]>
Commit | Line | Data |
---|---|---|
c63539ff ML |
1 | .. |
2 | Copyright 1988-2022 Free Software Foundation, Inc. | |
3 | This is part of the GCC manual. | |
4 | For copying conditions, see the copyright.rst file. | |
5 | ||
6 | .. index:: Specific, Specific installation notes, Target specific installation, Host specific installation, Target specific installation notes | |
7 | ||
8 | .. _specific: | |
9 | ||
10 | Host/target specific installation notes for GCC | |
11 | ----------------------------------------------- | |
12 | ||
13 | Please read this document carefully *before* installing the | |
14 | GNU Compiler Collection on your machine. | |
15 | ||
16 | Note that this list of install notes is *not* a list of supported | |
17 | hosts or targets. Not all supported hosts and targets are listed | |
18 | here, only the ones that require host-specific or target-specific | |
19 | information have to. | |
20 | ||
21 | aarch64\*-\*-\* | |
22 | =============== | |
23 | ||
24 | Binutils pre 2.24 does not have support for selecting :option:`-mabi` and | |
25 | does not support ILP32. If it is used to build GCC 4.9 or later, GCC will | |
26 | not support option :option:`-mabi=ilp32`. | |
27 | ||
28 | To enable a workaround for the Cortex-A53 erratum number 835769 by default | |
29 | (for all CPUs regardless of -mcpu option given) at configure time use the | |
30 | :option:`--enable-fix-cortex-a53-835769` option. This will enable the fix by | |
31 | default and can be explicitly disabled during compilation by passing the | |
32 | :option:`-mno-fix-cortex-a53-835769` option. Conversely, | |
33 | :option:`--disable-fix-cortex-a53-835769` will disable the workaround by | |
34 | default. The workaround is disabled by default if neither of | |
35 | :option:`--enable-fix-cortex-a53-835769` or | |
36 | :option:`--disable-fix-cortex-a53-835769` is given at configure time. | |
37 | ||
38 | To enable a workaround for the Cortex-A53 erratum number 843419 by default | |
39 | (for all CPUs regardless of -mcpu option given) at configure time use the | |
40 | :option:`--enable-fix-cortex-a53-843419` option. This workaround is applied at | |
41 | link time. Enabling the workaround will cause GCC to pass the relevant option | |
42 | to the linker. It can be explicitly disabled during compilation by passing the | |
43 | :option:`-mno-fix-cortex-a53-843419` option. Conversely, | |
44 | :option:`--disable-fix-cortex-a53-843419` will disable the workaround by default. | |
45 | The workaround is disabled by default if neither of | |
46 | :option:`--enable-fix-cortex-a53-843419` or | |
47 | :option:`--disable-fix-cortex-a53-843419` is given at configure time. | |
48 | ||
49 | To enable Branch Target Identification Mechanism and Return Address Signing by | |
50 | default at configure time use the :option:`--enable-standard-branch-protection` | |
51 | option. This is equivalent to having :option:`-mbranch-protection=standard` | |
52 | during compilation. This can be explicitly disabled during compilation by | |
53 | passing the :option:`-mbranch-protection=none` option which turns off all | |
54 | types of branch protections. Conversely, | |
55 | :option:`--disable-standard-branch-protection` will disable both the | |
56 | protections by default. This mechanism is turned off by default if neither | |
57 | of the options are given at configure time. | |
58 | ||
59 | alpha\*-\*-\* | |
60 | ============= | |
61 | ||
62 | This section contains general configuration information for all | |
63 | Alpha-based platforms using ELF. In addition to reading this | |
64 | section, please read all other sections that match your target. | |
65 | ||
66 | amd64-\*-solaris2\* | |
67 | =================== | |
68 | ||
69 | This is a synonym for :samp:`x86_64-*-solaris2*`. | |
70 | ||
71 | amdgcn-\*-amdhsa | |
72 | ================ | |
73 | ||
74 | AMD GCN GPU target. | |
75 | ||
76 | Instead of GNU Binutils, you will need to install LLVM 13.0.1, or later, and copy | |
77 | :samp:`bin/llvm-mc` to :samp:`amdgcn-amdhsa/bin/as`, | |
78 | :samp:`bin/lld` to :samp:`amdgcn-amdhsa/bin/ld`, | |
79 | :samp:`bin/llvm-nm` to :samp:`amdgcn-amdhsa/bin/nm`, and | |
80 | :samp:`bin/llvm-ar` to both :samp:`bin/amdgcn-amdhsa-ar` and | |
81 | :samp:`bin/amdgcn-amdhsa-ranlib`. | |
82 | ||
83 | Use Newlib (3.2.0, or newer). | |
84 | ||
85 | To run the binaries, install the HSA Runtime from the | |
86 | `ROCm Platform <https://rocm.github.io>`_, and use | |
87 | :samp:`libexec/gcc/amdhsa-amdhsa/{version}/gcn-run` to launch them | |
88 | on the GPU. | |
89 | ||
90 | arc-\*-elf32 | |
91 | ============ | |
92 | ||
93 | Use :samp:`configure --target=arc-elf32 --with-cpu={cpu} --enable-languages="c,c++"` | |
94 | to configure GCC, with :samp:`{cpu}` being one of :samp:`arc600`, :samp:`arc601`, | |
95 | or :samp:`arc700`. | |
96 | ||
97 | arc-linux-uclibc | |
98 | ================ | |
99 | ||
100 | Use :samp:`configure --target=arc-linux-uclibc --with-cpu=arc700 --enable-languages="c,c++"` to configure GCC. | |
101 | ||
102 | arm-\*-eabi | |
103 | =========== | |
104 | ||
105 | ARM-family processors. | |
106 | ||
107 | Building the Ada frontend commonly fails (an infinite loop executing | |
108 | ``xsinfo``) if the host compiler is GNAT 4.8. Host compilers built from the | |
109 | GNAT 4.6, 4.9 or 5 release branches are known to succeed. | |
110 | ||
111 | avr | |
112 | === | |
113 | ||
114 | ATMEL AVR-family micro controllers. These are used in embedded | |
115 | applications. There are no standard Unix configurations. | |
116 | See :ref:`gcc:avr-options`, | |
117 | for the list of supported MCU types. | |
118 | ||
119 | Use :samp:`configure --target=avr --enable-languages="c"` to configure GCC. | |
120 | ||
121 | Further installation notes and other useful information about AVR tools | |
122 | can also be obtained from: | |
123 | ||
124 | * `http://www.nongnu.org/avr/ <http://www.nongnu.org/avr/>`_ | |
125 | ||
126 | * `http://www.amelek.gda.pl/avr/ <http://www.amelek.gda.pl/avr/>`_ | |
127 | ||
128 | The following error: | |
129 | ||
130 | .. code-block:: bash | |
131 | ||
132 | Error: register required | |
133 | ||
134 | indicates that you should upgrade to a newer version of the binutils. | |
135 | ||
136 | Blackfin | |
137 | ======== | |
138 | ||
139 | The Blackfin processor, an Analog Devices DSP. | |
140 | See :ref:`gcc:blackfin-options`, | |
141 | ||
142 | More information, and a version of binutils with support for this processor, | |
143 | are available at https://sourceforge.net/projects/adi-toolchain/. | |
144 | ||
145 | CRIS | |
146 | ==== | |
147 | ||
148 | CRIS is a CPU architecture in Axis Communications systems-on-a-chip, for | |
149 | example the ETRAX series. These are used in embedded applications. | |
150 | ||
151 | See :ref:`gcc:cris-options`, | |
152 | for a list of CRIS-specific options. | |
153 | ||
154 | Use :samp:`configure --target=cris-elf` to configure GCCfor building | |
155 | a cross-compiler for CRIS. | |
156 | ||
157 | DOS | |
158 | === | |
159 | ||
160 | Please have a look at the `binaries page <https://gcc.gnu.org/install/binaries.html>`_. | |
161 | ||
162 | You cannot install GCC by itself on MSDOS; it will not compile under | |
163 | any MSDOS compiler except itself. You need to get the complete | |
164 | compilation package DJGPP, which includes binaries as well as sources, | |
165 | and includes all the necessary compilation tools and libraries. | |
166 | ||
167 | epiphany-\*-elf | |
168 | =============== | |
169 | ||
170 | Adapteva Epiphany. | |
171 | This configuration is intended for embedded systems. | |
172 | ||
173 | \*-\*-freebsd\* | |
174 | =============== | |
175 | ||
176 | In order to better utilize FreeBSD base system functionality and match | |
177 | the configuration of the system compiler, GCC 4.5 and above as well as | |
178 | GCC 4.4 past 2010-06-20 leverage SSP support in libc (which is present | |
179 | on FreeBSD 7 or later) and the use of ``__cxa_atexit`` by default | |
180 | (on FreeBSD 6 or later). The use of ``dl_iterate_phdr`` inside | |
181 | :samp:`libgcc_s.so.1` and boehm-gc (on FreeBSD 7 or later) is enabled | |
182 | by GCC 4.5 and above. | |
183 | ||
184 | We support FreeBSD using the ELF file format with DWARF 2 debugging | |
185 | for all CPU architectures. There are | |
186 | no known issues with mixing object files and libraries with different | |
187 | debugging formats. Otherwise, this release of GCC should now match | |
188 | more of the configuration used in the stock FreeBSD configuration of | |
189 | GCC. In particular, :option:`--enable-threads` is now configured by | |
190 | default. However, as a general user, do not attempt to replace the | |
191 | system compiler with this release. Known to bootstrap and check with | |
192 | good results on FreeBSD 7.2-STABLE. In the past, known to bootstrap | |
193 | and check with good results on FreeBSD 3.0, 3.4, 4.0, 4.2, 4.3, 4.4, | |
194 | 4.5, 4.8, 4.9 and 5-CURRENT. | |
195 | ||
196 | The version of binutils installed in :samp:`/usr/bin` probably works | |
197 | with this release of GCC. Bootstrapping against the latest GNU | |
198 | binutils and/or the version found in :samp:`/usr/ports/devel/binutils` has | |
199 | been known to enable additional features and improve overall testsuite | |
200 | results. However, it is currently known that boehm-gc may not configure | |
201 | properly on FreeBSD prior to the FreeBSD 7.0 release with GNU binutils | |
202 | after 2.16.1. | |
203 | ||
204 | ft32-\*-elf | |
205 | =========== | |
206 | ||
207 | The FT32 processor. | |
208 | This configuration is intended for embedded systems. | |
209 | ||
210 | h8300-hms | |
211 | ========= | |
212 | ||
213 | Renesas H8/300 series of processors. | |
214 | ||
215 | Please have a look at the `binaries page <https://gcc.gnu.org/install/binaries.html>`_. | |
216 | ||
217 | The calling convention and structure layout has changed in release 2.6. | |
218 | All code must be recompiled. The calling convention now passes the | |
219 | first three arguments in function calls in registers. Structures are no | |
220 | longer a multiple of 2 bytes. | |
221 | ||
222 | hppa\*-hp-hpux\* | |
223 | ================ | |
224 | ||
225 | Support for HP-UX version 9 and older was discontinued in GCC 3.4. | |
226 | ||
227 | We require using gas/binutils on all hppa platforms. Version 2.19 or | |
228 | later is recommended. | |
229 | ||
230 | It may be helpful to configure GCC with the :option:`--with-gnu-as` and | |
231 | :option:`--with-as=...` options to ensure that GCC can find GAS. | |
232 | ||
233 | The HP assembler should not be used with GCC. It is rarely tested and may | |
234 | not work. It shouldn't be used with any languages other than C due to its | |
235 | many limitations. | |
236 | ||
237 | Specifically, :option:`-g` does not work (HP-UX uses a peculiar debugging | |
238 | format which GCC does not know about). It also inserts timestamps | |
239 | into each object file it creates, causing the 3-stage comparison test to | |
240 | fail during a bootstrap. You should be able to continue by saying | |
241 | :samp:`make all-host all-target` after getting the failure from :samp:`make`. | |
242 | ||
243 | Various GCC features are not supported. For example, it does not support weak | |
244 | symbols or alias definitions. As a result, explicit template instantiations | |
245 | are required when using C++. This makes it difficult if not impossible to | |
246 | build many C++ applications. | |
247 | ||
248 | There are two default scheduling models for instructions. These are | |
249 | PROCESSOR_7100LC and PROCESSOR_8000. They are selected from the pa-risc | |
250 | architecture specified for the target machine when configuring. | |
251 | PROCESSOR_8000 is the default. PROCESSOR_7100LC is selected when | |
252 | the target is a :samp:`hppa1*` machine. | |
253 | ||
254 | The PROCESSOR_8000 model is not well suited to older processors. Thus, | |
255 | it is important to completely specify the machine architecture when | |
256 | configuring if you want a model other than PROCESSOR_8000. The macro | |
257 | TARGET_SCHED_DEFAULT can be defined in BOOT_CFLAGS if a different | |
258 | default scheduling model is desired. | |
259 | ||
260 | As of GCC 4.0, GCC uses the UNIX 95 namespace for HP-UX 10.10 | |
261 | through 11.00, and the UNIX 98 namespace for HP-UX 11.11 and later. | |
262 | This namespace change might cause problems when bootstrapping with | |
263 | an earlier version of GCC or the HP compiler as essentially the same | |
264 | namespace is required for an entire build. This problem can be avoided | |
265 | in a number of ways. With HP cc, :envvar:`UNIX_STD` can be set to :samp:`95` | |
266 | or :samp:`98`. Another way is to add an appropriate set of predefines | |
267 | to :envvar:`CC`. The description for the munix= option contains | |
268 | a list of the predefines used with each standard. | |
269 | ||
270 | More specific information to :samp:`hppa*-hp-hpux*` targets follows. | |
271 | ||
272 | hppa\*-hp-hpux10 | |
273 | ================ | |
274 | ||
275 | For hpux10.20, we *highly* recommend you pick up the latest sed patch | |
276 | ``PHCO_19798`` from HP. | |
277 | ||
278 | The C++ ABI has changed incompatibly in GCC 4.0. COMDAT subspaces are | |
279 | used for one-only code and data. This resolves many of the previous | |
280 | problems in using C++ on this target. However, the ABI is not compatible | |
281 | with the one implemented under HP-UX 11 using secondary definitions. | |
282 | ||
283 | hppa\*-hp-hpux11 | |
284 | ================ | |
285 | ||
286 | GCC 3.0 and up support HP-UX 11. GCC 2.95.x is not supported and cannot | |
287 | be used to compile GCC 3.0 and up. | |
288 | ||
289 | The libffi library haven't been ported to 64-bit HP-UXand doesn't build. | |
290 | ||
291 | Refer to `binaries page <https://gcc.gnu.org/install/binaries.html>`_ for information about obtaining | |
292 | precompiled GCC binaries for HP-UX. Precompiled binaries must be obtained | |
293 | to build the Ada language as it cannot be bootstrapped using C. Ada is | |
294 | only available for the 32-bit PA-RISC runtime. | |
295 | ||
296 | Starting with GCC 3.4 an ISO C compiler is required to bootstrap. The | |
297 | bundled compiler supports only traditional C; you will need either HP's | |
298 | unbundled compiler, or a binary distribution of GCC. | |
299 | ||
300 | It is possible to build GCC 3.3 starting with the bundled HP compiler, | |
301 | but the process requires several steps. GCC 3.3 can then be used to | |
302 | build later versions. | |
303 | ||
304 | There are several possible approaches to building the distribution. | |
305 | Binutils can be built first using the HP tools. Then, the GCC | |
306 | distribution can be built. The second approach is to build GCC | |
307 | first using the HP tools, then build binutils, then rebuild GCC. | |
308 | There have been problems with various binary distributions, so it | |
309 | is best not to start from a binary distribution. | |
310 | ||
311 | On 64-bit capable systems, there are two distinct targets. Different | |
312 | installation prefixes must be used if both are to be installed on | |
313 | the same system. The :samp:`hppa[1-2]*-hp-hpux11*` target generates code | |
314 | for the 32-bit PA-RISC runtime architecture and uses the HP linker. | |
315 | The :samp:`hppa64-hp-hpux11*` target generates 64-bit code for the | |
316 | PA-RISC 2.0 architecture. | |
317 | ||
318 | The script config.guess now selects the target type based on the compiler | |
319 | detected during configuration. You must define :envvar:`PATH` or :envvar:`CC` so | |
320 | that configure finds an appropriate compiler for the initial bootstrap. | |
321 | When :envvar:`CC` is used, the definition should contain the options that are | |
322 | needed whenever :envvar:`CC` is used. | |
323 | ||
324 | Specifically, options that determine the runtime architecture must be | |
325 | in :envvar:`CC` to correctly select the target for the build. It is also | |
326 | convenient to place many other compiler options in :envvar:`CC`. For example, | |
327 | :envvar:`CC="cc -Ac +DA2.0W -Wp,-H16376 -D_CLASSIC_TYPES -D_HPUX_SOURCE"` | |
328 | can be used to bootstrap the GCC 3.3 branch with the HP compiler in | |
329 | 64-bit K&R/bundled mode. The +DA2.0W option will result in | |
330 | the automatic selection of the :samp:`hppa64-hp-hpux11*` target. The | |
331 | macro definition table of cpp needs to be increased for a successful | |
332 | build with the HP compiler. _CLASSIC_TYPES and _HPUX_SOURCE need to | |
333 | be defined when building with the bundled compiler, or when using the | |
334 | :option:`-Ac` option. These defines aren't necessary with :option:`-Ae`. | |
335 | ||
336 | It is best to explicitly configure the :samp:`hppa64-hp-hpux11*` target | |
337 | with the :option:`--with-ld=...` option. This overrides the standard | |
338 | search for ld. The two linkers supported on this target require different | |
339 | commands. The default linker is determined during configuration. As a | |
340 | result, it's not possible to switch linkers in the middle of a GCC build. | |
341 | This has been reported to sometimes occur in unified builds of binutils | |
342 | and GCC. | |
343 | ||
344 | A recent linker patch must be installed for the correct operation of | |
345 | GCC 3.3 and later. ``PHSS_26559`` and ``PHSS_24304`` are the | |
346 | oldest linker patches that are known to work. They are for HP-UX | |
347 | 11.00 and 11.11, respectively. ``PHSS_24303``, the companion to | |
348 | ``PHSS_24304``, might be usable but it hasn't been tested. These | |
349 | patches have been superseded. Consult the HP patch database to obtain | |
350 | the currently recommended linker patch for your system. | |
351 | ||
352 | The patches are necessary for the support of weak symbols on the | |
353 | 32-bit port, and for the running of initializers and finalizers. Weak | |
354 | symbols are implemented using SOM secondary definition symbols. Prior | |
355 | to HP-UX 11, there are bugs in the linker support for secondary symbols. | |
356 | The patches correct a problem of linker core dumps creating shared | |
357 | libraries containing secondary symbols, as well as various other | |
358 | linking issues involving secondary symbols. | |
359 | ||
360 | GCC 3.3 uses the ELF DT_INIT_ARRAY and DT_FINI_ARRAY capabilities to | |
361 | run initializers and finalizers on the 64-bit port. The 32-bit port | |
362 | uses the linker +init and +fini options for the same | |
363 | purpose. The patches correct various problems with the +init/+fini | |
364 | options, including program core dumps. Binutils 2.14 corrects a | |
365 | problem on the 64-bit port resulting from HP's non-standard use of | |
366 | the .init and .fini sections for array initializers and finalizers. | |
367 | ||
368 | Although the HP and GNU linkers are both supported for the | |
369 | :samp:`hppa64-hp-hpux11*` target, it is strongly recommended that the | |
370 | HP linker be used for link editing on this target. | |
371 | ||
372 | At this time, the GNU linker does not support the creation of long | |
373 | branch stubs. As a result, it cannot successfully link binaries | |
374 | containing branch offsets larger than 8 megabytes. In addition, | |
375 | there are problems linking shared libraries, linking executables | |
376 | with :option:`-static`, and with dwarf2 unwind and exception support. | |
377 | It also doesn't provide stubs for internal calls to global functions | |
378 | in shared libraries, so these calls cannot be overloaded. | |
379 | ||
380 | The HP dynamic loader does not support GNU symbol versioning, so symbol | |
381 | versioning is not supported. It may be necessary to disable symbol | |
382 | versioning with :option:`--disable-symvers` when using GNU ld. | |
383 | ||
384 | POSIX threads are the default. The optional DCE thread library is not | |
385 | supported, so :option:`--enable-threads=dce` does not work. | |
386 | ||
387 | \*-\*-linux-gnu | |
388 | =============== | |
389 | ||
390 | The ``.init_array`` and ``.fini_array`` sections are enabled | |
391 | unconditionally which requires at least glibc 2.1 and binutils 2.12. | |
392 | ||
393 | Versions of libstdc++-v3 starting with 3.2.1 require bug fixes present | |
394 | in glibc 2.2.5 and later. More information is available in the | |
395 | libstdc++-v3 documentation. | |
396 | ||
397 | i?86-\*-linux\* | |
398 | =============== | |
399 | ||
400 | As of GCC 3.3, binutils 2.13.1 or later is required for this platform. | |
401 | See :pr:`10877` for more information. | |
402 | ||
403 | If you receive Signal 11 errors when building on GNU/Linux, then it is | |
404 | possible you have a hardware problem. Further information on this can be | |
405 | found on `www.bitwizard.nl <https://www.bitwizard.nl/sig11/>`_. | |
406 | ||
407 | i?86-\*-solaris2\* | |
408 | ================== | |
409 | ||
410 | Use this for Solaris 11.3 or later on x86 and x86-64 systems. Starting | |
411 | with GCC 4.7, there is also a 64-bit :samp:`amd64-*-solaris2*` or | |
412 | :samp:`x86_64-*-solaris2*` configuration that corresponds to | |
413 | :samp:`sparcv9-sun-solaris2*`. | |
414 | ||
415 | It is recommended that you configure GCC to use the GNU assembler. The | |
416 | versions included in Solaris 11.3, from GNU binutils 2.23.1 or | |
417 | newer (available as :samp:`/usr/bin/gas` and | |
418 | :samp:`/usr/gnu/bin/as`), work fine. The current version, from GNU | |
419 | binutils 2.34, is known to work. Recent versions of the Solaris assembler in | |
420 | :samp:`/usr/bin/as` work almost as well, though. | |
421 | ||
422 | For linking, the Solaris linker is preferred. If you want to use the GNU | |
423 | linker instead, the version in Solaris 11.3, from GNU binutils 2.23.1 or | |
424 | newer (in :samp:`/usr/gnu/bin/ld` and :samp:`/usr/bin/gld`), works, | |
425 | as does the latest version, from GNU binutils 2.34. | |
426 | ||
427 | To use GNU :command:`as`, configure with the options | |
428 | :option:`--with-gnu-as --with-as=/usr/gnu/bin/as`. It may be necessary | |
429 | to configure with :option:`--without-gnu-ld --with-ld=/usr/ccs/bin/ld` to | |
430 | guarantee use of Solaris :command:`ld`. | |
431 | ||
432 | .. todo:: why -without-gnu-ld -with-ld? | |
433 | ||
434 | ia64-\*-linux | |
435 | ============= | |
436 | ||
437 | IA-64 processor (also known as IPF, or Itanium Processor Family) | |
438 | running GNU/Linux. | |
439 | ||
440 | If you are using the installed system libunwind library with | |
441 | :option:`--with-system-libunwind`, then you must use libunwind 0.98 or | |
442 | later. | |
443 | ||
444 | ia64-\*-hpux\* | |
445 | ============== | |
446 | ||
447 | Building GCC on this target requires the GNU Assembler. The bundled HP | |
448 | assembler will not work. To prevent GCC from using the wrong assembler, | |
449 | the option :option:`--with-gnu-as` may be necessary. | |
450 | ||
451 | The GCC libunwind library has not been ported to HPUX. This means that for | |
452 | GCC versions 3.2.3 and earlier, :option:`--enable-libunwind-exceptions` | |
453 | is required to build GCC. For GCC 3.3 and later, this is the default. | |
454 | For gcc 3.4.3 and later, :option:`--enable-libunwind-exceptions` is | |
455 | removed and the system libunwind library will always be used. | |
456 | ||
457 | \*-ibm-aix\* | |
458 | ============ | |
459 | ||
460 | Support for AIX version 3 and older was discontinued in GCC 3.4. | |
461 | Support for AIX version 4.2 and older was discontinued in GCC 4.5. | |
462 | ||
463 | 'out of memory' bootstrap failures may indicate a problem with | |
464 | process resource limits (ulimit). Hard limits are configured in the | |
465 | :samp:`/etc/security/limits` system configuration file. | |
466 | ||
467 | GCC 4.9 and above require a C++ compiler for bootstrap. IBM VAC++ / xlC | |
468 | cannot bootstrap GCC. xlc can bootstrap an older version of GCC and | |
469 | G++ can bootstrap recent releases of GCC. | |
470 | ||
471 | GCC can bootstrap with recent versions of IBM XLC, but bootstrapping | |
472 | with an earlier release of GCC is recommended. Bootstrapping with XLC | |
473 | requires a larger data segment, which can be enabled through the | |
474 | :samp:`{LDR_CNTRL}` environment variable, e.g., | |
475 | ||
476 | .. code-block:: bash | |
477 | ||
478 | % LDR_CNTRL=MAXDATA=0x50000000 | |
479 | % export LDR_CNTRL | |
480 | ||
481 | One can start with a pre-compiled version of GCC to build from | |
482 | sources. One may delete GCC's 'fixed' header files when starting | |
483 | with a version of GCC built for an earlier release of AIX. | |
484 | ||
485 | To speed up the configuration phases of bootstrapping and installing GCC, | |
486 | one may use GNU Bash instead of AIX :command:`/bin/sh`, e.g., | |
487 | ||
488 | .. code-block:: bash | |
489 | ||
490 | % CONFIG_SHELL=/opt/freeware/bin/bash | |
491 | % export CONFIG_SHELL | |
492 | ||
493 | and then proceed as described in :ref:`building`, | |
494 | where we strongly recommend specifying an absolute path | |
495 | to invoke :samp:`{srcdir}` /configure. | |
496 | ||
497 | Because GCC on AIX is built as a 32-bit executable by default, | |
498 | (although it can generate 64-bit programs) the GMP and MPFR libraries | |
499 | required by gfortran must be 32-bit libraries. Building GMP and MPFR | |
500 | as static archive libraries works better than shared libraries. | |
501 | ||
502 | Errors involving ``alloca`` when building GCC generally are due | |
503 | to an incorrect definition of ``CC`` in the Makefile or mixing files | |
504 | compiled with the native C compiler and GCC. During the stage1 phase of | |
505 | the build, the native AIX compiler **must** be invoked as :command:`cc` | |
506 | (not :command:`xlc`). Once :command:`configure` has been informed of | |
507 | :command:`xlc`, one needs to use :samp:`make distclean` to remove the | |
508 | configure cache files and ensure that :envvar:`CC` environment variable | |
509 | does not provide a definition that will confuse :command:`configure`. | |
510 | If this error occurs during stage2 or later, then the problem most likely | |
511 | is the version of Make (see above). | |
512 | ||
513 | The native :command:`as` and :command:`ld` are recommended for | |
514 | bootstrapping on AIX. The GNU Assembler, GNU Linker, and GNU | |
515 | Binutils version 2.20 is the minimum level that supports bootstrap on | |
516 | AIX 5. The GNU Assembler has not been updated to support AIX 6or | |
517 | AIX 7. The native AIX tools do interoperate with GCC. | |
518 | ||
519 | AIX 7.1 added partial support for DWARF debugging, but full support | |
520 | requires AIX 7.1 TL03 SP7 that supports additional DWARF sections and | |
521 | fixes a bug in the assembler. AIX 7.1 TL03 SP5 distributed a version | |
522 | of libm.a missing important symbols; a fix for IV77796 will be | |
523 | included in SP6. | |
524 | ||
525 | AIX 5.3 TL10, AIX 6.1 TL05 and AIX 7.1 TL00 introduced an AIX | |
526 | assembler change that sometimes produces corrupt assembly files | |
527 | causing AIX linker errors. The bug breaks GCC bootstrap on AIX and | |
528 | can cause compilation failures with existing GCC installations. An | |
529 | AIX iFix for AIX 5.3 is available (APAR IZ98385 for AIX 5.3 TL10, APAR | |
530 | IZ98477 for AIX 5.3 TL11 and IZ98134 for AIX 5.3 TL12). AIX 5.3 TL11 SP8, | |
531 | AIX 5.3 TL12 SP5, AIX 6.1 TL04 SP11, AIX 6.1 TL05 SP7, AIX 6.1 TL06 SP6, | |
532 | AIX 6.1 TL07 and AIX 7.1 TL01 should include the fix. | |
533 | ||
534 | Building :samp:`libstdc++.a` requires a fix for an AIX Assembler bug | |
535 | APAR IY26685 (AIX 4.3) or APAR IY25528 (AIX 5.1). It also requires a | |
536 | fix for another AIX Assembler bug and a co-dependent AIX Archiver fix | |
537 | referenced as APAR IY53606 (AIX 5.2) or as APAR IY54774 (AIX 5.1) | |
538 | ||
539 | .. _transferaixshobj: | |
540 | ||
541 | :samp:`libstdc++` in GCC 3.4 increments the major version number of the | |
542 | shared object and GCC installation places the :samp:`libstdc++.a` | |
543 | shared library in a common location which will overwrite the and GCC | |
544 | 3.3 version of the shared library. Applications either need to be | |
545 | re-linked against the new shared library or the GCC 3.1 and GCC 3.3 | |
546 | versions of the :samp:`libstdc++` shared object needs to be available | |
547 | to the AIX runtime loader. The GCC 3.1 :samp:`libstdc++.so.4`, if | |
548 | present, and GCC 3.3 :samp:`libstdc++.so.5` shared objects can be | |
549 | installed for runtime dynamic loading using the following steps to set | |
550 | the :samp:`F_LOADONLY` flag in the shared object for *each* | |
551 | multilib :samp:`libstdc++.a` installed: | |
552 | ||
553 | Extract the shared objects from the currently installed | |
554 | :samp:`libstdc++.a` archive: | |
555 | ||
556 | .. code-block:: bash | |
557 | ||
558 | % ar -x libstdc++.a libstdc++.so.4 libstdc++.so.5 | |
559 | ||
560 | Enable the :samp:`F_LOADONLY` flag so that the shared object will be | |
561 | available for runtime dynamic loading, but not linking: | |
562 | ||
563 | .. code-block:: bash | |
564 | ||
565 | % strip -e libstdc++.so.4 libstdc++.so.5 | |
566 | ||
567 | Archive the runtime-only shared object in the GCC 3.4 | |
568 | :samp:`libstdc++.a` archive: | |
569 | ||
570 | .. code-block:: bash | |
571 | ||
572 | % ar -q libstdc++.a libstdc++.so.4 libstdc++.so.5 | |
573 | ||
574 | Eventually, the :option:`--with-aix-soname=svr4` | |
575 | configure option may drop the need for this procedure for libraries that | |
576 | support it. | |
577 | ||
578 | Linking executables and shared libraries may produce warnings of | |
579 | duplicate symbols. The assembly files generated by GCC for AIX always | |
580 | have included multiple symbol definitions for certain global variable | |
581 | and function declarations in the original program. The warnings should | |
582 | not prevent the linker from producing a correct library or runnable | |
583 | executable. | |
584 | ||
585 | AIX 4.3 utilizes a 'large format' archive to support both 32-bit and | |
586 | 64-bit object modules. The routines provided in AIX 4.3.0 and AIX 4.3.1 | |
587 | to parse archive libraries did not handle the new format correctly. | |
588 | These routines are used by GCC and result in error messages during | |
589 | linking such as 'not a COFF file'. The version of the routines shipped | |
590 | with AIX 4.3.1 should work for a 32-bit environment. The :option:`-g` | |
591 | option of the archive command may be used to create archives of 32-bit | |
592 | objects using the original 'small format'. A correct version of the | |
593 | routines is shipped with AIX 4.3.2 and above. | |
594 | ||
595 | Some versions of the AIX binder (linker) can fail with a relocation | |
596 | overflow severe error when the :option:`-bbigtoc` option is used to link | |
597 | GCC-produced object files into an executable that overflows the TOC. A fix | |
598 | for APAR IX75823 (OVERFLOW DURING LINK WHEN USING GCC AND -BBIGTOC) is | |
599 | available from IBM Customer Support and from its | |
600 | `techsupport.services.ibm.com <https://techsupport.services.ibm.com/>`_ | |
601 | website as PTF U455193. | |
602 | ||
603 | The AIX 4.3.2.1 linker (bos.rte.bind_cmds Level 4.3.2.1) will dump core | |
604 | with a segmentation fault when invoked by any version of GCC. A fix for | |
605 | APAR IX87327 is available from IBM Customer Support and from its | |
606 | `techsupport.services.ibm.com <https://techsupport.services.ibm.com/>`_ | |
607 | website as PTF U461879. This fix is incorporated in AIX 4.3.3 and above. | |
608 | ||
609 | The initial assembler shipped with AIX 4.3.0 generates incorrect object | |
610 | files. A fix for APAR IX74254 (64BIT DISASSEMBLED OUTPUT FROM COMPILER FAILS | |
611 | TO ASSEMBLE/BIND) is available from IBM Customer Support and from its | |
612 | `techsupport.services.ibm.com <https://techsupport.services.ibm.com/>`_ | |
613 | website as PTF U453956. This fix is incorporated in AIX 4.3.1 and above. | |
614 | ||
615 | AIX provides National Language Support (NLS). Compilers and assemblers | |
616 | use NLS to support locale-specific representations of various data | |
617 | formats including floating-point numbers (e.g., :samp:`.` vs :samp:`,` for | |
618 | separating decimal fractions). There have been problems reported where | |
619 | GCC does not produce the same floating-point formats that the assembler | |
620 | expects. If one encounters this problem, set the :envvar:`LANG` | |
621 | environment variable to :samp:`C` or :samp:`En_US`. | |
622 | ||
623 | A default can be specified with the :option:`-mcpu=cpu_type` | |
624 | switch and using the configure option :option:`--with-cpu-cpu_type`. | |
625 | ||
626 | iq2000-\*-elf | |
627 | ============= | |
628 | ||
629 | Vitesse IQ2000 processors. These are used in embedded | |
630 | applications. There are no standard Unix configurations. | |
631 | ||
632 | lm32-\*-elf | |
633 | =========== | |
634 | ||
635 | Lattice Mico32 processor. | |
636 | This configuration is intended for embedded systems. | |
637 | ||
638 | lm32-\*-uclinux | |
639 | =============== | |
640 | ||
641 | Lattice Mico32 processor. | |
642 | This configuration is intended for embedded systems running uClinux. | |
643 | ||
644 | LoongArch | |
645 | ========= | |
646 | ||
647 | LoongArch processor. | |
648 | The following LoongArch targets are available: | |
649 | ||
650 | ``loongarch64-linux-gnu*`` | |
651 | LoongArch processor running GNU/Linux. This target triplet may be coupled | |
652 | with a small set of possible suffixes to identify their default ABI type: | |
653 | ||
654 | ``f64`` | |
655 | Uses ``lp64d/base`` ABI by default. | |
656 | ||
657 | ``f32`` | |
658 | Uses ``lp64f/base`` ABI by default. | |
659 | ||
660 | ``sf`` | |
661 | Uses ``lp64s/base`` ABI by default. | |
662 | ||
663 | ``loongarch64-linux-gnu`` | |
664 | Same as ``loongarch64-linux-gnuf64``, but may be used with | |
665 | :option:`--with-abi=*` to configure the default ABI type. | |
666 | ||
667 | More information about LoongArch can be found at | |
668 | https://github.com/loongson/LoongArch-Documentation. | |
669 | ||
670 | m32c-\*-elf | |
671 | =========== | |
672 | ||
673 | Renesas M32C processor. | |
674 | This configuration is intended for embedded systems. | |
675 | ||
676 | m32r-\*-elf | |
677 | =========== | |
678 | ||
679 | Renesas M32R processor. | |
680 | This configuration is intended for embedded systems. | |
681 | ||
682 | m68k-\*-\* | |
683 | ========== | |
684 | ||
685 | By default, | |
686 | :samp:`m68k-*-elf*`, :samp:`m68k-*-rtems`, :samp:`m68k-*-uclinux` and | |
687 | :samp:`m68k-*-linux` | |
688 | build libraries for both M680x0 and ColdFire processors. If you only | |
689 | need the M680x0 libraries, you can omit the ColdFire ones by passing | |
690 | :option:`--with-arch=m68k` to :command:`configure`. Alternatively, you | |
691 | can omit the M680x0 libraries by passing :option:`--with-arch=cf` to | |
692 | :command:`configure`. These targets default to 5206 or 5475 code as | |
693 | appropriate for the target system when | |
694 | configured with :option:`--with-arch=cf` and 68020 code otherwise. | |
695 | ||
696 | The :samp:`m68k-*-netbsd` and | |
697 | :samp:`m68k-*-openbsd` targets also support the :option:`--with-arch` | |
698 | option. They will generate ColdFire CFV4e code when configured with | |
699 | :option:`--with-arch=cf` and 68020 code otherwise. | |
700 | ||
701 | You can override the default processors listed above by configuring | |
702 | with :option:`--with-cpu=target`. This :samp:`{target}` can either | |
703 | be a :option:`-mcpu` argument or one of the following values: | |
704 | :samp:`m68000`, :samp:`m68010`, :samp:`m68020`, :samp:`m68030`, | |
705 | :samp:`m68040`, :samp:`m68060`, :samp:`m68020-40` and :samp:`m68020-60`. | |
706 | ||
707 | GCC requires at least binutils version 2.17 on these targets. | |
708 | ||
709 | m68k-\*-uclinux | |
710 | =============== | |
711 | ||
712 | GCC 4.3 changed the uClinux configuration so that it uses the | |
713 | :samp:`m68k-linux-gnu` ABI rather than the :samp:`m68k-elf` ABI. | |
714 | It also added improved support for C++ and flat shared libraries, | |
715 | both of which were ABI changes. | |
716 | ||
717 | microblaze-\*-elf | |
718 | ================= | |
719 | ||
720 | Xilinx MicroBlaze processor. | |
721 | This configuration is intended for embedded systems. | |
722 | ||
723 | mips-\*-\* | |
724 | ========== | |
725 | ||
726 | If on a MIPS system you get an error message saying 'does not have gp | |
727 | sections for all it's [sic] sectons [sic]', don't worry about it. This | |
728 | happens whenever you use GAS with the MIPS linker, but there is not | |
729 | really anything wrong, and it is okay to use the output file. You can | |
730 | stop such warnings by installing the GNU linker. | |
731 | ||
732 | It would be nice to extend GAS to produce the gp tables, but they are | |
733 | optional, and there should not be a warning about their absence. | |
734 | ||
735 | The libstdc++ atomic locking routines for MIPS targets requires MIPS II | |
736 | and later. A patch went in just after the GCC 3.3 release to | |
737 | make :samp:`mips*-*-*` use the generic implementation instead. You can also | |
738 | configure for :samp:`mipsel-elf` as a workaround. The | |
739 | :samp:`mips*-*-linux*` target continues to use the MIPS II routines. More | |
740 | work on this is expected in future releases. | |
741 | ||
742 | .. If you make -with-llsc the default for another target, please also | |
743 | update the description of the -with-llsc option. | |
744 | ||
745 | The built-in ``__sync_*`` functions are available on MIPS II and | |
746 | later systems and others that support the :samp:`ll`, :samp:`sc` and | |
747 | :samp:`sync` instructions. This can be overridden by passing | |
748 | :option:`--with-llsc` or :option:`--without-llsc` when configuring GCC. | |
749 | Since the Linux kernel emulates these instructions if they are | |
750 | missing, the default for :samp:`mips*-*-linux*` targets is | |
751 | :option:`--with-llsc`. The :option:`--with-llsc` and | |
752 | :option:`--without-llsc` configure options may be overridden at compile | |
753 | time by passing the :option:`-mllsc` or :option:`-mno-llsc` options to | |
754 | the compiler. | |
755 | ||
756 | MIPS systems check for division by zero (unless | |
757 | :option:`-mno-check-zero-division` is passed to the compiler) by | |
758 | generating either a conditional trap or a break instruction. Using | |
759 | trap results in smaller code, but is only supported on MIPS II and | |
760 | later. Also, some versions of the Linux kernel have a bug that | |
761 | prevents trap from generating the proper signal (``SIGFPE``). To enable | |
762 | the use of break, use the :option:`--with-divide=breaks` | |
763 | :command:`configure` option when configuring GCC. The default is to | |
764 | use traps on systems that support them. | |
765 | ||
766 | moxie-\*-elf | |
767 | ============ | |
768 | ||
769 | The moxie processor. | |
770 | ||
771 | msp430-\*-elf\* | |
772 | =============== | |
773 | ||
774 | TI MSP430 processor. | |
775 | This configuration is intended for embedded systems. | |
776 | ||
777 | :samp:`msp430-*-elf` is the standard configuration with most GCC | |
778 | features enabled by default. | |
779 | ||
780 | :samp:`msp430-*-elfbare` is tuned for a bare-metal environment, and disables | |
781 | features related to shared libraries and other functionality not used for | |
782 | this device. This reduces code and data usage of the GCC libraries, resulting | |
783 | in a minimal run-time environment by default. | |
784 | ||
785 | Features disabled by default include: | |
786 | ||
787 | * transactional memory | |
788 | ||
789 | * __cxa_atexit | |
790 | ||
791 | nds32le-\*-elf | |
792 | ============== | |
793 | ||
794 | Andes NDS32 target in little endian mode. | |
795 | ||
796 | nds32be-\*-elf | |
797 | ============== | |
798 | ||
799 | Andes NDS32 target in big endian mode. | |
800 | ||
801 | nvptx-\*-none | |
802 | ============= | |
803 | ||
804 | Nvidia PTX target. | |
805 | ||
806 | Instead of GNU binutils, you will need to install | |
807 | `nvptx-tools <https://github.com/MentorEmbedded/nvptx-tools/>`_. | |
808 | Tell GCC where to find it: | |
809 | :option:`--with-build-time-tools=[install-nvptx-tools]/nvptx-none/bin`. | |
810 | ||
811 | You will need newlib 3.1.0 or later. It can be | |
812 | automatically built together with GCC. For this, add a symbolic link | |
813 | to nvptx-newlib's :samp:`newlib` directory to the directory containing | |
814 | the GCC sources. | |
815 | ||
816 | Use the :option:`--disable-sjlj-exceptions` and | |
817 | :option:`--enable-newlib-io-long-long` options when configuring. | |
818 | ||
819 | The :option:`--with-arch` option may be specified to override the | |
820 | default value for the :option:`-march` option, and to also build | |
821 | corresponding target libraries. | |
822 | The default is :option:`--with-arch=sm_30`. | |
823 | ||
824 | For example, if :option:`--with-arch=sm_70` is specified, | |
825 | :option:`-march=sm_30` and :option:`-march=sm_70` target libraries are | |
826 | built, and code generation defaults to :option:`-march=sm_70`. | |
827 | ||
828 | or1k-\*-elf | |
829 | =========== | |
830 | ||
831 | The OpenRISC 1000 32-bit processor with delay slots. | |
832 | This configuration is intended for embedded systems. | |
833 | ||
834 | or1k-\*-linux | |
835 | ============= | |
836 | ||
837 | The OpenRISC 1000 32-bit processor with delay slots. | |
838 | ||
839 | powerpc-\*-\* | |
840 | ============= | |
841 | ||
842 | You can specify a default version for the :option:`-mcpu=cpu_type` | |
843 | switch by using the configure option :option:`--with-cpu-cpu_type`. | |
844 | ||
845 | You will need GNU binutils 2.20 or newer. | |
846 | ||
847 | powerpc-\*-darwin\* | |
848 | =================== | |
849 | ||
850 | PowerPC running Darwin (Mac OS X kernel). | |
851 | ||
852 | Pre-installed versions of Mac OS X may not include any developer tools, | |
853 | meaning that you will not be able to build GCC from source. Tool | |
854 | binaries are available at | |
855 | https://opensource.apple.com. | |
856 | ||
857 | This version of GCC requires at least cctools-590.36. The | |
858 | cctools-590.36 package referenced from | |
859 | https://gcc.gnu.org/ml/gcc/2006-03/msg00507.html will not work | |
860 | on systems older than 10.3.9 (aka darwin7.9.0). | |
861 | ||
862 | powerpc-\*-elf | |
863 | ============== | |
864 | ||
865 | PowerPC system in big endian mode, running System V.4. | |
866 | ||
867 | powerpc\*-\*-linux-gnu\* | |
868 | ======================== | |
869 | ||
870 | PowerPC system in big endian mode running Linux. | |
871 | ||
872 | powerpc-\*-netbsd\* | |
873 | =================== | |
874 | ||
875 | PowerPC system in big endian mode running NetBSD. | |
876 | ||
877 | powerpc-\*-eabisim | |
878 | ================== | |
879 | ||
880 | Embedded PowerPC system in big endian mode for use in running under the | |
881 | PSIM simulator. | |
882 | ||
883 | powerpc-\*-eabi | |
884 | =============== | |
885 | ||
886 | Embedded PowerPC system in big endian mode. | |
887 | ||
888 | powerpcle-\*-elf | |
889 | ================ | |
890 | ||
891 | PowerPC system in little endian mode, running System V.4. | |
892 | ||
893 | powerpcle-\*-eabisim | |
894 | ==================== | |
895 | ||
896 | Embedded PowerPC system in little endian mode for use in running under | |
897 | the PSIM simulator. | |
898 | ||
899 | powerpcle-\*-eabi | |
900 | ================= | |
901 | ||
902 | Embedded PowerPC system in little endian mode. | |
903 | ||
904 | rl78-\*-elf | |
905 | =========== | |
906 | ||
907 | The Renesas RL78 processor. | |
908 | This configuration is intended for embedded systems. | |
909 | ||
910 | riscv32-\*-elf | |
911 | ============== | |
912 | ||
913 | The RISC-V RV32 instruction set. | |
914 | This configuration is intended for embedded systems. | |
915 | This (and all other RISC-V) targets require the binutils 2.30 release. | |
916 | ||
917 | riscv32-\*-linux | |
918 | ================ | |
919 | ||
920 | The RISC-V RV32 instruction set running GNU/Linux. | |
921 | This (and all other RISC-V) targets require the binutils 2.30 release. | |
922 | ||
923 | riscv64-\*-elf | |
924 | ============== | |
925 | ||
926 | The RISC-V RV64 instruction set. | |
927 | This configuration is intended for embedded systems. | |
928 | This (and all other RISC-V) targets require the binutils 2.30 release. | |
929 | ||
930 | riscv64-\*-linux | |
931 | ================ | |
932 | ||
933 | The RISC-V RV64 instruction set running GNU/Linux. | |
934 | This (and all other RISC-V) targets require the binutils 2.30 release. | |
935 | ||
936 | rx-\*-elf | |
937 | ========= | |
938 | ||
939 | The Renesas RX processor. | |
940 | ||
941 | s390-\*-linux\* | |
942 | =============== | |
943 | ||
944 | S/390 system running GNU/Linux for S/390. | |
945 | ||
946 | s390x-\*-linux\* | |
947 | ================ | |
948 | ||
949 | zSeries system (64-bit) running GNU/Linux for zSeries. | |
950 | ||
951 | s390x-ibm-tpf\* | |
952 | =============== | |
953 | ||
954 | zSeries system (64-bit) running TPF. This platform is | |
955 | supported as cross-compilation target only. | |
956 | ||
957 | .. Please use Solaris 2 to refer to all release of Solaris, starting | |
958 | with 2.0 until 2.6, 7, 8, etc. Solaris 1 was a marketing name for | |
959 | SunOS 4 releases which we don't use to avoid confusion. Solaris | |
960 | alone is too unspecific and must be avoided. | |
961 | ||
962 | \*-\*-solaris2\* | |
963 | ================ | |
964 | ||
965 | Support for Solaris 10 has been removed in GCC 10. Support for Solaris | |
966 | 9 has been removed in GCC 5. Support for Solaris 8 has been removed in | |
967 | GCC 4.8. Support for Solaris 7 has been removed in GCC 4.6. | |
968 | ||
969 | Solaris 11.3 provides GCC 4.5.2, 4.7.3, and 4.8.2 as | |
970 | :command:`/usr/gcc/4.5/bin/gcc` or similar. Newer Solaris versions | |
971 | provide one or more of GCC 5, 7, and 9. Alternatively, | |
972 | you can install a pre-built GCC to bootstrap and install GCC. See the | |
973 | :ref:`binaries` for details. | |
974 | ||
975 | The Solaris 2 :command:`/bin/sh` will often fail to configure | |
976 | :samp:`libstdc++-v3`. We therefore recommend using the | |
977 | following initial sequence of commands | |
978 | ||
979 | .. code-block:: bash | |
980 | ||
981 | % CONFIG_SHELL=/bin/ksh | |
982 | % export CONFIG_SHELL | |
983 | ||
984 | and proceed as described in :ref:`configuration` the configure instructions. | |
985 | In addition we strongly recommend specifying an absolute path to invoke | |
986 | :samp:`{srcdir}/configure`. | |
987 | ||
988 | In Solaris 11, you need to check for ``system/header``, | |
989 | ``system/linker``, and ``developer/assembler`` packages. | |
990 | ||
991 | Trying to use the linker and other tools in | |
992 | :samp:`/usr/ucb` to install GCC has been observed to cause trouble. | |
993 | For example, the linker may hang indefinitely. The fix is to remove | |
994 | :samp:`/usr/ucb` from your :envvar:`PATH`. | |
995 | ||
996 | The build process works more smoothly with the legacy Solaris tools so, if you | |
997 | have :samp:`/usr/xpg4/bin` in your :envvar:`PATH`, we recommend that you place | |
998 | :samp:`/usr/bin` before :samp:`/usr/xpg4/bin` for the duration of the build. | |
999 | ||
1000 | We recommend the use of the Solaris assembler or the GNU assembler, in | |
1001 | conjunction with the Solaris linker. The GNU :command:`as` | |
1002 | versions included in Solaris 11.3, | |
1003 | from GNU binutils 2.23.1 or newer (in :samp:`/usr/bin/gas` and | |
1004 | :samp:`/usr/gnu/bin/as`), are known to work. | |
1005 | The current version, from GNU binutils 2.34, | |
1006 | is known to work as well. Note that your mileage may vary | |
1007 | if you use a combination of the GNU tools and the Solaris tools: while the | |
1008 | combination GNU :command:`as` + Solaris :command:`ld` should reasonably work, | |
1009 | the reverse combination Solaris :command:`as` + GNU :command:`ld` may fail to | |
1010 | build or cause memory corruption at runtime in some cases for C++ programs. | |
1011 | ||
1012 | .. todo:: still? | |
1013 | ||
1014 | GNU :command:`ld` usually works as well. Again, the current | |
1015 | version (2.34) is known to work, but generally lacks platform specific | |
1016 | features, so better stay with Solaris :command:`ld`. To use the LTO linker | |
1017 | plugin (:option:`-fuse-linker-plugin`) with GNU :command:`ld`, GNU | |
1018 | binutils *must* be configured with :option:`--enable-largefile`. | |
1019 | ||
1020 | To enable symbol versioning in :samp:`libstdc++` with the Solaris linker, | |
1021 | you need to have any version of GNU :command:`c++filt`, which is part of | |
1022 | GNU binutils. :samp:`libstdc++` symbol versioning will be disabled if no | |
1023 | appropriate version is found. Solaris :command:`c++filt` from the Solaris | |
1024 | Studio compilers does *not* work. | |
1025 | ||
1026 | In order to build the GNU D compiler, GDC, a working :samp:`libphobos` is | |
1027 | needed. That library wasn't built by default in GCC 9--11 on SPARC, or | |
1028 | on x86 when the Solaris assembler is used, but can be enabled by | |
1029 | configuring with :option:`--enable-libphobos`. Also, GDC 9.4.0 is | |
1030 | required on x86, while GDC 9.3.0 is known to work on SPARC. | |
1031 | ||
1032 | The versions of the GNU Multiple Precision Library (GMP), the MPFR | |
1033 | library and the MPC library bundled with Solaris 11.3 and later are | |
1034 | usually recent enough to match GCC's requirements. There are two | |
1035 | caveats: | |
1036 | ||
1037 | * While the version of the GMP library in Solaris 11.3 works with GCC, you | |
1038 | need to configure with :option:`--with-gmp-include=/usr/include/gmp`. | |
1039 | ||
1040 | * The version of the MPFR libary included in Solaris 11.3 is too old; you | |
1041 | need to provide a more recent one. | |
1042 | ||
1043 | sparc\*-\*-\* | |
1044 | ============= | |
1045 | ||
1046 | This section contains general configuration information for all | |
1047 | SPARC-based platforms. In addition to reading this section, please | |
1048 | read all other sections that match your target. | |
1049 | ||
1050 | Newer versions of the GNU Multiple Precision Library (GMP), the MPFR | |
1051 | library and the MPC library are known to be miscompiled by earlier | |
1052 | versions of GCC on these platforms. We therefore recommend the use | |
1053 | of the exact versions of these libraries listed as minimal versions | |
1054 | in :ref:`prerequisites` the prerequisites. | |
1055 | ||
1056 | sparc-sun-solaris2\* | |
1057 | ==================== | |
1058 | ||
1059 | When GCC is configured to use GNU binutils 2.14 or later, the binaries | |
1060 | produced are smaller than the ones produced using Solaris native tools; | |
1061 | this difference is quite significant for binaries containing debugging | |
1062 | information. | |
1063 | ||
1064 | Starting with Solaris 7, the operating system is capable of executing | |
1065 | 64-bit SPARC V9 binaries. GCC 3.1 and later properly supports | |
1066 | this; the :option:`-m64` option enables 64-bit code generation. | |
1067 | However, if all you want is code tuned for the UltraSPARC CPU, you | |
1068 | should try the :option:`-mtune=ultrasparc` option instead, which produces | |
1069 | code that, unlike full 64-bit code, can still run on non-UltraSPARC | |
1070 | machines. | |
1071 | ||
1072 | When configuring the GNU Multiple Precision Library (GMP), the MPFR | |
1073 | library or the MPC library on a Solaris 7 or later system, the canonical | |
1074 | target triplet must be specified as the :command:`build` parameter on the | |
1075 | configure line. This target triplet can be obtained by invoking :command:`./config.guess` in the toplevel source directory of GCC (and | |
1076 | not that of GMP or MPFR or MPC). For example on a Solaris 11 system: | |
1077 | ||
1078 | .. code-block:: bash | |
1079 | ||
1080 | % ./configure --build=sparc-sun-solaris2.11 --prefix=xxx | |
1081 | ||
1082 | sparc-\*-linux\* | |
1083 | ================ | |
1084 | ||
1085 | sparc64-\*-solaris2\* | |
1086 | ===================== | |
1087 | ||
1088 | When configuring a 64-bit-default GCC on Solaris/SPARC, you must use a | |
1089 | build compiler that generates 64-bit code, either by default or by | |
1090 | specifying :samp:`CC='gcc -m64' CXX='gcc-m64'` to :command:`configure`. | |
1091 | Additionally, you *must* pass :option:`--build=sparc64-sun-solaris2.11` | |
1092 | or :option:`--build=sparcv9-sun-solaris2.11` because :samp:`config.guess` | |
1093 | misdetects this situation, which can cause build failures. | |
1094 | ||
1095 | When configuring the GNU Multiple Precision Library (GMP), the MPFR | |
1096 | library or the MPC library, the canonical target triplet must be specified | |
1097 | as the :command:`build` parameter on the configure line. For example | |
1098 | on a Solaris 11 system: | |
1099 | ||
1100 | .. code-block:: bash | |
1101 | ||
1102 | % ./configure --build=sparc64-sun-solaris2.11 --prefix=xxx | |
1103 | ||
1104 | sparcv9-\*-solaris2\* | |
1105 | ===================== | |
1106 | ||
1107 | This is a synonym for :samp:`sparc64-*-solaris2*`. | |
1108 | ||
1109 | c6x-\*-\* | |
1110 | ========= | |
1111 | ||
1112 | The C6X family of processors. This port requires binutils-2.22 or newer. | |
1113 | ||
1114 | visium-\*-elf | |
1115 | ============= | |
1116 | ||
1117 | CDS VISIUMcore processor. | |
1118 | This configuration is intended for embedded systems. | |
1119 | ||
1120 | \*-\*-vxworks\* | |
1121 | =============== | |
1122 | ||
1123 | Support for VxWorks is in flux. At present GCC supports *only* the | |
1124 | very recent VxWorks 5.5 (aka Tornado 2.2) release, and only on PowerPC. | |
1125 | We welcome patches for other architectures supported by VxWorks 5.5. | |
1126 | Support for VxWorks AE would also be welcome; we believe this is merely | |
1127 | a matter of writing an appropriate 'configlette' (see below). We are | |
1128 | not interested in supporting older, a.out or COFF-based, versions of | |
1129 | VxWorks in GCC 3. | |
1130 | ||
1131 | VxWorks comes with an older version of GCC installed in | |
1132 | :samp:`{$WIND_BASE}/host`; we recommend you do not overwrite it. | |
1133 | Choose an installation :samp:`{prefix}` entirely outside :samp:`{$WIND_BASE}`. | |
1134 | Before running :command:`configure`, create the directories :samp:`{prefix}` | |
1135 | and :samp:`{prefix}/bin`. Link or copy the appropriate assembler, | |
1136 | linker, etc. into :samp:`{prefix}/bin`, and set your :samp:`{PATH}` to | |
1137 | include that directory while running both :command:`configure` and | |
1138 | :command:`make`. | |
1139 | ||
1140 | You must give :command:`configure` the | |
1141 | :option:`--with-headers=$WIND_BASE/target/h` switch so that it can | |
1142 | find the VxWorks system headers. Since VxWorks is a cross compilation | |
1143 | target only, you must also specify :option:`--target=target`. | |
1144 | :command:`configure` will attempt to create the directory | |
1145 | :samp:`{prefix}/{target}/sys-include` and copy files into it; | |
1146 | make sure the user running :command:`configure` has sufficient privilege | |
1147 | to do so. | |
1148 | ||
1149 | GCC's exception handling runtime requires a special 'configlette' | |
1150 | module, :samp:`contrib/gthr_supp_vxw_5x.c`. Follow the instructions in | |
1151 | that file to add the module to your kernel build. (Future versions of | |
1152 | VxWorks will incorporate this module.) | |
1153 | ||
1154 | x86_64-\*-\*, amd64-\*-\* | |
1155 | ========================= | |
1156 | ||
1157 | GCC supports the x86-64 architecture implemented by the AMD64 processor | |
1158 | (amd64-\*-\* is an alias for x86_64-\*-\*) on GNU/Linux, FreeBSD and NetBSD. | |
1159 | On GNU/Linux the default is a bi-arch compiler which is able to generate | |
1160 | both 64-bit x86-64 and 32-bit x86 code (via the :option:`-m32` switch). | |
1161 | ||
1162 | x86_64-\*-solaris2\* | |
1163 | ==================== | |
1164 | ||
1165 | GCC also supports the x86-64 architecture implemented by the AMD64 | |
1166 | processor (:samp:`amd64-*-*` is an alias for :samp:`x86_64-*-*`) on | |
1167 | Solaris 10 or later. Unlike other systems, without special options a | |
1168 | bi-arch compiler is built which generates 32-bit code by default, but | |
1169 | can generate 64-bit x86-64 code with the :option:`-m64` switch. Since | |
1170 | GCC 4.7, there is also a configuration that defaults to 64-bit code, but | |
1171 | can generate 32-bit code with :option:`-m32`. To configure and build | |
1172 | this way, you have to provide all support libraries like :samp:`libgmp` | |
1173 | as 64-bit code, configure with :option:`--target=x86_64-pc-solaris2.11` | |
1174 | and :samp:`CC=gcc -m64`. | |
1175 | ||
1176 | xtensa\*-\*-elf | |
1177 | =============== | |
1178 | ||
1179 | This target is intended for embedded Xtensa systems using the | |
1180 | :samp:`newlib` C library. It uses ELF but does not support shared | |
1181 | objects. Designed-defined instructions specified via the | |
1182 | Tensilica Instruction Extension (TIE) language are only supported | |
1183 | through inline assembly. | |
1184 | ||
1185 | The Xtensa configuration information must be specified prior to | |
1186 | building GCC. The :samp:`include/xtensa-config.h` header | |
1187 | file contains the configuration information. If you created your | |
1188 | own Xtensa configuration with the Xtensa Processor Generator, the | |
1189 | downloaded files include a customized copy of this header file, | |
1190 | which you can use to replace the default header file. | |
1191 | ||
1192 | xtensa\*-\*-linux\* | |
1193 | =================== | |
1194 | ||
1195 | This target is for Xtensa systems running GNU/Linux. It supports ELF | |
1196 | shared objects and the GNU C library (glibc). It also generates | |
1197 | position-independent code (PIC) regardless of whether the | |
1198 | :option:`-fpic` or :option:`-fPIC` options are used. In other | |
1199 | respects, this target is the same as the | |
1200 | :samp:`xtensa*-*-elf` target. | |
1201 | ||
1202 | Microsoft Windows | |
1203 | ================= | |
1204 | ||
1205 | Intel 16-bit versions | |
1206 | ===================== | |
1207 | ||
1208 | The 16-bit versions of Microsoft Windows, such as Windows 3.1, are not | |
1209 | supported. | |
1210 | ||
1211 | However, the 32-bit port has limited support for Microsoft | |
1212 | Windows 3.11 in the Win32s environment, as a target only. See below. | |
1213 | ||
1214 | Intel 32-bit versions | |
1215 | ===================== | |
1216 | ||
1217 | The 32-bit versions of Windows, including Windows 95, Windows NT, Windows | |
1218 | XP, and Windows Vista, are supported by several different target | |
1219 | platforms. These targets differ in which Windows subsystem they target | |
1220 | and which C libraries are used. | |
1221 | ||
1222 | * Cygwin \*-\*-cygwin: Cygwin provides a user-space | |
1223 | Linux API emulation layer in the Win32 subsystem. | |
1224 | ||
1225 | * MinGW \*-\*-mingw32: MinGW is a native GCC port for | |
1226 | the Win32 subsystem that provides a subset of POSIX. | |
1227 | ||
1228 | * MKS i386-pc-mks: NuTCracker from MKS. See | |
1229 | https://www.mkssoftware.com for more information. | |
1230 | ||
1231 | Intel 64-bit versions | |
1232 | ===================== | |
1233 | ||
1234 | GCC contains support for x86-64 using the mingw-w64 | |
1235 | runtime library, available from https://www.mingw-w64.org/downloads/. | |
1236 | This library should be used with the target triple x86_64-pc-mingw32. | |
1237 | ||
1238 | Windows CE | |
1239 | ========== | |
1240 | ||
1241 | Windows CE is supported as a target only on Hitachi | |
1242 | SuperH (sh-wince-pe), and MIPS (mips-wince-pe). | |
1243 | ||
1244 | Other Windows Platforms | |
1245 | ======================= | |
1246 | ||
1247 | GCC no longer supports Windows NT on the Alpha or PowerPC. | |
1248 | ||
1249 | GCC no longer supports the Windows POSIX subsystem. However, it does | |
1250 | support the Interix subsystem. See above. | |
1251 | ||
1252 | Old target names including \*-\*-winnt and \*-\*-windowsnt are no longer used. | |
1253 | ||
1254 | PW32 (i386-pc-pw32) support was never completed, and the project seems to | |
1255 | be inactive. See http://pw32.sourceforge.net/ for more information. | |
1256 | ||
1257 | UWIN support has been removed due to a lack of maintenance. | |
1258 | ||
1259 | \*-\*-cygwin | |
1260 | ============ | |
1261 | ||
1262 | Ports of GCC are included with the | |
1263 | `Cygwin environment <http://www.cygwin.com/>`_. | |
1264 | ||
1265 | GCC will build under Cygwin without modification; it does not build | |
1266 | with Microsoft's C++ compiler and there are no plans to make it do so. | |
1267 | ||
1268 | The Cygwin native compiler can be configured to target any 32-bit x86 | |
1269 | cpu architecture desired; the default is i686-pc-cygwin. It should be | |
1270 | used with as up-to-date a version of binutils as possible; use either | |
1271 | the latest official GNU binutils release in the Cygwin distribution, | |
1272 | or version 2.20 or above if building your own. | |
1273 | ||
1274 | \*-\*-mingw32 | |
1275 | ============= | |
1276 | ||
1277 | GCC will build with and support only MinGW runtime 3.12 and later. | |
1278 | Earlier versions of headers are incompatible with the new default semantics | |
1279 | of ``extern inline`` in ``-std=c99`` and ``-std=gnu99`` modes. | |
1280 | ||
1281 | To support emitting DWARF debugging info you need to use GNU binutils | |
1282 | version 2.16 or above containing support for the ``.secrel32`` | |
1283 | assembler pseudo-op. | |
1284 | ||
1285 | Older systems | |
1286 | ============= | |
1287 | ||
1288 | GCC contains support files for many older (1980s and early | |
1289 | 1990s) Unix variants. For the most part, support for these systems | |
1290 | has not been deliberately removed, but it has not been maintained for | |
1291 | several years and may suffer from bitrot. | |
1292 | ||
1293 | Starting with GCC 3.1, each release has a list of 'obsoleted' systems. | |
1294 | Support for these systems is still present in that release, but | |
1295 | :command:`configure` will fail unless the :option:`--enable-obsolete` | |
1296 | option is given. Unless a maintainer steps forward, support for these | |
1297 | systems will be removed from the next release of GCC. | |
1298 | ||
1299 | Support for old systems as hosts for GCC can cause problems if the | |
1300 | workarounds for compiler, library and operating system bugs affect the | |
1301 | cleanliness or maintainability of the rest of GCC. In some cases, to | |
1302 | bring GCC up on such a system, if still possible with current GCC, may | |
1303 | require first installing an old version of GCC which did work on that | |
1304 | system, and using it to compile a more recent GCC, to avoid bugs in the | |
1305 | vendor compiler. Old releases of GCC 1 and GCC 2 are available in the | |
1306 | :samp:`old-releases` directory on the | |
1307 | `GCC mirror sites <https://gcc.gnu.org/mirrors.html>`_. | |
1308 | Header bugs may generally be avoided using | |
1309 | :command:`fixincludes`, but bugs or deficiencies in libraries and the | |
1310 | operating system may still cause problems. | |
1311 | ||
1312 | Support for older systems as targets for cross-compilation is less | |
1313 | problematic than support for them as hosts for GCC; if an enthusiast | |
1314 | wishes to make such a target work again (including resurrecting any of | |
1315 | the targets that never worked with GCC 2, starting from the last | |
1316 | version before they were removed), patches | |
1317 | https://gcc.gnu.org/contribute.html following the usual requirements would be | |
1318 | likely to be accepted, since they should not affect the support for more | |
1319 | modern targets. | |
1320 | ||
1321 | For some systems, old versions of GNU binutils may also be useful, | |
1322 | and are available from :samp:`pub/binutils/old-releases` on | |
1323 | `sourceware.org mirror sites <https://sourceware.org/mirrors.html>`_. | |
1324 | ||
1325 | Some of the information on specific systems above relates to | |
1326 | such older systems, but much of the information | |
1327 | about GCC on such systems (which may no longer be applicable to | |
1328 | current GCC) is to be found in the GCC texinfo manual. | |
1329 | ||
1330 | all ELF targets (SVR4, Solaris 2, etc.) | |
1331 | ======================================= | |
1332 | ||
1333 | C++ support is significantly better on ELF targets if you use the | |
1334 | GNU linker; duplicate copies of | |
1335 | inlines, vtables and template instantiations will be discarded | |
3ed1b4ce | 1336 | automatically. |