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
50 @settitle Using and Porting the GNU Compiler Collection (GCC)
53 @c seems reasonable to assume at least one of INTERNALS or USING is set...
55 @settitle Using the GNU Compiler Collection
58 @settitle Porting the GNU Compiler Collection
65 @c Use with @@smallbook.
67 @c Cause even numbered pages to be printed on the left hand side of
68 @c the page and odd numbered pages to be printed on the right hand
69 @c side of the page. Using this, you can print on both sides of a
70 @c sheet of paper and have the text on the same part of the sheet.
72 @c The text on right hand pages is pushed towards the right hand
73 @c margin and the text on left hand pages is pushed toward the left
75 @c (To provide the reverse effect, set bindingoffset to -0.75in.)
78 @c \global\bindingoffset=0.75in
79 @c \global\normaloffset =0.75in
83 @dircategory Programming
85 * gcc: (gcc). The GNU Compiler Collection.
89 This file documents the use and the internals of the GNU compiler.
93 This file documents the internals of the GNU compiler.
96 This file documents the use of the GNU compiler.
99 Published by the Free Software Foundation@*
100 59 Temple Place - Suite 330@*
101 Boston, MA 02111-1307 USA
103 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000 Free Software Foundation, Inc.
105 Permission is granted to make and distribute verbatim copies of
106 this manual provided the copyright notice and this permission notice
107 are preserved on all copies.
110 Permission is granted to process this file through Tex and print the
111 results, provided the printed document carries copying permission
112 notice identical to this one except for the removal of this paragraph
113 (this paragraph not being relevant to the printed manual).
116 Permission is granted to copy and distribute modified versions of this
117 manual under the conditions for verbatim copying, provided also that the
118 sections entitled ``GNU General Public License'' and ``Funding for Free
119 Software'' are included exactly as in the original, and provided that
120 the entire resulting derived work is distributed under the terms of a
121 permission notice identical to this one.
123 Permission is granted to copy and distribute translations of this manual
124 into another language, under the above conditions for modified versions,
125 except that the sections entitled ``GNU General Public License'' and
126 ``Funding for Free Software'', and this permission notice, may be
127 included in translations approved by the Free Software Foundation
128 instead of in the original English.
131 @setchapternewpage odd
136 @center @titlefont{Using and Porting the GNU Compiler Collection}
141 @title Using the GNU Compiler Collection
144 @title Porting the GNU Compiler Collection
147 @center Richard M. Stallman
149 @center Last updated 10 November 2000
151 @c The version number appears five times more in this file.
155 @vskip 0pt plus 1filll
156 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000 Free Software Foundation, Inc.
158 For GCC Version 2.97@*
160 Published by the Free Software Foundation @*
161 59 Temple Place - Suite 330@*
162 Boston, MA 02111-1307, USA@*
163 Last printed April, 1998.@*
164 Printed copies are available for $50 each.@*
167 Permission is granted to make and distribute verbatim copies of
168 this manual provided the copyright notice and this permission notice
169 are preserved on all copies.
171 Permission is granted to copy and distribute modified versions of this
172 manual under the conditions for verbatim copying, provided also that the
173 sections entitled ``GNU General Public License'' and ``Funding for Free
174 Software'' are included exactly as in the original, and provided that
175 the entire resulting derived work is distributed under the terms of a
176 permission notice identical to this one.
178 Permission is granted to copy and distribute translations of this manual
179 into another language, under the above conditions for modified versions,
180 except that the sections entitled ``GNU General Public License'' and
181 ``Funding for Free Software'', and this permission notice, may be
182 included in translations approved by the Free Software Foundation
183 instead of in the original English.
187 @node Top, G++ and GCC,, (DIR)
193 This manual documents how to run, install and port the GNU
194 compiler, as well as its new features and incompatibilities, and how to
195 report bugs. It corresponds to GCC version 2.97.
200 This manual documents how to run and install the GNU compiler,
201 as well as its new features and incompatibilities, and how to report
202 bugs. It corresponds to GCC version 2.97.
205 This manual documents how to port the GNU compiler,
206 as well as its new features and incompatibilities, and how to report
207 bugs. It corresponds to GCC version 2.97.
212 * G++ and GCC:: You can compile C or C++ programs.
213 * Standards:: Language standards supported by GCC.
214 * Invoking GCC:: Command options supported by @samp{gcc}.
215 * Installation:: How to configure, compile and install GCC.
216 * C Extensions:: GNU extensions to the C language family.
217 * C++ Extensions:: GNU extensions to the C++ language.
218 * Gcov:: gcov: a GCC test coverage program.
219 * Trouble:: If you have trouble installing GCC.
220 * Bugs:: How, why and where to report bugs.
221 * Service:: How to find suppliers of support for GCC.
222 * Contributing:: How to contribute to testing and developing GCC.
223 * VMS:: Using GCC on VMS.
226 * Portability:: Goals of GCC's portability features.
227 * Interface:: Function-call interface of GCC output.
228 * Passes:: Order of passes, what they do, and what each file is for.
229 * RTL:: The intermediate representation that most passes work on.
230 * Machine Desc:: How to write machine description instruction patterns.
231 * Target Macros:: How to write the machine description C macros.
232 * Config:: Writing the @file{xm-@var{machine}.h} file.
233 * Fragments:: Writing the @file{t-@var{target}} and @file{x-@var{host}} files.
236 * Funding:: How to help assure funding for free software.
237 * GNU/Linux:: Linux and the GNU Project
239 * Copying:: GNU General Public License says
240 how you can copy and share GCC.
241 * Contributors:: People who have contributed to GCC.
243 * Index:: Index of concepts and symbol names.
248 @chapter Compile C, C++, Objective C, Fortran, Java or CHILL
251 Several versions of the compiler (C, C++, Objective C, Fortran, Java
252 and CHILL) are integrated; this is why we use the name
253 ``GNU Compiler Collection''. GCC can compile programs written in any of these
254 languages. The Fortran and CHILL compilers are described in
255 separate manuals. The Java compiler currently has no manual documenting it.
258 ``GCC'' is a common shorthand term for the GNU Compiler Collection. This is both
259 the most general name for the compiler, and the name used when the
260 emphasis is on compiling C programs (as the abbreviation formerly
261 stood for ``GNU C Compiler'').
265 When referring to C++ compilation, it is usual to call the compiler
266 ``G++''. Since there is only one compiler, it is also accurate to call
267 it ``GCC'' no matter what the language context; however, the term
268 ``G++'' is more useful when the emphasis is on compiling C++ programs.
270 We use the name ``GCC'' to refer to the compilation system as a
271 whole, and more specifically to the language-independent part of the
272 compiler. For example, we refer to the optimization options as
273 affecting the behavior of ``GCC'' or sometimes just ``the compiler''.
275 Front ends for other languages, such as Ada 95 and Pascal exist but
276 have not yet been integrated into GCC. These front-ends, like that for C++,
277 are built in subdirectories of GCC and link to it. The result is an
278 integrated compiler that can compile programs written in C, C++,
279 Objective C, or any of the languages for which you have installed front
282 In this manual, we only discuss the options for the C, Objective-C, and
283 C++ compilers and those of the GCC core. Consult the documentation
284 of the other front ends for the options to use when compiling programs
285 written in other languages.
287 @cindex compiler compared to C++ preprocessor
288 @cindex intermediate C version, nonexistent
289 @cindex C intermediate output, nonexistent
290 G++ is a @emph{compiler}, not merely a preprocessor. G++ builds object
291 code directly from your C++ program source. There is no intermediate C
292 version of the program. (By contrast, for example, some other
293 implementations use a program that generates a C program from your C++
294 source.) Avoiding an intermediate C representation of the program means
295 that you get better object code, and better debugging information. The
296 GNU debugger, GDB, works with this information in the object code to
297 give you comprehensive C++ source-level editing capabilities
298 (@pxref{C,,C and C++,gdb.info, Debugging with GDB}).
300 @c FIXME! Someone who knows something about Objective C ought to put in
301 @c a paragraph or two about it here, and move the index entry down when
302 @c there is more to point to than the general mention in the 1st par.
305 @chapter Language Standards Supported by GCC
308 @cindex ANSI C standard
312 @cindex ANSI X3.159-1989
314 @cindex ISO C standard
329 @cindex Technical Corrigenda
331 @cindex Technical Corrigendum 1
333 @cindex Technical Corrigendum 2
335 @cindex freestanding implementation
336 @cindex freestanding environment
337 @cindex hosted implementation
338 @cindex hosted environment
339 @findex __STDC_HOSTED__
341 For each language compiled by GCC for which there is a standard, GCC
342 attempts to follow one or more versions of that standard, possibly
343 with some exceptions, and possibly with some extensions.
345 GCC supports three versions of the C standard, although support for
346 the most recent version is not yet complete.
348 The original ANSI C standard (X3.159-1989) was ratified in 1989 and
349 published in 1990. This standard was ratified as an ISO standard
350 (ISO/IEC 9899:1990) later in 1990. There were no technical
351 differences between these publications, although the sections of the
352 ANSI standard were renumbered and became clauses in the ISO standard.
353 This standard, in both its forms, is commonly known as @dfn{C89}, or
354 occasionally as @dfn{C90}, from the dates of ratification. The ANSI
355 standard, but not the ISO standard, also came with a Rationale
356 document. To select this standard in GCC, use one of the options
357 @samp{-ansi}, @samp{-std=c89} or @samp{-std=iso9899:1990}; to obtain
358 all the diagnostics required by the standard, you should also specify
359 @samp{-pedantic} (or @samp{-pedantic-errors} if you want them to be
360 errors rather than warnings). @xref{C Dialect Options,,Options
361 Controlling C Dialect}.
363 Errors in the 1990 ISO C standard were corrected in two Technical
364 Corrigenda published in 1994 and 1996. GCC does not support the
367 An amendment to the 1990 standard was published in 1995. This
368 amendment added digraphs and @code{__STDC_VERSION__} to the language,
369 but otherwise concerned the library. This amendment is commonly known
370 as @dfn{AMD1}; the amended standard is sometimes known as @dfn{C94} or
371 @dfn{C95}. To select this standard in GCC, use the option
372 @samp{-std=iso9899:199409} (with, as for other standard versions,
373 @samp{-pedantic} to receive all required diagnostics).
375 A new edition of the ISO C standard was published in 1999 as ISO/IEC
376 9899:1999, and is commonly known as @dfn{C99}. GCC has incomplete
377 support for this standard version; see
378 @uref{http://gcc.gnu.org/c99status.html} for details. To select this
379 standard, use @samp{-std=c99} or @samp{-std=iso9899:1999}. (While in
380 development, drafts of this standard version were referred to as
383 GCC also has some limited support for traditional (pre-ISO) C with the
384 @samp{-traditional} option. This support may be of use for compiling
385 some very old programs that have not been updated to ISO C, but should
386 not be used for new programs. It will not work with some modern C
387 libraries such as the GNU C library.
389 By default, GCC provides some extensions to the C language that on
390 rare occasions conflict with the C standard. @xref{C
391 Extensions,,Extensions to the C Language Family}. Use of the
392 @samp{-std} options listed above will disable these extensions where
393 they conflict with the C standard version selected. You may also
394 select an extended version of the C language explicitly with
395 @samp{-std=gnu89} (for C89 with GNU extensions) or @samp{-std=gnu99}
396 (for C99 with GNU extensions). The default, if no C language dialect
397 options are given, is @samp{-std=gnu89}; this will change to
398 @samp{-std=gnu99} in some future release when the C99 support is
399 complete. Some features that are part of the C99 standard are
400 accepted as extensions in C89 mode.
402 The ISO C standard defines (in clause 4) two classes of conforming
403 implementation. A @dfn{conforming hosted implementation} supports the
404 whole standard including all the library facilities; a @dfn{conforming
405 freestanding implementation} is only required to provide certain
406 library facilities: those in @code{<float.h>}, @code{<limits.h>},
407 @code{<stdarg.h>}, and @code{<stddef.h>}; since AMD1, also those in
408 @code{<iso646.h>}; and in C99, also those in @code{<stdbool.h>} and
409 @code{<stdint.h>}. In addition, complex types, added in C99, are not
410 required for freestanding implementations. The standard also defines
411 two environments for programs, a @dfn{freestanding environment},
412 required of all implementations and which may not have library
413 facilities beyond those required of freestanding implementations,
414 where the handling of program startup and termination are
415 implementation-defined, and a @dfn{hosted environment}, which is not
416 required, in which all the library facilities are provided and startup
417 is through a function @code{int main (void)} or @code{int main (int,
418 char *[])}. An OS kernel would be a freestanding environment; a
419 program using the facilities of an operating system would normally be
420 in a hosted implementation.
422 GNU CC aims towards being usable as a conforming freestanding
423 implementation, or as the compiler for a conforming hosted
424 implementation. By default, it will act as the compiler for a hosted
425 implementation, defining @code{__STDC_HOSTED__} as @code{1} and
426 presuming that when the names of ISO C functions are used, they have
427 the semantics defined in the standard. To make it act as a conforming
428 freestanding implementation for a freestanding environment, use the
429 option @samp{-ffreestanding}; it will then define
430 @code{__STDC_HOSTED__} to @code{0} and not make assumptions about the
431 meanings of function names from the standard library. To build an OS
432 kernel, you may well still need to make your own arrangements for
433 linking and startup. @xref{C Dialect Options,,Options Controlling C
436 GNU CC does not provide the library facilities required only of hosted
437 implementations, nor yet all the facilities required by C99 of
438 freestanding implementations; to use the facilities of a hosted
439 environment, you will need to find them elsewhere (for example, in the
440 GNU C library). @xref{Standard Libraries,,Standard Libraries}.
442 For references to Technical Corrigenda, Rationale documents and
443 information concerning the history of C that is available online, see
444 @uref{http://gcc.gnu.org/readings.html}
446 @c FIXME: details of C++ standard.
447 @c FIXME: definitions of Java and Objective C.
449 @xref{Language,,The GNU Fortran Language, g77, Using and Porting GNU
450 Fortran}, for details of the Fortran language supported by GCC.
452 @xref{References,,Language Definition References, chill, GNU Chill},
453 for details of the CHILL standard.
457 @include install.texi
464 @chapter Known Causes of Trouble with GCC
466 @cindex installation trouble
467 @cindex known causes of trouble
469 This section describes known problems that affect users of GCC. Most
470 of these are not GCC bugs per se---if they were, we would fix them.
471 But the result for a user may be like the result of a bug.
473 Some of these problems are due to bugs in other software, some are
474 missing features that are too much work to add, and some are places
475 where people's opinions differ as to what is best.
478 * Actual Bugs:: Bugs we will fix later.
479 * Installation Problems:: Problems that manifest when you install GCC.
480 * Cross-Compiler Problems:: Common problems of cross compiling with GCC.
481 * Interoperation:: Problems using GCC with other compilers,
482 and with certain linkers, assemblers and debuggers.
483 * External Bugs:: Problems compiling certain programs.
484 * Incompatibilities:: GCC is incompatible with traditional C.
485 * Fixed Headers:: GNU C uses corrected versions of system header files.
486 This is necessary, but doesn't always work smoothly.
487 * Standard Libraries:: GNU C uses the system C library, which might not be
488 compliant with the ISO C standard.
489 * Disappointments:: Regrettable things we can't change, but not quite bugs.
490 * C++ Misunderstandings:: Common misunderstandings with GNU C++.
491 * Protoize Caveats:: Things to watch out for when using @code{protoize}.
492 * Non-bugs:: Things we think are right, but some others disagree.
493 * Warnings and Errors:: Which problems in your code get warnings,
494 and which get errors.
498 @section Actual Bugs We Haven't Fixed Yet
502 The @code{fixincludes} script interacts badly with automounters; if the
503 directory of system header files is automounted, it tends to be
504 unmounted while @code{fixincludes} is running. This would seem to be a
505 bug in the automounter. We don't know any good way to work around it.
508 The @code{fixproto} script will sometimes add prototypes for the
509 @code{sigsetjmp} and @code{siglongjmp} functions that reference the
510 @code{jmp_buf} type before that type is defined. To work around this,
511 edit the offending file and place the typedef in front of the
515 There are several obscure case of mis-using struct, union, and
516 enum tags that are not detected as errors by the compiler.
519 When @samp{-pedantic-errors} is specified, GCC will incorrectly give
520 an error message when a function name is specified in an expression
521 involving the comma operator.
524 Loop unrolling doesn't work properly for certain C++ programs. This is
525 a bug in the C++ front end. It sometimes emits incorrect debug info, and
526 the loop unrolling code is unable to recover from this error.
529 @node Installation Problems
530 @section Installation Problems
532 This is a list of problems (and some apparent problems which don't
533 really mean anything is wrong) that show up during installation of GNU
538 On certain systems, defining certain environment variables such as
539 @code{CC} can interfere with the functioning of @code{make}.
542 If you encounter seemingly strange errors when trying to build the
543 compiler in a directory other than the source directory, it could be
544 because you have previously configured the compiler in the source
545 directory. Make sure you have done all the necessary preparations.
549 If you build GCC on a BSD system using a directory stored in a System
550 V file system, problems may occur in running @code{fixincludes} if the
551 System V file system doesn't support symbolic links. These problems
552 result in a failure to fix the declaration of @code{size_t} in
553 @file{sys/types.h}. If you find that @code{size_t} is a signed type and
554 that type mismatches occur, this could be the cause.
556 The solution is not to use such a directory for building GCC.
559 In previous versions of GCC, the @code{gcc} driver program looked for
560 @code{as} and @code{ld} in various places; for example, in files
561 beginning with @file{/usr/local/lib/gcc-}. GCC version 2 looks for
562 them in the directory
563 @file{/usr/local/lib/gcc-lib/@var{target}/@var{version}}.
565 Thus, to use a version of @code{as} or @code{ld} that is not the system
566 default, for example @code{gas} or GNU @code{ld}, you must put them in
567 that directory (or make links to them from that directory).
570 Some commands executed when making the compiler may fail (return a
571 non-zero status) and be ignored by @code{make}. These failures, which
572 are often due to files that were not found, are expected, and can safely
576 It is normal to have warnings in compiling certain files about
577 unreachable code and about enumeration type clashes. These files' names
578 begin with @samp{insn-}. Also, @file{real.c} may get some warnings that
582 Sometimes @code{make} recompiles parts of the compiler when installing
583 the compiler. In one case, this was traced down to a bug in
584 @code{make}. Either ignore the problem or switch to GNU Make.
587 If you have installed a program known as purify, you may find that it
588 causes errors while linking @code{enquire}, which is part of building
589 GCC. The fix is to get rid of the file @code{real-ld} which purify
590 installs---so that GCC won't try to use it.
593 On GNU/Linux SLS 1.01, there is a problem with @file{libc.a}: it does not
594 contain the obstack functions. However, GCC assumes that the obstack
595 functions are in @file{libc.a} when it is the GNU C library. To work
596 around this problem, change the @code{__GNU_LIBRARY__} conditional
597 around line 31 to @samp{#if 1}.
600 On some 386 systems, building the compiler never finishes because
601 @code{enquire} hangs due to a hardware problem in the motherboard---it
602 reports floating point exceptions to the kernel incorrectly. You can
603 install GCC except for @file{float.h} by patching out the command to
604 run @code{enquire}. You may also be able to fix the problem for real by
605 getting a replacement motherboard. This problem was observed in
606 Revision E of the Micronics motherboard, and is fixed in Revision F.
607 It has also been observed in the MYLEX MXA-33 motherboard.
609 If you encounter this problem, you may also want to consider removing
610 the FPU from the socket during the compilation. Alternatively, if you
611 are running SCO Unix, you can reboot and force the FPU to be ignored.
612 To do this, type @samp{hd(40)unix auto ignorefpu}.
615 On some 386 systems, GCC crashes trying to compile @file{enquire.c}.
616 This happens on machines that don't have a 387 FPU chip. On 386
617 machines, the system kernel is supposed to emulate the 387 when you
618 don't have one. The crash is due to a bug in the emulator.
620 One of these systems is the Unix from Interactive Systems: 386/ix.
621 On this system, an alternate emulator is provided, and it does work.
622 To use it, execute this command as super-user:
625 ln /etc/emulator.rel1 /etc/emulator
629 and then reboot the system. (The default emulator file remains present
630 under the name @file{emulator.dflt}.)
632 Try using @file{/etc/emulator.att}, if you have such a problem on the
635 Another system which has this problem is Esix. We don't know whether it
636 has an alternate emulator that works.
638 On NetBSD 0.8, a similar problem manifests itself as these error messages:
641 enquire.c: In function `fprop':
642 enquire.c:2328: floating overflow
646 On SCO systems, when compiling GCC with the system's compiler,
647 do not use @samp{-O}. Some versions of the system's compiler miscompile
650 @cindex @code{genflags}, crash on Sun 4
652 Sometimes on a Sun 4 you may observe a crash in the program
653 @code{genflags} or @code{genoutput} while building GCC. This is said to
654 be due to a bug in @code{sh}. You can probably get around it by running
655 @code{genflags} or @code{genoutput} manually and then retrying the
659 On Solaris 2, executables of GCC version 2.0.2 are commonly
660 available, but they have a bug that shows up when compiling current
661 versions of GCC: undefined symbol errors occur during assembly if you
664 The solution is to compile the current version of GCC without
665 @samp{-g}. That makes a working compiler which you can use to recompile
669 Solaris 2 comes with a number of optional OS packages. Some of these
670 packages are needed to use GCC fully. If you did not install all
671 optional packages when installing Solaris, you will need to verify that
672 the packages that GCC needs are installed.
674 To check whether an optional package is installed, use
675 the @code{pkginfo} command. To add an optional package, use the
676 @code{pkgadd} command. For further details, see the Solaris
679 For Solaris 2.0 and 2.1, GCC needs six packages: @samp{SUNWarc},
680 @samp{SUNWbtool}, @samp{SUNWesu}, @samp{SUNWhea}, @samp{SUNWlibm}, and
683 For Solaris 2.2, GCC needs an additional seventh package: @samp{SUNWsprot}.
686 On Solaris 2, trying to use the linker and other tools in
687 @file{/usr/ucb} to install GCC has been observed to cause trouble.
688 For example, the linker may hang indefinitely. The fix is to remove
689 @file{/usr/ucb} from your @code{PATH}.
692 If you use the 1.31 version of the MIPS assembler (such as was shipped
693 with Ultrix 3.1), you will need to use the -fno-delayed-branch switch
694 when optimizing floating point code. Otherwise, the assembler will
695 complain when the GCC compiler fills a branch delay slot with a
696 floating point instruction, such as @code{add.d}.
699 If on a MIPS system you get an error message saying ``does not have gp
700 sections for all it's [sic] sectons [sic]'', don't worry about it. This
701 happens whenever you use GAS with the MIPS linker, but there is not
702 really anything wrong, and it is okay to use the output file. You can
703 stop such warnings by installing the GNU linker.
705 It would be nice to extend GAS to produce the gp tables, but they are
706 optional, and there should not be a warning about their absence.
709 In Ultrix 4.0 on the MIPS machine, @file{stdio.h} does not work with GNU
710 CC at all unless it has been fixed with @code{fixincludes}. This causes
711 problems in building GCC. Once GCC is installed, the problems go
714 To work around this problem, when making the stage 1 compiler, specify
718 GCC_FOR_TARGET="./xgcc -B./ -I./include"
721 When making stage 2 and stage 3, specify this option:
724 CFLAGS="-g -I./include"
728 Users have reported some problems with version 2.0 of the MIPS
729 compiler tools that were shipped with Ultrix 4.1. Version 2.10
730 which came with Ultrix 4.2 seems to work fine.
732 Users have also reported some problems with version 2.20 of the
733 MIPS compiler tools that were shipped with RISC/os 4.x. The earlier
734 version 2.11 seems to work fine.
737 Some versions of the MIPS linker will issue an assertion failure
738 when linking code that uses @code{alloca} against shared
739 libraries on RISC-OS 5.0, and DEC's OSF/1 systems. This is a bug
740 in the linker, that is supposed to be fixed in future revisions.
741 To protect against this, GCC passes @samp{-non_shared} to the
742 linker unless you pass an explicit @samp{-shared} or
743 @samp{-call_shared} switch.
746 On System V release 3, you may get this error message
750 ld fatal: failed to write symbol name @var{something}
751 in strings table for file @var{whatever}
754 This probably indicates that the disk is full or your ULIMIT won't allow
755 the file to be as large as it needs to be.
757 This problem can also result because the kernel parameter @code{MAXUMEM}
758 is too small. If so, you must regenerate the kernel and make the value
759 much larger. The default value is reported to be 1024; a value of 32768
760 is said to work. Smaller values may also work.
763 On System V, if you get an error like this,
766 /usr/local/lib/bison.simple: In function `yyparse':
767 /usr/local/lib/bison.simple:625: virtual memory exhausted
771 that too indicates a problem with disk space, ULIMIT, or @code{MAXUMEM}.
774 Current GCC versions probably do not work on version 2 of the NeXT
778 On NeXTStep 3.0, the Objective C compiler does not work, due,
779 apparently, to a kernel bug that it happens to trigger. This problem
780 does not happen on 3.1.
783 On the Tower models 4@var{n}0 and 6@var{n}0, by default a process is not
784 allowed to have more than one megabyte of memory. GCC cannot compile
785 itself (or many other programs) with @samp{-O} in that much memory.
787 To solve this problem, reconfigure the kernel adding the following line
788 to the configuration file:
795 On HP 9000 series 300 or 400 running HP-UX release 8.0, there is a bug
796 in the assembler that must be fixed before GCC can be built. This
797 bug manifests itself during the first stage of compilation, while
798 building @file{libgcc2.a}:
802 cc1: warning: `-g' option not supported on this version of GCC
803 cc1: warning: `-g1' option not supported on this version of GCC
804 ./xgcc: Internal compiler error: program as got fatal signal 11
807 A patched version of the assembler is available as the file
808 @uref{ftp://altdorf.ai.mit.edu/archive/cph/hpux-8.0-assembler}. If you
809 have HP software support, the patch can also be obtained directly from
810 HP, as described in the following note:
813 This is the patched assembler, to patch SR#1653-010439, where the
814 assembler aborts on floating point constants.
816 The bug is not really in the assembler, but in the shared library
817 version of the function ``cvtnum(3c)''. The bug on ``cvtnum(3c)'' is
818 SR#4701-078451. Anyway, the attached assembler uses the archive
819 library version of ``cvtnum(3c)'' and thus does not exhibit the bug.
822 This patch is also known as PHCO_4484.
825 On HP-UX version 8.05, but not on 8.07 or more recent versions,
826 the @code{fixproto} shell script triggers a bug in the system shell.
827 If you encounter this problem, upgrade your operating system or
828 use BASH (the GNU shell) to run @code{fixproto}.
831 Some versions of the Pyramid C compiler are reported to be unable to
832 compile GCC. You must use an older version of GCC for
833 bootstrapping. One indication of this problem is if you get a crash
834 when GCC compiles the function @code{muldi3} in file @file{libgcc2.c}.
836 You may be able to succeed by getting GCC version 1, installing it,
837 and using it to compile GCC version 2. The bug in the Pyramid C
838 compiler does not seem to affect GCC version 1.
841 There may be similar problems on System V Release 3.1 on 386 systems.
844 On the Intel Paragon (an i860 machine), if you are using operating
845 system version 1.0, you will get warnings or errors about redefinition
846 of @code{va_arg} when you build GCC.
848 If this happens, then you need to link most programs with the library
849 @file{iclib.a}. You must also modify @file{stdio.h} as follows: before
853 #if defined(__i860__) && !defined(_VA_LIST)
868 extern int vprintf(const char *, va_list );
869 extern int vsprintf(char *, const char *, va_list );
880 These problems don't exist in operating system version 1.1.
883 On the Altos 3068, programs compiled with GCC won't work unless you
884 fix a kernel bug. This happens using system versions V.2.2 1.0gT1 and
885 V.2.2 1.0e and perhaps later versions as well. See the file
889 You will get several sorts of compilation and linking errors on the
890 we32k if you don't follow the special instructions. @xref{Configurations}.
893 A bug in the HP-UX 8.05 (and earlier) shell will cause the fixproto
894 program to report an error of the form:
897 ./fixproto: sh internal 1K buffer overflow
900 To fix this, change the first line of the fixproto script to look like:
907 @node Cross-Compiler Problems
908 @section Cross-Compiler Problems
910 You may run into problems with cross compilation on certain machines,
915 Cross compilation can run into trouble for certain machines because
916 some target machines' assemblers require floating point numbers to be
917 written as @emph{integer} constants in certain contexts.
919 The compiler writes these integer constants by examining the floating
920 point value as an integer and printing that integer, because this is
921 simple to write and independent of the details of the floating point
922 representation. But this does not work if the compiler is running on
923 a different machine with an incompatible floating point format, or
924 even a different byte-ordering.
926 In addition, correct constant folding of floating point values
927 requires representing them in the target machine's format.
928 (The C standard does not quite require this, but in practice
929 it is the only way to win.)
931 It is now possible to overcome these problems by defining macros such
932 as @code{REAL_VALUE_TYPE}. But doing so is a substantial amount of
933 work for each target machine.
935 @xref{Cross-compilation}.
938 @xref{Cross-compilation,,Cross Compilation and Floating Point Format,
939 gcc.info, Using and Porting GCC}.
943 At present, the program @file{mips-tfile} which adds debug
944 support to object files on MIPS systems does not work in a cross
949 @section Interoperation
951 This section lists various difficulties encountered in using GNU C or
952 GNU C++ together with other compilers or with the assemblers, linkers,
953 libraries and debuggers on certain systems.
957 Objective C does not work on the RS/6000.
960 GNU C++ does not do name mangling in the same way as other C++
961 compilers. This means that object files compiled with one compiler
962 cannot be used with another.
964 This effect is intentional, to protect you from more subtle problems.
965 Compilers differ as to many internal details of C++ implementation,
966 including: how class instances are laid out, how multiple inheritance is
967 implemented, and how virtual function calls are handled. If the name
968 encoding were made the same, your programs would link against libraries
969 provided from other compilers---but the programs would then crash when
970 run. Incompatible libraries are then detected at link time, rather than
974 Older GDB versions sometimes fail to read the output of GCC version
975 2. If you have trouble, get GDB version 4.4 or later.
979 DBX rejects some files produced by GCC, though it accepts similar
980 constructs in output from PCC. Until someone can supply a coherent
981 description of what is valid DBX input and what is not, there is
982 nothing I can do about these problems. You are on your own.
985 The GNU assembler (GAS) does not support PIC. To generate PIC code, you
986 must use some other assembler, such as @file{/bin/as}.
989 On some BSD systems, including some versions of Ultrix, use of profiling
990 causes static variable destructors (currently used only in C++) not to
994 Use of @samp{-I/usr/include} may cause trouble.
996 Many systems come with header files that won't work with GCC unless
997 corrected by @code{fixincludes}. The corrected header files go in a new
998 directory; GCC searches this directory before @file{/usr/include}.
999 If you use @samp{-I/usr/include}, this tells GCC to search
1000 @file{/usr/include} earlier on, before the corrected headers. The
1001 result is that you get the uncorrected header files.
1003 Instead, you should use these options (when compiling C programs):
1006 -I/usr/local/lib/gcc-lib/@var{target}/@var{version}/include -I/usr/include
1009 For C++ programs, GCC also uses a special directory that defines C++
1010 interfaces to standard C subroutines. This directory is meant to be
1011 searched @emph{before} other standard include directories, so that it
1012 takes precedence. If you are compiling C++ programs and specifying
1013 include directories explicitly, use this option first, then the two
1017 -I/usr/local/lib/g++-include
1021 @cindex @code{vfork}, for the Sun-4
1023 There is a bug in @code{vfork} on the Sun-4 which causes the registers
1024 of the child process to clobber those of the parent. Because of this,
1025 programs that call @code{vfork} are likely to lose when compiled
1026 optimized with GCC when the child code alters registers which contain
1027 C variables in the parent. This affects variables which are live in the
1028 parent across the call to @code{vfork}.
1030 If you encounter this, you can work around the problem by declaring
1031 variables @code{volatile} in the function that calls @code{vfork}, until
1032 the problem goes away, or by not declaring them @code{register} and not
1033 using @samp{-O} for those source files.
1037 On some SGI systems, when you use @samp{-lgl_s} as an option,
1038 it gets translated magically to @samp{-lgl_s -lX11_s -lc_s}.
1039 Naturally, this does not happen when you use GCC.
1040 You must specify all three options explicitly.
1043 On a Sparc, GCC aligns all values of type @code{double} on an 8-byte
1044 boundary, and it expects every @code{double} to be so aligned. The Sun
1045 compiler usually gives @code{double} values 8-byte alignment, with one
1046 exception: function arguments of type @code{double} may not be aligned.
1048 As a result, if a function compiled with Sun CC takes the address of an
1049 argument of type @code{double} and passes this pointer of type
1050 @code{double *} to a function compiled with GCC, dereferencing the
1051 pointer may cause a fatal signal.
1053 One way to solve this problem is to compile your entire program with GNU
1054 CC. Another solution is to modify the function that is compiled with
1055 Sun CC to copy the argument into a local variable; local variables
1056 are always properly aligned. A third solution is to modify the function
1057 that uses the pointer to dereference it via the following function
1058 @code{access_double} instead of directly with @samp{*}:
1062 access_double (double *unaligned_ptr)
1064 union d2i @{ double d; int i[2]; @};
1066 union d2i *p = (union d2i *) unaligned_ptr;
1077 Storing into the pointer can be done likewise with the same union.
1080 On Solaris, the @code{malloc} function in the @file{libmalloc.a} library
1081 may allocate memory that is only 4 byte aligned. Since GCC on the
1082 Sparc assumes that doubles are 8 byte aligned, this may result in a
1083 fatal signal if doubles are stored in memory allocated by the
1084 @file{libmalloc.a} library.
1086 The solution is to not use the @file{libmalloc.a} library. Use instead
1087 @code{malloc} and related functions from @file{libc.a}; they do not have
1091 Sun forgot to include a static version of @file{libdl.a} with some
1092 versions of SunOS (mainly 4.1). This results in undefined symbols when
1093 linking static binaries (that is, if you use @samp{-static}). If you
1094 see undefined symbols @code{_dlclose}, @code{_dlsym} or @code{_dlopen}
1095 when linking, compile and link against the file
1096 @file{mit/util/misc/dlsym.c} from the MIT version of X windows.
1099 The 128-bit long double format that the Sparc port supports currently
1100 works by using the architecturally defined quad-word floating point
1101 instructions. Since there is no hardware that supports these
1102 instructions they must be emulated by the operating system. Long
1103 doubles do not work in Sun OS versions 4.0.3 and earlier, because the
1104 kernel emulator uses an obsolete and incompatible format. Long doubles
1105 do not work in Sun OS version 4.1.1 due to a problem in a Sun library.
1106 Long doubles do work on Sun OS versions 4.1.2 and higher, but GCC
1107 does not enable them by default. Long doubles appear to work in Sun OS
1111 On HP-UX version 9.01 on the HP PA, the HP compiler @code{cc} does not
1112 compile GCC correctly. We do not yet know why. However, GCC
1113 compiled on earlier HP-UX versions works properly on HP-UX 9.01 and can
1114 compile itself properly on 9.01.
1117 On the HP PA machine, ADB sometimes fails to work on functions compiled
1118 with GCC. Specifically, it fails to work on functions that use
1119 @code{alloca} or variable-size arrays. This is because GCC doesn't
1120 generate HP-UX unwind descriptors for such functions. It may even be
1121 impossible to generate them.
1124 Debugging (@samp{-g}) is not supported on the HP PA machine, unless you use
1125 the preliminary GNU tools (@pxref{Installation}).
1128 Taking the address of a label may generate errors from the HP-UX
1129 PA assembler. GAS for the PA does not have this problem.
1132 Using floating point parameters for indirect calls to static functions
1133 will not work when using the HP assembler. There simply is no way for GCC
1134 to specify what registers hold arguments for static functions when using
1135 the HP assembler. GAS for the PA does not have this problem.
1138 In extremely rare cases involving some very large functions you may
1139 receive errors from the HP linker complaining about an out of bounds
1140 unconditional branch offset. This used to occur more often in previous
1141 versions of GCC, but is now exceptionally rare. If you should run
1142 into it, you can work around by making your function smaller.
1145 GCC compiled code sometimes emits warnings from the HP-UX assembler of
1149 (warning) Use of GR3 when
1150 frame >= 8192 may cause conflict.
1153 These warnings are harmless and can be safely ignored.
1156 The current version of the assembler (@file{/bin/as}) for the RS/6000
1157 has certain problems that prevent the @samp{-g} option in GCC from
1158 working. Note that @file{Makefile.in} uses @samp{-g} by default when
1159 compiling @file{libgcc2.c}.
1161 IBM has produced a fixed version of the assembler. The upgraded
1162 assembler unfortunately was not included in any of the AIX 3.2 update
1163 PTF releases (3.2.2, 3.2.3, or 3.2.3e). Users of AIX 3.1 should request
1164 PTF U403044 from IBM and users of AIX 3.2 should request PTF U416277.
1165 See the file @file{README.RS6000} for more details on these updates.
1167 You can test for the presence of a fixed assembler by using the
1175 If the command exits normally, the assembler fix already is installed.
1176 If the assembler complains that "-u" is an unknown flag, you need to
1180 On the IBM RS/6000, compiling code of the form
1191 will cause the linker to report an undefined symbol @code{foo}.
1192 Although this behavior differs from most other systems, it is not a
1193 bug because redefining an @code{extern} variable as @code{static}
1194 is undefined in ISO C.
1197 AIX on the RS/6000 provides support (NLS) for environments outside of
1198 the United States. Compilers and assemblers use NLS to support
1199 locale-specific representations of various objects including
1200 floating-point numbers ("." vs "," for separating decimal fractions).
1201 There have been problems reported where the library linked with GCC does
1202 not produce the same floating-point formats that the assembler accepts.
1203 If you have this problem, set the LANG environment variable to "C" or
1207 Even if you specify @samp{-fdollars-in-identifiers},
1208 you cannot successfully use @samp{$} in identifiers on the RS/6000 due
1209 to a restriction in the IBM assembler. GAS supports these
1213 On the RS/6000, XLC version 1.3.0.0 will miscompile @file{jump.c}. XLC
1214 version 1.3.0.1 or later fixes this problem. You can obtain XLC-1.3.0.2
1215 by requesting PTF 421749 from IBM.
1218 There is an assembler bug in versions of DG/UX prior to 5.4.2.01 that
1219 occurs when the @samp{fldcr} instruction is used. GCC uses
1220 @samp{fldcr} on the 88100 to serialize volatile memory references. Use
1221 the option @samp{-mno-serialize-volatile} if your version of the
1222 assembler has this bug.
1225 On VMS, GAS versions 1.38.1 and earlier may cause spurious warning
1226 messages from the linker. These warning messages complain of mismatched
1227 psect attributes. You can ignore them. @xref{VMS Install}.
1230 On NewsOS version 3, if you include both of the files @file{stddef.h}
1231 and @file{sys/types.h}, you get an error because there are two typedefs
1232 of @code{size_t}. You should change @file{sys/types.h} by adding these
1233 lines around the definition of @code{size_t}:
1238 @var{actual typedef here}
1244 On the Alliant, the system's own convention for returning structures
1245 and unions is unusual, and is not compatible with GCC no matter
1246 what options are used.
1251 On the IBM RT PC, the MetaWare HighC compiler (hc) uses a different
1252 convention for structure and union returning. Use the option
1253 @samp{-mhc-struct-return} to tell GCC to use a convention compatible
1256 @cindex Vax calling convention
1257 @cindex Ultrix calling convention
1259 On Ultrix, the Fortran compiler expects registers 2 through 5 to be saved
1260 by function calls. However, the C compiler uses conventions compatible
1261 with BSD Unix: registers 2 through 5 may be clobbered by function calls.
1263 GCC uses the same convention as the Ultrix C compiler. You can use
1264 these options to produce code compatible with the Fortran compiler:
1267 -fcall-saved-r2 -fcall-saved-r3 -fcall-saved-r4 -fcall-saved-r5
1271 On the WE32k, you may find that programs compiled with GCC do not
1272 work with the standard shared C library. You may need to link with
1273 the ordinary C compiler. If you do so, you must specify the following
1277 -L/usr/local/lib/gcc-lib/we32k-att-sysv/2.8.1 -lgcc -lc_s
1280 The first specifies where to find the library @file{libgcc.a}
1281 specified with the @samp{-lgcc} option.
1283 GCC does linking by invoking @code{ld}, just as @code{cc} does, and
1284 there is no reason why it @emph{should} matter which compilation program
1285 you use to invoke @code{ld}. If someone tracks this problem down,
1286 it can probably be fixed easily.
1289 On the Alpha, you may get assembler errors about invalid syntax as a
1290 result of floating point constants. This is due to a bug in the C
1291 library functions @code{ecvt}, @code{fcvt} and @code{gcvt}. Given valid
1292 floating point numbers, they sometimes print @samp{NaN}.
1295 On Irix 4.0.5F (and perhaps in some other versions), an assembler bug
1296 sometimes reorders instructions incorrectly when optimization is turned
1297 on. If you think this may be happening to you, try using the GNU
1298 assembler; GAS version 2.1 supports ECOFF on Irix.
1300 Or use the @samp{-noasmopt} option when you compile GCC with itself,
1301 and then again when you compile your program. (This is a temporary
1302 kludge to turn off assembler optimization on Irix.) If this proves to
1303 be what you need, edit the assembler spec in the file @file{specs} so
1304 that it unconditionally passes @samp{-O0} to the assembler, and never
1305 passes @samp{-O2} or @samp{-O3}.
1309 @section Problems Compiling Certain Programs
1311 @c prevent bad page break with this line
1312 Certain programs have problems compiling.
1316 Parse errors may occur compiling X11 on a Decstation running Ultrix 4.2
1317 because of problems in DEC's versions of the X11 header files
1318 @file{X11/Xlib.h} and @file{X11/Xutil.h}. People recommend adding
1319 @samp{-I/usr/include/mit} to use the MIT versions of the header files,
1320 using the @samp{-traditional} switch to turn off ISO C, or fixing the
1321 header files by adding this:
1325 #define NeedFunctionPrototypes 0
1330 On various 386 Unix systems derived from System V, including SCO, ISC,
1331 and ESIX, you may get error messages about running out of virtual memory
1332 while compiling certain programs.
1334 You can prevent this problem by linking GCC with the GNU malloc
1335 (which thus replaces the malloc that comes with the system). GNU malloc
1336 is available as a separate package, and also in the file
1337 @file{src/gmalloc.c} in the GNU Emacs 19 distribution.
1339 If you have installed GNU malloc as a separate library package, use this
1340 option when you relink GCC:
1343 MALLOC=/usr/local/lib/libgmalloc.a
1346 Alternatively, if you have compiled @file{gmalloc.c} from Emacs 19, copy
1347 the object file to @file{gmalloc.o} and use this option when you relink
1355 @node Incompatibilities
1356 @section Incompatibilities of GCC
1357 @cindex incompatibilities of GCC
1359 There are several noteworthy incompatibilities between GNU C and K&R
1360 (non-ISO) versions of C. The @samp{-traditional} option
1361 eliminates many of these incompatibilities, @emph{but not all}, by
1362 telling GNU C to behave like a K&R C compiler.
1365 @cindex string constants
1366 @cindex read-only strings
1367 @cindex shared strings
1369 GCC normally makes string constants read-only. If several
1370 identical-looking string constants are used, GCC stores only one
1373 @cindex @code{mktemp}, and constant strings
1374 One consequence is that you cannot call @code{mktemp} with a string
1375 constant argument. The function @code{mktemp} always alters the
1376 string its argument points to.
1378 @cindex @code{sscanf}, and constant strings
1379 @cindex @code{fscanf}, and constant strings
1380 @cindex @code{scanf}, and constant strings
1381 Another consequence is that @code{sscanf} does not work on some systems
1382 when passed a string constant as its format control string or input.
1383 This is because @code{sscanf} incorrectly tries to write into the string
1384 constant. Likewise @code{fscanf} and @code{scanf}.
1386 The best solution to these problems is to change the program to use
1387 @code{char}-array variables with initialization strings for these
1388 purposes instead of string constants. But if this is not possible,
1389 you can use the @samp{-fwritable-strings} flag, which directs GCC
1390 to handle string constants the same way most C compilers do.
1391 @samp{-traditional} also has this effect, among others.
1394 @code{-2147483648} is positive.
1396 This is because 2147483648 cannot fit in the type @code{int}, so
1397 (following the ISO C rules) its data type is @code{unsigned long int}.
1398 Negating this value yields 2147483648 again.
1401 GCC does not substitute macro arguments when they appear inside of
1402 string constants. For example, the following macro in GCC
1409 will produce output @code{"a"} regardless of what the argument @var{a} is.
1411 The @samp{-traditional} option directs GCC to handle such cases
1412 (among others) in the old-fashioned (non-ISO) fashion.
1414 @cindex @code{setjmp} incompatibilities
1415 @cindex @code{longjmp} incompatibilities
1417 When you use @code{setjmp} and @code{longjmp}, the only automatic
1418 variables guaranteed to remain valid are those declared
1419 @code{volatile}. This is a consequence of automatic register
1420 allocation. Consider this function:
1434 /* @r{@code{longjmp (j)} may occur in @code{fun3}.} */
1439 Here @code{a} may or may not be restored to its first value when the
1440 @code{longjmp} occurs. If @code{a} is allocated in a register, then
1441 its first value is restored; otherwise, it keeps the last value stored
1444 If you use the @samp{-W} option with the @samp{-O} option, you will
1445 get a warning when GCC thinks such a problem might be possible.
1447 The @samp{-traditional} option directs GNU C to put variables in
1448 the stack by default, rather than in registers, in functions that
1449 call @code{setjmp}. This results in the behavior found in
1450 traditional C compilers.
1453 Programs that use preprocessing directives in the middle of macro
1454 arguments do not work with GCC. For example, a program like this
1463 ISO C does not permit such a construct. It would make sense to support
1464 it when @samp{-traditional} is used, but it is too much work to
1468 K&R compilers allow comments to cross over an inclusion boundary (i.e.
1469 started in an include file and ended in the including file). I think
1470 this would be quite ugly and can't imagine it could be needed.
1472 @cindex external declaration scope
1473 @cindex scope of external declarations
1474 @cindex declaration scope
1476 Declarations of external variables and functions within a block apply
1477 only to the block containing the declaration. In other words, they
1478 have the same scope as any other declaration in the same place.
1480 In some other C compilers, a @code{extern} declaration affects all the
1481 rest of the file even if it happens within a block.
1483 The @samp{-traditional} option directs GNU C to treat all @code{extern}
1484 declarations as global, like traditional compilers.
1487 In traditional C, you can combine @code{long}, etc., with a typedef name,
1492 typedef long foo bar;
1495 In ISO C, this is not allowed: @code{long} and other type modifiers
1496 require an explicit @code{int}. Because this criterion is expressed
1497 by Bison grammar rules rather than C code, the @samp{-traditional}
1498 flag cannot alter it.
1500 @cindex typedef names as function parameters
1502 PCC allows typedef names to be used as function parameters. The
1503 difficulty described immediately above applies here too.
1506 When in @samp{-traditional} mode, GCC allows the following erroneous
1507 pair of declarations to appear together in a given scope:
1515 GCC treats all characters of identifiers as significant, even when in
1516 @samp{-traditional} mode. According to K&R-1 (2.2), ``No more than the
1517 first eight characters are significant, although more may be used.''.
1518 Also according to K&R-1 (2.2), ``An identifier is a sequence of letters
1519 and digits; the first character must be a letter. The underscore _
1520 counts as a letter.'', but GCC also allows dollar signs in identifiers.
1524 PCC allows whitespace in the middle of compound assignment operators
1525 such as @samp{+=}. GCC, following the ISO standard, does not
1526 allow this. The difficulty described immediately above applies here
1532 GCC complains about unterminated character constants inside of
1533 preprocessing conditionals that fail. Some programs have English
1534 comments enclosed in conditionals that are guaranteed to fail; if these
1535 comments contain apostrophes, GCC will probably report an error. For
1536 example, this code would produce an error:
1540 You can't expect this to work.
1544 The best solution to such a problem is to put the text into an actual
1545 C comment delimited by @samp{/*@dots{}*/}. However,
1546 @samp{-traditional} suppresses these error messages.
1549 Many user programs contain the declaration @samp{long time ();}. In the
1550 past, the system header files on many systems did not actually declare
1551 @code{time}, so it did not matter what type your program declared it to
1552 return. But in systems with ISO C headers, @code{time} is declared to
1553 return @code{time_t}, and if that is not the same as @code{long}, then
1554 @samp{long time ();} is erroneous.
1556 The solution is to change your program to use appropriate system headers
1557 (@code{<time.h>} on systems with ISO C headers) and not to declare
1558 @code{time} if the system header files declare it, or failing that to
1559 use @code{time_t} as the return type of @code{time}.
1561 @cindex @code{float} as function value type
1563 When compiling functions that return @code{float}, PCC converts it to
1564 a double. GCC actually returns a @code{float}. If you are concerned
1565 with PCC compatibility, you should declare your functions to return
1566 @code{double}; you might as well say what you mean.
1571 When compiling functions that return structures or unions, GCC
1572 output code normally uses a method different from that used on most
1573 versions of Unix. As a result, code compiled with GCC cannot call
1574 a structure-returning function compiled with PCC, and vice versa.
1576 The method used by GCC is as follows: a structure or union which is
1577 1, 2, 4 or 8 bytes long is returned like a scalar. A structure or union
1578 with any other size is stored into an address supplied by the caller
1579 (usually in a special, fixed register, but on some machines it is passed
1580 on the stack). The machine-description macros @code{STRUCT_VALUE} and
1581 @code{STRUCT_INCOMING_VALUE} tell GCC where to pass this address.
1583 By contrast, PCC on most target machines returns structures and unions
1584 of any size by copying the data into an area of static storage, and then
1585 returning the address of that storage as if it were a pointer value.
1586 The caller must copy the data from that memory area to the place where
1587 the value is wanted. GCC does not use this method because it is
1588 slower and nonreentrant.
1590 On some newer machines, PCC uses a reentrant convention for all
1591 structure and union returning. GCC on most of these machines uses a
1592 compatible convention when returning structures and unions in memory,
1593 but still returns small structures and unions in registers.
1595 You can tell GCC to use a compatible convention for all structure and
1596 union returning with the option @samp{-fpcc-struct-return}.
1598 @cindex preprocessing tokens
1599 @cindex preprocessing numbers
1601 GNU C complains about program fragments such as @samp{0x74ae-0x4000}
1602 which appear to be two hexadecimal constants separated by the minus
1603 operator. Actually, this string is a single @dfn{preprocessing token}.
1604 Each such token must correspond to one token in C. Since this does not,
1605 GNU C prints an error message. Although it may appear obvious that what
1606 is meant is an operator and two values, the ISO C standard specifically
1607 requires that this be treated as erroneous.
1609 A @dfn{preprocessing token} is a @dfn{preprocessing number} if it
1610 begins with a digit and is followed by letters, underscores, digits,
1611 periods and @samp{e+}, @samp{e-}, @samp{E+}, or @samp{E-} character
1614 To make the above program fragment valid, place whitespace in front of
1615 the minus sign. This whitespace will end the preprocessing number.
1619 @section Fixed Header Files
1621 GCC needs to install corrected versions of some system header files.
1622 This is because most target systems have some header files that won't
1623 work with GCC unless they are changed. Some have bugs, some are
1624 incompatible with ISO C, and some depend on special features of other
1627 Installing GCC automatically creates and installs the fixed header
1628 files, by running a program called @code{fixincludes} (or for certain
1629 targets an alternative such as @code{fixinc.svr4}). Normally, you
1630 don't need to pay attention to this. But there are cases where it
1631 doesn't do the right thing automatically.
1635 If you update the system's header files, such as by installing a new
1636 system version, the fixed header files of GCC are not automatically
1637 updated. The easiest way to update them is to reinstall GCC. (If
1638 you want to be clever, look in the makefile and you can find a
1642 On some systems, in particular SunOS 4, header file directories contain
1643 machine-specific symbolic links in certain places. This makes it
1644 possible to share most of the header files among hosts running the
1645 same version of SunOS 4 on different machine models.
1647 The programs that fix the header files do not understand this special
1648 way of using symbolic links; therefore, the directory of fixed header
1649 files is good only for the machine model used to build it.
1651 In SunOS 4, only programs that look inside the kernel will notice the
1652 difference between machine models. Therefore, for most purposes, you
1653 need not be concerned about this.
1655 It is possible to make separate sets of fixed header files for the
1656 different machine models, and arrange a structure of symbolic links so
1657 as to use the proper set, but you'll have to do this by hand.
1660 On Lynxos, GCC by default does not fix the header files. This is
1661 because bugs in the shell cause the @code{fixincludes} script to fail.
1663 This means you will encounter problems due to bugs in the system header
1664 files. It may be no comfort that they aren't GCC's fault, but it
1665 does mean that there's nothing for us to do about them.
1668 @node Standard Libraries
1669 @section Standard Libraries
1671 GCC by itself attempts to be a conforming freestanding implementation.
1672 @xref{Standards,,Language Standards Supported by GCC}, for details of
1673 what this means. Beyond the library facilities required of such an
1674 implementation, the rest of the C library is supplied by the vendor of
1675 the operating system. If that C library doesn't conform to the C
1676 standards, then your programs might get warnings (especially when using
1677 @samp{-Wall}) that you don't expect.
1679 For example, the @code{sprintf} function on SunOS 4.1.3 returns
1680 @code{char *} while the C standard says that @code{sprintf} returns an
1681 @code{int}. The @code{fixincludes} program could make the prototype for
1682 this function match the Standard, but that would be wrong, since the
1683 function will still return @code{char *}.
1685 If you need a Standard compliant library, then you need to find one, as
1686 GCC does not provide one. The GNU C library (called @code{glibc})
1687 provides ISO C, POSIX, BSD, SystemV and X/Open compatibility for
1688 GNU/Linux and HURD-based GNU systems; no recent version of it supports
1689 other systems, though some very old versions did. Version 2.2 of the
1690 GNU C library includes nearly complete C99 support. You could also ask
1691 your operating system vendor if newer libraries are available.
1693 @node Disappointments
1694 @section Disappointments and Misunderstandings
1696 These problems are perhaps regrettable, but we don't know any practical
1701 Certain local variables aren't recognized by debuggers when you compile
1704 This occurs because sometimes GCC optimizes the variable out of
1705 existence. There is no way to tell the debugger how to compute the
1706 value such a variable ``would have had'', and it is not clear that would
1707 be desirable anyway. So GCC simply does not mention the eliminated
1708 variable when it writes debugging information.
1710 You have to expect a certain amount of disagreement between the
1711 executable and your source code, when you use optimization.
1713 @cindex conflicting types
1714 @cindex scope of declaration
1716 Users often think it is a bug when GCC reports an error for code
1720 int foo (struct mumble *);
1722 struct mumble @{ @dots{} @};
1724 int foo (struct mumble *x)
1728 This code really is erroneous, because the scope of @code{struct
1729 mumble} in the prototype is limited to the argument list containing it.
1730 It does not refer to the @code{struct mumble} defined with file scope
1731 immediately below---they are two unrelated types with similar names in
1734 But in the definition of @code{foo}, the file-scope type is used
1735 because that is available to be inherited. Thus, the definition and
1736 the prototype do not match, and you get an error.
1738 This behavior may seem silly, but it's what the ISO standard specifies.
1739 It is easy enough for you to make your code work by moving the
1740 definition of @code{struct mumble} above the prototype. It's not worth
1741 being incompatible with ISO C just to avoid an error for the example
1745 Accesses to bitfields even in volatile objects works by accessing larger
1746 objects, such as a byte or a word. You cannot rely on what size of
1747 object is accessed in order to read or write the bitfield; it may even
1748 vary for a given bitfield according to the precise usage.
1750 If you care about controlling the amount of memory that is accessed, use
1751 volatile but do not use bitfields.
1754 GCC comes with shell scripts to fix certain known problems in system
1755 header files. They install corrected copies of various header files in
1756 a special directory where only GCC will normally look for them. The
1757 scripts adapt to various systems by searching all the system header
1758 files for the problem cases that we know about.
1760 If new system header files are installed, nothing automatically arranges
1761 to update the corrected header files. You will have to reinstall GCC
1762 to fix the new header files. More specifically, go to the build
1763 directory and delete the files @file{stmp-fixinc} and
1764 @file{stmp-headers}, and the subdirectory @code{include}; then do
1765 @samp{make install} again.
1768 @cindex floating point precision
1769 On 68000 and x86 systems, for instance, you can get paradoxical results
1770 if you test the precise values of floating point numbers. For example,
1771 you can find that a floating point value which is not a NaN is not equal
1772 to itself. This results from the fact that the floating point registers
1773 hold a few more bits of precision than fit in a @code{double} in memory.
1774 Compiled code moves values between memory and floating point registers
1775 at its convenience, and moving them into memory truncates them.
1777 You can partially avoid this problem by using the @samp{-ffloat-store}
1778 option (@pxref{Optimize Options}).
1781 On the MIPS, variable argument functions using @file{varargs.h}
1782 cannot have a floating point value for the first argument. The
1783 reason for this is that in the absence of a prototype in scope,
1784 if the first argument is a floating point, it is passed in a
1785 floating point register, rather than an integer register.
1787 If the code is rewritten to use the ISO standard @file{stdarg.h}
1788 method of variable arguments, and the prototype is in scope at
1789 the time of the call, everything will work fine.
1792 On the H8/300 and H8/300H, variable argument functions must be
1793 implemented using the ISO standard @file{stdarg.h} method of
1794 variable arguments. Furthermore, calls to functions using @file{stdarg.h}
1795 variable arguments must have a prototype for the called function
1796 in scope at the time of the call.
1799 @node C++ Misunderstandings
1800 @section Common Misunderstandings with GNU C++
1802 @cindex misunderstandings in C++
1803 @cindex surprises in C++
1804 @cindex C++ misunderstandings
1805 C++ is a complex language and an evolving one, and its standard
1806 definition (the ISO C++ standard) was only recently completed. As a
1807 result, your C++ compiler may occasionally surprise you, even when its
1808 behavior is correct. This section discusses some areas that frequently
1809 give rise to questions of this sort.
1812 * Static Definitions:: Static member declarations are not definitions
1813 * Temporaries:: Temporaries may vanish before you expect
1814 * Copy Assignment:: Copy Assignment operators copy virtual bases twice
1817 @node Static Definitions
1818 @subsection Declare @emph{and} Define Static Members
1820 @cindex C++ static data, declaring and defining
1821 @cindex static data in C++, declaring and defining
1822 @cindex declaring static data in C++
1823 @cindex defining static data in C++
1824 When a class has static data members, it is not enough to @emph{declare}
1825 the static member; you must also @emph{define} it. For example:
1836 This declaration only establishes that the class @code{Foo} has an
1837 @code{int} named @code{Foo::bar}, and a member function named
1838 @code{Foo::method}. But you still need to define @emph{both}
1839 @code{method} and @code{bar} elsewhere. According to the ISO
1840 standard, you must supply an initializer in one (and only one) source
1847 Other C++ compilers may not correctly implement the standard behavior.
1848 As a result, when you switch to @code{g++} from one of these compilers,
1849 you may discover that a program that appeared to work correctly in fact
1850 does not conform to the standard: @code{g++} reports as undefined
1851 symbols any static data members that lack definitions.
1854 @subsection Temporaries May Vanish Before You Expect
1856 @cindex temporaries, lifetime of
1857 @cindex portions of temporary objects, pointers to
1858 It is dangerous to use pointers or references to @emph{portions} of a
1859 temporary object. The compiler may very well delete the object before
1860 you expect it to, leaving a pointer to garbage. The most common place
1861 where this problem crops up is in classes like string classes,
1862 especially ones that define a conversion function to type @code{char *}
1863 or @code{const char *} -- which is one reason why the standard
1864 @code{string} class requires you to call the @code{c_str} member
1865 function. However, any class that returns a pointer to some internal
1866 structure is potentially subject to this problem.
1868 For example, a program may use a function @code{strfunc} that returns
1869 @code{string} objects, and another function @code{charfunc} that
1870 operates on pointers to @code{char}:
1874 void charfunc (const char *);
1879 const char *p = strfunc().c_str();
1888 In this situation, it may seem reasonable to save a pointer to the C
1889 string returned by the @code{c_str} member function and use that rather
1890 than call @code{c_str} repeatedly. However, the temporary string
1891 created by the call to @code{strfunc} is destroyed after @code{p} is
1892 initialized, at which point @code{p} is left pointing to freed memory.
1894 Code like this may run successfully under some other compilers,
1895 particularly obsolete cfront-based compilers that delete temporaries
1896 along with normal local variables. However, the GNU C++ behavior is
1897 standard-conforming, so if your program depends on late destruction of
1898 temporaries it is not portable.
1900 The safe way to write such code is to give the temporary a name, which
1901 forces it to remain until the end of the scope of the name. For
1905 string& tmp = strfunc ();
1906 charfunc (tmp.c_str ());
1909 @node Copy Assignment
1910 @subsection Implicit Copy-Assignment for Virtual Bases
1912 When a base class is virtual, only one subobject of the base class
1913 belongs to each full object. Also, the constructors and destructors are
1914 invoked only once, and called from the most-derived class. However, such
1915 objects behave unspecified when being assigned. For example:
1920 Base(char *n) : name(strdup(n))@{@}
1921 Base& operator= (const Base& other)@{
1923 name = strdup (other.name);
1927 struct A:virtual Base@{
1932 struct B:virtual Base@{
1937 struct Derived:public A, public B@{
1938 Derived():Base("Derived")@{@}
1941 void func(Derived &d1, Derived &d2)
1947 The C++ standard specifies that @samp{Base::Base} is only called once
1948 when constructing or copy-constructing a Derived object. It is
1949 unspecified whether @samp{Base::operator=} is called more than once when
1950 the implicit copy-assignment for Derived objects is invoked (as it is
1951 inside @samp{func} in the example).
1953 g++ implements the "intuitive" algorithm for copy-assignment: assign all
1954 direct bases, then assign all members. In that algorithm, the virtual
1955 base subobject can be encountered many times. In the example, copying
1956 proceeds in the following order: @samp{val}, @samp{name} (via
1957 @code{strdup}), @samp{bval}, and @samp{name} again.
1959 If application code relies on copy-assignment, a user-defined
1960 copy-assignment operator removes any uncertainties. With such an
1961 operator, the application can define whether and how the virtual base
1962 subobject is assigned.
1964 @node Protoize Caveats
1965 @section Caveats of using @code{protoize}
1967 The conversion programs @code{protoize} and @code{unprotoize} can
1968 sometimes change a source file in a way that won't work unless you
1973 @code{protoize} can insert references to a type name or type tag before
1974 the definition, or in a file where they are not defined.
1976 If this happens, compiler error messages should show you where the new
1977 references are, so fixing the file by hand is straightforward.
1980 There are some C constructs which @code{protoize} cannot figure out.
1981 For example, it can't determine argument types for declaring a
1982 pointer-to-function variable; this you must do by hand. @code{protoize}
1983 inserts a comment containing @samp{???} each time it finds such a
1984 variable; so you can find all such variables by searching for this
1985 string. ISO C does not require declaring the argument types of
1986 pointer-to-function types.
1989 Using @code{unprotoize} can easily introduce bugs. If the program
1990 relied on prototypes to bring about conversion of arguments, these
1991 conversions will not take place in the program without prototypes.
1992 One case in which you can be sure @code{unprotoize} is safe is when
1993 you are removing prototypes that were made with @code{protoize}; if
1994 the program worked before without any prototypes, it will work again
1997 You can find all the places where this problem might occur by compiling
1998 the program with the @samp{-Wconversion} option. It prints a warning
1999 whenever an argument is converted.
2002 Both conversion programs can be confused if there are macro calls in and
2003 around the text to be converted. In other words, the standard syntax
2004 for a declaration or definition must not result from expanding a macro.
2005 This problem is inherent in the design of C and cannot be fixed. If
2006 only a few functions have confusing macro calls, you can easily convert
2010 @code{protoize} cannot get the argument types for a function whose
2011 definition was not actually compiled due to preprocessing conditionals.
2012 When this happens, @code{protoize} changes nothing in regard to such
2013 a function. @code{protoize} tries to detect such instances and warn
2016 You can generally work around this problem by using @code{protoize} step
2017 by step, each time specifying a different set of @samp{-D} options for
2018 compilation, until all of the functions have been converted. There is
2019 no automatic way to verify that you have got them all, however.
2022 Confusion may result if there is an occasion to convert a function
2023 declaration or definition in a region of source code where there is more
2024 than one formal parameter list present. Thus, attempts to convert code
2025 containing multiple (conditionally compiled) versions of a single
2026 function header (in the same vicinity) may not produce the desired (or
2029 If you plan on converting source files which contain such code, it is
2030 recommended that you first make sure that each conditionally compiled
2031 region of source code which contains an alternative function header also
2032 contains at least one additional follower token (past the final right
2033 parenthesis of the function header). This should circumvent the
2037 @code{unprotoize} can become confused when trying to convert a function
2038 definition or declaration which contains a declaration for a
2039 pointer-to-function formal argument which has the same name as the
2040 function being defined or declared. We recommend you avoid such choices
2041 of formal parameter names.
2044 You might also want to correct some of the indentation by hand and break
2045 long lines. (The conversion programs don't write lines longer than
2046 eighty characters in any case.)
2050 @section Certain Changes We Don't Want to Make
2052 This section lists changes that people frequently request, but which
2053 we do not make because we think GCC is better without them.
2057 Checking the number and type of arguments to a function which has an
2058 old-fashioned definition and no prototype.
2060 Such a feature would work only occasionally---only for calls that appear
2061 in the same file as the called function, following the definition. The
2062 only way to check all calls reliably is to add a prototype for the
2063 function. But adding a prototype eliminates the motivation for this
2064 feature. So the feature is not worthwhile.
2067 Warning about using an expression whose type is signed as a shift count.
2069 Shift count operands are probably signed more often than unsigned.
2070 Warning about this would cause far more annoyance than good.
2073 Warning about assigning a signed value to an unsigned variable.
2075 Such assignments must be very common; warning about them would cause
2076 more annoyance than good.
2079 Warning when a non-void function value is ignored.
2081 Coming as I do from a Lisp background, I balk at the idea that there is
2082 something dangerous about discarding a value. There are functions that
2083 return values which some callers may find useful; it makes no sense to
2084 clutter the program with a cast to @code{void} whenever the value isn't
2088 Assuming (for optimization) that the address of an external symbol is
2091 This assumption is false on certain systems when @samp{#pragma weak} is
2095 Making @samp{-fshort-enums} the default.
2097 This would cause storage layout to be incompatible with most other C
2098 compilers. And it doesn't seem very important, given that you can get
2099 the same result in other ways. The case where it matters most is when
2100 the enumeration-valued object is inside a structure, and in that case
2101 you can specify a field width explicitly.
2104 Making bitfields unsigned by default on particular machines where ``the
2105 ABI standard'' says to do so.
2107 The ISO C standard leaves it up to the implementation whether a bitfield
2108 declared plain @code{int} is signed or not. This in effect creates two
2109 alternative dialects of C.
2111 The GNU C compiler supports both dialects; you can specify the signed
2112 dialect with @samp{-fsigned-bitfields} and the unsigned dialect with
2113 @samp{-funsigned-bitfields}. However, this leaves open the question of
2114 which dialect to use by default.
2116 Currently, the preferred dialect makes plain bitfields signed, because
2117 this is simplest. Since @code{int} is the same as @code{signed int} in
2118 every other context, it is cleanest for them to be the same in bitfields
2121 Some computer manufacturers have published Application Binary Interface
2122 standards which specify that plain bitfields should be unsigned. It is
2123 a mistake, however, to say anything about this issue in an ABI. This is
2124 because the handling of plain bitfields distinguishes two dialects of C.
2125 Both dialects are meaningful on every type of machine. Whether a
2126 particular object file was compiled using signed bitfields or unsigned
2127 is of no concern to other object files, even if they access the same
2128 bitfields in the same data structures.
2130 A given program is written in one or the other of these two dialects.
2131 The program stands a chance to work on most any machine if it is
2132 compiled with the proper dialect. It is unlikely to work at all if
2133 compiled with the wrong dialect.
2135 Many users appreciate the GNU C compiler because it provides an
2136 environment that is uniform across machines. These users would be
2137 inconvenienced if the compiler treated plain bitfields differently on
2140 Occasionally users write programs intended only for a particular machine
2141 type. On these occasions, the users would benefit if the GNU C compiler
2142 were to support by default the same dialect as the other compilers on
2143 that machine. But such applications are rare. And users writing a
2144 program to run on more than one type of machine cannot possibly benefit
2145 from this kind of compatibility.
2147 This is why GCC does and will treat plain bitfields in the same
2148 fashion on all types of machines (by default).
2150 There are some arguments for making bitfields unsigned by default on all
2151 machines. If, for example, this becomes a universal de facto standard,
2152 it would make sense for GCC to go along with it. This is something
2153 to be considered in the future.
2155 (Of course, users strongly concerned about portability should indicate
2156 explicitly in each bitfield whether it is signed or not. In this way,
2157 they write programs which have the same meaning in both C dialects.)
2160 Undefining @code{__STDC__} when @samp{-ansi} is not used.
2162 Currently, GCC defines @code{__STDC__} as long as you don't use
2163 @samp{-traditional}. This provides good results in practice.
2165 Programmers normally use conditionals on @code{__STDC__} to ask whether
2166 it is safe to use certain features of ISO C, such as function
2167 prototypes or ISO token concatenation. Since plain @samp{gcc} supports
2168 all the features of ISO C, the correct answer to these questions is
2171 Some users try to use @code{__STDC__} to check for the availability of
2172 certain library facilities. This is actually incorrect usage in an ISO
2173 C program, because the ISO C standard says that a conforming
2174 freestanding implementation should define @code{__STDC__} even though it
2175 does not have the library facilities. @samp{gcc -ansi -pedantic} is a
2176 conforming freestanding implementation, and it is therefore required to
2177 define @code{__STDC__}, even though it does not come with an ISO C
2180 Sometimes people say that defining @code{__STDC__} in a compiler that
2181 does not completely conform to the ISO C standard somehow violates the
2182 standard. This is illogical. The standard is a standard for compilers
2183 that claim to support ISO C, such as @samp{gcc -ansi}---not for other
2184 compilers such as plain @samp{gcc}. Whatever the ISO C standard says
2185 is relevant to the design of plain @samp{gcc} without @samp{-ansi} only
2186 for pragmatic reasons, not as a requirement.
2188 GCC normally defines @code{__STDC__} to be 1, and in addition
2189 defines @code{__STRICT_ANSI__} if you specify the @samp{-ansi} option.
2190 On some hosts, system include files use a different convention, where
2191 @code{__STDC__} is normally 0, but is 1 if the user specifies strict
2192 conformance to the C Standard. GCC follows the host convention when
2193 processing system include files, but when processing user files it follows
2194 the usual GNU C convention.
2197 Undefining @code{__STDC__} in C++.
2199 Programs written to compile with C++-to-C translators get the
2200 value of @code{__STDC__} that goes with the C compiler that is
2201 subsequently used. These programs must test @code{__STDC__}
2202 to determine what kind of C preprocessor that compiler uses:
2203 whether they should concatenate tokens in the ISO C fashion
2204 or in the traditional fashion.
2206 These programs work properly with GNU C++ if @code{__STDC__} is defined.
2207 They would not work otherwise.
2209 In addition, many header files are written to provide prototypes in ISO
2210 C but not in traditional C. Many of these header files can work without
2211 change in C++ provided @code{__STDC__} is defined. If @code{__STDC__}
2212 is not defined, they will all fail, and will all need to be changed to
2213 test explicitly for C++ as well.
2216 Deleting ``empty'' loops.
2218 Historically, GCC has not deleted ``empty'' loops under the
2219 assumption that the most likely reason you would put one in a program is
2220 to have a delay, so deleting them will not make real programs run any
2223 However, the rationale here is that optimization of a nonempty loop
2224 cannot produce an empty one, which holds for C but is not always the
2227 Moreover, with @samp{-funroll-loops} small ``empty'' loops are already
2228 removed, so the current behavior is both sub-optimal and inconsistent
2229 and will change in the future.
2232 Making side effects happen in the same order as in some other compiler.
2234 @cindex side effects, order of evaluation
2235 @cindex order of evaluation, side effects
2236 It is never safe to depend on the order of evaluation of side effects.
2237 For example, a function call like this may very well behave differently
2238 from one compiler to another:
2241 void func (int, int);
2247 There is no guarantee (in either the C or the C++ standard language
2248 definitions) that the increments will be evaluated in any particular
2249 order. Either increment might happen first. @code{func} might get the
2250 arguments @samp{2, 3}, or it might get @samp{3, 2}, or even @samp{2, 2}.
2253 Not allowing structures with volatile fields in registers.
2255 Strictly speaking, there is no prohibition in the ISO C standard
2256 against allowing structures with volatile fields in registers, but
2257 it does not seem to make any sense and is probably not what you wanted
2258 to do. So the compiler will give an error message in this case.
2261 Making certain warnings into errors by default.
2263 Some ISO C testsuites report failure when the compiler does not produce
2264 an error message for a certain program.
2266 ISO C requires a ``diagnostic'' message for certain kinds of invalid
2267 programs, but a warning is defined by GCC to count as a diagnostic. If
2268 GCC produces a warning but not an error, that is correct ISO C support.
2269 If test suites call this ``failure'', they should be run with the GCC
2270 option @samp{-pedantic-errors}, which will turn these warnings into
2275 @node Warnings and Errors
2276 @section Warning Messages and Error Messages
2278 @cindex error messages
2279 @cindex warnings vs errors
2280 @cindex messages, warning and error
2281 The GNU compiler can produce two kinds of diagnostics: errors and
2282 warnings. Each kind has a different purpose:
2286 @emph{Errors} report problems that make it impossible to compile your
2287 program. GCC reports errors with the source file name and line
2288 number where the problem is apparent.
2291 @emph{Warnings} report other unusual conditions in your code that
2292 @emph{may} indicate a problem, although compilation can (and does)
2293 proceed. Warning messages also report the source file name and line
2294 number, but include the text @samp{warning:} to distinguish them
2295 from error messages.
2298 Warnings may indicate danger points where you should check to make sure
2299 that your program really does what you intend; or the use of obsolete
2300 features; or the use of nonstandard features of GNU C or C++. Many
2301 warnings are issued only if you ask for them, with one of the @samp{-W}
2302 options (for instance, @samp{-Wall} requests a variety of useful
2305 GCC always tries to compile your program if possible; it never
2306 gratuitously rejects a program whose meaning is clear merely because
2307 (for instance) it fails to conform to a standard. In some cases,
2308 however, the C and C++ standards specify that certain extensions are
2309 forbidden, and a diagnostic @emph{must} be issued by a conforming
2310 compiler. The @samp{-pedantic} option tells GCC to issue warnings in
2311 such cases; @samp{-pedantic-errors} says to make them errors instead.
2312 This does not mean that @emph{all} non-ISO constructs get warnings
2315 @xref{Warning Options,,Options to Request or Suppress Warnings}, for
2316 more detail on these and related command-line options.
2319 @chapter Reporting Bugs
2321 @cindex reporting bugs
2323 Your bug reports play an essential role in making GCC reliable.
2325 When you encounter a problem, the first thing to do is to see if it is
2326 already known. @xref{Trouble}. If it isn't known, then you should
2329 Reporting a bug may help you by bringing a solution to your problem, or
2330 it may not. (If it does not, look in the service directory; see
2331 @ref{Service}.) In any case, the principal function of a bug report is
2332 to help the entire community by making the next version of GCC work
2333 better. Bug reports are your contribution to the maintenance of GCC.
2335 Since the maintainers are very overloaded, we cannot respond to every
2336 bug report. However, if the bug has not been fixed, we are likely to
2337 send you a patch and ask you to tell us whether it works.
2339 In order for a bug report to serve its purpose, you must include the
2340 information that makes for fixing the bug.
2343 * Criteria: Bug Criteria. Have you really found a bug?
2344 * Where: Bug Lists. Where to send your bug report.
2345 * Reporting: Bug Reporting. How to report a bug effectively.
2346 * GNATS: gccbug. You can use a bug reporting tool.
2347 * Patches: Sending Patches. How to send a patch for GCC.
2348 * Known: Trouble. Known problems.
2349 * Help: Service. Where to ask for help.
2352 @node Bug Criteria,Bug Lists,,Bugs
2353 @section Have You Found a Bug?
2354 @cindex bug criteria
2356 If you are not sure whether you have found a bug, here are some guidelines:
2359 @cindex fatal signal
2362 If the compiler gets a fatal signal, for any input whatever, that is a
2363 compiler bug. Reliable compilers never crash.
2365 @cindex invalid assembly code
2366 @cindex assembly code, invalid
2368 If the compiler produces invalid assembly code, for any input whatever
2369 (except an @code{asm} statement), that is a compiler bug, unless the
2370 compiler reports errors (not just warnings) which would ordinarily
2371 prevent the assembler from being run.
2373 @cindex undefined behavior
2374 @cindex undefined function value
2375 @cindex increment operators
2377 If the compiler produces valid assembly code that does not correctly
2378 execute the input source code, that is a compiler bug.
2380 However, you must double-check to make sure, because you may have run
2381 into an incompatibility between GNU C and traditional C
2382 (@pxref{Incompatibilities}). These incompatibilities might be considered
2383 bugs, but they are inescapable consequences of valuable features.
2385 Or you may have a program whose behavior is undefined, which happened
2386 by chance to give the desired results with another C or C++ compiler.
2388 For example, in many nonoptimizing compilers, you can write @samp{x;}
2389 at the end of a function instead of @samp{return x;}, with the same
2390 results. But the value of the function is undefined if @code{return}
2391 is omitted; it is not a bug when GCC produces different results.
2393 Problems often result from expressions with two increment operators,
2394 as in @code{f (*p++, *p++)}. Your previous compiler might have
2395 interpreted that expression the way you intended; GCC might
2396 interpret it another way. Neither compiler is wrong. The bug is
2399 After you have localized the error to a single source line, it should
2400 be easy to check for these things. If your program is correct and
2401 well defined, you have found a compiler bug.
2404 If the compiler produces an error message for valid input, that is a
2407 @cindex invalid input
2409 If the compiler does not produce an error message for invalid input,
2410 that is a compiler bug. However, you should note that your idea of
2411 ``invalid input'' might be my idea of ``an extension'' or ``support
2412 for traditional practice''.
2415 If you are an experienced user of one of the languages GCC supports, your
2416 suggestions for improvement of GCC are welcome in any case.
2419 @node Bug Lists,Bug Reporting,Bug Criteria,Bugs
2420 @section Where to Report Bugs
2421 @cindex bug report mailing lists
2422 @kindex gcc-bugs@@gcc.gnu.org or bug-gcc@@gnu.org
2423 Send bug reports for the GNU Compiler Collection to
2424 @email{gcc-bugs@@gcc.gnu.org}. In accordance with the GNU-wide
2425 convention, in which bug reports for tool ``foo'' are sent
2426 to @samp{bug-foo@@gnu.org}, the address @email{bug-gcc@@gnu.org}
2427 may also be used; it will forward to the address given above.
2429 Please read @uref{http://www.gnu.org/software/gcc/bugs.html} for
2430 bug reporting instructions before you post a bug report.
2432 Often people think of posting bug reports to the newsgroup instead of
2433 mailing them. This appears to work, but it has one problem which can be
2434 crucial: a newsgroup posting does not contain a mail path back to the
2435 sender. Thus, if maintainers need more information, they may be unable
2436 to reach you. For this reason, you should always send bug reports by
2437 mail to the proper mailing list.
2439 As a last resort, send bug reports on paper to:
2443 Free Software Foundation
2444 59 Temple Place - Suite 330
2445 Boston, MA 02111-1307, USA
2448 @node Bug Reporting,gccbug,Bug Lists,Bugs
2449 @section How to Report Bugs
2450 @cindex compiler bugs, reporting
2452 You may find additional and/or more up-to-date instructions at
2453 @uref{http://www.gnu.org/software/gcc/bugs.html}.
2455 The fundamental principle of reporting bugs usefully is this:
2456 @strong{report all the facts}. If you are not sure whether to state a
2457 fact or leave it out, state it!
2459 Often people omit facts because they think they know what causes the
2460 problem and they conclude that some details don't matter. Thus, you might
2461 assume that the name of the variable you use in an example does not matter.
2462 Well, probably it doesn't, but one cannot be sure. Perhaps the bug is a
2463 stray memory reference which happens to fetch from the location where that
2464 name is stored in memory; perhaps, if the name were different, the contents
2465 of that location would fool the compiler into doing the right thing despite
2466 the bug. Play it safe and give a specific, complete example. That is the
2467 easiest thing for you to do, and the most helpful.
2469 Keep in mind that the purpose of a bug report is to enable someone to
2470 fix the bug if it is not known. It isn't very important what happens if
2471 the bug is already known. Therefore, always write your bug reports on
2472 the assumption that the bug is not known.
2474 Sometimes people give a few sketchy facts and ask, ``Does this ring a
2475 bell?'' This cannot help us fix a bug, so it is basically useless. We
2476 respond by asking for enough details to enable us to investigate.
2477 You might as well expedite matters by sending them to begin with.
2479 Try to make your bug report self-contained. If we have to ask you for
2480 more information, it is best if you include all the previous information
2481 in your response, as well as the information that was missing.
2483 Please report each bug in a separate message. This makes it easier for
2484 us to track which bugs have been fixed and to forward your bugs reports
2485 to the appropriate maintainer.
2487 To enable someone to investigate the bug, you should include all these
2492 The version of GCC. You can get this by running it with the
2495 Without this, we won't know whether there is any point in looking for
2496 the bug in the current version of GCC.
2499 A complete input file that will reproduce the bug. If the bug is in the
2500 C preprocessor, send a source file and any header files that it
2501 requires. If the bug is in the compiler proper (@file{cc1}), send the
2502 preprocessor output generated by adding @samp{-save-temps} to the
2503 compilation command (@pxref{Debugging Options}). When you do this, use
2504 the same @samp{-I}, @samp{-D} or @samp{-U} options that you used in
2505 actual compilation. Then send the @var{input}.i or @var{input}.ii files
2508 A single statement is not enough of an example. In order to compile it,
2509 it must be embedded in a complete file of compiler input; and the bug
2510 might depend on the details of how this is done.
2512 Without a real example one can compile, all anyone can do about your bug
2513 report is wish you luck. It would be futile to try to guess how to
2514 provoke the bug. For example, bugs in register allocation and reloading
2515 frequently depend on every little detail of the function they happen in.
2517 Even if the input file that fails comes from a GNU program, you should
2518 still send the complete test case. Don't ask the GCC maintainers to
2519 do the extra work of obtaining the program in question---they are all
2520 overworked as it is. Also, the problem may depend on what is in the
2521 header files on your system; it is unreliable for the GCC maintainers
2522 to try the problem with the header files available to them. By sending
2523 CPP output, you can eliminate this source of uncertainty and save us
2524 a certain percentage of wild goose chases.
2527 The command arguments you gave GCC to compile that example
2528 and observe the bug. For example, did you use @samp{-O}? To guarantee
2529 you won't omit something important, list all the options.
2531 If we were to try to guess the arguments, we would probably guess wrong
2532 and then we would not encounter the bug.
2535 The type of machine you are using, and the operating system name and
2539 The operands you gave to the @code{configure} command when you installed
2543 A complete list of any modifications you have made to the compiler
2544 source. (We don't promise to investigate the bug unless it happens in
2545 an unmodified compiler. But if you've made modifications and don't tell
2546 us, then you are sending us on a wild goose chase.)
2548 Be precise about these changes. A description in English is not
2549 enough---send a context diff for them.
2551 Adding files of your own (such as a machine description for a machine we
2552 don't support) is a modification of the compiler source.
2555 Details of any other deviations from the standard procedure for installing
2559 A description of what behavior you observe that you believe is
2560 incorrect. For example, ``The compiler gets a fatal signal,'' or,
2561 ``The assembler instruction at line 208 in the output is incorrect.''
2563 Of course, if the bug is that the compiler gets a fatal signal, then one
2564 can't miss it. But if the bug is incorrect output, the maintainer might
2565 not notice unless it is glaringly wrong. None of us has time to study
2566 all the assembler code from a 50-line C program just on the chance that
2567 one instruction might be wrong. We need @emph{you} to do this part!
2569 Even if the problem you experience is a fatal signal, you should still
2570 say so explicitly. Suppose something strange is going on, such as, your
2571 copy of the compiler is out of synch, or you have encountered a bug in
2572 the C library on your system. (This has happened!) Your copy might
2573 crash and the copy here would not. If you @i{said} to expect a crash,
2574 then when the compiler here fails to crash, we would know that the bug
2575 was not happening. If you don't say to expect a crash, then we would
2576 not know whether the bug was happening. We would not be able to draw
2577 any conclusion from our observations.
2579 If the problem is a diagnostic when compiling GCC with some other
2580 compiler, say whether it is a warning or an error.
2582 Often the observed symptom is incorrect output when your program is run.
2583 Sad to say, this is not enough information unless the program is short
2584 and simple. None of us has time to study a large program to figure out
2585 how it would work if compiled correctly, much less which line of it was
2586 compiled wrong. So you will have to do that. Tell us which source line
2587 it is, and what incorrect result happens when that line is executed. A
2588 person who understands the program can find this as easily as finding a
2589 bug in the program itself.
2592 If you send examples of assembler code output from GCC,
2593 please use @samp{-g} when you make them. The debugging information
2594 includes source line numbers which are essential for correlating the
2595 output with the input.
2598 If you wish to mention something in the GCC source, refer to it by
2599 context, not by line number.
2601 The line numbers in the development sources don't match those in your
2602 sources. Your line numbers would convey no useful information to the
2606 Additional information from a debugger might enable someone to find a
2607 problem on a machine which he does not have available. However, you
2608 need to think when you collect this information if you want it to have
2609 any chance of being useful.
2611 @cindex backtrace for bug reports
2612 For example, many people send just a backtrace, but that is never
2613 useful by itself. A simple backtrace with arguments conveys little
2614 about GCC because the compiler is largely data-driven; the same
2615 functions are called over and over for different RTL insns, doing
2616 different things depending on the details of the insn.
2618 Most of the arguments listed in the backtrace are useless because they
2619 are pointers to RTL list structure. The numeric values of the
2620 pointers, which the debugger prints in the backtrace, have no
2621 significance whatever; all that matters is the contents of the objects
2622 they point to (and most of the contents are other such pointers).
2624 In addition, most compiler passes consist of one or more loops that
2625 scan the RTL insn sequence. The most vital piece of information about
2626 such a loop---which insn it has reached---is usually in a local variable,
2630 What you need to provide in addition to a backtrace are the values of
2631 the local variables for several stack frames up. When a local
2632 variable or an argument is an RTX, first print its value and then use
2633 the GDB command @code{pr} to print the RTL expression that it points
2634 to. (If GDB doesn't run on your machine, use your debugger to call
2635 the function @code{debug_rtx} with the RTX as an argument.) In
2636 general, whenever a variable is a pointer, its value is no use
2637 without the data it points to.
2640 Here are some things that are not necessary:
2644 A description of the envelope of the bug.
2646 Often people who encounter a bug spend a lot of time investigating
2647 which changes to the input file will make the bug go away and which
2648 changes will not affect it.
2650 This is often time consuming and not very useful, because the way we
2651 will find the bug is by running a single example under the debugger with
2652 breakpoints, not by pure deduction from a series of examples. You might
2653 as well save your time for something else.
2655 Of course, if you can find a simpler example to report @emph{instead} of
2656 the original one, that is a convenience. Errors in the output will be
2657 easier to spot, running under the debugger will take less time, etc.
2658 Most GCC bugs involve just one function, so the most straightforward
2659 way to simplify an example is to delete all the function definitions
2660 except the one where the bug occurs. Those earlier in the file may be
2661 replaced by external declarations if the crucial function depends on
2662 them. (Exception: inline functions may affect compilation of functions
2663 defined later in the file.)
2665 However, simplification is not vital; if you don't want to do this,
2666 report the bug anyway and send the entire test case you used.
2669 In particular, some people insert conditionals @samp{#ifdef BUG} around
2670 a statement which, if removed, makes the bug not happen. These are just
2671 clutter; we won't pay any attention to them anyway. Besides, you should
2672 send us cpp output, and that can't have conditionals.
2675 A patch for the bug.
2677 A patch for the bug is useful if it is a good one. But don't omit the
2678 necessary information, such as the test case, on the assumption that a
2679 patch is all we need. We might see problems with your patch and decide
2680 to fix the problem another way, or we might not understand it at all.
2682 Sometimes with a program as complicated as GCC it is very hard to
2683 construct an example that will make the program follow a certain path
2684 through the code. If you don't send the example, we won't be able to
2685 construct one, so we won't be able to verify that the bug is fixed.
2687 And if we can't understand what bug you are trying to fix, or why your
2688 patch should be an improvement, we won't install it. A test case will
2689 help us to understand.
2691 @xref{Sending Patches}, for guidelines on how to make it easy for us to
2692 understand and install your patches.
2695 A guess about what the bug is or what it depends on.
2697 Such guesses are usually wrong. Even I can't guess right about such
2698 things without first using the debugger to find the facts.
2703 We have no way of examining a core dump for your type of machine
2704 unless we have an identical system---and if we do have one,
2705 we should be able to reproduce the crash ourselves.
2708 @node gccbug,Sending Patches, Bug Reporting, Bugs
2709 @section The gccbug script
2710 @cindex gccbug script
2712 To simplify creation of bug reports, and to allow better tracking of
2713 reports, we use the GNATS bug tracking system. Part of that system is
2714 the @code{gccbug} script. This is a Unix shell script, so you need a
2715 shell to run it. It is normally installed in the same directory where
2716 @code{gcc} is installed.
2718 The gccbug script is derived from send-pr, @pxref{using
2719 send-pr,,Creating new Problem Reports,send-pr,Reporting Problems}. When
2720 invoked, it starts a text editor so you can fill out the various fields
2721 of the report. When the you quit the editor, the report is automatically
2722 send to the bug reporting address.
2724 A number of fields in this bug report form are specific to GCC, and are
2729 @cindex @code{Category} field
2730 @cindex @code{>Category:}
2732 The category of a GCC problem can be one of the following:
2736 A problem with the C compiler proper.
2740 A problem with the C++ compiler.
2744 A problem with the Fortran 77.
2747 A problem with the Java compiler.
2750 A problem with the Objective C compiler.
2753 A problem with the C++ standard library.
2756 A problem with the Fortran 77 library.
2759 A problem with the Objective C library.
2762 The problem occurs only when generating optimized code.
2765 The problem occurs only when generating code for debugging.
2768 The problem is specific to the target architecture.
2771 The problem is independent from target architecture and programming
2775 It is a problem in some other part of the GCC software.
2778 There is a problem with the GCC home page.
2782 @cindex @code{Class} field
2783 @cindex @code{>Class:}
2785 The class of a problem can be one of the following:
2788 @cindex @emph{doc-bug} class
2790 A problem with the documentation.
2792 @cindex @emph{accepts-illegal} class
2793 @item accepts-illegal
2794 GCC fails to reject erroneous code.
2796 @cindex @emph{rejects-legal} class
2798 GCC gives an error message for correct code.
2800 @cindex @emph{wrong-code} class
2802 The machine code generated by gcc is incorrect.
2804 @cindex @emph{ice-on-legal-code} class
2805 @item ice-on-legal-code
2806 GCC gives an Internal Compiler Error (ICE) for correct code.
2808 @cindex @emph{ice-on-illegal-code} class
2809 @item ice-on-illegal-code
2810 GCC gives an ICE instead of reporting an error
2812 @cindex @emph{pessimizes-code} class
2813 @item pessimizes-code
2814 GCC misses an important optimization opportunity.
2816 @cindex @emph{sw-bug} class
2818 A general product problem. (@samp{sw} stands for ``software''.)
2820 @cindex @emph{change-request} class
2821 @item change-request
2822 A request for a change in behavior, etc.
2824 @cindex @emph{support} class
2826 A support problem or question.
2828 @cindex @emph{duplicate} class
2829 @item duplicate (@var{pr-number})
2830 Duplicate PR. @var{pr-number} should be the number of the original PR.
2833 The default is @samp{sw-bug}.
2839 @node Sending Patches,, gccbug, Bugs
2840 @section Sending Patches for GCC
2842 If you would like to write bug fixes or improvements for the GNU C
2843 compiler, that is very helpful. Send suggested fixes to the patches
2844 mailing list, @email{gcc-patches@@gcc.gnu.org}.
2846 Please follow these guidelines so we can study your patches efficiently.
2847 If you don't follow these guidelines, your information might still be
2848 useful, but using it will take extra work. Maintaining GNU C is a lot
2849 of work in the best of circumstances, and we can't keep up unless you do
2854 Send an explanation with your changes of what problem they fix or what
2855 improvement they bring about. For a bug fix, just include a copy of the
2856 bug report, and explain why the change fixes the bug.
2858 (Referring to a bug report is not as good as including it, because then
2859 we will have to look it up, and we have probably already deleted it if
2860 we've already fixed the bug.)
2863 Always include a proper bug report for the problem you think you have
2864 fixed. We need to convince ourselves that the change is right before
2865 installing it. Even if it is right, we might have trouble judging it if
2866 we don't have a way to reproduce the problem.
2869 Include all the comments that are appropriate to help people reading the
2870 source in the future understand why this change was needed.
2873 Don't mix together changes made for different reasons.
2874 Send them @emph{individually}.
2876 If you make two changes for separate reasons, then we might not want to
2877 install them both. We might want to install just one. If you send them
2878 all jumbled together in a single set of diffs, we have to do extra work
2879 to disentangle them---to figure out which parts of the change serve
2880 which purpose. If we don't have time for this, we might have to ignore
2881 your changes entirely.
2883 If you send each change as soon as you have written it, with its own
2884 explanation, then the two changes never get tangled up, and we can
2885 consider each one properly without any extra work to disentangle them.
2887 Ideally, each change you send should be impossible to subdivide into
2888 parts that we might want to consider separately, because each of its
2889 parts gets its motivation from the other parts.
2892 Send each change as soon as that change is finished. Sometimes people
2893 think they are helping us by accumulating many changes to send them all
2894 together. As explained above, this is absolutely the worst thing you
2897 Since you should send each change separately, you might as well send it
2898 right away. That gives us the option of installing it immediately if it
2902 Use @samp{diff -c} to make your diffs. Diffs without context are hard
2903 for us to install reliably. More than that, they make it hard for us to
2904 study the diffs to decide whether we want to install them. Unidiff
2905 format is better than contextless diffs, but not as easy to read as
2908 If you have GNU diff, use @samp{diff -cp}, which shows the name of the
2909 function that each change occurs in.
2912 Write the change log entries for your changes. We get lots of changes,
2913 and we don't have time to do all the change log writing ourselves.
2915 Read the @file{ChangeLog} file to see what sorts of information to put
2916 in, and to learn the style that we use. The purpose of the change log
2917 is to show people where to find what was changed. So you need to be
2918 specific about what functions you changed; in large functions, it's
2919 often helpful to indicate where within the function the change was.
2921 On the other hand, once you have shown people where to find the change,
2922 you need not explain its purpose. Thus, if you add a new function, all
2923 you need to say about it is that it is new. If you feel that the
2924 purpose needs explaining, it probably does---but the explanation will be
2925 much more useful if you put it in comments in the code.
2927 If you would like your name to appear in the header line for who made
2928 the change, send us the header line.
2931 When you write the fix, keep in mind that we can't install a change that
2932 would break other systems.
2934 People often suggest fixing a problem by changing machine-independent
2935 files such as @file{toplev.c} to do something special that a particular
2936 system needs. Sometimes it is totally obvious that such changes would
2937 break GCC for almost all users. We can't possibly make a change like
2938 that. At best it might tell us how to write another patch that would
2939 solve the problem acceptably.
2941 Sometimes people send fixes that @emph{might} be an improvement in
2942 general---but it is hard to be sure of this. It's hard to install
2943 such changes because we have to study them very carefully. Of course,
2944 a good explanation of the reasoning by which you concluded the change
2945 was correct can help convince us.
2947 The safest changes are changes to the configuration files for a
2948 particular machine. These are safe because they can't create new bugs
2951 Please help us keep up with the workload by designing the patch in a
2952 form that is good to install.
2956 @chapter How To Get Help with GCC
2958 If you need help installing, using or changing GCC, there are two
2963 Send a message to a suitable network mailing list. First try
2964 @email{gcc-help@@gcc.gnu.org} (for help installing or using GCC), and if
2965 that brings no response, try @email{gcc@@gcc.gnu.org}. For help
2966 changing GCC, ask @email{gcc@@gcc.gnu.org}. If you think you have found
2967 a bug in GCC, please report it following the instructions at
2968 @uref{http://gcc.gnu.org/bugs.html}.
2971 Look in the service directory for someone who might help you for a fee.
2972 The service directory is found at
2973 @uref{http://www.gnu.org/prep/service.html}.
2976 @c For further information, see
2977 @c @uref{http://gcc.gnu.org/cgi-bin/fom.cgi?file=12}.
2978 @c FIXME: this URL may be too volatile, this FAQ entry needs to move to
2979 @c the regular web pages before we can uncomment the reference.
2982 @chapter Contributing to GCC Development
2984 If you would like to help pretest GCC releases to assure they work well,
2985 our current development sources are available by CVS (see
2986 @uref{http://gcc.gnu.org/cvs.html}). Source and binary snapshots are
2987 also available for FTP; see @uref{http://gcc.gnu.org/snapshots.html}.
2988 Remember that snapshots of the current development sources may not be of
2989 production quality; if you find problems (whether compiler crashes,
2990 miscompiled code, poor optimization or any other problem), please report
2991 them following our bug reporting instructions at
2992 @uref{http://gcc.gnu.org/bugs.html}.
2994 If you would like to work on improvements to GCC, please read
2995 @uref{http://gcc.gnu.org/contribute.html} and
2996 @uref{http://gcc.gnu.org/contributewhy.html} for information on how to
2997 make useful contributions and avoid duplication of effort. Suggested
2998 projects are listed at @uref{http://gcc.gnu.org/projects/}.
3001 @chapter Using GCC on VMS
3003 @c prevent bad page break with this line
3004 Here is how to use GCC on VMS.
3007 * Include Files and VMS:: Where the preprocessor looks for the include files.
3008 * Global Declarations:: How to do globaldef, globalref and globalvalue with
3010 * VMS Misc:: Misc information.
3013 @node Include Files and VMS
3014 @section Include Files and VMS
3016 @cindex include files and VMS
3017 @cindex VMS and include files
3018 @cindex header files and VMS
3019 Due to the differences between the filesystems of Unix and VMS, GCC
3020 attempts to translate file names in @samp{#include} into names that VMS
3021 will understand. The basic strategy is to prepend a prefix to the
3022 specification of the include file, convert the whole filename to a VMS
3023 filename, and then try to open the file. GCC tries various prefixes
3024 one by one until one of them succeeds:
3028 The first prefix is the @samp{GNU_CC_INCLUDE:} logical name: this is
3029 where GNU C header files are traditionally stored. If you wish to store
3030 header files in non-standard locations, then you can assign the logical
3031 @samp{GNU_CC_INCLUDE} to be a search list, where each element of the
3032 list is suitable for use with a rooted logical.
3035 The next prefix tried is @samp{SYS$SYSROOT:[SYSLIB.]}. This is where
3036 VAX-C header files are traditionally stored.
3039 If the include file specification by itself is a valid VMS filename, the
3040 preprocessor then uses this name with no prefix in an attempt to open
3044 If the file specification is not a valid VMS filename (i.e. does not
3045 contain a device or a directory specifier, and contains a @samp{/}
3046 character), the preprocessor tries to convert it from Unix syntax to
3049 Conversion works like this: the first directory name becomes a device,
3050 and the rest of the directories are converted into VMS-format directory
3051 names. For example, the name @file{X11/foobar.h} is
3052 translated to @file{X11:[000000]foobar.h} or @file{X11:foobar.h},
3053 whichever one can be opened. This strategy allows you to assign a
3054 logical name to point to the actual location of the header files.
3057 If none of these strategies succeeds, the @samp{#include} fails.
3060 Include directives of the form:
3067 are a common source of incompatibility between VAX-C and GCC. VAX-C
3068 treats this much like a standard @code{#include <foobar.h>} directive.
3069 That is incompatible with the ISO C behavior implemented by GCC: to
3070 expand the name @code{foobar} as a macro. Macro expansion should
3071 eventually yield one of the two standard formats for @code{#include}:
3074 #include "@var{file}"
3075 #include <@var{file}>
3078 If you have this problem, the best solution is to modify the source to
3079 convert the @code{#include} directives to one of the two standard forms.
3080 That will work with either compiler. If you want a quick and dirty fix,
3081 define the file names as macros with the proper expansion, like this:
3084 #define stdio <stdio.h>
3088 This will work, as long as the name doesn't conflict with anything else
3091 Another source of incompatibility is that VAX-C assumes that:
3098 is actually asking for the file @file{foobar.h}. GCC does not
3099 make this assumption, and instead takes what you ask for literally;
3100 it tries to read the file @file{foobar}. The best way to avoid this
3101 problem is to always specify the desired file extension in your include
3104 GCC for VMS is distributed with a set of include files that is
3105 sufficient to compile most general purpose programs. Even though the
3106 GCC distribution does not contain header files to define constants
3107 and structures for some VMS system-specific functions, there is no
3108 reason why you cannot use GCC with any of these functions. You first
3109 may have to generate or create header files, either by using the public
3110 domain utility @code{UNSDL} (which can be found on a DECUS tape), or by
3111 extracting the relevant modules from one of the system macro libraries,
3112 and using an editor to construct a C header file.
3114 A @code{#include} file name cannot contain a DECNET node name. The
3115 preprocessor reports an I/O error if you attempt to use a node name,
3116 whether explicitly, or implicitly via a logical name.
3118 @node Global Declarations
3119 @section Global Declarations and VMS
3123 @findex GLOBALVALUEDEF
3124 @findex GLOBALVALUEREF
3125 GCC does not provide the @code{globalref}, @code{globaldef} and
3126 @code{globalvalue} keywords of VAX-C. You can get the same effect with
3127 an obscure feature of GAS, the GNU assembler. (This requires GAS
3128 version 1.39 or later.) The following macros allow you to use this
3129 feature in a fairly natural way:
3133 #define GLOBALREF(TYPE,NAME) \
3135 asm ("_$$PsectAttributes_GLOBALSYMBOL$$" #NAME)
3136 #define GLOBALDEF(TYPE,NAME,VALUE) \
3138 asm ("_$$PsectAttributes_GLOBALSYMBOL$$" #NAME) \
3140 #define GLOBALVALUEREF(TYPE,NAME) \
3141 const TYPE NAME[1] \
3142 asm ("_$$PsectAttributes_GLOBALVALUE$$" #NAME)
3143 #define GLOBALVALUEDEF(TYPE,NAME,VALUE) \
3144 const TYPE NAME[1] \
3145 asm ("_$$PsectAttributes_GLOBALVALUE$$" #NAME) \
3148 #define GLOBALREF(TYPE,NAME) \
3150 #define GLOBALDEF(TYPE,NAME,VALUE) \
3151 globaldef TYPE NAME = VALUE
3152 #define GLOBALVALUEDEF(TYPE,NAME,VALUE) \
3153 globalvalue TYPE NAME = VALUE
3154 #define GLOBALVALUEREF(TYPE,NAME) \
3155 globalvalue TYPE NAME
3160 (The @code{_$$PsectAttributes_GLOBALSYMBOL} prefix at the start of the
3161 name is removed by the assembler, after it has modified the attributes
3162 of the symbol). These macros are provided in the VMS binaries
3163 distribution in a header file @file{GNU_HACKS.H}. An example of the
3167 GLOBALREF (int, ijk);
3168 GLOBALDEF (int, jkl, 0);
3171 The macros @code{GLOBALREF} and @code{GLOBALDEF} cannot be used
3172 straightforwardly for arrays, since there is no way to insert the array
3173 dimension into the declaration at the right place. However, you can
3174 declare an array with these macros if you first define a typedef for the
3175 array type, like this:
3178 typedef int intvector[10];
3179 GLOBALREF (intvector, foo);
3182 Array and structure initializers will also break the macros; you can
3183 define the initializer to be a macro of its own, or you can expand the
3184 @code{GLOBALDEF} macro by hand. You may find a case where you wish to
3185 use the @code{GLOBALDEF} macro with a large array, but you are not
3186 interested in explicitly initializing each element of the array. In
3187 such cases you can use an initializer like: @code{@{0,@}}, which will
3188 initialize the entire array to @code{0}.
3190 A shortcoming of this implementation is that a variable declared with
3191 @code{GLOBALVALUEREF} or @code{GLOBALVALUEDEF} is always an array. For
3192 example, the declaration:
3195 GLOBALVALUEREF(int, ijk);
3199 declares the variable @code{ijk} as an array of type @code{int [1]}.
3200 This is done because a globalvalue is actually a constant; its ``value''
3201 is what the linker would normally consider an address. That is not how
3202 an integer value works in C, but it is how an array works. So treating
3203 the symbol as an array name gives consistent results---with the
3204 exception that the value seems to have the wrong type. @strong{Don't
3205 try to access an element of the array.} It doesn't have any elements.
3206 The array ``address'' may not be the address of actual storage.
3208 The fact that the symbol is an array may lead to warnings where the
3209 variable is used. Insert type casts to avoid the warnings. Here is an
3210 example; it takes advantage of the ISO C feature allowing macros that
3211 expand to use the same name as the macro itself.
3214 GLOBALVALUEREF (int, ss$_normal);
3215 GLOBALVALUEDEF (int, xyzzy,123);
3217 #define ss$_normal ((int) ss$_normal)
3218 #define xyzzy ((int) xyzzy)
3222 Don't use @code{globaldef} or @code{globalref} with a variable whose
3223 type is an enumeration type; this is not implemented. Instead, make the
3224 variable an integer, and use a @code{globalvaluedef} for each of the
3225 enumeration values. An example of this would be:
3229 GLOBALDEF (int, color, 0);
3230 GLOBALVALUEDEF (int, RED, 0);
3231 GLOBALVALUEDEF (int, BLUE, 1);
3232 GLOBALVALUEDEF (int, GREEN, 3);
3234 enum globaldef color @{RED, BLUE, GREEN = 3@};
3239 @section Other VMS Issues
3241 @cindex exit status and VMS
3242 @cindex return value of @code{main}
3243 @cindex @code{main} and the exit status
3244 GCC automatically arranges for @code{main} to return 1 by default if
3245 you fail to specify an explicit return value. This will be interpreted
3246 by VMS as a status code indicating a normal successful completion.
3247 Version 1 of GCC did not provide this default.
3249 GCC on VMS works only with the GNU assembler, GAS. You need version
3250 1.37 or later of GAS in order to produce value debugging information for
3251 the VMS debugger. Use the ordinary VMS linker with the object files
3254 @cindex shared VMS run time system
3255 @cindex @file{VAXCRTL}
3256 Under previous versions of GCC, the generated code would occasionally
3257 give strange results when linked to the sharable @file{VAXCRTL} library.
3258 Now this should work.
3260 A caveat for use of @code{const} global variables: the @code{const}
3261 modifier must be specified in every external declaration of the variable
3262 in all of the source files that use that variable. Otherwise the linker
3263 will issue warnings about conflicting attributes for the variable. Your
3264 program will still work despite the warnings, but the variable will be
3265 placed in writable storage.
3267 @cindex name augmentation
3268 @cindex case sensitivity and VMS
3269 @cindex VMS and case sensitivity
3270 Although the VMS linker does distinguish between upper and lower case
3271 letters in global symbols, most VMS compilers convert all such symbols
3272 into upper case and most run-time library routines also have upper case
3273 names. To be able to reliably call such routines, GCC (by means of
3274 the assembler GAS) converts global symbols into upper case like other
3275 VMS compilers. However, since the usual practice in C is to distinguish
3276 case, GCC (via GAS) tries to preserve usual C behavior by augmenting
3277 each name that is not all lower case. This means truncating the name
3278 to at most 23 characters and then adding more characters at the end
3279 which encode the case pattern of those 23. Names which contain at
3280 least one dollar sign are an exception; they are converted directly into
3281 upper case without augmentation.
3283 Name augmentation yields bad results for programs that use precompiled
3284 libraries (such as Xlib) which were generated by another compiler. You
3285 can use the compiler option @samp{/NOCASE_HACK} to inhibit augmentation;
3286 it makes external C functions and variables case-independent as is usual
3287 on VMS. Alternatively, you could write all references to the functions
3288 and variables in such libraries using lower case; this will work on VMS,
3289 but is not portable to other systems. The compiler option @samp{/NAMES}
3290 also provides control over global name handling.
3292 Function and variable names are handled somewhat differently with GNU
3293 C++. The GNU C++ compiler performs @dfn{name mangling} on function
3294 names, which means that it adds information to the function name to
3295 describe the data types of the arguments that the function takes. One
3296 result of this is that the name of a function can become very long.
3297 Since the VMS linker only recognizes the first 31 characters in a name,
3298 special action is taken to ensure that each function and variable has a
3299 unique name that can be represented in 31 characters.
3301 If the name (plus a name augmentation, if required) is less than 32
3302 characters in length, then no special action is performed. If the name
3303 is longer than 31 characters, the assembler (GAS) will generate a
3304 hash string based upon the function name, truncate the function name to
3305 23 characters, and append the hash string to the truncated name. If the
3306 @samp{/VERBOSE} compiler option is used, the assembler will print both
3307 the full and truncated names of each symbol that is truncated.
3309 The @samp{/NOCASE_HACK} compiler option should not be used when you are
3310 compiling programs that use libg++. libg++ has several instances of
3311 objects (i.e. @code{Filebuf} and @code{filebuf}) which become
3312 indistinguishable in a case-insensitive environment. This leads to
3313 cases where you need to inhibit augmentation selectively (if you were
3314 using libg++ and Xlib in the same program, for example). There is no
3315 special feature for doing this, but you can get the result by defining a
3316 macro for each mixed case symbol for which you wish to inhibit
3317 augmentation. The macro should expand into the lower case equivalent of
3318 itself. For example:
3321 #define StuDlyCapS studlycaps
3324 These macro definitions can be placed in a header file to minimize the
3325 number of changes to your source code.
3330 @chapter GCC and Portability
3332 @cindex GCC and portability
3334 The main goal of GCC was to make a good, fast compiler for machines in
3335 the class that the GNU system aims to run on: 32-bit machines that address
3336 8-bit bytes and have several general registers. Elegance, theoretical
3337 power and simplicity are only secondary.
3339 GCC gets most of the information about the target machine from a machine
3340 description which gives an algebraic formula for each of the machine's
3341 instructions. This is a very clean way to describe the target. But when
3342 the compiler needs information that is difficult to express in this
3343 fashion, I have not hesitated to define an ad-hoc parameter to the machine
3344 description. The purpose of portability is to reduce the total work needed
3345 on the compiler; it was not of interest for its own sake.
3348 @cindex autoincrement addressing, availability
3350 GCC does not contain machine dependent code, but it does contain code
3351 that depends on machine parameters such as endianness (whether the most
3352 significant byte has the highest or lowest address of the bytes in a word)
3353 and the availability of autoincrement addressing. In the RTL-generation
3354 pass, it is often necessary to have multiple strategies for generating code
3355 for a particular kind of syntax tree, strategies that are usable for different
3356 combinations of parameters. Often I have not tried to address all possible
3357 cases, but only the common ones or only the ones that I have encountered.
3358 As a result, a new target may require additional strategies. You will know
3359 if this happens because the compiler will call @code{abort}. Fortunately,
3360 the new strategies can be added in a machine-independent fashion, and will
3361 affect only the target machines that need them.
3366 @chapter Interfacing to GCC Output
3367 @cindex interfacing to GCC output
3368 @cindex run-time conventions
3369 @cindex function call conventions
3370 @cindex conventions, run-time
3372 GCC is normally configured to use the same function calling convention
3373 normally in use on the target system. This is done with the
3374 machine-description macros described (@pxref{Target Macros}).
3376 @cindex unions, returning
3377 @cindex structures, returning
3378 @cindex returning structures and unions
3379 However, returning of structure and union values is done differently on
3380 some target machines. As a result, functions compiled with PCC
3381 returning such types cannot be called from code compiled with GCC,
3382 and vice versa. This does not cause trouble often because few Unix
3383 library routines return structures or unions.
3385 GCC code returns structures and unions that are 1, 2, 4 or 8 bytes
3386 long in the same registers used for @code{int} or @code{double} return
3387 values. (GCC typically allocates variables of such types in
3388 registers also.) Structures and unions of other sizes are returned by
3389 storing them into an address passed by the caller (usually in a
3390 register). The machine-description macros @code{STRUCT_VALUE} and
3391 @code{STRUCT_INCOMING_VALUE} tell GCC where to pass this address.
3393 By contrast, PCC on most target machines returns structures and unions
3394 of any size by copying the data into an area of static storage, and then
3395 returning the address of that storage as if it were a pointer value.
3396 The caller must copy the data from that memory area to the place where
3397 the value is wanted. This is slower than the method used by GCC, and
3398 fails to be reentrant.
3400 On some target machines, such as RISC machines and the 80386, the
3401 standard system convention is to pass to the subroutine the address of
3402 where to return the value. On these machines, GCC has been
3403 configured to be compatible with the standard compiler, when this method
3404 is used. It may not be compatible for structures of 1, 2, 4 or 8 bytes.
3406 @cindex argument passing
3407 @cindex passing arguments
3408 GCC uses the system's standard convention for passing arguments. On
3409 some machines, the first few arguments are passed in registers; in
3410 others, all are passed on the stack. It would be possible to use
3411 registers for argument passing on any machine, and this would probably
3412 result in a significant speedup. But the result would be complete
3413 incompatibility with code that follows the standard convention. So this
3414 change is practical only if you are switching to GCC as the sole C
3415 compiler for the system. We may implement register argument passing on
3416 certain machines once we have a complete GNU system so that we can
3417 compile the libraries with GCC.
3419 On some machines (particularly the Sparc), certain types of arguments
3420 are passed ``by invisible reference''. This means that the value is
3421 stored in memory, and the address of the memory location is passed to
3424 @cindex @code{longjmp} and automatic variables
3425 If you use @code{longjmp}, beware of automatic variables. ISO C says that
3426 automatic variables that are not declared @code{volatile} have undefined
3427 values after a @code{longjmp}. And this is all GCC promises to do,
3428 because it is very difficult to restore register variables correctly, and
3429 one of GCC's features is that it can put variables in registers without
3432 If you want a variable to be unaltered by @code{longjmp}, and you don't
3433 want to write @code{volatile} because old C compilers don't accept it,
3434 just take the address of the variable. If a variable's address is ever
3435 taken, even if just to compute it and ignore it, then the variable cannot
3446 @cindex arithmetic libraries
3447 @cindex math libraries
3448 Code compiled with GCC may call certain library routines. Most of
3449 them handle arithmetic for which there are no instructions. This
3450 includes multiply and divide on some machines, and floating point
3451 operations on any machine for which floating point support is disabled
3452 with @samp{-msoft-float}. Some standard parts of the C library, such as
3453 @code{bcopy} or @code{memcpy}, are also called automatically. The usual
3454 function call interface is used for calling the library routines.
3456 These library routines should be defined in the library @file{libgcc.a},
3457 which GCC automatically searches whenever it links a program. On
3458 machines that have multiply and divide instructions, if hardware
3459 floating point is in use, normally @file{libgcc.a} is not needed, but it
3460 is searched just in case.
3462 Each arithmetic function is defined in @file{libgcc1.c} to use the
3463 corresponding C arithmetic operator. As long as the file is compiled
3464 with another C compiler, which supports all the C arithmetic operators,
3465 this file will work portably. However, @file{libgcc1.c} does not work if
3466 compiled with GCC, because each arithmetic function would compile
3467 into a call to itself!
3472 @chapter Passes and Files of the Compiler
3473 @cindex passes and files of the compiler
3474 @cindex files and passes of the compiler
3475 @cindex compiler passes and files
3477 @cindex top level of compiler
3478 The overall control structure of the compiler is in @file{toplev.c}. This
3479 file is responsible for initialization, decoding arguments, opening and
3480 closing files, and sequencing the passes.
3482 @cindex parsing pass
3483 The parsing pass is invoked only once, to parse the entire input. The RTL
3484 intermediate code for a function is generated as the function is parsed, a
3485 statement at a time. Each statement is read in as a syntax tree and then
3486 converted to RTL; then the storage for the tree for the statement is
3487 reclaimed. Storage for types (and the expressions for their sizes),
3488 declarations, and a representation of the binding contours and how they nest,
3489 remain until the function is finished being compiled; these are all needed
3490 to output the debugging information.
3492 @findex rest_of_compilation
3493 @findex rest_of_decl_compilation
3494 Each time the parsing pass reads a complete function definition or
3495 top-level declaration, it calls either the function
3496 @code{rest_of_compilation}, or the function
3497 @code{rest_of_decl_compilation} in @file{toplev.c}, which are
3498 responsible for all further processing necessary, ending with output of
3499 the assembler language. All other compiler passes run, in sequence,
3500 within @code{rest_of_compilation}. When that function returns from
3501 compiling a function definition, the storage used for that function
3502 definition's compilation is entirely freed, unless it is an inline
3505 (@pxref{Inline,,An Inline Function is As Fast As a Macro}).
3508 (@pxref{Inline,,An Inline Function is As Fast As a Macro,gcc.texi,Using GCC}).
3511 Here is a list of all the passes of the compiler and their source files.
3512 Also included is a description of where debugging dumps can be requested
3513 with @samp{-d} options.
3517 Parsing. This pass reads the entire text of a function definition,
3518 constructing partial syntax trees. This and RTL generation are no longer
3519 truly separate passes (formerly they were), but it is easier to think
3520 of them as separate.
3522 The tree representation does not entirely follow C syntax, because it is
3523 intended to support other languages as well.
3525 Language-specific data type analysis is also done in this pass, and every
3526 tree node that represents an expression has a data type attached.
3527 Variables are represented as declaration nodes.
3529 @cindex constant folding
3530 @cindex arithmetic simplifications
3531 @cindex simplifications, arithmetic
3532 Constant folding and some arithmetic simplifications are also done
3535 The language-independent source files for parsing are
3536 @file{stor-layout.c}, @file{fold-const.c}, and @file{tree.c}.
3537 There are also header files @file{tree.h} and @file{tree.def}
3538 which define the format of the tree representation.@refill
3540 @c Avoiding overfull is tricky here.
3541 The source files to parse C are
3545 @file{c-aux-info.c},
3548 along with header files
3552 The source files for parsing C++ are in @file{cp/}.
3553 They are @file{parse.y},
3555 @file{cvt.c}, @file{decl.c}, @file{decl2.c},
3557 @file{expr.c}, @file{init.c}, @file{lex.c},
3558 @file{method.c}, @file{ptree.c},@*
3559 @file{search.c}, @file{tree.c},
3560 @file{typeck2.c}, and
3561 @file{typeck.c}, along with header files @file{cp-tree.def},
3562 @file{cp-tree.h}, and @file{decl.h}.
3564 The special source files for parsing Objective C are in @file{objc/}.
3565 They are @file{objc-parse.y}, @file{objc-act.c}, @file{objc-tree.def}, and
3566 @file{objc-act.h}. Certain C-specific files are used for this as
3569 The file @file{c-common.c} is also used for all of the above languages.
3571 @cindex RTL generation
3573 RTL generation. This is the conversion of syntax tree into RTL code.
3574 It is actually done statement-by-statement during parsing, but for
3575 most purposes it can be thought of as a separate pass.
3577 @cindex target-parameter-dependent code
3578 This is where the bulk of target-parameter-dependent code is found,
3579 since often it is necessary for strategies to apply only when certain
3580 standard kinds of instructions are available. The purpose of named
3581 instruction patterns is to provide this information to the RTL
3584 @cindex tail recursion optimization
3585 Optimization is done in this pass for @code{if}-conditions that are
3586 comparisons, boolean operations or conditional expressions. Tail
3587 recursion is detected at this time also. Decisions are made about how
3588 best to arrange loops and how to output @code{switch} statements.
3590 @c Avoiding overfull is tricky here.
3591 The source files for RTL generation include
3599 and @file{emit-rtl.c}.
3601 @file{insn-emit.c}, generated from the machine description by the
3602 program @code{genemit}, is used in this pass. The header file
3603 @file{expr.h} is used for communication within this pass.@refill
3607 The header files @file{insn-flags.h} and @file{insn-codes.h},
3608 generated from the machine description by the programs @code{genflags}
3609 and @code{gencodes}, tell this pass which standard names are available
3610 for use and which patterns correspond to them.@refill
3612 Aside from debugging information output, none of the following passes
3613 refers to the tree structure representation of the function (only
3614 part of which is saved).
3616 @cindex inline, automatic
3617 The decision of whether the function can and should be expanded inline
3618 in its subsequent callers is made at the end of rtl generation. The
3619 function must meet certain criteria, currently related to the size of
3620 the function and the types and number of parameters it has. Note that
3621 this function may contain loops, recursive calls to itself
3622 (tail-recursive functions can be inlined!), gotos, in short, all
3623 constructs supported by GCC. The file @file{integrate.c} contains
3624 the code to save a function's rtl for later inlining and to inline that
3625 rtl when the function is called. The header file @file{integrate.h}
3626 is also used for this purpose.
3628 The option @samp{-dr} causes a debugging dump of the RTL code after
3629 this pass. This dump file's name is made by appending @samp{.rtl} to
3630 the input file name.
3632 @cindex jump optimization
3633 @cindex unreachable code
3636 Jump optimization. This pass simplifies jumps to the following
3637 instruction, jumps across jumps, and jumps to jumps. It deletes
3638 unreferenced labels and unreachable code, except that unreachable code
3639 that contains a loop is not recognized as unreachable in this pass.
3640 (Such loops are deleted later in the basic block analysis.) It also
3641 converts some code originally written with jumps into sequences of
3642 instructions that directly set values from the results of comparisons,
3643 if the machine has such instructions.
3645 Jump optimization is performed two or three times. The first time is
3646 immediately following RTL generation. The second time is after CSE,
3647 but only if CSE says repeated jump optimization is needed. The
3648 last time is right before the final pass. That time, cross-jumping
3649 and deletion of no-op move instructions are done together with the
3650 optimizations described above.
3652 The source file of this pass is @file{jump.c}.
3654 The option @samp{-dj} causes a debugging dump of the RTL code after
3655 this pass is run for the first time. This dump file's name is made by
3656 appending @samp{.jump} to the input file name.
3658 @cindex register use analysis
3660 Register scan. This pass finds the first and last use of each
3661 register, as a guide for common subexpression elimination. Its source
3662 is in @file{regclass.c}.
3664 @cindex jump threading
3666 Jump threading. This pass detects a condition jump that branches to an
3667 identical or inverse test. Such jumps can be @samp{threaded} through
3668 the second conditional test. The source code for this pass is in
3669 @file{jump.c}. This optimization is only performed if
3670 @samp{-fthread-jumps} is enabled.
3672 @cindex common subexpression elimination
3673 @cindex constant propagation
3675 Common subexpression elimination. This pass also does constant
3676 propagation. Its source file is @file{cse.c}. If constant
3677 propagation causes conditional jumps to become unconditional or to
3678 become no-ops, jump optimization is run again when CSE is finished.
3680 The option @samp{-ds} causes a debugging dump of the RTL code after
3681 this pass. This dump file's name is made by appending @samp{.cse} to
3682 the input file name.
3684 @cindex global common subexpression elimination
3685 @cindex constant propagation
3686 @cindex copy propagation
3688 Global common subexpression elimination. This pass performs GCSE
3689 using Morel-Renvoise Partial Redundancy Elimination, with the exception
3690 that it does not try to move invariants out of loops - that is left to
3691 the loop optimization pass. This pass also performs global constant
3692 and copy propagation.
3694 The source file for this pass is gcse.c.
3696 The option @samp{-dG} causes a debugging dump of the RTL code after
3697 this pass. This dump file's name is made by appending @samp{.gcse} to
3698 the input file name.
3700 @cindex loop optimization
3702 @cindex strength-reduction
3704 Loop optimization. This pass moves constant expressions out of loops,
3705 and optionally does strength-reduction and loop unrolling as well.
3706 Its source files are @file{loop.c} and @file{unroll.c}, plus the header
3707 @file{loop.h} used for communication between them. Loop unrolling uses
3708 some functions in @file{integrate.c} and the header @file{integrate.h}.
3710 The option @samp{-dL} causes a debugging dump of the RTL code after
3711 this pass. This dump file's name is made by appending @samp{.loop} to
3712 the input file name.
3715 If @samp{-frerun-cse-after-loop} was enabled, a second common
3716 subexpression elimination pass is performed after the loop optimization
3717 pass. Jump threading is also done again at this time if it was specified.
3719 The option @samp{-dt} causes a debugging dump of the RTL code after
3720 this pass. This dump file's name is made by appending @samp{.cse2} to
3721 the input file name.
3723 @cindex data flow analysis
3724 @cindex analysis, data flow
3725 @cindex basic blocks
3727 Data flow analysis (@file{flow.c}). This pass divides the program
3728 into basic blocks (and in the process deletes unreachable loops); then
3729 it computes which pseudo-registers are live at each point in the
3730 program, and makes the first instruction that uses a value point at
3731 the instruction that computed the value.
3733 @cindex autoincrement/decrement analysis
3734 This pass also deletes computations whose results are never used, and
3735 combines memory references with add or subtract instructions to make
3736 autoincrement or autodecrement addressing.
3738 The option @samp{-df} causes a debugging dump of the RTL code after
3739 this pass. This dump file's name is made by appending @samp{.flow} to
3740 the input file name. If stupid register allocation is in use, this
3741 dump file reflects the full results of such allocation.
3743 @cindex instruction combination
3745 Instruction combination (@file{combine.c}). This pass attempts to
3746 combine groups of two or three instructions that are related by data
3747 flow into single instructions. It combines the RTL expressions for
3748 the instructions by substitution, simplifies the result using algebra,
3749 and then attempts to match the result against the machine description.
3751 The option @samp{-dc} causes a debugging dump of the RTL code after
3752 this pass. This dump file's name is made by appending @samp{.combine}
3753 to the input file name.
3755 @cindex register movement
3757 Register movement (@file{regmove.c}). This pass looks for cases where
3758 matching constraints would force an instruction to need a reload, and
3759 this reload would be a register to register move. It then attempts
3760 to change the registers used by the instruction to avoid the move
3763 The option @samp{-dN} causes a debugging dump of the RTL code after
3764 this pass. This dump file's name is made by appending @samp{.regmove}
3765 to the input file name.
3767 @cindex instruction scheduling
3768 @cindex scheduling, instruction
3770 Instruction scheduling (@file{sched.c}). This pass looks for
3771 instructions whose output will not be available by the time that it is
3772 used in subsequent instructions. (Memory loads and floating point
3773 instructions often have this behavior on RISC machines). It re-orders
3774 instructions within a basic block to try to separate the definition and
3775 use of items that otherwise would cause pipeline stalls.
3777 Instruction scheduling is performed twice. The first time is immediately
3778 after instruction combination and the second is immediately after reload.
3780 The option @samp{-dS} causes a debugging dump of the RTL code after this
3781 pass is run for the first time. The dump file's name is made by
3782 appending @samp{.sched} to the input file name.
3784 @cindex register class preference pass
3786 Register class preferencing. The RTL code is scanned to find out
3787 which register class is best for each pseudo register. The source
3788 file is @file{regclass.c}.
3790 @cindex register allocation
3791 @cindex local register allocation
3793 Local register allocation (@file{local-alloc.c}). This pass allocates
3794 hard registers to pseudo registers that are used only within one basic
3795 block. Because the basic block is linear, it can use fast and
3796 powerful techniques to do a very good job.
3798 The option @samp{-dl} causes a debugging dump of the RTL code after
3799 this pass. This dump file's name is made by appending @samp{.lreg} to
3800 the input file name.
3802 @cindex global register allocation
3804 Global register allocation (@file{global.c}). This pass
3805 allocates hard registers for the remaining pseudo registers (those
3806 whose life spans are not contained in one basic block).
3810 Reloading. This pass renumbers pseudo registers with the hardware
3811 registers numbers they were allocated. Pseudo registers that did not
3812 get hard registers are replaced with stack slots. Then it finds
3813 instructions that are invalid because a value has failed to end up in
3814 a register, or has ended up in a register of the wrong kind. It fixes
3815 up these instructions by reloading the problematical values
3816 temporarily into registers. Additional instructions are generated to
3819 The reload pass also optionally eliminates the frame pointer and inserts
3820 instructions to save and restore call-clobbered registers around calls.
3822 Source files are @file{reload.c} and @file{reload1.c}, plus the header
3823 @file{reload.h} used for communication between them.
3825 The option @samp{-dg} causes a debugging dump of the RTL code after
3826 this pass. This dump file's name is made by appending @samp{.greg} to
3827 the input file name.
3829 @cindex instruction scheduling
3830 @cindex scheduling, instruction
3832 Instruction scheduling is repeated here to try to avoid pipeline stalls
3833 due to memory loads generated for spilled pseudo registers.
3835 The option @samp{-dR} causes a debugging dump of the RTL code after
3836 this pass. This dump file's name is made by appending @samp{.sched2}
3837 to the input file name.
3839 @cindex cross-jumping
3840 @cindex no-op move instructions
3842 Jump optimization is repeated, this time including cross-jumping
3843 and deletion of no-op move instructions.
3845 The option @samp{-dJ} causes a debugging dump of the RTL code after
3846 this pass. This dump file's name is made by appending @samp{.jump2}
3847 to the input file name.
3849 @cindex delayed branch scheduling
3850 @cindex scheduling, delayed branch
3852 Delayed branch scheduling. This optional pass attempts to find
3853 instructions that can go into the delay slots of other instructions,
3854 usually jumps and calls. The source file name is @file{reorg.c}.
3856 The option @samp{-dd} causes a debugging dump of the RTL code after
3857 this pass. This dump file's name is made by appending @samp{.dbr}
3858 to the input file name.
3860 @cindex branch shortening
3862 Branch shortening. On many RISC machines, branch instructions have a
3863 limited range. Thus, longer sequences of instructions must be used for
3864 long branches. In this pass, the compiler figures out what how far each
3865 instruction will be from each other instruction, and therefore whether
3866 the usual instructions, or the longer sequences, must be used for each
3869 @cindex register-to-stack conversion
3871 Conversion from usage of some hard registers to usage of a register
3872 stack may be done at this point. Currently, this is supported only
3873 for the floating-point registers of the Intel 80387 coprocessor. The
3874 source file name is @file{reg-stack.c}.
3876 The options @samp{-dk} causes a debugging dump of the RTL code after
3877 this pass. This dump file's name is made by appending @samp{.stack}
3878 to the input file name.
3881 @cindex peephole optimization
3883 Final. This pass outputs the assembler code for the function. It is
3884 also responsible for identifying spurious test and compare
3885 instructions. Machine-specific peephole optimizations are performed
3886 at the same time. The function entry and exit sequences are generated
3887 directly as assembler code in this pass; they never exist as RTL.
3889 The source files are @file{final.c} plus @file{insn-output.c}; the
3890 latter is generated automatically from the machine description by the
3891 tool @file{genoutput}. The header file @file{conditions.h} is used
3892 for communication between these files.
3894 @cindex debugging information generation
3896 Debugging information output. This is run after final because it must
3897 output the stack slot offsets for pseudo registers that did not get
3898 hard registers. Source files are @file{dbxout.c} for DBX symbol table
3899 format, @file{sdbout.c} for SDB symbol table format, and
3900 @file{dwarfout.c} for DWARF symbol table format.
3903 Some additional files are used by all or many passes:
3907 Every pass uses @file{machmode.def} and @file{machmode.h} which define
3911 Several passes use @file{real.h}, which defines the default
3912 representation of floating point constants and how to operate on them.
3915 All the passes that work with RTL use the header files @file{rtl.h}
3916 and @file{rtl.def}, and subroutines in file @file{rtl.c}. The tools
3917 @code{gen*} also use these files to read and work with the machine
3922 Several passes refer to the header file @file{insn-config.h} which
3923 contains a few parameters (C macro definitions) generated
3924 automatically from the machine description RTL by the tool
3927 @cindex instruction recognizer
3929 Several passes use the instruction recognizer, which consists of
3930 @file{recog.c} and @file{recog.h}, plus the files @file{insn-recog.c}
3931 and @file{insn-extract.c} that are generated automatically from the
3932 machine description by the tools @file{genrecog} and
3933 @file{genextract}.@refill
3936 Several passes use the header files @file{regs.h} which defines the
3937 information recorded about pseudo register usage, and @file{basic-block.h}
3938 which defines the information recorded about basic blocks.
3941 @file{hard-reg-set.h} defines the type @code{HARD_REG_SET}, a bit-vector
3942 with a bit for each hard register, and some macros to manipulate it.
3943 This type is just @code{int} if the machine has few enough hard registers;
3944 otherwise it is an array of @code{int} and some of the macros expand
3948 Several passes use instruction attributes. A definition of the
3949 attributes defined for a particular machine is in file
3950 @file{insn-attr.h}, which is generated from the machine description by
3951 the program @file{genattr}. The file @file{insn-attrtab.c} contains
3952 subroutines to obtain the attribute values for insns. It is generated
3953 from the machine description by the program @file{genattrtab}.@refill
3965 @chapter The Configuration File
3966 @cindex configuration file
3967 @cindex @file{xm-@var{machine}.h}
3969 The configuration file @file{xm-@var{machine}.h} contains macro
3970 definitions that describe the machine and system on which the compiler
3971 is running, unlike the definitions in @file{@var{machine}.h}, which
3972 describe the machine for which the compiler is producing output. Most
3973 of the values in @file{xm-@var{machine}.h} are actually the same on all
3974 machines that GCC runs on, so large parts of all configuration files
3975 are identical. But there are some macros that vary:
3980 Define this macro if the host system is System V.
3984 Define this macro if the host system is VMS.
3986 @findex FATAL_EXIT_CODE
3987 @item FATAL_EXIT_CODE
3988 A C expression for the status code to be returned when the compiler
3989 exits after serious errors.
3991 @findex SUCCESS_EXIT_CODE
3992 @item SUCCESS_EXIT_CODE
3993 A C expression for the status code to be returned when the compiler
3994 exits without serious errors.
3996 @findex HOST_WORDS_BIG_ENDIAN
3997 @item HOST_WORDS_BIG_ENDIAN
3998 Defined if the host machine stores words of multi-word values in
3999 big-endian order. (GCC does not depend on the host byte ordering
4002 @findex HOST_FLOAT_WORDS_BIG_ENDIAN
4003 @item HOST_FLOAT_WORDS_BIG_ENDIAN
4004 Define this macro to be 1 if the host machine stores @code{DFmode},
4005 @code{XFmode} or @code{TFmode} floating point numbers in memory with the
4006 word containing the sign bit at the lowest address; otherwise, define it
4009 This macro need not be defined if the ordering is the same as for
4010 multi-word integers.
4012 @findex HOST_FLOAT_FORMAT
4013 @item HOST_FLOAT_FORMAT
4014 A numeric code distinguishing the floating point format for the host
4015 machine. See @code{TARGET_FLOAT_FORMAT} in @ref{Storage Layout} for the
4016 alternatives and default.
4018 @findex HOST_BITS_PER_CHAR
4019 @item HOST_BITS_PER_CHAR
4020 A C expression for the number of bits in @code{char} on the host
4023 @findex HOST_BITS_PER_SHORT
4024 @item HOST_BITS_PER_SHORT
4025 A C expression for the number of bits in @code{short} on the host
4028 @findex HOST_BITS_PER_INT
4029 @item HOST_BITS_PER_INT
4030 A C expression for the number of bits in @code{int} on the host
4033 @findex HOST_BITS_PER_LONG
4034 @item HOST_BITS_PER_LONG
4035 A C expression for the number of bits in @code{long} on the host
4038 @findex ONLY_INT_FIELDS
4039 @item ONLY_INT_FIELDS
4040 Define this macro to indicate that the host compiler only supports
4041 @code{int} bit fields, rather than other integral types, including
4042 @code{enum}, as do most C compilers.
4044 @findex OBSTACK_CHUNK_SIZE
4045 @item OBSTACK_CHUNK_SIZE
4046 A C expression for the size of ordinary obstack chunks.
4047 If you don't define this, a usually-reasonable default is used.
4049 @findex OBSTACK_CHUNK_ALLOC
4050 @item OBSTACK_CHUNK_ALLOC
4051 The function used to allocate obstack chunks.
4052 If you don't define this, @code{xmalloc} is used.
4054 @findex OBSTACK_CHUNK_FREE
4055 @item OBSTACK_CHUNK_FREE
4056 The function used to free obstack chunks.
4057 If you don't define this, @code{free} is used.
4059 @findex USE_C_ALLOCA
4061 Define this macro to indicate that the compiler is running with the
4062 @code{alloca} implemented in C. This version of @code{alloca} can be
4063 found in the file @file{alloca.c}; to use it, you must also alter the
4064 @file{Makefile} variable @code{ALLOCA}. (This is done automatically
4065 for the systems on which we know it is needed.)
4067 If you do define this macro, you should probably do it as follows:
4071 #define USE_C_ALLOCA
4073 #define alloca __builtin_alloca
4078 so that when the compiler is compiled with GCC it uses the more
4079 efficient built-in @code{alloca} function.
4081 @item FUNCTION_CONVERSION_BUG
4082 @findex FUNCTION_CONVERSION_BUG
4083 Define this macro to indicate that the host compiler does not properly
4084 handle converting a function value to a pointer-to-function when it is
4085 used in an expression.
4087 @findex MULTIBYTE_CHARS
4088 @item MULTIBYTE_CHARS
4089 Define this macro to enable support for multibyte characters in the
4090 input to GCC. This requires that the host system support the ISO C
4091 library functions for converting multibyte characters to wide
4096 Define this if your system is POSIX.1 compliant.
4098 @findex USE_PROTOTYPES
4099 @item USE_PROTOTYPES
4100 Define this to be 1 if you know that the host compiler supports
4101 prototypes, even if it doesn't define __STDC__, or define
4102 it to be 0 if you do not want any prototypes used in compiling
4103 GCC. If @samp{USE_PROTOTYPES} is not defined, it will be
4104 determined automatically whether your compiler supports
4105 prototypes by checking if @samp{__STDC__} is defined.
4107 @findex PATH_SEPARATOR
4108 @item PATH_SEPARATOR
4109 Define this macro to be a C character constant representing the
4110 character used to separate components in paths. The default value is
4113 @findex DIR_SEPARATOR
4115 If your system uses some character other than slash to separate
4116 directory names within a file specification, define this macro to be a C
4117 character constant specifying that character. When GCC displays file
4118 names, the character you specify will be used. GCC will test for
4119 both slash and the character you specify when parsing filenames.
4121 @findex OBJECT_SUFFIX
4123 Define this macro to be a C string representing the suffix for object
4124 files on your machine. If you do not define this macro, GCC will use
4125 @samp{.o} as the suffix for object files.
4127 @findex EXECUTABLE_SUFFIX
4128 @item EXECUTABLE_SUFFIX
4129 Define this macro to be a C string representing the suffix for executable
4130 files on your machine. If you do not define this macro, GCC will use
4131 the null string as the suffix for object files.
4133 @findex HOST_BIT_BUCKET
4134 @item HOST_BIT_BUCKET
4135 The name of a file or file-like object on the host system which acts as
4136 a ``bit bucket''. If you do not define this macro, GCC will use
4137 @samp{/dev/null} as the bit bucket. If the target does not support a
4138 bit bucket, this should be defined to the null string, or some other
4139 illegal filename. If the bit bucket is not writable, GCC will use a
4140 temporary file instead.
4142 @findex COLLECT_EXPORT_LIST
4143 @item COLLECT_EXPORT_LIST
4144 If defined, @code{collect2} will scan the individual object files
4145 specified on its command line and create an export list for the linker.
4146 Define this macro for systems like AIX, where the linker discards
4147 object files that are not referenced from @code{main} and uses export
4150 @findex COLLECT2_HOST_INITIALIZATION
4151 @item COLLECT2_HOST_INITIALIZATION
4152 If defined, a C statement (sans semicolon) that performs host-dependent
4153 initialization when @code{collect2} is being initialized.
4155 @findex GCC_DRIVER_HOST_INITIALIZATION
4156 @item GCC_DRIVER_HOST_INITIALIZATION
4157 If defined, a C statement (sans semicolon) that performs host-dependent
4158 initialization when a compilation driver is being initialized.
4160 @findex UPDATE_PATH_HOST_CANONICALIZE
4161 @item UPDATE_PATH_HOST_CANONICALIZE (@var{path}, @var{key})
4162 If defined, a C statement (sans semicolon) that performs host-dependent
4163 canonicalization when a path used in a compilation driver or preprocessor is
4164 canonicalized. @var{path} is the path to be canonicalized, and @var{key} is
4165 a translation prefix when its value isn't @code{NULL}. If the C statement
4166 does canonicalize @var{path}, the new path should be returned.
4171 In addition, configuration files for system V define @code{bcopy},
4172 @code{bzero} and @code{bcmp} as aliases. Some files define @code{alloca}
4173 as a macro when compiled with GCC, in order to take advantage of the
4174 benefit of GCC's built-in @code{alloca}.
4177 @chapter Makefile Fragments
4178 @cindex makefile fragment
4180 When you configure GCC using the @file{configure} script
4181 (@pxref{Installation}), it will construct the file @file{Makefile} from
4182 the template file @file{Makefile.in}. When it does this, it will
4183 incorporate makefile fragment files from the @file{config} directory,
4184 named @file{t-@var{target}} and @file{x-@var{host}}. If these files do
4185 not exist, it means nothing needs to be added for a given target or
4189 * Target Fragment:: Writing the @file{t-@var{target}} file.
4190 * Host Fragment:: Writing the @file{x-@var{host}} file.
4193 @node Target Fragment
4194 @section The Target Makefile Fragment
4195 @cindex target makefile fragment
4196 @cindex @file{t-@var{target}}
4198 The target makefile fragment, @file{t-@var{target}}, defines special
4199 target dependent variables and targets used in the @file{Makefile}:
4204 The rule to use to build @file{libgcc1.a}.
4205 If your target does not need to use the functions in @file{libgcc1.a},
4209 @findex CROSS_LIBGCC1
4211 The rule to use to build @file{libgcc1.a} when building a cross
4212 compiler. If your target does not need to use the functions in
4213 @file{libgcc1.a}, set this to empty. @xref{Cross Runtime}.
4215 @findex LIBGCC2_CFLAGS
4216 @item LIBGCC2_CFLAGS
4217 Compiler flags to use when compiling @file{libgcc2.c}.
4219 @findex LIB2FUNCS_EXTRA
4220 @item LIB2FUNCS_EXTRA
4221 A list of source file names to be compiled or assembled and inserted
4222 into @file{libgcc.a}.
4224 @findex Floating Point Emulation
4225 @item Floating Point Emulation
4226 To have GCC include software floating point libraries in @file{libgcc.a}
4227 define @code{FPBIT} and @code{DPBIT} along with a few rules as follows:
4229 # We want fine grained libraries, so use the new code to build the
4230 # floating point emulation libraries.
4235 fp-bit.c: $(srcdir)/config/fp-bit.c
4236 echo '#define FLOAT' > fp-bit.c
4237 cat $(srcdir)/config/fp-bit.c >> fp-bit.c
4239 dp-bit.c: $(srcdir)/config/fp-bit.c
4240 cat $(srcdir)/config/fp-bit.c > dp-bit.c
4243 You may need to provide additional #defines at the beginning of @file{fp-bit.c}
4244 and @file{dp-bit.c} to control target endianness and other options.
4247 @findex CRTSTUFF_T_CFLAGS
4248 @item CRTSTUFF_T_CFLAGS
4249 Special flags used when compiling @file{crtstuff.c}.
4250 @xref{Initialization}.
4252 @findex CRTSTUFF_T_CFLAGS_S
4253 @item CRTSTUFF_T_CFLAGS_S
4254 Special flags used when compiling @file{crtstuff.c} for shared
4255 linking. Used if you use @file{crtbeginS.o} and @file{crtendS.o}
4256 in @code{EXTRA-PARTS}.
4257 @xref{Initialization}.
4259 @findex MULTILIB_OPTIONS
4260 @item MULTILIB_OPTIONS
4261 For some targets, invoking GCC in different ways produces objects
4262 that can not be linked together. For example, for some targets GCC
4263 produces both big and little endian code. For these targets, you must
4264 arrange for multiple versions of @file{libgcc.a} to be compiled, one for
4265 each set of incompatible options. When GCC invokes the linker, it
4266 arranges to link in the right version of @file{libgcc.a}, based on
4267 the command line options used.
4269 The @code{MULTILIB_OPTIONS} macro lists the set of options for which
4270 special versions of @file{libgcc.a} must be built. Write options that
4271 are mutually incompatible side by side, separated by a slash. Write
4272 options that may be used together separated by a space. The build
4273 procedure will build all combinations of compatible options.
4275 For example, if you set @code{MULTILIB_OPTIONS} to @samp{m68000/m68020
4276 msoft-float}, @file{Makefile} will build special versions of
4277 @file{libgcc.a} using the following sets of options: @samp{-m68000},
4278 @samp{-m68020}, @samp{-msoft-float}, @samp{-m68000 -msoft-float}, and
4279 @samp{-m68020 -msoft-float}.
4281 @findex MULTILIB_DIRNAMES
4282 @item MULTILIB_DIRNAMES
4283 If @code{MULTILIB_OPTIONS} is used, this variable specifies the
4284 directory names that should be used to hold the various libraries.
4285 Write one element in @code{MULTILIB_DIRNAMES} for each element in
4286 @code{MULTILIB_OPTIONS}. If @code{MULTILIB_DIRNAMES} is not used, the
4287 default value will be @code{MULTILIB_OPTIONS}, with all slashes treated
4290 For example, if @code{MULTILIB_OPTIONS} is set to @samp{m68000/m68020
4291 msoft-float}, then the default value of @code{MULTILIB_DIRNAMES} is
4292 @samp{m68000 m68020 msoft-float}. You may specify a different value if
4293 you desire a different set of directory names.
4295 @findex MULTILIB_MATCHES
4296 @item MULTILIB_MATCHES
4297 Sometimes the same option may be written in two different ways. If an
4298 option is listed in @code{MULTILIB_OPTIONS}, GCC needs to know about
4299 any synonyms. In that case, set @code{MULTILIB_MATCHES} to a list of
4300 items of the form @samp{option=option} to describe all relevant
4301 synonyms. For example, @samp{m68000=mc68000 m68020=mc68020}.
4303 @findex MULTILIB_EXCEPTIONS
4304 @item MULTILIB_EXCEPTIONS
4305 Sometimes when there are multiple sets of @code{MULTILIB_OPTIONS} being
4306 specified, there are combinations that should not be built. In that
4307 case, set @code{MULTILIB_EXCEPTIONS} to be all of the switch exceptions
4308 in shell case syntax that should not be built.
4310 For example, in the PowerPC embedded ABI support, it is not desirable
4311 to build libraries compiled with the @samp{-mcall-aix} option
4312 and either of the @samp{-fleading-underscore} or @samp{-mlittle} options
4313 at the same time. Therefore @code{MULTILIB_EXCEPTIONS} is set to
4314 @code{*mcall-aix/*fleading-underscore* *mlittle/*mcall-aix*}.
4316 @findex MULTILIB_EXTRA_OPTS
4317 @item MULTILIB_EXTRA_OPTS
4318 Sometimes it is desirable that when building multiple versions of
4319 @file{libgcc.a} certain options should always be passed on to the
4320 compiler. In that case, set @code{MULTILIB_EXTRA_OPTS} to be the list
4321 of options to be used for all builds.
4325 @section The Host Makefile Fragment
4326 @cindex host makefile fragment
4327 @cindex @file{x-@var{host}}
4329 The host makefile fragment, @file{x-@var{host}}, defines special host
4330 dependent variables and targets used in the @file{Makefile}:
4335 The compiler to use when building the first stage.
4339 Additional host libraries to link with.
4343 The compiler to use when building @file{libgcc1.a} for a native
4348 The version of @code{ar} to use when building @file{libgcc1.a} for a native
4353 The install program to use.
4358 @unnumbered Funding Free Software
4360 If you want to have more free software a few years from now, it makes
4361 sense for you to help encourage people to contribute funds for its
4362 development. The most effective approach known is to encourage
4363 commercial redistributors to donate.
4365 Users of free software systems can boost the pace of development by
4366 encouraging for-a-fee distributors to donate part of their selling price
4367 to free software developers---the Free Software Foundation, and others.
4369 The way to convince distributors to do this is to demand it and expect
4370 it from them. So when you compare distributors, judge them partly by
4371 how much they give to free software development. Show distributors
4372 they must compete to be the one who gives the most.
4374 To make this approach work, you must insist on numbers that you can
4375 compare, such as, ``We will donate ten dollars to the Frobnitz project
4376 for each disk sold.'' Don't be satisfied with a vague promise, such as
4377 ``A portion of the profits are donated,'' since it doesn't give a basis
4380 Even a precise fraction ``of the profits from this disk'' is not very
4381 meaningful, since creative accounting and unrelated business decisions
4382 can greatly alter what fraction of the sales price counts as profit.
4383 If the price you pay is $50, ten percent of the profit is probably
4384 less than a dollar; it might be a few cents, or nothing at all.
4386 Some redistributors do development work themselves. This is useful too;
4387 but to keep everyone honest, you need to inquire how much they do, and
4388 what kind. Some kinds of development make much more long-term
4389 difference than others. For example, maintaining a separate version of
4390 a program contributes very little; maintaining the standard version of a
4391 program for the whole community contributes much. Easy new ports
4392 contribute little, since someone else would surely do them; difficult
4393 ports such as adding a new CPU to the GNU Compiler Collection contribute more;
4394 major new features or packages contribute the most.
4396 By establishing the idea that supporting further development is ``the
4397 proper thing to do'' when distributing free software for a fee, we can
4398 assure a steady flow of resources into making more free software.
4401 Copyright (C) 1994 Free Software Foundation, Inc.
4402 Verbatim copying and redistribution of this section is permitted
4403 without royalty; alteration is not permitted.
4407 @unnumbered Linux and the GNU Project
4409 Many computer users run a modified version of the GNU system every
4410 day, without realizing it. Through a peculiar turn of events, the
4411 version of GNU which is widely used today is more often known as
4412 ``Linux'', and many users are not aware of the extent of its
4413 connection with the GNU Project.
4415 There really is a Linux; it is a kernel, and these people are using
4416 it. But you can't use a kernel by itself; a kernel is useful only as
4417 part of a whole system. The system in which Linux is typically used
4418 is a modified variant of the GNU system---in other words, a Linux-based
4421 Many users are not fully aware of the distinction between the kernel,
4422 which is Linux, and the whole system, which they also call ``Linux''.
4423 The ambiguous use of the name doesn't promote understanding.
4425 Programmers generally know that Linux is a kernel. But since they
4426 have generally heard the whole system called ``Linux'' as well, they
4427 often envisage a history which fits that name. For example, many
4428 believe that once Linus Torvalds finished writing the kernel, his
4429 friends looked around for other free software, and for no particular
4430 reason most everything necessary to make a Unix-like system was
4433 What they found was no accident---it was the GNU system. The available
4434 free software added up to a complete system because the GNU Project
4435 had been working since 1984 to make one. The GNU Manifesto
4436 had set forth the goal of developing a free Unix-like system, called
4437 GNU. By the time Linux was written, the system was almost finished.
4439 Most free software projects have the goal of developing a particular
4440 program for a particular job. For example, Linus Torvalds set out to
4441 write a Unix-like kernel (Linux); Donald Knuth set out to write a text
4442 formatter (TeX); Bob Scheifler set out to develop a window system (X
4443 Windows). It's natural to measure the contribution of this kind of
4444 project by specific programs that came from the project.
4446 If we tried to measure the GNU Project's contribution in this way,
4447 what would we conclude? One CD-ROM vendor found that in their ``Linux
4448 distribution'', GNU software was the largest single contingent, around
4449 28% of the total source code, and this included some of the essential
4450 major components without which there could be no system. Linux itself
4451 was about 3%. So if you were going to pick a name for the system
4452 based on who wrote the programs in the system, the most appropriate
4453 single choice would be ``GNU''.
4455 But we don't think that is the right way to consider the question.
4456 The GNU Project was not, is not, a project to develop specific
4457 software packages. It was not a project to develop a C compiler,
4458 although we did. It was not a project to develop a text editor,
4459 although we developed one. The GNU Project's aim was to develop
4460 @emph{a complete free Unix-like system}.
4462 Many people have made major contributions to the free software in the
4463 system, and they all deserve credit. But the reason it is @emph{a
4464 system}---and not just a collection of useful programs---is because the
4465 GNU Project set out to make it one. We wrote the programs that were
4466 needed to make a @emph{complete} free system. We wrote essential but
4467 unexciting major components, such as the assembler and linker, because
4468 you can't have a system without them. A complete system needs more
4469 than just programming tools, so we wrote other components as well,
4470 such as the Bourne Again SHell, the PostScript interpreter
4471 Ghostscript, and the GNU C library.
4473 By the early 90s we had put together the whole system aside from the
4474 kernel (and we were also working on a kernel, the GNU Hurd, which runs
4475 on top of Mach). Developing this kernel has been a lot harder than we
4476 expected, and we are still working on finishing it.
4478 Fortunately, you don't have to wait for it, because Linux is working
4479 now. When Linus Torvalds wrote Linux, he filled the last major gap.
4480 People could then put Linux together with the GNU system to make a
4481 complete free system: a Linux-based GNU system (or GNU/Linux system,
4484 Putting them together sounds simple, but it was not a trivial job.
4485 The GNU C library (called glibc for short) needed substantial changes.
4486 Integrating a complete system as a distribution that would work ``out
4487 of the box'' was a big job, too. It required addressing the issue of
4488 how to install and boot the system---a problem we had not tackled,
4489 because we hadn't yet reached that point. The people who developed
4490 the various system distributions made a substantial contribution.
4492 The GNU Project supports GNU/Linux systems as well as @emph{the}
4493 GNU system---even with funds. We funded the rewriting of the
4494 Linux-related extensions to the GNU C library, so that now they are
4495 well integrated, and the newest GNU/Linux systems use the current
4496 library release with no changes. We also funded an early stage of the
4497 development of Debian GNU/Linux.
4499 We use Linux-based GNU systems today for most of our work, and we hope
4500 you use them too. But please don't confuse the public by using the
4501 name ``Linux'' ambiguously. Linux is the kernel, one of the essential
4502 major components of the system. The system as a whole is more or less
4506 @unnumbered GNU GENERAL PUBLIC LICENSE
4507 @center Version 2, June 1991
4510 Copyright @copyright{} 1989, 1991 Free Software Foundation, Inc.
4511 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
4513 Everyone is permitted to copy and distribute verbatim copies
4514 of this license document, but changing it is not allowed.
4517 @unnumberedsec Preamble
4519 The licenses for most software are designed to take away your
4520 freedom to share and change it. By contrast, the GNU General Public
4521 License is intended to guarantee your freedom to share and change free
4522 software---to make sure the software is free for all its users. This
4523 General Public License applies to most of the Free Software
4524 Foundation's software and to any other program whose authors commit to
4525 using it. (Some other Free Software Foundation software is covered by
4526 the GNU Library General Public License instead.) You can apply it to
4529 When we speak of free software, we are referring to freedom, not
4530 price. Our General Public Licenses are designed to make sure that you
4531 have the freedom to distribute copies of free software (and charge for
4532 this service if you wish), that you receive source code or can get it
4533 if you want it, that you can change the software or use pieces of it
4534 in new free programs; and that you know you can do these things.
4536 To protect your rights, we need to make restrictions that forbid
4537 anyone to deny you these rights or to ask you to surrender the rights.
4538 These restrictions translate to certain responsibilities for you if you
4539 distribute copies of the software, or if you modify it.
4541 For example, if you distribute copies of such a program, whether
4542 gratis or for a fee, you must give the recipients all the rights that
4543 you have. You must make sure that they, too, receive or can get the
4544 source code. And you must show them these terms so they know their
4547 We protect your rights with two steps: (1) copyright the software, and
4548 (2) offer you this license which gives you legal permission to copy,
4549 distribute and/or modify the software.
4551 Also, for each author's protection and ours, we want to make certain
4552 that everyone understands that there is no warranty for this free
4553 software. If the software is modified by someone else and passed on, we
4554 want its recipients to know that what they have is not the original, so
4555 that any problems introduced by others will not reflect on the original
4556 authors' reputations.
4558 Finally, any free program is threatened constantly by software
4559 patents. We wish to avoid the danger that redistributors of a free
4560 program will individually obtain patent licenses, in effect making the
4561 program proprietary. To prevent this, we have made it clear that any
4562 patent must be licensed for everyone's free use or not licensed at all.
4564 The precise terms and conditions for copying, distribution and
4565 modification follow.
4568 @unnumberedsec TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
4571 @center TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
4576 This License applies to any program or other work which contains
4577 a notice placed by the copyright holder saying it may be distributed
4578 under the terms of this General Public License. The ``Program'', below,
4579 refers to any such program or work, and a ``work based on the Program''
4580 means either the Program or any derivative work under copyright law:
4581 that is to say, a work containing the Program or a portion of it,
4582 either verbatim or with modifications and/or translated into another
4583 language. (Hereinafter, translation is included without limitation in
4584 the term ``modification''.) Each licensee is addressed as ``you''.
4586 Activities other than copying, distribution and modification are not
4587 covered by this License; they are outside its scope. The act of
4588 running the Program is not restricted, and the output from the Program
4589 is covered only if its contents constitute a work based on the
4590 Program (independent of having been made by running the Program).
4591 Whether that is true depends on what the Program does.
4594 You may copy and distribute verbatim copies of the Program's
4595 source code as you receive it, in any medium, provided that you
4596 conspicuously and appropriately publish on each copy an appropriate
4597 copyright notice and disclaimer of warranty; keep intact all the
4598 notices that refer to this License and to the absence of any warranty;
4599 and give any other recipients of the Program a copy of this License
4600 along with the Program.
4602 You may charge a fee for the physical act of transferring a copy, and
4603 you may at your option offer warranty protection in exchange for a fee.
4606 You may modify your copy or copies of the Program or any portion
4607 of it, thus forming a work based on the Program, and copy and
4608 distribute such modifications or work under the terms of Section 1
4609 above, provided that you also meet all of these conditions:
4613 You must cause the modified files to carry prominent notices
4614 stating that you changed the files and the date of any change.
4617 You must cause any work that you distribute or publish, that in
4618 whole or in part contains or is derived from the Program or any
4619 part thereof, to be licensed as a whole at no charge to all third
4620 parties under the terms of this License.
4623 If the modified program normally reads commands interactively
4624 when run, you must cause it, when started running for such
4625 interactive use in the most ordinary way, to print or display an
4626 announcement including an appropriate copyright notice and a
4627 notice that there is no warranty (or else, saying that you provide
4628 a warranty) and that users may redistribute the program under
4629 these conditions, and telling the user how to view a copy of this
4630 License. (Exception: if the Program itself is interactive but
4631 does not normally print such an announcement, your work based on
4632 the Program is not required to print an announcement.)
4635 These requirements apply to the modified work as a whole. If
4636 identifiable sections of that work are not derived from the Program,
4637 and can be reasonably considered independent and separate works in
4638 themselves, then this License, and its terms, do not apply to those
4639 sections when you distribute them as separate works. But when you
4640 distribute the same sections as part of a whole which is a work based
4641 on the Program, the distribution of the whole must be on the terms of
4642 this License, whose permissions for other licensees extend to the
4643 entire whole, and thus to each and every part regardless of who wrote it.
4645 Thus, it is not the intent of this section to claim rights or contest
4646 your rights to work written entirely by you; rather, the intent is to
4647 exercise the right to control the distribution of derivative or
4648 collective works based on the Program.
4650 In addition, mere aggregation of another work not based on the Program
4651 with the Program (or with a work based on the Program) on a volume of
4652 a storage or distribution medium does not bring the other work under
4653 the scope of this License.
4656 You may copy and distribute the Program (or a work based on it,
4657 under Section 2) in object code or executable form under the terms of
4658 Sections 1 and 2 above provided that you also do one of the following:
4662 Accompany it with the complete corresponding machine-readable
4663 source code, which must be distributed under the terms of Sections
4664 1 and 2 above on a medium customarily used for software interchange; or,
4667 Accompany it with a written offer, valid for at least three
4668 years, to give any third party, for a charge no more than your
4669 cost of physically performing source distribution, a complete
4670 machine-readable copy of the corresponding source code, to be
4671 distributed under the terms of Sections 1 and 2 above on a medium
4672 customarily used for software interchange; or,
4675 Accompany it with the information you received as to the offer
4676 to distribute corresponding source code. (This alternative is
4677 allowed only for noncommercial distribution and only if you
4678 received the program in object code or executable form with such
4679 an offer, in accord with Subsection b above.)
4682 The source code for a work means the preferred form of the work for
4683 making modifications to it. For an executable work, complete source
4684 code means all the source code for all modules it contains, plus any
4685 associated interface definition files, plus the scripts used to
4686 control compilation and installation of the executable. However, as a
4687 special exception, the source code distributed need not include
4688 anything that is normally distributed (in either source or binary
4689 form) with the major components (compiler, kernel, and so on) of the
4690 operating system on which the executable runs, unless that component
4691 itself accompanies the executable.
4693 If distribution of executable or object code is made by offering
4694 access to copy from a designated place, then offering equivalent
4695 access to copy the source code from the same place counts as
4696 distribution of the source code, even though third parties are not
4697 compelled to copy the source along with the object code.
4700 You may not copy, modify, sublicense, or distribute the Program
4701 except as expressly provided under this License. Any attempt
4702 otherwise to copy, modify, sublicense or distribute the Program is
4703 void, and will automatically terminate your rights under this License.
4704 However, parties who have received copies, or rights, from you under
4705 this License will not have their licenses terminated so long as such
4706 parties remain in full compliance.
4709 You are not required to accept this License, since you have not
4710 signed it. However, nothing else grants you permission to modify or
4711 distribute the Program or its derivative works. These actions are
4712 prohibited by law if you do not accept this License. Therefore, by
4713 modifying or distributing the Program (or any work based on the
4714 Program), you indicate your acceptance of this License to do so, and
4715 all its terms and conditions for copying, distributing or modifying
4716 the Program or works based on it.
4719 Each time you redistribute the Program (or any work based on the
4720 Program), the recipient automatically receives a license from the
4721 original licensor to copy, distribute or modify the Program subject to
4722 these terms and conditions. You may not impose any further
4723 restrictions on the recipients' exercise of the rights granted herein.
4724 You are not responsible for enforcing compliance by third parties to
4728 If, as a consequence of a court judgment or allegation of patent
4729 infringement or for any other reason (not limited to patent issues),
4730 conditions are imposed on you (whether by court order, agreement or
4731 otherwise) that contradict the conditions of this License, they do not
4732 excuse you from the conditions of this License. If you cannot
4733 distribute so as to satisfy simultaneously your obligations under this
4734 License and any other pertinent obligations, then as a consequence you
4735 may not distribute the Program at all. For example, if a patent
4736 license would not permit royalty-free redistribution of the Program by
4737 all those who receive copies directly or indirectly through you, then
4738 the only way you could satisfy both it and this License would be to
4739 refrain entirely from distribution of the Program.
4741 If any portion of this section is held invalid or unenforceable under
4742 any particular circumstance, the balance of the section is intended to
4743 apply and the section as a whole is intended to apply in other
4746 It is not the purpose of this section to induce you to infringe any
4747 patents or other property right claims or to contest validity of any
4748 such claims; this section has the sole purpose of protecting the
4749 integrity of the free software distribution system, which is
4750 implemented by public license practices. Many people have made
4751 generous contributions to the wide range of software distributed
4752 through that system in reliance on consistent application of that
4753 system; it is up to the author/donor to decide if he or she is willing
4754 to distribute software through any other system and a licensee cannot
4757 This section is intended to make thoroughly clear what is believed to
4758 be a consequence of the rest of this License.
4761 If the distribution and/or use of the Program is restricted in
4762 certain countries either by patents or by copyrighted interfaces, the
4763 original copyright holder who places the Program under this License
4764 may add an explicit geographical distribution limitation excluding
4765 those countries, so that distribution is permitted only in or among
4766 countries not thus excluded. In such case, this License incorporates
4767 the limitation as if written in the body of this License.
4770 The Free Software Foundation may publish revised and/or new versions
4771 of the General Public License from time to time. Such new versions will
4772 be similar in spirit to the present version, but may differ in detail to
4773 address new problems or concerns.
4775 Each version is given a distinguishing version number. If the Program
4776 specifies a version number of this License which applies to it and ``any
4777 later version'', you have the option of following the terms and conditions
4778 either of that version or of any later version published by the Free
4779 Software Foundation. If the Program does not specify a version number of
4780 this License, you may choose any version ever published by the Free Software
4784 If you wish to incorporate parts of the Program into other free
4785 programs whose distribution conditions are different, write to the author
4786 to ask for permission. For software which is copyrighted by the Free
4787 Software Foundation, write to the Free Software Foundation; we sometimes
4788 make exceptions for this. Our decision will be guided by the two goals
4789 of preserving the free status of all derivatives of our free software and
4790 of promoting the sharing and reuse of software generally.
4793 @heading NO WARRANTY
4800 BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
4801 FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
4802 OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
4803 PROVIDE THE PROGRAM ``AS IS'' WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
4804 OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
4805 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
4806 TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
4807 PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
4808 REPAIR OR CORRECTION.
4811 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
4812 WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
4813 REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
4814 INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
4815 OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
4816 TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
4817 YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
4818 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
4819 POSSIBILITY OF SUCH DAMAGES.
4823 @heading END OF TERMS AND CONDITIONS
4826 @center END OF TERMS AND CONDITIONS
4830 @unnumberedsec How to Apply These Terms to Your New Programs
4832 If you develop a new program, and you want it to be of the greatest
4833 possible use to the public, the best way to achieve this is to make it
4834 free software which everyone can redistribute and change under these terms.
4836 To do so, attach the following notices to the program. It is safest
4837 to attach them to the start of each source file to most effectively
4838 convey the exclusion of warranty; and each file should have at least
4839 the ``copyright'' line and a pointer to where the full notice is found.
4842 @var{one line to give the program's name and a brief idea of what it does.}
4843 Copyright (C) @var{yyyy} @var{name of author}
4845 This program is free software; you can redistribute it and/or modify
4846 it under the terms of the GNU General Public License as published by
4847 the Free Software Foundation; either version 2 of the License, or
4848 (at your option) any later version.
4850 This program is distributed in the hope that it will be useful,
4851 but WITHOUT ANY WARRANTY; without even the implied warranty of
4852 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
4853 GNU General Public License for more details.
4855 You should have received a copy of the GNU General Public License
4856 along with this program; if not, write to the Free Software
4857 Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
4860 Also add information on how to contact you by electronic and paper mail.
4862 If the program is interactive, make it output a short notice like this
4863 when it starts in an interactive mode:
4866 Gnomovision version 69, Copyright (C) @var{yyyy} @var{name of author}
4867 Gnomovision comes with ABSOLUTELY NO WARRANTY; for details
4869 This is free software, and you are welcome to redistribute it
4870 under certain conditions; type `show c' for details.
4873 The hypothetical commands @samp{show w} and @samp{show c} should show
4874 the appropriate parts of the General Public License. Of course, the
4875 commands you use may be called something other than @samp{show w} and
4876 @samp{show c}; they could even be mouse-clicks or menu items---whatever
4879 You should also get your employer (if you work as a programmer) or your
4880 school, if any, to sign a ``copyright disclaimer'' for the program, if
4881 necessary. Here is a sample; alter the names:
4884 Yoyodyne, Inc., hereby disclaims all copyright interest in the program
4885 `Gnomovision' (which makes passes at compilers) written by James Hacker.
4887 @var{signature of Ty Coon}, 1 April 1989
4888 Ty Coon, President of Vice
4891 This General Public License does not permit incorporating your program into
4892 proprietary programs. If your program is a subroutine library, you may
4893 consider it more useful to permit linking proprietary applications with the
4894 library. If this is what you want to do, use the GNU Library General
4895 Public License instead of this License.
4898 @unnumbered Contributors to GCC
4899 @cindex contributors
4900 @include contrib.texi