1 \input texinfo @c -*-texinfo-*-
4 @c @setfilename usegcc.info
5 @c @setfilename portgcc.info
6 @c To produce the full manual, use the "gcc.info" setfilename, and
7 @c make sure the following do NOT begin with '@c' (and the @clear lines DO)
10 @c To produce a user-only manual, use the "usegcc.info" setfilename, and
11 @c make sure the following does NOT begin with '@c':
13 @c To produce a porter-only manual, use the "portgcc.info" setfilename,
14 @c and make sure the following does NOT begin with '@c':
17 @c (For FSF printing, turn on smallbook, comment out finalout below;
18 @c that is all that is needed.)
20 @c 6/27/96 FSF DO wants smallbook fmt for 1st bound edition.
23 @c i also commented out the finalout command, so if there *are* any
24 @c overfulls, you'll (hopefully) see the rectangle in the right hand
25 @c margin. -mew 15june93
28 @c NOTE: checks/things to do:
30 @c -have bob do a search in all seven files for "mew" (ideally --mew,
31 @c but i may have forgotten the occasional "--"..).
32 @c Just checked... all have `--'! Bob 22Jul96
33 @c Use this to search: grep -n '\-\-mew' *.texi
34 @c -item/itemx, text after all (sub/sub)section titles, etc..
35 @c -consider putting the lists of options on pp 17--> etc in columns or
38 @c -continuity of phrasing; ie, bit-field vs bitfield in rtl.texi
39 @c -overfulls. do a search for "mew" in the files, and you will see
40 @c overfulls that i noted but could not deal with.
41 @c -have to add text: beginning of chapter 8
44 @c anything else? --mew 10feb93
46 @macro gcctabopt{body}
49 @macro gccoptlist{body}
54 @c Makeinfo handles the above macro OK, TeX needs manual line breaks;
55 @c they get lost at some point in handling the macro. But if @macro is
56 @c used here rather than @alias, it produces double line breaks.
67 @settitle Using and Porting the GNU Compiler Collection (GCC)
70 @c seems reasonable to assume at least one of INTERNALS or USING is set...
72 @settitle Using the GNU Compiler Collection
75 @settitle Porting the GNU Compiler Collection
86 @c Use with @@smallbook.
88 @c Cause even numbered pages to be printed on the left hand side of
89 @c the page and odd numbered pages to be printed on the right hand
90 @c side of the page. Using this, you can print on both sides of a
91 @c sheet of paper and have the text on the same part of the sheet.
93 @c The text on right hand pages is pushed towards the right hand
94 @c margin and the text on left hand pages is pushed toward the left
96 @c (To provide the reverse effect, set bindingoffset to -0.75in.)
99 @c \global\bindingoffset=0.75in
100 @c \global\normaloffset =0.75in
104 @dircategory Programming
106 * gcc: (gcc). The GNU Compiler Collection.
110 This file documents the use and the internals of the GNU compiler.
114 This file documents the internals of the GNU compiler.
117 This file documents the use of the GNU compiler.
120 Published by the Free Software Foundation@*
121 59 Temple Place - Suite 330@*
122 Boston, MA 02111-1307 USA
124 @c When you update the list of years below, search for copyright{} and
125 @c update the other copy too.
126 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
127 1999, 2000, 2001 Free Software Foundation, Inc.
129 Permission is granted to copy, distribute and/or modify this document
130 under the terms of the GNU Free Documentation License, Version 1.1 or
131 any later version published by the Free Software Foundation; with the
132 Invariant Sections being ``GNU General Public License'' and ``Funding
133 Free Software'', the Front-Cover texts being (a) (see below), and with
134 the Back-Cover Texts being (b) (see below). A copy of the license is
135 included in the section entitled ``GNU Free Documentation License''.
137 (a) The FSF's Front-Cover Text is:
141 (b) The FSF's Back-Cover Text is:
143 You have freedom to copy and modify this GNU Manual, like GNU
144 software. Copies published by the Free Software Foundation raise
145 funds for GNU development.
148 @setchapternewpage odd
153 @center @titlefont{Using and Porting the GNU Compiler Collection}
158 @title Using the GNU Compiler Collection
161 @title Porting the GNU Compiler Collection
164 @center Richard M. Stallman
166 @center Last updated 9 May 2001
168 @c The version number appears five times more in this file.
172 @vskip 0pt plus 1filll
173 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1998,
174 1999, 2000, 2001 Free Software Foundation, Inc.
176 For GCC Version 3.1@*
178 Published by the Free Software Foundation @*
179 59 Temple Place - Suite 330@*
180 Boston, MA 02111-1307, USA@*
181 Last printed April, 1998.@*
182 Printed copies are available for $50 each.@*
185 Permission is granted to copy, distribute and/or modify this document
186 under the terms of the GNU Free Documentation License, Version 1.1 or
187 any later version published by the Free Software Foundation; with the
188 Invariant Sections being ``GNU General Public License'', the Front-Cover
189 texts being (a) (see below), and with the Back-Cover Texts being (b)
190 (see below). A copy of the license is included in the section entitled
191 ``GNU Free Documentation License''.
193 (a) The FSF's Front-Cover Text is:
197 (b) The FSF's Back-Cover Text is:
199 You have freedom to copy and modify this GNU Manual, like GNU
200 software. Copies published by the Free Software Foundation raise
201 funds for GNU development.
207 @node Top, G++ and GCC,, (DIR)
213 This manual documents how to run, install and port the GNU
214 compiler, as well as its new features and incompatibilities, and how to
215 report bugs. It corresponds to GCC version 3.1.
220 This manual documents how to run and install the GNU compiler,
221 as well as its new features and incompatibilities, and how to report
222 bugs. It corresponds to GCC version 3.1.
225 This manual documents how to port the GNU compiler,
226 as well as its new features and incompatibilities, and how to report
227 bugs. It corresponds to GCC version 3.1.
232 * G++ and GCC:: You can compile C or C++ programs.
233 * Standards:: Language standards supported by GCC.
234 * Invoking GCC:: Command options supported by @samp{gcc}.
235 * Installation:: How to configure, compile and install GCC.
236 * C Extensions:: GNU extensions to the C language family.
237 * C++ Extensions:: GNU extensions to the C++ language.
238 * Objective C:: GNU Objective-C runtime features.
239 * Gcov:: gcov: a GCC test coverage program.
240 * Trouble:: If you have trouble installing GCC.
241 * Bugs:: How, why and where to report bugs.
242 * Service:: How to find suppliers of support for GCC.
243 * Contributing:: How to contribute to testing and developing GCC.
244 * VMS:: Using GCC on VMS.
245 * Makefile:: List of Makefile targets.
248 * Portability:: Goals of GCC's portability features.
249 * Interface:: Function-call interface of GCC output.
250 * Passes:: Order of passes, what they do, and what each file is for.
251 * Trees:: The source representation used by the C and C++ front-ends.
252 * RTL:: The intermediate representation that most passes work on.
253 * Machine Desc:: How to write machine description instruction patterns.
254 * Target Macros:: How to write the machine description C macros.
255 * Config:: Writing the @file{xm-@var{machine}.h} file.
256 * Fragments:: Writing the @file{t-@var{target}} and @file{x-@var{host}} files.
259 * Funding:: How to help assure funding for free software.
260 * GNU/Linux:: Linux and the GNU Project
262 * Copying:: GNU General Public License says
263 how you can copy and share GCC.
264 * GNU Free Documentation License:: How you can copy and share this manual.
265 * Contributors:: People who have contributed to GCC.
267 * Index:: Index of concepts and symbol names.
272 @chapter Compile C, C++, Objective C, Fortran, Java or CHILL
275 Several versions of the compiler (C, C++, Objective C, Fortran, Java
276 and CHILL) are integrated; this is why we use the name
277 ``GNU Compiler Collection''. GCC can compile programs written in any of these
278 languages. The Fortran, CHILL, and Java compilers are described in
282 ``GCC'' is a common shorthand term for the GNU Compiler Collection. This is both
283 the most general name for the compiler, and the name used when the
284 emphasis is on compiling C programs (as the abbreviation formerly
285 stood for ``GNU C Compiler'').
289 When referring to C++ compilation, it is usual to call the compiler
290 ``G++''. Since there is only one compiler, it is also accurate to call
291 it ``GCC'' no matter what the language context; however, the term
292 ``G++'' is more useful when the emphasis is on compiling C++ programs.
294 We use the name ``GCC'' to refer to the compilation system as a
295 whole, and more specifically to the language-independent part of the
296 compiler. For example, we refer to the optimization options as
297 affecting the behavior of ``GCC'' or sometimes just ``the compiler''.
299 Front ends for other languages, such as Ada 95 and Pascal exist but
300 have not yet been integrated into GCC. These front-ends, like that for C++,
301 are built in subdirectories of GCC and link to it. The result is an
302 integrated compiler that can compile programs written in C, C++,
303 Objective C, or any of the languages for which you have installed front
306 In this manual, we only discuss the options for the C, Objective-C, and
307 C++ compilers and those of the GCC core. Consult the documentation
308 of the other front ends for the options to use when compiling programs
309 written in other languages.
311 @cindex compiler compared to C++ preprocessor
312 @cindex intermediate C version, nonexistent
313 @cindex C intermediate output, nonexistent
314 G++ is a @emph{compiler}, not merely a preprocessor. G++ builds object
315 code directly from your C++ program source. There is no intermediate C
316 version of the program. (By contrast, for example, some other
317 implementations use a program that generates a C program from your C++
318 source.) Avoiding an intermediate C representation of the program means
319 that you get better object code, and better debugging information. The
320 GNU debugger, GDB, works with this information in the object code to
321 give you comprehensive C++ source-level editing capabilities
322 (@pxref{C,,C and C++,gdb.info, Debugging with GDB}).
324 @c FIXME! Someone who knows something about Objective C ought to put in
325 @c a paragraph or two about it here, and move the index entry down when
326 @c there is more to point to than the general mention in the 1st par.
329 @chapter Language Standards Supported by GCC
332 @cindex ANSI C standard
336 @cindex ANSI X3.159-1989
338 @cindex ISO C standard
353 @cindex Technical Corrigenda
355 @cindex Technical Corrigendum 1
357 @cindex Technical Corrigendum 2
359 @cindex freestanding implementation
360 @cindex freestanding environment
361 @cindex hosted implementation
362 @cindex hosted environment
363 @findex __STDC_HOSTED__
365 For each language compiled by GCC for which there is a standard, GCC
366 attempts to follow one or more versions of that standard, possibly
367 with some exceptions, and possibly with some extensions.
369 GCC supports three versions of the C standard, although support for
370 the most recent version is not yet complete.
372 The original ANSI C standard (X3.159-1989) was ratified in 1989 and
373 published in 1990. This standard was ratified as an ISO standard
374 (ISO/IEC 9899:1990) later in 1990. There were no technical
375 differences between these publications, although the sections of the
376 ANSI standard were renumbered and became clauses in the ISO standard.
377 This standard, in both its forms, is commonly known as @dfn{C89}, or
378 occasionally as @dfn{C90}, from the dates of ratification. The ANSI
379 standard, but not the ISO standard, also came with a Rationale
380 document. To select this standard in GCC, use one of the options
381 @samp{-ansi}, @samp{-std=c89} or @samp{-std=iso9899:1990}; to obtain
382 all the diagnostics required by the standard, you should also specify
383 @samp{-pedantic} (or @samp{-pedantic-errors} if you want them to be
384 errors rather than warnings). @xref{C Dialect Options,,Options
385 Controlling C Dialect}.
387 Errors in the 1990 ISO C standard were corrected in two Technical
388 Corrigenda published in 1994 and 1996. GCC does not support the
391 An amendment to the 1990 standard was published in 1995. This
392 amendment added digraphs and @code{__STDC_VERSION__} to the language,
393 but otherwise concerned the library. This amendment is commonly known
394 as @dfn{AMD1}; the amended standard is sometimes known as @dfn{C94} or
395 @dfn{C95}. To select this standard in GCC, use the option
396 @samp{-std=iso9899:199409} (with, as for other standard versions,
397 @samp{-pedantic} to receive all required diagnostics).
399 A new edition of the ISO C standard was published in 1999 as ISO/IEC
400 9899:1999, and is commonly known as @dfn{C99}. GCC has incomplete
401 support for this standard version; see
402 @uref{http://gcc.gnu.org/c99status.html} for details. To select this
403 standard, use @samp{-std=c99} or @samp{-std=iso9899:1999}. (While in
404 development, drafts of this standard version were referred to as
407 GCC also has some limited support for traditional (pre-ISO) C with the
408 @samp{-traditional} option. This support may be of use for compiling
409 some very old programs that have not been updated to ISO C, but should
410 not be used for new programs. It will not work with some modern C
411 libraries such as the GNU C library.
413 By default, GCC provides some extensions to the C language that on
414 rare occasions conflict with the C standard. @xref{C
415 Extensions,,Extensions to the C Language Family}. Use of the
416 @samp{-std} options listed above will disable these extensions where
417 they conflict with the C standard version selected. You may also
418 select an extended version of the C language explicitly with
419 @samp{-std=gnu89} (for C89 with GNU extensions) or @samp{-std=gnu99}
420 (for C99 with GNU extensions). The default, if no C language dialect
421 options are given, is @samp{-std=gnu89}; this will change to
422 @samp{-std=gnu99} in some future release when the C99 support is
423 complete. Some features that are part of the C99 standard are
424 accepted as extensions in C89 mode.
426 The ISO C standard defines (in clause 4) two classes of conforming
427 implementation. A @dfn{conforming hosted implementation} supports the
428 whole standard including all the library facilities; a @dfn{conforming
429 freestanding implementation} is only required to provide certain
430 library facilities: those in @code{<float.h>}, @code{<limits.h>},
431 @code{<stdarg.h>}, and @code{<stddef.h>}; since AMD1, also those in
432 @code{<iso646.h>}; and in C99, also those in @code{<stdbool.h>} and
433 @code{<stdint.h>}. In addition, complex types, added in C99, are not
434 required for freestanding implementations. The standard also defines
435 two environments for programs, a @dfn{freestanding environment},
436 required of all implementations and which may not have library
437 facilities beyond those required of freestanding implementations,
438 where the handling of program startup and termination are
439 implementation-defined, and a @dfn{hosted environment}, which is not
440 required, in which all the library facilities are provided and startup
441 is through a function @code{int main (void)} or @code{int main (int,
442 char *[])}. An OS kernel would be a freestanding environment; a
443 program using the facilities of an operating system would normally be
444 in a hosted implementation.
446 GNU CC aims towards being usable as a conforming freestanding
447 implementation, or as the compiler for a conforming hosted
448 implementation. By default, it will act as the compiler for a hosted
449 implementation, defining @code{__STDC_HOSTED__} as @code{1} and
450 presuming that when the names of ISO C functions are used, they have
451 the semantics defined in the standard. To make it act as a conforming
452 freestanding implementation for a freestanding environment, use the
453 option @samp{-ffreestanding}; it will then define
454 @code{__STDC_HOSTED__} to @code{0} and not make assumptions about the
455 meanings of function names from the standard library. To build an OS
456 kernel, you may well still need to make your own arrangements for
457 linking and startup. @xref{C Dialect Options,,Options Controlling C
460 GNU CC does not provide the library facilities required only of hosted
461 implementations, nor yet all the facilities required by C99 of
462 freestanding implementations; to use the facilities of a hosted
463 environment, you will need to find them elsewhere (for example, in the
464 GNU C library). @xref{Standard Libraries,,Standard Libraries}.
466 For references to Technical Corrigenda, Rationale documents and
467 information concerning the history of C that is available online, see
468 @uref{http://gcc.gnu.org/readings.html}
470 @c FIXME: details of C++ standard.
472 There is no formal written standard for Objective-C. The most
473 authoritative manual is ``Object-Oriented Programming and the
474 Objective-C Language'', available at a number of web sites;
475 @uref{http://developer.apple.com/techpubs/macosx/Cocoa/ObjectiveC} has a
476 recent version, while @uref{http://www.toodarkpark.org/computers/objc/}
477 is an older example. @uref{http://www.gnustep.org} includes useful
480 @xref{Language,,The GNU Fortran Language, g77, Using and Porting GNU
481 Fortran}, for details of the Fortran language supported by GCC.
483 @xref{Compatibility,,Compatibility with the Java Platform, gcj, GNU gcj},
484 for details of compatibility between @code{gcj} and the Java Platform.
486 @xref{References,,Language Definition References, chill, GNU Chill},
487 for details of the CHILL standard.
491 @include install-old.texi
500 @chapter Known Causes of Trouble with GCC
502 @cindex installation trouble
503 @cindex known causes of trouble
505 This section describes known problems that affect users of GCC. Most
506 of these are not GCC bugs per se---if they were, we would fix them.
507 But the result for a user may be like the result of a bug.
509 Some of these problems are due to bugs in other software, some are
510 missing features that are too much work to add, and some are places
511 where people's opinions differ as to what is best.
514 * Actual Bugs:: Bugs we will fix later.
515 * Cross-Compiler Problems:: Common problems of cross compiling with GCC.
516 * Interoperation:: Problems using GCC with other compilers,
517 and with certain linkers, assemblers and debuggers.
518 * External Bugs:: Problems compiling certain programs.
519 * Incompatibilities:: GCC is incompatible with traditional C.
520 * Fixed Headers:: GNU C uses corrected versions of system header files.
521 This is necessary, but doesn't always work smoothly.
522 * Standard Libraries:: GNU C uses the system C library, which might not be
523 compliant with the ISO C standard.
524 * Disappointments:: Regrettable things we can't change, but not quite bugs.
525 * C++ Misunderstandings:: Common misunderstandings with GNU C++.
526 * Protoize Caveats:: Things to watch out for when using @code{protoize}.
527 * Non-bugs:: Things we think are right, but some others disagree.
528 * Warnings and Errors:: Which problems in your code get warnings,
529 and which get errors.
533 @section Actual Bugs We Haven't Fixed Yet
537 The @code{fixincludes} script interacts badly with automounters; if the
538 directory of system header files is automounted, it tends to be
539 unmounted while @code{fixincludes} is running. This would seem to be a
540 bug in the automounter. We don't know any good way to work around it.
543 The @code{fixproto} script will sometimes add prototypes for the
544 @code{sigsetjmp} and @code{siglongjmp} functions that reference the
545 @code{jmp_buf} type before that type is defined. To work around this,
546 edit the offending file and place the typedef in front of the
550 When @samp{-pedantic-errors} is specified, GCC will incorrectly give
551 an error message when a function name is specified in an expression
552 involving the comma operator.
555 @node Cross-Compiler Problems
556 @section Cross-Compiler Problems
558 You may run into problems with cross compilation on certain machines,
563 Cross compilation can run into trouble for certain machines because
564 some target machines' assemblers require floating point numbers to be
565 written as @emph{integer} constants in certain contexts.
567 The compiler writes these integer constants by examining the floating
568 point value as an integer and printing that integer, because this is
569 simple to write and independent of the details of the floating point
570 representation. But this does not work if the compiler is running on
571 a different machine with an incompatible floating point format, or
572 even a different byte-ordering.
574 In addition, correct constant folding of floating point values
575 requires representing them in the target machine's format.
576 (The C standard does not quite require this, but in practice
577 it is the only way to win.)
579 It is now possible to overcome these problems by defining macros such
580 as @code{REAL_VALUE_TYPE}. But doing so is a substantial amount of
581 work for each target machine.
583 @xref{Cross-compilation}.
586 @xref{Cross-compilation,,Cross Compilation and Floating Point Format,
587 gcc.info, Using and Porting GCC}.
591 At present, the program @file{mips-tfile} which adds debug
592 support to object files on MIPS systems does not work in a cross
597 @section Interoperation
599 This section lists various difficulties encountered in using GNU C or
600 GNU C++ together with other compilers or with the assemblers, linkers,
601 libraries and debuggers on certain systems.
605 Objective C does not work on the RS/6000.
608 GNU C++ does not do name mangling in the same way as other C++
609 compilers. This means that object files compiled with one compiler
610 cannot be used with another.
612 This effect is intentional, to protect you from more subtle problems.
613 Compilers differ as to many internal details of C++ implementation,
614 including: how class instances are laid out, how multiple inheritance is
615 implemented, and how virtual function calls are handled. If the name
616 encoding were made the same, your programs would link against libraries
617 provided from other compilers---but the programs would then crash when
618 run. Incompatible libraries are then detected at link time, rather than
622 Older GDB versions sometimes fail to read the output of GCC version
623 2. If you have trouble, get GDB version 4.4 or later.
627 DBX rejects some files produced by GCC, though it accepts similar
628 constructs in output from PCC. Until someone can supply a coherent
629 description of what is valid DBX input and what is not, there is
630 nothing I can do about these problems. You are on your own.
633 The GNU assembler (GAS) does not support PIC. To generate PIC code, you
634 must use some other assembler, such as @file{/bin/as}.
637 On some BSD systems, including some versions of Ultrix, use of profiling
638 causes static variable destructors (currently used only in C++) not to
642 Use of @samp{-I/usr/include} may cause trouble.
644 Many systems come with header files that won't work with GCC unless
645 corrected by @code{fixincludes}. The corrected header files go in a new
646 directory; GCC searches this directory before @file{/usr/include}.
647 If you use @samp{-I/usr/include}, this tells GCC to search
648 @file{/usr/include} earlier on, before the corrected headers. The
649 result is that you get the uncorrected header files.
651 Instead, you should use these options (when compiling C programs):
654 -I/usr/local/lib/gcc-lib/@var{target}/@var{version}/include -I/usr/include
657 For C++ programs, GCC also uses a special directory that defines C++
658 interfaces to standard C subroutines. This directory is meant to be
659 searched @emph{before} other standard include directories, so that it
660 takes precedence. If you are compiling C++ programs and specifying
661 include directories explicitly, use this option first, then the two
665 -I/usr/local/lib/g++-include
669 @cindex @code{vfork}, for the Sun-4
671 There is a bug in @code{vfork} on the Sun-4 which causes the registers
672 of the child process to clobber those of the parent. Because of this,
673 programs that call @code{vfork} are likely to lose when compiled
674 optimized with GCC when the child code alters registers which contain
675 C variables in the parent. This affects variables which are live in the
676 parent across the call to @code{vfork}.
678 If you encounter this, you can work around the problem by declaring
679 variables @code{volatile} in the function that calls @code{vfork}, until
680 the problem goes away, or by not declaring them @code{register} and not
681 using @samp{-O} for those source files.
685 On some SGI systems, when you use @samp{-lgl_s} as an option,
686 it gets translated magically to @samp{-lgl_s -lX11_s -lc_s}.
687 Naturally, this does not happen when you use GCC.
688 You must specify all three options explicitly.
691 On a Sparc, GCC aligns all values of type @code{double} on an 8-byte
692 boundary, and it expects every @code{double} to be so aligned. The Sun
693 compiler usually gives @code{double} values 8-byte alignment, with one
694 exception: function arguments of type @code{double} may not be aligned.
696 As a result, if a function compiled with Sun CC takes the address of an
697 argument of type @code{double} and passes this pointer of type
698 @code{double *} to a function compiled with GCC, dereferencing the
699 pointer may cause a fatal signal.
701 One way to solve this problem is to compile your entire program with GNU
702 CC. Another solution is to modify the function that is compiled with
703 Sun CC to copy the argument into a local variable; local variables
704 are always properly aligned. A third solution is to modify the function
705 that uses the pointer to dereference it via the following function
706 @code{access_double} instead of directly with @samp{*}:
710 access_double (double *unaligned_ptr)
712 union d2i @{ double d; int i[2]; @};
714 union d2i *p = (union d2i *) unaligned_ptr;
725 Storing into the pointer can be done likewise with the same union.
728 On Solaris, the @code{malloc} function in the @file{libmalloc.a} library
729 may allocate memory that is only 4 byte aligned. Since GCC on the
730 Sparc assumes that doubles are 8 byte aligned, this may result in a
731 fatal signal if doubles are stored in memory allocated by the
732 @file{libmalloc.a} library.
734 The solution is to not use the @file{libmalloc.a} library. Use instead
735 @code{malloc} and related functions from @file{libc.a}; they do not have
739 Sun forgot to include a static version of @file{libdl.a} with some
740 versions of SunOS (mainly 4.1). This results in undefined symbols when
741 linking static binaries (that is, if you use @samp{-static}). If you
742 see undefined symbols @code{_dlclose}, @code{_dlsym} or @code{_dlopen}
743 when linking, compile and link against the file
744 @file{mit/util/misc/dlsym.c} from the MIT version of X windows.
747 The 128-bit long double format that the Sparc port supports currently
748 works by using the architecturally defined quad-word floating point
749 instructions. Since there is no hardware that supports these
750 instructions they must be emulated by the operating system. Long
751 doubles do not work in Sun OS versions 4.0.3 and earlier, because the
752 kernel emulator uses an obsolete and incompatible format. Long doubles
753 do not work in Sun OS version 4.1.1 due to a problem in a Sun library.
754 Long doubles do work on Sun OS versions 4.1.2 and higher, but GCC
755 does not enable them by default. Long doubles appear to work in Sun OS
759 On HP-UX version 9.01 on the HP PA, the HP compiler @code{cc} does not
760 compile GCC correctly. We do not yet know why. However, GCC
761 compiled on earlier HP-UX versions works properly on HP-UX 9.01 and can
762 compile itself properly on 9.01.
765 On the HP PA machine, ADB sometimes fails to work on functions compiled
766 with GCC. Specifically, it fails to work on functions that use
767 @code{alloca} or variable-size arrays. This is because GCC doesn't
768 generate HP-UX unwind descriptors for such functions. It may even be
769 impossible to generate them.
772 Debugging (@samp{-g}) is not supported on the HP PA machine, unless you use
773 the preliminary GNU tools (@pxref{Installation}).
776 Taking the address of a label may generate errors from the HP-UX
777 PA assembler. GAS for the PA does not have this problem.
780 Using floating point parameters for indirect calls to static functions
781 will not work when using the HP assembler. There simply is no way for GCC
782 to specify what registers hold arguments for static functions when using
783 the HP assembler. GAS for the PA does not have this problem.
786 In extremely rare cases involving some very large functions you may
787 receive errors from the HP linker complaining about an out of bounds
788 unconditional branch offset. This used to occur more often in previous
789 versions of GCC, but is now exceptionally rare. If you should run
790 into it, you can work around by making your function smaller.
793 GCC compiled code sometimes emits warnings from the HP-UX assembler of
797 (warning) Use of GR3 when
798 frame >= 8192 may cause conflict.
801 These warnings are harmless and can be safely ignored.
804 The current version of the assembler (@file{/bin/as}) for the RS/6000
805 has certain problems that prevent the @samp{-g} option in GCC from
806 working. Note that @file{Makefile.in} uses @samp{-g} by default when
807 compiling @file{libgcc2.c}.
809 IBM has produced a fixed version of the assembler. The upgraded
810 assembler unfortunately was not included in any of the AIX 3.2 update
811 PTF releases (3.2.2, 3.2.3, or 3.2.3e). Users of AIX 3.1 should request
812 PTF U403044 from IBM and users of AIX 3.2 should request PTF U416277.
813 See the file @file{README.RS6000} for more details on these updates.
815 You can test for the presence of a fixed assembler by using the
823 If the command exits normally, the assembler fix already is installed.
824 If the assembler complains that "-u" is an unknown flag, you need to
828 On the IBM RS/6000, compiling code of the form
839 will cause the linker to report an undefined symbol @code{foo}.
840 Although this behavior differs from most other systems, it is not a
841 bug because redefining an @code{extern} variable as @code{static}
842 is undefined in ISO C.
845 AIX on the RS/6000 provides support (NLS) for environments outside of
846 the United States. Compilers and assemblers use NLS to support
847 locale-specific representations of various objects including
848 floating-point numbers ("." vs "," for separating decimal fractions).
849 There have been problems reported where the library linked with GCC does
850 not produce the same floating-point formats that the assembler accepts.
851 If you have this problem, set the LANG environment variable to "C" or
855 Even if you specify @samp{-fdollars-in-identifiers},
856 you cannot successfully use @samp{$} in identifiers on the RS/6000 due
857 to a restriction in the IBM assembler. GAS supports these
861 On the RS/6000, XLC version 1.3.0.0 will miscompile @file{jump.c}. XLC
862 version 1.3.0.1 or later fixes this problem. You can obtain XLC-1.3.0.2
863 by requesting PTF 421749 from IBM.
866 There is an assembler bug in versions of DG/UX prior to 5.4.2.01 that
867 occurs when the @samp{fldcr} instruction is used. GCC uses
868 @samp{fldcr} on the 88100 to serialize volatile memory references. Use
869 the option @samp{-mno-serialize-volatile} if your version of the
870 assembler has this bug.
873 On VMS, GAS versions 1.38.1 and earlier may cause spurious warning
874 messages from the linker. These warning messages complain of mismatched
875 psect attributes. You can ignore them. @xref{VMS Install}.
878 On NewsOS version 3, if you include both of the files @file{stddef.h}
879 and @file{sys/types.h}, you get an error because there are two typedefs
880 of @code{size_t}. You should change @file{sys/types.h} by adding these
881 lines around the definition of @code{size_t}:
886 @var{actual typedef here}
892 On the Alliant, the system's own convention for returning structures
893 and unions is unusual, and is not compatible with GCC no matter
894 what options are used.
899 On the IBM RT PC, the MetaWare HighC compiler (hc) uses a different
900 convention for structure and union returning. Use the option
901 @samp{-mhc-struct-return} to tell GCC to use a convention compatible
904 @cindex Vax calling convention
905 @cindex Ultrix calling convention
907 On Ultrix, the Fortran compiler expects registers 2 through 5 to be saved
908 by function calls. However, the C compiler uses conventions compatible
909 with BSD Unix: registers 2 through 5 may be clobbered by function calls.
911 GCC uses the same convention as the Ultrix C compiler. You can use
912 these options to produce code compatible with the Fortran compiler:
915 -fcall-saved-r2 -fcall-saved-r3 -fcall-saved-r4 -fcall-saved-r5
919 On the WE32k, you may find that programs compiled with GCC do not
920 work with the standard shared C library. You may need to link with
921 the ordinary C compiler. If you do so, you must specify the following
925 -L/usr/local/lib/gcc-lib/we32k-att-sysv/2.8.1 -lgcc -lc_s
928 The first specifies where to find the library @file{libgcc.a}
929 specified with the @samp{-lgcc} option.
931 GCC does linking by invoking @code{ld}, just as @code{cc} does, and
932 there is no reason why it @emph{should} matter which compilation program
933 you use to invoke @code{ld}. If someone tracks this problem down,
934 it can probably be fixed easily.
937 On the Alpha, you may get assembler errors about invalid syntax as a
938 result of floating point constants. This is due to a bug in the C
939 library functions @code{ecvt}, @code{fcvt} and @code{gcvt}. Given valid
940 floating point numbers, they sometimes print @samp{NaN}.
943 On Irix 4.0.5F (and perhaps in some other versions), an assembler bug
944 sometimes reorders instructions incorrectly when optimization is turned
945 on. If you think this may be happening to you, try using the GNU
946 assembler; GAS version 2.1 supports ECOFF on Irix.
948 Or use the @samp{-noasmopt} option when you compile GCC with itself,
949 and then again when you compile your program. (This is a temporary
950 kludge to turn off assembler optimization on Irix.) If this proves to
951 be what you need, edit the assembler spec in the file @file{specs} so
952 that it unconditionally passes @samp{-O0} to the assembler, and never
953 passes @samp{-O2} or @samp{-O3}.
957 @section Problems Compiling Certain Programs
959 @c prevent bad page break with this line
960 Certain programs have problems compiling.
964 Parse errors may occur compiling X11 on a Decstation running Ultrix 4.2
965 because of problems in DEC's versions of the X11 header files
966 @file{X11/Xlib.h} and @file{X11/Xutil.h}. People recommend adding
967 @samp{-I/usr/include/mit} to use the MIT versions of the header files,
968 using the @samp{-traditional} switch to turn off ISO C, or fixing the
969 header files by adding this:
973 #define NeedFunctionPrototypes 0
978 On various 386 Unix systems derived from System V, including SCO, ISC,
979 and ESIX, you may get error messages about running out of virtual memory
980 while compiling certain programs.
982 You can prevent this problem by linking GCC with the GNU malloc
983 (which thus replaces the malloc that comes with the system). GNU malloc
984 is available as a separate package, and also in the file
985 @file{src/gmalloc.c} in the GNU Emacs 19 distribution.
987 If you have installed GNU malloc as a separate library package, use this
988 option when you relink GCC:
991 MALLOC=/usr/local/lib/libgmalloc.a
994 Alternatively, if you have compiled @file{gmalloc.c} from Emacs 19, copy
995 the object file to @file{gmalloc.o} and use this option when you relink
1003 @node Incompatibilities
1004 @section Incompatibilities of GCC
1005 @cindex incompatibilities of GCC
1007 There are several noteworthy incompatibilities between GNU C and K&R
1008 (non-ISO) versions of C. The @samp{-traditional} option
1009 eliminates many of these incompatibilities, @emph{but not all}, by
1010 telling GNU C to behave like a K&R C compiler.
1013 @cindex string constants
1014 @cindex read-only strings
1015 @cindex shared strings
1017 GCC normally makes string constants read-only. If several
1018 identical-looking string constants are used, GCC stores only one
1021 @cindex @code{mktemp}, and constant strings
1022 One consequence is that you cannot call @code{mktemp} with a string
1023 constant argument. The function @code{mktemp} always alters the
1024 string its argument points to.
1026 @cindex @code{sscanf}, and constant strings
1027 @cindex @code{fscanf}, and constant strings
1028 @cindex @code{scanf}, and constant strings
1029 Another consequence is that @code{sscanf} does not work on some systems
1030 when passed a string constant as its format control string or input.
1031 This is because @code{sscanf} incorrectly tries to write into the string
1032 constant. Likewise @code{fscanf} and @code{scanf}.
1034 The best solution to these problems is to change the program to use
1035 @code{char}-array variables with initialization strings for these
1036 purposes instead of string constants. But if this is not possible,
1037 you can use the @samp{-fwritable-strings} flag, which directs GCC
1038 to handle string constants the same way most C compilers do.
1039 @samp{-traditional} also has this effect, among others.
1042 @code{-2147483648} is positive.
1044 This is because 2147483648 cannot fit in the type @code{int}, so
1045 (following the ISO C rules) its data type is @code{unsigned long int}.
1046 Negating this value yields 2147483648 again.
1049 GCC does not substitute macro arguments when they appear inside of
1050 string constants. For example, the following macro in GCC
1057 will produce output @code{"a"} regardless of what the argument @var{a} is.
1059 The @samp{-traditional} option directs GCC to handle such cases
1060 (among others) in the old-fashioned (non-ISO) fashion.
1062 @cindex @code{setjmp} incompatibilities
1063 @cindex @code{longjmp} incompatibilities
1065 When you use @code{setjmp} and @code{longjmp}, the only automatic
1066 variables guaranteed to remain valid are those declared
1067 @code{volatile}. This is a consequence of automatic register
1068 allocation. Consider this function:
1082 /* @r{@code{longjmp (j)} may occur in @code{fun3}.} */
1087 Here @code{a} may or may not be restored to its first value when the
1088 @code{longjmp} occurs. If @code{a} is allocated in a register, then
1089 its first value is restored; otherwise, it keeps the last value stored
1092 If you use the @samp{-W} option with the @samp{-O} option, you will
1093 get a warning when GCC thinks such a problem might be possible.
1095 The @samp{-traditional} option directs GNU C to put variables in
1096 the stack by default, rather than in registers, in functions that
1097 call @code{setjmp}. This results in the behavior found in
1098 traditional C compilers.
1101 Programs that use preprocessing directives in the middle of macro
1102 arguments do not work with GCC. For example, a program like this
1111 ISO C does not permit such a construct. It would make sense to support
1112 it when @samp{-traditional} is used, but it is too much work to
1116 K&R compilers allow comments to cross over an inclusion boundary (i.e.
1117 started in an include file and ended in the including file). I think
1118 this would be quite ugly and can't imagine it could be needed.
1120 @cindex external declaration scope
1121 @cindex scope of external declarations
1122 @cindex declaration scope
1124 Declarations of external variables and functions within a block apply
1125 only to the block containing the declaration. In other words, they
1126 have the same scope as any other declaration in the same place.
1128 In some other C compilers, a @code{extern} declaration affects all the
1129 rest of the file even if it happens within a block.
1131 The @samp{-traditional} option directs GNU C to treat all @code{extern}
1132 declarations as global, like traditional compilers.
1135 In traditional C, you can combine @code{long}, etc., with a typedef name,
1140 typedef long foo bar;
1143 In ISO C, this is not allowed: @code{long} and other type modifiers
1144 require an explicit @code{int}. Because this criterion is expressed
1145 by Bison grammar rules rather than C code, the @samp{-traditional}
1146 flag cannot alter it.
1148 @cindex typedef names as function parameters
1150 PCC allows typedef names to be used as function parameters. The
1151 difficulty described immediately above applies here too.
1154 When in @samp{-traditional} mode, GCC allows the following erroneous
1155 pair of declarations to appear together in a given scope:
1163 GCC treats all characters of identifiers as significant, even when in
1164 @samp{-traditional} mode. According to K&R-1 (2.2), ``No more than the
1165 first eight characters are significant, although more may be used.''.
1166 Also according to K&R-1 (2.2), ``An identifier is a sequence of letters
1167 and digits; the first character must be a letter. The underscore _
1168 counts as a letter.'', but GCC also allows dollar signs in identifiers.
1172 PCC allows whitespace in the middle of compound assignment operators
1173 such as @samp{+=}. GCC, following the ISO standard, does not
1174 allow this. The difficulty described immediately above applies here
1180 GCC complains about unterminated character constants inside of
1181 preprocessing conditionals that fail. Some programs have English
1182 comments enclosed in conditionals that are guaranteed to fail; if these
1183 comments contain apostrophes, GCC will probably report an error. For
1184 example, this code would produce an error:
1188 You can't expect this to work.
1192 The best solution to such a problem is to put the text into an actual
1193 C comment delimited by @samp{/*@dots{}*/}. However,
1194 @samp{-traditional} suppresses these error messages.
1197 Many user programs contain the declaration @samp{long time ();}. In the
1198 past, the system header files on many systems did not actually declare
1199 @code{time}, so it did not matter what type your program declared it to
1200 return. But in systems with ISO C headers, @code{time} is declared to
1201 return @code{time_t}, and if that is not the same as @code{long}, then
1202 @samp{long time ();} is erroneous.
1204 The solution is to change your program to use appropriate system headers
1205 (@code{<time.h>} on systems with ISO C headers) and not to declare
1206 @code{time} if the system header files declare it, or failing that to
1207 use @code{time_t} as the return type of @code{time}.
1209 @cindex @code{float} as function value type
1211 When compiling functions that return @code{float}, PCC converts it to
1212 a double. GCC actually returns a @code{float}. If you are concerned
1213 with PCC compatibility, you should declare your functions to return
1214 @code{double}; you might as well say what you mean.
1219 When compiling functions that return structures or unions, GCC
1220 output code normally uses a method different from that used on most
1221 versions of Unix. As a result, code compiled with GCC cannot call
1222 a structure-returning function compiled with PCC, and vice versa.
1224 The method used by GCC is as follows: a structure or union which is
1225 1, 2, 4 or 8 bytes long is returned like a scalar. A structure or union
1226 with any other size is stored into an address supplied by the caller
1227 (usually in a special, fixed register, but on some machines it is passed
1228 on the stack). The machine-description macros @code{STRUCT_VALUE} and
1229 @code{STRUCT_INCOMING_VALUE} tell GCC where to pass this address.
1231 By contrast, PCC on most target machines returns structures and unions
1232 of any size by copying the data into an area of static storage, and then
1233 returning the address of that storage as if it were a pointer value.
1234 The caller must copy the data from that memory area to the place where
1235 the value is wanted. GCC does not use this method because it is
1236 slower and nonreentrant.
1238 On some newer machines, PCC uses a reentrant convention for all
1239 structure and union returning. GCC on most of these machines uses a
1240 compatible convention when returning structures and unions in memory,
1241 but still returns small structures and unions in registers.
1243 You can tell GCC to use a compatible convention for all structure and
1244 union returning with the option @samp{-fpcc-struct-return}.
1246 @cindex preprocessing tokens
1247 @cindex preprocessing numbers
1249 GNU C complains about program fragments such as @samp{0x74ae-0x4000}
1250 which appear to be two hexadecimal constants separated by the minus
1251 operator. Actually, this string is a single @dfn{preprocessing token}.
1252 Each such token must correspond to one token in C. Since this does not,
1253 GNU C prints an error message. Although it may appear obvious that what
1254 is meant is an operator and two values, the ISO C standard specifically
1255 requires that this be treated as erroneous.
1257 A @dfn{preprocessing token} is a @dfn{preprocessing number} if it
1258 begins with a digit and is followed by letters, underscores, digits,
1259 periods and @samp{e+}, @samp{e-}, @samp{E+}, @samp{E-}, @samp{p+},
1260 @samp{p-}, @samp{P+}, or @samp{P-} character sequences. (In strict C89
1261 mode, the sequences @samp{p+}, @samp{p-}, @samp{P+} and @samp{P-} cannot
1262 appear in preprocessing numbers.)
1264 To make the above program fragment valid, place whitespace in front of
1265 the minus sign. This whitespace will end the preprocessing number.
1269 @section Fixed Header Files
1271 GCC needs to install corrected versions of some system header files.
1272 This is because most target systems have some header files that won't
1273 work with GCC unless they are changed. Some have bugs, some are
1274 incompatible with ISO C, and some depend on special features of other
1277 Installing GCC automatically creates and installs the fixed header
1278 files, by running a program called @code{fixincludes} (or for certain
1279 targets an alternative such as @code{fixinc.svr4}). Normally, you
1280 don't need to pay attention to this. But there are cases where it
1281 doesn't do the right thing automatically.
1285 If you update the system's header files, such as by installing a new
1286 system version, the fixed header files of GCC are not automatically
1287 updated. The easiest way to update them is to reinstall GCC. (If
1288 you want to be clever, look in the makefile and you can find a
1292 On some systems, in particular SunOS 4, header file directories contain
1293 machine-specific symbolic links in certain places. This makes it
1294 possible to share most of the header files among hosts running the
1295 same version of SunOS 4 on different machine models.
1297 The programs that fix the header files do not understand this special
1298 way of using symbolic links; therefore, the directory of fixed header
1299 files is good only for the machine model used to build it.
1301 In SunOS 4, only programs that look inside the kernel will notice the
1302 difference between machine models. Therefore, for most purposes, you
1303 need not be concerned about this.
1305 It is possible to make separate sets of fixed header files for the
1306 different machine models, and arrange a structure of symbolic links so
1307 as to use the proper set, but you'll have to do this by hand.
1310 On Lynxos, GCC by default does not fix the header files. This is
1311 because bugs in the shell cause the @code{fixincludes} script to fail.
1313 This means you will encounter problems due to bugs in the system header
1314 files. It may be no comfort that they aren't GCC's fault, but it
1315 does mean that there's nothing for us to do about them.
1318 @node Standard Libraries
1319 @section Standard Libraries
1321 GCC by itself attempts to be a conforming freestanding implementation.
1322 @xref{Standards,,Language Standards Supported by GCC}, for details of
1323 what this means. Beyond the library facilities required of such an
1324 implementation, the rest of the C library is supplied by the vendor of
1325 the operating system. If that C library doesn't conform to the C
1326 standards, then your programs might get warnings (especially when using
1327 @samp{-Wall}) that you don't expect.
1329 For example, the @code{sprintf} function on SunOS 4.1.3 returns
1330 @code{char *} while the C standard says that @code{sprintf} returns an
1331 @code{int}. The @code{fixincludes} program could make the prototype for
1332 this function match the Standard, but that would be wrong, since the
1333 function will still return @code{char *}.
1335 If you need a Standard compliant library, then you need to find one, as
1336 GCC does not provide one. The GNU C library (called @code{glibc})
1337 provides ISO C, POSIX, BSD, SystemV and X/Open compatibility for
1338 GNU/Linux and HURD-based GNU systems; no recent version of it supports
1339 other systems, though some very old versions did. Version 2.2 of the
1340 GNU C library includes nearly complete C99 support. You could also ask
1341 your operating system vendor if newer libraries are available.
1343 @node Disappointments
1344 @section Disappointments and Misunderstandings
1346 These problems are perhaps regrettable, but we don't know any practical
1351 Certain local variables aren't recognized by debuggers when you compile
1354 This occurs because sometimes GCC optimizes the variable out of
1355 existence. There is no way to tell the debugger how to compute the
1356 value such a variable ``would have had'', and it is not clear that would
1357 be desirable anyway. So GCC simply does not mention the eliminated
1358 variable when it writes debugging information.
1360 You have to expect a certain amount of disagreement between the
1361 executable and your source code, when you use optimization.
1363 @cindex conflicting types
1364 @cindex scope of declaration
1366 Users often think it is a bug when GCC reports an error for code
1370 int foo (struct mumble *);
1372 struct mumble @{ @dots{} @};
1374 int foo (struct mumble *x)
1378 This code really is erroneous, because the scope of @code{struct
1379 mumble} in the prototype is limited to the argument list containing it.
1380 It does not refer to the @code{struct mumble} defined with file scope
1381 immediately below---they are two unrelated types with similar names in
1384 But in the definition of @code{foo}, the file-scope type is used
1385 because that is available to be inherited. Thus, the definition and
1386 the prototype do not match, and you get an error.
1388 This behavior may seem silly, but it's what the ISO standard specifies.
1389 It is easy enough for you to make your code work by moving the
1390 definition of @code{struct mumble} above the prototype. It's not worth
1391 being incompatible with ISO C just to avoid an error for the example
1395 Accesses to bitfields even in volatile objects works by accessing larger
1396 objects, such as a byte or a word. You cannot rely on what size of
1397 object is accessed in order to read or write the bitfield; it may even
1398 vary for a given bitfield according to the precise usage.
1400 If you care about controlling the amount of memory that is accessed, use
1401 volatile but do not use bitfields.
1404 GCC comes with shell scripts to fix certain known problems in system
1405 header files. They install corrected copies of various header files in
1406 a special directory where only GCC will normally look for them. The
1407 scripts adapt to various systems by searching all the system header
1408 files for the problem cases that we know about.
1410 If new system header files are installed, nothing automatically arranges
1411 to update the corrected header files. You will have to reinstall GCC
1412 to fix the new header files. More specifically, go to the build
1413 directory and delete the files @file{stmp-fixinc} and
1414 @file{stmp-headers}, and the subdirectory @code{include}; then do
1415 @samp{make install} again.
1418 @cindex floating point precision
1419 On 68000 and x86 systems, for instance, you can get paradoxical results
1420 if you test the precise values of floating point numbers. For example,
1421 you can find that a floating point value which is not a NaN is not equal
1422 to itself. This results from the fact that the floating point registers
1423 hold a few more bits of precision than fit in a @code{double} in memory.
1424 Compiled code moves values between memory and floating point registers
1425 at its convenience, and moving them into memory truncates them.
1427 You can partially avoid this problem by using the @samp{-ffloat-store}
1428 option (@pxref{Optimize Options}).
1431 On the MIPS, variable argument functions using @file{varargs.h}
1432 cannot have a floating point value for the first argument. The
1433 reason for this is that in the absence of a prototype in scope,
1434 if the first argument is a floating point, it is passed in a
1435 floating point register, rather than an integer register.
1437 If the code is rewritten to use the ISO standard @file{stdarg.h}
1438 method of variable arguments, and the prototype is in scope at
1439 the time of the call, everything will work fine.
1442 On the H8/300 and H8/300H, variable argument functions must be
1443 implemented using the ISO standard @file{stdarg.h} method of
1444 variable arguments. Furthermore, calls to functions using @file{stdarg.h}
1445 variable arguments must have a prototype for the called function
1446 in scope at the time of the call.
1449 @node C++ Misunderstandings
1450 @section Common Misunderstandings with GNU C++
1452 @cindex misunderstandings in C++
1453 @cindex surprises in C++
1454 @cindex C++ misunderstandings
1455 C++ is a complex language and an evolving one, and its standard
1456 definition (the ISO C++ standard) was only recently completed. As a
1457 result, your C++ compiler may occasionally surprise you, even when its
1458 behavior is correct. This section discusses some areas that frequently
1459 give rise to questions of this sort.
1462 * Static Definitions:: Static member declarations are not definitions
1463 * Temporaries:: Temporaries may vanish before you expect
1464 * Copy Assignment:: Copy Assignment operators copy virtual bases twice
1467 @node Static Definitions
1468 @subsection Declare @emph{and} Define Static Members
1470 @cindex C++ static data, declaring and defining
1471 @cindex static data in C++, declaring and defining
1472 @cindex declaring static data in C++
1473 @cindex defining static data in C++
1474 When a class has static data members, it is not enough to @emph{declare}
1475 the static member; you must also @emph{define} it. For example:
1486 This declaration only establishes that the class @code{Foo} has an
1487 @code{int} named @code{Foo::bar}, and a member function named
1488 @code{Foo::method}. But you still need to define @emph{both}
1489 @code{method} and @code{bar} elsewhere. According to the ISO
1490 standard, you must supply an initializer in one (and only one) source
1497 Other C++ compilers may not correctly implement the standard behavior.
1498 As a result, when you switch to @code{g++} from one of these compilers,
1499 you may discover that a program that appeared to work correctly in fact
1500 does not conform to the standard: @code{g++} reports as undefined
1501 symbols any static data members that lack definitions.
1504 @subsection Temporaries May Vanish Before You Expect
1506 @cindex temporaries, lifetime of
1507 @cindex portions of temporary objects, pointers to
1508 It is dangerous to use pointers or references to @emph{portions} of a
1509 temporary object. The compiler may very well delete the object before
1510 you expect it to, leaving a pointer to garbage. The most common place
1511 where this problem crops up is in classes like string classes,
1512 especially ones that define a conversion function to type @code{char *}
1513 or @code{const char *} -- which is one reason why the standard
1514 @code{string} class requires you to call the @code{c_str} member
1515 function. However, any class that returns a pointer to some internal
1516 structure is potentially subject to this problem.
1518 For example, a program may use a function @code{strfunc} that returns
1519 @code{string} objects, and another function @code{charfunc} that
1520 operates on pointers to @code{char}:
1524 void charfunc (const char *);
1529 const char *p = strfunc().c_str();
1538 In this situation, it may seem reasonable to save a pointer to the C
1539 string returned by the @code{c_str} member function and use that rather
1540 than call @code{c_str} repeatedly. However, the temporary string
1541 created by the call to @code{strfunc} is destroyed after @code{p} is
1542 initialized, at which point @code{p} is left pointing to freed memory.
1544 Code like this may run successfully under some other compilers,
1545 particularly obsolete cfront-based compilers that delete temporaries
1546 along with normal local variables. However, the GNU C++ behavior is
1547 standard-conforming, so if your program depends on late destruction of
1548 temporaries it is not portable.
1550 The safe way to write such code is to give the temporary a name, which
1551 forces it to remain until the end of the scope of the name. For
1555 string& tmp = strfunc ();
1556 charfunc (tmp.c_str ());
1559 @node Copy Assignment
1560 @subsection Implicit Copy-Assignment for Virtual Bases
1562 When a base class is virtual, only one subobject of the base class
1563 belongs to each full object. Also, the constructors and destructors are
1564 invoked only once, and called from the most-derived class. However, such
1565 objects behave unspecified when being assigned. For example:
1570 Base(char *n) : name(strdup(n))@{@}
1571 Base& operator= (const Base& other)@{
1573 name = strdup (other.name);
1577 struct A:virtual Base@{
1582 struct B:virtual Base@{
1587 struct Derived:public A, public B@{
1588 Derived():Base("Derived")@{@}
1591 void func(Derived &d1, Derived &d2)
1597 The C++ standard specifies that @samp{Base::Base} is only called once
1598 when constructing or copy-constructing a Derived object. It is
1599 unspecified whether @samp{Base::operator=} is called more than once when
1600 the implicit copy-assignment for Derived objects is invoked (as it is
1601 inside @samp{func} in the example).
1603 g++ implements the "intuitive" algorithm for copy-assignment: assign all
1604 direct bases, then assign all members. In that algorithm, the virtual
1605 base subobject can be encountered many times. In the example, copying
1606 proceeds in the following order: @samp{val}, @samp{name} (via
1607 @code{strdup}), @samp{bval}, and @samp{name} again.
1609 If application code relies on copy-assignment, a user-defined
1610 copy-assignment operator removes any uncertainties. With such an
1611 operator, the application can define whether and how the virtual base
1612 subobject is assigned.
1614 @node Protoize Caveats
1615 @section Caveats of using @code{protoize}
1617 The conversion programs @code{protoize} and @code{unprotoize} can
1618 sometimes change a source file in a way that won't work unless you
1623 @code{protoize} can insert references to a type name or type tag before
1624 the definition, or in a file where they are not defined.
1626 If this happens, compiler error messages should show you where the new
1627 references are, so fixing the file by hand is straightforward.
1630 There are some C constructs which @code{protoize} cannot figure out.
1631 For example, it can't determine argument types for declaring a
1632 pointer-to-function variable; this you must do by hand. @code{protoize}
1633 inserts a comment containing @samp{???} each time it finds such a
1634 variable; so you can find all such variables by searching for this
1635 string. ISO C does not require declaring the argument types of
1636 pointer-to-function types.
1639 Using @code{unprotoize} can easily introduce bugs. If the program
1640 relied on prototypes to bring about conversion of arguments, these
1641 conversions will not take place in the program without prototypes.
1642 One case in which you can be sure @code{unprotoize} is safe is when
1643 you are removing prototypes that were made with @code{protoize}; if
1644 the program worked before without any prototypes, it will work again
1647 You can find all the places where this problem might occur by compiling
1648 the program with the @samp{-Wconversion} option. It prints a warning
1649 whenever an argument is converted.
1652 Both conversion programs can be confused if there are macro calls in and
1653 around the text to be converted. In other words, the standard syntax
1654 for a declaration or definition must not result from expanding a macro.
1655 This problem is inherent in the design of C and cannot be fixed. If
1656 only a few functions have confusing macro calls, you can easily convert
1660 @code{protoize} cannot get the argument types for a function whose
1661 definition was not actually compiled due to preprocessing conditionals.
1662 When this happens, @code{protoize} changes nothing in regard to such
1663 a function. @code{protoize} tries to detect such instances and warn
1666 You can generally work around this problem by using @code{protoize} step
1667 by step, each time specifying a different set of @samp{-D} options for
1668 compilation, until all of the functions have been converted. There is
1669 no automatic way to verify that you have got them all, however.
1672 Confusion may result if there is an occasion to convert a function
1673 declaration or definition in a region of source code where there is more
1674 than one formal parameter list present. Thus, attempts to convert code
1675 containing multiple (conditionally compiled) versions of a single
1676 function header (in the same vicinity) may not produce the desired (or
1679 If you plan on converting source files which contain such code, it is
1680 recommended that you first make sure that each conditionally compiled
1681 region of source code which contains an alternative function header also
1682 contains at least one additional follower token (past the final right
1683 parenthesis of the function header). This should circumvent the
1687 @code{unprotoize} can become confused when trying to convert a function
1688 definition or declaration which contains a declaration for a
1689 pointer-to-function formal argument which has the same name as the
1690 function being defined or declared. We recommend you avoid such choices
1691 of formal parameter names.
1694 You might also want to correct some of the indentation by hand and break
1695 long lines. (The conversion programs don't write lines longer than
1696 eighty characters in any case.)
1700 @section Certain Changes We Don't Want to Make
1702 This section lists changes that people frequently request, but which
1703 we do not make because we think GCC is better without them.
1707 Checking the number and type of arguments to a function which has an
1708 old-fashioned definition and no prototype.
1710 Such a feature would work only occasionally---only for calls that appear
1711 in the same file as the called function, following the definition. The
1712 only way to check all calls reliably is to add a prototype for the
1713 function. But adding a prototype eliminates the motivation for this
1714 feature. So the feature is not worthwhile.
1717 Warning about using an expression whose type is signed as a shift count.
1719 Shift count operands are probably signed more often than unsigned.
1720 Warning about this would cause far more annoyance than good.
1723 Warning about assigning a signed value to an unsigned variable.
1725 Such assignments must be very common; warning about them would cause
1726 more annoyance than good.
1729 Warning when a non-void function value is ignored.
1731 Coming as I do from a Lisp background, I balk at the idea that there is
1732 something dangerous about discarding a value. There are functions that
1733 return values which some callers may find useful; it makes no sense to
1734 clutter the program with a cast to @code{void} whenever the value isn't
1738 Making @samp{-fshort-enums} the default.
1740 This would cause storage layout to be incompatible with most other C
1741 compilers. And it doesn't seem very important, given that you can get
1742 the same result in other ways. The case where it matters most is when
1743 the enumeration-valued object is inside a structure, and in that case
1744 you can specify a field width explicitly.
1747 Making bitfields unsigned by default on particular machines where ``the
1748 ABI standard'' says to do so.
1750 The ISO C standard leaves it up to the implementation whether a bitfield
1751 declared plain @code{int} is signed or not. This in effect creates two
1752 alternative dialects of C.
1754 The GNU C compiler supports both dialects; you can specify the signed
1755 dialect with @samp{-fsigned-bitfields} and the unsigned dialect with
1756 @samp{-funsigned-bitfields}. However, this leaves open the question of
1757 which dialect to use by default.
1759 Currently, the preferred dialect makes plain bitfields signed, because
1760 this is simplest. Since @code{int} is the same as @code{signed int} in
1761 every other context, it is cleanest for them to be the same in bitfields
1764 Some computer manufacturers have published Application Binary Interface
1765 standards which specify that plain bitfields should be unsigned. It is
1766 a mistake, however, to say anything about this issue in an ABI. This is
1767 because the handling of plain bitfields distinguishes two dialects of C.
1768 Both dialects are meaningful on every type of machine. Whether a
1769 particular object file was compiled using signed bitfields or unsigned
1770 is of no concern to other object files, even if they access the same
1771 bitfields in the same data structures.
1773 A given program is written in one or the other of these two dialects.
1774 The program stands a chance to work on most any machine if it is
1775 compiled with the proper dialect. It is unlikely to work at all if
1776 compiled with the wrong dialect.
1778 Many users appreciate the GNU C compiler because it provides an
1779 environment that is uniform across machines. These users would be
1780 inconvenienced if the compiler treated plain bitfields differently on
1783 Occasionally users write programs intended only for a particular machine
1784 type. On these occasions, the users would benefit if the GNU C compiler
1785 were to support by default the same dialect as the other compilers on
1786 that machine. But such applications are rare. And users writing a
1787 program to run on more than one type of machine cannot possibly benefit
1788 from this kind of compatibility.
1790 This is why GCC does and will treat plain bitfields in the same
1791 fashion on all types of machines (by default).
1793 There are some arguments for making bitfields unsigned by default on all
1794 machines. If, for example, this becomes a universal de facto standard,
1795 it would make sense for GCC to go along with it. This is something
1796 to be considered in the future.
1798 (Of course, users strongly concerned about portability should indicate
1799 explicitly in each bitfield whether it is signed or not. In this way,
1800 they write programs which have the same meaning in both C dialects.)
1803 Undefining @code{__STDC__} when @samp{-ansi} is not used.
1805 Currently, GCC defines @code{__STDC__} as long as you don't use
1806 @samp{-traditional}. This provides good results in practice.
1808 Programmers normally use conditionals on @code{__STDC__} to ask whether
1809 it is safe to use certain features of ISO C, such as function
1810 prototypes or ISO token concatenation. Since plain @samp{gcc} supports
1811 all the features of ISO C, the correct answer to these questions is
1814 Some users try to use @code{__STDC__} to check for the availability of
1815 certain library facilities. This is actually incorrect usage in an ISO
1816 C program, because the ISO C standard says that a conforming
1817 freestanding implementation should define @code{__STDC__} even though it
1818 does not have the library facilities. @samp{gcc -ansi -pedantic} is a
1819 conforming freestanding implementation, and it is therefore required to
1820 define @code{__STDC__}, even though it does not come with an ISO C
1823 Sometimes people say that defining @code{__STDC__} in a compiler that
1824 does not completely conform to the ISO C standard somehow violates the
1825 standard. This is illogical. The standard is a standard for compilers
1826 that claim to support ISO C, such as @samp{gcc -ansi}---not for other
1827 compilers such as plain @samp{gcc}. Whatever the ISO C standard says
1828 is relevant to the design of plain @samp{gcc} without @samp{-ansi} only
1829 for pragmatic reasons, not as a requirement.
1831 GCC normally defines @code{__STDC__} to be 1, and in addition
1832 defines @code{__STRICT_ANSI__} if you specify the @option{-ansi} option,
1833 or a @option{-std} option for strict conformance to some version of ISO C.
1834 On some hosts, system include files use a different convention, where
1835 @code{__STDC__} is normally 0, but is 1 if the user specifies strict
1836 conformance to the C Standard. GCC follows the host convention when
1837 processing system include files, but when processing user files it follows
1838 the usual GNU C convention.
1841 Undefining @code{__STDC__} in C++.
1843 Programs written to compile with C++-to-C translators get the
1844 value of @code{__STDC__} that goes with the C compiler that is
1845 subsequently used. These programs must test @code{__STDC__}
1846 to determine what kind of C preprocessor that compiler uses:
1847 whether they should concatenate tokens in the ISO C fashion
1848 or in the traditional fashion.
1850 These programs work properly with GNU C++ if @code{__STDC__} is defined.
1851 They would not work otherwise.
1853 In addition, many header files are written to provide prototypes in ISO
1854 C but not in traditional C. Many of these header files can work without
1855 change in C++ provided @code{__STDC__} is defined. If @code{__STDC__}
1856 is not defined, they will all fail, and will all need to be changed to
1857 test explicitly for C++ as well.
1860 Deleting ``empty'' loops.
1862 Historically, GCC has not deleted ``empty'' loops under the
1863 assumption that the most likely reason you would put one in a program is
1864 to have a delay, so deleting them will not make real programs run any
1867 However, the rationale here is that optimization of a nonempty loop
1868 cannot produce an empty one, which holds for C but is not always the
1871 Moreover, with @samp{-funroll-loops} small ``empty'' loops are already
1872 removed, so the current behavior is both sub-optimal and inconsistent
1873 and will change in the future.
1876 Making side effects happen in the same order as in some other compiler.
1878 @cindex side effects, order of evaluation
1879 @cindex order of evaluation, side effects
1880 It is never safe to depend on the order of evaluation of side effects.
1881 For example, a function call like this may very well behave differently
1882 from one compiler to another:
1885 void func (int, int);
1891 There is no guarantee (in either the C or the C++ standard language
1892 definitions) that the increments will be evaluated in any particular
1893 order. Either increment might happen first. @code{func} might get the
1894 arguments @samp{2, 3}, or it might get @samp{3, 2}, or even @samp{2, 2}.
1897 Not allowing structures with volatile fields in registers.
1899 Strictly speaking, there is no prohibition in the ISO C standard
1900 against allowing structures with volatile fields in registers, but
1901 it does not seem to make any sense and is probably not what you wanted
1902 to do. So the compiler will give an error message in this case.
1905 Making certain warnings into errors by default.
1907 Some ISO C testsuites report failure when the compiler does not produce
1908 an error message for a certain program.
1910 ISO C requires a ``diagnostic'' message for certain kinds of invalid
1911 programs, but a warning is defined by GCC to count as a diagnostic. If
1912 GCC produces a warning but not an error, that is correct ISO C support.
1913 If test suites call this ``failure'', they should be run with the GCC
1914 option @samp{-pedantic-errors}, which will turn these warnings into
1919 @node Warnings and Errors
1920 @section Warning Messages and Error Messages
1922 @cindex error messages
1923 @cindex warnings vs errors
1924 @cindex messages, warning and error
1925 The GNU compiler can produce two kinds of diagnostics: errors and
1926 warnings. Each kind has a different purpose:
1930 @emph{Errors} report problems that make it impossible to compile your
1931 program. GCC reports errors with the source file name and line
1932 number where the problem is apparent.
1935 @emph{Warnings} report other unusual conditions in your code that
1936 @emph{may} indicate a problem, although compilation can (and does)
1937 proceed. Warning messages also report the source file name and line
1938 number, but include the text @samp{warning:} to distinguish them
1939 from error messages.
1942 Warnings may indicate danger points where you should check to make sure
1943 that your program really does what you intend; or the use of obsolete
1944 features; or the use of nonstandard features of GNU C or C++. Many
1945 warnings are issued only if you ask for them, with one of the @samp{-W}
1946 options (for instance, @samp{-Wall} requests a variety of useful
1949 GCC always tries to compile your program if possible; it never
1950 gratuitously rejects a program whose meaning is clear merely because
1951 (for instance) it fails to conform to a standard. In some cases,
1952 however, the C and C++ standards specify that certain extensions are
1953 forbidden, and a diagnostic @emph{must} be issued by a conforming
1954 compiler. The @samp{-pedantic} option tells GCC to issue warnings in
1955 such cases; @samp{-pedantic-errors} says to make them errors instead.
1956 This does not mean that @emph{all} non-ISO constructs get warnings
1959 @xref{Warning Options,,Options to Request or Suppress Warnings}, for
1960 more detail on these and related command-line options.
1963 @chapter Reporting Bugs
1965 @cindex reporting bugs
1967 Your bug reports play an essential role in making GCC reliable.
1969 When you encounter a problem, the first thing to do is to see if it is
1970 already known. @xref{Trouble}. If it isn't known, then you should
1973 Reporting a bug may help you by bringing a solution to your problem, or
1974 it may not. (If it does not, look in the service directory; see
1975 @ref{Service}.) In any case, the principal function of a bug report is
1976 to help the entire community by making the next version of GCC work
1977 better. Bug reports are your contribution to the maintenance of GCC.
1979 Since the maintainers are very overloaded, we cannot respond to every
1980 bug report. However, if the bug has not been fixed, we are likely to
1981 send you a patch and ask you to tell us whether it works.
1983 In order for a bug report to serve its purpose, you must include the
1984 information that makes for fixing the bug.
1987 * Criteria: Bug Criteria. Have you really found a bug?
1988 * Where: Bug Lists. Where to send your bug report.
1989 * Reporting: Bug Reporting. How to report a bug effectively.
1990 * GNATS: gccbug. You can use a bug reporting tool.
1991 * Patches: Sending Patches. How to send a patch for GCC.
1992 * Known: Trouble. Known problems.
1993 * Help: Service. Where to ask for help.
1996 @node Bug Criteria,Bug Lists,,Bugs
1997 @section Have You Found a Bug?
1998 @cindex bug criteria
2000 If you are not sure whether you have found a bug, here are some guidelines:
2003 @cindex fatal signal
2006 If the compiler gets a fatal signal, for any input whatever, that is a
2007 compiler bug. Reliable compilers never crash.
2009 @cindex invalid assembly code
2010 @cindex assembly code, invalid
2012 If the compiler produces invalid assembly code, for any input whatever
2013 (except an @code{asm} statement), that is a compiler bug, unless the
2014 compiler reports errors (not just warnings) which would ordinarily
2015 prevent the assembler from being run.
2017 @cindex undefined behavior
2018 @cindex undefined function value
2019 @cindex increment operators
2021 If the compiler produces valid assembly code that does not correctly
2022 execute the input source code, that is a compiler bug.
2024 However, you must double-check to make sure, because you may have run
2025 into an incompatibility between GNU C and traditional C
2026 (@pxref{Incompatibilities}). These incompatibilities might be considered
2027 bugs, but they are inescapable consequences of valuable features.
2029 Or you may have a program whose behavior is undefined, which happened
2030 by chance to give the desired results with another C or C++ compiler.
2032 For example, in many nonoptimizing compilers, you can write @samp{x;}
2033 at the end of a function instead of @samp{return x;}, with the same
2034 results. But the value of the function is undefined if @code{return}
2035 is omitted; it is not a bug when GCC produces different results.
2037 Problems often result from expressions with two increment operators,
2038 as in @code{f (*p++, *p++)}. Your previous compiler might have
2039 interpreted that expression the way you intended; GCC might
2040 interpret it another way. Neither compiler is wrong. The bug is
2043 After you have localized the error to a single source line, it should
2044 be easy to check for these things. If your program is correct and
2045 well defined, you have found a compiler bug.
2048 If the compiler produces an error message for valid input, that is a
2051 @cindex invalid input
2053 If the compiler does not produce an error message for invalid input,
2054 that is a compiler bug. However, you should note that your idea of
2055 ``invalid input'' might be my idea of ``an extension'' or ``support
2056 for traditional practice''.
2059 If you are an experienced user of one of the languages GCC supports, your
2060 suggestions for improvement of GCC are welcome in any case.
2063 @node Bug Lists,Bug Reporting,Bug Criteria,Bugs
2064 @section Where to Report Bugs
2065 @cindex bug report mailing lists
2066 @kindex gcc-bugs@@gcc.gnu.org or bug-gcc@@gnu.org
2067 Send bug reports for the GNU Compiler Collection to
2068 @email{gcc-bugs@@gcc.gnu.org}. In accordance with the GNU-wide
2069 convention, in which bug reports for tool ``foo'' are sent
2070 to @samp{bug-foo@@gnu.org}, the address @email{bug-gcc@@gnu.org}
2071 may also be used; it will forward to the address given above.
2073 Please read @uref{http://gcc.gnu.org/bugs.html} for additional and/or
2074 more up-to-date bug reporting instructions before you post a bug report.
2076 @node Bug Reporting,gccbug,Bug Lists,Bugs
2077 @section How to Report Bugs
2078 @cindex compiler bugs, reporting
2080 The fundamental principle of reporting bugs usefully is this:
2081 @strong{report all the facts}. If you are not sure whether to state a
2082 fact or leave it out, state it!
2084 Often people omit facts because they think they know what causes the
2085 problem and they conclude that some details don't matter. Thus, you might
2086 assume that the name of the variable you use in an example does not matter.
2087 Well, probably it doesn't, but one cannot be sure. Perhaps the bug is a
2088 stray memory reference which happens to fetch from the location where that
2089 name is stored in memory; perhaps, if the name were different, the contents
2090 of that location would fool the compiler into doing the right thing despite
2091 the bug. Play it safe and give a specific, complete example. That is the
2092 easiest thing for you to do, and the most helpful.
2094 Keep in mind that the purpose of a bug report is to enable someone to
2095 fix the bug if it is not known. It isn't very important what happens if
2096 the bug is already known. Therefore, always write your bug reports on
2097 the assumption that the bug is not known.
2099 Sometimes people give a few sketchy facts and ask, ``Does this ring a
2100 bell?'' This cannot help us fix a bug, so it is basically useless. We
2101 respond by asking for enough details to enable us to investigate.
2102 You might as well expedite matters by sending them to begin with.
2104 Try to make your bug report self-contained. If we have to ask you for
2105 more information, it is best if you include all the previous information
2106 in your response, as well as the information that was missing.
2108 Please report each bug in a separate message. This makes it easier for
2109 us to track which bugs have been fixed and to forward your bugs reports
2110 to the appropriate maintainer.
2112 To enable someone to investigate the bug, you should include all these
2117 The version of GCC. You can get this by running it with the
2120 Without this, we won't know whether there is any point in looking for
2121 the bug in the current version of GCC.
2124 A complete input file that will reproduce the bug. If the bug is in the
2125 C preprocessor, send a source file and any header files that it
2126 requires. If the bug is in the compiler proper (@file{cc1}), send the
2127 preprocessor output generated by adding @samp{-save-temps} to the
2128 compilation command (@pxref{Debugging Options}). When you do this, use
2129 the same @samp{-I}, @samp{-D} or @samp{-U} options that you used in
2130 actual compilation. Then send the @var{input}.i or @var{input}.ii files
2133 A single statement is not enough of an example. In order to compile it,
2134 it must be embedded in a complete file of compiler input; and the bug
2135 might depend on the details of how this is done.
2137 Without a real example one can compile, all anyone can do about your bug
2138 report is wish you luck. It would be futile to try to guess how to
2139 provoke the bug. For example, bugs in register allocation and reloading
2140 frequently depend on every little detail of the function they happen in.
2142 Even if the input file that fails comes from a GNU program, you should
2143 still send the complete test case. Don't ask the GCC maintainers to
2144 do the extra work of obtaining the program in question---they are all
2145 overworked as it is. Also, the problem may depend on what is in the
2146 header files on your system; it is unreliable for the GCC maintainers
2147 to try the problem with the header files available to them. By sending
2148 CPP output, you can eliminate this source of uncertainty and save us
2149 a certain percentage of wild goose chases.
2152 The command arguments you gave GCC to compile that example
2153 and observe the bug. For example, did you use @samp{-O}? To guarantee
2154 you won't omit something important, list all the options.
2156 If we were to try to guess the arguments, we would probably guess wrong
2157 and then we would not encounter the bug.
2160 The type of machine you are using, and the operating system name and
2164 The operands you gave to the @code{configure} command when you installed
2168 A complete list of any modifications you have made to the compiler
2169 source. (We don't promise to investigate the bug unless it happens in
2170 an unmodified compiler. But if you've made modifications and don't tell
2171 us, then you are sending us on a wild goose chase.)
2173 Be precise about these changes. A description in English is not
2174 enough---send a context diff for them.
2176 Adding files of your own (such as a machine description for a machine we
2177 don't support) is a modification of the compiler source.
2180 Details of any other deviations from the standard procedure for installing
2184 A description of what behavior you observe that you believe is
2185 incorrect. For example, ``The compiler gets a fatal signal,'' or,
2186 ``The assembler instruction at line 208 in the output is incorrect.''
2188 Of course, if the bug is that the compiler gets a fatal signal, then one
2189 can't miss it. But if the bug is incorrect output, the maintainer might
2190 not notice unless it is glaringly wrong. None of us has time to study
2191 all the assembler code from a 50-line C program just on the chance that
2192 one instruction might be wrong. We need @emph{you} to do this part!
2194 Even if the problem you experience is a fatal signal, you should still
2195 say so explicitly. Suppose something strange is going on, such as, your
2196 copy of the compiler is out of synch, or you have encountered a bug in
2197 the C library on your system. (This has happened!) Your copy might
2198 crash and the copy here would not. If you @i{said} to expect a crash,
2199 then when the compiler here fails to crash, we would know that the bug
2200 was not happening. If you don't say to expect a crash, then we would
2201 not know whether the bug was happening. We would not be able to draw
2202 any conclusion from our observations.
2204 If the problem is a diagnostic when compiling GCC with some other
2205 compiler, say whether it is a warning or an error.
2207 Often the observed symptom is incorrect output when your program is run.
2208 Sad to say, this is not enough information unless the program is short
2209 and simple. None of us has time to study a large program to figure out
2210 how it would work if compiled correctly, much less which line of it was
2211 compiled wrong. So you will have to do that. Tell us which source line
2212 it is, and what incorrect result happens when that line is executed. A
2213 person who understands the program can find this as easily as finding a
2214 bug in the program itself.
2217 If you send examples of assembler code output from GCC,
2218 please use @samp{-g} when you make them. The debugging information
2219 includes source line numbers which are essential for correlating the
2220 output with the input.
2223 If you wish to mention something in the GCC source, refer to it by
2224 context, not by line number.
2226 The line numbers in the development sources don't match those in your
2227 sources. Your line numbers would convey no useful information to the
2231 Additional information from a debugger might enable someone to find a
2232 problem on a machine which he does not have available. However, you
2233 need to think when you collect this information if you want it to have
2234 any chance of being useful.
2236 @cindex backtrace for bug reports
2237 For example, many people send just a backtrace, but that is never
2238 useful by itself. A simple backtrace with arguments conveys little
2239 about GCC because the compiler is largely data-driven; the same
2240 functions are called over and over for different RTL insns, doing
2241 different things depending on the details of the insn.
2243 Most of the arguments listed in the backtrace are useless because they
2244 are pointers to RTL list structure. The numeric values of the
2245 pointers, which the debugger prints in the backtrace, have no
2246 significance whatever; all that matters is the contents of the objects
2247 they point to (and most of the contents are other such pointers).
2249 In addition, most compiler passes consist of one or more loops that
2250 scan the RTL insn sequence. The most vital piece of information about
2251 such a loop---which insn it has reached---is usually in a local variable,
2255 What you need to provide in addition to a backtrace are the values of
2256 the local variables for several stack frames up. When a local
2257 variable or an argument is an RTX, first print its value and then use
2258 the GDB command @code{pr} to print the RTL expression that it points
2259 to. (If GDB doesn't run on your machine, use your debugger to call
2260 the function @code{debug_rtx} with the RTX as an argument.) In
2261 general, whenever a variable is a pointer, its value is no use
2262 without the data it points to.
2265 Here are some things that are not necessary:
2269 A description of the envelope of the bug.
2271 Often people who encounter a bug spend a lot of time investigating
2272 which changes to the input file will make the bug go away and which
2273 changes will not affect it.
2275 This is often time consuming and not very useful, because the way we
2276 will find the bug is by running a single example under the debugger with
2277 breakpoints, not by pure deduction from a series of examples. You might
2278 as well save your time for something else.
2280 Of course, if you can find a simpler example to report @emph{instead} of
2281 the original one, that is a convenience. Errors in the output will be
2282 easier to spot, running under the debugger will take less time, etc.
2283 Most GCC bugs involve just one function, so the most straightforward
2284 way to simplify an example is to delete all the function definitions
2285 except the one where the bug occurs. Those earlier in the file may be
2286 replaced by external declarations if the crucial function depends on
2287 them. (Exception: inline functions may affect compilation of functions
2288 defined later in the file.)
2290 However, simplification is not vital; if you don't want to do this,
2291 report the bug anyway and send the entire test case you used.
2294 In particular, some people insert conditionals @samp{#ifdef BUG} around
2295 a statement which, if removed, makes the bug not happen. These are just
2296 clutter; we won't pay any attention to them anyway. Besides, you should
2297 send us cpp output, and that can't have conditionals.
2300 A patch for the bug.
2302 A patch for the bug is useful if it is a good one. But don't omit the
2303 necessary information, such as the test case, on the assumption that a
2304 patch is all we need. We might see problems with your patch and decide
2305 to fix the problem another way, or we might not understand it at all.
2307 Sometimes with a program as complicated as GCC it is very hard to
2308 construct an example that will make the program follow a certain path
2309 through the code. If you don't send the example, we won't be able to
2310 construct one, so we won't be able to verify that the bug is fixed.
2312 And if we can't understand what bug you are trying to fix, or why your
2313 patch should be an improvement, we won't install it. A test case will
2314 help us to understand.
2316 @xref{Sending Patches}, for guidelines on how to make it easy for us to
2317 understand and install your patches.
2320 A guess about what the bug is or what it depends on.
2322 Such guesses are usually wrong. Even I can't guess right about such
2323 things without first using the debugger to find the facts.
2328 We have no way of examining a core dump for your type of machine
2329 unless we have an identical system---and if we do have one,
2330 we should be able to reproduce the crash ourselves.
2333 @node gccbug,Sending Patches, Bug Reporting, Bugs
2334 @section The gccbug script
2335 @cindex gccbug script
2337 To simplify creation of bug reports, and to allow better tracking of
2338 reports, we use the GNATS bug tracking system. Part of that system is
2339 the @code{gccbug} script. This is a Unix shell script, so you need a
2340 shell to run it. It is normally installed in the same directory where
2341 @code{gcc} is installed.
2343 The gccbug script is derived from send-pr, @pxref{using
2344 send-pr,,Creating new Problem Reports,send-pr,Reporting Problems}. When
2345 invoked, it starts a text editor so you can fill out the various fields
2346 of the report. When the you quit the editor, the report is automatically
2347 send to the bug reporting address.
2349 A number of fields in this bug report form are specific to GCC, and are
2350 explained at @uref{http://gcc.gnu.org/gnats.html}.
2352 @node Sending Patches,, gccbug, Bugs
2353 @section Sending Patches for GCC
2355 If you would like to write bug fixes or improvements for the GNU C
2356 compiler, that is very helpful. Send suggested fixes to the patches
2357 mailing list, @email{gcc-patches@@gcc.gnu.org}.
2359 Please follow these guidelines so we can study your patches efficiently.
2360 If you don't follow these guidelines, your information might still be
2361 useful, but using it will take extra work. Maintaining GNU C is a lot
2362 of work in the best of circumstances, and we can't keep up unless you do
2367 Send an explanation with your changes of what problem they fix or what
2368 improvement they bring about. For a bug fix, just include a copy of the
2369 bug report, and explain why the change fixes the bug.
2371 (Referring to a bug report is not as good as including it, because then
2372 we will have to look it up, and we have probably already deleted it if
2373 we've already fixed the bug.)
2376 Always include a proper bug report for the problem you think you have
2377 fixed. We need to convince ourselves that the change is right before
2378 installing it. Even if it is right, we might have trouble judging it if
2379 we don't have a way to reproduce the problem.
2382 Include all the comments that are appropriate to help people reading the
2383 source in the future understand why this change was needed.
2386 Don't mix together changes made for different reasons.
2387 Send them @emph{individually}.
2389 If you make two changes for separate reasons, then we might not want to
2390 install them both. We might want to install just one. If you send them
2391 all jumbled together in a single set of diffs, we have to do extra work
2392 to disentangle them---to figure out which parts of the change serve
2393 which purpose. If we don't have time for this, we might have to ignore
2394 your changes entirely.
2396 If you send each change as soon as you have written it, with its own
2397 explanation, then the two changes never get tangled up, and we can
2398 consider each one properly without any extra work to disentangle them.
2400 Ideally, each change you send should be impossible to subdivide into
2401 parts that we might want to consider separately, because each of its
2402 parts gets its motivation from the other parts.
2405 Send each change as soon as that change is finished. Sometimes people
2406 think they are helping us by accumulating many changes to send them all
2407 together. As explained above, this is absolutely the worst thing you
2410 Since you should send each change separately, you might as well send it
2411 right away. That gives us the option of installing it immediately if it
2415 Use @samp{diff -c} to make your diffs. Diffs without context are hard
2416 for us to install reliably. More than that, they make it hard for us to
2417 study the diffs to decide whether we want to install them. Unidiff
2418 format is better than contextless diffs, but not as easy to read as
2421 If you have GNU diff, use @samp{diff -cp}, which shows the name of the
2422 function that each change occurs in.
2425 Write the change log entries for your changes. We get lots of changes,
2426 and we don't have time to do all the change log writing ourselves.
2428 Read the @file{ChangeLog} file to see what sorts of information to put
2429 in, and to learn the style that we use. The purpose of the change log
2430 is to show people where to find what was changed. So you need to be
2431 specific about what functions you changed; in large functions, it's
2432 often helpful to indicate where within the function the change was.
2434 On the other hand, once you have shown people where to find the change,
2435 you need not explain its purpose. Thus, if you add a new function, all
2436 you need to say about it is that it is new. If you feel that the
2437 purpose needs explaining, it probably does---but the explanation will be
2438 much more useful if you put it in comments in the code.
2440 If you would like your name to appear in the header line for who made
2441 the change, send us the header line.
2444 When you write the fix, keep in mind that we can't install a change that
2445 would break other systems.
2447 People often suggest fixing a problem by changing machine-independent
2448 files such as @file{toplev.c} to do something special that a particular
2449 system needs. Sometimes it is totally obvious that such changes would
2450 break GCC for almost all users. We can't possibly make a change like
2451 that. At best it might tell us how to write another patch that would
2452 solve the problem acceptably.
2454 Sometimes people send fixes that @emph{might} be an improvement in
2455 general---but it is hard to be sure of this. It's hard to install
2456 such changes because we have to study them very carefully. Of course,
2457 a good explanation of the reasoning by which you concluded the change
2458 was correct can help convince us.
2460 The safest changes are changes to the configuration files for a
2461 particular machine. These are safe because they can't create new bugs
2464 Please help us keep up with the workload by designing the patch in a
2465 form that is good to install.
2469 @chapter How To Get Help with GCC
2471 If you need help installing, using or changing GCC, there are two
2476 Send a message to a suitable network mailing list. First try
2477 @email{gcc-help@@gcc.gnu.org} (for help installing or using GCC), and if
2478 that brings no response, try @email{gcc@@gcc.gnu.org}. For help
2479 changing GCC, ask @email{gcc@@gcc.gnu.org}. If you think you have found
2480 a bug in GCC, please report it following the instructions at
2481 @pxref{Bug Reporting}.
2484 Look in the service directory for someone who might help you for a fee.
2485 The service directory is found at
2486 @uref{http://www.gnu.org/prep/service.html}.
2489 @c For further information, see
2490 @c @uref{http://gcc.gnu.org/cgi-bin/fom.cgi?file=12}.
2491 @c FIXME: this URL may be too volatile, this FAQ entry needs to move to
2492 @c the regular web pages before we can uncomment the reference.
2495 @chapter Contributing to GCC Development
2497 If you would like to help pretest GCC releases to assure they work well,
2498 our current development sources are available by CVS (see
2499 @uref{http://gcc.gnu.org/cvs.html}). Source and binary snapshots are
2500 also available for FTP; see @uref{http://gcc.gnu.org/snapshots.html}.
2502 If you would like to work on improvements to GCC, please read
2503 @uref{http://gcc.gnu.org/contribute.html} and
2504 @uref{http://gcc.gnu.org/contributewhy.html} for information on how to
2505 make useful contributions and avoid duplication of effort. Suggested
2506 projects are listed at @uref{http://gcc.gnu.org/projects/}.
2509 @chapter Using GCC on VMS
2511 @c prevent bad page break with this line
2512 Here is how to use GCC on VMS.
2515 * Include Files and VMS:: Where the preprocessor looks for the include files.
2516 * Global Declarations:: How to do globaldef, globalref and globalvalue with
2518 * VMS Misc:: Misc information.
2521 @node Include Files and VMS
2522 @section Include Files and VMS
2524 @cindex include files and VMS
2525 @cindex VMS and include files
2526 @cindex header files and VMS
2527 Due to the differences between the filesystems of Unix and VMS, GCC
2528 attempts to translate file names in @samp{#include} into names that VMS
2529 will understand. The basic strategy is to prepend a prefix to the
2530 specification of the include file, convert the whole filename to a VMS
2531 filename, and then try to open the file. GCC tries various prefixes
2532 one by one until one of them succeeds:
2536 The first prefix is the @samp{GNU_CC_INCLUDE:} logical name: this is
2537 where GNU C header files are traditionally stored. If you wish to store
2538 header files in non-standard locations, then you can assign the logical
2539 @samp{GNU_CC_INCLUDE} to be a search list, where each element of the
2540 list is suitable for use with a rooted logical.
2543 The next prefix tried is @samp{SYS$SYSROOT:[SYSLIB.]}. This is where
2544 VAX-C header files are traditionally stored.
2547 If the include file specification by itself is a valid VMS filename, the
2548 preprocessor then uses this name with no prefix in an attempt to open
2552 If the file specification is not a valid VMS filename (i.e. does not
2553 contain a device or a directory specifier, and contains a @samp{/}
2554 character), the preprocessor tries to convert it from Unix syntax to
2557 Conversion works like this: the first directory name becomes a device,
2558 and the rest of the directories are converted into VMS-format directory
2559 names. For example, the name @file{X11/foobar.h} is
2560 translated to @file{X11:[000000]foobar.h} or @file{X11:foobar.h},
2561 whichever one can be opened. This strategy allows you to assign a
2562 logical name to point to the actual location of the header files.
2565 If none of these strategies succeeds, the @samp{#include} fails.
2568 Include directives of the form:
2575 are a common source of incompatibility between VAX-C and GCC. VAX-C
2576 treats this much like a standard @code{#include <foobar.h>} directive.
2577 That is incompatible with the ISO C behavior implemented by GCC: to
2578 expand the name @code{foobar} as a macro. Macro expansion should
2579 eventually yield one of the two standard formats for @code{#include}:
2582 #include "@var{file}"
2583 #include <@var{file}>
2586 If you have this problem, the best solution is to modify the source to
2587 convert the @code{#include} directives to one of the two standard forms.
2588 That will work with either compiler. If you want a quick and dirty fix,
2589 define the file names as macros with the proper expansion, like this:
2592 #define stdio <stdio.h>
2596 This will work, as long as the name doesn't conflict with anything else
2599 Another source of incompatibility is that VAX-C assumes that:
2606 is actually asking for the file @file{foobar.h}. GCC does not
2607 make this assumption, and instead takes what you ask for literally;
2608 it tries to read the file @file{foobar}. The best way to avoid this
2609 problem is to always specify the desired file extension in your include
2612 GCC for VMS is distributed with a set of include files that is
2613 sufficient to compile most general purpose programs. Even though the
2614 GCC distribution does not contain header files to define constants
2615 and structures for some VMS system-specific functions, there is no
2616 reason why you cannot use GCC with any of these functions. You first
2617 may have to generate or create header files, either by using the public
2618 domain utility @code{UNSDL} (which can be found on a DECUS tape), or by
2619 extracting the relevant modules from one of the system macro libraries,
2620 and using an editor to construct a C header file.
2622 A @code{#include} file name cannot contain a DECNET node name. The
2623 preprocessor reports an I/O error if you attempt to use a node name,
2624 whether explicitly, or implicitly via a logical name.
2626 @node Global Declarations
2627 @section Global Declarations and VMS
2631 @findex GLOBALVALUEDEF
2632 @findex GLOBALVALUEREF
2633 GCC does not provide the @code{globalref}, @code{globaldef} and
2634 @code{globalvalue} keywords of VAX-C. You can get the same effect with
2635 an obscure feature of GAS, the GNU assembler. (This requires GAS
2636 version 1.39 or later.) The following macros allow you to use this
2637 feature in a fairly natural way:
2641 #define GLOBALREF(TYPE,NAME) \
2643 asm ("_$$PsectAttributes_GLOBALSYMBOL$$" #NAME)
2644 #define GLOBALDEF(TYPE,NAME,VALUE) \
2646 asm ("_$$PsectAttributes_GLOBALSYMBOL$$" #NAME) \
2648 #define GLOBALVALUEREF(TYPE,NAME) \
2649 const TYPE NAME[1] \
2650 asm ("_$$PsectAttributes_GLOBALVALUE$$" #NAME)
2651 #define GLOBALVALUEDEF(TYPE,NAME,VALUE) \
2652 const TYPE NAME[1] \
2653 asm ("_$$PsectAttributes_GLOBALVALUE$$" #NAME) \
2656 #define GLOBALREF(TYPE,NAME) \
2658 #define GLOBALDEF(TYPE,NAME,VALUE) \
2659 globaldef TYPE NAME = VALUE
2660 #define GLOBALVALUEDEF(TYPE,NAME,VALUE) \
2661 globalvalue TYPE NAME = VALUE
2662 #define GLOBALVALUEREF(TYPE,NAME) \
2663 globalvalue TYPE NAME
2668 (The @code{_$$PsectAttributes_GLOBALSYMBOL} prefix at the start of the
2669 name is removed by the assembler, after it has modified the attributes
2670 of the symbol). These macros are provided in the VMS binaries
2671 distribution in a header file @file{GNU_HACKS.H}. An example of the
2675 GLOBALREF (int, ijk);
2676 GLOBALDEF (int, jkl, 0);
2679 The macros @code{GLOBALREF} and @code{GLOBALDEF} cannot be used
2680 straightforwardly for arrays, since there is no way to insert the array
2681 dimension into the declaration at the right place. However, you can
2682 declare an array with these macros if you first define a typedef for the
2683 array type, like this:
2686 typedef int intvector[10];
2687 GLOBALREF (intvector, foo);
2690 Array and structure initializers will also break the macros; you can
2691 define the initializer to be a macro of its own, or you can expand the
2692 @code{GLOBALDEF} macro by hand. You may find a case where you wish to
2693 use the @code{GLOBALDEF} macro with a large array, but you are not
2694 interested in explicitly initializing each element of the array. In
2695 such cases you can use an initializer like: @code{@{0,@}}, which will
2696 initialize the entire array to @code{0}.
2698 A shortcoming of this implementation is that a variable declared with
2699 @code{GLOBALVALUEREF} or @code{GLOBALVALUEDEF} is always an array. For
2700 example, the declaration:
2703 GLOBALVALUEREF(int, ijk);
2707 declares the variable @code{ijk} as an array of type @code{int [1]}.
2708 This is done because a globalvalue is actually a constant; its ``value''
2709 is what the linker would normally consider an address. That is not how
2710 an integer value works in C, but it is how an array works. So treating
2711 the symbol as an array name gives consistent results---with the
2712 exception that the value seems to have the wrong type. @strong{Don't
2713 try to access an element of the array.} It doesn't have any elements.
2714 The array ``address'' may not be the address of actual storage.
2716 The fact that the symbol is an array may lead to warnings where the
2717 variable is used. Insert type casts to avoid the warnings. Here is an
2718 example; it takes advantage of the ISO C feature allowing macros that
2719 expand to use the same name as the macro itself.
2722 GLOBALVALUEREF (int, ss$_normal);
2723 GLOBALVALUEDEF (int, xyzzy,123);
2725 #define ss$_normal ((int) ss$_normal)
2726 #define xyzzy ((int) xyzzy)
2730 Don't use @code{globaldef} or @code{globalref} with a variable whose
2731 type is an enumeration type; this is not implemented. Instead, make the
2732 variable an integer, and use a @code{globalvaluedef} for each of the
2733 enumeration values. An example of this would be:
2737 GLOBALDEF (int, color, 0);
2738 GLOBALVALUEDEF (int, RED, 0);
2739 GLOBALVALUEDEF (int, BLUE, 1);
2740 GLOBALVALUEDEF (int, GREEN, 3);
2742 enum globaldef color @{RED, BLUE, GREEN = 3@};
2747 @section Other VMS Issues
2749 @cindex exit status and VMS
2750 @cindex return value of @code{main}
2751 @cindex @code{main} and the exit status
2752 GCC automatically arranges for @code{main} to return 1 by default if
2753 you fail to specify an explicit return value. This will be interpreted
2754 by VMS as a status code indicating a normal successful completion.
2755 Version 1 of GCC did not provide this default.
2757 GCC on VMS works only with the GNU assembler, GAS. You need version
2758 1.37 or later of GAS in order to produce value debugging information for
2759 the VMS debugger. Use the ordinary VMS linker with the object files
2762 @cindex shared VMS run time system
2763 @cindex @file{VAXCRTL}
2764 Under previous versions of GCC, the generated code would occasionally
2765 give strange results when linked to the sharable @file{VAXCRTL} library.
2766 Now this should work.
2768 A caveat for use of @code{const} global variables: the @code{const}
2769 modifier must be specified in every external declaration of the variable
2770 in all of the source files that use that variable. Otherwise the linker
2771 will issue warnings about conflicting attributes for the variable. Your
2772 program will still work despite the warnings, but the variable will be
2773 placed in writable storage.
2775 @cindex name augmentation
2776 @cindex case sensitivity and VMS
2777 @cindex VMS and case sensitivity
2778 Although the VMS linker does distinguish between upper and lower case
2779 letters in global symbols, most VMS compilers convert all such symbols
2780 into upper case and most run-time library routines also have upper case
2781 names. To be able to reliably call such routines, GCC (by means of
2782 the assembler GAS) converts global symbols into upper case like other
2783 VMS compilers. However, since the usual practice in C is to distinguish
2784 case, GCC (via GAS) tries to preserve usual C behavior by augmenting
2785 each name that is not all lower case. This means truncating the name
2786 to at most 23 characters and then adding more characters at the end
2787 which encode the case pattern of those 23. Names which contain at
2788 least one dollar sign are an exception; they are converted directly into
2789 upper case without augmentation.
2791 Name augmentation yields bad results for programs that use precompiled
2792 libraries (such as Xlib) which were generated by another compiler. You
2793 can use the compiler option @samp{/NOCASE_HACK} to inhibit augmentation;
2794 it makes external C functions and variables case-independent as is usual
2795 on VMS. Alternatively, you could write all references to the functions
2796 and variables in such libraries using lower case; this will work on VMS,
2797 but is not portable to other systems. The compiler option @samp{/NAMES}
2798 also provides control over global name handling.
2800 Function and variable names are handled somewhat differently with GNU
2801 C++. The GNU C++ compiler performs @dfn{name mangling} on function
2802 names, which means that it adds information to the function name to
2803 describe the data types of the arguments that the function takes. One
2804 result of this is that the name of a function can become very long.
2805 Since the VMS linker only recognizes the first 31 characters in a name,
2806 special action is taken to ensure that each function and variable has a
2807 unique name that can be represented in 31 characters.
2809 If the name (plus a name augmentation, if required) is less than 32
2810 characters in length, then no special action is performed. If the name
2811 is longer than 31 characters, the assembler (GAS) will generate a
2812 hash string based upon the function name, truncate the function name to
2813 23 characters, and append the hash string to the truncated name. If the
2814 @samp{/VERBOSE} compiler option is used, the assembler will print both
2815 the full and truncated names of each symbol that is truncated.
2817 The @samp{/NOCASE_HACK} compiler option should not be used when you are
2818 compiling programs that use libg++. libg++ has several instances of
2819 objects (i.e. @code{Filebuf} and @code{filebuf}) which become
2820 indistinguishable in a case-insensitive environment. This leads to
2821 cases where you need to inhibit augmentation selectively (if you were
2822 using libg++ and Xlib in the same program, for example). There is no
2823 special feature for doing this, but you can get the result by defining a
2824 macro for each mixed case symbol for which you wish to inhibit
2825 augmentation. The macro should expand into the lower case equivalent of
2826 itself. For example:
2829 #define StuDlyCapS studlycaps
2832 These macro definitions can be placed in a header file to minimize the
2833 number of changes to your source code.
2836 @chapter Makefile Targets
2837 @cindex makefile targets
2838 @cindex targets, makefile
2842 This is the default target. Depending on what your build/host/target
2843 configuration is, it coordinates all the things that need to be built.
2846 Produce info-formatted documentation. Also, @code{make dvi} is
2847 available for DVI-formatted documentation, and @code{make
2848 generated-manpages} to generate man pages.
2851 Delete the files made while building the compiler.
2854 That, and all the other files built by @code{make all}.
2857 That, and all the files created by @code{configure}.
2860 That, and any temporary or intermediate files, like emacs backup files.
2862 @item maintainer-clean
2863 Distclean plus any file that can be generated from other files. Note
2864 that additional tools may be required beyond what is normally needed to
2871 Deletes installed files.
2874 Run the testsuite. This creates a @file{testsuite} subdirectory that
2875 has various @file{.sum} and @file{.log} files containing the results of
2876 the testing. You can run subsets with, for example, @code{make check-gcc}.
2877 You can specify specific tests by setting RUNTESTFLAGS to be the name
2878 of the @file{.exp} file, optionally followed by (for some tests) an equals
2879 and a file wildcard, like:
2882 make check-gcc RUNTESTFLAGS="execute.exp=19980413-*"
2885 Note that running the testsuite may require additional tools be
2886 installed, such as TCL or dejagnu.
2889 Builds gcc three times - once with the native compiler, once with the
2890 native-built compiler it just built, and once with the compiler it built
2891 the second time. In theory, the last two should produce the same
2892 results, which @code{make compare} can check. Each step of this process
2893 is called a "stage", and the results of each stage N (N=1..3) are copied
2894 to a subdirectory @file{stageN/}.
2896 @item bootstrap-lean
2897 Like @code{bootstrap}, except that the various stages are removed once
2898 they're no longer needed. This saves disk space.
2901 Once bootstrapped, this incrementally rebuilds each of the three stages,
2902 one at a time. It does this by "bubbling" the stages up from their
2903 stubdirectories, rebuilding them, and copying them back to their
2904 subdirectories. This will allow you to, for example, quickly rebuild a
2905 bootstrapped compiler after changing the sources, without having to do a
2909 Rebuilds the most recently built stage. Since each stage requires
2910 special invocation, using this target means you don't have to keep track
2911 of which stage you're on or what invocation that stage needs.
2914 Removed everything (@code{make clean}) and rebuilds (@code{make bootstrap}).
2916 @item stageN (N=1..4)
2917 For each stage, moves the appropriate files to the stageN subdirectory.
2919 @item unstageN (N=1..4)
2920 Undoes the corresponding @code{stageN}.
2922 @item restageN (N=1..4)
2923 Undoes the corresponding @code{stageN} and rebuilds it with the
2927 Compares the results of stages 2 and 3. This ensures that the compiler
2928 is running properly, since it should produce the same object files
2929 regardless of how it itself was compiled.
2937 @chapter GCC and Portability
2939 @cindex GCC and portability
2941 The main goal of GCC was to make a good, fast compiler for machines in
2942 the class that the GNU system aims to run on: 32-bit machines that address
2943 8-bit bytes and have several general registers. Elegance, theoretical
2944 power and simplicity are only secondary.
2946 GCC gets most of the information about the target machine from a machine
2947 description which gives an algebraic formula for each of the machine's
2948 instructions. This is a very clean way to describe the target. But when
2949 the compiler needs information that is difficult to express in this
2950 fashion, I have not hesitated to define an ad-hoc parameter to the machine
2951 description. The purpose of portability is to reduce the total work needed
2952 on the compiler; it was not of interest for its own sake.
2955 @cindex autoincrement addressing, availability
2957 GCC does not contain machine dependent code, but it does contain code
2958 that depends on machine parameters such as endianness (whether the most
2959 significant byte has the highest or lowest address of the bytes in a word)
2960 and the availability of autoincrement addressing. In the RTL-generation
2961 pass, it is often necessary to have multiple strategies for generating code
2962 for a particular kind of syntax tree, strategies that are usable for different
2963 combinations of parameters. Often I have not tried to address all possible
2964 cases, but only the common ones or only the ones that I have encountered.
2965 As a result, a new target may require additional strategies. You will know
2966 if this happens because the compiler will call @code{abort}. Fortunately,
2967 the new strategies can be added in a machine-independent fashion, and will
2968 affect only the target machines that need them.
2973 @chapter Interfacing to GCC Output
2974 @cindex interfacing to GCC output
2975 @cindex run-time conventions
2976 @cindex function call conventions
2977 @cindex conventions, run-time
2979 GCC is normally configured to use the same function calling convention
2980 normally in use on the target system. This is done with the
2981 machine-description macros described (@pxref{Target Macros}).
2983 @cindex unions, returning
2984 @cindex structures, returning
2985 @cindex returning structures and unions
2986 However, returning of structure and union values is done differently on
2987 some target machines. As a result, functions compiled with PCC
2988 returning such types cannot be called from code compiled with GCC,
2989 and vice versa. This does not cause trouble often because few Unix
2990 library routines return structures or unions.
2992 GCC code returns structures and unions that are 1, 2, 4 or 8 bytes
2993 long in the same registers used for @code{int} or @code{double} return
2994 values. (GCC typically allocates variables of such types in
2995 registers also.) Structures and unions of other sizes are returned by
2996 storing them into an address passed by the caller (usually in a
2997 register). The machine-description macros @code{STRUCT_VALUE} and
2998 @code{STRUCT_INCOMING_VALUE} tell GCC where to pass this address.
3000 By contrast, PCC on most target machines returns structures and unions
3001 of any size by copying the data into an area of static storage, and then
3002 returning the address of that storage as if it were a pointer value.
3003 The caller must copy the data from that memory area to the place where
3004 the value is wanted. This is slower than the method used by GCC, and
3005 fails to be reentrant.
3007 On some target machines, such as RISC machines and the 80386, the
3008 standard system convention is to pass to the subroutine the address of
3009 where to return the value. On these machines, GCC has been
3010 configured to be compatible with the standard compiler, when this method
3011 is used. It may not be compatible for structures of 1, 2, 4 or 8 bytes.
3013 @cindex argument passing
3014 @cindex passing arguments
3015 GCC uses the system's standard convention for passing arguments. On
3016 some machines, the first few arguments are passed in registers; in
3017 others, all are passed on the stack. It would be possible to use
3018 registers for argument passing on any machine, and this would probably
3019 result in a significant speedup. But the result would be complete
3020 incompatibility with code that follows the standard convention. So this
3021 change is practical only if you are switching to GCC as the sole C
3022 compiler for the system. We may implement register argument passing on
3023 certain machines once we have a complete GNU system so that we can
3024 compile the libraries with GCC.
3026 On some machines (particularly the Sparc), certain types of arguments
3027 are passed ``by invisible reference''. This means that the value is
3028 stored in memory, and the address of the memory location is passed to
3031 @cindex @code{longjmp} and automatic variables
3032 If you use @code{longjmp}, beware of automatic variables. ISO C says that
3033 automatic variables that are not declared @code{volatile} have undefined
3034 values after a @code{longjmp}. And this is all GCC promises to do,
3035 because it is very difficult to restore register variables correctly, and
3036 one of GCC's features is that it can put variables in registers without
3039 If you want a variable to be unaltered by @code{longjmp}, and you don't
3040 want to write @code{volatile} because old C compilers don't accept it,
3041 just take the address of the variable. If a variable's address is ever
3042 taken, even if just to compute it and ignore it, then the variable cannot
3053 @cindex arithmetic libraries
3054 @cindex math libraries
3055 Code compiled with GCC may call certain library routines. Most of
3056 them handle arithmetic for which there are no instructions. This
3057 includes multiply and divide on some machines, and floating point
3058 operations on any machine for which floating point support is disabled
3059 with @samp{-msoft-float}. Some standard parts of the C library, such as
3060 @code{bcopy} or @code{memcpy}, are also called automatically. The usual
3061 function call interface is used for calling the library routines.
3063 Some of these routines can be defined in mostly machine-independent C;
3064 they appear in @file{libgcc2.c}. Others must be hand-written in
3065 assembly language for each processor. Wherever they are defined, they
3066 are compiled into the support library, @file{libgcc.a}, which is
3067 automatically searched when you link programs with GCC.
3072 @chapter Passes and Files of the Compiler
3073 @cindex passes and files of the compiler
3074 @cindex files and passes of the compiler
3075 @cindex compiler passes and files
3077 @cindex top level of compiler
3078 The overall control structure of the compiler is in @file{toplev.c}. This
3079 file is responsible for initialization, decoding arguments, opening and
3080 closing files, and sequencing the passes.
3082 @cindex parsing pass
3083 The parsing pass is invoked only once, to parse the entire input. The RTL
3084 intermediate code for a function is generated as the function is parsed, a
3085 statement at a time. Each statement is read in as a syntax tree and then
3086 converted to RTL; then the storage for the tree for the statement is
3087 reclaimed. Storage for types (and the expressions for their sizes),
3088 declarations, and a representation of the binding contours and how they nest,
3089 remain until the function is finished being compiled; these are all needed
3090 to output the debugging information.
3092 @findex rest_of_compilation
3093 @findex rest_of_decl_compilation
3094 Each time the parsing pass reads a complete function definition or
3095 top-level declaration, it calls either the function
3096 @code{rest_of_compilation}, or the function
3097 @code{rest_of_decl_compilation} in @file{toplev.c}, which are
3098 responsible for all further processing necessary, ending with output of
3099 the assembler language. All other compiler passes run, in sequence,
3100 within @code{rest_of_compilation}. When that function returns from
3101 compiling a function definition, the storage used for that function
3102 definition's compilation is entirely freed, unless it is an inline
3105 (@pxref{Inline,,An Inline Function is As Fast As a Macro}).
3108 (@pxref{Inline,,An Inline Function is As Fast As a Macro,gcc.texi,Using GCC}).
3111 Here is a list of all the passes of the compiler and their source files.
3112 Also included is a description of where debugging dumps can be requested
3113 with @samp{-d} options.
3117 Parsing. This pass reads the entire text of a function definition,
3118 constructing partial syntax trees. This and RTL generation are no longer
3119 truly separate passes (formerly they were), but it is easier to think
3120 of them as separate.
3122 The tree representation does not entirely follow C syntax, because it is
3123 intended to support other languages as well.
3125 Language-specific data type analysis is also done in this pass, and every
3126 tree node that represents an expression has a data type attached.
3127 Variables are represented as declaration nodes.
3129 @cindex constant folding
3130 @cindex arithmetic simplifications
3131 @cindex simplifications, arithmetic
3132 Constant folding and some arithmetic simplifications are also done
3135 The language-independent source files for parsing are
3136 @file{stor-layout.c}, @file{fold-const.c}, and @file{tree.c}.
3137 There are also header files @file{tree.h} and @file{tree.def}
3138 which define the format of the tree representation.@refill
3140 @c Avoiding overfull is tricky here.
3141 The source files to parse C are
3145 @file{c-aux-info.c},
3148 along with header files
3152 The source files for parsing C++ are in @file{cp/}.
3153 They are @file{parse.y},
3155 @file{cvt.c}, @file{decl.c}, @file{decl2.c},
3157 @file{expr.c}, @file{init.c}, @file{lex.c},
3158 @file{method.c}, @file{ptree.c},@*
3159 @file{search.c}, @file{tree.c},
3160 @file{typeck2.c}, and
3161 @file{typeck.c}, along with header files @file{cp-tree.def},
3162 @file{cp-tree.h}, and @file{decl.h}.
3164 The special source files for parsing Objective C are in @file{objc/}.
3165 They are @file{objc-parse.y}, @file{objc-act.c}, @file{objc-tree.def}, and
3166 @file{objc-act.h}. Certain C-specific files are used for this as
3169 The file @file{c-common.c} is also used for all of the above languages.
3171 @cindex RTL generation
3173 RTL generation. This is the conversion of syntax tree into RTL code.
3174 It is actually done statement-by-statement during parsing, but for
3175 most purposes it can be thought of as a separate pass.
3177 @cindex target-parameter-dependent code
3178 This is where the bulk of target-parameter-dependent code is found,
3179 since often it is necessary for strategies to apply only when certain
3180 standard kinds of instructions are available. The purpose of named
3181 instruction patterns is to provide this information to the RTL
3184 @cindex tail recursion optimization
3185 Optimization is done in this pass for @code{if}-conditions that are
3186 comparisons, boolean operations or conditional expressions. Tail
3187 recursion is detected at this time also. Decisions are made about how
3188 best to arrange loops and how to output @code{switch} statements.
3190 @c Avoiding overfull is tricky here.
3191 The source files for RTL generation include
3199 and @file{emit-rtl.c}.
3201 @file{insn-emit.c}, generated from the machine description by the
3202 program @code{genemit}, is used in this pass. The header file
3203 @file{expr.h} is used for communication within this pass.@refill
3207 The header files @file{insn-flags.h} and @file{insn-codes.h},
3208 generated from the machine description by the programs @code{genflags}
3209 and @code{gencodes}, tell this pass which standard names are available
3210 for use and which patterns correspond to them.@refill
3212 Aside from debugging information output, none of the following passes
3213 refers to the tree structure representation of the function (only
3214 part of which is saved).
3216 @cindex inline, automatic
3217 The decision of whether the function can and should be expanded inline
3218 in its subsequent callers is made at the end of rtl generation. The
3219 function must meet certain criteria, currently related to the size of
3220 the function and the types and number of parameters it has. Note that
3221 this function may contain loops, recursive calls to itself
3222 (tail-recursive functions can be inlined!), gotos, in short, all
3223 constructs supported by GCC. The file @file{integrate.c} contains
3224 the code to save a function's rtl for later inlining and to inline that
3225 rtl when the function is called. The header file @file{integrate.h}
3226 is also used for this purpose.
3228 The option @samp{-dr} causes a debugging dump of the RTL code after
3229 this pass. This dump file's name is made by appending @samp{.rtl} to
3230 the input file name.
3232 @cindex jump optimization
3233 @cindex unreachable code
3236 Jump optimization. This pass simplifies jumps to the following
3237 instruction, jumps across jumps, and jumps to jumps. It deletes
3238 unreferenced labels and unreachable code, except that unreachable code
3239 that contains a loop is not recognized as unreachable in this pass.
3240 (Such loops are deleted later in the basic block analysis.) It also
3241 converts some code originally written with jumps into sequences of
3242 instructions that directly set values from the results of comparisons,
3243 if the machine has such instructions.
3245 Jump optimization is performed two or three times. The first time is
3246 immediately following RTL generation. The second time is after CSE,
3247 but only if CSE says repeated jump optimization is needed. The
3248 last time is right before the final pass. That time, cross-jumping
3249 and deletion of no-op move instructions are done together with the
3250 optimizations described above.
3252 The source file of this pass is @file{jump.c}.
3254 The option @samp{-dj} causes a debugging dump of the RTL code after
3255 this pass is run for the first time. This dump file's name is made by
3256 appending @samp{.jump} to the input file name.
3258 @cindex register use analysis
3260 Register scan. This pass finds the first and last use of each
3261 register, as a guide for common subexpression elimination. Its source
3262 is in @file{regclass.c}.
3264 @cindex jump threading
3266 Jump threading. This pass detects a condition jump that branches to an
3267 identical or inverse test. Such jumps can be @samp{threaded} through
3268 the second conditional test. The source code for this pass is in
3269 @file{jump.c}. This optimization is only performed if
3270 @samp{-fthread-jumps} is enabled.
3272 @cindex common subexpression elimination
3273 @cindex constant propagation
3275 Common subexpression elimination. This pass also does constant
3276 propagation. Its source file is @file{cse.c}. If constant
3277 propagation causes conditional jumps to become unconditional or to
3278 become no-ops, jump optimization is run again when CSE is finished.
3280 The option @samp{-ds} causes a debugging dump of the RTL code after
3281 this pass. This dump file's name is made by appending @samp{.cse} to
3282 the input file name.
3284 @cindex global common subexpression elimination
3285 @cindex constant propagation
3286 @cindex copy propagation
3288 Global common subexpression elimination. This pass performs GCSE
3289 using Morel-Renvoise Partial Redundancy Elimination, with the exception
3290 that it does not try to move invariants out of loops - that is left to
3291 the loop optimization pass. This pass also performs global constant
3292 and copy propagation.
3294 The source file for this pass is gcse.c.
3296 The option @samp{-dG} causes a debugging dump of the RTL code after
3297 this pass. This dump file's name is made by appending @samp{.gcse} to
3298 the input file name.
3300 @cindex loop optimization
3302 @cindex strength-reduction
3304 Loop optimization. This pass moves constant expressions out of loops,
3305 and optionally does strength-reduction and loop unrolling as well.
3306 Its source files are @file{loop.c} and @file{unroll.c}, plus the header
3307 @file{loop.h} used for communication between them. Loop unrolling uses
3308 some functions in @file{integrate.c} and the header @file{integrate.h}.
3310 The option @samp{-dL} causes a debugging dump of the RTL code after
3311 this pass. This dump file's name is made by appending @samp{.loop} to
3312 the input file name.
3315 If @samp{-frerun-cse-after-loop} was enabled, a second common
3316 subexpression elimination pass is performed after the loop optimization
3317 pass. Jump threading is also done again at this time if it was specified.
3319 The option @samp{-dt} causes a debugging dump of the RTL code after
3320 this pass. This dump file's name is made by appending @samp{.cse2} to
3321 the input file name.
3323 @cindex data flow analysis
3324 @cindex analysis, data flow
3325 @cindex basic blocks
3327 Data flow analysis (@file{flow.c}). This pass divides the program
3328 into basic blocks (and in the process deletes unreachable loops); then
3329 it computes which pseudo-registers are live at each point in the
3330 program, and makes the first instruction that uses a value point at
3331 the instruction that computed the value.
3333 @cindex autoincrement/decrement analysis
3334 This pass also deletes computations whose results are never used, and
3335 combines memory references with add or subtract instructions to make
3336 autoincrement or autodecrement addressing.
3338 The option @samp{-df} causes a debugging dump of the RTL code after
3339 this pass. This dump file's name is made by appending @samp{.flow} to
3340 the input file name. If stupid register allocation is in use, this
3341 dump file reflects the full results of such allocation.
3343 @cindex instruction combination
3345 Instruction combination (@file{combine.c}). This pass attempts to
3346 combine groups of two or three instructions that are related by data
3347 flow into single instructions. It combines the RTL expressions for
3348 the instructions by substitution, simplifies the result using algebra,
3349 and then attempts to match the result against the machine description.
3351 The option @samp{-dc} causes a debugging dump of the RTL code after
3352 this pass. This dump file's name is made by appending @samp{.combine}
3353 to the input file name.
3355 @cindex register movement
3357 Register movement (@file{regmove.c}). This pass looks for cases where
3358 matching constraints would force an instruction to need a reload, and
3359 this reload would be a register to register move. It then attempts
3360 to change the registers used by the instruction to avoid the move
3363 The option @samp{-dN} causes a debugging dump of the RTL code after
3364 this pass. This dump file's name is made by appending @samp{.regmove}
3365 to the input file name.
3367 @cindex instruction scheduling
3368 @cindex scheduling, instruction
3370 Instruction scheduling (@file{sched.c}). This pass looks for
3371 instructions whose output will not be available by the time that it is
3372 used in subsequent instructions. (Memory loads and floating point
3373 instructions often have this behavior on RISC machines). It re-orders
3374 instructions within a basic block to try to separate the definition and
3375 use of items that otherwise would cause pipeline stalls.
3377 Instruction scheduling is performed twice. The first time is immediately
3378 after instruction combination and the second is immediately after reload.
3380 The option @samp{-dS} causes a debugging dump of the RTL code after this
3381 pass is run for the first time. The dump file's name is made by
3382 appending @samp{.sched} to the input file name.
3384 @cindex register class preference pass
3386 Register class preferencing. The RTL code is scanned to find out
3387 which register class is best for each pseudo register. The source
3388 file is @file{regclass.c}.
3390 @cindex register allocation
3391 @cindex local register allocation
3393 Local register allocation (@file{local-alloc.c}). This pass allocates
3394 hard registers to pseudo registers that are used only within one basic
3395 block. Because the basic block is linear, it can use fast and
3396 powerful techniques to do a very good job.
3398 The option @samp{-dl} causes a debugging dump of the RTL code after
3399 this pass. This dump file's name is made by appending @samp{.lreg} to
3400 the input file name.
3402 @cindex global register allocation
3404 Global register allocation (@file{global.c}). This pass
3405 allocates hard registers for the remaining pseudo registers (those
3406 whose life spans are not contained in one basic block).
3410 Reloading. This pass renumbers pseudo registers with the hardware
3411 registers numbers they were allocated. Pseudo registers that did not
3412 get hard registers are replaced with stack slots. Then it finds
3413 instructions that are invalid because a value has failed to end up in
3414 a register, or has ended up in a register of the wrong kind. It fixes
3415 up these instructions by reloading the problematical values
3416 temporarily into registers. Additional instructions are generated to
3419 The reload pass also optionally eliminates the frame pointer and inserts
3420 instructions to save and restore call-clobbered registers around calls.
3422 Source files are @file{reload.c} and @file{reload1.c}, plus the header
3423 @file{reload.h} used for communication between them.
3425 The option @samp{-dg} causes a debugging dump of the RTL code after
3426 this pass. This dump file's name is made by appending @samp{.greg} to
3427 the input file name.
3429 @cindex instruction scheduling
3430 @cindex scheduling, instruction
3432 Instruction scheduling is repeated here to try to avoid pipeline stalls
3433 due to memory loads generated for spilled pseudo registers.
3435 The option @samp{-dR} causes a debugging dump of the RTL code after
3436 this pass. This dump file's name is made by appending @samp{.sched2}
3437 to the input file name.
3439 @cindex cross-jumping
3440 @cindex no-op move instructions
3442 Jump optimization is repeated, this time including cross-jumping
3443 and deletion of no-op move instructions.
3445 The option @samp{-dJ} causes a debugging dump of the RTL code after
3446 this pass. This dump file's name is made by appending @samp{.jump2}
3447 to the input file name.
3449 @cindex delayed branch scheduling
3450 @cindex scheduling, delayed branch
3452 Delayed branch scheduling. This optional pass attempts to find
3453 instructions that can go into the delay slots of other instructions,
3454 usually jumps and calls. The source file name is @file{reorg.c}.
3456 The option @samp{-dd} causes a debugging dump of the RTL code after
3457 this pass. This dump file's name is made by appending @samp{.dbr}
3458 to the input file name.
3460 @cindex branch shortening
3462 Branch shortening. On many RISC machines, branch instructions have a
3463 limited range. Thus, longer sequences of instructions must be used for
3464 long branches. In this pass, the compiler figures out what how far each
3465 instruction will be from each other instruction, and therefore whether
3466 the usual instructions, or the longer sequences, must be used for each
3469 @cindex register-to-stack conversion
3471 Conversion from usage of some hard registers to usage of a register
3472 stack may be done at this point. Currently, this is supported only
3473 for the floating-point registers of the Intel 80387 coprocessor. The
3474 source file name is @file{reg-stack.c}.
3476 The options @samp{-dk} causes a debugging dump of the RTL code after
3477 this pass. This dump file's name is made by appending @samp{.stack}
3478 to the input file name.
3481 @cindex peephole optimization
3483 Final. This pass outputs the assembler code for the function. It is
3484 also responsible for identifying spurious test and compare
3485 instructions. Machine-specific peephole optimizations are performed
3486 at the same time. The function entry and exit sequences are generated
3487 directly as assembler code in this pass; they never exist as RTL.
3489 The source files are @file{final.c} plus @file{insn-output.c}; the
3490 latter is generated automatically from the machine description by the
3491 tool @file{genoutput}. The header file @file{conditions.h} is used
3492 for communication between these files.
3494 @cindex debugging information generation
3496 Debugging information output. This is run after final because it must
3497 output the stack slot offsets for pseudo registers that did not get
3498 hard registers. Source files are @file{dbxout.c} for DBX symbol table
3499 format, @file{sdbout.c} for SDB symbol table format, and
3500 @file{dwarfout.c} for DWARF symbol table format.
3503 Some additional files are used by all or many passes:
3507 Every pass uses @file{machmode.def} and @file{machmode.h} which define
3511 Several passes use @file{real.h}, which defines the default
3512 representation of floating point constants and how to operate on them.
3515 All the passes that work with RTL use the header files @file{rtl.h}
3516 and @file{rtl.def}, and subroutines in file @file{rtl.c}. The tools
3517 @code{gen*} also use these files to read and work with the machine
3522 Several passes refer to the header file @file{insn-config.h} which
3523 contains a few parameters (C macro definitions) generated
3524 automatically from the machine description RTL by the tool
3527 @cindex instruction recognizer
3529 Several passes use the instruction recognizer, which consists of
3530 @file{recog.c} and @file{recog.h}, plus the files @file{insn-recog.c}
3531 and @file{insn-extract.c} that are generated automatically from the
3532 machine description by the tools @file{genrecog} and
3533 @file{genextract}.@refill
3536 Several passes use the header files @file{regs.h} which defines the
3537 information recorded about pseudo register usage, and @file{basic-block.h}
3538 which defines the information recorded about basic blocks.
3541 @file{hard-reg-set.h} defines the type @code{HARD_REG_SET}, a bit-vector
3542 with a bit for each hard register, and some macros to manipulate it.
3543 This type is just @code{int} if the machine has few enough hard registers;
3544 otherwise it is an array of @code{int} and some of the macros expand
3548 Several passes use instruction attributes. A definition of the
3549 attributes defined for a particular machine is in file
3550 @file{insn-attr.h}, which is generated from the machine description by
3551 the program @file{genattr}. The file @file{insn-attrtab.c} contains
3552 subroutines to obtain the attribute values for insns. It is generated
3553 from the machine description by the program @file{genattrtab}.@refill
3558 @include c-tree.texi
3566 @chapter The Configuration File
3567 @cindex configuration file
3568 @cindex @file{xm-@var{machine}.h}
3570 The configuration file @file{xm-@var{machine}.h} contains macro
3571 definitions that describe the machine and system on which the compiler
3572 is running, unlike the definitions in @file{@var{machine}.h}, which
3573 describe the machine for which the compiler is producing output. Most
3574 of the values in @file{xm-@var{machine}.h} are actually the same on all
3575 machines that GCC runs on, so large parts of all configuration files
3576 are identical. But there are some macros that vary:
3581 Define this macro if the host system is System V.
3585 Define this macro if the host system is VMS.
3587 @findex FATAL_EXIT_CODE
3588 @item FATAL_EXIT_CODE
3589 A C expression for the status code to be returned when the compiler
3590 exits after serious errors. The default is the system-provided macro
3591 @samp{EXIT_FAILURE}, or @samp{1} if the system doesn't define that
3592 macro. Define this macro only if these defaults are incorrect.
3594 @findex SUCCESS_EXIT_CODE
3595 @item SUCCESS_EXIT_CODE
3596 A C expression for the status code to be returned when the compiler
3597 exits without serious errors. (Warnings are not serious errors.) The
3598 default is the system-provided macro @samp{EXIT_SUCCESS}, or @samp{0} if
3599 the system doesn't define that macro. Define this macro only if these
3600 defaults are incorrect.
3602 @findex HOST_WORDS_BIG_ENDIAN
3603 @item HOST_WORDS_BIG_ENDIAN
3604 Defined if the host machine stores words of multi-word values in
3605 big-endian order. (GCC does not depend on the host byte ordering
3608 @findex HOST_FLOAT_WORDS_BIG_ENDIAN
3609 @item HOST_FLOAT_WORDS_BIG_ENDIAN
3610 Define this macro to be 1 if the host machine stores @code{DFmode},
3611 @code{XFmode} or @code{TFmode} floating point numbers in memory with the
3612 word containing the sign bit at the lowest address; otherwise, define it
3615 This macro need not be defined if the ordering is the same as for
3616 multi-word integers.
3618 @findex HOST_FLOAT_FORMAT
3619 @item HOST_FLOAT_FORMAT
3620 A numeric code distinguishing the floating point format for the host
3621 machine. See @code{TARGET_FLOAT_FORMAT} in @ref{Storage Layout} for the
3622 alternatives and default.
3624 @findex HOST_BITS_PER_CHAR
3625 @item HOST_BITS_PER_CHAR
3626 A C expression for the number of bits in @code{char} on the host
3629 @findex HOST_BITS_PER_SHORT
3630 @item HOST_BITS_PER_SHORT
3631 A C expression for the number of bits in @code{short} on the host
3634 @findex HOST_BITS_PER_INT
3635 @item HOST_BITS_PER_INT
3636 A C expression for the number of bits in @code{int} on the host
3639 @findex HOST_BITS_PER_LONG
3640 @item HOST_BITS_PER_LONG
3641 A C expression for the number of bits in @code{long} on the host
3644 @findex ONLY_INT_FIELDS
3645 @item ONLY_INT_FIELDS
3646 Define this macro to indicate that the host compiler only supports
3647 @code{int} bit fields, rather than other integral types, including
3648 @code{enum}, as do most C compilers.
3650 @findex OBSTACK_CHUNK_SIZE
3651 @item OBSTACK_CHUNK_SIZE
3652 A C expression for the size of ordinary obstack chunks.
3653 If you don't define this, a usually-reasonable default is used.
3655 @findex OBSTACK_CHUNK_ALLOC
3656 @item OBSTACK_CHUNK_ALLOC
3657 The function used to allocate obstack chunks.
3658 If you don't define this, @code{xmalloc} is used.
3660 @findex OBSTACK_CHUNK_FREE
3661 @item OBSTACK_CHUNK_FREE
3662 The function used to free obstack chunks.
3663 If you don't define this, @code{free} is used.
3665 @findex USE_C_ALLOCA
3667 Define this macro to indicate that the compiler is running with the
3668 @code{alloca} implemented in C. This version of @code{alloca} can be
3669 found in the file @file{alloca.c}; to use it, you must also alter the
3670 @file{Makefile} variable @code{ALLOCA}. (This is done automatically
3671 for the systems on which we know it is needed.)
3673 If you do define this macro, you should probably do it as follows:
3677 #define USE_C_ALLOCA
3679 #define alloca __builtin_alloca
3684 so that when the compiler is compiled with GCC it uses the more
3685 efficient built-in @code{alloca} function.
3687 @item FUNCTION_CONVERSION_BUG
3688 @findex FUNCTION_CONVERSION_BUG
3689 Define this macro to indicate that the host compiler does not properly
3690 handle converting a function value to a pointer-to-function when it is
3691 used in an expression.
3693 @findex MULTIBYTE_CHARS
3694 @item MULTIBYTE_CHARS
3695 Define this macro to enable support for multibyte characters in the
3696 input to GCC. This requires that the host system support the ISO C
3697 library functions for converting multibyte characters to wide
3702 Define this if your system is POSIX.1 compliant.
3704 @findex PATH_SEPARATOR
3705 @item PATH_SEPARATOR
3706 Define this macro to be a C character constant representing the
3707 character used to separate components in paths. The default value is
3710 @findex DIR_SEPARATOR
3712 If your system uses some character other than slash to separate
3713 directory names within a file specification, define this macro to be a C
3714 character constant specifying that character. When GCC displays file
3715 names, the character you specify will be used. GCC will test for
3716 both slash and the character you specify when parsing filenames.
3718 @findex TARGET_OBJECT_SUFFIX
3719 @item TARGET_OBJECT_SUFFIX
3720 Define this macro to be a C string representing the suffix for object
3721 files on your target machine. If you do not define this macro, GCC will
3722 use @samp{.o} as the suffix for object files.
3724 @findex TARGET_EXECUTABLE_SUFFIX
3725 @item TARGET_EXECUTABLE_SUFFIX
3726 Define this macro to be a C string representing the suffix to be
3727 automatically added to executable files on your target machine. If you
3728 do not define this macro, GCC will use the null string as the suffix for
3731 @findex HOST_OBJECT_SUFFIX
3732 @item HOST_OBJECT_SUFFIX
3733 Define this macro to be a C string representing the suffix for object
3734 files on your host machine (@samp{xm-*.h}). If you do not define this
3735 macro, GCC will use @samp{.o} as the suffix for object files.
3737 @findex HOST_EXECUTABLE_SUFFIX
3738 @item HOST_EXECUTABLE_SUFFIX
3739 Define this macro to be a C string representing the suffix for
3740 executable files on your host machine (@samp{xm-*.h}). If you do not
3741 define this macro, GCC will use the null string as the suffix for
3744 @findex HOST_BIT_BUCKET
3745 @item HOST_BIT_BUCKET
3746 The name of a file or file-like object on the host system which acts as
3747 a ``bit bucket''. If you do not define this macro, GCC will use
3748 @samp{/dev/null} as the bit bucket. If the target does not support a
3749 bit bucket, this should be defined to the null string, or some other
3750 illegal filename. If the bit bucket is not writable, GCC will use a
3751 temporary file instead.
3753 @findex COLLECT_EXPORT_LIST
3754 @item COLLECT_EXPORT_LIST
3755 If defined, @code{collect2} will scan the individual object files
3756 specified on its command line and create an export list for the linker.
3757 Define this macro for systems like AIX, where the linker discards
3758 object files that are not referenced from @code{main} and uses export
3761 @findex COLLECT2_HOST_INITIALIZATION
3762 @item COLLECT2_HOST_INITIALIZATION
3763 If defined, a C statement (sans semicolon) that performs host-dependent
3764 initialization when @code{collect2} is being initialized.
3766 @findex GCC_DRIVER_HOST_INITIALIZATION
3767 @item GCC_DRIVER_HOST_INITIALIZATION
3768 If defined, a C statement (sans semicolon) that performs host-dependent
3769 initialization when a compilation driver is being initialized.
3771 @findex UPDATE_PATH_HOST_CANONICALIZE
3772 @item UPDATE_PATH_HOST_CANONICALIZE (@var{path}, @var{key})
3773 If defined, a C statement (sans semicolon) that performs host-dependent
3774 canonicalization when a path used in a compilation driver or preprocessor is
3775 canonicalized. @var{path} is the path to be canonicalized, and @var{key} is
3776 a translation prefix when its value isn't @code{NULL}. If the C statement
3777 does canonicalize @var{path}, the new path should be returned.
3782 In addition, configuration files for system V define @code{bcopy},
3783 @code{bzero} and @code{bcmp} as aliases. Some files define @code{alloca}
3784 as a macro when compiled with GCC, in order to take advantage of the
3785 benefit of GCC's built-in @code{alloca}.
3788 @chapter Makefile Fragments
3789 @cindex makefile fragment
3791 When you configure GCC using the @file{configure} script
3792 (@pxref{Installation}), it will construct the file @file{Makefile} from
3793 the template file @file{Makefile.in}. When it does this, it will
3794 incorporate makefile fragment files from the @file{config} directory,
3795 named @file{t-@var{target}} and @file{x-@var{host}}. If these files do
3796 not exist, it means nothing needs to be added for a given target or
3800 * Target Fragment:: Writing the @file{t-@var{target}} file.
3801 * Host Fragment:: Writing the @file{x-@var{host}} file.
3804 @node Target Fragment
3805 @section The Target Makefile Fragment
3806 @cindex target makefile fragment
3807 @cindex @file{t-@var{target}}
3809 The target makefile fragment, @file{t-@var{target}}, defines special
3810 target dependent variables and targets used in the @file{Makefile}:
3813 @findex LIBGCC2_CFLAGS
3814 @item LIBGCC2_CFLAGS
3815 Compiler flags to use when compiling @file{libgcc2.c}.
3817 @findex LIB2FUNCS_EXTRA
3818 @item LIB2FUNCS_EXTRA
3819 A list of source file names to be compiled or assembled and inserted
3820 into @file{libgcc.a}.
3822 @findex Floating Point Emulation
3823 @item Floating Point Emulation
3824 To have GCC include software floating point libraries in @file{libgcc.a}
3825 define @code{FPBIT} and @code{DPBIT} along with a few rules as follows:
3827 # We want fine grained libraries, so use the new code
3828 # to build the floating point emulation libraries.
3833 fp-bit.c: $(srcdir)/config/fp-bit.c
3834 echo '#define FLOAT' > fp-bit.c
3835 cat $(srcdir)/config/fp-bit.c >> fp-bit.c
3837 dp-bit.c: $(srcdir)/config/fp-bit.c
3838 cat $(srcdir)/config/fp-bit.c > dp-bit.c
3841 You may need to provide additional #defines at the beginning of @file{fp-bit.c}
3842 and @file{dp-bit.c} to control target endianness and other options.
3845 @findex CRTSTUFF_T_CFLAGS
3846 @item CRTSTUFF_T_CFLAGS
3847 Special flags used when compiling @file{crtstuff.c}.
3848 @xref{Initialization}.
3850 @findex CRTSTUFF_T_CFLAGS_S
3851 @item CRTSTUFF_T_CFLAGS_S
3852 Special flags used when compiling @file{crtstuff.c} for shared
3853 linking. Used if you use @file{crtbeginS.o} and @file{crtendS.o}
3854 in @code{EXTRA-PARTS}.
3855 @xref{Initialization}.
3857 @findex MULTILIB_OPTIONS
3858 @item MULTILIB_OPTIONS
3859 For some targets, invoking GCC in different ways produces objects
3860 that can not be linked together. For example, for some targets GCC
3861 produces both big and little endian code. For these targets, you must
3862 arrange for multiple versions of @file{libgcc.a} to be compiled, one for
3863 each set of incompatible options. When GCC invokes the linker, it
3864 arranges to link in the right version of @file{libgcc.a}, based on
3865 the command line options used.
3867 The @code{MULTILIB_OPTIONS} macro lists the set of options for which
3868 special versions of @file{libgcc.a} must be built. Write options that
3869 are mutually incompatible side by side, separated by a slash. Write
3870 options that may be used together separated by a space. The build
3871 procedure will build all combinations of compatible options.
3873 For example, if you set @code{MULTILIB_OPTIONS} to @samp{m68000/m68020
3874 msoft-float}, @file{Makefile} will build special versions of
3875 @file{libgcc.a} using the following sets of options: @samp{-m68000},
3876 @samp{-m68020}, @samp{-msoft-float}, @samp{-m68000 -msoft-float}, and
3877 @samp{-m68020 -msoft-float}.
3879 @findex MULTILIB_DIRNAMES
3880 @item MULTILIB_DIRNAMES
3881 If @code{MULTILIB_OPTIONS} is used, this variable specifies the
3882 directory names that should be used to hold the various libraries.
3883 Write one element in @code{MULTILIB_DIRNAMES} for each element in
3884 @code{MULTILIB_OPTIONS}. If @code{MULTILIB_DIRNAMES} is not used, the
3885 default value will be @code{MULTILIB_OPTIONS}, with all slashes treated
3888 For example, if @code{MULTILIB_OPTIONS} is set to @samp{m68000/m68020
3889 msoft-float}, then the default value of @code{MULTILIB_DIRNAMES} is
3890 @samp{m68000 m68020 msoft-float}. You may specify a different value if
3891 you desire a different set of directory names.
3893 @findex MULTILIB_MATCHES
3894 @item MULTILIB_MATCHES
3895 Sometimes the same option may be written in two different ways. If an
3896 option is listed in @code{MULTILIB_OPTIONS}, GCC needs to know about
3897 any synonyms. In that case, set @code{MULTILIB_MATCHES} to a list of
3898 items of the form @samp{option=option} to describe all relevant
3899 synonyms. For example, @samp{m68000=mc68000 m68020=mc68020}.
3901 @findex MULTILIB_EXCEPTIONS
3902 @item MULTILIB_EXCEPTIONS
3903 Sometimes when there are multiple sets of @code{MULTILIB_OPTIONS} being
3904 specified, there are combinations that should not be built. In that
3905 case, set @code{MULTILIB_EXCEPTIONS} to be all of the switch exceptions
3906 in shell case syntax that should not be built.
3908 For example, in the PowerPC embedded ABI support, it is not desirable
3909 to build libraries compiled with the @samp{-mcall-aix} option
3910 and either of the @samp{-fleading-underscore} or @samp{-mlittle} options
3911 at the same time. Therefore @code{MULTILIB_EXCEPTIONS} is set to
3912 @code{*mcall-aix/*fleading-underscore* *mlittle/*mcall-aix*}.
3914 @findex MULTILIB_EXTRA_OPTS
3915 @item MULTILIB_EXTRA_OPTS
3916 Sometimes it is desirable that when building multiple versions of
3917 @file{libgcc.a} certain options should always be passed on to the
3918 compiler. In that case, set @code{MULTILIB_EXTRA_OPTS} to be the list
3919 of options to be used for all builds.
3923 @section The Host Makefile Fragment
3924 @cindex host makefile fragment
3925 @cindex @file{x-@var{host}}
3927 The host makefile fragment, @file{x-@var{host}}, defines special host
3928 dependent variables and targets used in the @file{Makefile}:
3933 The compiler to use when building the first stage.
3937 The install program to use.
3942 @unnumbered Funding Free Software
3944 If you want to have more free software a few years from now, it makes
3945 sense for you to help encourage people to contribute funds for its
3946 development. The most effective approach known is to encourage
3947 commercial redistributors to donate.
3949 Users of free software systems can boost the pace of development by
3950 encouraging for-a-fee distributors to donate part of their selling price
3951 to free software developers---the Free Software Foundation, and others.
3953 The way to convince distributors to do this is to demand it and expect
3954 it from them. So when you compare distributors, judge them partly by
3955 how much they give to free software development. Show distributors
3956 they must compete to be the one who gives the most.
3958 To make this approach work, you must insist on numbers that you can
3959 compare, such as, ``We will donate ten dollars to the Frobnitz project
3960 for each disk sold.'' Don't be satisfied with a vague promise, such as
3961 ``A portion of the profits are donated,'' since it doesn't give a basis
3964 Even a precise fraction ``of the profits from this disk'' is not very
3965 meaningful, since creative accounting and unrelated business decisions
3966 can greatly alter what fraction of the sales price counts as profit.
3967 If the price you pay is $50, ten percent of the profit is probably
3968 less than a dollar; it might be a few cents, or nothing at all.
3970 Some redistributors do development work themselves. This is useful too;
3971 but to keep everyone honest, you need to inquire how much they do, and
3972 what kind. Some kinds of development make much more long-term
3973 difference than others. For example, maintaining a separate version of
3974 a program contributes very little; maintaining the standard version of a
3975 program for the whole community contributes much. Easy new ports
3976 contribute little, since someone else would surely do them; difficult
3977 ports such as adding a new CPU to the GNU Compiler Collection contribute more;
3978 major new features or packages contribute the most.
3980 By establishing the idea that supporting further development is ``the
3981 proper thing to do'' when distributing free software for a fee, we can
3982 assure a steady flow of resources into making more free software.
3985 Copyright (C) 1994 Free Software Foundation, Inc.
3986 Verbatim copying and redistribution of this section is permitted
3987 without royalty; alteration is not permitted.
3991 @unnumbered Linux and the GNU Project
3993 Many computer users run a modified version of the GNU system every
3994 day, without realizing it. Through a peculiar turn of events, the
3995 version of GNU which is widely used today is more often known as
3996 ``Linux'', and many users are not aware of the extent of its
3997 connection with the GNU Project.
3999 There really is a Linux; it is a kernel, and these people are using
4000 it. But you can't use a kernel by itself; a kernel is useful only as
4001 part of a whole system. The system in which Linux is typically used
4002 is a modified variant of the GNU system---in other words, a Linux-based
4005 Many users are not fully aware of the distinction between the kernel,
4006 which is Linux, and the whole system, which they also call ``Linux''.
4007 The ambiguous use of the name doesn't promote understanding.
4009 Programmers generally know that Linux is a kernel. But since they
4010 have generally heard the whole system called ``Linux'' as well, they
4011 often envisage a history which fits that name. For example, many
4012 believe that once Linus Torvalds finished writing the kernel, his
4013 friends looked around for other free software, and for no particular
4014 reason most everything necessary to make a Unix-like system was
4017 What they found was no accident---it was the GNU system. The available
4018 free software added up to a complete system because the GNU Project
4019 had been working since 1984 to make one. The GNU Manifesto
4020 had set forth the goal of developing a free Unix-like system, called
4021 GNU. By the time Linux was written, the system was almost finished.
4023 Most free software projects have the goal of developing a particular
4024 program for a particular job. For example, Linus Torvalds set out to
4025 write a Unix-like kernel (Linux); Donald Knuth set out to write a text
4026 formatter (TeX); Bob Scheifler set out to develop a window system (X
4027 Windows). It's natural to measure the contribution of this kind of
4028 project by specific programs that came from the project.
4030 If we tried to measure the GNU Project's contribution in this way,
4031 what would we conclude? One CD-ROM vendor found that in their ``Linux
4032 distribution'', GNU software was the largest single contingent, around
4033 28% of the total source code, and this included some of the essential
4034 major components without which there could be no system. Linux itself
4035 was about 3%. So if you were going to pick a name for the system
4036 based on who wrote the programs in the system, the most appropriate
4037 single choice would be ``GNU''.
4039 But we don't think that is the right way to consider the question.
4040 The GNU Project was not, is not, a project to develop specific
4041 software packages. It was not a project to develop a C compiler,
4042 although we did. It was not a project to develop a text editor,
4043 although we developed one. The GNU Project's aim was to develop
4044 @emph{a complete free Unix-like system}.
4046 Many people have made major contributions to the free software in the
4047 system, and they all deserve credit. But the reason it is @emph{a
4048 system}---and not just a collection of useful programs---is because the
4049 GNU Project set out to make it one. We wrote the programs that were
4050 needed to make a @emph{complete} free system. We wrote essential but
4051 unexciting major components, such as the assembler and linker, because
4052 you can't have a system without them. A complete system needs more
4053 than just programming tools, so we wrote other components as well,
4054 such as the Bourne Again SHell, the PostScript interpreter
4055 Ghostscript, and the GNU C library.
4057 By the early 90s we had put together the whole system aside from the
4058 kernel (and we were also working on a kernel, the GNU Hurd, which runs
4059 on top of Mach). Developing this kernel has been a lot harder than we
4060 expected, and we are still working on finishing it.
4062 Fortunately, you don't have to wait for it, because Linux is working
4063 now. When Linus Torvalds wrote Linux, he filled the last major gap.
4064 People could then put Linux together with the GNU system to make a
4065 complete free system: a Linux-based GNU system (or GNU/Linux system,
4068 Putting them together sounds simple, but it was not a trivial job.
4069 The GNU C library (called glibc for short) needed substantial changes.
4070 Integrating a complete system as a distribution that would work ``out
4071 of the box'' was a big job, too. It required addressing the issue of
4072 how to install and boot the system---a problem we had not tackled,
4073 because we hadn't yet reached that point. The people who developed
4074 the various system distributions made a substantial contribution.
4076 The GNU Project supports GNU/Linux systems as well as @emph{the}
4077 GNU system---even with funds. We funded the rewriting of the
4078 Linux-related extensions to the GNU C library, so that now they are
4079 well integrated, and the newest GNU/Linux systems use the current
4080 library release with no changes. We also funded an early stage of the
4081 development of Debian GNU/Linux.
4083 We use Linux-based GNU systems today for most of our work, and we hope
4084 you use them too. But please don't confuse the public by using the
4085 name ``Linux'' ambiguously. Linux is the kernel, one of the essential
4086 major components of the system. The system as a whole is more or less
4090 @unnumbered GNU GENERAL PUBLIC LICENSE
4091 @center Version 2, June 1991
4094 Copyright @copyright{} 1989, 1991 Free Software Foundation, Inc.
4095 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
4097 Everyone is permitted to copy and distribute verbatim copies
4098 of this license document, but changing it is not allowed.
4101 @unnumberedsec Preamble
4103 The licenses for most software are designed to take away your
4104 freedom to share and change it. By contrast, the GNU General Public
4105 License is intended to guarantee your freedom to share and change free
4106 software---to make sure the software is free for all its users. This
4107 General Public License applies to most of the Free Software
4108 Foundation's software and to any other program whose authors commit to
4109 using it. (Some other Free Software Foundation software is covered by
4110 the GNU Library General Public License instead.) You can apply it to
4113 When we speak of free software, we are referring to freedom, not
4114 price. Our General Public Licenses are designed to make sure that you
4115 have the freedom to distribute copies of free software (and charge for
4116 this service if you wish), that you receive source code or can get it
4117 if you want it, that you can change the software or use pieces of it
4118 in new free programs; and that you know you can do these things.
4120 To protect your rights, we need to make restrictions that forbid
4121 anyone to deny you these rights or to ask you to surrender the rights.
4122 These restrictions translate to certain responsibilities for you if you
4123 distribute copies of the software, or if you modify it.
4125 For example, if you distribute copies of such a program, whether
4126 gratis or for a fee, you must give the recipients all the rights that
4127 you have. You must make sure that they, too, receive or can get the
4128 source code. And you must show them these terms so they know their
4131 We protect your rights with two steps: (1) copyright the software, and
4132 (2) offer you this license which gives you legal permission to copy,
4133 distribute and/or modify the software.
4135 Also, for each author's protection and ours, we want to make certain
4136 that everyone understands that there is no warranty for this free
4137 software. If the software is modified by someone else and passed on, we
4138 want its recipients to know that what they have is not the original, so
4139 that any problems introduced by others will not reflect on the original
4140 authors' reputations.
4142 Finally, any free program is threatened constantly by software
4143 patents. We wish to avoid the danger that redistributors of a free
4144 program will individually obtain patent licenses, in effect making the
4145 program proprietary. To prevent this, we have made it clear that any
4146 patent must be licensed for everyone's free use or not licensed at all.
4148 The precise terms and conditions for copying, distribution and
4149 modification follow.
4152 @unnumberedsec TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
4155 @center TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
4160 This License applies to any program or other work which contains
4161 a notice placed by the copyright holder saying it may be distributed
4162 under the terms of this General Public License. The ``Program'', below,
4163 refers to any such program or work, and a ``work based on the Program''
4164 means either the Program or any derivative work under copyright law:
4165 that is to say, a work containing the Program or a portion of it,
4166 either verbatim or with modifications and/or translated into another
4167 language. (Hereinafter, translation is included without limitation in
4168 the term ``modification''.) Each licensee is addressed as ``you''.
4170 Activities other than copying, distribution and modification are not
4171 covered by this License; they are outside its scope. The act of
4172 running the Program is not restricted, and the output from the Program
4173 is covered only if its contents constitute a work based on the
4174 Program (independent of having been made by running the Program).
4175 Whether that is true depends on what the Program does.
4178 You may copy and distribute verbatim copies of the Program's
4179 source code as you receive it, in any medium, provided that you
4180 conspicuously and appropriately publish on each copy an appropriate
4181 copyright notice and disclaimer of warranty; keep intact all the
4182 notices that refer to this License and to the absence of any warranty;
4183 and give any other recipients of the Program a copy of this License
4184 along with the Program.
4186 You may charge a fee for the physical act of transferring a copy, and
4187 you may at your option offer warranty protection in exchange for a fee.
4190 You may modify your copy or copies of the Program or any portion
4191 of it, thus forming a work based on the Program, and copy and
4192 distribute such modifications or work under the terms of Section 1
4193 above, provided that you also meet all of these conditions:
4197 You must cause the modified files to carry prominent notices
4198 stating that you changed the files and the date of any change.
4201 You must cause any work that you distribute or publish, that in
4202 whole or in part contains or is derived from the Program or any
4203 part thereof, to be licensed as a whole at no charge to all third
4204 parties under the terms of this License.
4207 If the modified program normally reads commands interactively
4208 when run, you must cause it, when started running for such
4209 interactive use in the most ordinary way, to print or display an
4210 announcement including an appropriate copyright notice and a
4211 notice that there is no warranty (or else, saying that you provide
4212 a warranty) and that users may redistribute the program under
4213 these conditions, and telling the user how to view a copy of this
4214 License. (Exception: if the Program itself is interactive but
4215 does not normally print such an announcement, your work based on
4216 the Program is not required to print an announcement.)
4219 These requirements apply to the modified work as a whole. If
4220 identifiable sections of that work are not derived from the Program,
4221 and can be reasonably considered independent and separate works in
4222 themselves, then this License, and its terms, do not apply to those
4223 sections when you distribute them as separate works. But when you
4224 distribute the same sections as part of a whole which is a work based
4225 on the Program, the distribution of the whole must be on the terms of
4226 this License, whose permissions for other licensees extend to the
4227 entire whole, and thus to each and every part regardless of who wrote it.
4229 Thus, it is not the intent of this section to claim rights or contest
4230 your rights to work written entirely by you; rather, the intent is to
4231 exercise the right to control the distribution of derivative or
4232 collective works based on the Program.
4234 In addition, mere aggregation of another work not based on the Program
4235 with the Program (or with a work based on the Program) on a volume of
4236 a storage or distribution medium does not bring the other work under
4237 the scope of this License.
4240 You may copy and distribute the Program (or a work based on it,
4241 under Section 2) in object code or executable form under the terms of
4242 Sections 1 and 2 above provided that you also do one of the following:
4246 Accompany it with the complete corresponding machine-readable
4247 source code, which must be distributed under the terms of Sections
4248 1 and 2 above on a medium customarily used for software interchange; or,
4251 Accompany it with a written offer, valid for at least three
4252 years, to give any third party, for a charge no more than your
4253 cost of physically performing source distribution, a complete
4254 machine-readable copy of the corresponding source code, to be
4255 distributed under the terms of Sections 1 and 2 above on a medium
4256 customarily used for software interchange; or,
4259 Accompany it with the information you received as to the offer
4260 to distribute corresponding source code. (This alternative is
4261 allowed only for noncommercial distribution and only if you
4262 received the program in object code or executable form with such
4263 an offer, in accord with Subsection b above.)
4266 The source code for a work means the preferred form of the work for
4267 making modifications to it. For an executable work, complete source
4268 code means all the source code for all modules it contains, plus any
4269 associated interface definition files, plus the scripts used to
4270 control compilation and installation of the executable. However, as a
4271 special exception, the source code distributed need not include
4272 anything that is normally distributed (in either source or binary
4273 form) with the major components (compiler, kernel, and so on) of the
4274 operating system on which the executable runs, unless that component
4275 itself accompanies the executable.
4277 If distribution of executable or object code is made by offering
4278 access to copy from a designated place, then offering equivalent
4279 access to copy the source code from the same place counts as
4280 distribution of the source code, even though third parties are not
4281 compelled to copy the source along with the object code.
4284 You may not copy, modify, sublicense, or distribute the Program
4285 except as expressly provided under this License. Any attempt
4286 otherwise to copy, modify, sublicense or distribute the Program is
4287 void, and will automatically terminate your rights under this License.
4288 However, parties who have received copies, or rights, from you under
4289 this License will not have their licenses terminated so long as such
4290 parties remain in full compliance.
4293 You are not required to accept this License, since you have not
4294 signed it. However, nothing else grants you permission to modify or
4295 distribute the Program or its derivative works. These actions are
4296 prohibited by law if you do not accept this License. Therefore, by
4297 modifying or distributing the Program (or any work based on the
4298 Program), you indicate your acceptance of this License to do so, and
4299 all its terms and conditions for copying, distributing or modifying
4300 the Program or works based on it.
4303 Each time you redistribute the Program (or any work based on the
4304 Program), the recipient automatically receives a license from the
4305 original licensor to copy, distribute or modify the Program subject to
4306 these terms and conditions. You may not impose any further
4307 restrictions on the recipients' exercise of the rights granted herein.
4308 You are not responsible for enforcing compliance by third parties to
4312 If, as a consequence of a court judgment or allegation of patent
4313 infringement or for any other reason (not limited to patent issues),
4314 conditions are imposed on you (whether by court order, agreement or
4315 otherwise) that contradict the conditions of this License, they do not
4316 excuse you from the conditions of this License. If you cannot
4317 distribute so as to satisfy simultaneously your obligations under this
4318 License and any other pertinent obligations, then as a consequence you
4319 may not distribute the Program at all. For example, if a patent
4320 license would not permit royalty-free redistribution of the Program by
4321 all those who receive copies directly or indirectly through you, then
4322 the only way you could satisfy both it and this License would be to
4323 refrain entirely from distribution of the Program.
4325 If any portion of this section is held invalid or unenforceable under
4326 any particular circumstance, the balance of the section is intended to
4327 apply and the section as a whole is intended to apply in other
4330 It is not the purpose of this section to induce you to infringe any
4331 patents or other property right claims or to contest validity of any
4332 such claims; this section has the sole purpose of protecting the
4333 integrity of the free software distribution system, which is
4334 implemented by public license practices. Many people have made
4335 generous contributions to the wide range of software distributed
4336 through that system in reliance on consistent application of that
4337 system; it is up to the author/donor to decide if he or she is willing
4338 to distribute software through any other system and a licensee cannot
4341 This section is intended to make thoroughly clear what is believed to
4342 be a consequence of the rest of this License.
4345 If the distribution and/or use of the Program is restricted in
4346 certain countries either by patents or by copyrighted interfaces, the
4347 original copyright holder who places the Program under this License
4348 may add an explicit geographical distribution limitation excluding
4349 those countries, so that distribution is permitted only in or among
4350 countries not thus excluded. In such case, this License incorporates
4351 the limitation as if written in the body of this License.
4354 The Free Software Foundation may publish revised and/or new versions
4355 of the General Public License from time to time. Such new versions will
4356 be similar in spirit to the present version, but may differ in detail to
4357 address new problems or concerns.
4359 Each version is given a distinguishing version number. If the Program
4360 specifies a version number of this License which applies to it and ``any
4361 later version'', you have the option of following the terms and conditions
4362 either of that version or of any later version published by the Free
4363 Software Foundation. If the Program does not specify a version number of
4364 this License, you may choose any version ever published by the Free Software
4368 If you wish to incorporate parts of the Program into other free
4369 programs whose distribution conditions are different, write to the author
4370 to ask for permission. For software which is copyrighted by the Free
4371 Software Foundation, write to the Free Software Foundation; we sometimes
4372 make exceptions for this. Our decision will be guided by the two goals
4373 of preserving the free status of all derivatives of our free software and
4374 of promoting the sharing and reuse of software generally.
4377 @heading NO WARRANTY
4384 BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
4385 FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
4386 OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
4387 PROVIDE THE PROGRAM ``AS IS'' WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
4388 OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
4389 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
4390 TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
4391 PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
4392 REPAIR OR CORRECTION.
4395 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
4396 WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
4397 REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
4398 INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
4399 OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
4400 TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
4401 YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
4402 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
4403 POSSIBILITY OF SUCH DAMAGES.
4407 @heading END OF TERMS AND CONDITIONS
4410 @center END OF TERMS AND CONDITIONS
4414 @unnumberedsec How to Apply These Terms to Your New Programs
4416 If you develop a new program, and you want it to be of the greatest
4417 possible use to the public, the best way to achieve this is to make it
4418 free software which everyone can redistribute and change under these terms.
4420 To do so, attach the following notices to the program. It is safest
4421 to attach them to the start of each source file to most effectively
4422 convey the exclusion of warranty; and each file should have at least
4423 the ``copyright'' line and a pointer to where the full notice is found.
4426 @var{one line to give the program's name and a brief idea of what it does.}
4427 Copyright (C) @var{yyyy} @var{name of author}
4429 This program is free software; you can redistribute it and/or modify
4430 it under the terms of the GNU General Public License as published by
4431 the Free Software Foundation; either version 2 of the License, or
4432 (at your option) any later version.
4434 This program is distributed in the hope that it will be useful,
4435 but WITHOUT ANY WARRANTY; without even the implied warranty of
4436 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4437 GNU General Public License for more details.
4439 You should have received a copy of the GNU General Public License
4440 along with this program; if not, write to the Free Software
4441 Foundation, Inc., 59 Temple Place - Suite 330,
4442 Boston, MA 02111-1307, USA.
4445 Also add information on how to contact you by electronic and paper mail.
4447 If the program is interactive, make it output a short notice like this
4448 when it starts in an interactive mode:
4451 Gnomovision version 69, Copyright (C) @var{yyyy} @var{name of author}
4452 Gnomovision comes with ABSOLUTELY NO WARRANTY; for details
4454 This is free software, and you are welcome to redistribute it
4455 under certain conditions; type `show c' for details.
4458 The hypothetical commands @samp{show w} and @samp{show c} should show
4459 the appropriate parts of the General Public License. Of course, the
4460 commands you use may be called something other than @samp{show w} and
4461 @samp{show c}; they could even be mouse-clicks or menu items---whatever
4464 You should also get your employer (if you work as a programmer) or your
4465 school, if any, to sign a ``copyright disclaimer'' for the program, if
4466 necessary. Here is a sample; alter the names:
4469 Yoyodyne, Inc., hereby disclaims all copyright interest in the program
4470 `Gnomovision' (which makes passes at compilers) written by James Hacker.
4472 @var{signature of Ty Coon}, 1 April 1989
4473 Ty Coon, President of Vice
4476 This General Public License does not permit incorporating your program into
4477 proprietary programs. If your program is a subroutine library, you may
4478 consider it more useful to permit linking proprietary applications with the
4479 library. If this is what you want to do, use the GNU Library General
4480 Public License instead of this License.
4482 @c ---------------------------------------------------------------------
4484 @c ---------------------------------------------------------------------
4489 @unnumbered Contributors to GCC
4490 @cindex contributors
4491 @include contrib.texi
4493 @c ---------------------------------------------------------------------
4495 @c ---------------------------------------------------------------------
4502 @c ---------------------------------------------------------------------
4504 @c ---------------------------------------------------------------------