3 <title>Host/Target specific installation notes for GCC</title>
4 <meta http-equiv="Content-Type" content="text/html">
5 <meta name="description" content="Host/Target specific installation notes for GCC">
6 <meta name="generator" content="makeinfo 4.12">
7 <link title="Top" rel="top" href="#Top">
8 <link href="http://www.gnu.org/software/texinfo/" rel="generator-home" title="Texinfo Homepage">
10 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997,
11 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009,
12 2010 Free Software Foundation, Inc.
14 Permission is granted to copy, distribute and/or modify this document
15 under the terms of the GNU Free Documentation License, Version 1.3 or
16 any later version published by the Free Software Foundation; with no
17 Invariant Sections, the Front-Cover texts being (a) (see below), and
18 with the Back-Cover Texts being (b) (see below). A copy of the
19 license is included in the section entitled "GNU Free Documentation License".
21 (a) The FSF's Front-Cover Text is:
25 (b) The FSF's Back-Cover Text is:
27 You have freedom to copy and modify this GNU Manual, like GNU
28 software. Copies published by the Free Software Foundation raise
29 funds for GNU development.-->
30 <meta http-equiv="Content-Style-Type" content="text/css">
31 <style type="text/css"><!--
32 pre.display { font-family:inherit }
33 pre.format { font-family:inherit }
34 pre.smalldisplay { font-family:inherit; font-size:smaller }
35 pre.smallformat { font-family:inherit; font-size:smaller }
36 pre.smallexample { font-size:smaller }
37 pre.smalllisp { font-size:smaller }
38 span.sc { font-variant:small-caps }
39 span.roman { font-family:serif; font-weight:normal; }
40 span.sansserif { font-family:sans-serif; font-weight:normal; }
44 <h1 class="settitle">Host/Target specific installation notes for GCC</h1>
45 <a name="index-Specific-1"></a><a name="index-Specific-installation-notes-2"></a><a name="index-Target-specific-installation-3"></a><a name="index-Host-specific-installation-4"></a><a name="index-Target-specific-installation-notes-5"></a>
46 Please read this document carefully <em>before</em> installing the
47 GNU Compiler Collection on your machine.
49 <p>Note that this list of install notes is <em>not</em> a list of supported
50 hosts or targets. Not all supported hosts and targets are listed
51 here, only the ones that require host-specific or target-specific
55 <li><a href="#alpha-x-x">alpha*-*-*</a>
56 <li><a href="#alpha-dec-osf51">alpha*-dec-osf5.1</a>
57 <li><a href="#amd64-x-solaris210">amd64-*-solaris2.10</a>
58 <li><a href="#arm-x-eabi">arm-*-eabi</a>
59 <li><a href="#avr">avr</a>
60 <li><a href="#bfin">Blackfin</a>
61 <li><a href="#dos">DOS</a>
62 <li><a href="#x-x-freebsd">*-*-freebsd*</a>
63 <li><a href="#h8300-hms">h8300-hms</a>
64 <li><a href="#hppa-hp-hpux">hppa*-hp-hpux*</a>
65 <li><a href="#hppa-hp-hpux10">hppa*-hp-hpux10</a>
66 <li><a href="#hppa-hp-hpux11">hppa*-hp-hpux11</a>
67 <li><a href="#x-x-linux-gnu">*-*-linux-gnu</a>
68 <li><a href="#ix86-x-linux">i?86-*-linux*</a>
69 <li><a href="#ix86-x-solaris289">i?86-*-solaris2.[89]</a>
70 <li><a href="#ix86-x-solaris210">i?86-*-solaris2.10</a>
71 <li><a href="#ia64-x-linux">ia64-*-linux</a>
72 <li><a href="#ia64-x-hpux">ia64-*-hpux*</a>
73 <li><a href="#x-ibm-aix">*-ibm-aix*</a>
74 <li><a href="#iq2000-x-elf">iq2000-*-elf</a>
75 <li><a href="#lm32-x-elf">lm32-*-elf</a>
76 <li><a href="#lm32-x-uclinux">lm32-*-uclinux</a>
77 <li><a href="#m32c-x-elf">m32c-*-elf</a>
78 <li><a href="#m32r-x-elf">m32r-*-elf</a>
79 <li><a href="#m68k-x-x">m68k-*-*</a>
80 <li><a href="#m68k-uclinux">m68k-uclinux</a>
81 <li><a href="#mep-x-elf">mep-*-elf</a>
82 <li><a href="#microblaze-x-elf">microblaze-*-elf</a>
83 <li><a href="#mips-x-x">mips-*-*</a>
84 <li><a href="#mips-sgi-irix5">mips-sgi-irix5</a>
85 <li><a href="#mips-sgi-irix6">mips-sgi-irix6</a>
86 <li><a href="#powerpc-x-x">powerpc*-*-*</a>
87 <li><a href="#powerpc-x-darwin">powerpc-*-darwin*</a>
88 <li><a href="#powerpc-x-elf">powerpc-*-elf</a>
89 <li><a href="#powerpc-x-linux-gnu">powerpc*-*-linux-gnu*</a>
90 <li><a href="#powerpc-x-netbsd">powerpc-*-netbsd*</a>
91 <li><a href="#powerpc-x-eabisim">powerpc-*-eabisim</a>
92 <li><a href="#powerpc-x-eabi">powerpc-*-eabi</a>
93 <li><a href="#powerpcle-x-elf">powerpcle-*-elf</a>
94 <li><a href="#powerpcle-x-eabisim">powerpcle-*-eabisim</a>
95 <li><a href="#powerpcle-x-eabi">powerpcle-*-eabi</a>
96 <li><a href="#s390-x-linux">s390-*-linux*</a>
97 <li><a href="#s390x-x-linux">s390x-*-linux*</a>
98 <li><a href="#s390x-ibm-tpf">s390x-ibm-tpf*</a>
99 <li><a href="#x-x-solaris2">*-*-solaris2*</a>
100 <li><a href="#sparc-x-x">sparc*-*-*</a>
101 <li><a href="#sparc-sun-solaris2">sparc-sun-solaris2*</a>
102 <li><a href="#sparc-sun-solaris210">sparc-sun-solaris2.10</a>
103 <li><a href="#sparc-x-linux">sparc-*-linux*</a>
104 <li><a href="#sparc64-x-solaris2">sparc64-*-solaris2*</a>
105 <li><a href="#sparcv9-x-solaris2">sparcv9-*-solaris2*</a>
106 <li><a href="#c6x-x-x">c6x-*-*</a>
107 <li><a href="#tilegx-x-linux">tilegx-*-linux*</a>
108 <li><a href="#tilepro-x-linux">tilepro-*-linux*</a>
109 <li><a href="#x-x-vxworks">*-*-vxworks*</a>
110 <li><a href="#x86-64-x-x">x86_64-*-*</a> amd64-*-*
111 <li><a href="#x86-64-x-solaris210">x86_64-*-solaris2.1[0-9]*</a>
112 <li><a href="#xtensa-x-elf">xtensa*-*-elf</a>
113 <li><a href="#xtensa-x-linux">xtensa*-*-linux*</a>
114 <li><a href="#windows">Microsoft Windows</a>
115 <li><a href="#x-x-cygwin">*-*-cygwin</a>
116 <li><a href="#x-x-interix">*-*-interix</a>
117 <li><a href="#x-x-mingw32">*-*-mingw32</a>
118 <li><a href="#os2">OS/2</a>
119 <li><a href="#older">Older systems</a>
123 <li><a href="#elf">all ELF targets</a> (SVR4, Solaris 2, etc.)
126 <p><!-- -------- host/target specific issues start here ---------------- -->
129 <h3 class="heading"><a name="TOC0"></a><a name="alpha_002dx_002dx"></a>alpha*-*-*</h3>
131 <p>This section contains general configuration information for all
132 alpha-based platforms using ELF (in particular, ignore this section for
133 DEC OSF/1, Digital UNIX and Tru64 UNIX). In addition to reading this
134 section, please read all other sections that match your target.
136 <p>We require binutils 2.11.2 or newer.
137 Previous binutils releases had a number of problems with DWARF 2
138 debugging information, not the least of which is incorrect linking of
143 <h3 class="heading"><a name="TOC1"></a><a name="alpha_002ddec_002dosf51"></a>alpha*-dec-osf5.1</h3>
145 <p>Systems using processors that implement the DEC Alpha architecture and
146 are running the DEC/Compaq/HP Unix (DEC OSF/1, Digital UNIX, or Compaq/HP
147 Tru64 UNIX) operating system, for example the DEC Alpha AXP systems.
149 <p>Support for Tru64 UNIX V5.1 has been obsoleted in GCC 4.7, but can still
150 be enabled by configuring with <samp><span class="option">--enable-obsolete</span></samp>. Support will
151 be removed in GCC 4.8. As of GCC 4.6, support for Tru64 UNIX V4.0 and
152 V5.0 has been removed. As of GCC 3.2, versions before
153 <code>alpha*-dec-osf4</code> are no longer supported. (These are the versions
154 which identify themselves as DEC OSF/1.)
156 <p>On Tru64 UNIX, virtual memory exhausted bootstrap failures
157 may be fixed by reconfiguring Kernel Virtual Memory and Swap parameters
158 per the <samp><span class="command">/usr/sbin/sys_check</span></samp> Tuning Suggestions,
159 or applying the patch in
160 <a href="http://gcc.gnu.org/ml/gcc/2002-08/msg00822.html">http://gcc.gnu.org/ml/gcc/2002-08/msg00822.html</a>. Depending on
161 the OS version used, you need a data segment size between 512 MB and
162 1 GB, so simply use <samp><span class="command">ulimit -Sd unlimited</span></samp>.
164 <p>As of GNU binutils 2.22, neither GNU <samp><span class="command">as</span></samp> nor GNU <samp><span class="command">ld</span></samp>
165 are supported on Tru64 UNIX, so you must not configure GCC with
166 <samp><span class="option">--with-gnu-as</span></samp> or <samp><span class="option">--with-gnu-ld</span></samp>.
168 <p>Cross-compilers for the Tru64 UNIX target currently do not work because
169 the auxiliary programs <samp><span class="command">mips-tdump</span></samp> and <samp><span class="command">mips-tfile</span></samp> can't
170 be compiled on anything but Tru64 UNIX.
172 <p>GCC writes a ‘<samp><span class="samp">.verstamp</span></samp>’ directive to the assembler output file
173 unless it is built as a cross-compiler. It gets the version to use from
174 the system header file <samp><span class="file">/usr/include/stamp.h</span></samp>. If you install a
175 new version of Tru64 UNIX, you should rebuild GCC to pick up the new version
178 <p>GCC now supports both the native (ECOFF) debugging format used by DBX
179 and GDB and an encapsulated STABS format for use only with GDB. See the
180 discussion of the <samp><span class="option">--with-stabs</span></samp> option of <samp><span class="file">configure</span></samp> above
181 for more information on these formats and how to select them.
182 <!-- FIXME: does this work at all? If so, perhaps make default. -->
184 <p>There is a bug in DEC's assembler that produces incorrect line numbers
185 for ECOFF format when the ‘<samp><span class="samp">.align</span></samp>’ directive is used. To work
186 around this problem, GCC will not emit such alignment directives
187 while writing ECOFF format debugging information even if optimization is
188 being performed. Unfortunately, this has the very undesirable
189 side-effect that code addresses when <samp><span class="option">-O</span></samp> is specified are
190 different depending on whether or not <samp><span class="option">-g</span></samp> is also specified.
192 <p>To avoid this behavior, specify <samp><span class="option">-gstabs+</span></samp> and use GDB instead of
193 DBX. DEC is now aware of this problem with the assembler and hopes to
194 provide a fix shortly.
196 <!-- FIXME: still applicable? -->
199 <h3 class="heading"><a name="TOC2"></a><a name="amd64_002dx_002dsolaris210"></a>amd64-*-solaris2.1[0-9]*</h3>
201 <p>This is a synonym for ‘<samp><span class="samp">x86_64-*-solaris2.1[0-9]*</span></samp>’.
205 <h3 class="heading"><a name="TOC3"></a><a name="arm_002dx_002deabi"></a>arm-*-eabi</h3>
207 <p>ARM-family processors. Subtargets that use the ELF object format
208 require GNU binutils 2.13 or newer. Such subtargets include:
209 <code>arm-*-netbsdelf</code>, <code>arm-*-*linux-gnueabi*</code>
210 and <code>arm-*-rtemseabi</code>.
214 <h3 class="heading"><a name="TOC4"></a><a name="avr"></a>avr</h3>
216 <p>ATMEL AVR-family micro controllers. These are used in embedded
217 applications. There are no standard Unix configurations.
218 See “AVR Options” in the main manual
219 for the list of supported MCU types.
221 <p>Use ‘<samp><span class="samp">configure --target=avr --enable-languages="c"</span></samp>’ to configure GCC.
223 <p>Further installation notes and other useful information about AVR tools
224 can also be obtained from:
227 <li><a href="http://www.nongnu.org/avr/">http://www.nongnu.org/avr/</a>
228 <li><a href="http://www.amelek.gda.pl/avr/">http://www.amelek.gda.pl/avr/</a>
231 <p>We <em>strongly</em> recommend using binutils 2.13 or newer.
233 <p>The following error:
234 <pre class="smallexample"> Error: register required
236 <p>indicates that you should upgrade to a newer version of the binutils.
240 <h3 class="heading"><a name="TOC5"></a><a name="bfin"></a>Blackfin</h3>
242 <p>The Blackfin processor, an Analog Devices DSP.
243 See “Blackfin Options” in the main manual
245 <p>More information, and a version of binutils with support for this processor,
246 is available at <a href="http://blackfin.uclinux.org">http://blackfin.uclinux.org</a>
250 <h3 class="heading"><a name="TOC6"></a><a name="cr16"></a>CR16</h3>
252 <p>The CR16 CompactRISC architecture is a 16-bit architecture. This architecture is
253 used in embedded applications.
255 <p>See “CR16 Options” in the main manual for a list of CR16-specific options.
257 <p>Use ‘<samp><span class="samp">configure --target=cr16-elf --enable-languages=c,c++</span></samp>’ to configure
258 GCC for building a CR16 elf cross-compiler.
260 <p>Use ‘<samp><span class="samp">configure --target=cr16-uclinux --enable-languages=c,c++</span></samp>’ to configure
261 GCC for building a CR16 uclinux cross-compiler.
265 <h3 class="heading"><a name="TOC7"></a><a name="cris"></a>CRIS</h3>
267 <p>CRIS is the CPU architecture in Axis Communications ETRAX system-on-a-chip
268 series. These are used in embedded applications.
270 <p>See “CRIS Options” in the main manual
271 for a list of CRIS-specific options.
273 <p>There are a few different CRIS targets:
275 <dt><code>cris-axis-elf</code><dd>Mainly for monolithic embedded systems. Includes a multilib for the
276 ‘<samp><span class="samp">v10</span></samp>’ core used in ‘<samp><span class="samp">ETRAX 100 LX</span></samp>’.
277 <br><dt><code>cris-axis-linux-gnu</code><dd>A GNU/Linux port for the CRIS architecture, currently targeting
278 ‘<samp><span class="samp">ETRAX 100 LX</span></samp>’ by default.
281 <p>For <code>cris-axis-elf</code> you need binutils 2.11
282 or newer. For <code>cris-axis-linux-gnu</code> you need binutils 2.12 or newer.
284 <p>Pre-packaged tools can be obtained from
285 <a href="ftp://ftp.axis.com/pub/axis/tools/cris/compiler-kit/">ftp://ftp.axis.com/pub/axis/tools/cris/compiler-kit/</a>. More
286 information about this platform is available at
287 <a href="http://developer.axis.com/">http://developer.axis.com/</a>.
291 <h3 class="heading"><a name="TOC8"></a><a name="dos"></a>DOS</h3>
293 <p>Please have a look at the <a href="binaries.html">binaries page</a>.
295 <p>You cannot install GCC by itself on MSDOS; it will not compile under
296 any MSDOS compiler except itself. You need to get the complete
297 compilation package DJGPP, which includes binaries as well as sources,
298 and includes all the necessary compilation tools and libraries.
302 <h3 class="heading"><a name="TOC9"></a><a name="epiphany_002dx_002delf"></a>epiphany-*-elf</h3>
304 <p>Adapteva Epiphany.
305 This configuration is intended for embedded systems.
309 <h3 class="heading"><a name="TOC10"></a><a name="x_002dx_002dfreebsd"></a>*-*-freebsd*</h3>
311 <p>Support for FreeBSD 1 was discontinued in GCC 3.2. Support for
312 FreeBSD 2 (and any mutant a.out variants of FreeBSD 3) was
313 discontinued in GCC 4.0.
315 <p>In order to better utilize FreeBSD base system functionality and match
316 the configuration of the system compiler, GCC 4.5 and above as well as
317 GCC 4.4 past 2010-06-20 leverage SSP support in libc (which is present
318 on FreeBSD 7 or later) and the use of <code>__cxa_atexit</code> by default
319 (on FreeBSD 6 or later). The use of <code>dl_iterate_phdr</code> inside
320 <samp><span class="file">libgcc_s.so.1</span></samp> and boehm-gc (on FreeBSD 7 or later) is enabled
321 by GCC 4.5 and above.
323 <p>We support FreeBSD using the ELF file format with DWARF 2 debugging
324 for all CPU architectures. You may use <samp><span class="option">-gstabs</span></samp> instead of
325 <samp><span class="option">-g</span></samp>, if you really want the old debugging format. There are
326 no known issues with mixing object files and libraries with different
327 debugging formats. Otherwise, this release of GCC should now match
328 more of the configuration used in the stock FreeBSD configuration of
329 GCC. In particular, <samp><span class="option">--enable-threads</span></samp> is now configured by
330 default. However, as a general user, do not attempt to replace the
331 system compiler with this release. Known to bootstrap and check with
332 good results on FreeBSD 7.2-STABLE. In the past, known to bootstrap
333 and check with good results on FreeBSD 3.0, 3.4, 4.0, 4.2, 4.3, 4.4,
334 4.5, 4.8, 4.9 and 5-CURRENT.
336 <p>The version of binutils installed in <samp><span class="file">/usr/bin</span></samp> probably works
337 with this release of GCC. Bootstrapping against the latest GNU
338 binutils and/or the version found in <samp><span class="file">/usr/ports/devel/binutils</span></samp> has
339 been known to enable additional features and improve overall testsuite
340 results. However, it is currently known that boehm-gc (which itself
341 is required for java) may not configure properly on FreeBSD prior to
342 the FreeBSD 7.0 release with GNU binutils after 2.16.1.
346 <h3 class="heading"><a name="TOC11"></a><a name="h8300_002dhms"></a>h8300-hms</h3>
348 <p>Renesas H8/300 series of processors.
350 <p>Please have a look at the <a href="binaries.html">binaries page</a>.
352 <p>The calling convention and structure layout has changed in release 2.6.
353 All code must be recompiled. The calling convention now passes the
354 first three arguments in function calls in registers. Structures are no
355 longer a multiple of 2 bytes.
359 <h3 class="heading"><a name="TOC12"></a><a name="hppa_002dhp_002dhpux"></a>hppa*-hp-hpux*</h3>
361 <p>Support for HP-UX version 9 and older was discontinued in GCC 3.4.
363 <p>We require using gas/binutils on all hppa platforms. Version 2.19 or
364 later is recommended.
366 <p>It may be helpful to configure GCC with the
367 <a href="./configure.html#with-gnu-as"><samp><span class="option">--with-gnu-as</span></samp></a> and
368 <samp><span class="option">--with-as=...</span></samp> options to ensure that GCC can find GAS.
370 <p>The HP assembler should not be used with GCC. It is rarely tested and may
371 not work. It shouldn't be used with any languages other than C due to its
374 <p>Specifically, <samp><span class="option">-g</span></samp> does not work (HP-UX uses a peculiar debugging
375 format which GCC does not know about). It also inserts timestamps
376 into each object file it creates, causing the 3-stage comparison test to
377 fail during a bootstrap. You should be able to continue by saying
378 ‘<samp><span class="samp">make all-host all-target</span></samp>’ after getting the failure from ‘<samp><span class="samp">make</span></samp>’.
380 <p>Various GCC features are not supported. For example, it does not support weak
381 symbols or alias definitions. As a result, explicit template instantiations
382 are required when using C++. This makes it difficult if not impossible to
383 build many C++ applications.
385 <p>There are two default scheduling models for instructions. These are
386 PROCESSOR_7100LC and PROCESSOR_8000. They are selected from the pa-risc
387 architecture specified for the target machine when configuring.
388 PROCESSOR_8000 is the default. PROCESSOR_7100LC is selected when
389 the target is a ‘<samp><span class="samp">hppa1*</span></samp>’ machine.
391 <p>The PROCESSOR_8000 model is not well suited to older processors. Thus,
392 it is important to completely specify the machine architecture when
393 configuring if you want a model other than PROCESSOR_8000. The macro
394 TARGET_SCHED_DEFAULT can be defined in BOOT_CFLAGS if a different
395 default scheduling model is desired.
397 <p>As of GCC 4.0, GCC uses the UNIX 95 namespace for HP-UX 10.10
398 through 11.00, and the UNIX 98 namespace for HP-UX 11.11 and later.
399 This namespace change might cause problems when bootstrapping with
400 an earlier version of GCC or the HP compiler as essentially the same
401 namespace is required for an entire build. This problem can be avoided
402 in a number of ways. With HP cc, <samp><span class="env">UNIX_STD</span></samp> can be set to ‘<samp><span class="samp">95</span></samp>’
403 or ‘<samp><span class="samp">98</span></samp>’. Another way is to add an appropriate set of predefines
404 to <samp><span class="env">CC</span></samp>. The description for the <samp><span class="option">munix=</span></samp> option contains
405 a list of the predefines used with each standard.
407 <p>More specific information to ‘<samp><span class="samp">hppa*-hp-hpux*</span></samp>’ targets follows.
411 <h3 class="heading"><a name="TOC13"></a><a name="hppa_002dhp_002dhpux10"></a>hppa*-hp-hpux10</h3>
413 <p>For hpux10.20, we <em>highly</em> recommend you pick up the latest sed patch
414 <code>PHCO_19798</code> from HP.
416 <p>The C++ ABI has changed incompatibly in GCC 4.0. COMDAT subspaces are
417 used for one-only code and data. This resolves many of the previous
418 problems in using C++ on this target. However, the ABI is not compatible
419 with the one implemented under HP-UX 11 using secondary definitions.
423 <h3 class="heading"><a name="TOC14"></a><a name="hppa_002dhp_002dhpux11"></a>hppa*-hp-hpux11</h3>
425 <p>GCC 3.0 and up support HP-UX 11. GCC 2.95.x is not supported and cannot
426 be used to compile GCC 3.0 and up.
428 <p>The libffi and libjava libraries haven't been ported to 64-bit HP-UX and don't build.
430 <p>Refer to <a href="binaries.html">binaries</a> for information about obtaining
431 precompiled GCC binaries for HP-UX. Precompiled binaries must be obtained
432 to build the Ada language as it can't be bootstrapped using C. Ada is
433 only available for the 32-bit PA-RISC runtime.
435 <p>Starting with GCC 3.4 an ISO C compiler is required to bootstrap. The
436 bundled compiler supports only traditional C; you will need either HP's
437 unbundled compiler, or a binary distribution of GCC.
439 <p>It is possible to build GCC 3.3 starting with the bundled HP compiler,
440 but the process requires several steps. GCC 3.3 can then be used to
441 build later versions. The fastjar program contains ISO C code and
442 can't be built with the HP bundled compiler. This problem can be
443 avoided by not building the Java language. For example, use the
444 <samp><span class="option">--enable-languages="c,c++,f77,objc"</span></samp> option in your configure
447 <p>There are several possible approaches to building the distribution.
448 Binutils can be built first using the HP tools. Then, the GCC
449 distribution can be built. The second approach is to build GCC
450 first using the HP tools, then build binutils, then rebuild GCC.
451 There have been problems with various binary distributions, so it
452 is best not to start from a binary distribution.
454 <p>On 64-bit capable systems, there are two distinct targets. Different
455 installation prefixes must be used if both are to be installed on
456 the same system. The ‘<samp><span class="samp">hppa[1-2]*-hp-hpux11*</span></samp>’ target generates code
457 for the 32-bit PA-RISC runtime architecture and uses the HP linker.
458 The ‘<samp><span class="samp">hppa64-hp-hpux11*</span></samp>’ target generates 64-bit code for the
459 PA-RISC 2.0 architecture.
461 <p>The script config.guess now selects the target type based on the compiler
462 detected during configuration. You must define <samp><span class="env">PATH</span></samp> or <samp><span class="env">CC</span></samp> so
463 that configure finds an appropriate compiler for the initial bootstrap.
464 When <samp><span class="env">CC</span></samp> is used, the definition should contain the options that are
465 needed whenever <samp><span class="env">CC</span></samp> is used.
467 <p>Specifically, options that determine the runtime architecture must be
468 in <samp><span class="env">CC</span></samp> to correctly select the target for the build. It is also
469 convenient to place many other compiler options in <samp><span class="env">CC</span></samp>. For example,
470 <samp><span class="env">CC="cc -Ac +DA2.0W -Wp,-H16376 -D_CLASSIC_TYPES -D_HPUX_SOURCE"</span></samp>
471 can be used to bootstrap the GCC 3.3 branch with the HP compiler in
472 64-bit K&R/bundled mode. The <samp><span class="option">+DA2.0W</span></samp> option will result in
473 the automatic selection of the ‘<samp><span class="samp">hppa64-hp-hpux11*</span></samp>’ target. The
474 macro definition table of cpp needs to be increased for a successful
475 build with the HP compiler. _CLASSIC_TYPES and _HPUX_SOURCE need to
476 be defined when building with the bundled compiler, or when using the
477 <samp><span class="option">-Ac</span></samp> option. These defines aren't necessary with <samp><span class="option">-Ae</span></samp>.
479 <p>It is best to explicitly configure the ‘<samp><span class="samp">hppa64-hp-hpux11*</span></samp>’ target
480 with the <samp><span class="option">--with-ld=...</span></samp> option. This overrides the standard
481 search for ld. The two linkers supported on this target require different
482 commands. The default linker is determined during configuration. As a
483 result, it's not possible to switch linkers in the middle of a GCC build.
484 This has been reported to sometimes occur in unified builds of binutils
487 <p>A recent linker patch must be installed for the correct operation of
488 GCC 3.3 and later. <code>PHSS_26559</code> and <code>PHSS_24304</code> are the
489 oldest linker patches that are known to work. They are for HP-UX
490 11.00 and 11.11, respectively. <code>PHSS_24303</code>, the companion to
491 <code>PHSS_24304</code>, might be usable but it hasn't been tested. These
492 patches have been superseded. Consult the HP patch database to obtain
493 the currently recommended linker patch for your system.
495 <p>The patches are necessary for the support of weak symbols on the
496 32-bit port, and for the running of initializers and finalizers. Weak
497 symbols are implemented using SOM secondary definition symbols. Prior
498 to HP-UX 11, there are bugs in the linker support for secondary symbols.
499 The patches correct a problem of linker core dumps creating shared
500 libraries containing secondary symbols, as well as various other
501 linking issues involving secondary symbols.
503 <p>GCC 3.3 uses the ELF DT_INIT_ARRAY and DT_FINI_ARRAY capabilities to
504 run initializers and finalizers on the 64-bit port. The 32-bit port
505 uses the linker <samp><span class="option">+init</span></samp> and <samp><span class="option">+fini</span></samp> options for the same
506 purpose. The patches correct various problems with the +init/+fini
507 options, including program core dumps. Binutils 2.14 corrects a
508 problem on the 64-bit port resulting from HP's non-standard use of
509 the .init and .fini sections for array initializers and finalizers.
511 <p>Although the HP and GNU linkers are both supported for the
512 ‘<samp><span class="samp">hppa64-hp-hpux11*</span></samp>’ target, it is strongly recommended that the
513 HP linker be used for link editing on this target.
515 <p>At this time, the GNU linker does not support the creation of long
516 branch stubs. As a result, it can't successfully link binaries
517 containing branch offsets larger than 8 megabytes. In addition,
518 there are problems linking shared libraries, linking executables
519 with <samp><span class="option">-static</span></samp>, and with dwarf2 unwind and exception support.
520 It also doesn't provide stubs for internal calls to global functions
521 in shared libraries, so these calls can't be overloaded.
523 <p>The HP dynamic loader does not support GNU symbol versioning, so symbol
524 versioning is not supported. It may be necessary to disable symbol
525 versioning with <samp><span class="option">--disable-symvers</span></samp> when using GNU ld.
527 <p>POSIX threads are the default. The optional DCE thread library is not
528 supported, so <samp><span class="option">--enable-threads=dce</span></samp> does not work.
532 <h3 class="heading"><a name="TOC15"></a><a name="x_002dx_002dlinux_002dgnu"></a>*-*-linux-gnu</h3>
534 <p>Versions of libstdc++-v3 starting with 3.2.1 require bug fixes present
535 in glibc 2.2.5 and later. More information is available in the
536 libstdc++-v3 documentation.
540 <h3 class="heading"><a name="TOC16"></a><a name="ix86_002dx_002dlinux"></a>i?86-*-linux*</h3>
542 <p>As of GCC 3.3, binutils 2.13.1 or later is required for this platform.
543 See <a href="http://gcc.gnu.org/PR10877">bug 10877</a> for more information.
545 <p>If you receive Signal 11 errors when building on GNU/Linux, then it is
546 possible you have a hardware problem. Further information on this can be
547 found on <a href="http://www.bitwizard.nl/sig11/">www.bitwizard.nl</a>.
551 <h3 class="heading"><a name="TOC17"></a><a name="ix86_002dx_002dsolaris289"></a>i?86-*-solaris2.[89]</h3>
553 <p>The Sun assembler in Solaris 8 and 9 has several bugs and limitations.
554 While GCC works around them, several features are missing, so it is
555 <!-- FIXME: which ones? -->
556 recommended to use the GNU assembler instead. There is no bundled
557 version, but the current version, from GNU binutils 2.22, is known to
560 <p>Solaris 2/x86 doesn't support the execution of SSE/SSE2 instructions
561 before Solaris 9 4/04, even if the CPU supports them. Programs will
562 receive <code>SIGILL</code> if they try. The fix is available both in
563 Solaris 9 Update 6 and kernel patch 112234-12 or newer. There is no
564 corresponding patch for Solaris 8. To avoid this problem,
565 <samp><span class="option">-march</span></samp> defaults to ‘<samp><span class="samp">pentiumpro</span></samp>’ on Solaris 8 and 9. If
566 you have the patch installed, you can configure GCC with an appropriate
567 <samp><span class="option">--with-arch</span></samp> option, but need GNU <samp><span class="command">as</span></samp> for SSE2 support.
571 <h3 class="heading"><a name="TOC18"></a><a name="ix86_002dx_002dsolaris210"></a>i?86-*-solaris2.10</h3>
573 <p>Use this for Solaris 10 or later on x86 and x86-64 systems. Starting
574 with GCC 4.7, there is also a 64-bit ‘<samp><span class="samp">amd64-*-solaris2.1[0-9]*</span></samp>’ or
575 ‘<samp><span class="samp">x86_64-*-solaris2.1[0-9]*</span></samp>’ configuration that corresponds to
576 ‘<samp><span class="samp">sparcv9-sun-solaris2*</span></samp>’.
578 <p>It is recommended that you configure GCC to use the GNU assembler, in
579 <samp><span class="file">/usr/sfw/bin/gas</span></samp>. The versions included in Solaris 10, from GNU
580 binutils 2.15, and Solaris 11, from GNU binutils 2.19, work fine,
581 although the current version, from GNU binutils
582 2.22, is known to work, too. Recent versions of the Sun assembler in
583 <samp><span class="file">/usr/ccs/bin/as</span></samp> work almost as well, though.
584 <!-- FIXME: as patch requirements? -->
586 <p>For linking, the Sun linker, is preferred. If you want to use the GNU
587 linker instead, which is available in <samp><span class="file">/usr/sfw/bin/gld</span></samp>, note that
588 due to a packaging bug the version in Solaris 10, from GNU binutils
589 2.15, cannot be used, while the version in Solaris 11, from GNU binutils
590 2.19, works, as does the latest version, from GNU binutils 2.22.
592 <p>To use GNU <samp><span class="command">as</span></samp>, configure with the options
593 <samp><span class="option">--with-gnu-as --with-as=/usr/sfw/bin/gas</span></samp>. It may be necessary
594 to configure with <samp><span class="option">--without-gnu-ld --with-ld=/usr/ccs/bin/ld</span></samp> to
595 guarantee use of Sun <samp><span class="command">ld</span></samp>.
596 <!-- FIXME: why -without-gnu-ld -with-ld? -->
600 <h3 class="heading"><a name="TOC19"></a><a name="ia64_002dx_002dlinux"></a>ia64-*-linux</h3>
602 <p>IA-64 processor (also known as IPF, or Itanium Processor Family)
605 <p>If you are using the installed system libunwind library with
606 <samp><span class="option">--with-system-libunwind</span></samp>, then you must use libunwind 0.98 or
609 <p>None of the following versions of GCC has an ABI that is compatible
610 with any of the other versions in this list, with the exception that
611 Red Hat 2.96 and Trillian 000171 are compatible with each other:
612 3.1, 3.0.2, 3.0.1, 3.0, Red Hat 2.96, and Trillian 000717.
613 This primarily affects C++ programs and programs that create shared libraries.
614 GCC 3.1 or later is recommended for compiling linux, the kernel.
615 As of version 3.1 GCC is believed to be fully ABI compliant, and hence no
616 more major ABI changes are expected.
620 <h3 class="heading"><a name="TOC20"></a><a name="ia64_002dx_002dhpux"></a>ia64-*-hpux*</h3>
622 <p>Building GCC on this target requires the GNU Assembler. The bundled HP
623 assembler will not work. To prevent GCC from using the wrong assembler,
624 the option <samp><span class="option">--with-gnu-as</span></samp> may be necessary.
626 <p>The GCC libunwind library has not been ported to HPUX. This means that for
627 GCC versions 3.2.3 and earlier, <samp><span class="option">--enable-libunwind-exceptions</span></samp>
628 is required to build GCC. For GCC 3.3 and later, this is the default.
629 For gcc 3.4.3 and later, <samp><span class="option">--enable-libunwind-exceptions</span></samp> is
630 removed and the system libunwind library will always be used.
633 <!-- rs6000-ibm-aix*, powerpc-ibm-aix* -->
635 <h3 class="heading"><a name="TOC21"></a><a name="x_002dibm_002daix"></a>*-ibm-aix*</h3>
637 <p>Support for AIX version 3 and older was discontinued in GCC 3.4.
638 Support for AIX version 4.2 and older was discontinued in GCC 4.5.
640 <p>“out of memory” bootstrap failures may indicate a problem with
641 process resource limits (ulimit). Hard limits are configured in the
642 <samp><span class="file">/etc/security/limits</span></samp> system configuration file.
644 <p>GCC can bootstrap with recent versions of IBM XLC, but bootstrapping
645 with an earlier release of GCC is recommended. Bootstrapping with XLC
646 requires a larger data segment, which can be enabled through the
647 <var>LDR_CNTRL</var> environment variable, e.g.,
649 <pre class="smallexample"> % LDR_CNTRL=MAXDATA=0x50000000
652 <p>One can start with a pre-compiled version of GCC to build from
653 sources. One may delete GCC's “fixed” header files when starting
654 with a version of GCC built for an earlier release of AIX.
656 <p>To speed up the configuration phases of bootstrapping and installing GCC,
657 one may use GNU Bash instead of AIX <samp><span class="command">/bin/sh</span></samp>, e.g.,
659 <pre class="smallexample"> % CONFIG_SHELL=/opt/freeware/bin/bash
660 % export CONFIG_SHELL
662 <p>and then proceed as described in <a href="build.html">the build instructions</a>, where we strongly recommend specifying an absolute path
663 to invoke <var>srcdir</var>/configure.
665 <p>Because GCC on AIX is built as a 32-bit executable by default,
666 (although it can generate 64-bit programs) the GMP and MPFR libraries
667 required by gfortran must be 32-bit libraries. Building GMP and MPFR
668 as static archive libraries works better than shared libraries.
670 <p>Errors involving <code>alloca</code> when building GCC generally are due
671 to an incorrect definition of <code>CC</code> in the Makefile or mixing files
672 compiled with the native C compiler and GCC. During the stage1 phase of
673 the build, the native AIX compiler <strong>must</strong> be invoked as <samp><span class="command">cc</span></samp>
674 (not <samp><span class="command">xlc</span></samp>). Once <samp><span class="command">configure</span></samp> has been informed of
675 <samp><span class="command">xlc</span></samp>, one needs to use ‘<samp><span class="samp">make distclean</span></samp>’ to remove the
676 configure cache files and ensure that <samp><span class="env">CC</span></samp> environment variable
677 does not provide a definition that will confuse <samp><span class="command">configure</span></samp>.
678 If this error occurs during stage2 or later, then the problem most likely
679 is the version of Make (see above).
681 <p>The native <samp><span class="command">as</span></samp> and <samp><span class="command">ld</span></samp> are recommended for bootstrapping
682 on AIX. The GNU Assembler, GNU Linker, and GNU Binutils version 2.20
683 is required to bootstrap on AIX 5. The native AIX tools do
684 interoperate with GCC.
686 <p>AIX 5.3 TL10, AIX 6.1 TL05 and AIX 7.1 TL00 introduced an AIX
687 assembler change that sometimes produces corrupt assembly files
688 causing AIX linker errors. The bug breaks GCC bootstrap on AIX and
689 can cause compilation failures with existing GCC installations. An
690 AIX iFix for AIX 5.3 is available (APAR IZ98385 for AIX 5.3 TL10, APAR
691 IZ98477 for AIX 5.3 TL11 and IZ98134 for AIX 5.3 TL12). Fixes for AIX
692 6.1 (APAR IZ98732 for AIX 6.1 TL05 and APAR IZ98861 for AIX 6.1 TL06)
693 and AIX 7.1 are in verification and packaging phases.
695 <p>Building <samp><span class="file">libstdc++.a</span></samp> requires a fix for an AIX Assembler bug
696 APAR IY26685 (AIX 4.3) or APAR IY25528 (AIX 5.1). It also requires a
697 fix for another AIX Assembler bug and a co-dependent AIX Archiver fix
698 referenced as APAR IY53606 (AIX 5.2) or as APAR IY54774 (AIX 5.1)
700 <p>‘<samp><span class="samp">libstdc++</span></samp>’ in GCC 3.4 increments the major version number of the
701 shared object and GCC installation places the <samp><span class="file">libstdc++.a</span></samp>
702 shared library in a common location which will overwrite the and GCC
703 3.3 version of the shared library. Applications either need to be
704 re-linked against the new shared library or the GCC 3.1 and GCC 3.3
705 versions of the ‘<samp><span class="samp">libstdc++</span></samp>’ shared object needs to be available
706 to the AIX runtime loader. The GCC 3.1 ‘<samp><span class="samp">libstdc++.so.4</span></samp>’, if
707 present, and GCC 3.3 ‘<samp><span class="samp">libstdc++.so.5</span></samp>’ shared objects can be
708 installed for runtime dynamic loading using the following steps to set
709 the ‘<samp><span class="samp">F_LOADONLY</span></samp>’ flag in the shared object for <em>each</em>
710 multilib <samp><span class="file">libstdc++.a</span></samp> installed:
712 <p>Extract the shared objects from the currently installed
713 <samp><span class="file">libstdc++.a</span></samp> archive:
714 <pre class="smallexample"> % ar -x libstdc++.a libstdc++.so.4 libstdc++.so.5
716 <p>Enable the ‘<samp><span class="samp">F_LOADONLY</span></samp>’ flag so that the shared object will be
717 available for runtime dynamic loading, but not linking:
718 <pre class="smallexample"> % strip -e libstdc++.so.4 libstdc++.so.5
720 <p>Archive the runtime-only shared object in the GCC 3.4
721 <samp><span class="file">libstdc++.a</span></samp> archive:
722 <pre class="smallexample"> % ar -q libstdc++.a libstdc++.so.4 libstdc++.so.5
724 <p>Linking executables and shared libraries may produce warnings of
725 duplicate symbols. The assembly files generated by GCC for AIX always
726 have included multiple symbol definitions for certain global variable
727 and function declarations in the original program. The warnings should
728 not prevent the linker from producing a correct library or runnable
731 <p>AIX 4.3 utilizes a “large format” archive to support both 32-bit and
732 64-bit object modules. The routines provided in AIX 4.3.0 and AIX 4.3.1
733 to parse archive libraries did not handle the new format correctly.
734 These routines are used by GCC and result in error messages during
735 linking such as “not a COFF file”. The version of the routines shipped
736 with AIX 4.3.1 should work for a 32-bit environment. The <samp><span class="option">-g</span></samp>
737 option of the archive command may be used to create archives of 32-bit
738 objects using the original “small format”. A correct version of the
739 routines is shipped with AIX 4.3.2 and above.
741 <p>Some versions of the AIX binder (linker) can fail with a relocation
742 overflow severe error when the <samp><span class="option">-bbigtoc</span></samp> option is used to link
743 GCC-produced object files into an executable that overflows the TOC. A fix
744 for APAR IX75823 (OVERFLOW DURING LINK WHEN USING GCC AND -BBIGTOC) is
745 available from IBM Customer Support and from its
746 <a href="http://techsupport.services.ibm.com/">techsupport.services.ibm.com</a>
747 website as PTF U455193.
749 <p>The AIX 4.3.2.1 linker (bos.rte.bind_cmds Level 4.3.2.1) will dump core
750 with a segmentation fault when invoked by any version of GCC. A fix for
751 APAR IX87327 is available from IBM Customer Support and from its
752 <a href="http://techsupport.services.ibm.com/">techsupport.services.ibm.com</a>
753 website as PTF U461879. This fix is incorporated in AIX 4.3.3 and above.
755 <p>The initial assembler shipped with AIX 4.3.0 generates incorrect object
756 files. A fix for APAR IX74254 (64BIT DISASSEMBLED OUTPUT FROM COMPILER FAILS
757 TO ASSEMBLE/BIND) is available from IBM Customer Support and from its
758 <a href="http://techsupport.services.ibm.com/">techsupport.services.ibm.com</a>
759 website as PTF U453956. This fix is incorporated in AIX 4.3.1 and above.
761 <p>AIX provides National Language Support (NLS). Compilers and assemblers
762 use NLS to support locale-specific representations of various data
763 formats including floating-point numbers (e.g., ‘<samp><span class="samp">.</span></samp>’ vs ‘<samp><span class="samp">,</span></samp>’ for
764 separating decimal fractions). There have been problems reported where
765 GCC does not produce the same floating-point formats that the assembler
766 expects. If one encounters this problem, set the <samp><span class="env">LANG</span></samp>
767 environment variable to ‘<samp><span class="samp">C</span></samp>’ or ‘<samp><span class="samp">En_US</span></samp>’.
769 <p>A default can be specified with the <samp><span class="option">-mcpu=</span><var>cpu_type</var></samp>
770 switch and using the configure option <samp><span class="option">--with-cpu-</span><var>cpu_type</var></samp>.
774 <h3 class="heading"><a name="TOC22"></a><a name="iq2000_002dx_002delf"></a>iq2000-*-elf</h3>
776 <p>Vitesse IQ2000 processors. These are used in embedded
777 applications. There are no standard Unix configurations.
781 <h3 class="heading"><a name="TOC23"></a><a name="lm32_002dx_002delf"></a>lm32-*-elf</h3>
783 <p>Lattice Mico32 processor.
784 This configuration is intended for embedded systems.
788 <h3 class="heading"><a name="TOC24"></a><a name="lm32_002dx_002duclinux"></a>lm32-*-uclinux</h3>
790 <p>Lattice Mico32 processor.
791 This configuration is intended for embedded systems running uClinux.
795 <h3 class="heading"><a name="TOC25"></a><a name="m32c_002dx_002delf"></a>m32c-*-elf</h3>
797 <p>Renesas M32C processor.
798 This configuration is intended for embedded systems.
802 <h3 class="heading"><a name="TOC26"></a><a name="m32r_002dx_002delf"></a>m32r-*-elf</h3>
804 <p>Renesas M32R processor.
805 This configuration is intended for embedded systems.
809 <h3 class="heading"><a name="TOC27"></a><a name="m68k_002dx_002dx"></a>m68k-*-*</h3>
812 ‘<samp><span class="samp">m68k-*-elf*</span></samp>’, ‘<samp><span class="samp">m68k-*-rtems</span></samp>’, ‘<samp><span class="samp">m68k-*-uclinux</span></samp>’ and
813 ‘<samp><span class="samp">m68k-*-linux</span></samp>’
814 build libraries for both M680x0 and ColdFire processors. If you only
815 need the M680x0 libraries, you can omit the ColdFire ones by passing
816 <samp><span class="option">--with-arch=m68k</span></samp> to <samp><span class="command">configure</span></samp>. Alternatively, you
817 can omit the M680x0 libraries by passing <samp><span class="option">--with-arch=cf</span></samp> to
818 <samp><span class="command">configure</span></samp>. These targets default to 5206 or 5475 code as
819 appropriate for the target system when
820 configured with <samp><span class="option">--with-arch=cf</span></samp> and 68020 code otherwise.
822 <p>The ‘<samp><span class="samp">m68k-*-netbsd</span></samp>’ and
823 ‘<samp><span class="samp">m68k-*-openbsd</span></samp>’ targets also support the <samp><span class="option">--with-arch</span></samp>
824 option. They will generate ColdFire CFV4e code when configured with
825 <samp><span class="option">--with-arch=cf</span></samp> and 68020 code otherwise.
827 <p>You can override the default processors listed above by configuring
828 with <samp><span class="option">--with-cpu=</span><var>target</var></samp>. This <var>target</var> can either
829 be a <samp><span class="option">-mcpu</span></samp> argument or one of the following values:
830 ‘<samp><span class="samp">m68000</span></samp>’, ‘<samp><span class="samp">m68010</span></samp>’, ‘<samp><span class="samp">m68020</span></samp>’, ‘<samp><span class="samp">m68030</span></samp>’,
831 ‘<samp><span class="samp">m68040</span></samp>’, ‘<samp><span class="samp">m68060</span></samp>’, ‘<samp><span class="samp">m68020-40</span></samp>’ and ‘<samp><span class="samp">m68020-60</span></samp>’.
833 <p>GCC requires at least binutils version 2.17 on these targets.
837 <h3 class="heading"><a name="TOC28"></a><a name="m68k_002dx_002duclinux"></a>m68k-*-uclinux</h3>
839 <p>GCC 4.3 changed the uClinux configuration so that it uses the
840 ‘<samp><span class="samp">m68k-linux-gnu</span></samp>’ ABI rather than the ‘<samp><span class="samp">m68k-elf</span></samp>’ ABI.
841 It also added improved support for C++ and flat shared libraries,
842 both of which were ABI changes.
846 <h3 class="heading"><a name="TOC29"></a><a name="mep_002dx_002delf"></a>mep-*-elf</h3>
848 <p>Toshiba Media embedded Processor.
849 This configuration is intended for embedded systems.
853 <h3 class="heading"><a name="TOC30"></a><a name="microblaze_002dx_002delf"></a>microblaze-*-elf</h3>
855 <p>Xilinx MicroBlaze processor.
856 This configuration is intended for embedded systems.
860 <h3 class="heading"><a name="TOC31"></a><a name="mips_002dx_002dx"></a>mips-*-*</h3>
862 <p>If on a MIPS system you get an error message saying “does not have gp
863 sections for all it's [sic] sectons [sic]”, don't worry about it. This
864 happens whenever you use GAS with the MIPS linker, but there is not
865 really anything wrong, and it is okay to use the output file. You can
866 stop such warnings by installing the GNU linker.
868 <p>It would be nice to extend GAS to produce the gp tables, but they are
869 optional, and there should not be a warning about their absence.
871 <p>The libstdc++ atomic locking routines for MIPS targets requires MIPS II
872 and later. A patch went in just after the GCC 3.3 release to
873 make ‘<samp><span class="samp">mips*-*-*</span></samp>’ use the generic implementation instead. You can also
874 configure for ‘<samp><span class="samp">mipsel-elf</span></samp>’ as a workaround. The
875 ‘<samp><span class="samp">mips*-*-linux*</span></samp>’ target continues to use the MIPS II routines. More
876 work on this is expected in future releases.
878 <!-- If you make -with-llsc the default for another target, please also -->
879 <!-- update the description of the -with-llsc option. -->
880 <p>The built-in <code>__sync_*</code> functions are available on MIPS II and
881 later systems and others that support the ‘<samp><span class="samp">ll</span></samp>’, ‘<samp><span class="samp">sc</span></samp>’ and
882 ‘<samp><span class="samp">sync</span></samp>’ instructions. This can be overridden by passing
883 <samp><span class="option">--with-llsc</span></samp> or <samp><span class="option">--without-llsc</span></samp> when configuring GCC.
884 Since the Linux kernel emulates these instructions if they are
885 missing, the default for ‘<samp><span class="samp">mips*-*-linux*</span></samp>’ targets is
886 <samp><span class="option">--with-llsc</span></samp>. The <samp><span class="option">--with-llsc</span></samp> and
887 <samp><span class="option">--without-llsc</span></samp> configure options may be overridden at compile
888 time by passing the <samp><span class="option">-mllsc</span></samp> or <samp><span class="option">-mno-llsc</span></samp> options to
891 <p>MIPS systems check for division by zero (unless
892 <samp><span class="option">-mno-check-zero-division</span></samp> is passed to the compiler) by
893 generating either a conditional trap or a break instruction. Using
894 trap results in smaller code, but is only supported on MIPS II and
895 later. Also, some versions of the Linux kernel have a bug that
896 prevents trap from generating the proper signal (<code>SIGFPE</code>). To enable
897 the use of break, use the <samp><span class="option">--with-divide=breaks</span></samp>
898 <samp><span class="command">configure</span></samp> option when configuring GCC. The default is to
899 use traps on systems that support them.
901 <p>The assembler from GNU binutils 2.17 and earlier has a bug in the way
902 it sorts relocations for REL targets (o32, o64, EABI). This can cause
903 bad code to be generated for simple C++ programs. Also the linker
904 from GNU binutils versions prior to 2.17 has a bug which causes the
905 runtime linker stubs in very large programs, like <samp><span class="file">libgcj.so</span></samp>, to
906 be incorrectly generated. GNU Binutils 2.18 and later (and snapshots
907 made after Nov. 9, 2006) should be free from both of these problems.
911 <h3 class="heading"><a name="TOC32"></a><a name="mips_002dsgi_002dirix5"></a>mips-sgi-irix5</h3>
913 <p>Support for IRIX 5 has been removed in GCC 4.6.
917 <h3 class="heading"><a name="TOC33"></a><a name="mips_002dsgi_002dirix6"></a>mips-sgi-irix6</h3>
919 <p>Support for IRIX 6.5 has been obsoleted in GCC 4.7, but can still be
920 enabled by configuring with <samp><span class="option">--enable-obsolete</span></samp>. Support will be
921 removed in GCC 4.8. Support for IRIX 6 releases before 6.5 has been
922 removed in GCC 4.6, as well as support for the O32 ABI. It is
923 <em>strongly</em> recommended to upgrade to at least IRIX 6.5.18. This
924 release introduced full ISO C99 support, though for the N32 and N64 ABIs
927 <p>To build and use GCC on IRIX 6.5, you need the IRIX Development Foundation
928 (IDF) and IRIX Development Libraries (IDL). They are included with the
931 <p>If you are using SGI's MIPSpro <samp><span class="command">cc</span></samp> as your bootstrap compiler, you must
932 ensure that the N32 ABI is in use. To test this, compile a simple C
933 file with <samp><span class="command">cc</span></samp> and then run <samp><span class="command">file</span></samp> on the
934 resulting object file. The output should look like:
936 <pre class="smallexample"> test.o: ELF N32 MSB ...
938 <p class="noindent">If you see:
940 <pre class="smallexample"> test.o: ELF 32-bit MSB ...
942 <p class="noindent">or
944 <pre class="smallexample"> test.o: ELF 64-bit MSB ...
946 <p class="noindent">then your version of <samp><span class="command">cc</span></samp> uses the O32 or N64 ABI by default. You
947 should set the environment variable <samp><span class="env">CC</span></samp> to ‘<samp><span class="samp">cc -n32</span></samp>’
948 before configuring GCC.
950 <p>If you want the resulting <samp><span class="command">gcc</span></samp> to run on old 32-bit systems
951 with the MIPS R4400 CPU, you need to ensure that only code for the ‘<samp><span class="samp">mips3</span></samp>’
952 instruction set architecture (ISA) is generated. While GCC 3.x does
953 this correctly, both GCC 2.95 and SGI's MIPSpro <samp><span class="command">cc</span></samp> may change
954 the ISA depending on the machine where GCC is built. Using one of them
955 as the bootstrap compiler may result in ‘<samp><span class="samp">mips4</span></samp>’ code, which won't run at
956 all on ‘<samp><span class="samp">mips3</span></samp>’-only systems. For the test program above, you should see:
958 <pre class="smallexample"> test.o: ELF N32 MSB mips-3 ...
960 <p class="noindent">If you get:
962 <pre class="smallexample"> test.o: ELF N32 MSB mips-4 ...
964 <p class="noindent">instead, you should set the environment variable <samp><span class="env">CC</span></samp> to ‘<samp><span class="samp">cc
965 -n32 -mips3</span></samp>’ or ‘<samp><span class="samp">gcc -mips3</span></samp>’ respectively before configuring GCC.
967 <p>MIPSpro C 7.4 may cause bootstrap failures, due to a bug when inlining
968 <code>memcmp</code>. Either add <code>-U__INLINE_INTRINSICS</code> to the <samp><span class="env">CC</span></samp>
969 environment variable as a workaround or upgrade to MIPSpro C 7.4.1m.
971 <p>GCC on IRIX 6.5 is usually built to support the N32 and N64 ABIs. If
972 you build GCC on a system that doesn't have the N64 libraries installed
973 or cannot run 64-bit binaries,
974 you need to configure with <samp><span class="option">--disable-multilib</span></samp> so GCC doesn't
976 Look for <samp><span class="file">/usr/lib64/libc.so.1</span></samp> to see if you
977 have the 64-bit libraries installed.
979 <p>GCC must be configured with GNU <samp><span class="command">as</span></samp>. The latest version, from GNU
980 binutils 2.22, is known to work. On the other hand, bootstrap fails
981 with GNU <samp><span class="command">ld</span></samp> at least since GNU binutils 2.17.
983 <p>The <samp><span class="option">--enable-libgcj</span></samp>
984 option is disabled by default: IRIX 6 uses a very low default limit
985 (20480) for the command line length. Although <samp><span class="command">libtool</span></samp> contains a
986 workaround for this problem, at least the N64 ‘<samp><span class="samp">libgcj</span></samp>’ is known not
987 to build despite this, running into an internal error of the native
988 <samp><span class="command">ld</span></samp>. A sure fix is to increase this limit (‘<samp><span class="samp">ncargs</span></samp>’) to
989 its maximum of 262144 bytes. If you have root access, you can use the
990 <samp><span class="command">systune</span></samp> command to do this.
991 <!-- FIXME: does this work with current libtool? -->
993 <p><code>wchar_t</code> support in ‘<samp><span class="samp">libstdc++</span></samp>’ is not available for old
994 IRIX 6.5.x releases, x < 19. The problem cannot be autodetected
995 and in order to build GCC for such targets you need to configure with
996 <samp><span class="option">--disable-wchar_t</span></samp>.
1000 <h3 class="heading"><a name="TOC34"></a><a name="moxie_002dx_002delf"></a>moxie-*-elf</h3>
1002 <p>The moxie processor. See <a href="http://moxielogic.org/">http://moxielogic.org/</a> for more
1003 information about this processor.
1007 <h3 class="heading"><a name="TOC35"></a><a name="powerpc_002dx_002dx"></a>powerpc-*-*</h3>
1009 <p>You can specify a default version for the <samp><span class="option">-mcpu=</span><var>cpu_type</var></samp>
1010 switch by using the configure option <samp><span class="option">--with-cpu-</span><var>cpu_type</var></samp>.
1013 <a href="ftp://ftp.kernel.org/pub/linux/devel/binutils">binutils 2.15</a>
1014 or newer for a working GCC.
1018 <h3 class="heading"><a name="TOC36"></a><a name="powerpc_002dx_002ddarwin"></a>powerpc-*-darwin*</h3>
1020 <p>PowerPC running Darwin (Mac OS X kernel).
1022 <p>Pre-installed versions of Mac OS X may not include any developer tools,
1023 meaning that you will not be able to build GCC from source. Tool
1024 binaries are available at
1025 <a href="http://opensource.apple.com/">http://opensource.apple.com/</a>.
1027 <p>This version of GCC requires at least cctools-590.36. The
1028 cctools-590.36 package referenced from
1029 <a href="http://gcc.gnu.org/ml/gcc/2006-03/msg00507.html">http://gcc.gnu.org/ml/gcc/2006-03/msg00507.html</a> will not work
1030 on systems older than 10.3.9 (aka darwin7.9.0).
1034 <h3 class="heading"><a name="TOC37"></a><a name="powerpc_002dx_002delf"></a>powerpc-*-elf</h3>
1036 <p>PowerPC system in big endian mode, running System V.4.
1040 <h3 class="heading"><a name="TOC38"></a><a name="powerpc_002dx_002dlinux_002dgnu"></a>powerpc*-*-linux-gnu*</h3>
1042 <p>PowerPC system in big endian mode running Linux.
1046 <h3 class="heading"><a name="TOC39"></a><a name="powerpc_002dx_002dnetbsd"></a>powerpc-*-netbsd*</h3>
1048 <p>PowerPC system in big endian mode running NetBSD.
1052 <h3 class="heading"><a name="TOC40"></a><a name="powerpc_002dx_002deabisim"></a>powerpc-*-eabisim</h3>
1054 <p>Embedded PowerPC system in big endian mode for use in running under the
1059 <h3 class="heading"><a name="TOC41"></a><a name="powerpc_002dx_002deabi"></a>powerpc-*-eabi</h3>
1061 <p>Embedded PowerPC system in big endian mode.
1065 <h3 class="heading"><a name="TOC42"></a><a name="powerpcle_002dx_002delf"></a>powerpcle-*-elf</h3>
1067 <p>PowerPC system in little endian mode, running System V.4.
1071 <h3 class="heading"><a name="TOC43"></a><a name="powerpcle_002dx_002deabisim"></a>powerpcle-*-eabisim</h3>
1073 <p>Embedded PowerPC system in little endian mode for use in running under
1078 <h3 class="heading"><a name="TOC44"></a><a name="powerpcle_002dx_002deabi"></a>powerpcle-*-eabi</h3>
1080 <p>Embedded PowerPC system in little endian mode.
1084 <h3 class="heading"><a name="TOC45"></a><a name="rl78_002dx_002delf"></a>rl78-*-elf</h3>
1086 <p>The Renesas RL78 processor.
1087 This configuration is intended for embedded systems.
1091 <h3 class="heading"><a name="TOC46"></a><a name="rx_002dx_002delf"></a>rx-*-elf</h3>
1093 <p>The Renesas RX processor. See
1094 <a href="http://eu.renesas.com/fmwk.jsp?cnt=rx600_series_landing.jsp&fp=/products/mpumcu/rx_family/rx600_series">http://eu.renesas.com/fmwk.jsp?cnt=rx600_series_landing.jsp&fp=/products/mpumcu/rx_family/rx600_series</a>
1095 for more information about this processor.
1099 <h3 class="heading"><a name="TOC47"></a><a name="s390_002dx_002dlinux"></a>s390-*-linux*</h3>
1101 <p>S/390 system running GNU/Linux for S/390.
1105 <h3 class="heading"><a name="TOC48"></a><a name="s390x_002dx_002dlinux"></a>s390x-*-linux*</h3>
1107 <p>zSeries system (64-bit) running GNU/Linux for zSeries.
1111 <h3 class="heading"><a name="TOC49"></a><a name="s390x_002dibm_002dtpf"></a>s390x-ibm-tpf*</h3>
1113 <p>zSeries system (64-bit) running TPF. This platform is
1114 supported as cross-compilation target only.
1116 <p><hr /><!-- Please use Solaris 2 to refer to all release of Solaris, starting -->
1117 <!-- with 2.0 until 2.6, 7, 8, etc. Solaris 1 was a marketing name for -->
1118 <!-- SunOS 4 releases which we don't use to avoid confusion. Solaris -->
1119 <!-- alone is too unspecific and must be avoided. -->
1121 <h3 class="heading"><a name="TOC50"></a><a name="x_002dx_002dsolaris2"></a>*-*-solaris2*</h3>
1123 <p>Support for Solaris 8 has been obsoleted in GCC 4.7, but can still be
1124 enabled by configuring with <samp><span class="option">--enable-obsolete</span></samp>. Support will be
1125 removed in GCC 4.8. Support for Solaris 7 has been removed in GCC 4.6.
1127 <p>Sun does not ship a C compiler with Solaris 2 before Solaris 10, though
1128 you can download the Sun Studio compilers for free. In Solaris 10 and
1129 11, GCC 3.4.3 is available as <samp><span class="command">/usr/sfw/bin/gcc</span></samp>. Solaris 11
1130 also provides GCC 4.5.2 as <samp><span class="command">/usr/gcc/4.5/bin/gcc</span></samp>. Alternatively,
1131 you can install a pre-built GCC to bootstrap and install GCC. See the
1132 <a href="binaries.html">binaries page</a> for details.
1134 <p>The Solaris 2 <samp><span class="command">/bin/sh</span></samp> will often fail to configure
1135 ‘<samp><span class="samp">libstdc++-v3</span></samp>’, ‘<samp><span class="samp">boehm-gc</span></samp>’ or ‘<samp><span class="samp">libjava</span></samp>’. We therefore
1136 recommend using the following initial sequence of commands
1138 <pre class="smallexample"> % CONFIG_SHELL=/bin/ksh
1139 % export CONFIG_SHELL
1141 <p class="noindent">and proceed as described in <a href="configure.html">the configure instructions</a>.
1142 In addition we strongly recommend specifying an absolute path to invoke
1143 <samp><var>srcdir</var><span class="command">/configure</span></samp>.
1145 <p>Solaris 2 comes with a number of optional OS packages. Some of these
1146 are needed to use GCC fully, namely <code>SUNWarc</code>,
1147 <code>SUNWbtool</code>, <code>SUNWesu</code>, <code>SUNWhea</code>, <code>SUNWlibm</code>,
1148 <code>SUNWsprot</code>, and <code>SUNWtoo</code>. If you did not install all
1149 optional packages when installing Solaris 2, you will need to verify that
1150 the packages that GCC needs are installed.
1152 <p>To check whether an optional package is installed, use
1153 the <samp><span class="command">pkginfo</span></samp> command. To add an optional package, use the
1154 <samp><span class="command">pkgadd</span></samp> command. For further details, see the Solaris 2
1157 <p>Trying to use the linker and other tools in
1158 <samp><span class="file">/usr/ucb</span></samp> to install GCC has been observed to cause trouble.
1159 For example, the linker may hang indefinitely. The fix is to remove
1160 <samp><span class="file">/usr/ucb</span></samp> from your <samp><span class="env">PATH</span></samp>.
1162 <p>The build process works more smoothly with the legacy Sun tools so, if you
1163 have <samp><span class="file">/usr/xpg4/bin</span></samp> in your <samp><span class="env">PATH</span></samp>, we recommend that you place
1164 <samp><span class="file">/usr/bin</span></samp> before <samp><span class="file">/usr/xpg4/bin</span></samp> for the duration of the build.
1166 <p>We recommend the use of the Sun assembler or the GNU assembler, in
1167 conjunction with the Sun linker. The GNU <samp><span class="command">as</span></samp>
1168 versions included in Solaris 10, from GNU binutils 2.15, and Solaris 11,
1169 from GNU binutils 2.19, are known to work. They can be found in
1170 <samp><span class="file">/usr/sfw/bin/gas</span></samp>. Current versions of GNU binutils (2.22)
1171 are known to work as well. Note that your mileage may vary
1172 if you use a combination of the GNU tools and the Sun tools: while the
1173 combination GNU <samp><span class="command">as</span></samp> + Sun <samp><span class="command">ld</span></samp> should reasonably work,
1174 the reverse combination Sun <samp><span class="command">as</span></samp> + GNU <samp><span class="command">ld</span></samp> may fail to
1175 build or cause memory corruption at runtime in some cases for C++ programs.
1176 <!-- FIXME: still? -->
1177 GNU <samp><span class="command">ld</span></samp> usually works as well, although the version included in
1178 Solaris 10 cannot be used due to several bugs. Again, the current
1179 version (2.22) is known to work, but generally lacks platform specific
1180 features, so better stay with Sun <samp><span class="command">ld</span></samp>. To use the LTO linker
1181 plugin (<samp><span class="option">-fuse-linker-plugin</span></samp>) with GNU <samp><span class="command">ld</span></samp>, GNU
1182 binutils <em>must</em> be configured with <samp><span class="option">--enable-largefile</span></samp>.
1184 <p>To enable symbol versioning in ‘<samp><span class="samp">libstdc++</span></samp>’ with Sun <samp><span class="command">ld</span></samp>,
1185 you need to have any version of GNU <samp><span class="command">c++filt</span></samp>, which is part of
1186 GNU binutils. ‘<samp><span class="samp">libstdc++</span></samp>’ symbol versioning will be disabled if no
1187 appropriate version is found. Sun <samp><span class="command">c++filt</span></samp> from the Sun Studio
1188 compilers does <em>not</em> work.
1190 <p>Sun bug 4296832 turns up when compiling X11 headers with GCC 2.95 or
1191 newer: <samp><span class="command">g++</span></samp> will complain that types are missing. These headers
1192 assume that omitting the type means <code>int</code>; this assumption worked for
1193 C90 but is wrong for C++, and is now wrong for C99 also.
1195 <p><samp><span class="command">g++</span></samp> accepts such (invalid) constructs with the option
1196 <samp><span class="option">-fpermissive</span></samp>; it will assume that any missing type is <code>int</code>
1197 (as defined by C90).
1199 <p>There are patches for Solaris 8 (108652-24 or newer for SPARC,
1200 108653-22 for Intel) that fix this bug.
1202 <p>Sun bug 4927647 sometimes causes random spurious testsuite failures
1203 related to missing diagnostic output. This bug doesn't affect GCC
1204 itself, rather it is a kernel bug triggered by the <samp><span class="command">expect</span></samp>
1205 program which is used only by the GCC testsuite driver. When the bug
1206 causes the <samp><span class="command">expect</span></samp> program to miss anticipated output, extra
1207 testsuite failures appear.
1209 <p>There are patches for Solaris 8 (117350-12 or newer for SPARC,
1210 117351-12 or newer for Intel) and Solaris 9 (117171-11 or newer for
1211 SPARC, 117172-11 or newer for Intel) that address this problem.
1213 <p>Solaris 8 provides an alternate implementation of the thread
1214 library ‘<samp><span class="samp">libthread</span></samp>’. It is required for TLS support and has
1215 been made the default in Solaris 9, so it is always used on
1218 <p>Thread-local storage (TLS) is supported in Solaris 8 and 9, but requires
1219 some patches. The ‘<samp><span class="samp">libthread</span></samp>’ patches provide the
1220 <code>__tls_get_addr</code> (SPARC, 64-bit x86) resp. <code>___tls_get_addr</code>
1221 (32-bit x86) functions. On Solaris 8, you need 108993-26 or newer on
1222 SPARC, 108994-26 or newer on Intel. On Solaris 9, the necessary support
1223 on SPARC is present since FCS, while 114432-05 or newer is required on
1224 Intel. Additionally, on Solaris 8, patch 109147-14 or newer on SPARC or
1225 109148-22 or newer on Intel are required for the Sun <samp><span class="command">ld</span></samp> and
1226 runtime linker (<samp><span class="command">ld.so.1</span></samp>) support. Again, Solaris 9/SPARC
1227 works since FCS, while 113986-02 is required on Intel. The linker
1228 patches must be installed even if GNU <samp><span class="command">ld</span></samp> is used. Sun
1229 <samp><span class="command">as</span></samp> in Solaris 8 and 9 doesn't support the necessary
1230 relocations, so GNU <samp><span class="command">as</span></samp> must be used. The <samp><span class="command">configure</span></samp>
1231 script checks for those prerequisites and automatically enables TLS
1232 support if they are met. Although those minimal patch versions should
1233 work, it is recommended to use the latest patch versions which include
1234 additional bug fixes.
1238 <h3 class="heading"><a name="TOC51"></a><a name="sparc_002dx_002dx"></a>sparc*-*-*</h3>
1240 <p>This section contains general configuration information for all
1241 SPARC-based platforms. In addition to reading this section, please
1242 read all other sections that match your target.
1244 <p>Newer versions of the GNU Multiple Precision Library (GMP), the MPFR
1245 library and the MPC library are known to be miscompiled by earlier
1246 versions of GCC on these platforms. We therefore recommend the use
1247 of the exact versions of these libraries listed as minimal versions
1248 in <a href="prerequisites.html">the prerequisites</a>.
1252 <h3 class="heading"><a name="TOC52"></a><a name="sparc_002dsun_002dsolaris2"></a>sparc-sun-solaris2*</h3>
1254 <p>When GCC is configured to use GNU binutils 2.14 or later, the binaries
1255 produced are smaller than the ones produced using Sun's native tools;
1256 this difference is quite significant for binaries containing debugging
1259 <p>Starting with Solaris 7, the operating system is capable of executing
1260 64-bit SPARC V9 binaries. GCC 3.1 and later properly supports
1261 this; the <samp><span class="option">-m64</span></samp> option enables 64-bit code generation.
1262 However, if all you want is code tuned for the UltraSPARC CPU, you
1263 should try the <samp><span class="option">-mtune=ultrasparc</span></samp> option instead, which produces
1264 code that, unlike full 64-bit code, can still run on non-UltraSPARC
1267 <p>When configuring on a Solaris 7 or later system that is running a kernel
1268 that supports only 32-bit binaries, one must configure with
1269 <samp><span class="option">--disable-multilib</span></samp>, since we will not be able to build the
1270 64-bit target libraries.
1272 <p>GCC 3.3 and GCC 3.4 trigger code generation bugs in earlier versions of
1273 the GNU compiler (especially GCC 3.0.x versions), which lead to the
1274 miscompilation of the stage1 compiler and the subsequent failure of the
1275 bootstrap process. A workaround is to use GCC 3.2.3 as an intermediary
1276 stage, i.e. to bootstrap that compiler with the base compiler and then
1277 use it to bootstrap the final compiler.
1279 <p>GCC 3.4 triggers a code generation bug in versions 5.4 (Sun ONE Studio 7)
1280 and 5.5 (Sun ONE Studio 8) of the Sun compiler, which causes a bootstrap
1281 failure in form of a miscompilation of the stage1 compiler by the Sun
1282 compiler. This is Sun bug 4974440. This is fixed with patch 112760-07.
1284 <p>GCC 3.4 changed the default debugging format from Stabs to DWARF-2 for
1285 32-bit code on Solaris 7 and later. If you use the Sun assembler, this
1286 change apparently runs afoul of Sun bug 4910101 (which is referenced as
1287 an x86-only problem by Sun, probably because they do not use DWARF-2).
1288 A symptom of the problem is that you cannot compile C++ programs like
1289 <samp><span class="command">groff</span></samp> 1.19.1 without getting messages similar to the following:
1291 <pre class="smallexample"> ld: warning: relocation error: R_SPARC_UA32: ...
1292 external symbolic relocation against non-allocatable section
1293 .debug_info cannot be processed at runtime: relocation ignored.
1295 <p class="noindent">To work around this problem, compile with <samp><span class="option">-gstabs+</span></samp> instead of
1296 plain <samp><span class="option">-g</span></samp>.
1298 <p>When configuring the GNU Multiple Precision Library (GMP), the MPFR
1299 library or the MPC library on a Solaris 7 or later system, the canonical
1300 target triplet must be specified as the <samp><span class="command">build</span></samp> parameter on the
1301 configure line. This target triplet can be obtained by invoking <samp><span class="command">./config.guess</span></samp> in the toplevel source directory of GCC (and
1302 not that of GMP or MPFR or MPC). For example on a Solaris 9 system:
1304 <pre class="smallexample"> % ./configure --build=sparc-sun-solaris2.9 --prefix=xxx
1308 <h3 class="heading"><a name="TOC53"></a><a name="sparc_002dsun_002dsolaris210"></a>sparc-sun-solaris2.10</h3>
1310 <p>There is a bug in older versions of the Sun assembler which breaks
1311 thread-local storage (TLS). A typical error message is
1313 <pre class="smallexample"> ld: fatal: relocation error: R_SPARC_TLS_LE_HIX22: file /var/tmp//ccamPA1v.o:
1314 symbol <unknown>: bad symbol type SECT: symbol type must be TLS
1316 <p class="noindent">This bug is fixed in Sun patch 118683-03 or later.
1320 <h3 class="heading"><a name="TOC54"></a><a name="sparc_002dx_002dlinux"></a>sparc-*-linux*</h3>
1322 <p>GCC versions 3.0 and higher require binutils 2.11.2 and glibc 2.2.4
1323 or newer on this platform. All earlier binutils and glibc
1324 releases mishandled unaligned relocations on <code>sparc-*-*</code> targets.
1328 <h3 class="heading"><a name="TOC55"></a><a name="sparc64_002dx_002dsolaris2"></a>sparc64-*-solaris2*</h3>
1330 <p>When configuring the GNU Multiple Precision Library (GMP), the MPFR
1331 library or the MPC library, the canonical target triplet must be specified
1332 as the <samp><span class="command">build</span></samp> parameter on the configure line. For example
1333 on a Solaris 9 system:
1335 <pre class="smallexample"> % ./configure --build=sparc64-sun-solaris2.9 --prefix=xxx
1337 <p>The following compiler flags must be specified in the configure
1338 step in order to bootstrap this target with the Sun compiler:
1340 <pre class="smallexample"> % CC="cc -xarch=v9 -xildoff" <var>srcdir</var>/configure [<var>options</var>] [<var>target</var>]
1342 <p class="noindent"><samp><span class="option">-xarch=v9</span></samp> specifies the SPARC-V9 architecture to the Sun toolchain
1343 and <samp><span class="option">-xildoff</span></samp> turns off the incremental linker.
1347 <h3 class="heading"><a name="TOC56"></a><a name="sparcv9_002dx_002dsolaris2"></a>sparcv9-*-solaris2*</h3>
1349 <p>This is a synonym for ‘<samp><span class="samp">sparc64-*-solaris2*</span></samp>’.
1353 <h3 class="heading"><a name="TOC57"></a><a name="c6x_002dx_002dx"></a>c6x-*-*</h3>
1355 <p>The C6X family of processors. This port requires binutils-2.22 or newer.
1359 <h3 class="heading"><a name="TOC58"></a><a name="tilegx_002d_002a_002dlinux"></a>tilegx-*-linux*</h3>
1361 <p>The TILE-Gx processor running GNU/Linux. This port requires
1362 binutils-2.22 or newer.
1366 <h3 class="heading"><a name="TOC59"></a><a name="tilepro_002d_002a_002dlinux"></a>tilepro-*-linux*</h3>
1368 <p>The TILEPro processor running GNU/Linux. This port requires
1369 binutils-2.22 or newer.
1373 <h3 class="heading"><a name="TOC60"></a><a name="x_002dx_002dvxworks"></a>*-*-vxworks*</h3>
1375 <p>Support for VxWorks is in flux. At present GCC supports <em>only</em> the
1376 very recent VxWorks 5.5 (aka Tornado 2.2) release, and only on PowerPC.
1377 We welcome patches for other architectures supported by VxWorks 5.5.
1378 Support for VxWorks AE would also be welcome; we believe this is merely
1379 a matter of writing an appropriate “configlette” (see below). We are
1380 not interested in supporting older, a.out or COFF-based, versions of
1383 <p>VxWorks comes with an older version of GCC installed in
1384 <samp><var>$WIND_BASE</var><span class="file">/host</span></samp>; we recommend you do not overwrite it.
1385 Choose an installation <var>prefix</var> entirely outside <var>$WIND_BASE</var>.
1386 Before running <samp><span class="command">configure</span></samp>, create the directories <samp><var>prefix</var></samp>
1387 and <samp><var>prefix</var><span class="file">/bin</span></samp>. Link or copy the appropriate assembler,
1388 linker, etc. into <samp><var>prefix</var><span class="file">/bin</span></samp>, and set your <var>PATH</var> to
1389 include that directory while running both <samp><span class="command">configure</span></samp> and
1390 <samp><span class="command">make</span></samp>.
1392 <p>You must give <samp><span class="command">configure</span></samp> the
1393 <samp><span class="option">--with-headers=</span><var>$WIND_BASE</var><span class="option">/target/h</span></samp> switch so that it can
1394 find the VxWorks system headers. Since VxWorks is a cross compilation
1395 target only, you must also specify <samp><span class="option">--target=</span><var>target</var></samp>.
1396 <samp><span class="command">configure</span></samp> will attempt to create the directory
1397 <samp><var>prefix</var><span class="file">/</span><var>target</var><span class="file">/sys-include</span></samp> and copy files into it;
1398 make sure the user running <samp><span class="command">configure</span></samp> has sufficient privilege
1401 <p>GCC's exception handling runtime requires a special “configlette”
1402 module, <samp><span class="file">contrib/gthr_supp_vxw_5x.c</span></samp>. Follow the instructions in
1403 that file to add the module to your kernel build. (Future versions of
1404 VxWorks will incorporate this module.)
1408 <h3 class="heading"><a name="TOC61"></a><a name="x86_002d64_002dx_002dx"></a>x86_64-*-*, amd64-*-*</h3>
1410 <p>GCC supports the x86-64 architecture implemented by the AMD64 processor
1411 (amd64-*-* is an alias for x86_64-*-*) on GNU/Linux, FreeBSD and NetBSD.
1412 On GNU/Linux the default is a bi-arch compiler which is able to generate
1413 both 64-bit x86-64 and 32-bit x86 code (via the <samp><span class="option">-m32</span></samp> switch).
1417 <h3 class="heading"><a name="TOC62"></a><a name="x86_002d64_002dx_002dsolaris210"></a>x86_64-*-solaris2.1[0-9]*</h3>
1419 <p>GCC also supports the x86-64 architecture implemented by the AMD64
1420 processor (‘<samp><span class="samp">amd64-*-*</span></samp>’ is an alias for ‘<samp><span class="samp">x86_64-*-*</span></samp>’) on
1421 Solaris 10 or later. Unlike other systems, without special options a
1422 bi-arch compiler is built which generates 32-bit code by default, but
1423 can generate 64-bit x86-64 code with the <samp><span class="option">-m64</span></samp> switch. Since
1424 GCC 4.7, there is also configuration that defaults to 64-bit code, but
1425 can generate 32-bit code with <samp><span class="option">-m32</span></samp>. To configure and build
1426 this way, you have to provide all support libraries like <samp><span class="file">libgmp</span></samp>
1427 as 64-bit code, configure with <samp><span class="option">--target=x86_64-pc-solaris2.1x</span></samp>
1428 and ‘<samp><span class="samp">CC=gcc -m64</span></samp>’.
1432 <h3 class="heading"><a name="TOC63"></a><a name="xtensa_002dx_002delf"></a>xtensa*-*-elf</h3>
1434 <p>This target is intended for embedded Xtensa systems using the
1435 ‘<samp><span class="samp">newlib</span></samp>’ C library. It uses ELF but does not support shared
1436 objects. Designed-defined instructions specified via the
1437 Tensilica Instruction Extension (TIE) language are only supported
1438 through inline assembly.
1440 <p>The Xtensa configuration information must be specified prior to
1441 building GCC. The <samp><span class="file">include/xtensa-config.h</span></samp> header
1442 file contains the configuration information. If you created your
1443 own Xtensa configuration with the Xtensa Processor Generator, the
1444 downloaded files include a customized copy of this header file,
1445 which you can use to replace the default header file.
1449 <h3 class="heading"><a name="TOC64"></a><a name="xtensa_002dx_002dlinux"></a>xtensa*-*-linux*</h3>
1451 <p>This target is for Xtensa systems running GNU/Linux. It supports ELF
1452 shared objects and the GNU C library (glibc). It also generates
1453 position-independent code (PIC) regardless of whether the
1454 <samp><span class="option">-fpic</span></samp> or <samp><span class="option">-fPIC</span></samp> options are used. In other
1455 respects, this target is the same as the
1456 <a href="#xtensa*-*-elf">‘<samp><span class="samp">xtensa*-*-elf</span></samp>’</a> target.
1460 <h3 class="heading"><a name="TOC65"></a><a name="windows"></a>Microsoft Windows</h3>
1462 <h4 class="subheading"><a name="TOC66"></a>Intel 16-bit versions</h4>
1464 <p>The 16-bit versions of Microsoft Windows, such as Windows 3.1, are not
1467 <p>However, the 32-bit port has limited support for Microsoft
1468 Windows 3.11 in the Win32s environment, as a target only. See below.
1470 <h4 class="subheading"><a name="TOC67"></a>Intel 32-bit versions</h4>
1472 <p>The 32-bit versions of Windows, including Windows 95, Windows NT, Windows
1473 XP, and Windows Vista, are supported by several different target
1474 platforms. These targets differ in which Windows subsystem they target
1475 and which C libraries are used.
1478 <li>Cygwin <a href="#x-x-cygwin">*-*-cygwin</a>: Cygwin provides a user-space
1479 Linux API emulation layer in the Win32 subsystem.
1480 <li>Interix <a href="#x-x-interix">*-*-interix</a>: The Interix subsystem
1481 provides native support for POSIX.
1482 <li>MinGW <a href="#x-x-mingw32">*-*-mingw32</a>: MinGW is a native GCC port for
1483 the Win32 subsystem that provides a subset of POSIX.
1484 <li>MKS i386-pc-mks: NuTCracker from MKS. See
1485 <a href="http://www.mkssoftware.com/">http://www.mkssoftware.com/</a> for more information.
1488 <h4 class="subheading"><a name="TOC68"></a>Intel 64-bit versions</h4>
1490 <p>GCC contains support for x86-64 using the mingw-w64
1491 runtime library, available from <a href="http://mingw-w64.sourceforge.net/">http://mingw-w64.sourceforge.net/</a>.
1492 This library should be used with the target triple x86_64-pc-mingw32.
1494 <p>Presently Windows for Itanium is not supported.
1496 <h4 class="subheading"><a name="TOC69"></a>Windows CE</h4>
1498 <p>Windows CE is supported as a target only on Hitachi
1499 SuperH (sh-wince-pe), and MIPS (mips-wince-pe).
1501 <h4 class="subheading"><a name="TOC70"></a>Other Windows Platforms</h4>
1503 <p>GCC no longer supports Windows NT on the Alpha or PowerPC.
1505 <p>GCC no longer supports the Windows POSIX subsystem. However, it does
1506 support the Interix subsystem. See above.
1508 <p>Old target names including *-*-winnt and *-*-windowsnt are no longer used.
1510 <p>PW32 (i386-pc-pw32) support was never completed, and the project seems to
1511 be inactive. See <a href="http://pw32.sourceforge.net/">http://pw32.sourceforge.net/</a> for more information.
1513 <p>UWIN support has been removed due to a lack of maintenance.
1517 <h3 class="heading"><a name="TOC71"></a><a name="x_002dx_002dcygwin"></a>*-*-cygwin</h3>
1519 <p>Ports of GCC are included with the
1520 <a href="http://www.cygwin.com/">Cygwin environment</a>.
1522 <p>GCC will build under Cygwin without modification; it does not build
1523 with Microsoft's C++ compiler and there are no plans to make it do so.
1525 <p>The Cygwin native compiler can be configured to target any 32-bit x86
1526 cpu architecture desired; the default is i686-pc-cygwin. It should be
1527 used with as up-to-date a version of binutils as possible; use either
1528 the latest official GNU binutils release in the Cygwin distribution,
1529 or version 2.20 or above if building your own.
1533 <h3 class="heading"><a name="TOC72"></a><a name="x_002dx_002dinterix"></a>*-*-interix</h3>
1535 <p>The Interix target is used by OpenNT, Interix, Services For UNIX (SFU),
1536 and Subsystem for UNIX-based Applications (SUA). Applications compiled
1537 with this target run in the Interix subsystem, which is separate from
1538 the Win32 subsystem. This target was last known to work in GCC 3.3.
1542 <h3 class="heading"><a name="TOC73"></a><a name="x_002dx_002dmingw32"></a>*-*-mingw32</h3>
1544 <p>GCC will build with and support only MinGW runtime 3.12 and later.
1545 Earlier versions of headers are incompatible with the new default semantics
1546 of <code>extern inline</code> in <code>-std=c99</code> and <code>-std=gnu99</code> modes.
1550 <h3 class="heading"><a name="TOC74"></a><a name="older"></a>Older systems</h3>
1552 <p>GCC contains support files for many older (1980s and early
1553 1990s) Unix variants. For the most part, support for these systems
1554 has not been deliberately removed, but it has not been maintained for
1555 several years and may suffer from bitrot.
1557 <p>Starting with GCC 3.1, each release has a list of “obsoleted” systems.
1558 Support for these systems is still present in that release, but
1559 <samp><span class="command">configure</span></samp> will fail unless the <samp><span class="option">--enable-obsolete</span></samp>
1560 option is given. Unless a maintainer steps forward, support for these
1561 systems will be removed from the next release of GCC.
1563 <p>Support for old systems as hosts for GCC can cause problems if the
1564 workarounds for compiler, library and operating system bugs affect the
1565 cleanliness or maintainability of the rest of GCC. In some cases, to
1566 bring GCC up on such a system, if still possible with current GCC, may
1567 require first installing an old version of GCC which did work on that
1568 system, and using it to compile a more recent GCC, to avoid bugs in the
1569 vendor compiler. Old releases of GCC 1 and GCC 2 are available in the
1570 <samp><span class="file">old-releases</span></samp> directory on the <a href="../mirrors.html">GCC mirror sites</a>. Header bugs may generally be avoided using
1571 <samp><span class="command">fixincludes</span></samp>, but bugs or deficiencies in libraries and the
1572 operating system may still cause problems.
1574 <p>Support for older systems as targets for cross-compilation is less
1575 problematic than support for them as hosts for GCC; if an enthusiast
1576 wishes to make such a target work again (including resurrecting any of
1577 the targets that never worked with GCC 2, starting from the last
1578 version before they were removed), patches
1579 <a href="../contribute.html">following the usual requirements</a> would be
1580 likely to be accepted, since they should not affect the support for more
1583 <p>For some systems, old versions of GNU binutils may also be useful,
1584 and are available from <samp><span class="file">pub/binutils/old-releases</span></samp> on
1585 <a href="http://sourceware.org/mirrors.html">sourceware.org mirror sites</a>.
1587 <p>Some of the information on specific systems above relates to
1588 such older systems, but much of the information
1589 about GCC on such systems (which may no longer be applicable to
1590 current GCC) is to be found in the GCC texinfo manual.
1594 <h3 class="heading"><a name="TOC75"></a><a name="elf"></a>all ELF targets (SVR4, Solaris 2, etc.)</h3>
1596 <p>C++ support is significantly better on ELF targets if you use the
1597 <a href="./configure.html#with-gnu-ld">GNU linker</a>; duplicate copies of
1598 inlines, vtables and template instantiations will be discarded
1602 <p><a href="./index.html">Return to the GCC Installation page</a>
1604 <!-- ***Old documentation****************************************************** -->
1605 <!-- ***GFDL******************************************************************** -->
1606 <!-- *************************************************************************** -->
1607 <!-- Part 6 The End of the Document -->