1 \input texinfo.tex @c -*-texinfo-*-
4 @setfilename gccinstall.info
5 @settitle Installing GCC
10 @c Specify title for specific html page
12 @settitle Installing GCC
15 @settitle Host/Target specific installation notes for GCC
17 @ifset prerequisiteshtml
18 @settitle Prerequisites for GCC
21 @settitle Downloading GCC
24 @settitle Installing GCC: Configuration
27 @settitle Installing GCC: Building
30 @settitle Installing GCC: Testing
32 @ifset finalinstallhtml
33 @settitle Installing GCC: Final installation
36 @settitle Installing GCC: Binaries
39 @settitle Installing GCC: Old documentation
42 @settitle Installing GCC: GNU Free Documentation License
45 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
46 @c 1999, 2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
47 @c *** Converted to texinfo by Dean Wakerley, dean@wakerley.com
49 @c Include everything if we're not making html
53 @set prerequisiteshtml
64 @c Part 2 Summary Description and Copyright
66 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
67 1999, 2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
69 Permission is granted to copy, distribute and/or modify this document
70 under the terms of the GNU Free Documentation License, Version 1.2 or
71 any later version published by the Free Software Foundation; with no
72 Invariant Sections, the Front-Cover texts being (a) (see below), and
73 with the Back-Cover Texts being (b) (see below). A copy of the
74 license is included in the section entitled ``@uref{./gfdl.html,,GNU
75 Free Documentation License}''.
77 (a) The FSF's Front-Cover Text is:
81 (b) The FSF's Back-Cover Text is:
83 You have freedom to copy and modify this GNU Manual, like GNU
84 software. Copies published by the Free Software Foundation raise
85 funds for GNU development.
90 @dircategory Programming
92 * gccinstall: (gccinstall). Installing the GNU Compiler Collection.
95 @c Part 3 Titlepage and Copyright
98 @comment The title is printed in a large font.
99 @center @titlefont{Installing GCC}
101 @c The following two commands start the copyright page.
103 @vskip 0pt plus 1filll
107 @c Part 4 Top node and Master Menu
110 @comment node-name, next, Previous, up
113 * Installing GCC:: This document describes the generic installation
114 procedure for GCC as well as detailing some target
115 specific installation instructions.
117 * Specific:: Host/target specific installation notes for GCC.
118 * Binaries:: Where to get pre-compiled binaries.
120 * Old:: Old installation documentation.
122 * GNU Free Documentation License:: How you can copy and share this manual.
123 * Concept Index:: This index has two entries.
127 @c Part 5 The Body of the Document
128 @c ***Installing GCC**********************************************************
130 @comment node-name, next, previous, up
131 @node Installing GCC, Binaries, , Top
135 @chapter Installing GCC
138 The latest version of this document is always available at
139 @uref{http://gcc.gnu.org/install/,,http://gcc.gnu.org/install/}.
141 This document describes the generic installation procedure for GCC as well
142 as detailing some target specific installation instructions.
144 GCC includes several components that previously were separate distributions
145 with their own installation instructions. This document supersedes all
146 package specific installation instructions.
148 @emph{Before} starting the build/install procedure please check the
150 @ref{Specific, host/target specific installation notes}.
153 @uref{specific.html,,host/target specific installation notes}.
155 We recommend you browse the entire generic installation instructions before
158 Lists of successful builds for released versions of GCC are
159 available at @uref{http://gcc.gnu.org/buildstat.html}.
160 These lists are updated as new information becomes available.
162 The installation procedure itself is broken into five steps.
167 * Downloading the source::
170 * Testing:: (optional)
177 @uref{prerequisites.html,,Prerequisites}
179 @uref{download.html,,Downloading the source}
181 @uref{configure.html,,Configuration}
183 @uref{build.html,,Building}
185 @uref{test.html,,Testing} (optional)
187 @uref{finalinstall.html,,Final install}
191 Please note that GCC does not support @samp{make uninstall} and probably
192 won't do so in the near future as this would open a can of worms. Instead,
193 we suggest that you install GCC into a directory of its own and simply
194 remove that directory when you do not need that specific version of GCC
195 any longer, and, if shared libraries are installed there as well, no
196 more binaries exist that use them.
199 There are also some @uref{old.html,,old installation instructions},
200 which are mostly obsolete but still contain some information which has
201 not yet been merged into the main part of this manual.
209 @uref{./index.html,,Return to the GCC Installation page}
215 @c ***Prerequisites**************************************************
217 @comment node-name, next, previous, up
218 @node Prerequisites, Downloading the source, , Installing GCC
220 @ifset prerequisiteshtml
222 @chapter Prerequisites
224 @cindex Prerequisites
226 GCC requires that various tools and packages be available for use in the
227 build procedure. Modifying GCC sources requires additional tools
230 @heading Tools/packages necessary for building GCC
232 @item ISO C90 compiler
233 Necessary to bootstrap the GCC package, although versions of GCC prior
234 to 3.4 also allow bootstrapping with a traditional (K&R) C compiler.
236 To make all languages in a cross-compiler or other configuration where
237 3-stage bootstrap is not performed, you need to start with an existing
238 GCC binary (version 2.95 or later) because source code for language
239 frontends other than C might use GCC extensions.
243 In order to build the Ada compiler (GNAT) you must already have GNAT
244 installed because portions of the Ada frontend are written in Ada (with
245 GNAT extensions.) Refer to the Ada installation instructions for more
246 specific information.
248 @item A ``working'' POSIX compatible shell, or GNU bash
250 Necessary when running @command{configure} because some
251 @command{/bin/sh} shells have bugs and may crash when configuring the
252 target libraries. In other cases, @command{/bin/sh} or even some
253 @command{ksh} have disastrous corner-case performance problems. This
254 can cause target @command{configure} runs to literally take days to
255 complete in some cases.
257 So on some platforms @command{/bin/ksh} is sufficient, on others it
258 isn't. See the host/target specific instructions for your platform, or
259 use @command{bash} to be sure. Then set @env{CONFIG_SHELL} in your
260 environment to your ``good'' shell prior to running
261 @command{configure}/@command{make}.
263 @command{zsh} is not a fully compliant POSIX shell and will not
264 work when configuring GCC.
268 Necessary in some circumstances, optional in others. See the
269 host/target specific instructions for your platform for the exact
272 @item gzip version 1.2.4 (or later) or
273 @itemx bzip2 version 1.0.2 (or later)
275 Necessary to uncompress GCC @command{tar} files when source code is
276 obtained via FTP mirror sites.
278 @item GNU make version 3.79.1 (or later)
280 You must have GNU make installed to build GCC.
282 @item GNU tar version 1.12 (or later)
284 Necessary (only on some platforms) to untar the source code. Many
285 systems' @command{tar} programs will also work, only try GNU
286 @command{tar} if you have problems.
291 @heading Tools/packages necessary for modifying GCC
293 @item autoconf versions 2.13 and 2.57
294 @itemx GNU m4 version 1.4 (or later)
296 Necessary when modifying @file{configure.in}, @file{aclocal.m4}, etc.@:
297 to regenerate @file{configure} and @file{config.in} files. Most
298 directories require autoconf 2.13 (exactly), but @file{libiberty},
299 @file{fastjar}, @file{libstdc++-v3}, @file{libjava/libltdl}, and @file{gcc}
300 require autoconf 2.57 (exactly).
302 @item automake versions 1.4-gcj and 1.7.9
304 Necessary when modifying a @file{Makefile.am} file to regenerate its
305 associated @file{Makefile.in}.
307 Much of GCC does not use automake, so directly edit the @file{Makefile.in}
308 file. Specifically this applies to the @file{gcc}, @file{intl},
309 @file{libf2c}, @file{libiberty}, @file{libobjc} directories as well as any
310 of their subdirectories.
312 The @file{libstdc++-v3}, @file{libjava/libltdl}, and @file{fastjar}
313 directories require automake 1.7.9. However, the Java directories, which
314 include @file{boehm-gc}, @file{libffi}, @file{libjava}, and @file{zlib},
315 require a modified version of automake 1.4 downloadable from
316 @uref{ftp://gcc.gnu.org/pub/java/automake-gcj-1.4.tar.gz}.
318 @item gettext version 0.12 (or later)
320 Needed to regenerate @file{gcc.pot}.
322 @item gperf version 2.7.2 (or later)
324 Necessary when modifying @command{gperf} input files, e.g.@:
325 @file{gcc/cp/cfns.gperf} to regenerate its associated header file, e.g.@:
326 @file{gcc/cp/cfns.h}.
328 @item expect version ???
329 @itemx tcl version ???
330 @itemx dejagnu version ???
332 Necessary to run the GCC testsuite.
334 @item autogen version 5.5.4 (or later) and
335 @itemx guile version 1.4.1 (or later)
337 Necessary to regenerate @file{fixinc/fixincl.x} from
338 @file{fixinc/inclhack.def} and @file{fixinc/*.tpl}.
340 Necessary to run the @file{fixinc} @command{make check}.
342 Necessary to regenerate the top level @file{Makefile.in} file from
343 @file{Makefile.tpl} and @file{Makefile.def}.
345 @item GNU Bison version 1.28 (or later)
346 Berkeley @command{yacc} (@command{byacc}) is also reported to work other
349 Necessary when modifying @file{*.y} files.
351 Necessary to build GCC during development because the generated output
352 files are not included in the CVS repository. They are included in
355 @item Flex version 2.5.4 (or later)
357 Necessary when modifying @file{*.l} files.
359 Necessary to build GCC during development because the generated output
360 files are not included in the CVS repository. They are included in
363 @item Texinfo version 4.2 (or later)
365 Necessary for running @command{makeinfo} when modifying @file{*.texi}
366 files to test your changes.
368 Necessary to build GCC documentation during development because the
369 generated output files are not included in the CVS repository. They are
370 included in releases.
372 @item @TeX{} (any working version)
374 Necessary for running @command{texi2dvi}, used when running
375 @command{make dvi} to create DVI files.
377 @item cvs version 1.10 (or later)
378 @itemx ssh (any version)
380 Necessary to access the CVS repository. Public releases and weekly
381 snapshots of the development sources are also available via FTP.
383 @item perl version 5.6.1 (or later)
385 Necessary when regenerating @file{Makefile} dependencies in libiberty.
386 Necessary when regenerating @file{libiberty/functions.texi}.
387 Necessary when generating manpages from Texinfo manuals.
388 Used by various scripts to generate some files included in CVS (mainly
389 Unicode-related and rarely changing) from source tables.
391 @item GNU diffutils version 2.7 (or later)
393 Necessary when creating changes to GCC source code to submit for review.
395 @item patch version 2.5.4 (or later)
397 Necessary when applying patches, created with @command{diff}, to one's
407 @uref{./index.html,,Return to the GCC Installation page}
411 @c ***Downloading the source**************************************************
413 @comment node-name, next, previous, up
414 @node Downloading the source, Configuration, Prerequisites, Installing GCC
418 @chapter Downloading GCC
420 @cindex Downloading GCC
421 @cindex Downloading the Source
423 GCC is distributed via @uref{http://gcc.gnu.org/cvs.html,,CVS} and FTP
424 tarballs compressed with @command{gzip} or
425 @command{bzip2}. It is possible to download a full distribution or specific
428 Please refer to the @uref{http://gcc.gnu.org/releases.html,,releases web page}
429 for information on how to obtain GCC@.
431 The full distribution includes the C, C++, Objective-C, Fortran, Java,
432 and Ada (in case of GCC 3.1 and later) compilers. The full distribution
433 also includes runtime libraries for C++, Objective-C, Fortran, and Java.
434 In GCC 3.0 and later versions, GNU compiler testsuites are also included
435 in the full distribution.
437 If you choose to download specific components, you must download the core
438 GCC distribution plus any language specific distributions you wish to
439 use. The core distribution includes the C language front end as well as the
440 shared components. Each language has a tarball which includes the language
441 front end as well as the language runtime (when appropriate).
443 Unpack the core distribution as well as any language specific
444 distributions in the same directory.
446 If you also intend to build binutils (either to upgrade an existing
447 installation or for use in place of the corresponding tools of your
448 OS), unpack the binutils distribution either in the same directory or
449 a separate one. In the latter case, add symbolic links to any
450 components of the binutils you intend to build alongside the compiler
451 (@file{bfd}, @file{binutils}, @file{gas}, @file{gprof}, @file{ld},
452 @file{opcodes}, @dots{}) to the directory containing the GCC sources.
459 @uref{./index.html,,Return to the GCC Installation page}
463 @c ***Configuration***********************************************************
465 @comment node-name, next, previous, up
466 @node Configuration, Building, Downloading the source, Installing GCC
470 @chapter Installing GCC: Configuration
472 @cindex Configuration
473 @cindex Installing GCC: Configuration
475 Like most GNU software, GCC must be configured before it can be built.
476 This document describes the recommended configuration procedure
477 for both native and cross targets.
479 We use @var{srcdir} to refer to the toplevel source directory for
480 GCC; we use @var{objdir} to refer to the toplevel build/object directory.
482 If you obtained the sources via CVS, @var{srcdir} must refer to the top
483 @file{gcc} directory, the one where the @file{MAINTAINERS} can be found,
484 and not its @file{gcc} subdirectory, otherwise the build will fail.
486 If either @var{srcdir} or @var{objdir} is located on an automounted NFS
487 file system, the shell's built-in @command{pwd} command will return
488 temporary pathnames. Using these can lead to various sorts of build
489 problems. To avoid this issue, set the @env{PWDCMD} environment
490 variable to an automounter-aware @command{pwd} command, e.g.,
491 @command{pawd} or @samp{amq -w}, during the configuration and build
494 First, we @strong{highly} recommend that GCC be built into a
495 separate directory than the sources which does @strong{not} reside
496 within the source tree. This is how we generally build GCC; building
497 where @var{srcdir} == @var{objdir} should still work, but doesn't
498 get extensive testing; building where @var{objdir} is a subdirectory
499 of @var{srcdir} is unsupported.
501 If you have previously built GCC in the same directory for a
502 different target machine, do @samp{make distclean} to delete all files
503 that might be invalid. One of the files this deletes is @file{Makefile};
504 if @samp{make distclean} complains that @file{Makefile} does not exist
505 or issues a message like ``don't know how to make distclean'' it probably
506 means that the directory is already suitably clean. However, with the
507 recommended method of building in a separate @var{objdir}, you should
508 simply use a different @var{objdir} for each target.
510 Second, when configuring a native system, either @command{cc} or
511 @command{gcc} must be in your path or you must set @env{CC} in
512 your environment before running configure. Otherwise the configuration
515 Note that the bootstrap compiler and the resulting GCC must be link
516 compatible, else the bootstrap will fail with linker errors about
517 incompatible object file formats. Several multilibed targets are
518 affected by this requirement, see
520 @ref{Specific, host/target specific installation notes}.
523 @uref{specific.html,,host/target specific installation notes}.
531 % @var{srcdir}/configure [@var{options}] [@var{target}]
535 @heading Target specification
538 GCC has code to correctly determine the correct value for @var{target}
539 for nearly all native systems. Therefore, we highly recommend you not
540 provide a configure target when configuring a native compiler.
543 @var{target} must be specified as @option{--target=@var{target}}
544 when configuring a cross compiler; examples of valid targets would be
545 i960-rtems, m68k-coff, sh-elf, etc.
548 Specifying just @var{target} instead of @option{--target=@var{target}}
549 implies that the host defaults to @var{target}.
553 @heading Options specification
555 Use @var{options} to override several configure time options for
556 GCC@. A list of supported @var{options} follows; @samp{configure
557 --help} may list other options, but those not listed below may not
558 work and should not normally be used.
560 Note that each @option{--enable} option has a corresponding
561 @option{--disable} option and that each @option{--with} option has a
562 corresponding @option{--without} option.
565 @item --prefix=@var{dirname}
566 Specify the toplevel installation
567 directory. This is the recommended way to install the tools into a directory
568 other than the default. The toplevel installation directory defaults to
571 We @strong{highly} recommend against @var{dirname} being the same or a
572 subdirectory of @var{objdir} or vice versa. If specifying a directory
573 beneath a user's home directory tree, some shells will not expand
574 @var{dirname} correctly if it contains the @samp{~} metacharacter; use
577 The following standard @command{autoconf} options are supported. Normally you
578 should not need to use these options.
580 @item --exec-prefix=@var{dirname}
581 Specify the toplevel installation directory for architecture-dependent
582 files. The default is @file{@var{prefix}}.
584 @item --bindir=@var{dirname}
585 Specify the installation directory for the executables called by users
586 (such as @command{gcc} and @command{g++}). The default is
587 @file{@var{exec-prefix}/bin}.
589 @item --libdir=@var{dirname}
590 Specify the installation directory for object code libraries and
591 internal data files of GCC@. The default is @file{@var{exec-prefix}/lib}.
593 @item --libexecdir=@var{dirname}
594 Specify the installation directory for internal executables of GCC@.
595 The default is @file{@var{exec-prefix}/libexec}.
597 @item --with-slibdir=@var{dirname}
598 Specify the installation directory for the shared libgcc library. The
599 default is @file{@var{libdir}}.
601 @item --infodir=@var{dirname}
602 Specify the installation directory for documentation in info format.
603 The default is @file{@var{prefix}/info}.
605 @item --datadir=@var{dirname}
606 Specify the installation directory for some architecture-independent
607 data files referenced by GCC@. The default is @file{@var{prefix}/share}.
609 @item --mandir=@var{dirname}
610 Specify the installation directory for manual pages. The default is
611 @file{@var{prefix}/man}. (Note that the manual pages are only extracts from
612 the full GCC manuals, which are provided in Texinfo format. The manpages
613 are derived by an automatic conversion process from parts of the full
616 @item --with-gxx-include-dir=@var{dirname}
618 the installation directory for G++ header files. The default is
619 @file{@var{prefix}/include/c++/@var{version}}.
623 @item --program-prefix=@var{prefix}
624 GCC supports some transformations of the names of its programs when
625 installing them. This option prepends @var{prefix} to the names of
626 programs to install in @var{bindir} (see above). For example, specifying
627 @option{--program-prefix=foo-} would result in @samp{gcc}
628 being installed as @file{/usr/local/bin/foo-gcc}.
630 @item --program-suffix=@var{suffix}
631 Appends @var{suffix} to the names of programs to install in @var{bindir}
632 (see above). For example, specifying @option{--program-suffix=-3.1}
633 would result in @samp{gcc} being installed as
634 @file{/usr/local/bin/gcc-3.1}.
636 @item --program-transform-name=@var{pattern}
637 Applies the @samp{sed} script @var{pattern} to be applied to the names
638 of programs to install in @var{bindir} (see above). @var{pattern} has to
639 consist of one or more basic @samp{sed} editing commands, separated by
640 semicolons. For example, if you want the @samp{gcc} program name to be
641 transformed to the installed program @file{/usr/local/bin/myowngcc} and
642 the @samp{g++} program name to be transformed to
643 @file{/usr/local/bin/gspecial++} without changing other program names,
644 you could use the pattern
645 @option{--program-transform-name='s/^gcc$/myowngcc/; s/^g++$/gspecial++/'}
646 to achieve this effect.
648 All three options can be combined and used together, resulting in more
649 complex conversion patterns. As a basic rule, @var{prefix} (and
650 @var{suffix}) are prepended (appended) before further transformations
651 can happen with a special transformation script @var{pattern}.
653 As currently implemented, this option only takes effect for native
654 builds; cross compiler binaries' names are not transformed even when a
655 transformation is explicitly asked for by one of these options.
657 For native builds, some of the installed programs are also installed
658 with the target alias in front of their name, as in
659 @samp{i686-pc-linux-gnu-gcc}. All of the above transformations happen
660 before the target alias is prepended to the name - so, specifying
661 @option{--program-prefix=foo-} and @option{program-suffix=-3.1}, the
662 resulting binary would be installed as
663 @file{/usr/local/bin/i686-pc-linux-gnu-foo-gcc-3.1}.
665 As a last shortcoming, none of the installed Ada programs are
666 transformed yet, which will be fixed in some time.
668 @item --with-local-prefix=@var{dirname}
670 installation directory for local include files. The default is
671 @file{/usr/local}. Specify this option if you want the compiler to
672 search directory @file{@var{dirname}/include} for locally installed
673 header files @emph{instead} of @file{/usr/local/include}.
675 You should specify @option{--with-local-prefix} @strong{only} if your
676 site has a different convention (not @file{/usr/local}) for where to put
679 The default value for @option{--with-local-prefix} is @file{/usr/local}
680 regardless of the value of @option{--prefix}. Specifying
681 @option{--prefix} has no effect on which directory GCC searches for
682 local header files. This may seem counterintuitive, but actually it is
685 The purpose of @option{--prefix} is to specify where to @emph{install
686 GCC}. The local header files in @file{/usr/local/include}---if you put
687 any in that directory---are not part of GCC@. They are part of other
688 programs---perhaps many others. (GCC installs its own header files in
689 another directory which is based on the @option{--prefix} value.)
691 Both the local-prefix include directory and the GCC-prefix include
692 directory are part of GCC's "system include" directories. Although these
693 two directories are not fixed, they need to be searched in the proper
694 order for the correct processing of the include_next directive. The
695 local-prefix include directory is searched before the GCC-prefix
696 include directory. Another characteristic of system include directories
697 is that pedantic warnings are turned off for headers in these directories.
699 Some autoconf macros add @option{-I @var{directory}} options to the
700 compiler command line, to ensure that directories containing installed
701 packages' headers are searched. When @var{directory} is one of GCC's
702 system include directories, GCC will ignore the option so that system
703 directories continue to be processed in the correct order. This
704 may result in a search order different from what was specified but the
705 directory will still be searched.
707 GCC automatically searches for ordinary libraries using
708 @env{GCC_EXEC_PREFIX}. Thus, when the same installation prefix is
709 used for both GCC and packages, GCC will automatically search for
710 both headers and libraries. This provides a configuration that is
711 easy to use. GCC behaves in a manner similar to that when it is
712 installed as a system compiler in @file{/usr}.
714 Sites that need to install multiple versions of GCC may not want to
715 use the above simple configuration. It is possible to use the
716 @option{--program-prefix}, @option{--program-suffix} and
717 @option{--program-transform-name} options to install multiple versions
718 into a single directory, but it may be simpler to use different prefixes
719 and the @option{--with-local-prefix} option to specify the location of the
720 site-specific files for each version. It will then be necessary for
721 users to specify explicitly the location of local site libraries
722 (e.g., with @env{LIBRARY_PATH}).
724 The same value can be used for both @option{--with-local-prefix} and
725 @option{--prefix} provided it is not @file{/usr}. This can be used
726 to avoid the default search of @file{/usr/local/include}.
728 @strong{Do not} specify @file{/usr} as the @option{--with-local-prefix}!
729 The directory you use for @option{--with-local-prefix} @strong{must not}
730 contain any of the system's standard header files. If it did contain
731 them, certain programs would be miscompiled (including GNU Emacs, on
732 certain targets), because this would override and nullify the header
733 file corrections made by the @command{fixincludes} script.
735 Indications are that people who use this option use it based on mistaken
736 ideas of what it is for. People use it as if it specified where to
737 install part of GCC@. Perhaps they make this assumption because
738 installing GCC creates the directory.
740 @item --enable-shared[=@var{package}[,@dots{}]]
741 Build shared versions of libraries, if shared libraries are supported on
742 the target platform. Unlike GCC 2.95.x and earlier, shared libraries
743 are enabled by default on all platforms that support shared libraries,
744 except for @samp{libobjc} which is built as a static library only by
747 If a list of packages is given as an argument, build shared libraries
748 only for the listed packages. For other packages, only static libraries
749 will be built. Package names currently recognized in the GCC tree are
750 @samp{libgcc} (also known as @samp{gcc}), @samp{libstdc++} (not
751 @samp{libstdc++-v3}), @samp{libffi}, @samp{zlib}, @samp{boehm-gc} and
752 @samp{libjava}. Note that @samp{libobjc} does not recognize itself by
753 any name, so, if you list package names in @option{--enable-shared},
754 you will only get static Objective-C libraries. @samp{libf2c} and
755 @samp{libiberty} do not support shared libraries at all.
757 Use @option{--disable-shared} to build only static libraries. Note that
758 @option{--disable-shared} does not accept a list of package names as
759 argument, only @option{--enable-shared} does.
761 @item @anchor{with-gnu-as}--with-gnu-as
762 Specify that the compiler should assume that the
763 assembler it finds is the GNU assembler. However, this does not modify
764 the rules to find an assembler and will result in confusion if the
765 assembler found is not actually the GNU assembler. (Confusion may also
766 result if the compiler finds the GNU assembler but has not been
767 configured with @option{--with-gnu-as}.) If you have more than one
768 assembler installed on your system, you may want to use this option in
769 connection with @option{--with-as=@var{pathname}}.
771 The following systems are the only ones where it makes a difference
772 whether you use the GNU assembler. On any other system,
773 @option{--with-gnu-as} has no effect.
776 @item @samp{hppa1.0-@var{any}-@var{any}}
777 @item @samp{hppa1.1-@var{any}-@var{any}}
778 @item @samp{i386-@var{any}-sysv}
779 @item @samp{m68k-bull-sysv}
780 @item @samp{m68k-hp-hpux}
781 @item @samp{m68000-hp-hpux}
782 @item @samp{m68000-att-sysv}
783 @item @samp{@var{any}-lynx-lynxos}
784 @item @samp{mips-@var{any}}
785 @item @samp{sparc-sun-solaris2.@var{any}}
786 @item @samp{sparc64-@var{any}-solaris2.@var{any}}
789 On the systems listed above (except for the HP-PA, the SPARC, for ISC on
790 the 386, and for @samp{mips-sgi-irix5.*}), if you use the GNU assembler,
791 you should also use the GNU linker (and specify @option{--with-gnu-ld}).
793 @item @anchor{with-as}--with-as=@var{pathname}
795 compiler should use the assembler pointed to by @var{pathname}, rather
796 than the one found by the standard rules to find an assembler, which
800 Check the @file{@var{libexec}/gcc/@var{target}/@var{version}}
801 directory, where @var{libexec} defaults to
802 @file{@var{exec-prefix}/libexec} and @var{exec-prefix} defaults to
803 @var{prefix} which defaults to @file{/usr/local} unless overridden by
804 the @option{--prefix=@var{pathname}} switch described
805 above. @var{target} is the target system triple, such as
806 @samp{sparc-sun-solaris2.7}, and @var{version} denotes the GCC
807 version, such as 3.0.
809 Check operating system specific directories (e.g.@: @file{/usr/ccs/bin} on
812 Note that these rules do not check for the value of @env{PATH}. You may
813 want to use @option{--with-as} if no assembler is installed in the
814 directories listed above, or if you have multiple assemblers installed
815 and want to choose one that is not found by the above rules.
817 @item @anchor{with-gnu-ld}--with-gnu-ld
818 Same as @uref{#with-gnu-as,,@option{--with-gnu-as}}
821 @item --with-ld=@var{pathname}
822 Same as @uref{#with-as,,@option{--with-as}}
826 Specify that stabs debugging
827 information should be used instead of whatever format the host normally
828 uses. Normally GCC uses the same debug format as the host system.
830 On MIPS based systems and on Alphas, you must specify whether you want
831 GCC to create the normal ECOFF debugging format, or to use BSD-style
832 stabs passed through the ECOFF symbol table. The normal ECOFF debug
833 format cannot fully handle languages other than C@. BSD stabs format can
834 handle other languages, but it only works with the GNU debugger GDB@.
836 Normally, GCC uses the ECOFF debugging format by default; if you
837 prefer BSD stabs, specify @option{--with-stabs} when you configure GCC@.
839 No matter which default you choose when you configure GCC, the user
840 can use the @option{-gcoff} and @option{-gstabs+} options to specify explicitly
841 the debug format for a particular compilation.
843 @option{--with-stabs} is meaningful on the ISC system on the 386, also, if
844 @option{--with-gas} is used. It selects use of stabs debugging
845 information embedded in COFF output. This kind of debugging information
846 supports C++ well; ordinary COFF debugging information does not.
848 @option{--with-stabs} is also meaningful on 386 systems running SVR4. It
849 selects use of stabs debugging information embedded in ELF output. The
850 C++ compiler currently (2.6.0) does not support the DWARF debugging
851 information normally used on 386 SVR4 platforms; stabs provide a
852 workable alternative. This requires gas and gdb, as the normal SVR4
853 tools can not generate or interpret stabs.
855 @item --disable-multilib
856 Specify that multiple target
857 libraries to support different target variants, calling
858 conventions, etc should not be built. The default is to build a
859 predefined set of them.
861 Some targets provide finer-grained control over which multilibs are built
862 (e.g., @option{--disable-softfloat}):
868 fpu, 26bit, underscore, interwork, biendian, nofmult.
871 softfloat, m68881, m68000, m68020.
874 single-float, biendian, softfloat.
876 @item powerpc*-*-*, rs6000*-*-*
877 aix64, pthread, softfloat, powercpu, powerpccpu, powerpcos, biendian,
882 @item --enable-threads
883 Specify that the target
884 supports threads. This affects the Objective-C compiler and runtime
885 library, and exception handling for other languages like C++ and Java.
886 On some systems, this is the default.
888 In general, the best (and, in many cases, the only known) threading
889 model available will be configured for use. Beware that on some
890 systems, GCC has not been taught what threading models are generally
891 available for the system. In this case, @option{--enable-threads} is an
892 alias for @option{--enable-threads=single}.
894 @item --disable-threads
895 Specify that threading support should be disabled for the system.
896 This is an alias for @option{--enable-threads=single}.
898 @item --enable-threads=@var{lib}
900 @var{lib} is the thread support library. This affects the Objective-C
901 compiler and runtime library, and exception handling for other languages
902 like C++ and Java. The possibilities for @var{lib} are:
910 Ada tasking support. For non-Ada programs, this setting is equivalent
911 to @samp{single}. When used in conjunction with the Ada run time, it
912 causes GCC to use the same thread primitives as Ada uses. This option
913 is necessary when using both Ada and the back end exception handling,
914 which is the default for most Ada targets.
916 Generic MACH thread support, known to work on NeXTSTEP@. (Please note
917 that the file needed to support this configuration, @file{gthr-mach.h}, is
918 missing and thus this setting will cause a known bootstrap failure.)
920 This is an alias for @samp{single}.
922 Generic POSIX thread support.
924 RTEMS thread support.
926 Disable thread support, should work for all platforms.
928 Sun Solaris 2 thread support.
930 VxWorks thread support.
932 Microsoft Win32 API thread support.
935 @item --with-cpu=@var{cpu}
936 Specify which cpu variant the compiler should generate code for by default.
937 @var{cpu} will be used as the default value of the @option{-mcpu=} switch.
938 This option is only supported on some targets, including ARM, i386, PowerPC,
941 @item --with-schedule=@var{cpu}
942 @itemx --with-arch=@var{cpu}
943 @itemx --with-tune=@var{cpu}
944 @itemx --with-abi=@var{abi}
945 @itemx --with-float=@var{type}
946 These configure options provide default values for the @option{-mschedule=},
947 @option{-march=}, @option{-mtune=}, and @option{-mabi=} options and for
948 @option{-mhard-float} or @option{-msoft-float}. As with @option{--with-cpu},
949 which switches will be accepted and acceptable values of the arguments depend
952 @item --enable-altivec
953 Specify that the target supports AltiVec vector enhancements. This
954 option will adjust the ABI for AltiVec enhancements, as well as generate
955 AltiVec code when appropriate. This option is only available for
958 @item --enable-__cxa_atexit
959 Define if you want to use __cxa_atexit, rather than atexit, to
960 register C++ destructors for local statics and global objects.
961 This is essential for fully standards-compliant handling of
962 destructors, but requires __cxa_atexit in libc. This option is currently
963 only available on systems with GNU libc. When enabled, this will cause
964 @option{-fuse-cxa-exit} to be passed by default.
966 @item --enable-target-optspace
968 libraries should be optimized for code space instead of code speed.
969 This is the default for the m32r platform.
972 Specify that a user visible @command{cpp} program should not be installed.
974 @item --with-cpp-install-dir=@var{dirname}
975 Specify that the user visible @command{cpp} program should be installed
976 in @file{@var{prefix}/@var{dirname}/cpp}, in addition to @var{bindir}.
978 @item --enable-initfini-array
979 Force the use of sections @code{.init_array} and @code{.fini_array}
980 (instead of @code{.init} and @code{.fini}) for constructors and
981 destructors. Option @option{--disable-initfini-array} has the
982 opposite effect. If neither option is specified, the configure script
983 will try to guess whether the @code{.init_array} and
984 @code{.fini_array} sections are supported and, if they are, use them.
986 @item --enable-maintainer-mode
988 regenerate the GCC master message catalog @file{gcc.pot} are normally
989 disabled. This is because it can only be rebuilt if the complete source
990 tree is present. If you have changed the sources and want to rebuild the
991 catalog, configuring with @option{--enable-maintainer-mode} will enable
992 this. Note that you need a recent version of the @code{gettext} tools
995 @item --enable-generated-files-in-srcdir
996 Neither the .c and .h files that are generated from bison and flex nor the
997 info manuals and man pages that are built from the .texi files are present
998 in the CVS development tree. When building GCC from that development tree,
999 or from a snapshot which are created from CVS, then those generated files
1000 are placed in your build directory, which allows for the source to be in a
1003 If you configure with @option{--enable-generated-files-in-srcdir} then those
1004 generated files will go into the source directory. This is mainly intended
1005 for generating release or prerelease tarballs of the GCC sources, since it
1006 is not a requirement that the users of source releases to have flex, bison, or
1009 @item --enable-version-specific-runtime-libs
1011 that runtime libraries should be installed in the compiler specific
1012 subdirectory (@file{@var{libdir}/gcc}) rather than the usual places. In
1013 addition, @samp{libstdc++}'s include files will be installed into
1014 @file{@var{libdir}} unless you overruled it by using
1015 @option{--with-gxx-include-dir=@var{dirname}}. Using this option is
1016 particularly useful if you intend to use several versions of GCC in
1017 parallel. This is currently supported by @samp{libf2c} and
1018 @samp{libstdc++}, and is the default for @samp{libobjc} which cannot be
1019 changed in this case.
1021 @item --enable-languages=@var{lang1},@var{lang2},@dots{}
1022 Specify that only a particular subset of compilers and
1023 their runtime libraries should be built. For a list of valid values for
1024 @var{langN} you can issue the following command in the
1025 @file{gcc} directory of your GCC source tree:@*
1027 grep language= */config-lang.in
1029 Currently, you can use any of the following:
1030 @code{ada}, @code{c}, @code{c++}, @code{f77}, @code{java}, @code{objc}.
1031 Building the Ada compiler has special requirements, see below.@*
1032 If you do not pass this flag, all languages available in the @file{gcc}
1033 sub-tree will be configured. Re-defining @code{LANGUAGES} when calling
1034 @samp{make bootstrap} @strong{does not} work anymore, as those
1035 language sub-directories might not have been configured!
1038 Specify that the compiler should
1039 use DWARF 2 debugging information as the default.
1041 @item --enable-win32-registry
1042 @itemx --enable-win32-registry=@var{key}
1043 @itemx --disable-win32-registry
1044 The @option{--enable-win32-registry} option enables Microsoft Windows-hosted GCC
1045 to look up installations paths in the registry using the following key:
1048 @code{HKEY_LOCAL_MACHINE\SOFTWARE\Free Software Foundation\@var{key}}
1051 @var{key} defaults to GCC version number, and can be overridden by the
1052 @option{--enable-win32-registry=@var{key}} option. Vendors and distributors
1053 who use custom installers are encouraged to provide a different key,
1054 perhaps one comprised of vendor name and GCC version number, to
1055 avoid conflict with existing installations. This feature is enabled
1056 by default, and can be disabled by @option{--disable-win32-registry}
1057 option. This option has no effect on the other hosts.
1060 Specify that the machine does not have a floating point unit. This
1061 option only applies to @samp{m68k-sun-sunos@var{n}}. On any other
1062 system, @option{--nfp} has no effect.
1064 @item --enable-werror
1065 @itemx --disable-werror
1066 @itemx --enable-werror=yes
1067 @itemx --enable-werror=no
1068 When you specify this option, it controls whether certain files in the
1069 compiler are built with @option{-Werror} in bootstrap stage2 and later.
1070 If you don't specify it, @option{-Werror} is turned on for the main
1071 development trunk. However it defaults to off for release branches and
1072 final releases. The specific files which get @option{-Werror} are
1073 controlled by the Makefiles.
1075 @item --enable-checking
1076 @itemx --enable-checking=@var{list}
1077 When you specify this option, the compiler is built to perform checking
1078 of tree node types when referencing fields of that node, and some other
1079 internal consistency checks. This does not change the generated code,
1080 but adds error checking within the compiler. This will slow down the
1081 compiler and may only work properly if you are building the compiler
1082 with GCC@. This is on by default when building from CVS or snapshots,
1083 but off for releases. More control over the checks may be had by
1084 specifying @var{list}; the categories of checks available are
1085 @samp{misc}, @samp{tree}, @samp{gc}, @samp{rtl}, @samp{rtlflag},
1086 @samp{fold}, @samp{gcac} and @samp{valgrind}. The check @samp{valgrind}
1087 requires the external @command{valgrind} simulator, available from
1088 @uref{http://valgrind.kde.org/}. The default when @var{list} is
1089 not specified is @samp{misc,tree,gc,rtlflag}; the checks @samp{rtl},
1090 @samp{gcac} and @samp{valgrind} are very expensive.
1092 @item --enable-coverage
1093 @itemx --enable-coverage=@var{level}
1094 With this option, the compiler is built to collect self coverage
1095 information, every time it is run. This is for internal development
1096 purposes, and only works when the compiler is being built with gcc. The
1097 @var{level} argument controls whether the compiler is built optimized or
1098 not, values are @samp{opt} and @samp{noopt}. For coverage analysis you
1099 want to disable optimization, for performance analysis you want to
1100 enable optimization. When coverage is enabled, the default level is
1101 without optimization.
1103 @item --enable-gather-detailed-mem-stats
1104 When this option is specified more detailed information on memory
1105 allocation is gathered. This information is printed when using
1106 @option{-fmem-report}.
1109 @itemx --disable-nls
1110 The @option{--enable-nls} option enables Native Language Support (NLS),
1111 which lets GCC output diagnostics in languages other than American
1112 English. Native Language Support is enabled by default if not doing a
1113 canadian cross build. The @option{--disable-nls} option disables NLS@.
1115 @item --with-included-gettext
1116 If NLS is enabled, the @option{--with-included-gettext} option causes the build
1117 procedure to prefer its copy of GNU @command{gettext}.
1119 @item --with-catgets
1120 If NLS is enabled, and if the host lacks @code{gettext} but has the
1121 inferior @code{catgets} interface, the GCC build procedure normally
1122 ignores @code{catgets} and instead uses GCC's copy of the GNU
1123 @code{gettext} library. The @option{--with-catgets} option causes the
1124 build procedure to use the host's @code{catgets} in this situation.
1126 @item --with-libiconv-prefix=@var{dir}
1127 Search for libiconv header files in @file{@var{dir}/include} and
1128 libiconv library files in @file{@var{dir}/lib}.
1130 @item --enable-obsolete
1131 Enable configuration for an obsoleted system. If you attempt to
1132 configure GCC for a system (build, host, or target) which has been
1133 obsoleted, and you do not specify this flag, configure will halt with an
1136 All support for systems which have been obsoleted in one release of GCC
1137 is removed entirely in the next major release, unless someone steps
1138 forward to maintain the port.
1141 @subheading Cross-Compiler-Specific Options
1142 The following options only apply to building cross compilers.
1144 @item --with-sysroot
1145 @itemx --with-sysroot=@var{dir}
1146 Tells GCC to consider @var{dir} as the root of a tree that contains a
1147 (subset of) the root filesystem of the target operating system.
1148 Target system headers, libraries and run-time object files will be
1149 searched in there. The specified directory is not copied into the
1150 install tree, unlike the options @option{--with-headers} and
1151 @option{--with-libs} that this option obsoletes. The default value,
1152 in case @option{--with-sysroot} is not given an argument, is
1153 @option{$@{gcc_tooldir@}/sys-root}. If the specified directory is a
1154 subdirectory of @option{$@{exec_prefix@}}, then it will be found relative to
1155 the GCC binaries if the installation tree is moved.
1157 @item --with-headers
1158 @itemx --with-headers=@var{dir}
1159 Deprecated in favor of @option{--with-sysroot}.
1160 Specifies that target headers are available when building a cross compiler.
1161 The @var{dir} argument specifies a directory which has the target include
1162 files. These include files will be copied into the @file{gcc} install
1163 directory. @emph{This option with the @var{dir} argument is required} when
1164 building a cross compiler, if @file{@var{prefix}/@var{target}/sys-include}
1165 doesn't pre-exist. If @file{@var{prefix}/@var{target}/sys-include} does
1166 pre-exist, the @var{dir} argument may be omitted. @command{fixincludes}
1167 will be run on these files to make them compatible with GCC.
1169 @item --without-headers
1170 Tells GCC not use any target headers from a libc when building a cross
1171 compiler. When crossing to GNU/Linux, you need the headers so GCC
1172 can build the exception handling for libgcc.
1173 See @uref{http://www.objsw.com/CrossGCC/,,CrossGCC} for more information
1177 @itemx --with-libs=``@var{dir1} @var{dir2} @dots{} @var{dirN}''
1178 Deprecated in favor of @option{--with-sysroot}.
1179 Specifies a list of directories which contain the target runtime
1180 libraries. These libraries will be copied into the @file{gcc} install
1181 directory. If the directory list is omitted, this option has no
1184 Specifies that @samp{newlib} is
1185 being used as the target C library. This causes @code{__eprintf} to be
1186 omitted from @file{libgcc.a} on the assumption that it will be provided by
1190 @subheading Java-Specific Options
1192 The following option applies to the build of the Java front end.
1195 @item --disable-libgcj
1196 Specify that the run-time libraries
1197 used by GCJ should not be built. This is useful in case you intend
1198 to use GCJ with some other run-time, or you're going to install it
1199 separately, or it just happens not to build on your particular
1200 machine. In general, if the Java front end is enabled, the GCJ
1201 libraries will be enabled too, unless they're known to not work on
1202 the target platform. If GCJ is enabled but @samp{libgcj} isn't built, you
1203 may need to port it; in this case, before modifying the top-level
1204 @file{configure.in} so that @samp{libgcj} is enabled by default on this platform,
1205 you may use @option{--enable-libgcj} to override the default.
1209 The following options apply to building @samp{libgcj}.
1211 @subsubheading General Options
1214 @item --disable-getenv-properties
1215 Don't set system properties from @env{GCJ_PROPERTIES}.
1217 @item --enable-hash-synchronization
1218 Use a global hash table for monitor locks. Ordinarily,
1219 @samp{libgcj}'s @samp{configure} script automatically makes
1220 the correct choice for this option for your platform. Only use
1221 this if you know you need the library to be configured differently.
1223 @item --enable-interpreter
1224 Enable the Java interpreter. The interpreter is automatically
1225 enabled by default on all platforms that support it. This option
1226 is really only useful if you want to disable the interpreter
1227 (using @option{--disable-interpreter}).
1229 @item --disable-java-net
1230 Disable java.net. This disables the native part of java.net only,
1231 using non-functional stubs for native method implementations.
1233 @item --disable-jvmpi
1234 Disable JVMPI support.
1237 Enable runtime eCos target support.
1239 @item --without-libffi
1240 Don't use @samp{libffi}. This will disable the interpreter and JNI
1241 support as well, as these require @samp{libffi} to work.
1243 @item --enable-libgcj-debug
1244 Enable runtime debugging code.
1246 @item --enable-libgcj-multifile
1247 If specified, causes all @file{.java} source files to be
1248 compiled into @file{.class} files in one invocation of
1249 @samp{gcj}. This can speed up build time, but is more
1250 resource-intensive. If this option is unspecified or
1251 disabled, @samp{gcj} is invoked once for each @file{.java}
1252 file to compile into a @file{.class} file.
1254 @item --with-libiconv-prefix=DIR
1255 Search for libiconv in @file{DIR/include} and @file{DIR/lib}.
1257 @item --enable-sjlj-exceptions
1258 Force use of @code{builtin_setjmp} for exceptions. @samp{configure}
1259 ordinarily picks the correct value based on the platform. Only use
1260 this option if you are sure you need a different setting.
1262 @item --with-system-zlib
1263 Use installed @samp{zlib} rather than that included with GCC@.
1265 @item --with-win32-nlsapi=ansi, unicows or unicode
1266 Indicates how MinGW @samp{libgcj} translates between UNICODE
1267 characters and the Win32 API.
1270 Use the single-byte @code{char} and the Win32 A functions natively,
1271 translating to and from UNICODE when using these functions. If
1272 unspecified, this is the default.
1275 Use the @code{WCHAR} and Win32 W functions natively. Adds
1276 @code{-lunicows} to @file{libgcj.spec} to link with @samp{libunicows}.
1277 @file{unicows.dll} needs to be deployed on Microsoft Windows 9X machines
1278 running built executables. @file{libunicows.a}, an open-source
1279 import library around Microsoft's @code{unicows.dll}, is obtained from
1280 @uref{http://libunicows.sourceforge.net/}, which also gives details
1281 on getting @file{unicows.dll} from Microsoft.
1284 Use the @code{WCHAR} and Win32 W functions natively. Does @emph{not}
1285 add @code{-lunicows} to @file{libgcj.spec}. The built executables will
1286 only run on Microsoft Windows NT and above.
1290 @subsubheading AWT-Specific Options
1294 Use the X Window System.
1296 @item --enable-java-awt=PEER(S)
1297 Specifies the AWT peer library or libraries to build alongside
1298 @samp{libgcj}. If this option is unspecified or disabled, AWT
1299 will be non-functional. Current valid values are @option{gtk} and
1300 @option{xlib}. Multiple libraries should be separated by a
1301 comma (i.e. @option{--enable-java-awt=gtk,xlib}).
1303 @item --enable-gtk-cairo
1304 Build the cairo Graphics2D implementation on GTK.
1306 @item --enable-java-gc=TYPE
1307 Choose garbage collector. Defaults to @option{boehm} if unspecified.
1309 @item --disable-gtktest
1310 Do not try to compile and run a test GTK+ program.
1312 @item --disable-glibtest
1313 Do not try to compile and run a test GLIB program.
1315 @item --with-libart-prefix=PFX
1316 Prefix where libart is installed (optional).
1318 @item --with-libart-exec-prefix=PFX
1319 Exec prefix where libart is installed (optional).
1321 @item --disable-libarttest
1322 Do not try to compile and run a test libart program.
1331 @uref{./index.html,,Return to the GCC Installation page}
1335 @c ***Building****************************************************************
1337 @comment node-name, next, previous, up
1338 @node Building, Testing, Configuration, Installing GCC
1344 @cindex Installing GCC: Building
1346 Now that GCC is configured, you are ready to build the compiler and
1349 Some commands executed when making the compiler may fail (return a
1350 nonzero status) and be ignored by @command{make}. These failures, which
1351 are often due to files that were not found, are expected, and can safely
1354 It is normal to have compiler warnings when compiling certain files.
1355 Unless you are a GCC developer, you can generally ignore these warnings
1356 unless they cause compilation to fail. Developers should attempt to fix
1357 any warnings encountered, however they can temporarily continue past
1358 warnings-as-errors by specifying the configure flag
1359 @option{--disable-werror}.
1361 On certain old systems, defining certain environment variables such as
1362 @env{CC} can interfere with the functioning of @command{make}.
1364 If you encounter seemingly strange errors when trying to build the
1365 compiler in a directory other than the source directory, it could be
1366 because you have previously configured the compiler in the source
1367 directory. Make sure you have done all the necessary preparations.
1369 If you build GCC on a BSD system using a directory stored in an old System
1370 V file system, problems may occur in running @command{fixincludes} if the
1371 System V file system doesn't support symbolic links. These problems
1372 result in a failure to fix the declaration of @code{size_t} in
1373 @file{sys/types.h}. If you find that @code{size_t} is a signed type and
1374 that type mismatches occur, this could be the cause.
1376 The solution is not to use such a directory for building GCC@.
1378 When building from CVS or snapshots, or if you modify parser sources,
1379 you need the Bison parser generator installed. Any version 1.25 or
1380 later should work; older versions may also work. If you do not modify
1381 parser sources, releases contain the Bison-generated files and you do
1382 not need Bison installed to build them.
1384 When building from CVS or snapshots, or if you modify Texinfo
1385 documentation, you need version 4.2 or later of Texinfo installed if you
1386 want Info documentation to be regenerated. Releases contain Info
1387 documentation pre-built for the unmodified documentation in the release.
1389 @section Building a native compiler
1391 For a native build issue the command @samp{make bootstrap}. This
1392 will build the entire GCC system, which includes the following steps:
1396 Build host tools necessary to build the compiler such as texinfo, bison,
1400 Build target tools for use by the compiler such as binutils (bfd,
1401 binutils, gas, gprof, ld, and opcodes)
1402 if they have been individually linked
1403 or moved into the top level GCC source tree before configuring.
1406 Perform a 3-stage bootstrap of the compiler.
1409 Perform a comparison test of the stage2 and stage3 compilers.
1412 Build runtime libraries using the stage3 compiler from the previous step.
1416 If you are short on disk space you might consider @samp{make
1417 bootstrap-lean} instead. This is identical to @samp{make
1418 bootstrap} except that object files from the stage1 and
1419 stage2 of the 3-stage bootstrap of the compiler are deleted as
1420 soon as they are no longer needed.
1422 If you want to save additional space during the bootstrap and in
1423 the final installation as well, you can build the compiler binaries
1424 without debugging information as in the following example. This will save
1425 roughly 40% of disk space both for the bootstrap and the final installation.
1426 (Libraries will still contain debugging information.)
1429 make CFLAGS='-O' LIBCFLAGS='-g -O2' \
1430 LIBCXXFLAGS='-g -O2 -fno-implicit-templates' bootstrap
1433 If you wish to use non-default GCC flags when compiling the stage2 and
1434 stage3 compilers, set @code{BOOT_CFLAGS} on the command line when doing
1435 @samp{make bootstrap}. Non-default optimization flags are less well
1436 tested here than the default of @samp{-g -O2}, but should still work.
1437 In a few cases, you may find that you need to specify special flags such
1438 as @option{-msoft-float} here to complete the bootstrap; or, if the
1439 native compiler miscompiles the stage1 compiler, you may need to work
1440 around this, by choosing @code{BOOT_CFLAGS} to avoid the parts of the
1441 stage1 compiler that were miscompiled, or by using @samp{make
1442 bootstrap4} to increase the number of stages of bootstrap.
1444 If you used the flag @option{--enable-languages=@dots{}} to restrict
1445 the compilers to be built, only those you've actually enabled will be
1446 built. This will of course only build those runtime libraries, for
1447 which the particular compiler has been built. Please note,
1448 that re-defining @env{LANGUAGES} when calling @samp{make bootstrap}
1449 @strong{does not} work anymore!
1451 If the comparison of stage2 and stage3 fails, this normally indicates
1452 that the stage2 compiler has compiled GCC incorrectly, and is therefore
1453 a potentially serious bug which you should investigate and report. (On
1454 a few systems, meaningful comparison of object files is impossible; they
1455 always appear ``different''. If you encounter this problem, you will
1456 need to disable comparison in the @file{Makefile}.)
1458 @section Building a cross compiler
1460 We recommend reading the
1461 @uref{http://www.objsw.com/CrossGCC/,,crossgcc FAQ}
1462 for information about building cross compilers.
1464 When building a cross compiler, it is not generally possible to do a
1465 3-stage bootstrap of the compiler. This makes for an interesting problem
1466 as parts of GCC can only be built with GCC@.
1468 To build a cross compiler, we first recommend building and installing a
1469 native compiler. You can then use the native GCC compiler to build the
1470 cross compiler. The installed native compiler needs to be GCC version
1473 Assuming you have already installed a native copy of GCC and configured
1474 your cross compiler, issue the command @command{make}, which performs the
1479 Build host tools necessary to build the compiler such as texinfo, bison,
1483 Build target tools for use by the compiler such as binutils (bfd,
1484 binutils, gas, gprof, ld, and opcodes)
1485 if they have been individually linked or moved into the top level GCC source
1486 tree before configuring.
1489 Build the compiler (single stage only).
1492 Build runtime libraries using the compiler from the previous step.
1495 Note that if an error occurs in any step the make process will exit.
1497 If you are not building GNU binutils in the same source tree as GCC,
1498 you will need a cross-assembler and cross-linker installed before
1499 configuring GCC@. Put them in the directory
1500 @file{@var{prefix}/@var{target}/bin}. Here is a table of the tools
1501 you should put in this directory:
1505 This should be the cross-assembler.
1508 This should be the cross-linker.
1511 This should be the cross-archiver: a program which can manipulate
1512 archive files (linker libraries) in the target machine's format.
1515 This should be a program to construct a symbol table in an archive file.
1518 The installation of GCC will find these programs in that directory,
1519 and copy or link them to the proper place to for the cross-compiler to
1520 find them when run later.
1522 The easiest way to provide these files is to build the Binutils package.
1523 Configure it with the same @option{--host} and @option{--target}
1524 options that you use for configuring GCC, then build and install
1525 them. They install their executables automatically into the proper
1526 directory. Alas, they do not support all the targets that GCC
1529 If you are not building a C library in the same source tree as GCC,
1530 you should also provide the target libraries and headers before
1531 configuring GCC, specifying the directories with
1532 @option{--with-sysroot} or @option{--with-headers} and
1533 @option{--with-libs}. Many targets also require ``start files'' such
1534 as @file{crt0.o} and
1535 @file{crtn.o} which are linked into each executable. There may be several
1536 alternatives for @file{crt0.o}, for use with profiling or other
1537 compilation options. Check your target's definition of
1538 @code{STARTFILE_SPEC} to find out what start files it uses.
1540 @section Building in parallel
1542 You can use @samp{make bootstrap MAKE="make -j 2" -j 2}, or just
1543 @samp{make -j 2 bootstrap} for GNU Make 3.79 and above, instead of
1544 @samp{make bootstrap} to build GCC in parallel.
1545 You can also specify a bigger number, and in most cases using a value
1546 greater than the number of processors in your machine will result in
1547 fewer and shorter I/O latency hits, thus improving overall throughput;
1548 this is especially true for slow drives and network filesystems.
1550 @section Building the Ada compiler
1552 In order to build GNAT, the Ada compiler, you need a working GNAT
1553 compiler (GNAT version 3.14 or later, or GCC version 3.1 or later),
1554 including GNAT tools such as @command{gnatmake} and @command{gnatlink},
1555 since the Ada front end is written in Ada (with some
1556 GNAT-specific extensions), and GNU make.
1558 @command{configure} does not test whether the GNAT installation works
1559 and has a sufficiently recent version; if too old a GNAT version is
1560 installed, the build will fail unless @option{--enable-languages} is
1561 used to disable building the Ada front end.
1563 At the moment, the GNAT library and several tools for GNAT are not built
1564 by @samp{make bootstrap}. For a native build, you have to invoke
1565 @samp{make gnatlib_and_tools} in the @file{@var{objdir}/gcc}
1566 subdirectory before proceeding with the next steps.
1567 For a cross build, you need to invoke
1568 @samp{make gnatlib cross-gnattools ada.all.cross}. For a canadian
1569 cross you only need to invoke @samp{make cross-gnattools}; the GNAT
1570 library would be the same as the one built for the cross compiler.
1572 For example, you can build a native Ada compiler by issuing the
1573 following commands (assuming @command{make} is GNU make):
1577 @var{srcdir}/configure --enable-languages=c,ada
1581 make gnatlib_and_tools
1585 Currently, when compiling the Ada front end, you cannot use the parallel
1586 build feature described in the previous section.
1588 @section Building with profile feedback
1590 It is possible to use profile feedback to optimize the compiler itself. This
1591 should result in a faster compiler binary. Experiments done on x86 using gcc
1592 3.3 showed approximately 7 percent speedup on compiling C programs. To
1593 bootstrap compiler with profile feedback, use @code{make profiledbootstrap}.
1595 When @samp{make profiledbootstrap} is run, it will first build a @code{stage1}
1596 compiler. This compiler is used to build a @code{stageprofile} compiler
1597 instrumented to collect execution counts of instruction and branch
1598 probabilities. Then runtime libraries are compiled with profile collected.
1599 Finally a @code{stagefeedback} compiler is built using the information collected.
1601 Unlike @samp{make bootstrap} several additional restrictions apply. The
1602 compiler used to build @code{stage1} needs to support a 64-bit integral type.
1603 It is recommended to only use GCC for this. Also parallel make is currently
1604 not supported since collisions in profile collecting may occur.
1611 @uref{./index.html,,Return to the GCC Installation page}
1615 @c ***Testing*****************************************************************
1617 @comment node-name, next, previous, up
1618 @node Testing, Final install, Building, Installing GCC
1622 @chapter Installing GCC: Testing
1625 @cindex Installing GCC: Testing
1628 Before you install GCC, we encourage you to run the testsuites and to
1629 compare your results with results from a similar configuration that have
1630 been submitted to the
1631 @uref{http://gcc.gnu.org/ml/gcc-testresults/,,gcc-testresults mailing list}.
1632 Some of these archived results are linked from the build status lists
1633 at @uref{http://gcc.gnu.org/buildstat.html}, although not everyone who
1634 reports a successful build runs the testsuites and submits the results.
1635 This step is optional and may require you to download additional software,
1636 but it can give you confidence in your new GCC installation or point out
1637 problems before you install and start using your new GCC.
1639 First, you must have @uref{download.html,,downloaded the testsuites}.
1640 These are part of the full distribution, but if you downloaded the
1641 ``core'' compiler plus any front ends, you must download the testsuites
1644 Second, you must have the testing tools installed. This includes
1645 @uref{http://www.gnu.org/software/dejagnu/,,DejaGnu} 1.4.1 or 1.4.3
1646 and later, Tcl, and Expect; the DejaGnu site has links to these.
1648 If the directories where @command{runtest} and @command{expect} were
1649 installed are not in the @env{PATH}, you may need to set the following
1650 environment variables appropriately, as in the following example (which
1651 assumes that DejaGnu has been installed under @file{/usr/local}):
1654 TCL_LIBRARY = /usr/local/share/tcl8.0
1655 DEJAGNULIBS = /usr/local/share/dejagnu
1658 (On systems such as Cygwin, these paths are required to be actual
1659 paths, not mounts or links; presumably this is due to some lack of
1660 portability in the DejaGnu code.)
1663 Finally, you can run the testsuite (which may take a long time):
1665 cd @var{objdir}; make -k check
1668 This will test various components of GCC, such as compiler
1669 front ends and runtime libraries. While running the testsuite, DejaGnu
1670 might emit some harmless messages resembling
1671 @samp{WARNING: Couldn't find the global config file.} or
1672 @samp{WARNING: Couldn't find tool init file} that can be ignored.
1674 @section How can you run the testsuite on selected tests?
1676 In order to run sets of tests selectively, there are targets
1677 @samp{make check-gcc} and @samp{make check-g++}
1678 in the @file{gcc} subdirectory of the object directory. You can also
1679 just run @samp{make check} in a subdirectory of the object directory.
1682 A more selective way to just run all @command{gcc} execute tests in the
1686 make check-gcc RUNTESTFLAGS="execute.exp @var{other-options}"
1689 Likewise, in order to run only the @command{g++} ``old-deja'' tests in
1690 the testsuite with filenames matching @samp{9805*}, you would use
1693 make check-g++ RUNTESTFLAGS="old-deja.exp=9805* @var{other-options}"
1696 The @file{*.exp} files are located in the testsuite directories of the GCC
1697 source, the most important ones being @file{compile.exp},
1698 @file{execute.exp}, @file{dg.exp} and @file{old-deja.exp}.
1699 To get a list of the possible @file{*.exp} files, pipe the
1700 output of @samp{make check} into a file and look at the
1701 @samp{Running @dots{} .exp} lines.
1703 @section Passing options and running multiple testsuites
1705 You can pass multiple options to the testsuite using the
1706 @samp{--target_board} option of DejaGNU, either passed as part of
1707 @samp{RUNTESTFLAGS}, or directly to @command{runtest} if you prefer to
1708 work outside the makefiles. For example,
1711 make check-g++ RUNTESTFLAGS="--target_board=unix/-O3/-fno-strength-reduce"
1714 will run the standard @command{g++} testsuites (``unix'' is the target name
1715 for a standard native testsuite situation), passing
1716 @samp{-O3 -fno-strength-reduce} to the compiler on every test, i.e.,
1717 slashes separate options.
1719 You can run the testsuites multiple times using combinations of options
1720 with a syntax similar to the brace expansion of popular shells:
1723 @dots{}"--target_board=arm-sim@{-mhard-float,-msoft-float@}@{-O1,-O2,-O3,@}"
1726 (Note the empty option caused by the trailing comma in the final group.)
1727 The following will run each testsuite eight times using the @samp{arm-sim}
1728 target, as if you had specified all possible combinations yourself:
1731 --target_board=arm-sim/-mhard-float/-O1
1732 --target_board=arm-sim/-mhard-float/-O2
1733 --target_board=arm-sim/-mhard-float/-O3
1734 --target_board=arm-sim/-mhard-float
1735 --target_board=arm-sim/-msoft-float/-O1
1736 --target_board=arm-sim/-msoft-float/-O2
1737 --target_board=arm-sim/-msoft-float/-O3
1738 --target_board=arm-sim/-msoft-float
1741 They can be combined as many times as you wish, in arbitrary ways. This
1745 @dots{}"--target_board=unix/-Wextra@{-O3,-fno-strength-reduce@}@{-fomit-frame-pointer,@}"
1748 will generate four combinations, all involving @samp{-Wextra}.
1750 The disadvantage to this method is that the testsuites are run in serial,
1751 which is a waste on multiprocessor systems. For users with GNU Make and
1752 a shell which performs brace expansion, you can run the testsuites in
1753 parallel by having the shell perform the combinations and @command{make}
1754 do the parallel runs. Instead of using @samp{--target_board}, use a
1755 special makefile target:
1758 make -j@var{N} check-@var{testsuite}//@var{test-target}/@var{option1}/@var{option2}/@dots{}
1764 make -j3 check-gcc//sh-hms-sim/@{-m1,-m2,-m3,-m3e,-m4@}/@{,-nofpu@}
1767 will run three concurrent ``make-gcc'' testsuites, eventually testing all
1768 ten combinations as described above. Note that this is currently only
1769 supported in the @file{gcc} subdirectory. (To see how this works, try
1770 typing @command{echo} before the example given here.)
1773 @section Additional testing for Java Class Libraries
1775 The Java runtime tests can be executed via @samp{make check}
1776 in the @file{@var{target}/libjava/testsuite} directory in
1779 The @uref{http://sources.redhat.com/mauve/,,Mauve Project} provides
1780 a suite of tests for the Java Class Libraries. This suite can be run
1781 as part of libgcj testing by placing the Mauve tree within the libjava
1782 testsuite at @file{libjava/testsuite/libjava.mauve/mauve}, or by
1783 specifying the location of that tree when invoking @samp{make}, as in
1784 @samp{make MAUVEDIR=~/mauve check}.
1786 @uref{http://www-124.ibm.com/developerworks/oss/cvs/jikes/~checkout~/jacks/jacks.html,,Jacks}
1787 is a free testsuite that tests Java compiler front ends. This suite
1788 can be run as part of libgcj testing by placing the Jacks tree within
1789 the libjava testsuite at @file{libjava/testsuite/libjava.jacks/jacks}.
1791 @section How to interpret test results
1793 The result of running the testsuite are various @file{*.sum} and @file{*.log}
1794 files in the testsuite subdirectories. The @file{*.log} files contain a
1795 detailed log of the compiler invocations and the corresponding
1796 results, the @file{*.sum} files summarize the results. These summaries
1797 contain status codes for all tests:
1801 PASS: the test passed as expected
1803 XPASS: the test unexpectedly passed
1805 FAIL: the test unexpectedly failed
1807 XFAIL: the test failed as expected
1809 UNSUPPORTED: the test is not supported on this platform
1811 ERROR: the testsuite detected an error
1813 WARNING: the testsuite detected a possible problem
1816 It is normal for some tests to report unexpected failures. At the
1817 current time the testing harness does not allow fine grained control
1818 over whether or not a test is expected to fail. This problem should
1819 be fixed in future releases.
1822 @section Submitting test results
1824 If you want to report the results to the GCC project, use the
1825 @file{contrib/test_summary} shell script. Start it in the @var{objdir} with
1828 @var{srcdir}/contrib/test_summary -p your_commentary.txt \
1829 -m gcc-testresults@@gcc.gnu.org |sh
1832 This script uses the @command{Mail} program to send the results, so
1833 make sure it is in your @env{PATH}. The file @file{your_commentary.txt} is
1834 prepended to the testsuite summary and should contain any special
1835 remarks you have on your results or your build environment. Please
1836 do not edit the testsuite result block or the subject line, as these
1837 messages may be automatically processed.
1844 @uref{./index.html,,Return to the GCC Installation page}
1848 @c ***Final install***********************************************************
1850 @comment node-name, next, previous, up
1851 @node Final install, , Testing, Installing GCC
1853 @ifset finalinstallhtml
1855 @chapter Installing GCC: Final installation
1858 Now that GCC has been built (and optionally tested), you can install it with
1860 cd @var{objdir}; make install
1863 We strongly recommend to install into a target directory where there is
1864 no previous version of GCC present.
1866 That step completes the installation of GCC; user level binaries can
1867 be found in @file{@var{prefix}/bin} where @var{prefix} is the value
1868 you specified with the @option{--prefix} to configure (or
1869 @file{/usr/local} by default). (If you specified @option{--bindir},
1870 that directory will be used instead; otherwise, if you specified
1871 @option{--exec-prefix}, @file{@var{exec-prefix}/bin} will be used.)
1872 Headers for the C++ and Java libraries are installed in
1873 @file{@var{prefix}/include}; libraries in @file{@var{libdir}}
1874 (normally @file{@var{prefix}/lib}); internal parts of the compiler in
1875 @file{@var{libdir}/gcc} and @file{@var{libexecdir}/gcc}; documentation
1876 in info format in @file{@var{infodir}} (normally
1877 @file{@var{prefix}/info}).
1879 When installing cross-compilers, GCC's executables
1880 are not only installed into @file{@var{bindir}}, that
1881 is, @file{@var{exec-prefix}/bin}, but additionally into
1882 @file{@var{exec-prefix}/@var{target-alias}/bin}, if that directory
1883 exists. Typically, such @dfn{tooldirs} hold target-specific
1884 binutils, including assembler and linker.
1886 Installation into a temporary staging area or into a @command{chroot}
1887 jail can be achieved with the command
1890 make DESTDIR=@var{path-to-rootdir} install
1893 @noindent where @var{path-to-rootdir} is the absolute path of
1894 a directory relative to which all installation paths will be
1895 interpreted. Note that the directory specified by @code{DESTDIR}
1896 need not exist yet; it will be created if necessary.
1898 There is a subtle point with tooldirs and @code{DESTDIR}:
1899 If you relocate a cross-compiler installation with
1900 e.g.@: @samp{DESTDIR=@var{rootdir}}, then the directory
1901 @file{@var{rootdir}/@var{exec-prefix}/@var{target-alias}/bin} will
1902 be filled with duplicated GCC executables only if it already exists,
1903 it will not be created otherwise. This is regarded as a feature,
1904 not as a bug, because it gives slightly more control to the packagers
1905 using the @code{DESTDIR} feature.
1907 If you built a released version of GCC using @samp{make bootstrap} then please
1908 quickly review the build status page for your release, available from
1909 @uref{http://gcc.gnu.org/buildstat.html}.
1910 If your system is not listed for the version of GCC that you built,
1912 @email{gcc@@gcc.gnu.org} indicating
1913 that you successfully built and installed GCC.
1914 Include the following information:
1918 Output from running @file{@var{srcdir}/config.guess}. Do not send
1919 that file itself, just the one-line output from running it.
1922 The output of @samp{gcc -v} for your newly installed @command{gcc}.
1923 This tells us which version of GCC you built and the options you passed to
1927 Whether you enabled all languages or a subset of them. If you used a
1928 full distribution then this information is part of the configure
1929 options in the output of @samp{gcc -v}, but if you downloaded the
1930 ``core'' compiler plus additional front ends then it isn't apparent
1931 which ones you built unless you tell us about it.
1934 If the build was for GNU/Linux, also include:
1937 The distribution name and version (e.g., Red Hat 7.1 or Debian 2.2.3);
1938 this information should be available from @file{/etc/issue}.
1941 The version of the Linux kernel, available from @samp{uname --version}
1945 The version of glibc you used; for RPM-based systems like Red Hat,
1946 Mandrake, and SuSE type @samp{rpm -q glibc} to get the glibc version,
1947 and on systems like Debian and Progeny use @samp{dpkg -l libc6}.
1949 For other systems, you can include similar information if you think it is
1953 Any other information that you think would be useful to people building
1954 GCC on the same configuration. The new entry in the build status list
1955 will include a link to the archived copy of your message.
1958 We'd also like to know if the
1960 @ref{Specific, host/target specific installation notes}
1963 @uref{specific.html,,host/target specific installation notes}
1965 didn't include your host/target information or if that information is
1966 incomplete or out of date. Send a note to
1967 @email{gcc@@gcc.gnu.org} detailing how the information should be changed.
1969 If you find a bug, please report it following the
1970 @uref{../bugs.html,,bug reporting guidelines}.
1972 If you want to print the GCC manuals, do @samp{cd @var{objdir}; make
1973 dvi}. You will need to have @command{texi2dvi} (version at least 4.2)
1974 and @TeX{} installed. This creates a number of @file{.dvi} files in
1975 subdirectories of @file{@var{objdir}}; these may be converted for
1976 printing with programs such as @command{dvips}. You can also
1977 @uref{http://www.gnu.org/order/order.html,,buy printed manuals from the
1978 Free Software Foundation}, though such manuals may not be for the most
1979 recent version of GCC@.
1986 @uref{./index.html,,Return to the GCC Installation page}
1990 @c ***Binaries****************************************************************
1992 @comment node-name, next, previous, up
1993 @node Binaries, Specific, Installing GCC, Top
1997 @chapter Installing GCC: Binaries
2000 @cindex Installing GCC: Binaries
2002 We are often asked about pre-compiled versions of GCC@. While we cannot
2003 provide these for all platforms, below you'll find links to binaries for
2004 various platforms where creating them by yourself is not easy due to various
2007 Please note that we did not create these binaries, nor do we
2008 support them. If you have any problems installing them, please
2009 contact their makers.
2016 @uref{http://www.bullfreeware.com,,Bull's Freeware and Shareware Archive for AIX};
2019 @uref{http://aixpdslib.seas.ucla.edu,,UCLA Software Library for AIX}.
2023 DOS---@uref{http://www.delorie.com/djgpp/,,DJGPP}.
2026 Renesas H8/300[HS]---@uref{http://h8300-hms.sourceforge.net/,,GNU
2027 Development Tools for the Renesas H8/300[HS] Series}.
2033 @uref{http://hpux.cae.wisc.edu/,,HP-UX Porting Center};
2036 @uref{ftp://sunsite.informatik.rwth-aachen.de/pub/packages/gcc_hpux/,,Binaries for HP-UX 11.00 at Aachen University of Technology}.
2040 Motorola 68HC11/68HC12---@uref{http://www.gnu-m68hc11.org,,GNU
2041 Development Tools for the Motorola 68HC11/68HC12}.
2044 @uref{http://www.sco.com/skunkware/devtools/index.html#gcc,,SCO
2045 OpenServer/Unixware}.
2048 Sinix/Reliant Unix---@uref{ftp://ftp.fujitsu-siemens.com/pub/pd/gnu/gcc/,,Siemens}.
2051 Solaris 2 (SPARC, Intel)---@uref{http://www.sunfreeware.com/,,Sunfreeware}.
2054 SGI---@uref{http://freeware.sgi.com/,,SGI Freeware}.
2060 The @uref{http://sources.redhat.com/cygwin/,,Cygwin} project;
2062 The @uref{http://www.mingw.org/,,MinGW} project.
2066 @uref{ftp://ftp.thewrittenword.com/packages/by-name/,,The
2067 Written Word} offers binaries for
2070 Digital UNIX 4.0D and 5.1,
2072 HP-UX 10.20, 11.00, and 11.11, and
2073 Solaris/SPARC 2.5.1, 2.6, 2.7, 8, and 9,
2076 @uref{http://www.openpkg.org/,,OpenPKG} offers binaries for quite a
2077 number of platforms.
2080 In addition to those specific offerings, you can get a binary
2081 distribution CD-ROM from the
2082 @uref{http://www.fsf.org/order/order.html,,Free Software Foundation}.
2083 It contains binaries for a number of platforms, and
2084 includes not only GCC, but other stuff as well. The current CD does
2085 not contain the latest version of GCC, but it should allow
2086 bootstrapping the compiler. An updated version of that disk is in the
2094 @uref{./index.html,,Return to the GCC Installation page}
2098 @c ***Specific****************************************************************
2100 @comment node-name, next, previous, up
2101 @node Specific, Old, Binaries, Top
2105 @chapter Host/target specific installation notes for GCC
2108 @cindex Specific installation notes
2109 @cindex Target specific installation
2110 @cindex Host specific installation
2111 @cindex Target specific installation notes
2113 Please read this document carefully @emph{before} installing the
2114 GNU Compiler Collection on your machine.
2119 @uref{#alpha-x-x,,alpha*-*-*}
2121 @uref{#alpha-dec-osf,,alpha*-dec-osf*}
2123 @uref{#alphaev5-cray-unicosmk,,alphaev5-cray-unicosmk*}
2125 @uref{#arc-x-elf,,arc-*-elf}
2127 @uref{#arm-x-elf,,arm-*-elf}
2128 @uref{#arm-x-coff,,arm-*-coff}
2129 @uref{#arm-x-aout,,arm-*-aout}
2131 @uref{#xscale-x-x,,xscale-*-*}
2139 @uref{#dsp16xx,,dsp16xx}
2141 @uref{#x-x-freebsd,,*-*-freebsd*}
2143 @uref{#h8300-hms,,h8300-hms}
2145 @uref{#hppa-hp-hpux,,hppa*-hp-hpux*}
2147 @uref{#hppa-hp-hpux10,,hppa*-hp-hpux10}
2149 @uref{#hppa-hp-hpux11,,hppa*-hp-hpux11}
2151 @uref{#i370-x-x,,i370-*-*}
2153 @uref{#x-x-linux-gnu,,*-*-linux-gnu}
2155 @uref{#ix86-x-linuxaout,,i?86-*-linux*aout}
2157 @uref{#ix86-x-linux,,i?86-*-linux*}
2159 @uref{#ix86-x-sco32v5,,i?86-*-sco3.2v5*}
2161 @uref{#ix86-x-udk,,i?86-*-udk}
2163 @uref{#ia64-x-linux,,ia64-*-linux}
2165 @uref{#ia64-x-hpux,,ia64-*-hpux*}
2167 @uref{#x-ibm-aix,,*-ibm-aix*}
2169 @uref{#ip2k-x-elf,,ip2k-*-elf}
2171 @uref{#iq2000-x-elf,,iq2000-*-elf}
2173 @uref{#m32r-x-elf,,m32r-*-elf}
2175 @uref{#m6811-elf,,m6811-elf}
2177 @uref{#m6812-elf,,m6812-elf}
2179 @uref{#m68k-hp-hpux,,m68k-hp-hpux}
2181 @uref{#mips-x-x,,mips-*-*}
2183 @uref{#mips-sgi-irix5,,mips-sgi-irix5}
2185 @uref{#mips-sgi-irix6,,mips-sgi-irix6}
2187 @uref{#powerpc-x-x,,powerpc*-*-*, powerpc-*-sysv4}
2189 @uref{#powerpc-x-darwin,,powerpc-*-darwin*}
2191 @uref{#powerpc-x-elf,,powerpc-*-elf, powerpc-*-sysv4}
2193 @uref{#powerpc-x-linux-gnu,,powerpc*-*-linux-gnu*}
2195 @uref{#powerpc-x-netbsd,,powerpc-*-netbsd*}
2197 @uref{#powerpc-x-eabisim,,powerpc-*-eabisim}
2199 @uref{#powerpc-x-eabi,,powerpc-*-eabi}
2201 @uref{#powerpcle-x-elf,,powerpcle-*-elf, powerpcle-*-sysv4}
2203 @uref{#powerpcle-x-eabisim,,powerpcle-*-eabisim}
2205 @uref{#powerpcle-x-eabi,,powerpcle-*-eabi}
2207 @uref{#s390-x-linux,,s390-*-linux*}
2209 @uref{#s390x-x-linux,,s390x-*-linux*}
2211 @uref{#s390x-ibm-tpf,,s390x-ibm-tpf*}
2213 @uref{#x-x-solaris2,,*-*-solaris2*}
2215 @uref{#sparc-sun-solaris2,,sparc-sun-solaris2*}
2217 @uref{#sparc-sun-solaris27,,sparc-sun-solaris2.7}
2219 @uref{#sparc-x-linux,,sparc-*-linux*}
2221 @uref{#sparc64-x-solaris2,,sparc64-*-solaris2*}
2223 @uref{#sparcv9-x-solaris2,,sparcv9-*-solaris2*}
2225 @uref{#x-x-sysv,,*-*-sysv*}
2227 @uref{#vax-dec-ultrix,,vax-dec-ultrix}
2229 @uref{#x-x-vxworks,,*-*-vxworks*}
2231 @uref{#x86-64-x-x,,x86_64-*-*, amd64-*-*}
2233 @uref{#xtensa-x-elf,,xtensa-*-elf}
2235 @uref{#xtensa-x-linux,,xtensa-*-linux*}
2237 @uref{#windows,,Microsoft Windows}
2241 @uref{#older,,Older systems}
2246 @uref{#elf,,all ELF targets} (SVR4, Solaris 2, etc.)
2252 <!-- -------- host/target specific issues start here ---------------- -->
2255 @heading @anchor{alpha-x-x}alpha*-*-*
2257 This section contains general configuration information for all
2258 alpha-based platforms using ELF (in particular, ignore this section for
2259 DEC OSF/1, Digital UNIX and Tru64 UNIX)@. In addition to reading this
2260 section, please read all other sections that match your target.
2262 We require binutils 2.11.2 or newer.
2263 Previous binutils releases had a number of problems with DWARF 2
2264 debugging information, not the least of which is incorrect linking of
2270 @heading @anchor{alpha-dec-osf}alpha*-dec-osf*
2271 Systems using processors that implement the DEC Alpha architecture and
2272 are running the DEC/Compaq Unix (DEC OSF/1, Digital UNIX, or Compaq
2273 Tru64 UNIX) operating system, for example the DEC Alpha AXP systems.
2275 As of GCC 3.2, versions before @code{alpha*-dec-osf4} are no longer
2276 supported. (These are the versions which identify themselves as DEC
2279 In Digital Unix V4.0, virtual memory exhausted bootstrap failures
2280 may be fixed by configuring with @option{--with-gc=simple},
2281 reconfiguring Kernel Virtual Memory and Swap parameters
2282 per the @command{/usr/sbin/sys_check} Tuning Suggestions,
2283 or applying the patch in
2284 @uref{http://gcc.gnu.org/ml/gcc/2002-08/msg00822.html}.
2286 In Tru64 UNIX V5.1, Compaq introduced a new assembler that does not
2287 currently (2001-06-13) work with @command{mips-tfile}. As a workaround,
2288 we need to use the old assembler, invoked via the barely documented
2289 @option{-oldas} option. To bootstrap GCC, you either need to use the
2293 % CC=cc @var{srcdir}/configure [@var{options}] [@var{target}]
2296 or you can use a copy of GCC 2.95.3 or higher built on Tru64 UNIX V4.0:
2299 % CC=gcc -Wa,-oldas @var{srcdir}/configure [@var{options}] [@var{target}]
2302 As of GNU binutils 2.11.2, neither GNU @command{as} nor GNU @command{ld}
2303 are supported on Tru64 UNIX, so you must not configure GCC with
2304 @option{--with-gnu-as} or @option{--with-gnu-ld}.
2306 GCC writes a @samp{.verstamp} directive to the assembler output file
2307 unless it is built as a cross-compiler. It gets the version to use from
2308 the system header file @file{/usr/include/stamp.h}. If you install a
2309 new version of DEC Unix, you should rebuild GCC to pick up the new version
2312 Note that since the Alpha is a 64-bit architecture, cross-compilers from
2313 32-bit machines will not generate code as efficient as that generated
2314 when the compiler is running on a 64-bit machine because many
2315 optimizations that depend on being able to represent a word on the
2316 target in an integral value on the host cannot be performed. Building
2317 cross-compilers on the Alpha for 32-bit machines has only been tested in
2318 a few cases and may not work properly.
2320 @samp{make compare} may fail on old versions of DEC Unix unless you add
2321 @option{-save-temps} to @code{CFLAGS}. On these systems, the name of the
2322 assembler input file is stored in the object file, and that makes
2323 comparison fail if it differs between the @code{stage1} and
2324 @code{stage2} compilations. The option @option{-save-temps} forces a
2325 fixed name to be used for the assembler input file, instead of a
2326 randomly chosen name in @file{/tmp}. Do not add @option{-save-temps}
2327 unless the comparisons fail without that option. If you add
2328 @option{-save-temps}, you will have to manually delete the @samp{.i} and
2329 @samp{.s} files after each series of compilations.
2331 GCC now supports both the native (ECOFF) debugging format used by DBX
2332 and GDB and an encapsulated STABS format for use only with GDB@. See the
2333 discussion of the @option{--with-stabs} option of @file{configure} above
2334 for more information on these formats and how to select them.
2336 There is a bug in DEC's assembler that produces incorrect line numbers
2337 for ECOFF format when the @samp{.align} directive is used. To work
2338 around this problem, GCC will not emit such alignment directives
2339 while writing ECOFF format debugging information even if optimization is
2340 being performed. Unfortunately, this has the very undesirable
2341 side-effect that code addresses when @option{-O} is specified are
2342 different depending on whether or not @option{-g} is also specified.
2344 To avoid this behavior, specify @option{-gstabs+} and use GDB instead of
2345 DBX@. DEC is now aware of this problem with the assembler and hopes to
2346 provide a fix shortly.
2351 @heading @anchor{alphaev5-cray-unicosmk}alphaev5-cray-unicosmk*
2352 Cray T3E systems running Unicos/Mk.
2354 This port is incomplete and has many known bugs. We hope to improve the
2355 support for this target soon. Currently, only the C front end is supported,
2356 and it is not possible to build parallel applications. Cray modules are not
2357 supported; in particular, Craylibs are assumed to be in
2358 @file{/opt/ctl/craylibs/craylibs}.
2360 You absolutely @strong{must} use GNU make on this platform. Also, you
2361 need to tell GCC where to find the assembler and the linker. The
2362 simplest way to do so is by providing @option{--with-as} and
2363 @option{--with-ld} to @file{configure}, e.g.@:
2366 configure --with-as=/opt/ctl/bin/cam --with-ld=/opt/ctl/bin/cld \
2367 --enable-languages=c
2370 The comparison test during @samp{make bootstrap} fails on Unicos/Mk
2371 because the assembler inserts timestamps into object files. You should
2372 be able to work around this by doing @samp{make all} after getting this
2378 @heading @anchor{arc-x-elf}arc-*-elf
2379 Argonaut ARC processor.
2380 This configuration is intended for embedded systems.
2385 @heading @anchor{arm-x-elf}arm-*-elf
2386 @heading @anchor{xscale-x-x}xscale-*-*
2387 ARM-family processors. Subtargets that use the ELF object format
2388 require GNU binutils 2.13 or newer. Such subtargets include:
2389 @code{arm-*-freebsd}, @code{arm-*-netbsdelf}, @code{arm-*-*linux},
2390 @code{arm-*-rtems} and @code{arm-*-kaos}.
2395 @heading @anchor{arm-x-coff}arm-*-coff
2396 ARM-family processors. Note that there are two different varieties
2397 of PE format subtarget supported: @code{arm-wince-pe} and
2398 @code{arm-pe} as well as a standard COFF target @code{arm-*-coff}.
2403 @heading @anchor{arm-x-aout}arm-*-aout
2404 ARM-family processors. These targets support the AOUT file format:
2405 @code{arm-*-aout}, @code{arm-*-netbsd}.
2410 @heading @anchor{avr}avr
2412 ATMEL AVR-family micro controllers. These are used in embedded
2413 applications. There are no standard Unix configurations.
2415 @xref{AVR Options,, AVR Options, gcc, Using and Porting the GNU Compiler
2419 See ``AVR Options'' in the main manual
2421 for the list of supported MCU types.
2423 Use @samp{configure --target=avr --enable-languages="c"} to configure GCC@.
2425 Further installation notes and other useful information about AVR tools
2426 can also be obtained from:
2430 @uref{http://www.openavr.org,,http://www.openavr.org}
2432 @uref{http://home.overta.ru/users/denisc/,,http://home.overta.ru/users/denisc/}
2434 @uref{http://www.amelek.gda.pl/avr/,,http://www.amelek.gda.pl/avr/}
2437 We @emph{strongly} recommend using binutils 2.13 or newer.
2439 The following error:
2441 Error: register required
2444 indicates that you should upgrade to a newer version of the binutils.
2449 @heading @anchor{c4x}c4x
2451 Texas Instruments TMS320C3x and TMS320C4x Floating Point Digital Signal
2452 Processors. These are used in embedded applications. There are no
2453 standard Unix configurations.
2455 @xref{TMS320C3x/C4x Options,, TMS320C3x/C4x Options, gcc, Using and
2456 Porting the GNU Compiler Collection (GCC)},
2459 See ``TMS320C3x/C4x Options'' in the main manual
2461 for the list of supported MCU types.
2463 GCC can be configured as a cross compiler for both the C3x and C4x
2464 architectures on the same system. Use @samp{configure --target=c4x
2465 --enable-languages="c,c++"} to configure.
2468 Further installation notes and other useful information about C4x tools
2469 can also be obtained from:
2473 @uref{http://www.elec.canterbury.ac.nz/c4x/,,http://www.elec.canterbury.ac.nz/c4x/}
2479 @heading @anchor{cris}CRIS
2481 CRIS is the CPU architecture in Axis Communications ETRAX system-on-a-chip
2482 series. These are used in embedded applications.
2485 @xref{CRIS Options,, CRIS Options, gcc, Using and Porting the GNU Compiler
2489 See ``CRIS Options'' in the main manual
2491 for a list of CRIS-specific options.
2493 There are a few different CRIS targets:
2495 @item cris-axis-aout
2496 Old target. Includes a multilib for the @samp{elinux} a.out-based
2497 target. No multilibs for newer architecture variants.
2499 Mainly for monolithic embedded systems. Includes a multilib for the
2500 @samp{v10} core used in @samp{ETRAX 100 LX}.
2501 @item cris-axis-linux-gnu
2502 A GNU/Linux port for the CRIS architecture, currently targeting
2503 @samp{ETRAX 100 LX} by default.
2506 For @code{cris-axis-aout} and @code{cris-axis-elf} you need binutils 2.11
2507 or newer. For @code{cris-axis-linux-gnu} you need binutils 2.12 or newer.
2509 Pre-packaged tools can be obtained from
2510 @uref{ftp://ftp.axis.com/pub/axis/tools/cris/compiler-kit/}. More
2511 information about this platform is available at
2512 @uref{http://developer.axis.com/}.
2517 @heading @anchor{dos}DOS
2519 Please have a look at the @uref{binaries.html,,binaries page}.
2521 You cannot install GCC by itself on MSDOS; it will not compile under
2522 any MSDOS compiler except itself. You need to get the complete
2523 compilation package DJGPP, which includes binaries as well as sources,
2524 and includes all the necessary compilation tools and libraries.
2529 @heading @anchor{dsp16xx}dsp16xx
2530 A port to the AT&T DSP1610 family of processors.
2535 @heading @anchor{x-x-freebsd}*-*-freebsd*
2537 The version of binutils installed in @file{/usr/bin} probably works with
2538 this release of GCC. However, on FreeBSD 4, bootstrapping against the
2539 latest FSF binutils is known to improve overall testsuite results; and,
2540 on FreeBSD/alpha, using binutils 2.14 or later is required to build libjava.
2542 Support for FreeBSD 1 was discontinued in GCC 3.2.
2544 Support for FreeBSD 2 will be discontinued after GCC 3.4. The
2545 following was true for GCC 3.1 but the current status is unknown.
2546 For FreeBSD 2 or any mutant a.out versions of FreeBSD 3: All
2547 configuration support and files as shipped with GCC 2.95 are still in
2548 place. FreeBSD 2.2.7 has been known to bootstrap completely; however,
2549 it is unknown which version of binutils was used (it is assumed that it
2550 was the system copy in @file{/usr/bin}) and C++ EH failures were noted.
2552 For FreeBSD using the ELF file format: DWARF 2 debugging is now the
2553 default for all CPU architectures. It had been the default on
2554 FreeBSD/alpha since its inception. You may use @option{-gstabs} instead
2555 of @option{-g}, if you really want the old debugging format. There are
2556 no known issues with mixing object files and libraries with different
2557 debugging formats. Otherwise, this release of GCC should now match more
2558 of the configuration used in the stock FreeBSD configuration of GCC. In
2559 particular, @option{--enable-threads} is now configured by default.
2560 However, as a general user, do not attempt to replace the system
2561 compiler with this release. Known to bootstrap and check with good
2562 results on FreeBSD 4.9-STABLE and 5-CURRENT@. In the past, known to
2563 bootstrap and check with good results on FreeBSD 3.0, 3.4, 4.0, 4.2,
2564 4.3, 4.4, 4.5, 4.8-STABLE@.
2566 In principle, @option{--enable-threads} is now compatible with
2567 @option{--enable-libgcj} on FreeBSD@. However, it has only been built
2568 and tested on @samp{i386-*-freebsd[45]} and @samp{alpha-*-freebsd[45]}.
2570 library may be incorrectly built (symbols are missing at link time).
2571 There is a rare timing-based startup hang (probably involves an
2572 assumption about the thread library). Multi-threaded boehm-gc (required for
2573 libjava) exposes severe threaded signal-handling bugs on FreeBSD before
2574 4.5-RELEASE@. Other CPU architectures
2575 supported by FreeBSD will require additional configuration tuning in, at
2576 the very least, both boehm-gc and libffi.
2578 Shared @file{libgcc_s.so} is now built and installed by default.
2583 @heading @anchor{h8300-hms}h8300-hms
2584 Renesas H8/300 series of processors.
2586 Please have a look at the @uref{binaries.html,,binaries page}.
2588 The calling convention and structure layout has changed in release 2.6.
2589 All code must be recompiled. The calling convention now passes the
2590 first three arguments in function calls in registers. Structures are no
2591 longer a multiple of 2 bytes.
2596 @heading @anchor{hppa-hp-hpux}hppa*-hp-hpux*
2597 Support for HP-UX version 9 and older was discontinued in GCC 3.4.
2599 We @emph{highly} recommend using gas/binutils 2.8 or newer on all hppa
2600 platforms; you may encounter a variety of problems when using the HP
2603 Specifically, @option{-g} does not work on HP-UX (since that system
2604 uses a peculiar debugging format which GCC does not know about), unless you
2605 use GAS and GDB and configure GCC with the
2606 @uref{./configure.html#with-gnu-as,,@option{--with-gnu-as}} and
2607 @option{--with-as=@dots{}} options.
2609 If you wish to use the pa-risc 2.0 architecture support with a 32-bit
2610 runtime, you must use either the HP assembler, or gas/binutils 2.11
2613 There are two default scheduling models for instructions. These are
2614 PROCESSOR_7100LC and PROCESSOR_8000. They are selected from the pa-risc
2615 architecture specified for the target machine when configuring.
2616 PROCESSOR_8000 is the default. PROCESSOR_7100LC is selected when
2617 the target is a @samp{hppa1*} machine.
2619 The PROCESSOR_8000 model is not well suited to older processors. Thus,
2620 it is important to completely specify the machine architecture when
2621 configuring if you want a model other than PROCESSOR_8000. The macro
2622 TARGET_SCHED_DEFAULT can be defined in BOOT_CFLAGS if a different
2623 default scheduling model is desired.
2625 More specific information to @samp{hppa*-hp-hpux*} targets follows.
2630 @heading @anchor{hppa-hp-hpux10}hppa*-hp-hpux10
2632 For hpux10.20, we @emph{highly} recommend you pick up the latest sed patch
2633 @code{PHCO_19798} from HP@. HP has two sites which provide patches free of
2639 <a href="http://us.itrc.hp.com/service/home/home.do">US, Canada, Asia-Pacific, and
2643 @uref{http://us.itrc.hp.com/service/home/home.do,,} US, Canada, Asia-Pacific,
2647 @uref{http://europe.itrc.hp.com/service/home/home.do,,} Europe.
2650 The HP assembler on these systems has some problems. Most notably the
2651 assembler inserts timestamps into each object file it creates, causing
2652 the 3-stage comparison test to fail during a @samp{make bootstrap}.
2653 You should be able to continue by saying @samp{make all} after getting
2654 the failure from @samp{make bootstrap}.
2660 @heading @anchor{hppa-hp-hpux11}hppa*-hp-hpux11
2662 GCC 3.0 and up support HP-UX 11. GCC 2.95.x is not supported and cannot
2663 be used to compile GCC 3.0 and up.
2665 Refer to @uref{binaries.html,,binaries} for information about obtaining
2666 precompiled GCC binaries for HP-UX. Precompiled binaries must be obtained
2667 to build the Ada language as it can't be bootstrapped using C. Ada is
2668 only available for the 32-bit PA-RISC runtime. The libffi and libjava
2669 haven't been ported to HP-UX and don't build.
2671 It is possible to build GCC 3.3 starting with the bundled HP compiler,
2672 but the process requires several steps. GCC 3.3 can then be used to
2673 build later versions. The fastjar program contains ISO C code and
2674 can't be built with the HP bundled compiler. This problem can be
2675 avoided by not building the Java language. For example, use the
2676 @option{--enable-languages="c,c++,f77,objc"} option in your configure
2679 Starting with GCC 3.4 an ISO C compiler is required to bootstrap. The
2680 bundled compiler supports only traditional C; you will need either HP's
2681 unbundled compiler, or a binary distribution of GCC@.
2683 There are several possible approaches to building the distribution.
2684 Binutils can be built first using the HP tools. Then, the GCC
2685 distribution can be built. The second approach is to build GCC
2686 first using the HP tools, then build binutils, then rebuild GCC.
2687 There have been problems with various binary distributions, so it
2688 is best not to start from a binary distribution.
2690 On 64-bit capable systems, there are two distinct targets. Different
2691 installation prefixes must be used if both are to be installed on
2692 the same system. The @samp{hppa[1-2]*-hp-hpux11*} target generates code
2693 for the 32-bit PA-RISC runtime architecture and uses the HP linker.
2694 The @samp{hppa64-hp-hpux11*} target generates 64-bit code for the
2695 PA-RISC 2.0 architecture. The HP and GNU linkers are both supported
2698 The script config.guess now selects the target type based on the compiler
2699 detected during configuration. You must define @env{PATH} or @env{CC} so
2700 that configure finds an appropriate compiler for the initial bootstrap.
2701 When @env{CC} is used, the definition should contain the options that are
2702 needed whenever @env{CC} is used.
2704 Specifically, options that determine the runtime architecture must be
2705 in @env{CC} to correctly select the target for the build. It is also
2706 convenient to place many other compiler options in @env{CC}. For example,
2707 @env{CC="cc -Ac +DA2.0W -Wp,-H16376 -D_CLASSIC_TYPES -D_HPUX_SOURCE"}
2708 can be used to bootstrap the GCC 3.3 branch with the HP compiler in
2709 64-bit K&R/bundled mode. The @option{+DA2.0W} option will result in
2710 the automatic selection of the @samp{hppa64-hp-hpux11*} target. The
2711 macro definition table of cpp needs to be increased for a successful
2712 build with the HP compiler. _CLASSIC_TYPES and _HPUX_SOURCE need to
2713 be defined when building with the bundled compiler, or when using the
2714 @option{-Ac} option. These defines aren't necessary with @option{-Ae}.
2716 It is best to explicitly configure the @samp{hppa64-hp-hpux11*} target
2717 with the @option{--with-ld=@dots{}} option. This overrides the standard
2718 search for ld. The two linkers supported on this target require different
2719 commands. The default linker is determined during configuration. As a
2720 result, it's not possible to switch linkers in the middle of a GCC build.
2721 This has been been reported to sometimes occur in unified builds of
2724 With GCC 3.0 through 3.2, you must use binutils 2.11 or above. As of
2725 GCC 3.3, binutils 2.14 or later is required.
2727 Although the HP assembler can be used for an initial build, it shouldn't
2728 be used with any languages other than C and perhaps Fortran due to its
2729 many limitations. For example, it does not support weak symbols or alias
2730 definitions. As a result, explicit template instantiations are required
2731 when using C++. This makes it difficult if not impossible to build many
2732 C++ applications. You can't generate debugging information when using
2733 the HP assembler. Finally, @samp{make bootstrap} fails in the final
2734 comparison of object modules due to the time stamps that it inserts into
2735 the modules. The bootstrap can be continued from this point with
2738 A recent linker patch must be installed for the correct operation of
2739 GCC 3.3 and later. @code{PHSS_26559} and @code{PHSS_24304} are the
2740 oldest linker patches that are known to work. They are for HP-UX
2741 11.00 and 11.11, respectively. @code{PHSS_24303}, the companion to
2742 @code{PHSS_24304}, might be usable but it hasn't been tested. These
2743 patches have been superseded. Consult the HP patch database to obtain
2744 the currently recommended linker patch for your system.
2746 The patches are necessary for the support of weak symbols on the
2747 32-bit port, and for the running of initializers and finalizers. Weak
2748 symbols are implemented using SOM secondary definition symbols. Prior
2749 to HP-UX 11, there are bugs in the linker support for secondary symbols.
2750 The patches correct a problem of linker core dumps creating shared
2751 libraries containing secondary symbols, as well as various other
2752 linking issues involving secondary symbols.
2754 GCC 3.3 uses the ELF DT_INIT_ARRAY and DT_FINI_ARRAY capabilities to
2755 run initializers and finalizers on the 64-bit port. The 32-bit port
2756 uses the linker @option{+init} and @option{+fini} options for the same
2757 purpose. The patches correct various problems with the +init/+fini
2758 options, including program core dumps. Binutils 2.14 corrects a
2759 problem on the 64-bit port resulting from HP's non-standard use of
2760 the .init and .fini sections for array initializers and finalizers.
2762 There are a number of issues to consider in selecting which linker to
2763 use with the 64-bit port. The GNU 64-bit linker can only create dynamic
2764 binaries. The @option{-static} option causes linking with archive
2765 libraries but doesn't produce a truly static binary. Dynamic binaries
2766 still require final binding by the dynamic loader to resolve a set of
2767 dynamic-loader-defined symbols. The default behavior of the HP linker
2768 is the same as the GNU linker. However, it can generate true 64-bit
2769 static binaries using the @option{+compat} option.
2771 The HP 64-bit linker doesn't support linkonce semantics. As a
2772 result, C++ programs have many more sections than they should.
2774 The GNU 64-bit linker has some issues with shared library support
2775 and exceptions. As a result, we only support libgcc in archive
2776 format. For similar reasons, dwarf2 unwind and exception support
2777 are disabled. The GNU linker also has problems creating binaries
2778 with @option{-static}. It doesn't provide stubs for internal
2779 calls to global functions in shared libraries, so these calls
2780 can't be overloaded.
2782 Thread support is not implemented in GCC 3.0 through 3.2, so the
2783 @option{--enable-threads} configure option does not work. In 3.3
2784 and later, POSIX threads are supported. The optional DCE thread
2785 library is not supported.
2787 This port still is undergoing significant development.
2792 @heading @anchor{i370-x-x}i370-*-*
2793 This port is very preliminary and has many known bugs. We hope to
2794 have a higher-quality port for this machine soon.
2799 @heading @anchor{x-x-linux-gnu}*-*-linux-gnu
2801 Versions of libstdc++-v3 starting with 3.2.1 require bugfixes present
2802 in glibc 2.2.5 and later. More information is available in the
2803 libstdc++-v3 documentation.
2808 @heading @anchor{ix86-x-linuxaout}i?86-*-linux*aout
2809 Use this configuration to generate @file{a.out} binaries on Linux-based
2810 GNU systems. This configuration is being superseded.
2815 @heading @anchor{ix86-x-linux}i?86-*-linux*
2817 As of GCC 3.3, binutils 2.13.1 or later is required for this platform.
2818 See @uref{http://gcc.gnu.org/PR10877,,bug 10877} for more information.
2820 If you receive Signal 11 errors when building on GNU/Linux, then it is
2821 possible you have a hardware problem. Further information on this can be
2822 found on @uref{http://www.bitwizard.nl/sig11/,,www.bitwizard.nl}.
2824 Some recent versions of GNU/Linux, such as Fedora Core 1, support
2825 exec-shield-randomize. Turning this on interferes with precompiled
2826 headers. If you need to use precompiled headers,
2827 exec-shield-randomize can be turned off for the entire system by
2828 editing @file{/etc/sysctl.conf} and adding a line:
2830 kernel.exec-shield-randomize = 0
2832 You may then need to run @samp{sysctl -p}.
2834 This will be fixed in future releases of GCC.
2839 @heading @anchor{ix86-x-sco32v5}i?86-*-sco3.2v5*
2840 Use this for the SCO OpenServer Release 5 family of operating systems.
2842 Unlike earlier versions of GCC, the ability to generate COFF with this
2843 target is no longer provided.
2845 Earlier versions of GCC emitted DWARF 1 when generating ELF to allow
2846 the system debugger to be used. That support was too burdensome to
2847 maintain. GCC now emits only DWARF 2 for this target. This means you
2848 may use either the UDK debugger or GDB to debug programs built by this
2851 GCC is now only supported on releases 5.0.4 and later, and requires that
2852 you install Support Level Supplement OSS646B or later, and Support Level
2853 Supplement OSS631C or later. If you are using release 5.0.7 of
2854 OpenServer, you must have at least the first maintenance pack installed
2855 (this includes the relevant portions of OSS646). OSS646, also known as
2856 the "Execution Environment Update", provides updated link editors and
2857 assemblers, as well as updated standard C and math libraries. The C
2858 startup modules are also updated to support the System V gABI draft, and
2859 GCC relies on that behavior. OSS631 provides a collection of commonly
2860 used open source libraries, some of which GCC depends on (such as GNU
2861 gettext and zlib). SCO OpenServer Release 5.0.7 has all of this built
2862 in by default, but OSS631C and later also apply to that release. Please
2864 @uref{ftp://ftp.sco.com/pub/openserver5,,ftp://ftp.sco.com/pub/openserver5}
2865 for the latest versions of these (and other potentially useful)
2868 Although there is support for using the native assembler, it is
2869 recommended that you configure GCC to use the GNU assembler. You do
2870 this by using the flags
2871 @uref{./configure.html#with-gnu-as,,@option{--with-gnu-as}}. You should
2872 use a modern version of GNU binutils. Version 2.13.2.1 was used for all
2873 testing. In general, only the @option{--with-gnu-as} option is tested.
2874 A modern bintuils (as well as a plethora of other development related
2875 GNU utilities) can be found in Support Level Supplement OSS658A, the
2876 "GNU Development Tools" package. See the SCO web and ftp sites for details.
2877 That package also contains the currently "officially supported" version of
2878 GCC, version 2.95.3. It is useful for bootstrapping this version.
2883 @heading @anchor{ix86-x-udk}i?86-*-udk
2885 This target emulates the SCO Universal Development Kit and requires that
2886 package be installed. (If it is installed, you will have a
2887 @file{/udk/usr/ccs/bin/cc} file present.) It's very much like the
2888 @samp{i?86-*-unixware7*} target
2889 but is meant to be used when hosting on a system where UDK isn't the
2890 default compiler such as OpenServer 5 or Unixware 2. This target will
2891 generate binaries that will run on OpenServer, Unixware 2, or Unixware 7,
2892 with the same warnings and caveats as the SCO UDK@.
2894 This target is a little tricky to build because we have to distinguish
2895 it from the native tools (so it gets headers, startups, and libraries
2896 from the right place) while making the tools not think we're actually
2897 building a cross compiler. The easiest way to do this is with a configure
2901 CC=/udk/usr/ccs/bin/cc @var{/your/path/to}/gcc/configure \
2902 --host=i686-pc-udk --target=i686-pc-udk --program-prefix=udk-
2905 @emph{You should substitute @samp{i686} in the above command with the appropriate
2906 processor for your host.}
2908 After the usual @samp{make bootstrap} and
2909 @samp{make install}, you can then access the UDK-targeted GCC
2910 tools by adding @command{udk-} before the commonly known name. For
2911 example, to invoke the C compiler, you would use @command{udk-gcc}.
2912 They will coexist peacefully with any native-target GCC tools you may
2919 @heading @anchor{ia64-x-linux}ia64-*-linux
2920 IA-64 processor (also known as IPF, or Itanium Processor Family)
2923 If you are using the installed system libunwind library with
2924 @option{--with-system-libunwind}, then you must use libunwind 0.98 or
2927 None of the following versions of GCC has an ABI that is compatible
2928 with any of the other versions in this list, with the exception that
2929 Red Hat 2.96 and Trillian 000171 are compatible with each other:
2930 3.1, 3.0.2, 3.0.1, 3.0, Red Hat 2.96, and Trillian 000717.
2931 This primarily affects C++ programs and programs that create shared libraries.
2932 GCC 3.1 or later is recommended for compiling linux, the kernel.
2933 As of version 3.1 GCC is believed to be fully ABI compliant, and hence no
2934 more major ABI changes are expected.
2939 @heading @anchor{ia64-x-hpux}ia64-*-hpux*
2940 Building GCC on this target requires the GNU Assembler. The bundled HP
2941 assembler will not work. To prevent GCC from using the wrong assembler,
2942 the option @option{--with-gnu-as} may be necessary.
2944 The GCC libunwind library has not been ported to HPUX. This means that for
2945 GCC versions 3.2.3 and earlier, @option{--enable-libunwind-exceptions}
2946 is required to build GCC. For GCC 3.3 and later, this is the default.
2947 For gcc 3.4.3 and later, @option{--enable-libunwind-exceptions} is
2948 removed and the system libunwind library will always be used.
2952 <!-- rs6000-ibm-aix*, powerpc-ibm-aix* -->
2954 @heading @anchor{x-ibm-aix}*-ibm-aix*
2955 Support for AIX version 3 and older was discontinued in GCC 3.4.
2957 AIX Make frequently has problems with GCC makefiles. GNU Make 3.79.1 or
2958 newer is recommended to build on this platform.
2960 To speed up the configuration phases of bootstrapping and installing GCC,
2961 one may use GNU Bash instead of AIX @command{/bin/sh}, e.g.,
2964 % CONFIG_SHELL=/opt/freeware/bin/bash
2965 % export CONFIG_SHELL
2968 and then proceed as described in @uref{build.html,,the build instructions},
2969 where we strongly recommend using GNU make and specifying an absolute path
2970 to invoke @var{srcdir}/configure.
2972 Errors involving @code{alloca} when building GCC generally are due
2973 to an incorrect definition of @code{CC} in the Makefile or mixing files
2974 compiled with the native C compiler and GCC@. During the stage1 phase of
2975 the build, the native AIX compiler @strong{must} be invoked as @command{cc}
2976 (not @command{xlc}). Once @command{configure} has been informed of
2977 @command{xlc}, one needs to use @samp{make distclean} to remove the
2978 configure cache files and ensure that @env{CC} environment variable
2979 does not provide a definition that will confuse @command{configure}.
2980 If this error occurs during stage2 or later, then the problem most likely
2981 is the version of Make (see above).
2983 The native @command{as} and @command{ld} are recommended for bootstrapping
2984 on AIX 4 and required for bootstrapping on AIX 5L. The GNU Assembler
2985 reports that it supports WEAK symbols on AIX 4, which causes GCC to try to
2986 utilize weak symbol functionality although it is not supported. The GNU
2987 Assembler and Linker do not support AIX 5L sufficiently to bootstrap GCC.
2988 The native AIX tools do interoperate with GCC@.
2990 Building @file{libstdc++.a} requires a fix for an AIX Assembler bug
2991 APAR IY26685 (AIX 4.3) or APAR IY25528 (AIX 5.1). It also requires a
2992 fix for another AIX Assembler bug and a co-dependent AIX Archiver fix
2993 referenced as APAR IY53606 (AIX 5.2) or a APAR IY54774 (AIX 5.1)
2995 @samp{libstdc++} in GCC 3.4 increments the major version number of the
2996 shared object and GCC installation places the @file{libstdc++.a}
2997 shared library in a common location which will overwrite the and GCC
2998 3.3 version of the shared library. Applications either need to be
2999 re-linked against the new shared library or the GCC 3.1 and GCC 3.3
3000 versions of the @samp{libstdc++} shared object needs to be available
3001 to the AIX runtime loader. The GCC 3.1 @samp{libstdc++.so.4}, if
3002 present, and GCC 3.3 @samp{libstdc++.so.5} shared objects can be
3003 installed for runtime dynamic loading using the following steps to set
3004 the @samp{F_LOADONLY} flag in the shared object for @emph{each}
3005 multilib @file{libstdc++.a} installed:
3007 Extract the shared objects from the currently installed
3008 @file{libstdc++.a} archive:
3010 % ar -x libstdc++.a libstdc++.so.4 libstdc++.so.5
3013 Enable the @samp{F_LOADONLY} flag so that the shared object will be
3014 available for runtime dynamic loading, but not linking:
3016 % strip -e libstdc++.so.4 libstdc++.so.5
3019 Archive the runtime-only shared object in the GCC 3.4
3020 @file{libstdc++.a} archive:
3022 % ar -q libstdc++.a libstdc++.so.4 libstdc++.so.5
3025 Linking executables and shared libraries may produce warnings of
3026 duplicate symbols. The assembly files generated by GCC for AIX always
3027 have included multiple symbol definitions for certain global variable
3028 and function declarations in the original program. The warnings should
3029 not prevent the linker from producing a correct library or runnable
3032 AIX 4.3 utilizes a ``large format'' archive to support both 32-bit and
3033 64-bit object modules. The routines provided in AIX 4.3.0 and AIX 4.3.1
3034 to parse archive libraries did not handle the new format correctly.
3035 These routines are used by GCC and result in error messages during
3036 linking such as ``not a COFF file''. The version of the routines shipped
3037 with AIX 4.3.1 should work for a 32-bit environment. The @option{-g}
3038 option of the archive command may be used to create archives of 32-bit
3039 objects using the original ``small format''. A correct version of the
3040 routines is shipped with AIX 4.3.2 and above.
3042 Some versions of the AIX binder (linker) can fail with a relocation
3043 overflow severe error when the @option{-bbigtoc} option is used to link
3044 GCC-produced object files into an executable that overflows the TOC@. A fix
3045 for APAR IX75823 (OVERFLOW DURING LINK WHEN USING GCC AND -BBIGTOC) is
3046 available from IBM Customer Support and from its
3047 @uref{http://techsupport.services.ibm.com/,,techsupport.services.ibm.com}
3048 website as PTF U455193.
3050 The AIX 4.3.2.1 linker (bos.rte.bind_cmds Level 4.3.2.1) will dump core
3051 with a segmentation fault when invoked by any version of GCC@. A fix for
3052 APAR IX87327 is available from IBM Customer Support and from its
3053 @uref{http://techsupport.services.ibm.com/,,techsupport.services.ibm.com}
3054 website as PTF U461879. This fix is incorporated in AIX 4.3.3 and above.
3056 The initial assembler shipped with AIX 4.3.0 generates incorrect object
3057 files. A fix for APAR IX74254 (64BIT DISASSEMBLED OUTPUT FROM COMPILER FAILS
3058 TO ASSEMBLE/BIND) is available from IBM Customer Support and from its
3059 @uref{http://techsupport.services.ibm.com/,,techsupport.services.ibm.com}
3060 website as PTF U453956. This fix is incorporated in AIX 4.3.1 and above.
3062 AIX provides National Language Support (NLS)@. Compilers and assemblers
3063 use NLS to support locale-specific representations of various data
3064 formats including floating-point numbers (e.g., @samp{.} vs @samp{,} for
3065 separating decimal fractions). There have been problems reported where
3066 GCC does not produce the same floating-point formats that the assembler
3067 expects. If one encounters this problem, set the @env{LANG}
3068 environment variable to @samp{C} or @samp{En_US}.
3070 By default, GCC for AIX 4.1 and above produces code that can be used on
3071 both Power or PowerPC processors.
3073 A default can be specified with the @option{-mcpu=@var{cpu_type}}
3074 switch and using the configure option @option{--with-cpu-@var{cpu_type}}.
3079 @heading @anchor{ip2k-x-elf}ip2k-*-elf
3080 Ubicom IP2022 micro controller.
3081 This configuration is intended for embedded systems.
3082 There are no standard Unix configurations.
3084 Use @samp{configure --target=ip2k-elf --enable-languages=c} to configure GCC@.
3089 @heading @anchor{iq2000-x-elf}iq2000-*-elf
3090 Vitesse IQ2000 processors. These are used in embedded
3091 applications. There are no standard Unix configurations.
3096 @heading @anchor{m32r-x-elf}m32r-*-elf
3097 Renesas M32R processor.
3098 This configuration is intended for embedded systems.
3103 @heading @anchor{m6811-elf}m6811-elf
3104 Motorola 68HC11 family micro controllers. These are used in embedded
3105 applications. There are no standard Unix configurations.
3110 @heading @anchor{m6812-elf}m6812-elf
3111 Motorola 68HC12 family micro controllers. These are used in embedded
3112 applications. There are no standard Unix configurations.
3117 @heading @anchor{m68k-hp-hpux}m68k-hp-hpux
3118 HP 9000 series 300 or 400 running HP-UX@. HP-UX version 8.0 has a bug in
3119 the assembler that prevents compilation of GCC@. This
3120 bug manifests itself during the first stage of compilation, while
3121 building @file{libgcc2.a}:
3125 cc1: warning: `-g' option not supported on this version of GCC
3126 cc1: warning: `-g1' option not supported on this version of GCC
3127 ./xgcc: Internal compiler error: program as got fatal signal 11
3130 A patched version of the assembler is available as the file
3131 @uref{ftp://altdorf.ai.mit.edu/archive/cph/hpux-8.0-assembler}. If you
3132 have HP software support, the patch can also be obtained directly from
3133 HP, as described in the following note:
3136 This is the patched assembler, to patch SR#1653-010439, where the
3137 assembler aborts on floating point constants.
3139 The bug is not really in the assembler, but in the shared library
3140 version of the function ``cvtnum(3c)''. The bug on ``cvtnum(3c)'' is
3141 SR#4701-078451. Anyway, the attached assembler uses the archive
3142 library version of ``cvtnum(3c)'' and thus does not exhibit the bug.
3145 This patch is also known as PHCO_4484.
3147 In addition gdb does not understand that native HP-UX format, so
3148 you must use gas if you wish to use gdb.
3150 On HP-UX version 8.05, but not on 8.07 or more recent versions, the
3151 @command{fixproto} shell script triggers a bug in the system shell. If you
3152 encounter this problem, upgrade your operating system or use BASH (the
3153 GNU shell) to run @command{fixproto}. This bug will cause the fixproto
3154 program to report an error of the form:
3157 ./fixproto: sh internal 1K buffer overflow
3160 To fix this, you can also change the first line of the fixproto script
3170 @heading @anchor{mips-x-x}mips-*-*
3171 If on a MIPS system you get an error message saying ``does not have gp
3172 sections for all it's [sic] sectons [sic]'', don't worry about it. This
3173 happens whenever you use GAS with the MIPS linker, but there is not
3174 really anything wrong, and it is okay to use the output file. You can
3175 stop such warnings by installing the GNU linker.
3177 It would be nice to extend GAS to produce the gp tables, but they are
3178 optional, and there should not be a warning about their absence.
3180 The libstdc++ atomic locking routines for MIPS targets requires MIPS II
3181 and later. A patch went in just after the GCC 3.3 release to
3182 make @samp{mips*-*-*} use the generic implementation instead. You can also
3183 configure for @samp{mipsel-elf} as a workaround. The
3184 @samp{mips*-*-linux*} target continues to use the MIPS II routines. More
3185 work on this is expected in future releases.
3187 Cross-compilers for the Mips as target using the Mips assembler
3188 currently do not work, because the auxiliary programs
3189 @file{mips-tdump.c} and @file{mips-tfile.c} can't be compiled on
3190 anything but a Mips. It does work to cross compile for a Mips
3191 if you use the GNU assembler and linker.
3196 @heading @anchor{mips-sgi-irix5}mips-sgi-irix5
3198 In order to compile GCC on an SGI running IRIX 5, the @samp{compiler_dev.hdr}
3199 subsystem must be installed from the IDO CD-ROM supplied by SGI@.
3200 It is also available for download from
3201 @uref{ftp://ftp.sgi.com/sgi/IRIX5.3/iris-development-option-5.3.tardist}.
3203 If you use the MIPS C compiler to bootstrap, it may be necessary
3204 to increase its table size for switch statements with the
3205 @option{-Wf,-XNg1500} option. If you use the @option{-O2}
3206 optimization option, you also need to use @option{-Olimit 3000}.
3208 To enable debugging under IRIX 5, you must use GNU binutils 2.15 or
3209 later, and use the @option{--with-gnu-as} and @option{--with-gnu-ld}
3210 @command{configure} options when configuring GCC@. You need to use GNU
3211 @command{ar} and @command{nm}, also distributed with GNU binutils.
3213 Some users have reported that @command{/bin/sh} will hang during bootstrap.
3214 This problem can be avoided by running the commands:
3217 % CONFIG_SHELL=/bin/ksh
3218 % export CONFIG_SHELL
3221 before starting the build.
3226 @heading @anchor{mips-sgi-irix6}mips-sgi-irix6
3228 If you are using SGI's MIPSpro @command{cc} as your bootstrap compiler, you must
3229 ensure that the N32 ABI is in use. To test this, compile a simple C
3230 file with @command{cc} and then run @command{file} on the
3231 resulting object file. The output should look like:
3234 test.o: ELF N32 MSB @dots{}
3240 test.o: ELF 32-bit MSB @dots{}
3246 test.o: ELF 64-bit MSB @dots{}
3249 then your version of @command{cc} uses the O32 or N64 ABI by default. You
3250 should set the environment variable @env{CC} to @samp{cc -n32}
3251 before configuring GCC@. SGI's MIPSpro 7.2 assembler may misassemble
3252 parts of the compiler, causing bootstrap failures. MIPSpro 7.3 is
3253 known to work. MIPSpro C 7.4 may cause bootstrap failures, too, due
3254 to a bug when inlining @code{memcmp}. Either add @code{-U__INLINE_INTRINSICS}
3255 to the @env{CC} environment variable as a workaround or upgrade to
3258 If you want the resulting @command{gcc} to run on old 32-bit systems
3259 with the MIPS R4400 CPU, you need to ensure that only code for the @samp{mips3}
3260 instruction set architecture (ISA) is generated. While GCC 3.x does
3261 this correctly, both GCC 2.95 and SGI's MIPSpro @command{cc} may change
3262 the ISA depending on the machine where GCC is built. Using one of them
3263 as the bootstrap compiler may result in @samp{mips4} code, which won't run at
3264 all on @samp{mips3}-only systems. For the test program above, you should see:
3267 test.o: ELF N32 MSB mips-3 @dots{}
3273 test.o: ELF N32 MSB mips-4 @dots{}
3276 instead, you should set the environment variable @env{CC} to @samp{cc
3277 -n32 -mips3} or @samp{gcc -mips3} respectively before configuring GCC@.
3279 GCC on IRIX 6 is usually built to support the N32, O32 and N64 ABIs. If
3280 you build GCC on a system that doesn't have the N64 libraries installed
3281 or cannot run 64-bit binaries,
3282 you need to configure with @option{--disable-multilib} so GCC doesn't
3283 try to use them. This will disable building the O32 libraries, too.
3284 Look for @file{/usr/lib64/libc.so.1} to see if you
3285 have the 64-bit libraries installed.
3287 To enable debugging for the O32 ABI, you must use GNU @command{as} from
3288 GNU binutils 2.15 or later. You may also use GNU @command{ld}, but
3289 this is not required and currently causes some problems with Ada.
3291 The @option{--enable-threads} option doesn't currently work, a patch is
3292 in preparation for a future release. The @option{--enable-libgcj}
3293 option is disabled by default: IRIX 6 uses a very low default limit
3294 (20480) for the command line length. Although @command{libtool} contains a
3295 workaround for this problem, at least the N64 @samp{libgcj} is known not
3296 to build despite this, running into an internal error of the native
3297 @command{ld}. A sure fix is to increase this limit (@samp{ncargs}) to
3298 its maximum of 262144 bytes. If you have root access, you can use the
3299 @command{systune} command to do this.
3301 See @uref{http://freeware.sgi.com/} for more
3302 information about using GCC on IRIX platforms.
3307 @heading @anchor{powerpc-x-x}powerpc-*-*
3309 You can specify a default version for the @option{-mcpu=@var{cpu_type}}
3310 switch by using the configure option @option{--with-cpu-@var{cpu_type}}.
3315 @heading @anchor{powerpc-x-darwin}powerpc-*-darwin*
3316 PowerPC running Darwin (Mac OS X kernel).
3318 Pre-installed versions of Mac OS X may not include any developer tools,
3319 meaning that you will not be able to build GCC from source. Tool
3320 binaries are available at
3321 @uref{http://developer.apple.com/darwin/projects/compiler/} (free
3322 registration required).
3324 The default stack limit of 512K is too small, which may cause compiles
3325 to fail with 'Bus error'. Set the stack larger, for instance
3326 by doing @samp{limit stack 800}. It's a good idea to use the GNU
3327 preprocessor instead of Apple's @file{cpp-precomp} during the first stage of
3328 bootstrapping; this is automatic when doing @samp{make bootstrap}, but
3329 to do it from the toplevel objdir you will need to say @samp{make
3330 CC='cc -no-cpp-precomp' bootstrap}.
3332 The version of GCC shipped by Apple typically includes a number of
3333 extensions not available in a standard GCC release. These extensions
3334 are generally specific to Mac programming.
3339 @heading @anchor{powerpc-x-elf}powerpc-*-elf, powerpc-*-sysv4
3340 PowerPC system in big endian mode, running System V.4.
3345 @heading @anchor{powerpc-x-linux-gnu}powerpc*-*-linux-gnu*
3348 @uref{ftp://ftp.kernel.org/pub/linux/devel/binutils,,binutils 2.15}
3349 or newer for a working GCC@.
3354 @heading @anchor{powerpc-x-netbsd}powerpc-*-netbsd*
3355 PowerPC system in big endian mode running NetBSD@. To build the
3356 documentation you will need Texinfo version 4.2 (NetBSD 1.5.1 included
3357 Texinfo version 3.12).
3362 @heading @anchor{powerpc-x-eabisim}powerpc-*-eabisim
3363 Embedded PowerPC system in big endian mode for use in running under the
3369 @heading @anchor{powerpc-x-eabi}powerpc-*-eabi
3370 Embedded PowerPC system in big endian mode.
3375 @heading @anchor{powerpcle-x-elf}powerpcle-*-elf, powerpcle-*-sysv4
3376 PowerPC system in little endian mode, running System V.4.
3381 @heading @anchor{powerpcle-x-eabisim}powerpcle-*-eabisim
3382 Embedded PowerPC system in little endian mode for use in running under
3388 @heading @anchor{powerpcle-x-eabi}powerpcle-*-eabi
3389 Embedded PowerPC system in little endian mode.
3394 @heading @anchor{s390-x-linux}s390-*-linux*
3395 S/390 system running GNU/Linux for S/390@.
3400 @heading @anchor{s390x-x-linux}s390x-*-linux*
3401 zSeries system (64-bit) running GNU/Linux for zSeries@.
3406 @heading @anchor{s390x-ibm-tpf}s390x-ibm-tpf*
3407 zSeries system (64-bit) running TPF. This platform is
3408 supported as cross-compilation target only.
3413 @c Please use Solaris 2 to refer to all release of Solaris, starting
3414 @c with 2.0 until 2.6, 7, 8, etc. Solaris 1 was a marketing name for
3415 @c SunOS 4 releases which we don't use to avoid confusion. Solaris
3416 @c alone is too unspecific and must be avoided.
3417 @heading @anchor{x-x-solaris2}*-*-solaris2*
3419 Sun does not ship a C compiler with Solaris 2. To bootstrap and install
3420 GCC you first have to install a pre-built compiler, see the
3421 @uref{binaries.html,,binaries page} for details.
3423 The Solaris 2 @command{/bin/sh} will often fail to configure
3424 @file{libstdc++-v3}, @file{boehm-gc} or @file{libjava}. We therefore
3425 recommend to use the following sequence of commands to bootstrap and
3429 % CONFIG_SHELL=/bin/ksh
3430 % export CONFIG_SHELL
3433 and then proceed as described in @uref{build.html,,the build instructions}.
3434 In addition we strongly recommend specifying an absolute path to invoke
3435 @var{srcdir}/configure.
3437 Solaris 2 comes with a number of optional OS packages. Some of these
3438 are needed to use GCC fully, namely @code{SUNWarc},
3439 @code{SUNWbtool}, @code{SUNWesu}, @code{SUNWhea}, @code{SUNWlibm},
3440 @code{SUNWsprot}, and @code{SUNWtoo}. If you did not install all
3441 optional packages when installing Solaris 2, you will need to verify that
3442 the packages that GCC needs are installed.
3444 To check whether an optional package is installed, use
3445 the @command{pkginfo} command. To add an optional package, use the
3446 @command{pkgadd} command. For further details, see the Solaris 2
3449 Trying to use the linker and other tools in
3450 @file{/usr/ucb} to install GCC has been observed to cause trouble.
3451 For example, the linker may hang indefinitely. The fix is to remove
3452 @file{/usr/ucb} from your @env{PATH}.
3454 The build process works more smoothly with the legacy Sun tools so, if you
3455 have @file{/usr/xpg4/bin} in your @env{PATH}, we recommend that you place
3456 @file{/usr/bin} before @file{/usr/xpg4/bin} for the duration of the build.
3458 All releases of GNU binutils prior to 2.11.2 have known bugs on this
3459 platform. We recommend the use of GNU binutils 2.11.2 or later, or the
3460 vendor tools (Sun @command{as}, Sun @command{ld}). Note that your mileage
3461 may vary if you use a combination of the GNU tools and the Sun tools: while
3462 the combination GNU @command{as} + Sun @command{ld} should reasonably work,
3463 the reverse combination Sun @command{as} + GNU @command{ld} is known to
3464 cause memory corruption at runtime in some cases for C++ programs.
3466 The stock GNU binutils 2.15 release is broken on this platform because of a
3467 single bug. It has been fixed on the 2.15 branch in the CVS repository.
3468 You can obtain a working version by checking out the binutils-2_15-branch
3469 from the CVS repository or applying the patch
3470 @uref{http://sources.redhat.com/ml/binutils-cvs/2004-09/msg00036.html} to the
3473 Sun bug 4296832 turns up when compiling X11 headers with GCC 2.95 or
3474 newer: @command{g++} will complain that types are missing. These headers assume
3475 that omitting the type means @code{int}; this assumption worked for C89 but
3476 is wrong for C++, and is now wrong for C99 also.
3478 @command{g++} accepts such (invalid) constructs with the option
3479 @option{-fpermissive}; it
3480 will assume that any missing type is @code{int} (as defined by C89).
3482 There are patches for Solaris 2.6 (105633-56 or newer for SPARC,
3483 106248-42 or newer for Intel), Solaris 7 (108376-21 or newer for SPARC,
3484 108377-20 for Intel), and Solaris 8 (108652-24 or newer for SPARC,
3485 108653-22 for Intel) that fix this bug.
3487 Sun bug 4927647 sometimes causes random spurious testsuite failures
3488 related to missing diagnostic output. This bug doesn't affect GCC
3489 itself, rather it is a kernel bug triggered by the @command{expect}
3490 program which is used only by the GCC testsuite driver. When the bug
3491 causes the @command{expect} program to miss anticipated output, extra
3492 testsuite failures appear.
3494 There are patches for Solaris 8 (117350-12 or newer for SPARC,
3495 117351-12 or newer for Intel) and Solaris 9 (117171-11 or newer for
3496 SPARC, 117172-11 or newer for Intel) that address this problem.
3501 @heading @anchor{sparc-sun-solaris2}sparc-sun-solaris2*
3503 When GCC is configured to use binutils 2.11.2 or later the binaries
3504 produced are smaller than the ones produced using Sun's native tools;
3505 this difference is quite significant for binaries containing debugging
3508 Sun @command{as} 4.x is broken in that it cannot cope with long symbol names.
3509 A typical error message might look similar to the following:
3512 /usr/ccs/bin/as: "/var/tmp/ccMsw135.s", line 11041: error:
3513 can't compute value of an expression involving an external symbol.
3516 This is Sun bug 4237974. This is fixed with patch 108908-02 for Solaris
3517 2.6 and has been fixed in later (5.x) versions of the assembler,
3518 starting with Solaris 7.
3520 Starting with Solaris 7, the operating system is capable of executing
3521 64-bit SPARC V9 binaries. GCC 3.1 and later properly supports
3522 this; the @option{-m64} option enables 64-bit code generation.
3523 However, if all you want is code tuned for the UltraSPARC CPU, you
3524 should try the @option{-mtune=ultrasparc} option instead, which produces
3525 code that, unlike full 64-bit code, can still run on non-UltraSPARC
3528 When configuring on a Solaris 7 or later system that is running a kernel
3529 that supports only 32-bit binaries, one must configure with
3530 @option{--disable-multilib}, since we will not be able to build the
3531 64-bit target libraries.
3533 GCC 3.3 and GCC 3.4 trigger code generation bugs in earlier versions of
3534 the GNU compiler (especially GCC 3.0.x versions), which lead to the
3535 miscompilation of the stage1 compiler and the subsequent failure of the
3536 bootstrap process. A workaround is to use GCC 3.2.3 as an intermediary
3537 stage, i.e. to bootstrap that compiler with the base compiler and then
3538 use it to bootstrap the final compiler.
3540 GCC 3.4 triggers a code generation bug in versions 5.4 (Sun ONE Studio 7)
3541 and 5.5 (Sun ONE Studio 8) of the Sun compiler, which causes a bootstrap
3542 failure in form of a miscompilation of the stage1 compiler by the Sun
3543 compiler. This is Sun bug 4974440. This is fixed with patch 112760-07.
3545 GCC 3.4 changed the default debugging format from STABS to DWARF-2 for
3546 32-bit code on Solaris 7 and later. If you are using the Sun
3547 assembler, this change apparently runs afoul of Sun bug 4910101, for
3548 which (as of 2004-05-23) there is no fix. A symptom of the problem is
3549 that you cannot compile C++ programs like @command{groff} 1.19.1
3550 without getting messages similar to the following:
3553 ld: warning: relocation error: R_SPARC_UA32: @dots{}
3554 external symbolic relocation against non-allocatable section
3555 .debug_info cannot be processed at runtime: relocation ignored.
3558 To work around this problem, compile with @option{-gstabs+} instead of
3564 @heading @anchor{sparc-sun-solaris27}sparc-sun-solaris2.7
3566 Sun patch 107058-01 (1999-01-13) for Solaris 7/SPARC triggers a bug in
3567 the dynamic linker. This problem (Sun bug 4210064) affects GCC 2.8
3568 and later, including all EGCS releases. Sun formerly recommended
3569 107058-01 for all Solaris 7 users, but around 1999-09-01 it started to
3570 recommend it only for people who use Sun's compilers.
3572 Here are some workarounds to this problem:
3575 Do not install Sun patch 107058-01 until after Sun releases a
3576 complete patch for bug 4210064. This is the simplest course to take,
3577 unless you must also use Sun's C compiler. Unfortunately 107058-01
3578 is preinstalled on some new Solaris 7-based hosts, so you may have to
3582 Copy the original, unpatched Solaris 7
3583 @command{/usr/ccs/bin/as} into
3584 @command{/usr/local/libexec/gcc/sparc-sun-solaris2.7/3.4/as},
3585 adjusting the latter name to fit your local conventions and software
3589 Install Sun patch 106950-03 (1999-05-25) or later. Nobody with
3590 both 107058-01 and 106950-03 installed has reported the bug with GCC
3591 and Sun's dynamic linker. This last course of action is riskiest,
3592 for two reasons. First, you must install 106950 on all hosts that
3593 run code generated by GCC; it doesn't suffice to install it only on
3594 the hosts that run GCC itself. Second, Sun says that 106950-03 is
3595 only a partial fix for bug 4210064, but Sun doesn't know whether the
3596 partial fix is adequate for GCC@. Revision -08 or later should fix
3597 the bug. The current (as of 2004-05-23) revision is -24, and is included in
3598 the Solaris 7 Recommended Patch Cluster.
3601 GCC 3.3 triggers a bug in version 5.0 Alpha 03/27/98 of the Sun assembler,
3602 which causes a bootstrap failure when linking the 64-bit shared version of
3603 libgcc. A typical error message is:
3606 ld: fatal: relocation error: R_SPARC_32: file libgcc/sparcv9/_muldi3.o:
3607 symbol <unknown>: offset 0xffffffff7ec133e7 is non-aligned.
3610 This bug has been fixed in the final 5.0 version of the assembler.
3615 @heading @anchor{sparc-x-linux}sparc-*-linux*
3617 GCC versions 3.0 and higher require binutils 2.11.2 and glibc 2.2.4
3618 or newer on this platform. All earlier binutils and glibc
3619 releases mishandled unaligned relocations on @code{sparc-*-*} targets.
3625 @heading @anchor{sparc64-x-solaris2}sparc64-*-solaris2*
3627 The following compiler flags must be specified in the configure
3628 step in order to bootstrap this target with the Sun compiler:
3631 % CC="cc -xildoff -xarch=v9" @var{srcdir}/configure [@var{options}] [@var{target}]
3634 @option{-xildoff} turns off the incremental linker, and @option{-xarch=v9}
3635 specifies the SPARC-V9 architecture to the Sun linker and assembler.
3640 @heading @anchor{sparcv9-x-solaris2}sparcv9-*-solaris2*
3642 This is a synonym for sparc64-*-solaris2*.
3647 @heading @anchor{x-x-sysv}*-*-sysv*
3648 On System V release 3, you may get this error message
3652 ld fatal: failed to write symbol name @var{something}
3653 in strings table for file @var{whatever}
3656 This probably indicates that the disk is full or your ulimit won't allow
3657 the file to be as large as it needs to be.
3659 This problem can also result because the kernel parameter @code{MAXUMEM}
3660 is too small. If so, you must regenerate the kernel and make the value
3661 much larger. The default value is reported to be 1024; a value of 32768
3662 is said to work. Smaller values may also work.
3664 On System V, if you get an error like this,
3667 /usr/local/lib/bison.simple: In function `yyparse':
3668 /usr/local/lib/bison.simple:625: virtual memory exhausted
3672 that too indicates a problem with disk space, ulimit, or @code{MAXUMEM}.
3674 On a System V release 4 system, make sure @file{/usr/bin} precedes
3675 @file{/usr/ucb} in @code{PATH}. The @command{cc} command in
3676 @file{/usr/ucb} uses libraries which have bugs.
3681 @heading @anchor{vax-dec-ultrix}vax-dec-ultrix
3682 Don't try compiling with VAX C (@command{vcc}). It produces incorrect code
3683 in some cases (for example, when @code{alloca} is used).
3688 @heading @anchor{x-x-vxworks}*-*-vxworks*
3689 Support for VxWorks is in flux. At present GCC supports @emph{only} the
3690 very recent VxWorks 5.5 (aka Tornado 2.2) release, and only on PowerPC.
3691 We welcome patches for other architectures supported by VxWorks 5.5.
3692 Support for VxWorks AE would also be welcome; we believe this is merely
3693 a matter of writing an appropriate ``configlette'' (see below). We are
3694 not interested in supporting older, a.out or COFF-based, versions of
3697 VxWorks comes with an older version of GCC installed in
3698 @file{@var{$WIND_BASE}/host}; we recommend you do not overwrite it.
3699 Choose an installation @var{prefix} entirely outside @var{$WIND_BASE}.
3700 Before running @command{configure}, create the directories @file{@var{prefix}}
3701 and @file{@var{prefix}/bin}. Link or copy the appropriate assembler,
3702 linker, etc. into @file{@var{prefix}/bin}, and set your @var{PATH} to
3703 include that directory while running both @command{configure} and
3706 You must give @command{configure} the
3707 @option{--with-headers=@var{$WIND_BASE}/target/h} switch so that it can
3708 find the VxWorks system headers. Since VxWorks is a cross compilation
3709 target only, you must also specify @option{--target=@var{target}}.
3710 @command{configure} will attempt to create the directory
3711 @file{@var{prefix}/@var{target}/sys-include} and copy files into it;
3712 make sure the user running @command{configure} has sufficient privilege
3715 GCC's exception handling runtime requires a special ``configlette''
3716 module, @file{contrib/gthr_supp_vxw_5x.c}. Follow the instructions in
3717 that file to add the module to your kernel build. (Future versions of
3718 VxWorks will incorporate this module.)
3723 @heading @anchor{x86-64-x-x}x86_64-*-*, amd64-*-*
3725 GCC supports the x86-64 architecture implemented by the AMD64 processor
3726 (amd64-*-* is an alias for x86_64-*-*) on GNU/Linux, FreeBSD and NetBSD.
3727 On GNU/Linux the default is a bi-arch compiler which is able to generate
3728 both 64-bit x86-64 and 32-bit x86 code (via the @option{-m32} switch).
3733 @heading @anchor{xtensa-x-elf}xtensa-*-elf
3735 This target is intended for embedded Xtensa systems using the
3736 @samp{newlib} C library. It uses ELF but does not support shared
3737 objects. Designed-defined instructions specified via the
3738 Tensilica Instruction Extension (TIE) language are only supported
3739 through inline assembly.
3741 The Xtensa configuration information must be specified prior to
3742 building GCC@. The @file{include/xtensa-config.h} header
3743 file contains the configuration information. If you created your
3744 own Xtensa configuration with the Xtensa Processor Generator, the
3745 downloaded files include a customized copy of this header file,
3746 which you can use to replace the default header file.
3751 @heading @anchor{xtensa-x-linux}xtensa-*-linux*
3753 This target is for Xtensa systems running GNU/Linux. It supports ELF
3754 shared objects and the GNU C library (glibc). It also generates
3755 position-independent code (PIC) regardless of whether the
3756 @option{-fpic} or @option{-fPIC} options are used. In other
3757 respects, this target is the same as the
3758 @uref{#xtensa-*-elf,,@samp{xtensa-*-elf}} target.
3763 @heading @anchor{windows}Microsoft Windows (32-bit)
3765 A port of GCC 2.95.2 and 3.x is included with the
3766 @uref{http://www.cygwin.com/,,Cygwin environment}.
3768 Current (as of early 2001) snapshots of GCC will build under Cygwin
3769 without modification.
3771 GCC does not currently build with Microsoft's C++ compiler and there
3772 are no plans to make it do so.
3777 @heading @anchor{os2}OS/2
3779 GCC does not currently support OS/2. However, Andrew Zabolotny has been
3780 working on a generic OS/2 port with pgcc. The current code can be found
3781 at @uref{http://www.goof.com/pcg/os2/,,http://www.goof.com/pcg/os2/}.
3783 An older copy of GCC 2.8.1 is included with the EMX tools available at
3784 @uref{ftp://ftp.leo.org/pub/comp/os/os2/leo/devtools/emx+gcc/,,
3785 ftp://ftp.leo.org/pub/comp/os/os2/leo/devtools/emx+gcc/}.
3790 @heading @anchor{older}Older systems
3792 GCC contains support files for many older (1980s and early
3793 1990s) Unix variants. For the most part, support for these systems
3794 has not been deliberately removed, but it has not been maintained for
3795 several years and may suffer from bitrot.
3797 Starting with GCC 3.1, each release has a list of ``obsoleted'' systems.
3798 Support for these systems is still present in that release, but
3799 @command{configure} will fail unless the @option{--enable-obsolete}
3800 option is given. Unless a maintainer steps forward, support for these
3801 systems will be removed from the next release of GCC@.
3803 Support for old systems as hosts for GCC can cause problems if the
3804 workarounds for compiler, library and operating system bugs affect the
3805 cleanliness or maintainability of the rest of GCC@. In some cases, to
3806 bring GCC up on such a system, if still possible with current GCC, may
3807 require first installing an old version of GCC which did work on that
3808 system, and using it to compile a more recent GCC, to avoid bugs in the
3809 vendor compiler. Old releases of GCC 1 and GCC 2 are available in the
3810 @file{old-releases} directory on the @uref{../mirrors.html,,GCC mirror
3811 sites}. Header bugs may generally be avoided using
3812 @command{fixincludes}, but bugs or deficiencies in libraries and the
3813 operating system may still cause problems.
3815 Support for older systems as targets for cross-compilation is less
3816 problematic than support for them as hosts for GCC; if an enthusiast
3817 wishes to make such a target work again (including resurrecting any of
3818 the targets that never worked with GCC 2, starting from the last CVS
3819 version before they were removed), patches
3820 @uref{../contribute.html,,following the usual requirements} would be
3821 likely to be accepted, since they should not affect the support for more
3824 For some systems, old versions of GNU binutils may also be useful,
3825 and are available from @file{pub/binutils/old-releases} on
3826 @uref{http://sources.redhat.com/mirrors.html,,sources.redhat.com mirror sites}.
3828 Some of the information on specific systems above relates to
3829 such older systems, but much of the information
3830 about GCC on such systems (which may no longer be applicable to
3831 current GCC) is to be found in the GCC texinfo manual.
3836 @heading @anchor{elf}all ELF targets (SVR4, Solaris 2, etc.)
3838 C++ support is significantly better on ELF targets if you use the
3839 @uref{./configure.html#with-gnu-ld,,GNU linker}; duplicate copies of
3840 inlines, vtables and template instantiations will be discarded
3849 @uref{./index.html,,Return to the GCC Installation page}
3853 @c ***Old documentation******************************************************
3855 @include install-old.texi
3861 @uref{./index.html,,Return to the GCC Installation page}
3865 @c ***GFDL********************************************************************
3873 @uref{./index.html,,Return to the GCC Installation page}
3877 @c ***************************************************************************
3878 @c Part 6 The End of the Document
3880 @comment node-name, next, previous, up
3881 @node Concept Index, , GNU Free Documentation License, Top
3885 @unnumbered Concept Index