end what `-v' does for the back end.
     The system prints the name of each unit, either a compilation unit
     or nested unit, as it is being analyzed.

  4. Finally, you can start `gdb' directly on the `gnat1' executable.
     `gnat1' is the front-end of GNAT, and can be run independently
     (normally it is just called from `gcc'). You can use `gdb' on
     `gnat1' as you would on a C program (but *note The GNAT Debugger
     GDB::. for caveats). The `where' command is the first line of
     attack; the variable `lineno' (seen by `print lineno'), used by
     the second phase of `gnat1' and by the `gcc' backend, indicates
     the source line at which the execution stopped, and `input_file
     name' indicates the name of the source file.

Naming Conventions for GNAT Source Files
========================================

In order to examine the workings of the GNAT system, the following
brief description of its organization may be helpful:

   * Files with prefix `sc' contain the lexical scanner.

   * All files prefixed with `par' are components of the parser. The
     numbers correspond to chapters of the Ada 95 Reference Manual. For
     example, parsing of select statements can be found in
     `par-ch9.adb'.

   * All files prefixed with `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.

   * All files prefixed with `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'.

   * The files prefixed with `bind' implement the binder, which
     verifies the consistency of the compilation, determines an order of
     elaboration, and generates the bind file.

   * The files `atree.ads' and `atree.adb' detail the low-level data
     structures used by the front-end.

   * The files `sinfo.ads' and `sinfo.adb' detail the structure of the
     abstract syntax tree as produced by the parser.

   * The files `einfo.ads' and `einfo.adb' detail the attributes of all
     entities, computed during semantic analysis.

   * Library management issues are dealt with in files with prefix
     `lib'.

   * Ada files with the prefix `a-' are children of `Ada', as defined
     in Annex A.

   * Files with prefix `i-' are children of `Interfaces', as defined in
     Annex B.

   * Files with prefix `s-' are children of `System'. This includes
     both language-defined children and GNAT run-time routines.

   * Files with prefix `g-' are children of `GNAT'. These are useful
     general-purpose packages, fully documented in their
     specifications. All the other `.c' files are modifications of
     common `gcc' files.

Getting Internal Debugging Information
======================================

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.

Performance Considerations
**************************

The GNAT system provides a number of options that allow a trade-off
between

   * performance of the generated code

   * speed of compilation

   * minimization of dependences and recompilation

   * the degree of run-time checking.

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:

   * no optimization

   * no inlining of subprogram calls

   * all run-time checks enabled except overflow and elaboration checks

These options are suitable for most program development purposes. This
chapter describes how you can modify these choices, and also provides
some guidelines on debugging optimized code.

Controlling Run-Time Checks
===========================

By default, GNAT produces all run-time checks, except arithmetic
overflow checking for integer operations (that includes division by
zero) and checks for access before elaboration on subprogram calls.
Two gnat switches, `-gnatp' and `-gnato' allow this default to be
modified. *Note Run-Time Checks::.

   Our experience is that the default is suitable for most development
purposes.

   We treat integer overflow specially because these are quite
expensive and in our experience are not as important as other run-time
checks in the development process.

   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.

   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.

Optimization Levels
===================

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 `-ON' switch, where N is an
integer from 0 to 3, on the `gcc' command line to control the
optimization level:

`-O0'
     no optimization (the default)

`-O1'
     medium level optimization

`-O2'
     full optimization

`-O3'
     full optimization, and also attempt automatic inlining of small
     subprograms within a unit (*note Inlining of Subprograms::.).

   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.

   Note: 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.

Debugging Optimized Code
========================

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.  This section describes the most common cases.

  1. The "hopping Program Counter":  Repeated 'step' or 'next' commands
     show the PC bouncing back and forth in the code.  This may result
     from any of the following optimizations:

        * Common subexpression elimination: using a single instance of
          code for a quantity that the source computes several times.
          As a result you may not be able to stop on what looks like a
          statement.

        * Invariant code motion: moving an expression that does not
          change within a loop, to the beginning of the loop.

        * Instruction scheduling: moving instructions so as to overlap
          loads and stores (typically) with other code, or in general
          to move computations of values closer to their uses. Often
          this causes you to pass an assignment statement without the
          assignment happening and then later bounce back to the
          statement when the value is actually needed.  Placing a
          breakpoint on a line of code and then stepping over it may,
          therefore, not always cause all the expected side-effects.

  2. The "big leap": More commonly known as cross-jumping, in which two
     identical pieces of code are merged and the program counter
     suddenly jumps to a statement that is not supposed to be executed,
     simply because it (and the code following) translates to the same
     thing as the code that *was* supposed to be executed.  This effect
     is typically seen in sequences that end in a jump, such as a
     `goto', a `return', or a `break' in a C `switch' statement.

  3. The "roving variable": The symptom is an unexpected value in a
     variable.  There are various reasons for this effect:

        * In a subprogram prologue, a parameter may not yet have been
          moved to its "home".

        * A variable may be dead, and its register re-used.  This is
          probably the most common cause.

        * As mentioned above, the assignment of a value to a variable
          may have been moved.

        * A variable may be eliminated entirely by value propagation or
          other means.  In this case, GCC may incorrectly generate
          debugging information for the variable

     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.

Inlining of Subprograms
=======================

A call to a subprogram in the current unit is inlined if all the
following conditions are met:

   * The optimization level is at least `-O1'.

   * The called subprogram is suitable for inlining: It must be small
     enough and not contain nested subprograms or anything else that
     `gcc' cannot support in inlined subprograms.

   * The call occurs after the definition of the body of the subprogram.

   * Either `pragma Inline' applies to the subprogram or it is small
     and automatic inlining (optimization level `-O3') is specified.

Calls to subprograms in `with''ed units are normally not inlined.  To
achieve this level of inlining, the following conditions must all be
true:

   * The optimization level is at least `-O1'.

   * The called subprogram is suitable for inlining: It must be small
     enough and not contain nested subprograms or anything else `gcc'
     cannot support in inlined subprograms.

   * The call appears in a body (not in a package spec).

   * There is a `pragma Inline' for the subprogram.

   * The `-gnatn' switch is used in the `gcc' command line

   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 can be used to prevent all inlining.
This switch overrides all other conditions and ensures that no inlining
occurs. The extra dependences resulting from `-gnatn' will still be
active, even if this switch is used to suppress the resulting inlining
actions.

Index
*****

* Menu:

* --GCC=compiler_name (gnatlink):        Switches for gnatlink.
* --GCC=compiler_name (gnatmake):        Switches for gnatmake.
* --GNATBIND=binder_name (gnatmake):     Switches for gnatmake.
* --GNATLINK=linker_name (gnatmake):     Switches for gnatmake.
* --LINK= (gnatlink):                    Switches for gnatlink.
* -83 (gnathtml):                        Converting Ada Files to html with gnathtml.
* -A (gnatbind):                         Output Control.
* -A (gnatlink):                         Switches for gnatlink.
* -a (gnatls):                           Switches for gnatls.
* -A (gnatmake):                         Switches for gnatmake.
* -a (gnatmake):                         Switches for gnatmake.
* -aI (gnatmake):                        Switches for gnatmake.
* -aL (gnatmake):                        Switches for gnatmake.
* -aO (gnatmake):                        Switches for gnatmake.
* -b (gcc):                              Switches for gcc.
* -B (gcc):                              Switches for gcc.
* -b (gnatbind):                         Binder Error Message Control.
* -B (gnatlink):                         Switches for gnatlink.
* -b (gnatlink):                         Switches for gnatlink.
* -bargs (gnatmake):                     Mode Switches for gnatmake.
* -c (gcc):                              Switches for gcc.
* -C (gnatbind):                         Output Control.
* -c (gnatbind):                         Output Control.
* -c (gnatchop):                         Switches for gnatchop.
* -C (gnatlink):                         Switches for gnatlink.
* -c (gnatmake):                         Switches for gnatmake.
* -cargs (gnatmake):                     Mode Switches for gnatmake.
* -d (gnathtml):                         Converting Ada Files to html with gnathtml.
* -d (gnatls):                           Switches for gnatls.
* -e (gnatbind):                         Output Control.
* -f (gnatbind):                         Elaboration Control.
* -f (gnathtml):                         Converting Ada Files to html with gnathtml.
* -f (gnatlink):                         Summary of Binder Switches.
* -f (gnatmake):                         Switches for gnatmake.
* -fno-inline (gcc):                     Inlining of Subprograms.
* -fstack-check:                         Stack Overflow Checking.
* -funwind-tables (gcc):                 Switches for gcc.
* -g (gcc):                              Switches for gcc.
* -g (gnatlink):                         Switches for gnatlink.
* -gnat83 (gcc):                         Compiling Ada 83 Programs.
* -gnata (gcc):                          Debugging and Assertion Control.
* -gnatb (gcc):                          Output and Error Message Control.
* -gnatc (gcc):                          Using gcc for Semantic Checking.
* -gnatD (gcc):                          Debugging Control.
* -gnatdc switch:                        GNAT Abnormal Termination or Failure to Terminate.
* -gnatE (gcc) <1>:                      Debugging Control.
* -gnatE (gcc):                          Run-Time Checks.
* -gnatf (gcc):                          Output and Error Message Control.
* -gnatG (gcc):                          Debugging Control.
* -gnati (gcc):                          Character Set Control.
* -gnatk (gcc):                          File Naming Control.
* -gnatl (gcc):                          Output and Error Message Control.
* -gnatm (gcc):                          Output and Error Message Control.
* -gnatN (gcc):                          Subprogram Inlining Control.
* -gnatn (gcc) <1>:                      Inlining of Subprograms.
* -gnatn (gcc):                          Subprogram Inlining Control.
* -gnatN switch:                         Source Dependencies.
* -gnatn switch:                         Source Dependencies.
* -gnato (gcc) <1>:                      Controlling Run-Time Checks.
* -gnato (gcc):                          Run-Time Checks.
* -gnatp (gcc) <1>:                      Run-Time Checks.
* -gnatp (gcc):                          Controlling Run-Time Checks.
* -gnatq (gcc):                          Output and Error Message Control.
* -gnatR (gcc):                          Debugging Control.
* -gnats (gcc):                          Using gcc for Syntax Checking.
* -gnatT (gcc):                          Run-Time Control.
* -gnatt (gcc):                          Auxiliary Output Control.
* -gnatu (gcc):                          Auxiliary Output Control.
* -gnatU (gcc):                          Output and Error Message Control.
* -gnatv (gcc):                          Output and Error Message Control.
* -gnatW (gcc):                          Character Set Control.
* -gnatwA (gcc):                         Output and Error Message Control.
* -gnatwa (gcc):                         Output and Error Message Control.
* -gnatwc (gcc):                         Output and Error Message Control.
* -gnatwC (gcc):                         Output and Error Message Control.
* -gnatwe (gcc):                         Output and Error Message Control.
* -gnatwH (gcc):                         Output and Error Message Control.
* -gnatwh (gcc):                         Output and Error Message Control.
* -gnatwI (gcc):                         Output and Error Message Control.
* -gnatwi (gcc):                         Output and Error Message Control.
* -gnatwl (gcc):                         Output and Error Message Control.
* -gnatwL (gcc):                         Output and Error Message Control.
* -gnatwo (gcc):                         Output and Error Message Control.
* -gnatwO (gcc):                         Output and Error Message Control.
* -gnatwP (gcc):                         Output and Error Message Control.
* -gnatwp (gcc):                         Output and Error Message Control.
* -gnatwr (gcc):                         Output and Error Message Control.
* -gnatwR (gcc):                         Output and Error Message Control.
* -gnatws (gcc):                         Output and Error Message Control.
* -gnatwU (gcc):                         Output and Error Message Control.
* -gnatwu (gcc):                         Output and Error Message Control.
* -gnatx (gcc):                          Debugging Control.
* -h (gnatbind) <1>:                     Output Control.
* -h (gnatbind):                         Elaboration Control.
* -h (gnatls):                           Switches for gnatls.
* -I (gcc):                              Switches for gcc.
* -I (gnathtml):                         Converting Ada Files to html with gnathtml.
* -I (gnatmake):                         Switches for gnatmake.
* -i (gnatmake):                         Switches for gnatmake.
* -i (gnatmem):                          Switches for gnatmem.
* -I- (gcc):                             Switches for gcc.
* -I- (gnatmake):                        Switches for gnatmake.
* -j (gnatmake):                         Switches for gnatmake.
* -k (gnatchop):                         Switches for gnatchop.
* -k (gnatmake):                         Switches for gnatmake.
* -l (gnatbind):                         Output Control.
* -l (gnathtml):                         Converting Ada Files to html with gnathtml.
* -L (gnatmake):                         Switches for gnatmake.
* -largs (gnatmake):                     Mode Switches for gnatmake.
* -m (gnatbind):                         Binder Error Message Control.
* -M (gnatbind):                         Binder Error Message Control.
* -m (gnatmake):                         Switches for gnatmake.
* -M (gnatmake):                         Switches for gnatmake.
* -n (gnatbind):                         Binding with Non-Ada Main Programs.
* -n (gnatlink):                         Switches for gnatlink.
* -n (gnatmake):                         Switches for gnatmake.
* -nostdinc (gnatmake):                  Switches for gnatmake.
* -nostdlib (gnatmake):                  Switches for gnatmake.
* -O (gcc):                              Optimization Levels.
* -o (gcc):                              Switches for gcc.
* -O (gcc):                              Switches for gcc.
* -O (gnatbind):                         Output Control.
* -o (gnatbind):                         Output Control.
* -o (gnathtml):                         Converting Ada Files to html with gnathtml.
* -o (gnatlink):                         Switches for gnatlink.
* -o (gnatls):                           Switches for gnatls.
* -o (gnatmake):                         Switches for gnatmake.
* -o (gnatmem):                          Switches for gnatmem.
* -p (gnathtml):                         Converting Ada Files to html with gnathtml.
* -pass-exit-codes (gcc):                Auxiliary Output Control.
* -q (gnatchop):                         Switches for gnatchop.
* -q (gnatmake):                         Switches for gnatmake.
* -q (gnatmem):                          Switches for gnatmem.
* -r (gnatchop):                         Switches for gnatchop.
* -S (gcc):                              Switches for gcc.
* -s (gnatbind):                         Consistency-Checking Modes.
* -s (gnatls):                           Switches for gnatls.
* -s (gnatmake):                         Switches for gnatmake.
* -sc (gnathtml):                        Converting Ada Files to html with gnathtml.
* -t (gnatbind):                         Binder Error Message Control.
* -t (gnathtml):                         Converting Ada Files to html with gnathtml.
* -u (gnatls):                           Switches for gnatls.
* -u (gnatmake):                         Switches for gnatmake.
* -v (gcc):                              Switches for gcc.
* -V (gcc):                              Switches for gcc.
* -v (gnatbind):                         Binder Error Message Control.
* -v (gnatchop):                         Switches for gnatchop.
* -v (gnatlink):                         Switches for gnatlink.
* -v (gnatmake):                         Switches for gnatmake.
* -v -v (gnatlink):                      Switches for gnatlink.
* -w:                                    Output and Error Message Control.
* -w (gnatchop):                         Switches for gnatchop.
* -we (gnatbind):                        Binder Error Message Control.
* -ws (gnatbind):                        Binder Error Message Control.
* -x (gnatbind):                         Consistency-Checking Modes.
* -z (gnatbind):                         Binding Programs with No Main Subprogram.
* -z (gnatmake):                         Switches for gnatmake.
* __gnat_finalize:                       Running gnatbind.
* __gnat_initialize:                     Running gnatbind.
* __gnat_set_globals:                    Running gnatbind.
* _main:                                 The External Symbol Naming Scheme of GNAT.
* Access before elaboration:             Run-Time Checks.
* Access-to-subprogram:                  Elaboration for Access-to-Subprogram Values.
* ACVC, Ada 83 tests:                    Compiling Ada 83 Programs.
* Ada <1>:                               Search Paths for gnatbind.
* Ada:                                   Naming Conventions for GNAT Source Files.
* Ada 83 compatibility:                  Compiling Ada 83 Programs.
* Ada 95 Language Reference Manual:      What You Should Know before Reading This Guide.
* Ada expressions:                       Using Ada Expressions.
* Ada Library Information files:         The Ada Library Information Files.
* Ada.Characters.Latin_1:                Latin-1.
* ADA_INCLUDE_PATH:                      Search Paths and the Run-Time Library (RTL).
* ADA_OBJECTS_PATH:                      Search Paths for gnatbind.
* adafinal <1>:                          Running gnatbind.
* adafinal:                              Binding with Non-Ada Main Programs.
* adainit <1>:                           Running gnatbind.
* adainit:                               Binding with Non-Ada Main Programs.
* ali files:                             The Ada Library Information Files.
* Annex A:                               Naming Conventions for GNAT Source Files.
* Annex B:                               Naming Conventions for GNAT Source Files.
* Asm:                                   Calling Conventions.
* Assert:                                Debugging and Assertion Control.
* Assertions:                            Debugging and Assertion Control.
* Binder consistency checks:             Binder Error Message Control.
* Binder output file:                    Interfacing to C.
* Binder, multiple input files:          Binding with Non-Ada Main Programs.
* Breakpoints and tasks:                 Ada Tasks.
* C:                                     Calling Conventions.
* C++:                                   Calling Conventions.
* Calling Conventions:                   Calling Conventions.
* Check, elaboration:                    Run-Time Checks.
* Check, overflow:                       Run-Time Checks.
* Checks, access before elaboration:     Run-Time Checks.
* Checks, division by zero:              Run-Time Checks.
* Checks, elaboration:                   Checking the Elaboration Order in Ada 95.
* Checks, overflow:                      Controlling Run-Time Checks.
* Checks, suppressing:                   Run-Time Checks.
* COBOL:                                 Calling Conventions.
* code page 437:                         Other 8-Bit Codes.
* code page 850:                         Other 8-Bit Codes.
* Combining GNAT switches:               Switches for gcc.
* Compilation model:                     The GNAT Compilation Model.
* Configuration pragmas:                 Configuration Pragmas.
* Consistency checks, in binder:         Binder Error Message Control.
* Convention Ada:                        Calling Conventions.
* Convention Asm:                        Calling Conventions.
* Convention Assembler:                  Calling Conventions.
* Convention C:                          Calling Conventions.
* Convention C++:                        Calling Conventions.
* Convention COBOL:                      Calling Conventions.
* Convention Fortran:                    Calling Conventions.
* Convention Stdcall:                    Calling Conventions.
* Convention Stubbed:                    Calling Conventions.
* Conventions:                           Conventions.
* CR:                                    Source Representation.
* Debug:                                 Debugging and Assertion Control.
* Debug Pool:                            Finding Memory Problems with GNAT Debug Pool.
* Debugger:                              Running and Debugging Ada Programs.
* Debugging:                             Running and Debugging Ada Programs.
* Debugging Generic Units:               Debugging Generic Units.
* Debugging information, including:      Switches for gnatlink.
* Debugging options:                     Debugging Control.
* Dependencies, producing list:          Switches for gnatmake.
* Dependency rules:                      The GNAT Make Program gnatmake.
* Division by zero:                      Run-Time Checks.
* Elaborate:                             Controlling the Elaboration Order in Ada 95.
* Elaborate_All:                         Controlling the Elaboration Order in Ada 95.
* Elaborate_Body:                        Controlling the Elaboration Order in Ada 95.
* Elaboration checks <1>:                Run-Time Checks.
* Elaboration checks:                    Checking the Elaboration Order in Ada 95.
* Elaboration control <1>:               Summary of Procedures for Elaboration Control.
* Elaboration control:                   Elaboration Order Handling in GNAT.
* Elaboration of library tasks:          Elaboration Issues for Library Tasks.
* Elaboration order control:             Comparison between GNAT and C/C++ Compilation Models.
* Eliminate:                             Eliminate Pragma.
* End of source file:                    Source Representation.
* Error messages, suppressing:           Output and Error Message Control.
* EUC Coding:                            Wide Character Encodings.
* Exceptions:                            Ada Exceptions.
* Export:                                The External Symbol Naming Scheme of GNAT.
* FF:                                    Source Representation.
* File names <1>:                        Using Other File Names.
* File names:                            Alternative File Naming Schemes.
* File naming schemes, alternative:      Alternative File Naming Schemes.
* Foreign Languages:                     Calling Conventions.
* Fortran:                               Calling Conventions.
* gdb:                                   Running and Debugging Ada Programs.
* Generic formal parameters:             Compiling Ada 83 Programs.
* Generics <1>:                          Generating Object Files.
* Generics:                              Debugging Generic Units.
* GMEM (gnatmem):                        Running gnatmem (GMEM Mode).
* GNAT <1>:                              Naming Conventions for GNAT Source Files.
* GNAT:                                  Search Paths for gnatbind.
* GNAT Abnormal Termination or Failure to Terminate: GNAT Abnormal Termination or Failure to Terminate.
* GNAT compilation model:                The GNAT Compilation Model.
* GNAT library:                          Comparison between GNAT and Conventional Ada Library Models.
* gnat.adc <1>:                          Using Other File Names.
* gnat.adc:                              The Configuration Pragmas File.
* gnat1:                                 Compiling Programs.
* gnat_argc:                             Command-Line Access.
* gnat_argv:                             Command-Line Access.
* GNAT_STACK_LIMIT:                      Stack Overflow Checking.
* gnatbind:                              Binding Using gnatbind.
* gnatchop:                              Renaming Files Using gnatchop.
* gnatelim:                              Reducing the Size of Ada Executables with gnatelim.
* gnatfind:                              The Cross-Referencing Tools gnatxref and gnatfind.
* gnatkr:                                File Name Krunching Using gnatkr.
* gnatlink:                              Linking Using gnatlink.
* gnatls:                                The GNAT Library Browser gnatls.
* gnatmake:                              The GNAT Make Program gnatmake.
* gnatmem:                               Finding Memory Problems with gnatmem.
* gnatprep:                              Preprocessing Using gnatprep.
* gnatstub:                              Creating Sample Bodies Using gnatstub.
* gnatxref:                              The Cross-Referencing Tools gnatxref and gnatfind.
* GNU make:                              Using gnatmake in a Makefile.
* HT:                                    Source Representation.
* Inline <1>:                            Source Dependencies.
* Inline:                                Inlining of Subprograms.
* Inlining:                              Comparison between GNAT and Conventional Ada Library Models.
* Interfaces <1>:                        Naming Conventions for GNAT Source Files.
* Interfaces:                            Search Paths for gnatbind.
* Interfacing to Ada:                    Calling Conventions.
* Interfacing to Assembler:              Calling Conventions.
* Interfacing to C:                      Calling Conventions.
* Interfacing to C++:                    Calling Conventions.
* Interfacing to COBOL:                  Calling Conventions.
* Interfacing to Fortran:                Calling Conventions.
* Internal trees, writing to file:       Auxiliary Output Control.
* Latin-1 <1>:                           Source Representation.
* Latin-1:                               Latin-1.
* Latin-2:                               Other 8-Bit Codes.
* Latin-3:                               Other 8-Bit Codes.
* Latin-4:                               Other 8-Bit Codes.
* LF:                                    Source Representation.
* Library browser:                       The GNAT Library Browser gnatls.
* Library tasks, elaboration issues:     Elaboration Issues for Library Tasks.
* Library, building, installing:         GNAT and Libraries.
* Linker libraries:                      Switches for gnatmake.
* Machine_Overflows:                     Run-Time Checks.
* Main Program:                          Running gnatbind.
* make:                                  Using the GNU make Utility.
* makefile:                              Using gnatmake in a Makefile.
* Mixed Language Programming:            Mixed Language Programming.
* Multiple units, syntax checking:       Using gcc for Syntax Checking.
* n (gnatmem):                           Switches for gnatmem.
* No code generated:                     Compiling Programs.
* No_Entry_Calls_In_Elaboration_Code:    Elaboration Issues for Library Tasks.
* Object file list:                      Running gnatbind.
* Order of elaboration:                  Elaboration Order Handling in GNAT.
* Other Ada compilers:                   Calling Conventions.
* Overflow checks <1>:                   Run-Time Checks.
* Overflow checks:                       Controlling Run-Time Checks.
* Parallel make:                         Switches for gnatmake.
* Performance:                           Performance Considerations.
* pragma Elaborate:                      Controlling the Elaboration Order in Ada 95.
* pragma Elaborate_All:                  Controlling the Elaboration Order in Ada 95.
* pragma Elaborate_Body:                 Controlling the Elaboration Order in Ada 95.
* pragma Inline:                         Inlining of Subprograms.
* pragma Preelaborate:                   Controlling the Elaboration Order in Ada 95.
* pragma Pure:                           Controlling the Elaboration Order in Ada 95.
* pragma Suppress:                       Controlling Run-Time Checks.
* pragma Unsuppress:                     Controlling Run-Time Checks.
* Pragmas, configuration:                Configuration Pragmas.
* Preelaborate:                          Controlling the Elaboration Order in Ada 95.
* Pure:                                  Controlling the Elaboration Order in Ada 95.
* Recompilation, by gnatmake:            Notes on the Command Line.
* RTL:                                   Switches for gcc.
* SDP_Table_Build:                       Running gnatbind.
* Search paths, for gnatmake:            Switches for gnatmake.
* Shift JIS Coding:                      Wide Character Encodings.
* Source file, end:                      Source Representation.
* Source files, suppressing search:      Switches for gnatmake.
* Source files, use by binder:           Running gnatbind.
* Source_File_Name pragma <1>:           Alternative File Naming Schemes.
* Source_File_Name pragma:               Using Other File Names.
* Source_Reference:                      Switches for gnatchop.
* Stack Overflow Checking:               Stack Overflow Checking.
* Stdcall:                               Calling Conventions.
* stderr:                                Output and Error Message Control.
* stdout:                                Output and Error Message Control.
* storage, pool, memory corruption:      Finding Memory Problems with GNAT Debug Pool.
* Stubbed:                               Calling Conventions.
* Style checking:                        Style Checking.
* SUB:                                   Source Representation.
* Subunits:                              Generating Object Files.
* Suppress <1>:                          Run-Time Checks.
* Suppress:                              Controlling Run-Time Checks.
* Suppressing checks:                    Run-Time Checks.
* System <1>:                            Naming Conventions for GNAT Source Files.
* System:                                Search Paths for gnatbind.
* System.IO:                             Search Paths and the Run-Time Library (RTL).
* Task switching:                        Ada Tasks.
* Tasks:                                 Ada Tasks.
* Time Slicing:                          Run-Time Control.
* Time stamp checks, in binder:          Binder Error Message Control.
* Tree file:                             Tree Files.
* Typographical conventions:             Conventions.
* Unsuppress <1>:                        Run-Time Checks.
* Unsuppress:                            Controlling Run-Time Checks.
* Upper-Half Coding:                     Wide Character Encodings.
* VT:                                    Source Representation.
* Warning messages:                      Output and Error Message Control.
* Warnings:                              Binder Error Message Control.
* Writing internal trees:                Auxiliary Output Control.
* Zero Cost Exceptions:                  Running gnatbind.

