configure.html revision 1.1.1.10
1 <html lang="en"> 2 <head> 3 <title>Installing GCC: Configuration</title> 4 <meta http-equiv="Content-Type" content="text/html"> 5 <meta name="description" content="Installing GCC: Configuration"> 6 <meta name="generator" content="makeinfo 4.13"> 7 <link title="Top" rel="top" href="#Top"> 8 <link href="http://www.gnu.org/software/texinfo/" rel="generator-home" title="Texinfo Homepage"> 9 <!-- 10 Copyright (C) 1988-2017 Free Software Foundation, Inc. 11 12 Permission is granted to copy, distribute and/or modify this document 13 under the terms of the GNU Free Documentation License, Version 1.3 or 14 any later version published by the Free Software Foundation; with no 15 Invariant Sections, the Front-Cover texts being (a) (see below), and 16 with the Back-Cover Texts being (b) (see below). A copy of the 17 license is included in the section entitled "GNU Free Documentation License". 18 19 (a) The FSF's Front-Cover Text is: 20 21 A GNU Manual 22 23 (b) The FSF's Back-Cover Text is: 24 25 You have freedom to copy and modify this GNU Manual, like GNU 26 software. Copies published by the Free Software Foundation raise 27 funds for GNU development.--> 28 <meta http-equiv="Content-Style-Type" content="text/css"> 29 <style type="text/css"><!-- 30 pre.display { font-family:inherit } 31 pre.format { font-family:inherit } 32 pre.smalldisplay { font-family:inherit; font-size:smaller } 33 pre.smallformat { font-family:inherit; font-size:smaller } 34 pre.smallexample { font-size:smaller } 35 pre.smalllisp { font-size:smaller } 36 span.sc { font-variant:small-caps } 37 span.roman { font-family:serif; font-weight:normal; } 38 span.sansserif { font-family:sans-serif; font-weight:normal; } 39 --></style> 40 </head> 41 <body> 42 <h1 class="settitle">Installing GCC: Configuration</h1> 43 <a name="index-Configuration-1"></a><a name="index-Installing-GCC_003a-Configuration-2"></a> 44 Like most GNU software, GCC must be configured before it can be built. 45 This document describes the recommended configuration procedure 46 for both native and cross targets. 47 48 <p>We use <var>srcdir</var> to refer to the toplevel source directory for 49 GCC; we use <var>objdir</var> to refer to the toplevel build/object directory. 50 51 <p>If you obtained the sources via SVN, <var>srcdir</var> must refer to the top 52 <samp><span class="file">gcc</span></samp> directory, the one where the <samp><span class="file">MAINTAINERS</span></samp> file can be 53 found, and not its <samp><span class="file">gcc</span></samp> subdirectory, otherwise the build will fail. 54 55 <p>If either <var>srcdir</var> or <var>objdir</var> is located on an automounted NFS 56 file system, the shell's built-in <samp><span class="command">pwd</span></samp> command will return 57 temporary pathnames. Using these can lead to various sorts of build 58 problems. To avoid this issue, set the <samp><span class="env">PWDCMD</span></samp> environment 59 variable to an automounter-aware <samp><span class="command">pwd</span></samp> command, e.g., 60 <samp><span class="command">pawd</span></samp> or ‘<samp><span class="samp">amq -w</span></samp>’, during the configuration and build 61 phases. 62 63 <p>First, we <strong>highly</strong> recommend that GCC be built into a 64 separate directory from the sources which does <strong>not</strong> reside 65 within the source tree. This is how we generally build GCC; building 66 where <var>srcdir</var> == <var>objdir</var> should still work, but doesn't 67 get extensive testing; building where <var>objdir</var> is a subdirectory 68 of <var>srcdir</var> is unsupported. 69 70 <p>If you have previously built GCC in the same directory for a 71 different target machine, do ‘<samp><span class="samp">make distclean</span></samp>’ to delete all files 72 that might be invalid. One of the files this deletes is <samp><span class="file">Makefile</span></samp>; 73 if ‘<samp><span class="samp">make distclean</span></samp>’ complains that <samp><span class="file">Makefile</span></samp> does not exist 74 or issues a message like “don't know how to make distclean” it probably 75 means that the directory is already suitably clean. However, with the 76 recommended method of building in a separate <var>objdir</var>, you should 77 simply use a different <var>objdir</var> for each target. 78 79 <p>Second, when configuring a native system, either <samp><span class="command">cc</span></samp> or 80 <samp><span class="command">gcc</span></samp> must be in your path or you must set <samp><span class="env">CC</span></samp> in 81 your environment before running configure. Otherwise the configuration 82 scripts may fail. 83 84 <p>To configure GCC: 85 86 <pre class="smallexample"> % mkdir <var>objdir</var> 87 % cd <var>objdir</var> 88 % <var>srcdir</var>/configure [<var>options</var>] [<var>target</var>] 89 </pre> 90 <h3 class="heading"><a name="TOC0"></a>Distributor options</h3> 91 92 <p>If you will be distributing binary versions of GCC, with modifications 93 to the source code, you should use the options described in this 94 section to make clear that your version contains modifications. 95 96 <dl> 97 <dt><code>--with-pkgversion=</code><var>version</var><dd>Specify a string that identifies your package. You may wish 98 to include a build number or build date. This version string will be 99 included in the output of <samp><span class="command">gcc --version</span></samp>. This suffix does 100 not replace the default version string, only the ‘<samp><span class="samp">GCC</span></samp>’ part. 101 102 <p>The default value is ‘<samp><span class="samp">GCC</span></samp>’. 103 104 <br><dt><code>--with-bugurl=</code><var>url</var><dd>Specify the URL that users should visit if they wish to report a bug. 105 You are of course welcome to forward bugs reported to you to the FSF, 106 if you determine that they are not bugs in your modifications. 107 108 <p>The default value refers to the FSF's GCC bug tracker. 109 110 </dl> 111 112 <h3 class="heading"><a name="TOC1"></a>Target specification</h3> 113 114 <ul> 115 <li>GCC has code to correctly determine the correct value for <var>target</var> 116 for nearly all native systems. Therefore, we highly recommend you do 117 not provide a configure target when configuring a native compiler. 118 119 <li><var>target</var> must be specified as <samp><span class="option">--target=</span><var>target</var></samp> 120 when configuring a cross compiler; examples of valid targets would be 121 m68k-elf, sh-elf, etc. 122 123 <li>Specifying just <var>target</var> instead of <samp><span class="option">--target=</span><var>target</var></samp> 124 implies that the host defaults to <var>target</var>. 125 </ul> 126 127 <h3 class="heading"><a name="TOC2"></a>Options specification</h3> 128 129 <p>Use <var>options</var> to override several configure time options for 130 GCC. A list of supported <var>options</var> follows; ‘<samp><span class="samp">configure 131 --help</span></samp>’ may list other options, but those not listed below may not 132 work and should not normally be used. 133 134 <p>Note that each <samp><span class="option">--enable</span></samp> option has a corresponding 135 <samp><span class="option">--disable</span></samp> option and that each <samp><span class="option">--with</span></samp> option has a 136 corresponding <samp><span class="option">--without</span></samp> option. 137 138 <dl> 139 <dt><code>--prefix=</code><var>dirname</var><dd>Specify the toplevel installation 140 directory. This is the recommended way to install the tools into a directory 141 other than the default. The toplevel installation directory defaults to 142 <samp><span class="file">/usr/local</span></samp>. 143 144 <p>We <strong>highly</strong> recommend against <var>dirname</var> being the same or a 145 subdirectory of <var>objdir</var> or vice versa. If specifying a directory 146 beneath a user's home directory tree, some shells will not expand 147 <var>dirname</var> correctly if it contains the ‘<samp><span class="samp">~</span></samp>’ metacharacter; use 148 <samp><span class="env">$HOME</span></samp> instead. 149 150 <p>The following standard <samp><span class="command">autoconf</span></samp> options are supported. Normally you 151 should not need to use these options. 152 <dl> 153 <dt><code>--exec-prefix=</code><var>dirname</var><dd>Specify the toplevel installation directory for architecture-dependent 154 files. The default is <samp><var>prefix</var></samp>. 155 156 <br><dt><code>--bindir=</code><var>dirname</var><dd>Specify the installation directory for the executables called by users 157 (such as <samp><span class="command">gcc</span></samp> and <samp><span class="command">g++</span></samp>). The default is 158 <samp><var>exec-prefix</var><span class="file">/bin</span></samp>. 159 160 <br><dt><code>--libdir=</code><var>dirname</var><dd>Specify the installation directory for object code libraries and 161 internal data files of GCC. The default is <samp><var>exec-prefix</var><span class="file">/lib</span></samp>. 162 163 <br><dt><code>--libexecdir=</code><var>dirname</var><dd>Specify the installation directory for internal executables of GCC. 164 The default is <samp><var>exec-prefix</var><span class="file">/libexec</span></samp>. 165 166 <br><dt><code>--with-slibdir=</code><var>dirname</var><dd>Specify the installation directory for the shared libgcc library. The 167 default is <samp><var>libdir</var></samp>. 168 169 <br><dt><code>--datarootdir=</code><var>dirname</var><dd>Specify the root of the directory tree for read-only architecture-independent 170 data files referenced by GCC. The default is <samp><var>prefix</var><span class="file">/share</span></samp>. 171 172 <br><dt><code>--infodir=</code><var>dirname</var><dd>Specify the installation directory for documentation in info format. 173 The default is <samp><var>datarootdir</var><span class="file">/info</span></samp>. 174 175 <br><dt><code>--datadir=</code><var>dirname</var><dd>Specify the installation directory for some architecture-independent 176 data files referenced by GCC. The default is <samp><var>datarootdir</var></samp>. 177 178 <br><dt><code>--docdir=</code><var>dirname</var><dd>Specify the installation directory for documentation files (other 179 than Info) for GCC. The default is <samp><var>datarootdir</var><span class="file">/doc</span></samp>. 180 181 <br><dt><code>--htmldir=</code><var>dirname</var><dd>Specify the installation directory for HTML documentation files. 182 The default is <samp><var>docdir</var></samp>. 183 184 <br><dt><code>--pdfdir=</code><var>dirname</var><dd>Specify the installation directory for PDF documentation files. 185 The default is <samp><var>docdir</var></samp>. 186 187 <br><dt><code>--mandir=</code><var>dirname</var><dd>Specify the installation directory for manual pages. The default is 188 <samp><var>datarootdir</var><span class="file">/man</span></samp>. (Note that the manual pages are only extracts 189 from the full GCC manuals, which are provided in Texinfo format. The manpages 190 are derived by an automatic conversion process from parts of the full 191 manual.) 192 193 <br><dt><code>--with-gxx-include-dir=</code><var>dirname</var><dd>Specify 194 the installation directory for G++ header files. The default depends 195 on other configuration options, and differs between cross and native 196 configurations. 197 198 <br><dt><code>--with-specs=</code><var>specs</var><dd>Specify additional command line driver SPECS. 199 This can be useful if you need to turn on a non-standard feature by 200 default without modifying the compiler's source code, for instance 201 <samp><span class="option">--with-specs=%{!fcommon:%{!fno-common:-fno-common}}</span></samp>. 202 See “Spec Files” in the main manual 203 204 </dl> 205 206 <br><dt><code>--program-prefix=</code><var>prefix</var><dd>GCC supports some transformations of the names of its programs when 207 installing them. This option prepends <var>prefix</var> to the names of 208 programs to install in <var>bindir</var> (see above). For example, specifying 209 <samp><span class="option">--program-prefix=foo-</span></samp> would result in ‘<samp><span class="samp">gcc</span></samp>’ 210 being installed as <samp><span class="file">/usr/local/bin/foo-gcc</span></samp>. 211 212 <br><dt><code>--program-suffix=</code><var>suffix</var><dd>Appends <var>suffix</var> to the names of programs to install in <var>bindir</var> 213 (see above). For example, specifying <samp><span class="option">--program-suffix=-3.1</span></samp> 214 would result in ‘<samp><span class="samp">gcc</span></samp>’ being installed as 215 <samp><span class="file">/usr/local/bin/gcc-3.1</span></samp>. 216 217 <br><dt><code>--program-transform-name=</code><var>pattern</var><dd>Applies the ‘<samp><span class="samp">sed</span></samp>’ script <var>pattern</var> to be applied to the names 218 of programs to install in <var>bindir</var> (see above). <var>pattern</var> has to 219 consist of one or more basic ‘<samp><span class="samp">sed</span></samp>’ editing commands, separated by 220 semicolons. For example, if you want the ‘<samp><span class="samp">gcc</span></samp>’ program name to be 221 transformed to the installed program <samp><span class="file">/usr/local/bin/myowngcc</span></samp> and 222 the ‘<samp><span class="samp">g++</span></samp>’ program name to be transformed to 223 <samp><span class="file">/usr/local/bin/gspecial++</span></samp> without changing other program names, 224 you could use the pattern 225 <samp><span class="option">--program-transform-name='s/^gcc$/myowngcc/; s/^g++$/gspecial++/'</span></samp> 226 to achieve this effect. 227 228 <p>All three options can be combined and used together, resulting in more 229 complex conversion patterns. As a basic rule, <var>prefix</var> (and 230 <var>suffix</var>) are prepended (appended) before further transformations 231 can happen with a special transformation script <var>pattern</var>. 232 233 <p>As currently implemented, this option only takes effect for native 234 builds; cross compiler binaries' names are not transformed even when a 235 transformation is explicitly asked for by one of these options. 236 237 <p>For native builds, some of the installed programs are also installed 238 with the target alias in front of their name, as in 239 ‘<samp><span class="samp">i686-pc-linux-gnu-gcc</span></samp>’. All of the above transformations happen 240 before the target alias is prepended to the name—so, specifying 241 <samp><span class="option">--program-prefix=foo-</span></samp> and <samp><span class="option">program-suffix=-3.1</span></samp>, the 242 resulting binary would be installed as 243 <samp><span class="file">/usr/local/bin/i686-pc-linux-gnu-foo-gcc-3.1</span></samp>. 244 245 <p>As a last shortcoming, none of the installed Ada programs are 246 transformed yet, which will be fixed in some time. 247 248 <br><dt><code>--with-local-prefix=</code><var>dirname</var><dd>Specify the 249 installation directory for local include files. The default is 250 <samp><span class="file">/usr/local</span></samp>. Specify this option if you want the compiler to 251 search directory <samp><var>dirname</var><span class="file">/include</span></samp> for locally installed 252 header files <em>instead</em> of <samp><span class="file">/usr/local/include</span></samp>. 253 254 <p>You should specify <samp><span class="option">--with-local-prefix</span></samp> <strong>only</strong> if your 255 site has a different convention (not <samp><span class="file">/usr/local</span></samp>) for where to put 256 site-specific files. 257 258 <p>The default value for <samp><span class="option">--with-local-prefix</span></samp> is <samp><span class="file">/usr/local</span></samp> 259 regardless of the value of <samp><span class="option">--prefix</span></samp>. Specifying 260 <samp><span class="option">--prefix</span></samp> has no effect on which directory GCC searches for 261 local header files. This may seem counterintuitive, but actually it is 262 logical. 263 264 <p>The purpose of <samp><span class="option">--prefix</span></samp> is to specify where to <em>install 265 GCC</em>. The local header files in <samp><span class="file">/usr/local/include</span></samp>—if you put 266 any in that directory—are not part of GCC. They are part of other 267 programs—perhaps many others. (GCC installs its own header files in 268 another directory which is based on the <samp><span class="option">--prefix</span></samp> value.) 269 270 <p>Both the local-prefix include directory and the GCC-prefix include 271 directory are part of GCC's “system include” directories. Although these 272 two directories are not fixed, they need to be searched in the proper 273 order for the correct processing of the include_next directive. The 274 local-prefix include directory is searched before the GCC-prefix 275 include directory. Another characteristic of system include directories 276 is that pedantic warnings are turned off for headers in these directories. 277 278 <p>Some autoconf macros add <samp><span class="option">-I </span><var>directory</var></samp> options to the 279 compiler command line, to ensure that directories containing installed 280 packages' headers are searched. When <var>directory</var> is one of GCC's 281 system include directories, GCC will ignore the option so that system 282 directories continue to be processed in the correct order. This 283 may result in a search order different from what was specified but the 284 directory will still be searched. 285 286 <p>GCC automatically searches for ordinary libraries using 287 <samp><span class="env">GCC_EXEC_PREFIX</span></samp>. Thus, when the same installation prefix is 288 used for both GCC and packages, GCC will automatically search for 289 both headers and libraries. This provides a configuration that is 290 easy to use. GCC behaves in a manner similar to that when it is 291 installed as a system compiler in <samp><span class="file">/usr</span></samp>. 292 293 <p>Sites that need to install multiple versions of GCC may not want to 294 use the above simple configuration. It is possible to use the 295 <samp><span class="option">--program-prefix</span></samp>, <samp><span class="option">--program-suffix</span></samp> and 296 <samp><span class="option">--program-transform-name</span></samp> options to install multiple versions 297 into a single directory, but it may be simpler to use different prefixes 298 and the <samp><span class="option">--with-local-prefix</span></samp> option to specify the location of the 299 site-specific files for each version. It will then be necessary for 300 users to specify explicitly the location of local site libraries 301 (e.g., with <samp><span class="env">LIBRARY_PATH</span></samp>). 302 303 <p>The same value can be used for both <samp><span class="option">--with-local-prefix</span></samp> and 304 <samp><span class="option">--prefix</span></samp> provided it is not <samp><span class="file">/usr</span></samp>. This can be used 305 to avoid the default search of <samp><span class="file">/usr/local/include</span></samp>. 306 307 <p><strong>Do not</strong> specify <samp><span class="file">/usr</span></samp> as the <samp><span class="option">--with-local-prefix</span></samp>! 308 The directory you use for <samp><span class="option">--with-local-prefix</span></samp> <strong>must not</strong> 309 contain any of the system's standard header files. If it did contain 310 them, certain programs would be miscompiled (including GNU Emacs, on 311 certain targets), because this would override and nullify the header 312 file corrections made by the <samp><span class="command">fixincludes</span></samp> script. 313 314 <p>Indications are that people who use this option use it based on mistaken 315 ideas of what it is for. People use it as if it specified where to 316 install part of GCC. Perhaps they make this assumption because 317 installing GCC creates the directory. 318 319 <br><dt><code>--with-gcc-major-version-only</code><dd>Specifies that GCC should use only the major number rather than 320 <var>major</var>.<var>minor</var>.<var>patchlevel</var> in filesystem paths. 321 322 <br><dt><code>--with-native-system-header-dir=</code><var>dirname</var><dd>Specifies that <var>dirname</var> is the directory that contains native system 323 header files, rather than <samp><span class="file">/usr/include</span></samp>. This option is most useful 324 if you are creating a compiler that should be isolated from the system 325 as much as possible. It is most commonly used with the 326 <samp><span class="option">--with-sysroot</span></samp> option and will cause GCC to search 327 <var>dirname</var> inside the system root specified by that option. 328 329 <br><dt><code>--enable-shared[=</code><var>package</var><code>[,...]]</code><dd>Build shared versions of libraries, if shared libraries are supported on 330 the target platform. Unlike GCC 2.95.x and earlier, shared libraries 331 are enabled by default on all platforms that support shared libraries. 332 333 <p>If a list of packages is given as an argument, build shared libraries 334 only for the listed packages. For other packages, only static libraries 335 will be built. Package names currently recognized in the GCC tree are 336 ‘<samp><span class="samp">libgcc</span></samp>’ (also known as ‘<samp><span class="samp">gcc</span></samp>’), ‘<samp><span class="samp">libstdc++</span></samp>’ (not 337 ‘<samp><span class="samp">libstdc++-v3</span></samp>’), ‘<samp><span class="samp">libffi</span></samp>’, ‘<samp><span class="samp">zlib</span></samp>’, ‘<samp><span class="samp">boehm-gc</span></samp>’, 338 ‘<samp><span class="samp">ada</span></samp>’, ‘<samp><span class="samp">libada</span></samp>’, ‘<samp><span class="samp">libgo</span></samp>’, and ‘<samp><span class="samp">libobjc</span></samp>’. 339 Note ‘<samp><span class="samp">libiberty</span></samp>’ does not support shared libraries at all. 340 341 <p>Use <samp><span class="option">--disable-shared</span></samp> to build only static libraries. Note that 342 <samp><span class="option">--disable-shared</span></samp> does not accept a list of package names as 343 argument, only <samp><span class="option">--enable-shared</span></samp> does. 344 345 <p>Contrast with <samp><span class="option">--enable-host-shared</span></samp>, which affects <em>host</em> 346 code. 347 348 <br><dt><code>--enable-host-shared</code><dd>Specify that the <em>host</em> code should be built into position-independent 349 machine code (with -fPIC), allowing it to be used within shared libraries, 350 but yielding a slightly slower compiler. 351 352 <p>This option is required when building the libgccjit.so library. 353 354 <p>Contrast with <samp><span class="option">--enable-shared</span></samp>, which affects <em>target</em> 355 libraries. 356 357 <br><dt><code><a name="with-gnu-as"></a>--with-gnu-as</code><dd>Specify that the compiler should assume that the 358 assembler it finds is the GNU assembler. However, this does not modify 359 the rules to find an assembler and will result in confusion if the 360 assembler found is not actually the GNU assembler. (Confusion may also 361 result if the compiler finds the GNU assembler but has not been 362 configured with <samp><span class="option">--with-gnu-as</span></samp>.) If you have more than one 363 assembler installed on your system, you may want to use this option in 364 connection with <samp><span class="option">--with-as=</span><var>pathname</var></samp> or 365 <samp><span class="option">--with-build-time-tools=</span><var>pathname</var></samp>. 366 367 <p>The following systems are the only ones where it makes a difference 368 whether you use the GNU assembler. On any other system, 369 <samp><span class="option">--with-gnu-as</span></samp> has no effect. 370 371 <ul> 372 <li>‘<samp><span class="samp">hppa1.0-</span><var>any</var><span class="samp">-</span><var>any</var></samp>’ 373 <li>‘<samp><span class="samp">hppa1.1-</span><var>any</var><span class="samp">-</span><var>any</var></samp>’ 374 <li>‘<samp><span class="samp">sparc-sun-solaris2.</span><var>any</var></samp>’ 375 <li>‘<samp><span class="samp">sparc64-</span><var>any</var><span class="samp">-solaris2.</span><var>any</var></samp>’ 376 </ul> 377 378 <br><dt><code><a name="with-as"></a>--with-as=</code><var>pathname</var><dd>Specify that the compiler should use the assembler pointed to by 379 <var>pathname</var>, rather than the one found by the standard rules to find 380 an assembler, which are: 381 <ul> 382 <li>Unless GCC is being built with a cross compiler, check the 383 <samp><var>libexec</var><span class="file">/gcc/</span><var>target</var><span class="file">/</span><var>version</var></samp> directory. 384 <var>libexec</var> defaults to <samp><var>exec-prefix</var><span class="file">/libexec</span></samp>; 385 <var>exec-prefix</var> defaults to <var>prefix</var>, which 386 defaults to <samp><span class="file">/usr/local</span></samp> unless overridden by the 387 <samp><span class="option">--prefix=</span><var>pathname</var></samp> switch described above. <var>target</var> 388 is the target system triple, such as ‘<samp><span class="samp">sparc-sun-solaris2.7</span></samp>’, and 389 <var>version</var> denotes the GCC version, such as 3.0. 390 391 <li>If the target system is the same that you are building on, check 392 operating system specific directories (e.g. <samp><span class="file">/usr/ccs/bin</span></samp> on 393 Sun Solaris 2). 394 395 <li>Check in the <samp><span class="env">PATH</span></samp> for a tool whose name is prefixed by the 396 target system triple. 397 398 <li>Check in the <samp><span class="env">PATH</span></samp> for a tool whose name is not prefixed by the 399 target system triple, if the host and target system triple are 400 the same (in other words, we use a host tool if it can be used for 401 the target as well). 402 </ul> 403 404 <p>You may want to use <samp><span class="option">--with-as</span></samp> if no assembler 405 is installed in the directories listed above, or if you have multiple 406 assemblers installed and want to choose one that is not found by the 407 above rules. 408 409 <br><dt><code><a name="with-gnu-ld"></a>--with-gnu-ld</code><dd>Same as <a href="#with-gnu-as"><samp><span class="option">--with-gnu-as</span></samp></a> 410 but for the linker. 411 412 <br><dt><code>--with-ld=</code><var>pathname</var><dd>Same as <a href="#with-as"><samp><span class="option">--with-as</span></samp></a> 413 but for the linker. 414 415 <br><dt><code>--with-stabs</code><dd>Specify that stabs debugging 416 information should be used instead of whatever format the host normally 417 uses. Normally GCC uses the same debug format as the host system. 418 419 <p>On MIPS based systems and on Alphas, you must specify whether you want 420 GCC to create the normal ECOFF debugging format, or to use BSD-style 421 stabs passed through the ECOFF symbol table. The normal ECOFF debug 422 format cannot fully handle languages other than C. BSD stabs format can 423 handle other languages, but it only works with the GNU debugger GDB. 424 425 <p>Normally, GCC uses the ECOFF debugging format by default; if you 426 prefer BSD stabs, specify <samp><span class="option">--with-stabs</span></samp> when you configure GCC. 427 428 <p>No matter which default you choose when you configure GCC, the user 429 can use the <samp><span class="option">-gcoff</span></samp> and <samp><span class="option">-gstabs+</span></samp> options to specify explicitly 430 the debug format for a particular compilation. 431 432 <p><samp><span class="option">--with-stabs</span></samp> is meaningful on the ISC system on the 386, also, if 433 <samp><span class="option">--with-gas</span></samp> is used. It selects use of stabs debugging 434 information embedded in COFF output. This kind of debugging information 435 supports C++ well; ordinary COFF debugging information does not. 436 437 <p><samp><span class="option">--with-stabs</span></samp> is also meaningful on 386 systems running SVR4. It 438 selects use of stabs debugging information embedded in ELF output. The 439 C++ compiler currently (2.6.0) does not support the DWARF debugging 440 information normally used on 386 SVR4 platforms; stabs provide a 441 workable alternative. This requires gas and gdb, as the normal SVR4 442 tools can not generate or interpret stabs. 443 444 <br><dt><code>--with-tls=</code><var>dialect</var><dd>Specify the default TLS dialect, for systems were there is a choice. 445 For ARM targets, possible values for <var>dialect</var> are <code>gnu</code> or 446 <code>gnu2</code>, which select between the original GNU dialect and the GNU TLS 447 descriptor-based dialect. 448 449 <br><dt><code>--enable-multiarch</code><dd>Specify whether to enable or disable multiarch support. The default is 450 to check for glibc start files in a multiarch location, and enable it 451 if the files are found. The auto detection is enabled for native builds, 452 and for cross builds configured with <samp><span class="option">--with-sysroot</span></samp>, and without 453 <samp><span class="option">--with-native-system-header-dir</span></samp>. 454 More documentation about multiarch can be found at 455 <a href="https://wiki.debian.org/Multiarch">https://wiki.debian.org/Multiarch</a>. 456 457 <br><dt><code>--enable-sjlj-exceptions</code><dd>Force use of the <code>setjmp</code>/<code>longjmp</code>-based scheme for exceptions. 458 ‘<samp><span class="samp">configure</span></samp>’ ordinarily picks the correct value based on the platform. 459 Only use this option if you are sure you need a different setting. 460 461 <br><dt><code>--enable-vtable-verify</code><dd>Specify whether to enable or disable the vtable verification feature. 462 Enabling this feature causes libstdc++ to be built with its virtual calls 463 in verifiable mode. This means that, when linked with libvtv, every 464 virtual call in libstdc++ will verify the vtable pointer through which the 465 call will be made before actually making the call. If not linked with libvtv, 466 the verifier will call stub functions (in libstdc++ itself) and do nothing. 467 If vtable verification is disabled, then libstdc++ is not built with its 468 virtual calls in verifiable mode at all. However the libvtv library will 469 still be built (see <samp><span class="option">--disable-libvtv</span></samp> to turn off building libvtv). 470 <samp><span class="option">--disable-vtable-verify</span></samp> is the default. 471 472 <br><dt><code>--disable-multilib</code><dd>Specify that multiple target 473 libraries to support different target variants, calling 474 conventions, etc. should not be built. The default is to build a 475 predefined set of them. 476 477 <p>Some targets provide finer-grained control over which multilibs are built 478 (e.g., <samp><span class="option">--disable-softfloat</span></samp>): 479 <dl> 480 <dt><code>arm-*-*</code><dd>fpu, 26bit, underscore, interwork, biendian, nofmult. 481 482 <br><dt><code>m68*-*-*</code><dd>softfloat, m68881, m68000, m68020. 483 484 <br><dt><code>mips*-*-*</code><dd>single-float, biendian, softfloat. 485 486 <br><dt><code>powerpc*-*-*, rs6000*-*-*</code><dd>aix64, pthread, softfloat, powercpu, powerpccpu, powerpcos, biendian, 487 sysv, aix. 488 489 </dl> 490 491 <br><dt><code>--with-multilib-list=</code><var>list</var><dt><code>--without-multilib-list</code><dd>Specify what multilibs to build. <var>list</var> is a comma separated list of 492 values, possibly consisting of a single value. Currently only implemented 493 for arm*-*-*, sh*-*-* and x86-64-*-linux*. The accepted values and meaning 494 for each target is given below. 495 496 <dl> 497 <dt><code>arm*-*-*</code><dd><var>list</var> is one of<code>default</code>, <code>aprofile</code> or <code>rmprofile</code>. 498 Specifying <code>default</code> is equivalent to omitting this option, ie. only the 499 default runtime library will be enabled. Specifying <code>aprofile</code> or 500 <code>rmprofile</code> builds multilibs for a combination of ISA, architecture, 501 FPU available and floating-point ABI. 502 503 <p>The table below gives the combination of ISAs, architectures, FPUs and 504 floating-point ABIs for which multilibs are built for each accepted value. 505 506 <p><table summary=""><tr align="left"><td valign="top" width="15%">Option </td><td valign="top" width="28%">aprofile </td><td valign="top" width="30%">rmprofile 507 <br></td></tr><tr align="left"><td valign="top" width="15%">ISAs 508 </td><td valign="top" width="28%"><code>-marm</code> and <code>-mthumb</code> 509 </td><td valign="top" width="30%"><code>-mthumb</code> 510 <br></td></tr><tr align="left"><td valign="top" width="15%">Architectures<br><br><br><br><br><br> 511 </td><td valign="top" width="28%">default architecture<br> 512 <code>-march=armv7-a</code><br> 513 <code>-march=armv7ve</code><br> 514 <code>-march=armv8-a</code><br><br><br> 515 </td><td valign="top" width="30%">default architecture<br> 516 <code>-march=armv6s-m</code><br> 517 <code>-march=armv7-m</code><br> 518 <code>-march=armv7e-m</code><br> 519 <code>-march=armv8-m.base</code><br> 520 <code>-march=armv8-m.main</code><br> 521 <code>-march=armv7</code> 522 <br></td></tr><tr align="left"><td valign="top" width="15%">FPUs<br><br><br><br><br> 523 </td><td valign="top" width="28%">none<br> 524 <code>-mfpu=vfpv3-d16</code><br> 525 <code>-mfpu=neon</code><br> 526 <code>-mfpu=vfpv4-d16</code><br> 527 <code>-mfpu=neon-vfpv4</code><br> 528 <code>-mfpu=neon-fp-armv8</code> 529 </td><td valign="top" width="30%">none<br> 530 <code>-mfpu=vfpv3-d16</code><br> 531 <code>-mfpu=fpv4-sp-d16</code><br> 532 <code>-mfpu=fpv5-sp-d16</code><br> 533 <code>-mfpu=fpv5-d16</code><br> 534 <br></td></tr><tr align="left"><td valign="top" width="15%">floating-point ABIs<br><br> 535 </td><td valign="top" width="28%"><code>-mfloat-abi=soft</code><br> 536 <code>-mfloat-abi=softfp</code><br> 537 <code>-mfloat-abi=hard</code> 538 </td><td valign="top" width="30%"><code>-mfloat-abi=soft</code><br> 539 <code>-mfloat-abi=softfp</code><br> 540 <code>-mfloat-abi=hard</code> 541 <br></td></tr></table> 542 543 <br><dt><code>sh*-*-*</code><dd><var>list</var> is a comma separated list of CPU names. These must be of the 544 form <code>sh*</code> or <code>m*</code> (in which case they match the compiler option 545 for that processor). The list should not contain any endian options - 546 these are handled by <samp><span class="option">--with-endian</span></samp>. 547 548 <p>If <var>list</var> is empty, then there will be no multilibs for extra 549 processors. The multilib for the secondary endian remains enabled. 550 551 <p>As a special case, if an entry in the list starts with a <code>!</code> 552 (exclamation point), then it is added to the list of excluded multilibs. 553 Entries of this sort should be compatible with ‘<samp><span class="samp">MULTILIB_EXCLUDES</span></samp>’ 554 (once the leading <code>!</code> has been stripped). 555 556 <p>If <samp><span class="option">--with-multilib-list</span></samp> is not given, then a default set of 557 multilibs is selected based on the value of <samp><span class="option">--target</span></samp>. This is 558 usually the complete set of libraries, but some targets imply a more 559 specialized subset. 560 561 <p>Example 1: to configure a compiler for SH4A only, but supporting both 562 endians, with little endian being the default: 563 <pre class="smallexample"> --with-cpu=sh4a --with-endian=little,big --with-multilib-list= 564 </pre> 565 <p>Example 2: to configure a compiler for both SH4A and SH4AL-DSP, but with 566 only little endian SH4AL: 567 <pre class="smallexample"> --with-cpu=sh4a --with-endian=little,big \ 568 --with-multilib-list=sh4al,!mb/m4al 569 </pre> 570 <br><dt><code>x86-64-*-linux*</code><dd><var>list</var> is a comma separated list of <code>m32</code>, <code>m64</code> and 571 <code>mx32</code> to enable 32-bit, 64-bit and x32 run-time libraries, 572 respectively. If <var>list</var> is empty, then there will be no multilibs 573 and only the default run-time library will be enabled. 574 575 <p>If <samp><span class="option">--with-multilib-list</span></samp> is not given, then only 32-bit and 576 64-bit run-time libraries will be enabled. 577 </dl> 578 579 <br><dt><code>--with-endian=</code><var>endians</var><dd>Specify what endians to use. 580 Currently only implemented for sh*-*-*. 581 582 <p><var>endians</var> may be one of the following: 583 <dl> 584 <dt><code>big</code><dd>Use big endian exclusively. 585 <br><dt><code>little</code><dd>Use little endian exclusively. 586 <br><dt><code>big,little</code><dd>Use big endian by default. Provide a multilib for little endian. 587 <br><dt><code>little,big</code><dd>Use little endian by default. Provide a multilib for big endian. 588 </dl> 589 590 <br><dt><code>--enable-threads</code><dd>Specify that the target 591 supports threads. This affects the Objective-C compiler and runtime 592 library, and exception handling for other languages like C++. 593 On some systems, this is the default. 594 595 <p>In general, the best (and, in many cases, the only known) threading 596 model available will be configured for use. Beware that on some 597 systems, GCC has not been taught what threading models are generally 598 available for the system. In this case, <samp><span class="option">--enable-threads</span></samp> is an 599 alias for <samp><span class="option">--enable-threads=single</span></samp>. 600 601 <br><dt><code>--disable-threads</code><dd>Specify that threading support should be disabled for the system. 602 This is an alias for <samp><span class="option">--enable-threads=single</span></samp>. 603 604 <br><dt><code>--enable-threads=</code><var>lib</var><dd>Specify that 605 <var>lib</var> is the thread support library. This affects the Objective-C 606 compiler and runtime library, and exception handling for other languages 607 like C++. The possibilities for <var>lib</var> are: 608 609 <dl> 610 <dt><code>aix</code><dd>AIX thread support. 611 <br><dt><code>dce</code><dd>DCE thread support. 612 <br><dt><code>lynx</code><dd>LynxOS thread support. 613 <br><dt><code>mipssde</code><dd>MIPS SDE thread support. 614 <br><dt><code>no</code><dd>This is an alias for ‘<samp><span class="samp">single</span></samp>’. 615 <br><dt><code>posix</code><dd>Generic POSIX/Unix98 thread support. 616 <br><dt><code>rtems</code><dd>RTEMS thread support. 617 <br><dt><code>single</code><dd>Disable thread support, should work for all platforms. 618 <br><dt><code>tpf</code><dd>TPF thread support. 619 <br><dt><code>vxworks</code><dd>VxWorks thread support. 620 <br><dt><code>win32</code><dd>Microsoft Win32 API thread support. 621 </dl> 622 623 <br><dt><code>--enable-tls</code><dd>Specify that the target supports TLS (Thread Local Storage). Usually 624 configure can correctly determine if TLS is supported. In cases where 625 it guesses incorrectly, TLS can be explicitly enabled or disabled with 626 <samp><span class="option">--enable-tls</span></samp> or <samp><span class="option">--disable-tls</span></samp>. This can happen if 627 the assembler supports TLS but the C library does not, or if the 628 assumptions made by the configure test are incorrect. 629 630 <br><dt><code>--disable-tls</code><dd>Specify that the target does not support TLS. 631 This is an alias for <samp><span class="option">--enable-tls=no</span></samp>. 632 633 <br><dt><code>--with-cpu=</code><var>cpu</var><dt><code>--with-cpu-32=</code><var>cpu</var><dt><code>--with-cpu-64=</code><var>cpu</var><dd>Specify which cpu variant the compiler should generate code for by default. 634 <var>cpu</var> will be used as the default value of the <samp><span class="option">-mcpu=</span></samp> switch. 635 This option is only supported on some targets, including ARC, ARM, i386, M68k, 636 PowerPC, and SPARC. It is mandatory for ARC. The <samp><span class="option">--with-cpu-32</span></samp> and 637 <samp><span class="option">--with-cpu-64</span></samp> options specify separate default CPUs for 638 32-bit and 64-bit modes; these options are only supported for i386, 639 x86-64, PowerPC, and SPARC. 640 641 <br><dt><code>--with-schedule=</code><var>cpu</var><dt><code>--with-arch=</code><var>cpu</var><dt><code>--with-arch-32=</code><var>cpu</var><dt><code>--with-arch-64=</code><var>cpu</var><dt><code>--with-tune=</code><var>cpu</var><dt><code>--with-tune-32=</code><var>cpu</var><dt><code>--with-tune-64=</code><var>cpu</var><dt><code>--with-abi=</code><var>abi</var><dt><code>--with-fpu=</code><var>type</var><dt><code>--with-float=</code><var>type</var><dd>These configure options provide default values for the <samp><span class="option">-mschedule=</span></samp>, 642 <samp><span class="option">-march=</span></samp>, <samp><span class="option">-mtune=</span></samp>, <samp><span class="option">-mabi=</span></samp>, and <samp><span class="option">-mfpu=</span></samp> 643 options and for <samp><span class="option">-mhard-float</span></samp> or <samp><span class="option">-msoft-float</span></samp>. As with 644 <samp><span class="option">--with-cpu</span></samp>, which switches will be accepted and acceptable values 645 of the arguments depend on the target. 646 647 <br><dt><code>--with-mode=</code><var>mode</var><dd>Specify if the compiler should default to <samp><span class="option">-marm</span></samp> or <samp><span class="option">-mthumb</span></samp>. 648 This option is only supported on ARM targets. 649 650 <br><dt><code>--with-stack-offset=</code><var>num</var><dd>This option sets the default for the -mstack-offset=<var>num</var> option, 651 and will thus generally also control the setting of this option for 652 libraries. This option is only supported on Epiphany targets. 653 654 <br><dt><code>--with-fpmath=</code><var>isa</var><dd>This options sets <samp><span class="option">-mfpmath=sse</span></samp> by default and specifies the default 655 ISA for floating-point arithmetics. You can select either ‘<samp><span class="samp">sse</span></samp>’ which 656 enables <samp><span class="option">-msse2</span></samp> or ‘<samp><span class="samp">avx</span></samp>’ which enables <samp><span class="option">-mavx</span></samp> by default. 657 This option is only supported on i386 and x86-64 targets. 658 659 <br><dt><code>--with-fp-32=</code><var>mode</var><dd>On MIPS targets, set the default value for the <samp><span class="option">-mfp</span></samp> option when using 660 the o32 ABI. The possibilities for <var>mode</var> are: 661 <dl> 662 <dt><code>32</code><dd>Use the o32 FP32 ABI extension, as with the <samp><span class="option">-mfp32</span></samp> command-line 663 option. 664 <br><dt><code>xx</code><dd>Use the o32 FPXX ABI extension, as with the <samp><span class="option">-mfpxx</span></samp> command-line 665 option. 666 <br><dt><code>64</code><dd>Use the o32 FP64 ABI extension, as with the <samp><span class="option">-mfp64</span></samp> command-line 667 option. 668 </dl> 669 In the absence of this configuration option the default is to use the o32 670 FP32 ABI extension. 671 672 <br><dt><code>--with-odd-spreg-32</code><dd>On MIPS targets, set the <samp><span class="option">-modd-spreg</span></samp> option by default when using 673 the o32 ABI. 674 675 <br><dt><code>--without-odd-spreg-32</code><dd>On MIPS targets, set the <samp><span class="option">-mno-odd-spreg</span></samp> option by default when using 676 the o32 ABI. This is normally used in conjunction with 677 <samp><span class="option">--with-fp-32=64</span></samp> in order to target the o32 FP64A ABI extension. 678 679 <br><dt><code>--with-nan=</code><var>encoding</var><dd>On MIPS targets, set the default encoding convention to use for the 680 special not-a-number (NaN) IEEE 754 floating-point data. The 681 possibilities for <var>encoding</var> are: 682 <dl> 683 <dt><code>legacy</code><dd>Use the legacy encoding, as with the <samp><span class="option">-mnan=legacy</span></samp> command-line 684 option. 685 <br><dt><code>2008</code><dd>Use the 754-2008 encoding, as with the <samp><span class="option">-mnan=2008</span></samp> command-line 686 option. 687 </dl> 688 To use this configuration option you must have an assembler version 689 installed that supports the <samp><span class="option">-mnan=</span></samp> command-line option too. 690 In the absence of this configuration option the default convention is 691 the legacy encoding, as when neither of the <samp><span class="option">-mnan=2008</span></samp> and 692 <samp><span class="option">-mnan=legacy</span></samp> command-line options has been used. 693 694 <br><dt><code>--with-divide=</code><var>type</var><dd>Specify how the compiler should generate code for checking for 695 division by zero. This option is only supported on the MIPS target. 696 The possibilities for <var>type</var> are: 697 <dl> 698 <dt><code>traps</code><dd>Division by zero checks use conditional traps (this is the default on 699 systems that support conditional traps). 700 <br><dt><code>breaks</code><dd>Division by zero checks use the break instruction. 701 </dl> 702 703 <!-- If you make -with-llsc the default for additional targets, --> 704 <!-- update the -with-llsc description in the MIPS section below. --> 705 <br><dt><code>--with-llsc</code><dd>On MIPS targets, make <samp><span class="option">-mllsc</span></samp> the default when no 706 <samp><span class="option">-mno-llsc</span></samp> option is passed. This is the default for 707 Linux-based targets, as the kernel will emulate them if the ISA does 708 not provide them. 709 710 <br><dt><code>--without-llsc</code><dd>On MIPS targets, make <samp><span class="option">-mno-llsc</span></samp> the default when no 711 <samp><span class="option">-mllsc</span></samp> option is passed. 712 713 <br><dt><code>--with-synci</code><dd>On MIPS targets, make <samp><span class="option">-msynci</span></samp> the default when no 714 <samp><span class="option">-mno-synci</span></samp> option is passed. 715 716 <br><dt><code>--without-synci</code><dd>On MIPS targets, make <samp><span class="option">-mno-synci</span></samp> the default when no 717 <samp><span class="option">-msynci</span></samp> option is passed. This is the default. 718 719 <br><dt><code>--with-lxc1-sxc1</code><dd>On MIPS targets, make <samp><span class="option">-mlxc1-sxc1</span></samp> the default when no 720 <samp><span class="option">-mno-lxc1-sxc1</span></samp> option is passed. This is the default. 721 722 <br><dt><code>--without-lxc1-sxc1</code><dd>On MIPS targets, make <samp><span class="option">-mno-lxc1-sxc1</span></samp> the default when no 723 <samp><span class="option">-mlxc1-sxc1</span></samp> option is passed. The indexed load/store 724 instructions are not directly a problem but can lead to unexpected 725 behaviour when deployed in an application intended for a 32-bit address 726 space but run on a 64-bit processor. The issue is seen because all 727 known MIPS 64-bit Linux kernels execute o32 and n32 applications 728 with 64-bit addressing enabled which affects the overflow behaviour 729 of the indexed addressing mode. GCC will assume that ordinary 730 32-bit arithmetic overflow behaviour is the same whether performed 731 as an <code>addu</code> instruction or as part of the address calculation 732 in <code>lwxc1</code> type instructions. This assumption holds true in a 733 pure 32-bit environment and can hold true in a 64-bit environment if 734 the address space is accurately set to be 32-bit for o32 and n32. 735 736 <br><dt><code>--with-madd4</code><dd>On MIPS targets, make <samp><span class="option">-mmadd4</span></samp> the default when no 737 <samp><span class="option">-mno-madd4</span></samp> option is passed. This is the default. 738 739 <br><dt><code>--without-madd4</code><dd>On MIPS targets, make <samp><span class="option">-mno-madd4</span></samp> the default when no 740 <samp><span class="option">-mmadd4</span></samp> option is passed. The <code>madd4</code> instruction 741 family can be problematic when targeting a combination of cores that 742 implement these instructions differently. There are two known cores 743 that implement these as fused operations instead of unfused (where 744 unfused is normally expected). Disabling these instructions is the 745 only way to ensure compatible code is generated; this will incur 746 a performance penalty. 747 748 <br><dt><code>--with-mips-plt</code><dd>On MIPS targets, make use of copy relocations and PLTs. 749 These features are extensions to the traditional 750 SVR4-based MIPS ABIs and require support from GNU binutils 751 and the runtime C library. 752 753 <br><dt><code>--enable-__cxa_atexit</code><dd>Define if you want to use __cxa_atexit, rather than atexit, to 754 register C++ destructors for local statics and global objects. 755 This is essential for fully standards-compliant handling of 756 destructors, but requires __cxa_atexit in libc. This option is currently 757 only available on systems with GNU libc. When enabled, this will cause 758 <samp><span class="option">-fuse-cxa-atexit</span></samp> to be passed by default. 759 760 <br><dt><code>--enable-gnu-indirect-function</code><dd>Define if you want to enable the <code>ifunc</code> attribute. This option is 761 currently only available on systems with GNU libc on certain targets. 762 763 <br><dt><code>--enable-target-optspace</code><dd>Specify that target 764 libraries should be optimized for code space instead of code speed. 765 This is the default for the m32r platform. 766 767 <br><dt><code>--with-cpp-install-dir=</code><var>dirname</var><dd>Specify that the user visible <samp><span class="command">cpp</span></samp> program should be installed 768 in <samp><var>prefix</var><span class="file">/</span><var>dirname</var><span class="file">/cpp</span></samp>, in addition to <var>bindir</var>. 769 770 <br><dt><code>--enable-comdat</code><dd>Enable COMDAT group support. This is primarily used to override the 771 automatically detected value. 772 773 <br><dt><code>--enable-initfini-array</code><dd>Force the use of sections <code>.init_array</code> and <code>.fini_array</code> 774 (instead of <code>.init</code> and <code>.fini</code>) for constructors and 775 destructors. Option <samp><span class="option">--disable-initfini-array</span></samp> has the 776 opposite effect. If neither option is specified, the configure script 777 will try to guess whether the <code>.init_array</code> and 778 <code>.fini_array</code> sections are supported and, if they are, use them. 779 780 <br><dt><code>--enable-link-mutex</code><dd>When building GCC, use a mutex to avoid linking the compilers for 781 multiple languages at the same time, to avoid thrashing on build 782 systems with limited free memory. The default is not to use such a mutex. 783 784 <br><dt><code>--enable-maintainer-mode</code><dd>The build rules that regenerate the Autoconf and Automake output files as 785 well as the GCC master message catalog <samp><span class="file">gcc.pot</span></samp> are normally 786 disabled. This is because it can only be rebuilt if the complete source 787 tree is present. If you have changed the sources and want to rebuild the 788 catalog, configuring with <samp><span class="option">--enable-maintainer-mode</span></samp> will enable 789 this. Note that you need a recent version of the <code>gettext</code> tools 790 to do so. 791 792 <br><dt><code>--disable-bootstrap</code><dd>For a native build, the default configuration is to perform 793 a 3-stage bootstrap of the compiler when ‘<samp><span class="samp">make</span></samp>’ is invoked, 794 testing that GCC can compile itself correctly. If you want to disable 795 this process, you can configure with <samp><span class="option">--disable-bootstrap</span></samp>. 796 797 <br><dt><code>--enable-bootstrap</code><dd>In special cases, you may want to perform a 3-stage build 798 even if the target and host triplets are different. 799 This is possible when the host can run code compiled for 800 the target (e.g. host is i686-linux, target is i486-linux). 801 Starting from GCC 4.2, to do this you have to configure explicitly 802 with <samp><span class="option">--enable-bootstrap</span></samp>. 803 804 <br><dt><code>--enable-generated-files-in-srcdir</code><dd>Neither the .c and .h files that are generated from Bison and flex nor the 805 info manuals and man pages that are built from the .texi files are present 806 in the SVN development tree. When building GCC from that development tree, 807 or from one of our snapshots, those generated files are placed in your 808 build directory, which allows for the source to be in a readonly 809 directory. 810 811 <p>If you configure with <samp><span class="option">--enable-generated-files-in-srcdir</span></samp> then those 812 generated files will go into the source directory. This is mainly intended 813 for generating release or prerelease tarballs of the GCC sources, since it 814 is not a requirement that the users of source releases to have flex, Bison, 815 or makeinfo. 816 817 <br><dt><code>--enable-version-specific-runtime-libs</code><dd>Specify 818 that runtime libraries should be installed in the compiler specific 819 subdirectory (<samp><var>libdir</var><span class="file">/gcc</span></samp>) rather than the usual places. In 820 addition, ‘<samp><span class="samp">libstdc++</span></samp>’'s include files will be installed into 821 <samp><var>libdir</var></samp> unless you overruled it by using 822 <samp><span class="option">--with-gxx-include-dir=</span><var>dirname</var></samp>. Using this option is 823 particularly useful if you intend to use several versions of GCC in 824 parallel. This is currently supported by ‘<samp><span class="samp">libgfortran</span></samp>’, 825 ‘<samp><span class="samp">libstdc++</span></samp>’, and ‘<samp><span class="samp">libobjc</span></samp>’. 826 827 <br><dt><code><a name="WithAixSoname"></a>--with-aix-soname=‘</code><samp><span class="samp">aix</span></samp><code>’, ‘</code><samp><span class="samp">svr4</span></samp><code>’ or ‘</code><samp><span class="samp">both</span></samp><code>’</code><dd>Traditional AIX shared library versioning (versioned <code>Shared Object</code> 828 files as members of unversioned <code>Archive Library</code> files named 829 ‘<samp><span class="samp">lib.a</span></samp>’) causes numerous headaches for package managers. However, 830 <code>Import Files</code> as members of <code>Archive Library</code> files allow for 831 <strong>filename-based versioning</strong> of shared libraries as seen on Linux/SVR4, 832 where this is called the "SONAME". But as they prevent static linking, 833 <code>Import Files</code> may be used with <code>Runtime Linking</code> only, where the 834 linker does search for ‘<samp><span class="samp">libNAME.so</span></samp>’ before ‘<samp><span class="samp">libNAME.a</span></samp>’ library 835 filenames with the ‘<samp><span class="samp">-lNAME</span></samp>’ linker flag. 836 837 <p><a name="AixLdCommand"></a>For detailed information please refer to the AIX 838 <a href="https://www.ibm.com/support/knowledgecenter/search/%22the%20ld%20command%2C%20also%20called%20the%20linkage%20editor%20or%20binder%22">ld Command</a> reference. 839 840 <p>As long as shared library creation is enabled, upon: 841 <dl> 842 <dt><code>--with-aix-soname=aix</code><br><dt><code>--with-aix-soname=both</code><dd> A (traditional AIX) <code>Shared Archive Library</code> file is created: 843 <ul> 844 <li>using the ‘<samp><span class="samp">libNAME.a</span></samp>’ filename scheme 845 <li>with the <code>Shared Object</code> file as archive member named 846 ‘<samp><span class="samp">libNAME.so.V</span></samp>’ (except for ‘<samp><span class="samp">libgcc_s</span></samp>’, where the <code>Shared 847 Object</code> file is named ‘<samp><span class="samp">shr.o</span></samp>’ for backwards compatibility), which 848 <ul> 849 <li>is used for runtime loading from inside the ‘<samp><span class="samp">libNAME.a</span></samp>’ file 850 <li>is used for dynamic loading via 851 <code>dlopen("libNAME.a(libNAME.so.V)", RTLD_MEMBER)</code> 852 <li>is used for shared linking 853 <li>is used for static linking, so no separate <code>Static Archive 854 Library</code> file is needed 855 </ul> 856 </ul> 857 <br><dt><code>--with-aix-soname=both</code><br><dt><code>--with-aix-soname=svr4</code><dd> A (second) <code>Shared Archive Library</code> file is created: 858 <ul> 859 <li>using the ‘<samp><span class="samp">libNAME.so.V</span></samp>’ filename scheme 860 <li>with the <code>Shared Object</code> file as archive member named 861 ‘<samp><span class="samp">shr.o</span></samp>’, which 862 <ul> 863 <li>is created with the <code>-G linker flag</code> 864 <li>has the <code>F_LOADONLY</code> flag set 865 <li>is used for runtime loading from inside the ‘<samp><span class="samp">libNAME.so.V</span></samp>’ file 866 <li>is used for dynamic loading via <code>dlopen("libNAME.so.V(shr.o)", 867 RTLD_MEMBER)</code> 868 </ul> 869 <li>with the <code>Import File</code> as archive member named ‘<samp><span class="samp">shr.imp</span></samp>’, 870 which 871 <ul> 872 <li>refers to ‘<samp><span class="samp">libNAME.so.V(shr.o)</span></samp>’ as the "SONAME", to be recorded 873 in the <code>Loader Section</code> of subsequent binaries 874 <li>indicates whether ‘<samp><span class="samp">libNAME.so.V(shr.o)</span></samp>’ is 32 or 64 bit 875 <li>lists all the public symbols exported by ‘<samp><span class="samp">lib.so.V(shr.o)</span></samp>’, 876 eventually decorated with the <code>‘</code><samp><span class="samp">weak</span></samp><code>’ Keyword</code> 877 <li>is necessary for shared linking against ‘<samp><span class="samp">lib.so.V(shr.o)</span></samp>’ 878 </ul> 879 </ul> 880 A symbolic link using the ‘<samp><span class="samp">libNAME.so</span></samp>’ filename scheme is created: 881 <ul> 882 <li>pointing to the ‘<samp><span class="samp">libNAME.so.V</span></samp>’ <code>Shared Archive Library</code> file 883 <li>to permit the <code>ld Command</code> to find ‘<samp><span class="samp">lib.so.V(shr.imp)</span></samp>’ via 884 the ‘<samp><span class="samp">-lNAME</span></samp>’ argument (requires <code>Runtime Linking</code> to be enabled) 885 <li>to permit dynamic loading of ‘<samp><span class="samp">lib.so.V(shr.o)</span></samp>’ without the need 886 to specify the version number via <code>dlopen("libNAME.so(shr.o)", 887 RTLD_MEMBER)</code> 888 </ul> 889 </dl> 890 891 <p>As long as static library creation is enabled, upon: 892 <dl> 893 <dt><code>--with-aix-soname=svr4</code><dd> A <code>Static Archive Library</code> is created: 894 <ul> 895 <li>using the ‘<samp><span class="samp">libNAME.a</span></samp>’ filename scheme 896 <li>with all the <code>Static Object</code> files as archive members, which 897 <ul> 898 <li>are used for static linking 899 </ul> 900 </ul> 901 </dl> 902 903 <p>While the aix-soname=‘<samp><span class="samp">svr4</span></samp>’ option does not create <code>Shared Object</code> 904 files as members of unversioned <code>Archive Library</code> files any more, package 905 managers still are responsible to 906 <a href="./specific.html#TransferAixShobj">transfer</a> <code>Shared Object</code> files 907 found as member of a previously installed unversioned <code>Archive Library</code> 908 file into the newly installed <code>Archive Library</code> file with the same 909 filename. 910 911 <p><em>WARNING:</em> Creating <code>Shared Object</code> files with <code>Runtime Linking</code> 912 enabled may bloat the TOC, eventually leading to <code>TOC overflow</code> errors, 913 requiring the use of either the <samp><span class="option">-Wl,-bbigtoc</span></samp> linker flag (seen to 914 break with the <code>GDB</code> debugger) or some of the TOC-related compiler flags, 915 see “RS/6000 and PowerPC Options” in the main manual. 916 917 <p><samp><span class="option">--with-aix-soname</span></samp> is currently supported by ‘<samp><span class="samp">libgcc_s</span></samp>’ only, so 918 this option is still experimental and not for normal use yet. 919 920 <p>Default is the traditional behavior <samp><span class="option">--with-aix-soname=‘</span><samp><span class="samp">aix</span></samp><span class="option">’</span></samp>. 921 922 <br><dt><code>--enable-languages=</code><var>lang1</var><code>,</code><var>lang2</var><code>,...</code><dd>Specify that only a particular subset of compilers and 923 their runtime libraries should be built. For a list of valid values for 924 <var>langN</var> you can issue the following command in the 925 <samp><span class="file">gcc</span></samp> directory of your GCC source tree:<br> 926 <pre class="smallexample"> grep ^language= */config-lang.in 927 </pre> 928 <p>Currently, you can use any of the following: 929 <code>all</code>, <code>ada</code>, <code>c</code>, <code>c++</code>, <code>fortran</code>, 930 <code>go</code>, <code>jit</code>, <code>lto</code>, <code>objc</code>, <code>obj-c++</code>. 931 Building the Ada compiler has special requirements, see below. 932 If you do not pass this flag, or specify the option <code>all</code>, then all 933 default languages available in the <samp><span class="file">gcc</span></samp> sub-tree will be configured. 934 Ada, Go, Jit, and Objective-C++ are not default languages. LTO is not a 935 default language, but is built by default because <samp><span class="option">--enable-lto</span></samp> is 936 enabled by default. The other languages are default languages. 937 938 <br><dt><code>--enable-stage1-languages=</code><var>lang1</var><code>,</code><var>lang2</var><code>,...</code><dd>Specify that a particular subset of compilers and their runtime 939 libraries should be built with the system C compiler during stage 1 of 940 the bootstrap process, rather than only in later stages with the 941 bootstrapped C compiler. The list of valid values is the same as for 942 <samp><span class="option">--enable-languages</span></samp>, and the option <code>all</code> will select all 943 of the languages enabled by <samp><span class="option">--enable-languages</span></samp>. This option is 944 primarily useful for GCC development; for instance, when a development 945 version of the compiler cannot bootstrap due to compiler bugs, or when 946 one is debugging front ends other than the C front end. When this 947 option is used, one can then build the target libraries for the 948 specified languages with the stage-1 compiler by using <samp><span class="command">make 949 stage1-bubble all-target</span></samp>, or run the testsuite on the stage-1 compiler 950 for the specified languages using <samp><span class="command">make stage1-start check-gcc</span></samp>. 951 952 <br><dt><code>--disable-libada</code><dd>Specify that the run-time libraries and tools used by GNAT should not 953 be built. This can be useful for debugging, or for compatibility with 954 previous Ada build procedures, when it was required to explicitly 955 do a ‘<samp><span class="samp">make -C gcc gnatlib_and_tools</span></samp>’. 956 957 <br><dt><code>--disable-libsanitizer</code><dd>Specify that the run-time libraries for the various sanitizers should 958 not be built. 959 960 <br><dt><code>--disable-libssp</code><dd>Specify that the run-time libraries for stack smashing protection 961 should not be built. 962 963 <br><dt><code>--disable-libquadmath</code><dd>Specify that the GCC quad-precision math library should not be built. 964 On some systems, the library is required to be linkable when building 965 the Fortran front end, unless <samp><span class="option">--disable-libquadmath-support</span></samp> 966 is used. 967 968 <br><dt><code>--disable-libquadmath-support</code><dd>Specify that the Fortran front end and <code>libgfortran</code> do not add 969 support for <code>libquadmath</code> on systems supporting it. 970 971 <br><dt><code>--disable-libgomp</code><dd>Specify that the GNU Offloading and Multi Processing Runtime Library 972 should not be built. 973 974 <br><dt><code>--disable-libvtv</code><dd>Specify that the run-time libraries used by vtable verification 975 should not be built. 976 977 <br><dt><code>--with-dwarf2</code><dd>Specify that the compiler should 978 use DWARF 2 debugging information as the default. 979 980 <br><dt><code>--with-advance-toolchain=</code><var>at</var><dd>On 64-bit PowerPC Linux systems, configure the compiler to use the 981 header files, library files, and the dynamic linker from the Advance 982 Toolchain release <var>at</var> instead of the default versions that are 983 provided by the Linux distribution. In general, this option is 984 intended for the developers of GCC, and it is not intended for general 985 use. 986 987 <br><dt><code>--enable-targets=all</code><dt><code>--enable-targets=</code><var>target_list</var><dd>Some GCC targets, e.g. powerpc64-linux, build bi-arch compilers. 988 These are compilers that are able to generate either 64-bit or 32-bit 989 code. Typically, the corresponding 32-bit target, e.g. 990 powerpc-linux for powerpc64-linux, only generates 32-bit code. This 991 option enables the 32-bit target to be a bi-arch compiler, which is 992 useful when you want a bi-arch compiler that defaults to 32-bit, and 993 you are building a bi-arch or multi-arch binutils in a combined tree. 994 On mips-linux, this will build a tri-arch compiler (ABI o32/n32/64), 995 defaulted to o32. 996 Currently, this option only affects sparc-linux, powerpc-linux, x86-linux, 997 mips-linux and s390-linux. 998 999 <br><dt><code>--enable-default-pie</code><dd>Turn on <samp><span class="option">-fPIE</span></samp> and <samp><span class="option">-pie</span></samp> by default. 1000 1001 <br><dt><code>--enable-secureplt</code><dd>This option enables <samp><span class="option">-msecure-plt</span></samp> by default for powerpc-linux. 1002 See “RS/6000 and PowerPC Options” in the main manual 1003 1004 <br><dt><code>--enable-default-ssp</code><dd>Turn on <samp><span class="option">-fstack-protector-strong</span></samp> by default. 1005 1006 <br><dt><code>--enable-cld</code><dd>This option enables <samp><span class="option">-mcld</span></samp> by default for 32-bit x86 targets. 1007 See “i386 and x86-64 Options” in the main manual 1008 1009 <br><dt><code>--enable-win32-registry</code><dt><code>--enable-win32-registry=</code><var>key</var><dt><code>--disable-win32-registry</code><dd>The <samp><span class="option">--enable-win32-registry</span></samp> option enables Microsoft Windows-hosted GCC 1010 to look up installations paths in the registry using the following key: 1011 1012 <pre class="smallexample"> <code>HKEY_LOCAL_MACHINE\SOFTWARE\Free Software Foundation\</code><var>key</var> 1013 </pre> 1014 <p><var>key</var> defaults to GCC version number, and can be overridden by the 1015 <samp><span class="option">--enable-win32-registry=</span><var>key</var></samp> option. Vendors and distributors 1016 who use custom installers are encouraged to provide a different key, 1017 perhaps one comprised of vendor name and GCC version number, to 1018 avoid conflict with existing installations. This feature is enabled 1019 by default, and can be disabled by <samp><span class="option">--disable-win32-registry</span></samp> 1020 option. This option has no effect on the other hosts. 1021 1022 <br><dt><code>--nfp</code><dd>Specify that the machine does not have a floating point unit. This 1023 option only applies to ‘<samp><span class="samp">m68k-sun-sunos</span><var>n</var></samp>’. On any other 1024 system, <samp><span class="option">--nfp</span></samp> has no effect. 1025 1026 <br><dt><code>--enable-werror</code><dt><code>--disable-werror</code><dt><code>--enable-werror=yes</code><dt><code>--enable-werror=no</code><dd>When you specify this option, it controls whether certain files in the 1027 compiler are built with <samp><span class="option">-Werror</span></samp> in bootstrap stage2 and later. 1028 If you don't specify it, <samp><span class="option">-Werror</span></samp> is turned on for the main 1029 development trunk. However it defaults to off for release branches and 1030 final releases. The specific files which get <samp><span class="option">-Werror</span></samp> are 1031 controlled by the Makefiles. 1032 1033 <br><dt><code>--enable-checking</code><dt><code>--enable-checking=</code><var>list</var><dd>When you specify this option, the compiler is built to perform internal 1034 consistency checks of the requested complexity. This does not change the 1035 generated code, but adds error checking within the compiler. This will 1036 slow down the compiler and may only work properly if you are building 1037 the compiler with GCC. This is ‘<samp><span class="samp">yes,extra</span></samp>’ by default when building 1038 from SVN or snapshots, but ‘<samp><span class="samp">release</span></samp>’ for releases. The default 1039 for building the stage1 compiler is ‘<samp><span class="samp">yes</span></samp>’. More control 1040 over the checks may be had by specifying <var>list</var>. The categories of 1041 checks available are ‘<samp><span class="samp">yes</span></samp>’ (most common checks 1042 ‘<samp><span class="samp">assert,misc,tree,gc,rtlflag,runtime</span></samp>’), ‘<samp><span class="samp">no</span></samp>’ (no checks at 1043 all), ‘<samp><span class="samp">all</span></samp>’ (all but ‘<samp><span class="samp">valgrind</span></samp>’), ‘<samp><span class="samp">release</span></samp>’ (cheapest 1044 checks ‘<samp><span class="samp">assert,runtime</span></samp>’) or ‘<samp><span class="samp">none</span></samp>’ (same as ‘<samp><span class="samp">no</span></samp>’). 1045 Individual checks can be enabled with these flags ‘<samp><span class="samp">assert</span></samp>’, 1046 ‘<samp><span class="samp">df</span></samp>’, ‘<samp><span class="samp">fold</span></samp>’, ‘<samp><span class="samp">gc</span></samp>’, ‘<samp><span class="samp">gcac</span></samp>’, ‘<samp><span class="samp">misc</span></samp>’, ‘<samp><span class="samp">rtl</span></samp>’, 1047 ‘<samp><span class="samp">rtlflag</span></samp>’, ‘<samp><span class="samp">runtime</span></samp>’, ‘<samp><span class="samp">tree</span></samp>’, ‘<samp><span class="samp">extra</span></samp>’ and ‘<samp><span class="samp">valgrind</span></samp>’. 1048 ‘<samp><span class="samp">extra</span></samp>’ adds for ‘<samp><span class="samp">misc</span></samp>’ checking extra checks that might affect 1049 code generation and should therefore not differ between stage1 and later 1050 stages. 1051 1052 <p>The ‘<samp><span class="samp">valgrind</span></samp>’ check requires the external <samp><span class="command">valgrind</span></samp> 1053 simulator, available from <a href="http://valgrind.org/">http://valgrind.org/</a>. The 1054 ‘<samp><span class="samp">df</span></samp>’, ‘<samp><span class="samp">rtl</span></samp>’, ‘<samp><span class="samp">gcac</span></samp>’ and ‘<samp><span class="samp">valgrind</span></samp>’ checks are very expensive. 1055 To disable all checking, ‘<samp><span class="samp">--disable-checking</span></samp>’ or 1056 ‘<samp><span class="samp">--enable-checking=none</span></samp>’ must be explicitly requested. Disabling 1057 assertions will make the compiler and runtime slightly faster but 1058 increase the risk of undetected internal errors causing wrong code to be 1059 generated. 1060 1061 <br><dt><code>--disable-stage1-checking</code><dt><code>--enable-stage1-checking</code><dt><code>--enable-stage1-checking=</code><var>list</var><dd>If no <samp><span class="option">--enable-checking</span></samp> option is specified the stage1 1062 compiler will be built with ‘<samp><span class="samp">yes</span></samp>’ checking enabled, otherwise 1063 the stage1 checking flags are the same as specified by 1064 <samp><span class="option">--enable-checking</span></samp>. To build the stage1 compiler with 1065 different checking options use <samp><span class="option">--enable-stage1-checking</span></samp>. 1066 The list of checking options is the same as for <samp><span class="option">--enable-checking</span></samp>. 1067 If your system is too slow or too small to bootstrap a released compiler 1068 with checking for stage1 enabled, you can use ‘<samp><span class="samp">--disable-stage1-checking</span></samp>’ 1069 to disable checking for the stage1 compiler. 1070 1071 <br><dt><code>--enable-coverage</code><dt><code>--enable-coverage=</code><var>level</var><dd>With this option, the compiler is built to collect self coverage 1072 information, every time it is run. This is for internal development 1073 purposes, and only works when the compiler is being built with gcc. The 1074 <var>level</var> argument controls whether the compiler is built optimized or 1075 not, values are ‘<samp><span class="samp">opt</span></samp>’ and ‘<samp><span class="samp">noopt</span></samp>’. For coverage analysis you 1076 want to disable optimization, for performance analysis you want to 1077 enable optimization. When coverage is enabled, the default level is 1078 without optimization. 1079 1080 <br><dt><code>--enable-gather-detailed-mem-stats</code><dd>When this option is specified more detailed information on memory 1081 allocation is gathered. This information is printed when using 1082 <samp><span class="option">-fmem-report</span></samp>. 1083 1084 <br><dt><code>--enable-valgrind-annotations</code><dd>Mark selected memory related operations in the compiler when run under 1085 valgrind to suppress false positives. 1086 1087 <br><dt><code>--enable-nls</code><dt><code>--disable-nls</code><dd>The <samp><span class="option">--enable-nls</span></samp> option enables Native Language Support (NLS), 1088 which lets GCC output diagnostics in languages other than American 1089 English. Native Language Support is enabled by default if not doing a 1090 canadian cross build. The <samp><span class="option">--disable-nls</span></samp> option disables NLS. 1091 1092 <br><dt><code>--with-included-gettext</code><dd>If NLS is enabled, the <samp><span class="option">--with-included-gettext</span></samp> option causes the build 1093 procedure to prefer its copy of GNU <samp><span class="command">gettext</span></samp>. 1094 1095 <br><dt><code>--with-catgets</code><dd>If NLS is enabled, and if the host lacks <code>gettext</code> but has the 1096 inferior <code>catgets</code> interface, the GCC build procedure normally 1097 ignores <code>catgets</code> and instead uses GCC's copy of the GNU 1098 <code>gettext</code> library. The <samp><span class="option">--with-catgets</span></samp> option causes the 1099 build procedure to use the host's <code>catgets</code> in this situation. 1100 1101 <br><dt><code>--with-libiconv-prefix=</code><var>dir</var><dd>Search for libiconv header files in <samp><var>dir</var><span class="file">/include</span></samp> and 1102 libiconv library files in <samp><var>dir</var><span class="file">/lib</span></samp>. 1103 1104 <br><dt><code>--enable-obsolete</code><dd>Enable configuration for an obsoleted system. If you attempt to 1105 configure GCC for a system (build, host, or target) which has been 1106 obsoleted, and you do not specify this flag, configure will halt with an 1107 error message. 1108 1109 <p>All support for systems which have been obsoleted in one release of GCC 1110 is removed entirely in the next major release, unless someone steps 1111 forward to maintain the port. 1112 1113 <br><dt><code>--enable-decimal-float</code><dt><code>--enable-decimal-float=yes</code><dt><code>--enable-decimal-float=no</code><dt><code>--enable-decimal-float=bid</code><dt><code>--enable-decimal-float=dpd</code><dt><code>--disable-decimal-float</code><dd>Enable (or disable) support for the C decimal floating point extension 1114 that is in the IEEE 754-2008 standard. This is enabled by default only 1115 on PowerPC, i386, and x86_64 GNU/Linux systems. Other systems may also 1116 support it, but require the user to specifically enable it. You can 1117 optionally control which decimal floating point format is used (either 1118 ‘<samp><span class="samp">bid</span></samp>’ or ‘<samp><span class="samp">dpd</span></samp>’). The ‘<samp><span class="samp">bid</span></samp>’ (binary integer decimal) 1119 format is default on i386 and x86_64 systems, and the ‘<samp><span class="samp">dpd</span></samp>’ 1120 (densely packed decimal) format is default on PowerPC systems. 1121 1122 <br><dt><code>--enable-fixed-point</code><dt><code>--disable-fixed-point</code><dd>Enable (or disable) support for C fixed-point arithmetic. 1123 This option is enabled by default for some targets (such as MIPS) which 1124 have hardware-support for fixed-point operations. On other targets, you 1125 may enable this option manually. 1126 1127 <br><dt><code>--with-long-double-128</code><dd>Specify if <code>long double</code> type should be 128-bit by default on selected 1128 GNU/Linux architectures. If using <code>--without-long-double-128</code>, 1129 <code>long double</code> will be by default 64-bit, the same as <code>double</code> type. 1130 When neither of these configure options are used, the default will be 1131 128-bit <code>long double</code> when built against GNU C Library 2.4 and later, 1132 64-bit <code>long double</code> otherwise. 1133 1134 <br><dt><code>--enable-fdpic</code><dd>On SH Linux systems, generate ELF FDPIC code. 1135 1136 <br><dt><code>--with-gmp=</code><var>pathname</var><dt><code>--with-gmp-include=</code><var>pathname</var><dt><code>--with-gmp-lib=</code><var>pathname</var><dt><code>--with-mpfr=</code><var>pathname</var><dt><code>--with-mpfr-include=</code><var>pathname</var><dt><code>--with-mpfr-lib=</code><var>pathname</var><dt><code>--with-mpc=</code><var>pathname</var><dt><code>--with-mpc-include=</code><var>pathname</var><dt><code>--with-mpc-lib=</code><var>pathname</var><dd>If you want to build GCC but do not have the GMP library, the MPFR 1137 library and/or the MPC library installed in a standard location and 1138 do not have their sources present in the GCC source tree then you 1139 can explicitly specify the directory where they are installed 1140 (‘<samp><span class="samp">--with-gmp=</span><var>gmpinstalldir</var></samp>’, 1141 ‘<samp><span class="samp">--with-mpfr=</span><var>mpfrinstalldir</var></samp>’, 1142 ‘<samp><span class="samp">--with-mpc=</span><var>mpcinstalldir</var></samp>’). The 1143 <samp><span class="option">--with-gmp=</span><var>gmpinstalldir</var></samp> option is shorthand for 1144 <samp><span class="option">--with-gmp-lib=</span><var>gmpinstalldir</var><span class="option">/lib</span></samp> and 1145 <samp><span class="option">--with-gmp-include=</span><var>gmpinstalldir</var><span class="option">/include</span></samp>. Likewise the 1146 <samp><span class="option">--with-mpfr=</span><var>mpfrinstalldir</var></samp> option is shorthand for 1147 <samp><span class="option">--with-mpfr-lib=</span><var>mpfrinstalldir</var><span class="option">/lib</span></samp> and 1148 <samp><span class="option">--with-mpfr-include=</span><var>mpfrinstalldir</var><span class="option">/include</span></samp>, also the 1149 <samp><span class="option">--with-mpc=</span><var>mpcinstalldir</var></samp> option is shorthand for 1150 <samp><span class="option">--with-mpc-lib=</span><var>mpcinstalldir</var><span class="option">/lib</span></samp> and 1151 <samp><span class="option">--with-mpc-include=</span><var>mpcinstalldir</var><span class="option">/include</span></samp>. If these 1152 shorthand assumptions are not correct, you can use the explicit 1153 include and lib options directly. You might also need to ensure the 1154 shared libraries can be found by the dynamic linker when building and 1155 using GCC, for example by setting the runtime shared library path 1156 variable (<samp><span class="env">LD_LIBRARY_PATH</span></samp> on GNU/Linux and Solaris systems). 1157 1158 <p>These flags are applicable to the host platform only. When building 1159 a cross compiler, they will not be used to configure target libraries. 1160 1161 <br><dt><code>--with-isl=</code><var>pathname</var><dt><code>--with-isl-include=</code><var>pathname</var><dt><code>--with-isl-lib=</code><var>pathname</var><dd>If you do not have the isl library installed in a standard location and you 1162 want to build GCC, you can explicitly specify the directory where it is 1163 installed (‘<samp><span class="samp">--with-isl=</span><var>islinstalldir</var></samp>’). The 1164 <samp><span class="option">--with-isl=</span><var>islinstalldir</var></samp> option is shorthand for 1165 <samp><span class="option">--with-isl-lib=</span><var>islinstalldir</var><span class="option">/lib</span></samp> and 1166 <samp><span class="option">--with-isl-include=</span><var>islinstalldir</var><span class="option">/include</span></samp>. If this 1167 shorthand assumption is not correct, you can use the explicit 1168 include and lib options directly. 1169 1170 <p>These flags are applicable to the host platform only. When building 1171 a cross compiler, they will not be used to configure target libraries. 1172 1173 <br><dt><code>--with-stage1-ldflags=</code><var>flags</var><dd>This option may be used to set linker flags to be used when linking 1174 stage 1 of GCC. These are also used when linking GCC if configured with 1175 <samp><span class="option">--disable-bootstrap</span></samp>. If <samp><span class="option">--with-stage1-libs</span></samp> is not set to a 1176 value, then the default is ‘<samp><span class="samp">-static-libstdc++ -static-libgcc</span></samp>’, if 1177 supported. 1178 1179 <br><dt><code>--with-stage1-libs=</code><var>libs</var><dd>This option may be used to set libraries to be used when linking stage 1 1180 of GCC. These are also used when linking GCC if configured with 1181 <samp><span class="option">--disable-bootstrap</span></samp>. 1182 1183 <br><dt><code>--with-boot-ldflags=</code><var>flags</var><dd>This option may be used to set linker flags to be used when linking 1184 stage 2 and later when bootstrapping GCC. If –with-boot-libs 1185 is not is set to a value, then the default is 1186 ‘<samp><span class="samp">-static-libstdc++ -static-libgcc</span></samp>’. 1187 1188 <br><dt><code>--with-boot-libs=</code><var>libs</var><dd>This option may be used to set libraries to be used when linking stage 2 1189 and later when bootstrapping GCC. 1190 1191 <br><dt><code>--with-debug-prefix-map=</code><var>map</var><dd>Convert source directory names using <samp><span class="option">-fdebug-prefix-map</span></samp> when 1192 building runtime libraries. ‘<samp><var>map</var></samp>’ is a space-separated 1193 list of maps of the form ‘<samp><var>old</var><span class="samp">=</span><var>new</var></samp>’. 1194 1195 <br><dt><code>--enable-linker-build-id</code><dd>Tells GCC to pass <samp><span class="option">--build-id</span></samp> option to the linker for all final 1196 links (links performed without the <samp><span class="option">-r</span></samp> or <samp><span class="option">--relocatable</span></samp> 1197 option), if the linker supports it. If you specify 1198 <samp><span class="option">--enable-linker-build-id</span></samp>, but your linker does not 1199 support <samp><span class="option">--build-id</span></samp> option, a warning is issued and the 1200 <samp><span class="option">--enable-linker-build-id</span></samp> option is ignored. The default is off. 1201 1202 <br><dt><code>--with-linker-hash-style=</code><var>choice</var><dd>Tells GCC to pass <samp><span class="option">--hash-style=</span><var>choice</var></samp> option to the 1203 linker for all final links. <var>choice</var> can be one of 1204 ‘<samp><span class="samp">sysv</span></samp>’, ‘<samp><span class="samp">gnu</span></samp>’, and ‘<samp><span class="samp">both</span></samp>’ where ‘<samp><span class="samp">sysv</span></samp>’ is the default. 1205 1206 <br><dt><code>--enable-gnu-unique-object</code><dt><code>--disable-gnu-unique-object</code><dd>Tells GCC to use the gnu_unique_object relocation for C++ template 1207 static data members and inline function local statics. Enabled by 1208 default for a toolchain with an assembler that accepts it and 1209 GLIBC 2.11 or above, otherwise disabled. 1210 1211 <br><dt><code>--with-diagnostics-color=</code><var>choice</var><dd>Tells GCC to use <var>choice</var> as the default for <samp><span class="option">-fdiagnostics-color=</span></samp> 1212 option (if not used explicitly on the command line). <var>choice</var> 1213 can be one of ‘<samp><span class="samp">never</span></samp>’, ‘<samp><span class="samp">auto</span></samp>’, ‘<samp><span class="samp">always</span></samp>’, and ‘<samp><span class="samp">auto-if-env</span></samp>’ 1214 where ‘<samp><span class="samp">auto</span></samp>’ is the default. ‘<samp><span class="samp">auto-if-env</span></samp>’ means that 1215 <samp><span class="option">-fdiagnostics-color=auto</span></samp> will be the default if <code>GCC_COLORS</code> 1216 is present and non-empty in the environment, and 1217 <samp><span class="option">-fdiagnostics-color=never</span></samp> otherwise. 1218 1219 <br><dt><code>--enable-lto</code><dt><code>--disable-lto</code><dd>Enable support for link-time optimization (LTO). This is enabled by 1220 default, and may be disabled using <samp><span class="option">--disable-lto</span></samp>. 1221 1222 <br><dt><code>--enable-linker-plugin-configure-flags=FLAGS</code><dt><code>--enable-linker-plugin-flags=FLAGS</code><dd>By default, linker plugins (such as the LTO plugin) are built for the 1223 host system architecture. For the case that the linker has a 1224 different (but run-time compatible) architecture, these flags can be 1225 specified to build plugins that are compatible to the linker. For 1226 example, if you are building GCC for a 64-bit x86_64 1227 (‘<samp><span class="samp">x86_64-unknown-linux-gnu</span></samp>’) host system, but have a 32-bit x86 1228 GNU/Linux (‘<samp><span class="samp">i686-pc-linux-gnu</span></samp>’) linker executable (which is 1229 executable on the former system), you can configure GCC as follows for 1230 getting compatible linker plugins: 1231 1232 <pre class="smallexample"> % <var>srcdir</var>/configure \ 1233 --host=x86_64-unknown-linux-gnu \ 1234 --enable-linker-plugin-configure-flags=--host=i686-pc-linux-gnu \ 1235 --enable-linker-plugin-flags='CC=gcc\ -m32\ -Wl,-rpath,[...]/i686-pc-linux-gnu/lib' 1236 </pre> 1237 <br><dt><code>--with-plugin-ld=</code><var>pathname</var><dd>Enable an alternate linker to be used at link-time optimization (LTO) 1238 link time when <samp><span class="option">-fuse-linker-plugin</span></samp> is enabled. 1239 This linker should have plugin support such as gold starting with 1240 version 2.20 or GNU ld starting with version 2.21. 1241 See <samp><span class="option">-fuse-linker-plugin</span></samp> for details. 1242 1243 <br><dt><code>--enable-canonical-system-headers</code><dt><code>--disable-canonical-system-headers</code><dd>Enable system header path canonicalization for <samp><span class="file">libcpp</span></samp>. This can 1244 produce shorter header file paths in diagnostics and dependency output 1245 files, but these changed header paths may conflict with some compilation 1246 environments. Enabled by default, and may be disabled using 1247 <samp><span class="option">--disable-canonical-system-headers</span></samp>. 1248 1249 <br><dt><code>--with-glibc-version=</code><var>major</var><code>.</code><var>minor</var><dd>Tell GCC that when the GNU C Library (glibc) is used on the target it 1250 will be version <var>major</var>.<var>minor</var> or later. Normally this can 1251 be detected from the C library's header files, but this option may be 1252 needed when bootstrapping a cross toolchain without the header files 1253 available for building the initial bootstrap compiler. 1254 1255 <p>If GCC is configured with some multilibs that use glibc and some that 1256 do not, this option applies only to the multilibs that use glibc. 1257 However, such configurations may not work well as not all the relevant 1258 configuration in GCC is on a per-multilib basis. 1259 1260 <br><dt><code>--enable-as-accelerator-for=</code><var>target</var><dd>Build as offload target compiler. Specify offload host triple by <var>target</var>. 1261 1262 <br><dt><code>--enable-offload-targets=</code><var>target1</var><code>[=</code><var>path1</var><code>],...,</code><var>targetN</var><code>[=</code><var>pathN</var><code>]</code><dd>Enable offloading to targets <var>target1</var>, <small class="dots">...</small>, <var>targetN</var>. 1263 Offload compilers are expected to be already installed. Default search 1264 path for them is <samp><var>exec-prefix</var></samp>, but it can be changed by 1265 specifying paths <var>path1</var>, <small class="dots">...</small>, <var>pathN</var>. 1266 1267 <pre class="smallexample"> % <var>srcdir</var>/configure \ 1268 --enable-offload-target=i686-unknown-linux-gnu=/path/to/i686/compiler,x86_64-pc-linux-gnu 1269 </pre> 1270 <p>If ‘<samp><span class="samp">hsa</span></samp>’ is specified as one of the targets, the compiler will be 1271 built with support for HSA GPU accelerators. Because the same 1272 compiler will emit the accelerator code, no path should be specified. 1273 1274 <br><dt><code>--with-hsa-runtime=</code><var>pathname</var><dt><code>--with-hsa-runtime-include=</code><var>pathname</var><dt><code>--with-hsa-runtime-lib=</code><var>pathname</var><dd> 1275 If you configure GCC with HSA offloading but do not have the HSA 1276 run-time library installed in a standard location then you can 1277 explicitly specify the directory where they are installed. The 1278 <samp><span class="option">--with-hsa-runtime=</span><var>hsainstalldir</var></samp> option is a 1279 shorthand for 1280 <samp><span class="option">--with-hsa-runtime-lib=</span><var>hsainstalldir</var><span class="option">/lib</span></samp> and 1281 <samp><span class="option">--with-hsa-runtime-include=</span><var>hsainstalldir</var><span class="option">/include</span></samp>. 1282 </dl> 1283 1284 <h4 class="subheading"><a name="TOC3"></a>Cross-Compiler-Specific Options</h4> 1285 1286 <p>The following options only apply to building cross compilers. 1287 1288 <dl> 1289 <dt><code>--with-sysroot</code><dt><code>--with-sysroot=</code><var>dir</var><dd>Tells GCC to consider <var>dir</var> as the root of a tree that contains 1290 (a subset of) the root filesystem of the target operating system. 1291 Target system headers, libraries and run-time object files will be 1292 searched for in there. More specifically, this acts as if 1293 <samp><span class="option">--sysroot=</span><var>dir</var></samp> was added to the default options of the built 1294 compiler. The specified directory is not copied into the 1295 install tree, unlike the options <samp><span class="option">--with-headers</span></samp> and 1296 <samp><span class="option">--with-libs</span></samp> that this option obsoletes. The default value, 1297 in case <samp><span class="option">--with-sysroot</span></samp> is not given an argument, is 1298 <samp><span class="option">${gcc_tooldir}/sys-root</span></samp>. If the specified directory is a 1299 subdirectory of <samp><span class="option">${exec_prefix}</span></samp>, then it will be found relative to 1300 the GCC binaries if the installation tree is moved. 1301 1302 <p>This option affects the system root for the compiler used to build 1303 target libraries (which runs on the build system) and the compiler newly 1304 installed with <code>make install</code>; it does not affect the compiler which is 1305 used to build GCC itself. 1306 1307 <p>If you specify the <samp><span class="option">--with-native-system-header-dir=</span><var>dirname</var></samp> 1308 option then the compiler will search that directory within <var>dirname</var> for 1309 native system headers rather than the default <samp><span class="file">/usr/include</span></samp>. 1310 1311 <br><dt><code>--with-build-sysroot</code><dt><code>--with-build-sysroot=</code><var>dir</var><dd>Tells GCC to consider <var>dir</var> as the system root (see 1312 <samp><span class="option">--with-sysroot</span></samp>) while building target libraries, instead of 1313 the directory specified with <samp><span class="option">--with-sysroot</span></samp>. This option is 1314 only useful when you are already using <samp><span class="option">--with-sysroot</span></samp>. You 1315 can use <samp><span class="option">--with-build-sysroot</span></samp> when you are configuring with 1316 <samp><span class="option">--prefix</span></samp> set to a directory that is different from the one in 1317 which you are installing GCC and your target libraries. 1318 1319 <p>This option affects the system root for the compiler used to build 1320 target libraries (which runs on the build system); it does not affect 1321 the compiler which is used to build GCC itself. 1322 1323 <p>If you specify the <samp><span class="option">--with-native-system-header-dir=</span><var>dirname</var></samp> 1324 option then the compiler will search that directory within <var>dirname</var> for 1325 native system headers rather than the default <samp><span class="file">/usr/include</span></samp>. 1326 1327 <br><dt><code>--with-headers</code><dt><code>--with-headers=</code><var>dir</var><dd>Deprecated in favor of <samp><span class="option">--with-sysroot</span></samp>. 1328 Specifies that target headers are available when building a cross compiler. 1329 The <var>dir</var> argument specifies a directory which has the target include 1330 files. These include files will be copied into the <samp><span class="file">gcc</span></samp> install 1331 directory. <em>This option with the </em><var>dir</var><em> argument is required</em> when 1332 building a cross compiler, if <samp><var>prefix</var><span class="file">/</span><var>target</var><span class="file">/sys-include</span></samp> 1333 doesn't pre-exist. If <samp><var>prefix</var><span class="file">/</span><var>target</var><span class="file">/sys-include</span></samp> does 1334 pre-exist, the <var>dir</var> argument may be omitted. <samp><span class="command">fixincludes</span></samp> 1335 will be run on these files to make them compatible with GCC. 1336 1337 <br><dt><code>--without-headers</code><dd>Tells GCC not use any target headers from a libc when building a cross 1338 compiler. When crossing to GNU/Linux, you need the headers so GCC 1339 can build the exception handling for libgcc. 1340 1341 <br><dt><code>--with-libs</code><dt><code>--with-libs="</code><var>dir1</var> <var>dir2</var><code> ... </code><var>dirN</var><code>"</code><dd>Deprecated in favor of <samp><span class="option">--with-sysroot</span></samp>. 1342 Specifies a list of directories which contain the target runtime 1343 libraries. These libraries will be copied into the <samp><span class="file">gcc</span></samp> install 1344 directory. If the directory list is omitted, this option has no 1345 effect. 1346 1347 <br><dt><code>--with-newlib</code><dd>Specifies that ‘<samp><span class="samp">newlib</span></samp>’ is 1348 being used as the target C library. This causes <code>__eprintf</code> to be 1349 omitted from <samp><span class="file">libgcc.a</span></samp> on the assumption that it will be provided by 1350 ‘<samp><span class="samp">newlib</span></samp>’. 1351 1352 <br><dt><code>--with-avrlibc</code><dd>Specifies that ‘<samp><span class="samp">AVR-Libc</span></samp>’ is 1353 being used as the target C library. This causes float support 1354 functions like <code>__addsf3</code> to be omitted from <samp><span class="file">libgcc.a</span></samp> on 1355 the assumption that it will be provided by <samp><span class="file">libm.a</span></samp>. For more 1356 technical details, cf. <a href="http://gcc.gnu.org/PR54461">PR54461</a>. 1357 This option is only supported for the AVR target. It is not supported for 1358 RTEMS configurations, which currently use newlib. The option is 1359 supported since version 4.7.2 and is the default in 4.8.0 and newer. 1360 1361 <br><dt><code>--with-nds32-lib=</code><var>library</var><dd>Specifies that <var>library</var> setting is used for building <samp><span class="file">libgcc.a</span></samp>. 1362 Currently, the valid <var>library</var> is ‘<samp><span class="samp">newlib</span></samp>’ or ‘<samp><span class="samp">mculib</span></samp>’. 1363 This option is only supported for the NDS32 target. 1364 1365 <br><dt><code>--with-build-time-tools=</code><var>dir</var><dd>Specifies where to find the set of target tools (assembler, linker, etc.) 1366 that will be used while building GCC itself. This option can be useful 1367 if the directory layouts are different between the system you are building 1368 GCC on, and the system where you will deploy it. 1369 1370 <p>For example, on an ‘<samp><span class="samp">ia64-hp-hpux</span></samp>’ system, you may have the GNU 1371 assembler and linker in <samp><span class="file">/usr/bin</span></samp>, and the native tools in a 1372 different path, and build a toolchain that expects to find the 1373 native tools in <samp><span class="file">/usr/bin</span></samp>. 1374 1375 <p>When you use this option, you should ensure that <var>dir</var> includes 1376 <samp><span class="command">ar</span></samp>, <samp><span class="command">as</span></samp>, <samp><span class="command">ld</span></samp>, <samp><span class="command">nm</span></samp>, 1377 <samp><span class="command">ranlib</span></samp> and <samp><span class="command">strip</span></samp> if necessary, and possibly 1378 <samp><span class="command">objdump</span></samp>. Otherwise, GCC may use an inconsistent set of 1379 tools. 1380 </dl> 1381 1382 <h5 class="subsubheading"><a name="TOC4"></a>Overriding <samp><span class="command">configure</span></samp> test results</h5> 1383 1384 <p>Sometimes, it might be necessary to override the result of some 1385 <samp><span class="command">configure</span></samp> test, for example in order to ease porting to a new 1386 system or work around a bug in a test. The toplevel <samp><span class="command">configure</span></samp> 1387 script provides three variables for this: 1388 1389 <dl> 1390 <dt><code>build_configargs</code><dd><a name="index-g_t_0040code_007bbuild_005fconfigargs_007d-3"></a>The contents of this variable is passed to all build <samp><span class="command">configure</span></samp> 1391 scripts. 1392 1393 <br><dt><code>host_configargs</code><dd><a name="index-g_t_0040code_007bhost_005fconfigargs_007d-4"></a>The contents of this variable is passed to all host <samp><span class="command">configure</span></samp> 1394 scripts. 1395 1396 <br><dt><code>target_configargs</code><dd><a name="index-g_t_0040code_007btarget_005fconfigargs_007d-5"></a>The contents of this variable is passed to all target <samp><span class="command">configure</span></samp> 1397 scripts. 1398 1399 </dl> 1400 1401 <p>In order to avoid shell and <samp><span class="command">make</span></samp> quoting issues for complex 1402 overrides, you can pass a setting for <samp><span class="env">CONFIG_SITE</span></samp> and set 1403 variables in the site file. 1404 1405 <h4 class="subheading"><a name="TOC5"></a>Objective-C-Specific Options</h4> 1406 1407 <p>The following options apply to the build of the Objective-C runtime library. 1408 1409 <dl> 1410 <dt><code>--enable-objc-gc</code><dd>Specify that an additional variant of the GNU Objective-C runtime library 1411 is built, using an external build of the Boehm-Demers-Weiser garbage 1412 collector (<a href="http://www.hboehm.info/gc/">http://www.hboehm.info/gc/</a>). This library needs to be 1413 available for each multilib variant, unless configured with 1414 <samp><span class="option">--enable-objc-gc=‘</span><samp><span class="samp">auto</span></samp><span class="option">’</span></samp> in which case the build of the 1415 additional runtime library is skipped when not available and the build 1416 continues. 1417 1418 <br><dt><code>--with-target-bdw-gc=</code><var>list</var><dt><code>--with-target-bdw-gc-include=</code><var>list</var><dt><code>--with-target-bdw-gc-lib=</code><var>list</var><dd>Specify search directories for the garbage collector header files and 1419 libraries. <var>list</var> is a comma separated list of key value pairs of the 1420 form ‘<samp><var>multilibdir</var><span class="samp">=</span><var>path</var></samp>’, where the default multilib key 1421 is named as ‘<samp><span class="samp">.</span></samp>’ (dot), or is omitted (e.g. 1422 ‘<samp><span class="samp">--with-target-bdw-gc=/opt/bdw-gc,32=/opt-bdw-gc32</span></samp>’). 1423 1424 <p>The options <samp><span class="option">--with-target-bdw-gc-include</span></samp> and 1425 <samp><span class="option">--with-target-bdw-gc-lib</span></samp> must always be specified together 1426 for each multilib variant and they take precedence over 1427 <samp><span class="option">--with-target-bdw-gc</span></samp>. If <samp><span class="option">--with-target-bdw-gc-include</span></samp> 1428 is missing values for a multilib, then the value for the default 1429 multilib is used (e.g. ‘<samp><span class="samp">--with-target-bdw-gc-include=/opt/bdw-gc/include</span></samp>’ 1430 ‘<samp><span class="samp">--with-target-bdw-gc-lib=/opt/bdw-gc/lib64,32=/opt-bdw-gc/lib32</span></samp>’). 1431 If none of these options are specified, the library is assumed in 1432 default locations. 1433 </dl> 1434 1435 <p><hr /> 1436 <p><a href="./index.html">Return to the GCC Installation page</a> 1437 1438 <!-- ***Building**************************************************************** --> 1439 <!-- ***Testing***************************************************************** --> 1440 <!-- ***Final install*********************************************************** --> 1441 <!-- ***Binaries**************************************************************** --> 1442 <!-- ***Specific**************************************************************** --> 1443 <!-- ***Old documentation****************************************************** --> 1444 <!-- ***GFDL******************************************************************** --> 1445 <!-- *************************************************************************** --> 1446 <!-- Part 6 The End of the Document --> 1447 </body></html> 1448 1449
Indexes created Thu Jul 16 00:26:27 UTC 2026