]>
Commit | Line | Data |
---|---|---|
61952351 UD |
1 | Frequently Asked Questions about the GNU C Library |
2 | ||
3 | This document tries to answer questions a user might have when | |
4 | installing and using glibc. Please make sure you read this before | |
5 | sending questions or bug reports to the maintainers. | |
6 | ||
7 | The GNU C library is very complex. The installation process has not | |
8 | been completely automated; there are too many variables. You can do | |
9 | substantial damage to your system by installing the library | |
10 | incorrectly. Make sure you understand what you are undertaking before | |
11 | you begin. | |
12 | ||
13 | If you have any questions you think should be answered in this document, | |
14 | please let me know. | |
15 | ||
16 | --drepper@cygnus.com | |
17 | \f | |
18 | ? Compiling glibc | |
19 | ||
20 | ?? What systems does the GNU C Library run on? | |
21 | ||
22 | {UD} This is difficult to answer. The file `README' lists the | |
23 | architectures GNU libc was known to run on *at some time*. This does | |
24 | not mean that it still can be compiled and run on them now. | |
25 | ||
26 | The systems glibc is known to work on as of this release, and most | |
27 | probably in the future, are: | |
28 | ||
29 | *-*-gnu GNU Hurd | |
30 | i[3456]86-*-linux-gnu Linux-2.0 on Intel | |
31 | m68k-*-linux-gnu Linux-2.0 on Motorola 680x0 | |
32 | alpha-*-linux-gnu Linux-2.0 on DEC Alpha | |
33 | powerpc-*-linux-gnu Linux and MkLinux on PowerPC systems | |
34 | sparc-*-linux-gnu Linux-2.0 on SPARC | |
35 | sparc64-*-linux-gnu Linux-2.0 on UltraSPARC | |
36 | ||
37 | Ports to other Linux platforms are in development, and may in fact | |
38 | work already, but no one has sent us success reports for them. | |
39 | Currently no ports to other operating systems are underway, although a | |
40 | few people have expressed interest. | |
41 | ||
42 | If you have a system not listed above (or in the `README' file) and | |
43 | you are really interested in porting it, contact | |
44 | ||
45 | <bug-glibc@prep.ai.mit.edu> | |
46 | ||
47 | ?? What compiler do I need to build GNU libc? | |
48 | ||
49 | {UD} You must use GNU CC to compile GNU libc. A lot of extensions of | |
50 | GNU CC are used to increase portability and speed. | |
51 | ||
52 | GNU CC is found, like all other GNU packages, on | |
53 | ftp://prep.ai.mit.edu/pub/gnu | |
54 | and the many mirror sites. prep is always overloaded, so try to find | |
55 | a local mirror first. | |
56 | ||
57 | You always should try to use the latest official release. Older | |
58 | versions may not have all the features GNU libc requires. On most | |
59 | supported platforms, 2.7.2.3 is the earliest version that works at all. | |
60 | ||
61 | ?? When I try to compile glibc I get only error messages. | |
62 | What's wrong? | |
63 | ||
64 | {UD} You definitely need GNU make to translate GNU libc. No | |
65 | other make program has the needed functionality. | |
66 | ||
67 | We recommend version GNU make version 3.75. Versions 3.76 and 3.76.1 | |
68 | have bugs which appear when building big projects like GNU libc. | |
69 | Versions before 3.74 have bugs and/or are missing features. | |
70 | ||
71 | ?? Do I need a special linker or archiver? | |
72 | ||
73 | {UD} You may be able to use your system linker, but GNU libc works | |
74 | best with GNU binutils. | |
75 | ||
76 | On systems where the native linker does not support weak symbols you | |
77 | will not get a fully ISO C compliant C library. Generally speaking | |
78 | you should use the GNU binutils if they provide at least the same | |
79 | functionality as your system's tools. | |
80 | ||
81 | Always get the newest release of GNU binutils available. Older | |
82 | releases are known to have bugs that prevent a successful compilation. | |
83 | ||
84 | {ZW} As of release 2.1 a linker supporting symbol versions is | |
85 | required. For Linux, get binutils-2.8.1.0.17 or later. Other systems | |
86 | may have native linker support, but it's moot right now, because glibc | |
87 | has not been ported to them. | |
88 | ||
89 | ?? Do I need some more things to compile GNU C Library? | |
90 | ||
91 | {UD} Yes, there are some more :-). | |
92 | ||
93 | * GNU gettext. This package contains the tools needed to construct | |
94 | `message catalog' files containing translated versions of system | |
95 | messages. See ftp://prep.ai.mit.edu/pub/gnu or better any mirror | |
96 | site. (We distribute compiled message catalogs, but they may not be | |
97 | updated in patches.) | |
98 | ||
99 | * Some files depend on special tools. E.g., files ending in .gperf | |
100 | need a `gperf' program. The GNU version (part of libg++) is known | |
101 | to work while some vendor versions do not. | |
102 | ||
103 | You should not need these tools unless you change the source files. | |
104 | ||
105 | * When compiling for Linux, the header files of the Linux kernel must | |
106 | be available to the compiler as <linux/*.h> and <asm/*.h>. | |
107 | ||
108 | * lots of disk space (~170MB for i?86-linux; more for RISC platforms). | |
109 | ||
110 | * plenty of time. Compiling just the shared and static libraries for | |
111 | i?86-linux takes approximately 1h on an i586@133, or 2.5h on | |
112 | i486@66, or 4.5h on i486@33. Multiply this by 1.5 or 2.0 if you | |
113 | build profiling and/or the highly optimized version as well. For | |
114 | Hurd systems times are much higher. | |
115 | ||
116 | You should avoid compiling in a NFS mounted filesystem. This is | |
117 | very slow. | |
118 | ||
119 | James Troup <J.J.Troup@comp.brad.ac.uk> reports a compile time of | |
120 | 45h34m for a full build (shared, static, and profiled) on | |
121 | Atari Falcon (Motorola 68030 @ 16 Mhz, 14 Mb memory) and 22h48m | |
122 | on Atari TT030 (Motorola 68030 @ 32 Mhz, 34 Mb memory) | |
123 | ||
124 | If you have some more measurements let me know. | |
125 | ||
126 | ?? When I run `nm -u libc.so' on the produced library I still | |
127 | find unresolved symbols. Can this be ok? | |
128 | ||
129 | {UD} Yes, this is ok. There can be several kinds of unresolved | |
130 | symbols: | |
131 | ||
132 | * magic symbols automatically generated by the linker. These have names | |
133 | like __start_* and __stop_* | |
134 | ||
135 | * symbols starting with _dl_* come from the dynamic linker | |
136 | ||
137 | * symbols resolved by using libgcc.a | |
138 | (__udivdi3, __umoddi3, or similar) | |
139 | ||
140 | * weak symbols, which need not be resolved at all (fabs for example) | |
141 | ||
142 | Generally, you should make sure you find a real program which produces | |
143 | errors while linking before deciding there is a problem. | |
144 | ||
145 | ??addon What are these `add-ons'? | |
146 | ||
147 | {UD} To avoid complications with export rules or external source | |
148 | code some optional parts of the libc are distributed as separate | |
149 | packages (e.g., the crypt package, see ?crypt). | |
150 | ||
151 | To use these packages as part of GNU libc, just unpack the tarfiles in | |
152 | the libc source directory and tell the configuration script about them | |
153 | using the --enable-add-ons option. If you give just --enable-add-ons | |
154 | configure tries to find all the add-on packages in your source tree. | |
155 | This may not work. If it doesn't, or if you want to select only a | |
156 | subset of the add-ons, give a comma-separated list of the add-ons to | |
157 | enable: | |
158 | ||
159 | configure --enable-add-ons=crypt,linuxthreads | |
160 | ||
161 | for example. | |
162 | ||
163 | Add-ons can add features (including entirely new shared libraries), | |
164 | override files, provide support for additional architectures, and | |
165 | just about anything else. The existing makefiles do most of the work; | |
166 | only some few stub rules must be written to get everything running. | |
167 | ||
168 | ?? My XXX kernel emulates a floating-point coprocessor for me. | |
169 | Should I enable --with-fp? | |
170 | ||
171 | {ZW} An emulated FPU is just as good as a real one, as far as the C | |
172 | library is concerned. You only need to say --without-fp if your | |
173 | machine has no way to execute floating-point instructions. | |
174 | ||
175 | People who are interested in squeezing the last drop of performance | |
176 | out of their machine may wish to avoid the trap overhead, but this is | |
177 | far more trouble than it's worth: you then have to compile | |
178 | *everything* this way, including the compiler's internal libraries | |
179 | (libgcc.a for GNU C), because the calling conventions change. | |
180 | ||
181 | ?? When compiling GNU libc I get lots of errors saying functions | |
182 | in glibc are duplicated in libgcc. | |
183 | ||
184 | {EY} This is *exactly* the same problem that I was having. The | |
185 | problem was due to the fact that configure didn't correctly detect | |
186 | that the linker flag --no-whole-archive was supported in my linker. | |
187 | In my case it was because I had run ./configure with bogus CFLAGS, and | |
188 | the test failed. | |
189 | ||
190 | One thing that is particularly annoying about this problem is that | |
191 | once this is misdetected, running configure again won't fix it unless | |
192 | you first delete config.cache. | |
193 | ||
194 | {UD} Starting with glibc-2.0.3 there should be a better test to avoid | |
195 | some problems of this kind. The setting of CFLAGS is checked at the | |
196 | very beginning and if it is not usable `configure' will bark. | |
197 | ||
198 | ?? What's the problem with configure --enable-omitfp? | |
199 | ||
200 | {AJ} When --enable-omitfp is set the libraries are built without frame | |
201 | pointers. Some compilers produce buggy code for this model and | |
202 | therefore we don't advise using it at the moment. | |
203 | ||
204 | If you use --enable-omitfp, you're on your own. If you encounter | |
205 | problems with a library that was build this way, we advise you to | |
206 | rebuild the library without --enable-omitfp. If the problem vanishes | |
207 | consider tracking the problem down and report it as compiler failure. | |
208 | ||
209 | Since a library build with --enable-omitfp is undebuggable on most | |
210 | systems, debuggable libraries are also built - you can use it by | |
211 | appending "_g" to the library names. | |
212 | ||
213 | The compilation of these extra libraries and the compiler optimizations | |
214 | slow down the build process and need more disk space. | |
215 | ||
216 | ? Installation and configuration issues | |
217 | ||
218 | ?? Can I replace the libc on my Linux system with GNU libc? | |
219 | ||
220 | {UD} You cannot replace any existing libc for Linux with GNU | |
221 | libc. It is binary incompatible and therefore has a different major | |
222 | version. You can, however, install it alongside your existing libc. | |
223 | ||
224 | For Linux there are three major libc versions: | |
225 | libc-4 a.out libc | |
226 | libc-5 original ELF libc | |
227 | libc-6 GNU libc | |
228 | ||
229 | You can have any combination of these three installed. For more | |
230 | information consult documentation for shared library handling. The | |
231 | Makefiles of GNU libc will automatically generate the needed symbolic | |
232 | links which the linker will use. | |
233 | ||
234 | ?? How do I configure GNU libc so that the essential libraries | |
235 | like libc.so go into /lib and the other into /usr/lib? | |
236 | ||
237 | {UD,AJ} Like all other GNU packages GNU libc is designed to use a base | |
238 | directory and install all files relative to this. The default is | |
239 | /usr/local, because this is safe (it will not damage the system if | |
240 | installed there). If you wish to install GNU libc as the primary C | |
241 | library on your system, set the base directory to /usr (i.e. run | |
242 | configure --prefix=/usr <other_options>). Note that this can damage | |
243 | your system; see ?safety for details. | |
244 | ||
245 | Some systems like Linux have a filesystem standard which makes a | |
246 | difference between essential libraries and others. Essential | |
247 | libraries are placed in /lib because this directory is required to be | |
248 | located on the same disk partition as /. The /usr subtree might be | |
249 | found on another partition/disk. If you configure for Linux with | |
250 | --prefix=/usr, then this will be done automatically. | |
251 | ||
252 | To install the essential libraries which come with GNU libc in /lib on | |
253 | systems other than Linux one must explicitly request it. Autoconf has | |
254 | no option for this so you have to use a `configparms' file (see the | |
255 | `INSTALL' file for details). It should contain: | |
256 | ||
257 | slibdir=/lib | |
258 | sysconfdir=/etc | |
259 | ||
260 | The first line specifies the directory for the essential libraries, | |
261 | the second line the directory for system configuration files. | |
262 | ||
263 | ??safety How should I avoid damaging my system when I install GNU libc? | |
264 | ||
265 | {ZW} If you wish to be cautious, do not configure with --prefix=/usr. | |
266 | If you don't specify a prefix, glibc will be installed in /usr/local, | |
267 | where it will probably not break anything. (If you wish to be | |
268 | certain, set the prefix to something like /usr/local/glibc2 which is | |
269 | not used for anything.) | |
270 | ||
271 | The dangers when installing glibc in /usr are twofold: | |
272 | ||
273 | * glibc will overwrite the headers in /usr/include. Other C libraries | |
274 | install a different but overlapping set of headers there, so the | |
275 | effect will probably be that you can't compile anything. You need to | |
276 | rename /usr/include out of the way first. (Do not throw it away; you | |
277 | will then lose the ability to compile programs against your old libc.) | |
278 | ||
279 | * None of your old libraries, static or shared, can be used with a | |
280 | different C library major version. For shared libraries this is not a | |
281 | problem, because the filenames are different and the dynamic linker | |
282 | will enforce the restriction. But static libraries have no version | |
283 | information. You have to evacuate all the static libraries in | |
284 | /usr/lib to a safe location. | |
285 | ||
286 | The situation is rather similar to the move from a.out to ELF which | |
287 | long-time Linux users will remember. | |
288 | ||
289 | ?? Do I need to use GNU CC to compile programs that will use the | |
290 | GNU C Library? | |
291 | ||
292 | {ZW} In theory, no; the linker does not care, and the headers are | |
293 | supposed to check for GNU CC before using its extensions to the C | |
294 | language. | |
295 | ||
296 | However, there are currently no ports of glibc to systems where | |
297 | another compiler is the default, so no one has tested the headers | |
298 | extensively against another compiler. You may therefore encounter | |
299 | difficulties. If you do, please report them as bugs. | |
300 | ||
301 | Also, in several places GNU extensions provide large benefits in code | |
302 | quality. For example, the library has hand-optimized, inline assembly | |
303 | versions of some string functions. These can only be used with GCC. | |
304 | See ?string for details. | |
305 | ||
306 | ??crypt When linking with the new libc I get unresolved symbols | |
307 | `crypt' and `setkey'. Why aren't these functions in the | |
308 | libc anymore? | |
309 | ||
310 | {UD} The US places restrictions on exporting cryptographic programs | |
311 | and source code. Until this law gets abolished we cannot ship the | |
312 | cryptographic functions together with glibc. | |
313 | ||
314 | The functions are available, as an add-on (see ?addon). People in the | |
315 | US may get it from the same place they got GNU libc from. People | |
316 | outside the US should get the code from ftp://ftp.ifi.uio.no/pub/gnu, | |
317 | or another archive site outside the USA. The README explains how to | |
318 | install the sources. | |
319 | ||
320 | If you already have the crypt code on your system the reason for the | |
321 | failure is probably that you did not link with -lcrypt. The crypto | |
322 | functions are in a separate library to make it possible to export GNU | |
323 | libc binaries from the US. | |
324 | ||
325 | ?? When I use GNU libc on my Linux system by linking against | |
326 | the libc.so which comes with glibc all I get is a core dump. | |
327 | ||
328 | {UD} On Linux, gcc sets the dynamic linker to /lib/ld-linux.so.1 | |
329 | unless the user specifies a -dynamic-linker argument. This is the | |
330 | name of the libc5 dynamic linker, which does not work with glibc. | |
331 | ||
332 | For casual use of GNU libc you can just specify | |
333 | -dynamic-linker=/lib/ld-linux.so.2 | |
334 | ||
335 | which is the glibc dynamic linker, on Linux systems. On other systems | |
336 | the name is /lib/ld.so.1. | |
337 | ||
338 | To change your environment to use GNU libc for compiling you need to | |
339 | change the `specs' file of your gcc. This file is normally found at | |
340 | ||
341 | /usr/lib/gcc-lib/<arch>/<version>/specs | |
342 | ||
343 | In this file you have to change a few things: | |
344 | ||
345 | - change `ld-linux.so.1' to `ld-linux.so.2' | |
346 | ||
347 | - remove all expression `%{...:-lgmon}'; there is no libgmon in glibc | |
348 | ||
349 | - fix a minor bug by changing %{pipe:-} to %| | |
350 | ||
351 | Here is what the gcc-2.7.2 specs file should look like when GNU libc | |
352 | is installed at /usr: | |
353 | ||
354 | ----------------------------------------------------------------------- | |
355 | *asm: | |
356 | %{V} %{v:%{!V:-V}} %{Qy:} %{!Qn:-Qy} %{n} %{T} %{Ym,*} %{Yd,*} %{Wa,*:%*} | |
357 | ||
358 | *asm_final: | |
359 | %| | |
360 | ||
361 | *cpp: | |
362 | %{fPIC:-D__PIC__ -D__pic__} %{fpic:-D__PIC__ -D__pic__} %{!m386:-D__i486__} %{posix:-D_POSIX_SOURCE} %{pthread:-D_REENTRANT} | |
363 | ||
364 | *cc1: | |
365 | %{profile:-p} | |
366 | ||
367 | *cc1plus: | |
368 | ||
369 | ||
370 | *endfile: | |
371 | %{!shared:crtend.o%s} %{shared:crtendS.o%s} crtn.o%s | |
372 | ||
373 | *link: | |
374 | -m elf_i386 %{shared:-shared} %{!shared: %{!ibcs: %{!static: %{rdynamic:-export-dynamic} %{!dynamic-linker:-dynamic-linker /lib/ld-linux.so.2}} %{static:-static}}} | |
375 | ||
376 | *lib: | |
377 | %{!shared: %{pthread:-lpthread} %{profile:-lc_p} %{!profile: -lc}} | |
378 | ||
379 | *libgcc: | |
380 | -lgcc | |
381 | ||
382 | *startfile: | |
383 | %{!shared: %{pg:gcrt1.o%s} %{!pg:%{p:gcrt1.o%s} %{!p:%{profile:gcrt1.o%s} %{!profile:crt1.o%s}}}} crti.o%s %{!shared:crtbegin.o%s} %{shared:crtbeginS.o%s} | |
384 | ||
385 | *switches_need_spaces: | |
386 | ||
387 | ||
388 | *signed_char: | |
389 | %{funsigned-char:-D__CHAR_UNSIGNED__} | |
390 | ||
391 | *predefines: | |
392 | -D__ELF__ -Dunix -Di386 -Dlinux -Asystem(unix) -Asystem(posix) -Acpu(i386) -Amachine(i386) | |
393 | ||
394 | *cross_compile: | |
395 | 0 | |
396 | ||
397 | *multilib: | |
398 | . ; | |
399 | ||
400 | ----------------------------------------------------------------------- | |
401 | ||
402 | Things get a bit more complicated if you have GNU libc installed in | |
403 | some other place than /usr, i.e., if you do not want to use it instead | |
404 | of the old libc. In this case the needed startup files and libraries | |
405 | are not found in the regular places. So the specs file must tell the | |
406 | compiler and linker exactly what to use. | |
407 | ||
408 | Version 2.7.2.3 does and future versions of GCC will automatically | |
409 | provide the correct specs. | |
410 | ||
411 | ?? Looking through the shared libc file I haven't found the | |
412 | functions `stat', `lstat', `fstat', and `mknod' and while | |
413 | linking on my Linux system I get error messages. How is | |
414 | this supposed to work? | |
415 | ||
416 | {RM} Believe it or not, stat and lstat (and fstat, and mknod) | |
417 | are supposed to be undefined references in libc.so.6! Your problem is | |
418 | probably a missing or incorrect /usr/lib/libc.so file; note that this | |
419 | is a small text file now, not a symlink to libc.so.6. It should look | |
420 | something like this: | |
421 | ||
422 | GROUP ( libc.so.6 ld.so.1 libc.a ) | |
423 | ||
424 | or in ix86/Linux and alpha/Linux: | |
425 | ||
426 | GROUP ( libc.so.6 ld-linux.so.2 libc.a ) | |
427 | ||
428 | ?? How can I compile gcc 2.7.2.1 from the gcc source code using | |
429 | glibc 2.x? | |
430 | ||
431 | {AJ} There's only correct support for glibc 2.0.x in gcc 2.7.2.3 | |
432 | or later. You should get at least gcc 2.7.2.3. All previous versions | |
433 | had problems with glibc support. | |
434 | ||
435 | ?? The `gencat' utility cannot process the catalog sources which | |
436 | were used on my Linux libc5 based system. Why? | |
437 | ||
438 | {UD} The `gencat' utility provided with glibc complies to the XPG | |
439 | standard. The older Linux version did not obey the standard, so they | |
440 | are not compatible. | |
441 | ||
442 | To ease the transition from the Linux version some of the non-standard | |
443 | features are also present in the `gencat' program of GNU libc. This | |
444 | mainly includes the use of symbols for the message number and the automatic | |
445 | generation of header files which contain the needed #defines to map the | |
446 | symbols to integers. | |
447 | ||
448 | Here is a simple SED script to convert at least some Linux specific | |
449 | catalog files to the XPG4 form: | |
450 | ||
451 | ----------------------------------------------------------------------- | |
452 | # Change catalog source in Linux specific format to standard XPG format. | |
453 | # Ulrich Drepper <drepper@cygnus.com>, 1996. | |
454 | # | |
455 | /^\$ #/ { | |
456 | h | |
457 | s/\$ #\([^ ]*\).*/\1/ | |
458 | x | |
459 | s/\$ #[^ ]* *\(.*\)/\$ \1/ | |
460 | } | |
461 | ||
462 | /^# / { | |
463 | s/^# \(.*\)/\1/ | |
464 | G | |
465 | s/\(.*\)\n\(.*\)/\2 \1/ | |
466 | } | |
467 | ----------------------------------------------------------------------- | |
468 | ||
469 | ?? I have set up /etc/nis.conf, and the Linux libc 5 with NYS | |
470 | works great. But the glibc NIS+ doesn't seem to work. | |
471 | ||
d71b808a | 472 | {TK} The glibc NIS+ implementation uses a /var/nis/NIS_COLD_START |
61952351 UD |
473 | file for storing information about the NIS+ server and their public |
474 | keys, because the nis.conf file does not contain all the necessary | |
475 | information. You have to copy a NIS_COLD_START file from a Solaris | |
476 | client (the NIS_COLD_START file is byte order independent) or generate | |
477 | it with nisinit from the nis-tools package (available at | |
478 | http://www-vt.uni-paderborn.de/~kukuk/linux/nisplus.html). | |
479 | ||
480 | ?? After installing glibc name resolving doesn't work properly. | |
481 | ||
482 | {AJ} You probably should read the manual section describing | |
483 | nsswitch.conf (just type `info libc "NSS Configuration File"'). | |
484 | The NSS configuration file is usually the culprit. | |
485 | ||
486 | ?? I have /usr/include/net and /usr/include/scsi as symlinks | |
487 | into my Linux source tree. Is that wrong? | |
488 | ||
489 | {PB} This was necessary for libc5, but is not correct when using | |
490 | glibc. Including the kernel header files directly in user programs | |
491 | usually does not work (see ?kerhdr). glibc provides its own <net/*> | |
492 | and <scsi/*> header files to replace them, and you may have to remove | |
493 | any symlink that you have in place before you install glibc. However, | |
494 | /usr/include/asm and /usr/include/linux should remain as they were. | |
495 | ||
496 | ?? Programs like `logname', `top', `uptime' `users', `w' and | |
497 | `who', show incorrect information about the (number of) | |
498 | users on my system. Why? | |
499 | ||
500 | {MK} See ?getlog. | |
501 | ||
502 | ?? After upgrading to glibc 2.1 with symbol versioning I get | |
503 | errors about undefined symbols. What went wrong? | |
504 | ||
505 | {AJ} The problem is caused either by wrong program code or tools. In | |
506 | the versioned libc a lot of symbols are now local that were global | |
507 | symbols in previous versions. It seems that programs linked against | |
508 | older versions often accidentally used libc global variables -- | |
509 | something that should not happen. | |
510 | ||
511 | The only way to fix this is to recompile your program. Sorry, that's | |
512 | the price you might have to pay once for quite a number of advantages | |
513 | with symbol versioning. | |
514 | ||
515 | ?? When I start the program XXX after upgrading the library | |
516 | I get | |
517 | XXX: Symbol `_sys_errlist' has different size in shared | |
518 | object, consider re-linking | |
519 | Why? What should I do? | |
520 | ||
521 | {UD} As the message says, relink the binary. The problem is that | |
522 | a few symbols from the library can change in size and there is no way | |
523 | to avoid this. _sys_errlist is a good example. Occasionally there are | |
524 | new error numbers added to the kernel and this must be reflected at user | |
525 | level, breaking programs that refer to them directly. | |
526 | ||
527 | Such symbols should normally not be used at all. There are mechanisms | |
528 | to avoid using them. In the case of _sys_errlist, there is the | |
529 | strerror() function which should _always_ be used instead. So the | |
530 | correct fix is to rewrite that part of the application. | |
531 | ||
532 | In some situations (especially when testing a new library release) it | |
533 | might be possible that a symbol changed size when that should not have | |
534 | happened. So in case of doubt report such a warning message as a | |
535 | problem. | |
536 | ||
537 | ? Source and binary incompatibilities, and what to do about them | |
538 | ||
539 | ?? I expect GNU libc to be 100% source code compatible with | |
540 | the old Linux based GNU libc. Why isn't it like this? | |
541 | ||
542 | {DMT,UD} Not every extension in Linux libc's history was well | |
543 | thought-out. In fact it had a lot of problems with standards compliance | |
544 | and with cleanliness. With the introduction of a new version number these | |
545 | errors can now be corrected. Here is a list of the known source code | |
546 | incompatibilities: | |
547 | ||
548 | * _GNU_SOURCE: glibc does not make the GNU extensions available | |
549 | automatically. If a program depends on GNU extensions or some | |
550 | other non-standard functionality, it is necessary to compile it | |
551 | with the C compiler option -D_GNU_SOURCE, or better, to put | |
552 | `#define _GNU_SOURCE' at the beginning of your source files, before | |
553 | any C library header files are included. This difference normally | |
554 | manifests itself in the form of missing prototypes and/or data type | |
555 | definitions. Thus, if you get such errors, the first thing you | |
556 | should do is try defining _GNU_SOURCE and see if that makes the | |
557 | problem go away. | |
558 | ||
559 | For more information consult the file `NOTES' in the GNU C library | |
560 | sources. | |
561 | ||
562 | * reboot(): GNU libc sanitizes the interface of reboot() to be more | |
563 | compatible with the interface used on other OSes. reboot() as | |
564 | implemented in glibc takes just one argument. This argument | |
565 | corresponds to the third argument of the Linux reboot system call. | |
566 | That is, a call of the form reboot(a, b, c) needs to be changed into | |
567 | reboot(c). Beside this the header <sys/reboot.h> defines the needed | |
568 | constants for the argument. These RB_* constants should be used | |
569 | instead of the cryptic magic numbers. | |
570 | ||
571 | * swapon(): the interface of this function didn't change, but the | |
572 | prototype is in a separate header file <sys/swap.h>. This header | |
573 | file also provides the SWAP_* constants defined by <linux/swap.h>; | |
574 | you should use them for the second argument to swapon(). | |
575 | ||
576 | * errno: If a program uses the variable "errno", then it _must_ | |
577 | include <errno.h>. The old libc often (erroneously) declared this | |
578 | variable implicitly as a side-effect of including other libc header | |
579 | files. glibc is careful to avoid such namespace pollution, which, | |
580 | in turn, means that you really need to include the header files that | |
581 | you depend on. This difference normally manifests itself in the | |
582 | form of the compiler complaining about references to an undeclared | |
583 | symbol "errno". | |
584 | ||
585 | * Linux-specific syscalls: All Linux system calls now have appropriate | |
586 | library wrappers and corresponding declarations in various header files. | |
587 | This is because the syscall() macro that was traditionally used to | |
588 | work around missing syscall wrappers are inherently non-portable and | |
589 | error-prone. The following table lists all the new syscall stubs, | |
590 | the header-file declaring their interface and the system call name. | |
591 | ||
592 | syscall name: wrapper name: declaring header file: | |
593 | ------------- ------------- ---------------------- | |
594 | bdflush bdflush <sys/kdaemon.h> | |
595 | syslog ksyslog_ctl <sys/klog.h> | |
596 | ||
597 | * lpd: Older versions of lpd depend on a routine called _validuser(). | |
598 | The library does not provide this function, but instead provides | |
599 | __ivaliduser() which has a slightly different interface. Simply | |
600 | upgrading to a newer lpd should fix this problem (e.g., the 4.4BSD | |
601 | lpd is known to be working). | |
602 | ||
603 | * resolver functions/BIND: like on many other systems the functions of | |
604 | the resolver library are not included in libc itself. There is a | |
605 | separate library libresolv. If you get undefined symbol errors for | |
606 | symbols starting with `res_*' simply add -lresolv to your linker | |
607 | command line. | |
608 | ||
609 | * the `signal' function's behavior corresponds to the BSD semantic and | |
610 | not the SysV semantic as it was in libc-5. The interface on all GNU | |
611 | systems shall be the same and BSD is the semantic of choice. To use | |
612 | the SysV behavior simply use `sysv_signal', or define _XOPEN_SOURCE. | |
613 | See ?signal for details. | |
614 | ||
615 | ??getlog Why does getlogin() always return NULL on my Linux box? | |
616 | ||
617 | {UD} The GNU C library has a format for the UTMP and WTMP file which | |
618 | differs from what your system currently has. It was extended to | |
619 | fulfill the needs of the next years when IPv6 is introduced. The | |
620 | record size is different and some fields have different positions. | |
621 | The files written by functions from the one library cannot be read by | |
622 | functions from the other library. Sorry, but this is what a major | |
623 | release is for. It's better to have a cut now than having no means to | |
624 | support the new techniques later. | |
625 | ||
626 | {MK} There is however a (partial) solution for this problem. Please | |
627 | take a look at the file `README.utmpd'. | |
628 | ||
629 | ?? Where are the DST_* constants found in <sys/time.h> on many | |
630 | systems? | |
631 | ||
632 | {UD} These constants come from the old BSD days and are not used | |
633 | anymore (libc5 does not actually implement the handling although the | |
634 | constants are defined). | |
635 | ||
636 | Instead GNU libc contains zone database support and compatibility code | |
637 | for POSIX TZ environment variable handling. | |
638 | ||
639 | ?? The prototypes for `connect', `accept', `getsockopt', | |
640 | `setsockopt', `getsockname', `getpeername', `send', | |
641 | `sendto', and `recvfrom' are different in GNU libc from | |
642 | any other system I saw. This is a bug, isn't it? | |
643 | ||
644 | {UD} No, this is no bug. This version of GNU libc already follows the | |
645 | new Single Unix specifications (and I think the POSIX.1g draft which | |
646 | adopted the solution). The type for a parameter describing a size is | |
647 | now `socklen_t', a new type. | |
648 | ||
649 | ??kerhdr On Linux I've got problems with the declarations in Linux | |
650 | kernel headers. | |
651 | ||
652 | {UD,AJ} On Linux, the use of kernel headers is reduced to the minimum. | |
653 | This gives Linus the ability to change the headers more freely. Also, | |
654 | user programs are now insulated from changes in the size of kernel | |
655 | data structures. | |
656 | ||
657 | For example, the sigset_t type is 32 or 64 bits wide in the kernel. | |
658 | In glibc it is 1024 bits wide. This guarantees that when the kernel | |
659 | gets a bigger sigset_t (for POSIX.1e realtime support, say) user | |
660 | programs will not have to be recompiled. Consult the header files for | |
661 | more information about the changes. | |
662 | ||
663 | Therefore you shouldn't include Linux kernel header files directly if | |
664 | glibc has defined a replacement. Otherwise you might get undefined | |
665 | results because of type conflicts. | |
666 | ||
667 | ?? I don't include any kernel headers myself but the compiler | |
668 | still complains about redeclarations of types in the kernel | |
669 | headers. | |
670 | ||
671 | {UD} The kernel headers before Linux 2.1.61 don't work correctly with | |
672 | glibc. Compiling C programs is possible in most cases but C++ | |
673 | programs have (due to the change of the name lookups for `struct's) | |
674 | problems. One prominent example is `struct fd_set'. | |
675 | ||
676 | There might be some problems left but 2.1.61 fixes most of the known | |
677 | ones. See the BUGS file for other known problems. | |
678 | ||
679 | ??signal Why don't signals interrupt system calls anymore? | |
680 | ||
681 | {ZW} By default GNU libc uses the BSD semantics for signal(), | |
682 | unlike Linux libc 5 which used System V semantics. This is partially | |
683 | for compatibility with other systems and partially because the BSD | |
684 | semantics tend to make programming with signals easier. | |
685 | ||
686 | There are three differences: | |
687 | ||
688 | * BSD-style signals that occur in the middle of a system call do not | |
689 | affect the system call; System V signals cause the system call to | |
690 | fail and set errno to EINTR. | |
691 | ||
692 | * BSD signal handlers remain installed once triggered. System V signal | |
693 | handlers work only once, so one must reinstall them each time. | |
694 | ||
695 | * A BSD signal is blocked during the execution of its handler. In other | |
696 | words, a handler for SIGCHLD (for example) does not need to worry about | |
697 | being interrupted by another SIGCHLD. It may, however, be interrupted | |
698 | by other signals. | |
699 | ||
700 | There is general consensus that for `casual' programming with signals, the | |
701 | BSD semantics are preferable. You don't need to worry about system calls | |
702 | returning EINTR, and you don't need to worry about the race conditions | |
703 | associated with one-shot signal handlers. | |
704 | ||
705 | If you are porting an old program that relies on the old semantics, you can | |
706 | quickly fix the problem by changing signal() to sysv_signal() throughout. | |
707 | Alternatively, define _XOPEN_SOURCE before including <signal.h>. | |
708 | ||
709 | For new programs, the sigaction() function allows you to specify precisely | |
710 | how you want your signals to behave. All three differences listed above are | |
711 | individually switchable on a per-signal basis with this function. | |
712 | ||
713 | If all you want is for one specific signal to cause system calls to fail | |
714 | and return EINTR (for example, to implement a timeout) you can do this with | |
715 | siginterrupt(). | |
716 | ||
717 | ||
718 | ??string I've got errors compiling code that uses certain string | |
719 | functions. Why? | |
720 | ||
721 | {AJ} glibc 2.1 has special string functions that are faster | |
722 | than the normal library functions. Some of the functions are | |
723 | implemented as inline functions and others as macros. | |
724 | ||
725 | The optimized string functions are only used when compiling with | |
726 | optimizations (-O1 or higher). The behavior can be changed with two | |
727 | feature macros: | |
728 | ||
729 | * __NO_STRING_INLINES: Don't do any string optimizations. | |
730 | * __USE_STRING_INLINES: Use assembly language inline functions (might | |
731 | increase code size dramatically). | |
732 | ||
733 | Since some of these string functions are now additionally defined as | |
734 | macros, code like "char *strncpy();" doesn't work anymore (and is | |
735 | unnecessary, since <string.h> has the necessary declarations). Either | |
736 | change your code or define __NO_STRING_INLINES. | |
737 | ||
738 | {UD} Another problem in this area is that gcc still has problems on | |
739 | machines with very few registers (e.g., ix86). The inline assembler | |
740 | code can require almost all the registers and the register allocator | |
741 | cannot always handle this situation. | |
742 | ||
743 | One can disable the string optimizations selectively. Instead of writing | |
744 | ||
745 | cp = strcpy (foo, "lkj"); | |
746 | ||
747 | one can write | |
748 | ||
749 | cp = (strcpy) (foo, "lkj"); | |
750 | ||
751 | This disables the optimization for that specific call. | |
752 | ||
753 | ? Miscellaneous | |
754 | ||
755 | ?? After I changed configure.in I get `Autoconf version X.Y. | |
756 | or higher is required for this script'. What can I do? | |
757 | ||
758 | {UD} You have to get the specified autoconf version (or a later one) | |
759 | from your favorite mirror of prep.ai.mit.edu. | |
760 | ||
761 | ?? When I try to compile code which uses IPv6 headers and | |
762 | definitions on my Linux 2.x.y system I am in trouble. | |
763 | Nothing seems to work. | |
764 | ||
765 | {UD} The problem is that IPv6 development still has not reached a | |
766 | point where the headers are stable. There are still lots of | |
767 | incompatible changes made and the libc headers have to follow. | |
768 | ||
769 | Also, make sure you have a suitably recent kernel. As of the 970401 | |
770 | snapshot, according to Philip Blundell <philb@gnu.ai.mit.edu>, the | |
771 | required kernel version is 2.1.30. | |
772 | ||
773 | \f | |
774 | Answers were given by: | |
775 | {UD} Ulrich Drepper, <drepper@cygnus.com> | |
776 | {DMT} David Mosberger-Tang, <davidm@AZStarNet.com> | |
777 | {RM} Roland McGrath, <roland@gnu.org> | |
778 | {AJ} Andreas Jaeger, <aj@arthur.rhein-neckar.de> | |
779 | {EY} Eric Youngdale, <eric@andante.jic.com> | |
780 | {PB} Phil Blundell, <Philip.Blundell@pobox.com> | |
781 | {MK} Mark Kettenis, <kettenis@phys.uva.nl> | |
782 | {ZW} Zack Weinberg, <zack@rabi.phys.columbia.edu> | |
d71b808a | 783 | {TK} Thorsten Kukuk, <kukuk@vt.uni-paderborn.de> |
61952351 UD |
784 | \f |
785 | Local Variables: | |
786 | mode:outline | |
787 | outline-regexp:"\\?" | |
788 | End: |