![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to gcc.info-6 CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Apple XNU] / GNUtools / cc|
Return to gcc.info-6 CVS log | Up to [Apple XNU] / GNUtools / cc |
1.1 root 1: This is Info file gcc.info, produced by Makeinfo-1.54 from the input
2: file gcc.texi.
3:
4: This file documents the use and the internals of the GNU compiler.
5:
6: Published by the Free Software Foundation 675 Massachusetts Avenue
7: Cambridge, MA 02139 USA
8:
9: Copyright (C) 1988, 1989, 1992, 1993 Free Software Foundation, Inc.
10:
11: Permission is granted to make and distribute verbatim copies of this
12: manual provided the copyright notice and this permission notice are
13: preserved on all copies.
14:
15: Permission is granted to copy and distribute modified versions of
16: this manual under the conditions for verbatim copying, provided also
17: that the sections entitled "GNU General Public License" and "Protect
18: Your Freedom--Fight `Look And Feel'" are included exactly as in the
19: original, and provided that the entire resulting derived work is
20: distributed under the terms of a permission notice identical to this
21: one.
22:
23: Permission is granted to copy and distribute translations of this
24: manual into another language, under the above conditions for modified
25: versions, except that the sections entitled "GNU General Public
26: License" and "Protect Your Freedom--Fight `Look And Feel'", and this
27: permission notice, may be included in translations approved by the Free
28: Software Foundation instead of in the original English.
29:
30:
31: File: gcc.info, Node: PA Install, Next: Sun Install, Prev: Cross-Compiler, Up: Installation
32:
33: Installing on the HP Precision Architecture
34: ===========================================
35:
36: There are two variants of this CPU, called 1.0 and 1.1, which have
37: different machine descriptions. You must use the right one for your
38: machine. All 7NN machines and 8N7 machines use 1.1, while all other
39: 8NN machines use 1.0.
40:
41: The easiest way to handle this problem is to use `configure hpNNN'
42: or `configure hpNNN-hpux', where NNN is the model number of the
43: machine. Then `configure' will figure out if the machine is a 1.0 or
44: 1.1. Use `uname -a' to find out the model number of your machine.
45:
46: `-g' does not work on HP-UX, since that system uses a peculiar
47: debugging format which GNU CC does not know about. There are
48: preliminary versions of GAS and GDB for the HP-PA which do work with
49: GNU CC for debugging. You can get them by anonymous ftp from
50: `jaguar.cs.utah.edu' `dist' subdirectory. You would need to install
51: GAS in the file
52:
53: /usr/local/lib/gcc-lib/CONFIGURATION/GCCVERSION/as
54:
55: where CONFIGURATION is the configuration name (perhaps `hpNNN-hpux')
56: and GCCVERSION is the GNU CC version number. Do this *before* starting
57: the build process, otherwise you will get errors from the HPUX
58: assembler while building `libgcc2.a'. The command
59:
60: make install-dir
61:
62: will create the necessary directory hierarchy so you can install GAS
63: before building GCC.
64:
65: If you obtained GAS before October 6, 1992 it is highly recommended
66: you get a new one to avoid several bugs which have been discovered
67: recently.
68:
69: To enable debugging, configure GNU CC with the `--gas' option before
70: building.
71:
72: It has been reported that GNU CC produces invalid assembly code for
73: 1.1 machines running HP-UX 8.02 when using the HP assembler. Typically
74: the errors look like this:
75: as: bug.s @line#15 [err#1060]
76: Argument 0 or 2 in FARG upper
77: - lookahead = ARGW1=FR,RTNVAL=GR
78: as: foo.s @line#28 [err#1060]
79: Argument 0 or 2 in FARG upper
80: - lookahead = ARGW1=FR
81:
82: You can check the version of HP-UX you are running by executing the
83: command `uname -r'. If you are indeed running HP-UX 8.02 on a 1.1
84: machine and using the HP assembler then configure GCC with
85: "hp700-hpux8.02".
86:
87:
88: File: gcc.info, Node: Sun Install, Next: 3b1 Install, Prev: PA Install, Up: Installation
89:
90: Installing GNU CC on the Sun
91: ============================
92:
93: On Solaris (version 2.1), do not use the linker or other tools in
94: `/usr/ucb' to build GNU CC. Use `/usr/ccs/bin'.
95:
96: Make sure the environment variable `FLOAT_OPTION' is not set when
97: you compile `libgcc.a'. If this option were set to `f68881' when
98: `libgcc.a' is compiled, the resulting code would demand to be linked
99: with a special startup file and would not link properly without special
100: pains.
101:
102: The GNU compiler does not really support the Super SPARC processor
103: that is used in SPARC Station 10 and similar class machines. You can
104: get code that runs by specifying `sparc' as the cpu type; however, its
105: performance is not very good, and may vary widely according to the
106: compiler version and optimization options used. This is because the
107: instruction scheduling parameters designed for the Sparc are not correct
108: for the Super SPARC. Implementing scheduling parameters for the Super
109: SPARC might be a good project for someone who is willing to learn a
110: great deal about instruction scheduling in GNU CC.
111:
112: There is a bug in `alloca' in certain versions of the Sun library.
113: To avoid this bug, install the binaries of GNU CC that were compiled by
114: GNU CC. They use `alloca' as a built-in function and never the one in
115: the library.
116:
117: Some versions of the Sun compiler crash when compiling GNU CC. The
118: problem is a segmentation fault in cpp. This problem seems to be due to
119: the bulk of data in the environment variables. You may be able to avoid
120: it by using the following command to compile GNU CC with Sun CC:
121:
122: make CC="TERMCAP=x OBJS=x LIBFUNCS=x STAGESTUFF=x cc"
123:
124:
125: File: gcc.info, Node: 3b1 Install, Next: Unos Install, Prev: Sun Install, Up: Installation
126:
127: Installing GNU CC on the 3b1
128: ============================
129:
130: Installing GNU CC on the 3b1 is difficult if you do not already have
131: GNU CC running, due to bugs in the installed C compiler. However, the
132: following procedure might work. We are unable to test it.
133:
134: 1. Comment out the `#include "config.h"' line on line 37 of `cccp.c'
135: and do `make cpp'. This makes a preliminary version of GNU cpp.
136:
137: 2. Save the old `/lib/cpp' and copy the preliminary GNU cpp to that
138: file name.
139:
140: 3. Undo your change in `cccp.c', or reinstall the original version,
141: and do `make cpp' again.
142:
143: 4. Copy this final version of GNU cpp into `/lib/cpp'.
144:
145: 5. Replace every occurrence of `obstack_free' in the file `tree.c'
146: with `_obstack_free'.
147:
148: 6. Run `make' to get the first-stage GNU CC.
149:
150: 7. Reinstall the original version of `/lib/cpp'.
151:
152: 8. Now you can compile GNU CC with itself and install it in the normal
153: fashion.
154:
155:
156: File: gcc.info, Node: Unos Install, Next: VMS Install, Prev: 3b1 Install, Up: Installation
157:
158: Installing GNU CC on Unos
159: =========================
160:
161: Use `configure unos' for building on Unos.
162:
163: The Unos assembler is named `casm' instead of `as'. For some
164: strange reason linking `/bin/as' to `/bin/casm' changes the behavior,
165: and does not work. So, when installing GNU CC, you should install the
166: following script as `as' in the subdirectory where the passes of GCC
167: are installed:
168:
169: #!/bin/sh
170: casm $*
171:
172: The default Unos library is named `libunos.a' instead of `libc.a'.
173: To allow GNU CC to function, either change all references to `-lc' in
174: `gcc.c' to `-lunos' or link `/lib/libc.a' to `/lib/libunos.a'.
175:
176: When compiling GNU CC with the standard compiler, to overcome bugs in
177: the support of `alloca', do not use `-O' when making stage 2. Then use
178: the stage 2 compiler with `-O' to make the stage 3 compiler. This
179: compiler will have the same characteristics as the usual stage 2
180: compiler on other systems. Use it to make a stage 4 compiler and
181: compare that with stage 3 to verify proper compilation.
182:
183: (Perhaps simply defining `ALLOCA' in `x-crds' as described in the
184: comments there will make the above paragraph superfluous. Please
185: inform us of whether this works.)
186:
187: Unos uses memory segmentation instead of demand paging, so you will
188: need a lot of memory. 5 Mb is barely enough if no other tasks are
189: running. If linking `cc1' fails, try putting the object files into a
190: library and linking from that library.
191:
192:
193: File: gcc.info, Node: VMS Install, Next: WE32K Install, Prev: Unos Install, Up: Installation
194:
195: Installing GNU CC on VMS
196: ========================
197:
198: The VMS version of GNU CC is distributed in a backup saveset
199: containing both source code and precompiled binaries.
200:
201: To install the `gcc' command so you can use the compiler easily, in
202: the same manner as you use the VMS C compiler, you must install the VMS
203: CLD file for GNU CC as follows:
204:
205: 1. Define the VMS logical names `GNU_CC' and `GNU_CC_INCLUDE' to
206: point to the directories where the GNU CC executables
207: (`gcc-cpp.exe', `gcc-cc1.exe', etc.) and the C include files are
208: kept respectively. This should be done with the commands:
209:
210: $ assign /system /translation=concealed -
211: disk:[gcc.] gnu_cc
212: $ assign /system /translation=concealed -
213: disk:[gcc.include.] gnu_cc_include
214:
215: with the appropriate disk and directory names. These commands can
216: be placed in your system startup file so they will be executed
217: whenever the machine is rebooted. You may, if you choose, do this
218: via the `GCC_INSTALL.COM' script in the `[GCC]' directory.
219:
220: 2. Install the `GCC' command with the command line:
221:
222: $ set command /table=sys$common:[syslib]dcltables -
223: /output=sys$common:[syslib]dcltables gnu_cc:[000000]gcc
224: $ install replace sys$common:[syslib]dcltables
225:
226: 3. To install the help file, do the following:
227:
228: $ library/help sys$library:helplib.hlb gcc.hlp
229:
230: Now you can invoke the compiler with a command like `gcc /verbose
231: file.c', which is equivalent to the command `gcc -v -c file.c' in
232: Unix.
233:
234: If you wish to use GNU C++ you must first install GNU CC, and then
235: perform the following steps:
236:
237: 1. Define the VMS logical name `GNU_GXX_INCLUDE' to point to the
238: directory where the preprocessor will search for the C++ header
239: files. This can be done with the command:
240:
241: $ assign /system /translation=concealed -
242: disk:[gcc.gxx_include.] gnu_gxx_include
243:
244: with the appropriate disk and directory name. If you are going to
245: be using libg++, this is where the libg++ install procedure will
246: install the libg++ header files.
247:
248: 2. Obtain the file `gcc-cc1plus.exe', and place this in the same
249: directory that `gcc-cc1.exe' is kept.
250:
251: The GNU C++ compiler can be invoked with a command like `gcc /plus
252: /verbose file.cc', which is equivalent to the command `g++ -v -c
253: file.cc' in Unix.
254:
255: We try to put corresponding binaries and sources on the VMS
256: distribution tape. But sometimes the binaries will be from an older
257: version than the sources, because we don't always have time to update
258: them. (Use the `/version' option to determine the version number of
259: the binaries and compare it with the source file `version.c' to tell
260: whether this is so.) In this case, you should use the binaries you get
261: to recompile the sources. If you must recompile, here is how:
262:
263: 1. Execute the command procedure `vmsconfig.com' to set up the files
264: `tm.h', `config.h', `aux-output.c', and `md.', and to create files
265: `tconfig.h' and `hconfig.h'. This procedure also creates several
266: linker option files used by `make-cc1.com' and a data file used by
267: `make-l2.com'.
268:
269: $ @vmsconfig.com
270:
271: 2. Setup the logical names and command tables as defined above. In
272: addition, define the VMS logical name `GNU_BISON' to point at the
273: to the directories where the Bison executable is kept. This
274: should be done with the command:
275:
276: $ assign /system /translation=concealed -
277: disk:[bison.] gnu_bison
278:
279: You may, if you choose, use the `INSTALL_BISON.COM' script in the
280: `[BISON]' directory.
281:
282: 3. Install the `BISON' command with the command line:
283:
284: $ set command /table=sys$common:[syslib]dcltables -
285: /output=sys$common:[syslib]dcltables -
286: gnu_bison:[000000]bison
287: $ install replace sys$common:[syslib]dcltables
288:
289: 4. Type `@make-gcc' to recompile everything (alternatively, submit
290: the file `make-gcc.com' to a batch queue). If you wish to build
291: the GNU C++ compiler as well as the GNU CC compiler, you must
292: first edit `make-gcc.com' and follow the instructions that appear
293: in the comments.
294:
295: 5. In order to use GCC, you need a library of functions which GCC
296: compiled code will call to perform certain tasks, and these
297: functions are defined in the file `libgcc2.c'. To compile this
298: you should use the command procedure `make-l2.com', which will
299: generate the library `libgcc2.olb'. `libgcc2.olb' should be built
300: using the compiler built from the same distribution that
301: `libgcc2.c' came from, and `make-gcc.com' will automatically do
302: all of this for you.
303:
304: To install the library, use the following commands:
305:
306: $ library gnu_cc:[000000]gcclib/delete=(new,eprintf)
307: $ library gnu_cc:[000000]gcclib/delete=L_*
308: $ library libgcc2/extract=*/output=libgcc2.obj
309: $ library gnu_cc:[000000]gcclib libgcc2.obj
310:
311: The first command simply removes old modules that will be replaced
312: with modules from `libgcc2' under different module names. The
313: modules `new' and `eprintf' may not actually be present in your
314: `gcclib.olb'--if the VMS librarian complains about those modules
315: not being present, simply ignore the message and continue on with
316: the next command. The second command removes the modules that
317: came from the previous version of the library `libgcc2.c'.
318:
319: Whenever you update the compiler on your system, you should also
320: update the library with the above procedure.
321:
322: 6. You may wish to build GCC in such a way that no files are written
323: to the directory where the source files reside. An example would
324: be the when the source files are on a read-only disk. In these
325: cases, execute the following DCL commands (substituting your
326: actual path names):
327:
328: $ assign dua0:[gcc.build_dir.]/translation=concealed, -
329: dua1:[gcc.source_dir.]/translation=concealed gcc_build
330: $ set default gcc_build:[000000]
331:
332: where the directory `dua1:[gcc.source_dir]' contains the source
333: code, and the directory `dua0:[gcc.build_dir]' is meant to contain
334: all of the generated object files and executables. Once you have
335: done this, you can proceed building GCC as described above. (Keep
336: in mind that `gcc_build' is a rooted logical name, and thus the
337: device names in each element of the search list must be an actual
338: physical device name rather than another rooted logical name).
339:
340: 7. *If you are building GNU CC with a previous version of GNU CC, you
341: also should check to see that you have the newest version of the
342: assembler*. In particular, GNU CC version 2 treats global constant
343: variables slightly differently from GNU CC version 1, and GAS
344: version 1.38.1 does not have the patches required to work with GCC
345: version 2. If you use GAS 1.38.1, then `extern const' variables
346: will not have the read-only bit set, and the linker will generate
347: warning messages about mismatched psect attributes for these
348: variables. These warning messages are merely a nuisance, and can
349: safely be ignored.
350:
351: If you are compiling with a version of GNU CC older than 1.33,
352: specify `/DEFINE=("inline=")' as an option in all the
353: compilations. This requires editing all the `gcc' commands in
354: `make-cc1.com'. (The older versions had problems supporting
355: `inline'.) Once you have a working 1.33 or newer GNU CC, you can
356: change this file back.
357:
358: 8. If you want to build GNU CC with the VAX C compiler, you will need
359: to make minor changes in `make-cccp.com' and `make-cc1.com' to
360: choose alternate definitions of `CC', `CFLAGS', and `LIBS'. See
361: comments in those files. However, you must also have a working
362: version of the GNU assembler (GNU as, aka GAS) as it is used as
363: the back-end for GNU CC to produce binary object modules and is
364: not included in the GNU CC sources. GAS is also needed to compile
365: `libgcc2' in order to build `gcclib' (see above); `make-l2.com'
366: expects to be able to find it operational in
367: `gnu_cc:[000000]gnu-as.exe'.
368:
369: To use GNU CC on VMS, you need the VMS driver programs `gcc.exe',
370: `gcc.com', and `gcc.cld'. They are distributed with the VMS
371: binaries (`gcc-vms') rather than the GNU CC sources. GAS is also
372: included in `gcc-vms', as is Bison.
373:
374: Once you have successfully built GNU CC with VAX C, you should use
375: the resulting compiler to rebuild itself. Before doing this, be
376: sure to restore the `CC', `CFLAGS', and `LIBS' definitions in
377: `make-cccp.com' and `make-cc1.com'. The second generation
378: compiler will be able to take advantage of many optimizations that
379: must be suppressed when building with other compilers.
380:
381: Under previous versions of GNU CC, the generated code would
382: occasionally give strange results when linked with the sharable
383: `VAXCRTL' library. Now this should work.
384:
385: Even with this version, however, GNU CC itself should not be linked
386: with the sharable `VAXCRTL'. The version of `qsort' in `VAXCRTL' has a
387: bug (known to be present in VMS versions V4.6 through V5.5) which
388: causes the compiler to fail.
389:
390: The executables are generated by `make-cc1.com' and `make-cccp.com'
391: use the object library version of `VAXCRTL' in order to make use of the
392: `qsort' routine in `gcclib.olb'. If you wish to link the compiler
393: executables with the shareable image version of `VAXCRTL', you should
394: edit the file `tm.h' (created by `vmsconfig.com') to define the macro
395: `QSORT_WORKAROUND'.
396:
397: `QSORT_WORKAROUND' is always defined when GNU CC is compiled with
398: VAX C, to avoid a problem in case `gcclib.olb' is not yet available.
399:
400:
401: File: gcc.info, Node: WE32K Install, Next: MIPS Install, Prev: VMS Install, Up: Installation
402:
403: Installing GNU CC on the WE32K
404: ==============================
405:
406: These computers are also known as the 3b2, 3b5, 3b20 and other
407: similar names. (However, the 3b1 is actually a 68000; see *Note 3b1
408: Install::.)
409:
410: Don't use `-g' when compiling with the system's compiler. The
411: system's linker seems to be unable to handle such a large program with
412: debugging information.
413:
414: The system's compiler runs out of capacity when compiling `stmt.c'
415: in GNU CC. You can work around this by building `cpp' in GNU CC first,
416: then use that instead of the system's preprocessor with the system's C
417: compiler to compile `stmt.c'. Here is how:
418:
419: mv /lib/cpp /lib/cpp.att
420: cp cpp /lib/cpp.gnu
421: echo '/lib/cpp.gnu -traditional ${1+"$@"}' > /lib/cpp
422: chmod +x /lib/cpp
423:
424: The system's compiler produces bad code for some of the GNU CC
425: optimization files. So you must build the stage 2 compiler without
426: optimization. Then build a stage 3 compiler with optimization. That
427: executable should work. Here are the necessary commands:
428:
429: make LANGUAGES=c CC=stage1/xgcc CFLAGS="-Bstage1/ -g"
430: make stage2
431: make CC=stage2/xgcc CFLAGS="-Bstage2/ -g -O"
432:
433: You may need to raise the ULIMIT setting to build a C++ compiler, as
434: the file `cc1plus' is larger than one megabyte.
435:
436:
437: File: gcc.info, Node: MIPS Install, Next: Collect2, Prev: WE32K Install, Up: Installation
438:
439: Installing GNU CC on the MIPS
440: =============================
441:
442: See *Note Installation:: about whether to use either of the options
443: `--with-stabs' or `--with-gnu-as'.
444:
445: The MIPS C compiler needs to be told to increase its table size for
446: switch statements with the `-Wf,-XNg1500' option in order to compile
447: `cp-parse.c'. If you use the `-O2' optimization option, you also need
448: to use `-Olimit 3000'. Both of these options are automatically
449: generated in the `Makefile' that the shell script `configure' builds.
450: If you override the `CC' make variable and use the MIPS compilers, you
451: may need to add `-Wf,-XNg1500 -Olimit 3000'.
452:
453: MIPS computers running RISC-OS can support four different
454: personalities: default, BSD 4.3, System V.3, and System V.4 (older
455: versions of RISC-OS don't support V.4). To configure GCC for these
456: platforms use the following configurations:
457:
458: `mips-mips-riscos`rev''
459: Default configuration for RISC-OS, revision `rev'.
460:
461: `mips-mips-riscos`rev'bsd'
462: BSD 4.3 configuration for RISC-OS, revision `rev'.
463:
464: `mips-mips-riscos`rev'sysv4'
465: System V.4 configuration for RISC-OS, revision `rev'.
466:
467: `mips-mips-riscos`rev'sysv'
468: System V.3 configuration for RISC-OS, revision `rev'.
469:
470: The revision `rev' mentioned above is the revision of RISC-OS to
471: use. You must reconfigure GCC when going from a RISC-OS revision 4 to
472: RISC-OS revision 5. This has the effect of avoiding a linker bug (see
473: *Note Installation Problems:: for more details).
474:
475: DECstations can support three different personalities: Ultrix, DEC
476: OSF/1, and OSF/rose. To configure GCC for these platforms use the
477: following configurations:
478:
479: `decstation-ultrix'
480: Ultrix configuration.
481:
482: `decstation-osf1'
483: Dec's version of OSF/1.
484:
485: `decstation-osfrose'
486: Open Software Foundation reference port of OSF/1 which uses the
487: OSF/rose object file format instead of ECOFF. Normally, you would
488: not select this configuration.
489:
490: On Irix version 4.0.5F, and perhaps on some other versions as well,
491: there is an assembler bug that reorders instructions incorrectly. To
492: work around it, specify the target configuration `mips-sgi-irix4loser'.
493: This configuration inhibits assembler optimization.
494:
495: You can turn off assembler optimization in a compiler configured with
496: target `mips-sgi-irix4' using the `-noasmopt' option. This compiler
497: option passes the option `-O0' to the assembler, to inhibit reordering.
498:
499: The `-noasmopt' option can be useful for testing whether a problem
500: is due to erroneous assembler reordering. Even if a problem does not go
501: away with `-noasmopt', it may still be due to assembler
502: reordering--perhaps GNU CC itself was miscompiled as a result.
503:
504: We know this is inconvenient, but it's the best that can be done at
505: the last minute.
506:
507:
508: File: gcc.info, Node: Collect2, Next: Header Dirs, Prev: MIPS Install, Up: Installation
509:
510: `collect2'
511: ==========
512:
513: Many target systems do not have support in the assembler and linker
514: for "constructors"--initialization functions to be called before the
515: official "start" of `main'. On such systems, GNU CC uses a utility
516: called `collect2' to arrange to call these functions at start time.
517:
518: The program `collect2' works by linking the program once and looking
519: through the linker output file for symbols with particular names
520: indicating they are constructor functions. If it finds any, it creates
521: a new temporary `.c' file containing a table of them, compiles it, and
522: links the program a second time including that file.
523:
524: The actual calls to the constructors are carried out by a subroutine
525: called `__main', which is called (automatically) at the beginning of
526: the body of `main' (provided `main' was compiled with GNU CC).
527:
528: The program `collect2' is installed as `ld' in the directory where
529: the passes of the compiler are installed. When `collect2' needs to
530: find the *real* `ld', it tries the following file names:
531:
532: * `gld' in the directories listed in the compiler's search
533: directories.
534:
535: * `gld' in the directories listed in the environment variable `PATH'.
536:
537: * `real-ld' in the compiler's search directories.
538:
539: * `real-ld' in `PATH'.
540:
541: * `ld' in `PATH'.
542:
543: "The compiler's search directories" means all the directories where
544: `gcc' searches for passes of the compiler. This includes directories
545: that you specify with `-B'.
546:
547: Cross-compilers search a little differently:
548:
549: * `gld' in the compiler's search directories.
550:
551: * `TARGET-gld' in `PATH'.
552:
553: * `real-ld' in the compiler's search directories.
554:
555: * `TARGET-real-ld' in `PATH'.
556:
557: * `TARGET-ld' in `PATH'.
558:
559: `collect2' does not search for `ld' using the compiler's search
560: directories, because if it did, it would find itself--not the real
561: `ld'--and this could lead to infinite recursion. However, the
562: directory where `collect2' is installed might happen to be in `PATH'.
563: That could lead `collect2' to invoke itself anyway. when looking for
564: `ld'.
565:
566: To prevent this, `collect2' explicitly avoids running `ld' using the
567: file name under which `collect2' itself was invoked. In fact, it
568: remembers up to two such names--in case one copy of `collect2' finds
569: another copy (or version) of `collect2' installed as `ld' in a second
570: place in the search path.
571:
572: If two file names to avoid are not sufficient, you may still
573: encounter an infinite recursion of `collect2' processes. When this
574: happens. check all the files installed as `ld' in any of the
575: directories searched, and straighten out the situation.
576:
577: (In a future version, we will probably change `collect2' to avoid
578: any reinvocation of a file from which any parent `collect2' was run.)
579:
580:
581: File: gcc.info, Node: Header Dirs, Prev: Collect2, Up: Installation
582:
583: Standard Header File Directories
584: ================================
585:
586: `GCC_INCLUDE_DIR' means the same thing for native and cross. It is
587: where GNU CC stores its private include files, and also where GNU CC
588: stores the fixed include files. A cross compiled GNU CC runs
589: `fixincludes' on the header files in `$(tooldir)/include'. (If the
590: cross compilation header files need to be fixed, they must be installed
591: before GNU CC is built. If the cross compilation header files are
592: already suitable for ANSI C and GNU CC, nothing special need be done).
593:
594: `GPLUS_INCLUDE_DIR' means the same thing for native and cross. It
595: is where `g++' looks first for header files. `libg++' installs only
596: target independent header files in that directory.
597:
598: `LOCAL_INCLUDE_DIR' is used only for a native compiler. It is
599: normally `/usr/local/include'. GNU CC searches this directory so that
600: users can install header files in `/usr/local/include'.
601:
602: `CROSS_INCLUDE_DIR' is used only for a cross compiler. GNU CC
603: doesn't install anything there.
604:
605: `TOOL_INCLUDE_DIR' is used for both native and cross compilers. It
606: is the place for other packages to install header files that GNU CC will
607: use. For a cross-compiler, this is the equivalent of `/usr/include'.
608: When you build a cross-compiler, `fixincludes' processes any header
609: files in this directory.
610:
611:
612: File: gcc.info, Node: C Extensions, Next: C++ Extensions, Prev: Installation, Up: Top
613:
614: Extensions to the C Language Family
615: ***********************************
616:
617: GNU C provides several language features not found in ANSI standard
618: C. (The `-pedantic' option directs GNU CC to print a warning message if
619: any of these features is used.) To test for the availability of these
620: features in conditional compilation, check for a predefined macro
621: `__GNUC__', which is always defined under GNU CC.
622:
623: These extensions are available in C and in the languages derived from
624: it, C++ and Objective C. *Note Extensions to the C++ Language: C++
625: Extensions, for extensions that apply *only* to C++.
626:
627: * Menu:
628:
629: * Statement Exprs:: Putting statements and declarations inside expressions.
630: * Local Labels:: Labels local to a statement-expression.
631: * Labels as Values:: Getting pointers to labels, and computed gotos.
632: * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
633: * Constructing Calls:: Dispatching a call to another function.
634: * Naming Types:: Giving a name to the type of some expression.
635: * Typeof:: `typeof': referring to the type of an expression.
636: * Lvalues:: Using `?:', `,' and casts in lvalues.
637: * Conditionals:: Omitting the middle operand of a `?:' expression.
638: * Long Long:: Double-word integers--`long long int'.
639: * Complex:: Data types for complex numbers.
640: * Zero Length:: Zero-length arrays.
641: * Variable Length:: Arrays whose length is computed at run time.
642: * Macro Varargs:: Macros with variable number of arguments.
643: * Subscripting:: Any array can be subscripted, even if not an lvalue.
644: * Pointer Arith:: Arithmetic on `void'-pointers and function pointers.
645: * Initializers:: Non-constant initializers.
646: * Constructors:: Constructor expressions give structures, unions
647: or arrays as values.
648: * Labeled Elements:: Labeling elements of initializers.
649: * Cast to Union:: Casting to union type from any member of the union.
650: * Case Ranges:: `case 1 ... 9' and such.
651: * Function Attributes:: Declaring that functions have no side effects,
652: or that they can never return.
653: * Function Prototypes:: Prototype declarations and old-style definitions.
654: * Dollar Signs:: Dollar sign is allowed in identifiers.
655: * Character Escapes:: `\e' stands for the character ESC.
656: * Variable Attributes:: Specifying attributes of variables.
657: * Alignment:: Inquiring about the alignment of a type or variable.
658: * Inline:: Defining inline functions (as fast as macros).
659: * Extended Asm:: Assembler instructions with C expressions as operands.
660: (With them you can define "built-in" functions.)
661: * Asm Labels:: Specifying the assembler name to use for a C symbol.
662: * Explicit Reg Vars:: Defining variables residing in specified registers.
663: * Alternate Keywords:: `__const__', `__asm__', etc., for header files.
664: * Incomplete Enums:: `enum foo;', with details to follow.
665: * Function Names:: Printable strings which are the name of the current
666: function.
667:
668:
669: File: gcc.info, Node: Statement Exprs, Next: Local Labels, Up: C Extensions
670:
671: Statements and Declarations in Expressions
672: ==========================================
673:
674: A compound statement enclosed in parentheses may appear as an
675: expression in GNU C. This allows you to use loops, switches, and local
676: variables within an expression.
677:
678: Recall that a compound statement is a sequence of statements
679: surrounded by braces; in this construct, parentheses go around the
680: braces. For example:
681:
682: ({ int y = foo (); int z;
683: if (y > 0) z = y;
684: else z = - y;
685: z; })
686:
687: is a valid (though slightly more complex than necessary) expression for
688: the absolute value of `foo ()'.
689:
690: The last thing in the compound statement should be an expression
691: followed by a semicolon; the value of this subexpression serves as the
692: value of the entire construct. (If you use some other kind of statement
693: last within the braces, the construct has type `void', and thus
694: effectively no value.)
695:
696: This feature is especially useful in making macro definitions "safe"
697: (so that they evaluate each operand exactly once). For example, the
698: "maximum" function is commonly defined as a macro in standard C as
699: follows:
700:
701: #define max(a,b) ((a) > (b) ? (a) : (b))
702:
703: But this definition computes either A or B twice, with bad results if
704: the operand has side effects. In GNU C, if you know the type of the
705: operands (here let's assume `int'), you can define the macro safely as
706: follows:
707:
708: #define maxint(a,b) \
709: ({int _a = (a), _b = (b); _a > _b ? _a : _b; })
710:
711: Embedded statements are not allowed in constant expressions, such as
712: the value of an enumeration constant, the width of a bit field, or the
713: initial value of a static variable.
714:
715: If you don't know the type of the operand, you can still do this,
716: but you must use `typeof' (*note Typeof::.) or type naming (*note
717: Naming Types::.).
718:
719:
720: File: gcc.info, Node: Local Labels, Next: Labels as Values, Prev: Statement Exprs, Up: C Extensions
721:
722: Locally Declared Labels
723: =======================
724:
725: Each statement expression is a scope in which "local labels" can be
726: declared. A local label is simply an identifier; you can jump to it
727: with an ordinary `goto' statement, but only from within the statement
728: expression it belongs to.
729:
730: A local label declaration looks like this:
731:
732: __label__ LABEL;
733:
734: or
735:
736: __label__ LABEL1, LABEL2, ...;
737:
738: Local label declarations must come at the beginning of the statement
739: expression, right after the `({', before any ordinary declarations.
740:
741: The label declaration defines the label *name*, but does not define
742: the label itself. You must do this in the usual way, with `LABEL:',
743: within the statements of the statement expression.
744:
745: The local label feature is useful because statement expressions are
746: often used in macros. If the macro contains nested loops, a `goto' can
747: be useful for breaking out of them. However, an ordinary label whose
748: scope is the whole function cannot be used: if the macro can be
749: expanded several times in one function, the label will be multiply
750: defined in that function. A local label avoids this problem. For
751: example:
752:
753: #define SEARCH(array, target) \
754: ({ \
755: __label__ found; \
756: typeof (target) _SEARCH_target = (target); \
757: typeof (*(array)) *_SEARCH_array = (array); \
758: int i, j; \
759: int value; \
760: for (i = 0; i < max; i++) \
761: for (j = 0; j < max; j++) \
762: if (_SEARCH_array[i][j] == _SEARCH_target) \
763: { value = i; goto found; } \
764: value = -1; \
765: found: \
766: value; \
767: })
768:
769:
770: File: gcc.info, Node: Labels as Values, Next: Nested Functions, Prev: Local Labels, Up: C Extensions
771:
772: Labels as Values
773: ================
774:
775: You can get the address of a label defined in the current function
776: (or a containing function) with the unary operator `&&'. The value has
777: type `void *'. This value is a constant and can be used wherever a
778: constant of that type is valid. For example:
779:
780: void *ptr;
781: ...
782: ptr = &&foo;
783:
784: To use these values, you need to be able to jump to one. This is
785: done with the computed goto statement(1), `goto *EXP;'. For example,
786:
787: goto *ptr;
788:
789: Any expression of type `void *' is allowed.
790:
791: One way of using these constants is in initializing a static array
792: that will serve as a jump table:
793:
794: static void *array[] = { &&foo, &&bar, &&hack };
795:
796: Then you can select a label with indexing, like this:
797:
798: goto *array[i];
799:
800: Note that this does not check whether the subscript is in bounds--array
801: indexing in C never does that.
802:
803: Such an array of label values serves a purpose much like that of the
804: `switch' statement. The `switch' statement is cleaner, so use that
805: rather than an array unless the problem does not fit a `switch'
806: statement very well.
807:
808: Another use of label values is in an interpreter for threaded code.
809: The labels within the interpreter function can be stored in the
810: threaded code for super-fast dispatching.
811:
812: You can use this mechanism to jump to code in a different function.
813: If you do that, totally unpredictable things will happen. The best way
814: to avoid this is to store the label address only in automatic variables
815: and never pass it as an argument.
816:
817: ---------- Footnotes ----------
818:
819: (1) The analogous feature in Fortran is called an assigned goto,
820: but that name seems inappropriate in C, where one can do more than
821: simply store label addresses in label variables.
822:
823:
824: File: gcc.info, Node: Nested Functions, Next: Constructing Calls, Prev: Labels as Values, Up: C Extensions
825:
826: Nested Functions
827: ================
828:
829: A "nested function" is a function defined inside another function.
830: (Nested functions are not supported for GNU C++.) The nested function's
831: name is local to the block where it is defined. For example, here we
832: define a nested function named `square', and call it twice:
833:
834: foo (double a, double b)
835: {
836: double square (double z) { return z * z; }
837:
838: return square (a) + square (b);
839: }
840:
841: The nested function can access all the variables of the containing
842: function that are visible at the point of its definition. This is
843: called "lexical scoping". For example, here we show a nested function
844: which uses an inherited variable named `offset':
845:
846: bar (int *array, int offset, int size)
847: {
848: int access (int *array, int index)
849: { return array[index + offset]; }
850: int i;
851: ...
852: for (i = 0; i < size; i++)
853: ... access (array, i) ...
854: }
855:
856: Nested function definitions are permitted within functions in the
857: places where variable definitions are allowed; that is, in any block,
858: before the first statement in the block.
859:
860: It is possible to call the nested function from outside the scope of
861: its name by storing its address or passing the address to another
862: function:
863:
864: hack (int *array, int size)
865: {
866: void store (int index, int value)
867: { array[index] = value; }
868:
869: intermediate (store, size);
870: }
871:
872: Here, the function `intermediate' receives the address of `store' as
873: an argument. If `intermediate' calls `store', the arguments given to
874: `store' are used to store into `array'. But this technique works only
875: so long as the containing function (`hack', in this example) does not
876: exit. If you try to call the nested function through its address after
877: the containing function has exited, all hell will break loose.
878:
879: GNU CC implements taking the address of a nested function using a
880: technique called "trampolines". A paper describing them is available
881: from `maya.idiap.ch' in directory `pub/tmb', file `usenix88-lexic.ps.Z'.
882:
883: A nested function can jump to a label inherited from a containing
884: function, provided the label was explicitly declared in the containing
885: function (*note Local Labels::.). Such a jump returns instantly to the
886: containing function, exiting the nested function which did the `goto'
887: and any intermediate functions as well. Here is an example:
888:
889: bar (int *array, int offset, int size)
890: {
891: __label__ failure;
892: int access (int *array, int index)
893: {
894: if (index > size)
895: goto failure;
896: return array[index + offset];
897: }
898: int i;
899: ...
900: for (i = 0; i < size; i++)
901: ... access (array, i) ...
902: ...
903: return 0;
904:
905: /* Control comes here from `access'
906: if it detects an error. */
907: failure:
908: return -1;
909: }
910:
911: A nested function always has internal linkage. Declaring one with
912: `extern' is erroneous. If you need to declare the nested function
913: before its definition, use `auto' (which is otherwise meaningless for
914: function declarations).
915:
916: bar (int *array, int offset, int size)
917: {
918: __label__ failure;
919: auto int access (int *, int);
920: ...
921: int access (int *array, int index)
922: {
923: if (index > size)
924: goto failure;
925: return array[index + offset];
926: }
927: ...
928: }
929:
930:
931: File: gcc.info, Node: Constructing Calls, Next: Naming Types, Prev: Nested Functions, Up: C Extensions
932:
933: Constructing Function Calls
934: ===========================
935:
936: Using the built-in functions described below, you can record the
937: arguments a function received, and call another function with the same
938: arguments, without knowing the number or types of the arguments.
939:
940: You can also record the return value of that function call, and
941: later return that value, without knowing what data type the function
942: tried to return (as long as your caller expects that data type).
943:
944: `__builtin_apply_args ()'
945: This built-in function returns a pointer of type `void *' to data
946: describing how to perform a call with the same arguments as were
947: passed to the current function.
948:
949: The function saves the arg pointer register, structure value
950: address, and all registers that might be used to pass arguments to
951: a function into a block of memory allocated on the stack. Then it
952: returns the address of that block.
953:
954: `__builtin_apply (FUNCTION, ARGUMENTS, SIZE)'
955: This built-in function invokes FUNCTION (type `void (*)()') with a
956: copy of the parameters described by ARGUMENTS (type `void *') and
957: SIZE (type `int').
958:
959: The value of ARGUMENTS should be the value returned by
960: `__builtin_apply_args'. The argument SIZE specifies the size of
961: the stack argument data, in bytes.
962:
963: This function returns a pointer of type `void *' to data describing
964: how to return whatever value was returned by FUNCTION. The data
965: is saved in a block of memory allocated on the stack.
966:
967: It is not always simple to compute the proper value for SIZE. The
968: value is used by `__builtin_apply' to compute the amount of data
969: that should be pushed on the stack and copied from the incoming
970: argument area.
971:
972: `__builtin_return (RESULT)'
973: This built-in function returns the value described by RESULT from
974: the containing function. You should specify, for RESULT, a value
975: returned by `__builtin_apply'.
976:
977:
978: File: gcc.info, Node: Naming Types, Next: Typeof, Prev: Constructing Calls, Up: C Extensions
979:
980: Naming an Expression's Type
981: ===========================
982:
983: You can give a name to the type of an expression using a `typedef'
984: declaration with an initializer. Here is how to define NAME as a type
985: name for the type of EXP:
986:
987: typedef NAME = EXP;
988:
989: This is useful in conjunction with the statements-within-expressions
990: feature. Here is how the two together can be used to define a safe
991: "maximum" macro that operates on any arithmetic type:
992:
993: #define max(a,b) \
994: ({typedef _ta = (a), _tb = (b); \
995: _ta _a = (a); _tb _b = (b); \
996: _a > _b ? _a : _b; })
997:
998: The reason for using names that start with underscores for the local
999: variables is to avoid conflicts with variable names that occur within
1000: the expressions that are substituted for `a' and `b'. Eventually we
1001: hope to design a new form of declaration syntax that allows you to
1002: declare variables whose scopes start only after their initializers;
1003: this will be a more reliable way to prevent such conflicts.
1004:
1005:
1006: File: gcc.info, Node: Typeof, Next: Lvalues, Prev: Naming Types, Up: C Extensions
1007:
1008: Referring to a Type with `typeof'
1009: =================================
1010:
1011: Another way to refer to the type of an expression is with `typeof'.
1012: The syntax of using of this keyword looks like `sizeof', but the
1013: construct acts semantically like a type name defined with `typedef'.
1014:
1015: There are two ways of writing the argument to `typeof': with an
1016: expression or with a type. Here is an example with an expression:
1017:
1018: typeof (x[0](1))
1019:
1020: This assumes that `x' is an array of functions; the type described is
1021: that of the values of the functions.
1022:
1023: Here is an example with a typename as the argument:
1024:
1025: typeof (int *)
1026:
1027: Here the type described is that of pointers to `int'.
1028:
1029: If you are writing a header file that must work when included in
1030: ANSI C programs, write `__typeof__' instead of `typeof'. *Note
1031: Alternate Keywords::.
1032:
1033: A `typeof'-construct can be used anywhere a typedef name could be
1034: used. For example, you can use it in a declaration, in a cast, or
1035: inside of `sizeof' or `typeof'.
1036:
1037: * This declares `y' with the type of what `x' points to.
1038:
1039: typeof (*x) y;
1040:
1041: * This declares `y' as an array of such values.
1042:
1043: typeof (*x) y[4];
1044:
1045: * This declares `y' as an array of pointers to characters:
1046:
1047: typeof (typeof (char *)[4]) y;
1048:
1049: It is equivalent to the following traditional C declaration:
1050:
1051: char *y[4];
1052:
1053: To see the meaning of the declaration using `typeof', and why it
1054: might be a useful way to write, let's rewrite it with these macros:
1055:
1056: #define pointer(T) typeof(T *)
1057: #define array(T, N) typeof(T [N])
1058:
1059: Now the declaration can be rewritten this way:
1060:
1061: array (pointer (char), 4) y;
1062:
1063: Thus, `array (pointer (char), 4)' is the type of arrays of 4
1064: pointers to `char'.
1065:
1066:
1067: File: gcc.info, Node: Lvalues, Next: Conditionals, Prev: Typeof, Up: C Extensions
1068:
1069: Generalized Lvalues
1070: ===================
1071:
1072: Compound expressions, conditional expressions and casts are allowed
1073: as lvalues provided their operands are lvalues. This means that you
1074: can take their addresses or store values into them.
1075:
1076: For example, a compound expression can be assigned, provided the last
1077: expression in the sequence is an lvalue. These two expressions are
1078: equivalent:
1079:
1080: (a, b) += 5
1081: a, (b += 5)
1082:
1083: Similarly, the address of the compound expression can be taken.
1084: These two expressions are equivalent:
1085:
1086: &(a, b)
1087: a, &b
1088:
1089: A conditional expression is a valid lvalue if its type is not void
1090: and the true and false branches are both valid lvalues. For example,
1091: these two expressions are equivalent:
1092:
1093: (a ? b : c) = 5
1094: (a ? b = 5 : (c = 5))
1095:
1096: A cast is a valid lvalue if its operand is an lvalue. A simple
1097: assignment whose left-hand side is a cast works by converting the
1098: right-hand side first to the specified type, then to the type of the
1099: inner left-hand side expression. After this is stored, the value is
1100: converted back to the specified type to become the value of the
1101: assignment. Thus, if `a' has type `char *', the following two
1102: expressions are equivalent:
1103:
1104: (int)a = 5
1105: (int)(a = (char *)(int)5)
1106:
1107: An assignment-with-arithmetic operation such as `+=' applied to a
1108: cast performs the arithmetic using the type resulting from the cast,
1109: and then continues as in the previous case. Therefore, these two
1110: expressions are equivalent:
1111:
1112: (int)a += 5
1113: (int)(a = (char *)(int) ((int)a + 5))
1114:
1115: You cannot take the address of an lvalue cast, because the use of its
1116: address would not work out coherently. Suppose that `&(int)f' were
1117: permitted, where `f' has type `float'. Then the following statement
1118: would try to store an integer bit-pattern where a floating point number
1119: belongs:
1120:
1121: *&(int)f = 1;
1122:
1123: This is quite different from what `(int)f = 1' would do--that would
1124: convert 1 to floating point and store it. Rather than cause this
1125: inconsistency, we think it is better to prohibit use of `&' on a cast.
1126:
1127: If you really do want an `int *' pointer with the address of `f',
1128: you can simply write `(int *)&f'.
1129:
1130:
1131: File: gcc.info, Node: Conditionals, Next: Long Long, Prev: Lvalues, Up: C Extensions
1132:
1133: Conditionals with Omitted Operands
1134: ==================================
1135:
1136: The middle operand in a conditional expression may be omitted. Then
1137: if the first operand is nonzero, its value is the value of the
1138: conditional expression.
1139:
1140: Therefore, the expression
1141:
1142: x ? : y
1143:
1144: has the value of `x' if that is nonzero; otherwise, the value of `y'.
1145:
1146: This example is perfectly equivalent to
1147:
1148: x ? x : y
1149:
1150: In this simple case, the ability to omit the middle operand is not
1151: especially useful. When it becomes useful is when the first operand
1152: does, or may (if it is a macro argument), contain a side effect. Then
1153: repeating the operand in the middle would perform the side effect
1154: twice. Omitting the middle operand uses the value already computed
1155: without the undesirable effects of recomputing it.
1156:
1157:
1158: File: gcc.info, Node: Long Long, Next: Complex, Prev: Conditionals, Up: C Extensions
1159:
1160: Double-Word Integers
1161: ====================
1162:
1163: GNU C supports data types for integers that are twice as long as
1164: `long int'. Simply write `long long int' for a signed integer, or
1165: `unsigned long long int' for an unsigned integer. To make an integer
1166: constant of type `long long int', add the suffix `LL' to the integer.
1167: To make an integer constant of type `unsigned long long int', add the
1168: suffix `ULL' to the integer.
1169:
1170: You can use these types in arithmetic like any other integer types.
1171: Addition, subtraction, and bitwise boolean operations on these types
1172: are open-coded on all types of machines. Multiplication is open-coded
1173: if the machine supports fullword-to-doubleword a widening multiply
1174: instruction. Division and shifts are open-coded only on machines that
1175: provide special support. The operations that are not open-coded use
1176: special library routines that come with GNU CC.
1177:
1178: There may be pitfalls when you use `long long' types for function
1179: arguments, unless you declare function prototypes. If a function
1180: expects type `int' for its argument, and you pass a value of type `long
1181: long int', confusion will result because the caller and the subroutine
1182: will disagree about the number of bytes for the argument. Likewise, if
1183: the function expects `long long int' and you pass `int'. The best way
1184: to avoid such problems is to use prototypes.
1185:
1186:
1187: File: gcc.info, Node: Complex, Next: Zero Length, Prev: Long Long, Up: C Extensions
1188:
1189: Complex Numbers
1190: ===============
1191:
1192: GNU C supports complex data types. You can declare both complex
1193: integer types and complex floating types, using the keyword
1194: `__complex__'.
1195:
1196: For example, `__complex__ double x;' declares `x' as a variable
1197: whose real part and imaginary part are both of type `double'.
1198: `__complex__ short int y;' declares `y' to have real and imaginary
1199: parts of type `short int'; this is not likely to be useful, but it
1200: shows that the set of complex types is complete.
1201:
1202: To write a constant with a complex data type, use the suffix `i' or
1203: `j' (either one; they are equivalent). For example, `2.5fi' has type
1204: `__complex__ float' and `3i' has type `__complex__ int'. Such a
1205: constant always has a pure imaginary value, but you can form any
1206: complex value you like by adding one to a real constant.
1207:
1208: To extract the real part of a complex-valued expression EXP, write
1209: `__real__ EXP'. Likewise, use `__imag__' to extract the imaginary part.
1210:
1211: The operator `~' performs complex conjugation when used on a value
1212: with a complex type.
1213:
1214: GNU CC can allocate complex automatic variables in a noncontiguous
1215: fashion; it's even possible for the real part to be in a register while
1216: the imaginary part is on the stack (or vice-versa). None of the
1217: supported debugging info formats has a way to represent noncontiguous
1218: allocation like this, so GNU CC describes a noncontiguous complex
1219: variable as if it were two separate variables of noncomplex type. If
1220: the variable's actual name is `foo', the two fictitious variables are
1221: named `foo$real' and `foo$imag'. You can examine and set these two
1222: fictitious variables with your debugger.
1223:
1224: A future version of GDB will know how to recognize such pairs and
1225: treat them as a single variable with a complex type.
1226:
1227:
1228: File: gcc.info, Node: Zero Length, Next: Variable Length, Prev: Complex, Up: C Extensions
1229:
1230: Arrays of Length Zero
1231: =====================
1232:
1233: Zero-length arrays are allowed in GNU C. They are very useful as
1234: the last element of a structure which is really a header for a
1235: variable-length object:
1236:
1237: struct line {
1238: int length;
1239: char contents[0];
1240: };
1241:
1242: {
1243: struct line *thisline = (struct line *)
1244: malloc (sizeof (struct line) + this_length);
1245: thisline->length = this_length;
1246: }
1247:
1248: In standard C, you would have to give `contents' a length of 1, which
1249: means either you waste space or complicate the argument to `malloc'.
1250:
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.