]>
Commit | Line | Data |
---|---|---|
61952351 UD |
1 | Frequently Asked Questions about the GNU C Library |
2 | ||
f12944ec UD |
3 | This document tries to answer questions a user might have when installing |
4 | and using glibc. Please make sure you read this before sending questions or | |
5 | bug reports to the maintainers. | |
61952351 | 6 | |
f12944ec | 7 | The GNU C library is very complex. The installation process has not been |
fdacb17d | 8 | completely automated; there are too many variables. You can do substantial |
f12944ec UD |
9 | damage to your system by installing the library incorrectly. Make sure you |
10 | understand what you are undertaking before you begin. | |
61952351 UD |
11 | |
12 | If you have any questions you think should be answered in this document, | |
13 | please let me know. | |
14 | ||
15 | --drepper@cygnus.com | |
16 | \f | |
17 | ? Compiling glibc | |
18 | ||
19 | ?? What systems does the GNU C Library run on? | |
20 | ||
f12944ec UD |
21 | {UD} This is difficult to answer. The file `README' lists the architectures |
22 | GNU libc was known to run on *at some time*. This does not mean that it | |
23 | still can be compiled and run on them now. | |
61952351 | 24 | |
f12944ec UD |
25 | The systems glibc is known to work on as of this release, and most probably |
26 | in the future, are: | |
61952351 UD |
27 | |
28 | *-*-gnu GNU Hurd | |
bd355af0 UD |
29 | i[3456]86-*-linux-gnu Linux-2.x on Intel |
30 | m68k-*-linux-gnu Linux-2.x on Motorola 680x0 | |
31 | alpha-*-linux-gnu Linux-2.x on DEC Alpha | |
61952351 | 32 | powerpc-*-linux-gnu Linux and MkLinux on PowerPC systems |
bd355af0 UD |
33 | sparc-*-linux-gnu Linux-2.x on SPARC |
34 | sparc64-*-linux-gnu Linux-2.x on UltraSPARC | |
a35cb74d | 35 | arm-*-none ARM standalone systems |
cb0509a8 | 36 | arm-*-linux Linux-2.x on ARM |
a35cb74d | 37 | arm-*-linuxaout Linux-2.x on ARM using a.out binaries |
61952351 | 38 | |
f12944ec UD |
39 | Ports to other Linux platforms are in development, and may in fact work |
40 | already, but no one has sent us success reports for them. Currently no | |
41 | ports to other operating systems are underway, although a few people have | |
42 | expressed interest. | |
61952351 | 43 | |
f12944ec UD |
44 | If you have a system not listed above (or in the `README' file) and you are |
45 | really interested in porting it, contact | |
61952351 | 46 | |
b9b49b44 | 47 | <bug-glibc@gnu.org> |
61952351 UD |
48 | |
49 | ?? What compiler do I need to build GNU libc? | |
50 | ||
f12944ec UD |
51 | {UD} You must use GNU CC to compile GNU libc. A lot of extensions of GNU CC |
52 | are used to increase portability and speed. | |
61952351 UD |
53 | |
54 | GNU CC is found, like all other GNU packages, on | |
f12944ec | 55 | |
2eb45444 | 56 | ftp://ftp.gnu.org/pub/gnu |
f12944ec | 57 | |
2eb45444 | 58 | and the many mirror sites. ftp.gnu.org is always overloaded, so try to find |
61952351 UD |
59 | a local mirror first. |
60 | ||
ceb27555 | 61 | You should always try to use the latest official release. Older versions |
f12944ec | 62 | may not have all the features GNU libc requires. The current releases of |
a379e56a | 63 | egcs (1.0.3 and 1.1) and GNU CC (2.8.1) should work with the GNU C library |
c882585f | 64 | (for powerpc see ?powerpc; for ARM see ?arm). |
61952351 | 65 | |
d89e7a96 UD |
66 | {ZW} You may have problems if you try to mix code compiled with |
67 | EGCS and with GCC 2.8.1. See ?exception for details. | |
68 | ||
61952351 UD |
69 | ?? When I try to compile glibc I get only error messages. |
70 | What's wrong? | |
71 | ||
f12944ec UD |
72 | {UD} You definitely need GNU make to translate GNU libc. No other make |
73 | program has the needed functionality. | |
61952351 | 74 | |
d89e7a96 UD |
75 | We recommend version GNU make version 3.75 or 3.77. Versions before 3.75 |
76 | have bugs and/or are missing features. Version 3.76 has bugs which | |
77 | appear when building big projects like GNU libc. 3.76.1 appears to work but | |
d8a167a5 | 78 | some people have reported problems. If you build GNU make 3.77 from source, |
c882585f | 79 | please read ?make first. |
61952351 | 80 | |
d89e7a96 | 81 | ?? Do I need a special linker or assembler? |
61952351 | 82 | |
d89e7a96 UD |
83 | {ZW} If you want a shared library, you need a linker and assembler that |
84 | understand all the features of ELF, including weak and versioned symbols. | |
85 | The static library can be compiled with less featureful tools, but lacks key | |
86 | features such as NSS. | |
61952351 | 87 | |
d89e7a96 UD |
88 | For Linux or Hurd, you want binutils 2.8.1.0.23, 2.9.1, or 2.9.1.0.15 or |
89 | higher. These are the only versions we've tested and found reliable. Other | |
90 | versions after 2.8.1.0.23 may work but we don't recommend them, especially | |
91 | not when C++ is involved. Earlier versions do not work at all. | |
7fd18ea2 | 92 | |
d89e7a96 UD |
93 | Other operating systems may come with system tools that have all the |
94 | necessary features, but this is moot because glibc hasn't been ported to | |
95 | them. | |
61952351 | 96 | |
8619129f | 97 | ??powerpc Which compiler should I use for powerpc? |
4775243a | 98 | |
f304af81 | 99 | {GK} You want to use egcs 1.1 or later (together with the right versions |
f12944ec | 100 | of all the other tools, of course). |
4775243a | 101 | |
f304af81 UD |
102 | In fact, egcs 1.1 has a bug that causes linuxthreads to be |
103 | miscompiled, resulting in segmentation faults when using condition | |
104 | variables. There is a temporary patch at: | |
4775243a | 105 | |
f304af81 | 106 | <http://discus.anu.edu.au/~geoffk/egcs-3.diff> |
4775243a | 107 | |
f304af81 | 108 | Later versions of egcs may fix this problem. |
4775243a | 109 | |
cb0509a8 UD |
110 | ??arm Which tools should I use for ARM? |
111 | ||
112 | {PB} You should use egcs 1.1 or a later version. For ELF systems some | |
113 | changes are needed to the compiler; a patch against egcs-1.1.x can be found | |
114 | at: | |
115 | ||
116 | <ftp://ftp.netwinder.org/users/p/philb/egcs-1.1.1pre2-diff-981126> | |
117 | ||
118 | Binutils 2.9.1.0.16 or later is also required. | |
119 | ||
d89e7a96 | 120 | ?? Do I need some more things to compile the GNU C Library? |
61952351 UD |
121 | |
122 | {UD} Yes, there are some more :-). | |
123 | ||
124 | * GNU gettext. This package contains the tools needed to construct | |
125 | `message catalog' files containing translated versions of system | |
2eb45444 | 126 | messages. See ftp://ftp.gnu.org/pub/gnu or better any mirror |
61952351 UD |
127 | site. (We distribute compiled message catalogs, but they may not be |
128 | updated in patches.) | |
129 | ||
d89e7a96 UD |
130 | * Some files are built with special tools. E.g., files ending in .gperf |
131 | need a `gperf' program. The GNU version (now available in a separate | |
132 | package, formerly only as part of libg++) is known to work while some | |
133 | vendor versions do not. | |
61952351 UD |
134 | |
135 | You should not need these tools unless you change the source files. | |
136 | ||
d89e7a96 UD |
137 | * Perl 5 is needed if you wish to test an installation of GNU libc |
138 | as the primary C library. | |
bd355af0 | 139 | |
61952351 UD |
140 | * When compiling for Linux, the header files of the Linux kernel must |
141 | be available to the compiler as <linux/*.h> and <asm/*.h>. | |
142 | ||
8619129f UD |
143 | * lots of disk space (~170MB for i?86-linux; more for RISC platforms, |
144 | as much as 400MB). | |
61952351 UD |
145 | |
146 | * plenty of time. Compiling just the shared and static libraries for | |
147 | i?86-linux takes approximately 1h on an i586@133, or 2.5h on | |
148 | i486@66, or 4.5h on i486@33. Multiply this by 1.5 or 2.0 if you | |
149 | build profiling and/or the highly optimized version as well. For | |
150 | Hurd systems times are much higher. | |
151 | ||
152 | You should avoid compiling in a NFS mounted filesystem. This is | |
153 | very slow. | |
154 | ||
155 | James Troup <J.J.Troup@comp.brad.ac.uk> reports a compile time of | |
bd355af0 UD |
156 | 45h34m for a full build (shared, static, and profiled) on Atari |
157 | Falcon (Motorola 68030 @ 16 Mhz, 14 Mb memory) and Jan Barte | |
158 | <yann@plato.uni-paderborn.de> reports 22h48m on Atari TT030 | |
159 | (Motorola 68030 @ 32 Mhz, 34 Mb memory) | |
61952351 UD |
160 | |
161 | If you have some more measurements let me know. | |
162 | ||
d111572f UD |
163 | ?? What version of the Linux kernel headers should be used? |
164 | ||
f12944ec UD |
165 | {AJ,UD} The headers from the most recent Linux kernel should be used. The |
166 | headers used while compiling the GNU C library and the kernel binary used | |
167 | when using the library do not need to match. The GNU C library runs without | |
168 | problems on kernels that are older than the kernel headers used. The other | |
169 | way round (compiling the GNU C library with old kernel headers and running | |
170 | on a recent kernel) does not necessarily work. For example you can't use | |
171 | new kernel features when using old kernel headers for compiling the GNU C | |
172 | library. | |
173 | ||
ceb27555 | 174 | {ZW} Even if you are using a 2.0 kernel on your machine, we recommend you |
440d13e2 UD |
175 | compile GNU libc with 2.2 kernel headers. That way you won't have to |
176 | recompile libc if you ever upgrade to kernel 2.2. To tell libc which | |
ceb27555 | 177 | headers to use, give configure the --with-headers switch |
440d13e2 | 178 | (e.g. --with-headers=/usr/src/linux-2.2.0/include). |
ceb27555 | 179 | |
440d13e2 | 180 | Note that you must configure the 2.2 kernel if you do this, otherwise libc |
62595351 | 181 | will be unable to find <linux/version.h>. Just change the current directory |
440d13e2 | 182 | to the root of the 2.2 tree and do `make include/linux/version.h'. |
ceb27555 | 183 | |
f12944ec UD |
184 | ?? The compiler hangs while building iconvdata modules. What's |
185 | wrong? | |
186 | ||
d89e7a96 UD |
187 | {ZW} This is a problem with old versions of GCC. Initialization of large |
188 | static arrays is very slow. The compiler will eventually finish; give it | |
189 | time. | |
f12944ec | 190 | |
a379e56a | 191 | The problem is fixed in egcs 1.1 but not in earlier releases. |
d111572f | 192 | |
61952351 UD |
193 | ?? When I run `nm -u libc.so' on the produced library I still |
194 | find unresolved symbols. Can this be ok? | |
195 | ||
f12944ec | 196 | {UD} Yes, this is ok. There can be several kinds of unresolved symbols: |
61952351 UD |
197 | |
198 | * magic symbols automatically generated by the linker. These have names | |
199 | like __start_* and __stop_* | |
200 | ||
201 | * symbols starting with _dl_* come from the dynamic linker | |
202 | ||
61952351 UD |
203 | * weak symbols, which need not be resolved at all (fabs for example) |
204 | ||
205 | Generally, you should make sure you find a real program which produces | |
206 | errors while linking before deciding there is a problem. | |
207 | ||
208 | ??addon What are these `add-ons'? | |
209 | ||
f12944ec UD |
210 | {UD} To avoid complications with export rules or external source code some |
211 | optional parts of the libc are distributed as separate packages (e.g., the | |
212 | crypt package, see ?crypt). | |
61952351 | 213 | |
f12944ec UD |
214 | To use these packages as part of GNU libc, just unpack the tarfiles in the |
215 | libc source directory and tell the configuration script about them using the | |
216 | --enable-add-ons option. If you give just --enable-add-ons configure tries | |
217 | to find all the add-on packages in your source tree. This may not work. If | |
218 | it doesn't, or if you want to select only a subset of the add-ons, give a | |
219 | comma-separated list of the add-ons to enable: | |
61952351 UD |
220 | |
221 | configure --enable-add-ons=crypt,linuxthreads | |
222 | ||
223 | for example. | |
224 | ||
f12944ec UD |
225 | Add-ons can add features (including entirely new shared libraries), override |
226 | files, provide support for additional architectures, and just about anything | |
227 | else. The existing makefiles do most of the work; only some few stub rules | |
228 | must be written to get everything running. | |
61952351 UD |
229 | |
230 | ?? My XXX kernel emulates a floating-point coprocessor for me. | |
231 | Should I enable --with-fp? | |
232 | ||
f12944ec UD |
233 | {ZW} An emulated FPU is just as good as a real one, as far as the C library |
234 | is concerned. You only need to say --without-fp if your machine has no way | |
235 | to execute floating-point instructions. | |
61952351 UD |
236 | |
237 | People who are interested in squeezing the last drop of performance | |
238 | out of their machine may wish to avoid the trap overhead, but this is | |
239 | far more trouble than it's worth: you then have to compile | |
240 | *everything* this way, including the compiler's internal libraries | |
241 | (libgcc.a for GNU C), because the calling conventions change. | |
242 | ||
243 | ?? When compiling GNU libc I get lots of errors saying functions | |
244 | in glibc are duplicated in libgcc. | |
245 | ||
f12944ec UD |
246 | {EY} This is *exactly* the same problem that I was having. The problem was |
247 | due to the fact that configure didn't correctly detect that the linker flag | |
248 | --no-whole-archive was supported in my linker. In my case it was because I | |
249 | had run ./configure with bogus CFLAGS, and the test failed. | |
61952351 | 250 | |
f12944ec UD |
251 | One thing that is particularly annoying about this problem is that once this |
252 | is misdetected, running configure again won't fix it unless you first delete | |
253 | config.cache. | |
61952351 | 254 | |
f12944ec UD |
255 | {UD} Starting with glibc-2.0.3 there should be a better test to avoid some |
256 | problems of this kind. The setting of CFLAGS is checked at the very | |
257 | beginning and if it is not usable `configure' will bark. | |
61952351 | 258 | |
74015205 | 259 | ?? Why do I get messages about missing thread functions when I use |
da2d1bc5 | 260 | librt? I don't even use threads. |
74015205 | 261 | |
da2d1bc5 UD |
262 | {UD} In this case you probably mixed up your installation. librt uses |
263 | threads internally and has implicit references to the thread library. | |
f12944ec UD |
264 | Normally these references are satisfied automatically but if the thread |
265 | library is not in the expected place you must tell the linker where it is. | |
266 | When using GNU ld it works like this: | |
74015205 UD |
267 | |
268 | gcc -o foo foo.c -Wl,-rpath-link=/some/other/dir -lrt | |
269 | ||
f12944ec UD |
270 | The `/some/other/dir' should contain the thread library. `ld' will use the |
271 | given path to find the implicitly referenced library while not disturbing | |
272 | any other link path. | |
74015205 | 273 | |
61952351 UD |
274 | ?? What's the problem with configure --enable-omitfp? |
275 | ||
276 | {AJ} When --enable-omitfp is set the libraries are built without frame | |
fdacb17d | 277 | pointers. Some compilers produce buggy code for this model and therefore we |
f12944ec | 278 | don't advise using it at the moment. |
61952351 | 279 | |
fdacb17d | 280 | If you use --enable-omitfp, you're on your own. If you encounter problems |
f12944ec UD |
281 | with a library that was build this way, we advise you to rebuild the library |
282 | without --enable-omitfp. If the problem vanishes consider tracking the | |
283 | problem down and report it as compiler failure. | |
61952351 | 284 | |
f12944ec UD |
285 | Since a library build with --enable-omitfp is undebuggable on most systems, |
286 | debuggable libraries are also built - you can use it by appending "_g" to | |
287 | the library names. | |
61952351 | 288 | |
f12944ec UD |
289 | The compilation of these extra libraries and the compiler optimizations slow |
290 | down the build process and need more disk space. | |
61952351 | 291 | |
b0610668 UD |
292 | ?? I get failures during `make check'. What shall I do? |
293 | ||
294 | {AJ} The testsuite should compile and run cleanly on your system, every | |
295 | failure should be looked into. Depending on the failure I wouldn't advise | |
296 | installing the library at all. | |
297 | ||
298 | You should consider using the `glibcbug' script to report the failure, | |
299 | providing as much detail as possible. If you run a test directly, please | |
300 | remember to set up the environment correctly. You want to test the compiled | |
301 | library - and not your installed one. The best way is to copy the exact | |
302 | command line which failed and run the test from the subdirectory for this | |
303 | test in the sources. | |
304 | ||
305 | There are some failures which are not directly related to the GNU libc: | |
a379e56a UD |
306 | - Some compiler produce buggy code. The egcs 1.1 release should be ok. gcc |
307 | 2.8.1 might cause some failures, gcc 2.7.2.x is so buggy, that explicit | |
308 | checks have been used so that you can't build with it. | |
b0610668 UD |
309 | - The kernel might have bugs. For example on Linux/Alpha 2.0.34 the |
310 | floating point handling has quite a number of bugs and therefore most of | |
440d13e2 UD |
311 | the test cases in the math subdirectory will fail. Linux 2.2 has |
312 | fixes for the floating point support on Alpha. | |
b0610668 | 313 | |
7fd18ea2 UD |
314 | ?? What is symbol versioning good for? Do I need it? |
315 | ||
316 | {AJ} Symbol versioning solves problems that are related to interface | |
317 | changes. One version of an interface might have been introduced in a | |
318 | previous version of the GNU C library but the interface or the semantics of | |
319 | the function has been changed in the meantime. For binary compatibility | |
320 | with the old library, a newer library needs to still have the old interface | |
321 | for old programs. On the other hand new programs should use the new | |
322 | interface. Symbol versioning is the solution for this problem. The GNU | |
323 | libc version 2.1 uses by default symbol versioning if the binutils support | |
324 | it. | |
325 | ||
326 | We don't advise to build without symbol versioning since you lose binary | |
327 | compatibility if you do - for ever! The binary compatibility you lose is | |
328 | not only against the previous version of the GNU libc (version 2.0) but also | |
329 | against future versions. | |
330 | ||
b0610668 | 331 | |
61952351 UD |
332 | ? Installation and configuration issues |
333 | ||
334 | ?? Can I replace the libc on my Linux system with GNU libc? | |
335 | ||
f12944ec UD |
336 | {UD} You cannot replace any existing libc for Linux with GNU libc. It is |
337 | binary incompatible and therefore has a different major version. You can, | |
338 | however, install it alongside your existing libc. | |
61952351 UD |
339 | |
340 | For Linux there are three major libc versions: | |
341 | libc-4 a.out libc | |
342 | libc-5 original ELF libc | |
343 | libc-6 GNU libc | |
344 | ||
f12944ec UD |
345 | You can have any combination of these three installed. For more information |
346 | consult documentation for shared library handling. The Makefiles of GNU | |
347 | libc will automatically generate the needed symbolic links which the linker | |
348 | will use. | |
61952351 UD |
349 | |
350 | ?? How do I configure GNU libc so that the essential libraries | |
351 | like libc.so go into /lib and the other into /usr/lib? | |
352 | ||
353 | {UD,AJ} Like all other GNU packages GNU libc is designed to use a base | |
354 | directory and install all files relative to this. The default is | |
f12944ec UD |
355 | /usr/local, because this is safe (it will not damage the system if installed |
356 | there). If you wish to install GNU libc as the primary C library on your | |
357 | system, set the base directory to /usr (i.e. run configure --prefix=/usr | |
358 | <other_options>). Note that this can damage your system; see ?safety for | |
359 | details. | |
360 | ||
361 | Some systems like Linux have a filesystem standard which makes a difference | |
362 | between essential libraries and others. Essential libraries are placed in | |
363 | /lib because this directory is required to be located on the same disk | |
364 | partition as /. The /usr subtree might be found on another | |
365 | partition/disk. If you configure for Linux with --prefix=/usr, then this | |
366 | will be done automatically. | |
61952351 UD |
367 | |
368 | To install the essential libraries which come with GNU libc in /lib on | |
f12944ec UD |
369 | systems other than Linux one must explicitly request it. Autoconf has no |
370 | option for this so you have to use a `configparms' file (see the `INSTALL' | |
371 | file for details). It should contain: | |
61952351 UD |
372 | |
373 | slibdir=/lib | |
374 | sysconfdir=/etc | |
375 | ||
f12944ec UD |
376 | The first line specifies the directory for the essential libraries, the |
377 | second line the directory for system configuration files. | |
61952351 UD |
378 | |
379 | ??safety How should I avoid damaging my system when I install GNU libc? | |
380 | ||
f12944ec UD |
381 | {ZW} If you wish to be cautious, do not configure with --prefix=/usr. If |
382 | you don't specify a prefix, glibc will be installed in /usr/local, where it | |
383 | will probably not break anything. (If you wish to be certain, set the | |
384 | prefix to something like /usr/local/glibc2 which is not used for anything.) | |
61952351 UD |
385 | |
386 | The dangers when installing glibc in /usr are twofold: | |
387 | ||
388 | * glibc will overwrite the headers in /usr/include. Other C libraries | |
389 | install a different but overlapping set of headers there, so the | |
390 | effect will probably be that you can't compile anything. You need to | |
391 | rename /usr/include out of the way first. (Do not throw it away; you | |
392 | will then lose the ability to compile programs against your old libc.) | |
393 | ||
394 | * None of your old libraries, static or shared, can be used with a | |
395 | different C library major version. For shared libraries this is not a | |
396 | problem, because the filenames are different and the dynamic linker | |
397 | will enforce the restriction. But static libraries have no version | |
398 | information. You have to evacuate all the static libraries in | |
399 | /usr/lib to a safe location. | |
400 | ||
401 | The situation is rather similar to the move from a.out to ELF which | |
402 | long-time Linux users will remember. | |
403 | ||
404 | ?? Do I need to use GNU CC to compile programs that will use the | |
405 | GNU C Library? | |
406 | ||
f12944ec UD |
407 | {ZW} In theory, no; the linker does not care, and the headers are supposed |
408 | to check for GNU CC before using its extensions to the C language. | |
61952351 | 409 | |
f12944ec UD |
410 | However, there are currently no ports of glibc to systems where another |
411 | compiler is the default, so no one has tested the headers extensively | |
412 | against another compiler. You may therefore encounter difficulties. If you | |
413 | do, please report them as bugs. | |
61952351 UD |
414 | |
415 | Also, in several places GNU extensions provide large benefits in code | |
416 | quality. For example, the library has hand-optimized, inline assembly | |
f12944ec UD |
417 | versions of some string functions. These can only be used with GCC. See |
418 | ?string for details. | |
61952351 UD |
419 | |
420 | ??crypt When linking with the new libc I get unresolved symbols | |
421 | `crypt' and `setkey'. Why aren't these functions in the | |
422 | libc anymore? | |
423 | ||
f12944ec UD |
424 | {UD} The US places restrictions on exporting cryptographic programs and |
425 | source code. Until this law gets abolished we cannot ship the cryptographic | |
426 | functions together with glibc. | |
61952351 | 427 | |
f12944ec UD |
428 | The functions are available, as an add-on (see ?addon). People in the US |
429 | may get it from the same place they got GNU libc from. People outside the | |
430 | US should get the code from ftp://ftp.ifi.uio.no/pub/gnu, or another archive | |
431 | site outside the USA. The README explains how to install the sources. | |
61952351 | 432 | |
f12944ec UD |
433 | If you already have the crypt code on your system the reason for the failure |
434 | is probably that you did not link with -lcrypt. The crypto functions are in | |
435 | a separate library to make it possible to export GNU libc binaries from the | |
436 | US. | |
61952351 UD |
437 | |
438 | ?? When I use GNU libc on my Linux system by linking against | |
439 | the libc.so which comes with glibc all I get is a core dump. | |
440 | ||
f12944ec UD |
441 | {UD} On Linux, gcc sets the dynamic linker to /lib/ld-linux.so.1 unless the |
442 | user specifies a -dynamic-linker argument. This is the name of the libc5 | |
443 | dynamic linker, which does not work with glibc. | |
61952351 | 444 | |
a379e56a UD |
445 | For casual use of GNU libc you can just specify to the linker |
446 | --dynamic-linker=/lib/ld-linux.so.2 | |
61952351 | 447 | |
f12944ec | 448 | which is the glibc dynamic linker, on Linux systems. On other systems the |
a379e56a UD |
449 | name is /lib/ld.so.1. When linking via gcc, you've got to add |
450 | -Wl,--dynamic-linker=/lib/ld-linux.so.2 | |
451 | ||
452 | to the gcc command line. | |
61952351 | 453 | |
f12944ec UD |
454 | To change your environment to use GNU libc for compiling you need to change |
455 | the `specs' file of your gcc. This file is normally found at | |
61952351 UD |
456 | |
457 | /usr/lib/gcc-lib/<arch>/<version>/specs | |
458 | ||
459 | In this file you have to change a few things: | |
460 | ||
461 | - change `ld-linux.so.1' to `ld-linux.so.2' | |
462 | ||
463 | - remove all expression `%{...:-lgmon}'; there is no libgmon in glibc | |
464 | ||
465 | - fix a minor bug by changing %{pipe:-} to %| | |
466 | ||
f12944ec UD |
467 | Here is what the gcc-2.7.2 specs file should look like when GNU libc is |
468 | installed at /usr: | |
61952351 UD |
469 | |
470 | ----------------------------------------------------------------------- | |
471 | *asm: | |
472 | %{V} %{v:%{!V:-V}} %{Qy:} %{!Qn:-Qy} %{n} %{T} %{Ym,*} %{Yd,*} %{Wa,*:%*} | |
473 | ||
474 | *asm_final: | |
475 | %| | |
476 | ||
477 | *cpp: | |
478 | %{fPIC:-D__PIC__ -D__pic__} %{fpic:-D__PIC__ -D__pic__} %{!m386:-D__i486__} %{posix:-D_POSIX_SOURCE} %{pthread:-D_REENTRANT} | |
479 | ||
480 | *cc1: | |
481 | %{profile:-p} | |
482 | ||
483 | *cc1plus: | |
484 | ||
485 | ||
486 | *endfile: | |
487 | %{!shared:crtend.o%s} %{shared:crtendS.o%s} crtn.o%s | |
488 | ||
489 | *link: | |
490 | -m elf_i386 %{shared:-shared} %{!shared: %{!ibcs: %{!static: %{rdynamic:-export-dynamic} %{!dynamic-linker:-dynamic-linker /lib/ld-linux.so.2}} %{static:-static}}} | |
491 | ||
492 | *lib: | |
493 | %{!shared: %{pthread:-lpthread} %{profile:-lc_p} %{!profile: -lc}} | |
494 | ||
495 | *libgcc: | |
496 | -lgcc | |
497 | ||
498 | *startfile: | |
499 | %{!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} | |
500 | ||
501 | *switches_need_spaces: | |
502 | ||
503 | ||
504 | *signed_char: | |
505 | %{funsigned-char:-D__CHAR_UNSIGNED__} | |
506 | ||
507 | *predefines: | |
508 | -D__ELF__ -Dunix -Di386 -Dlinux -Asystem(unix) -Asystem(posix) -Acpu(i386) -Amachine(i386) | |
509 | ||
510 | *cross_compile: | |
511 | 0 | |
512 | ||
513 | *multilib: | |
514 | . ; | |
515 | ||
516 | ----------------------------------------------------------------------- | |
517 | ||
f12944ec UD |
518 | Things get a bit more complicated if you have GNU libc installed in some |
519 | other place than /usr, i.e., if you do not want to use it instead of the old | |
520 | libc. In this case the needed startup files and libraries are not found in | |
521 | the regular places. So the specs file must tell the compiler and linker | |
522 | exactly what to use. | |
61952351 UD |
523 | |
524 | Version 2.7.2.3 does and future versions of GCC will automatically | |
525 | provide the correct specs. | |
526 | ||
527 | ?? Looking through the shared libc file I haven't found the | |
528 | functions `stat', `lstat', `fstat', and `mknod' and while | |
529 | linking on my Linux system I get error messages. How is | |
530 | this supposed to work? | |
531 | ||
f12944ec UD |
532 | {RM} Believe it or not, stat and lstat (and fstat, and mknod) are supposed |
533 | to be undefined references in libc.so.6! Your problem is probably a missing | |
534 | or incorrect /usr/lib/libc.so file; note that this is a small text file now, | |
535 | not a symlink to libc.so.6. It should look something like this: | |
61952351 | 536 | |
71bedb76 | 537 | GROUP ( libc.so.6 libc_nonshared.a ) |
61952351 | 538 | |
d89e7a96 UD |
539 | ??exception When I run an executable on one system which I compiled on |
540 | another, I get dynamic linker errors. Both systems have the same | |
541 | version of glibc installed. What's wrong? | |
542 | ||
543 | {ZW} Glibc on one of these systems was compiled with gcc 2.7 or 2.8, the | |
544 | other with egcs (any version). Egcs has functions in its internal | |
545 | `libgcc.a' to support exception handling with C++. They are linked into | |
546 | any program or dynamic library compiled with egcs, whether it needs them or | |
547 | not. Dynamic libraries then turn around and export those functions again | |
548 | unless special steps are taken to prevent them. | |
549 | ||
550 | When you link your program, it resolves its references to the exception | |
551 | functions to the ones exported accidentally by libc.so. That works fine as | |
552 | long as libc has those functions. On the other system, libc doesn't have | |
553 | those functions because it was compiled by gcc 2.8, and you get undefined | |
554 | symbol errors. The symbols in question are named things like | |
555 | `__register_frame_info'. | |
556 | ||
557 | For glibc 2.0, the workaround is to not compile libc with egcs. We've also | |
558 | incorporated a patch which should prevent the EH functions sneaking into | |
559 | libc. It doesn't matter what compiler you use to compile your program. | |
560 | ||
561 | For glibc 2.1, we've chosen to do it the other way around: libc.so | |
562 | explicitly provides the EH functions. This is to prevent other shared | |
563 | libraries from doing it. You must therefore compile glibc 2.1 with EGCS | |
564 | unless you don't care about ever importing binaries from other systems. | |
565 | Again, it doesn't matter what compiler you use for your programs. | |
566 | ||
61952351 UD |
567 | ?? How can I compile gcc 2.7.2.1 from the gcc source code using |
568 | glibc 2.x? | |
569 | ||
f12944ec UD |
570 | {AJ} There's only correct support for glibc 2.0.x in gcc 2.7.2.3 or later. |
571 | But you should get at least gcc 2.8.1 or egcs 1.0.2 (or later versions) | |
572 | instead. | |
61952351 UD |
573 | |
574 | ?? The `gencat' utility cannot process the catalog sources which | |
575 | were used on my Linux libc5 based system. Why? | |
576 | ||
f12944ec UD |
577 | {UD} The `gencat' utility provided with glibc complies to the XPG standard. |
578 | The older Linux version did not obey the standard, so they are not | |
579 | compatible. | |
61952351 UD |
580 | |
581 | To ease the transition from the Linux version some of the non-standard | |
f12944ec UD |
582 | features are also present in the `gencat' program of GNU libc. This mainly |
583 | includes the use of symbols for the message number and the automatic | |
61952351 UD |
584 | generation of header files which contain the needed #defines to map the |
585 | symbols to integers. | |
586 | ||
f12944ec UD |
587 | Here is a simple SED script to convert at least some Linux specific catalog |
588 | files to the XPG4 form: | |
61952351 UD |
589 | |
590 | ----------------------------------------------------------------------- | |
591 | # Change catalog source in Linux specific format to standard XPG format. | |
592 | # Ulrich Drepper <drepper@cygnus.com>, 1996. | |
593 | # | |
594 | /^\$ #/ { | |
595 | h | |
596 | s/\$ #\([^ ]*\).*/\1/ | |
597 | x | |
598 | s/\$ #[^ ]* *\(.*\)/\$ \1/ | |
599 | } | |
600 | ||
601 | /^# / { | |
602 | s/^# \(.*\)/\1/ | |
603 | G | |
604 | s/\(.*\)\n\(.*\)/\2 \1/ | |
605 | } | |
606 | ----------------------------------------------------------------------- | |
607 | ||
da2d1bc5 UD |
608 | ?? Programs using libc have their messages translated, but other |
609 | behavior is not localized (e.g. collating order); why? | |
610 | ||
611 | {ZW} Translated messages are automatically installed, but the locale | |
f12944ec UD |
612 | database that controls other behaviors is not. You need to run localedef to |
613 | install this database, after you have run `make install'. For example, to | |
614 | set up the French Canadian locale, simply issue the command | |
da2d1bc5 UD |
615 | |
616 | localedef -i fr_CA -f ISO-8859-1 fr_CA | |
617 | ||
618 | Please see localedata/README in the source tree for further details. | |
619 | ||
61952351 UD |
620 | ?? I have set up /etc/nis.conf, and the Linux libc 5 with NYS |
621 | works great. But the glibc NIS+ doesn't seem to work. | |
622 | ||
f12944ec UD |
623 | {TK} The glibc NIS+ implementation uses a /var/nis/NIS_COLD_START file for |
624 | storing information about the NIS+ server and their public keys, because the | |
625 | nis.conf file does not contain all the necessary information. You have to | |
626 | copy a NIS_COLD_START file from a Solaris client (the NIS_COLD_START file is | |
627 | byte order independent) or generate it with nisinit from the nis-tools | |
628 | package; available at | |
629 | ||
630 | http://www-vt.uni-paderborn.de/~kukuk/linux/nisplus.html | |
61952351 | 631 | |
da2d1bc5 | 632 | ?? I have killed ypbind to stop using NIS, but glibc |
3dcf8ea6 | 633 | continues using NIS. |
4d06461a | 634 | |
f12944ec UD |
635 | {TK} For faster NIS lookups, glibc uses the /var/yp/binding/ files from |
636 | ypbind. ypbind 3.3 and older versions don't always remove these files, so | |
637 | glibc will continue to use them. Other BSD versions seem to work correctly. | |
638 | Until ypbind 3.4 is released, you can find a patch at | |
639 | ||
05f732b3 | 640 | ftp://ftp.kernel.org/pub/linux/utils/net/NIS/ypbind-3.3-glibc4.diff.gz |
a788b6c2 | 641 | |
3dcf8ea6 UD |
642 | ?? Under Linux/Alpha, I always get "do_ypcall: clnt_call: |
643 | RPC: Unable to receive; errno = Connection refused" when using NIS. | |
a788b6c2 | 644 | |
f12944ec UD |
645 | {TK} You need a ypbind version which is 64bit clean. Some versions are not |
646 | 64bit clean. A 64bit clean implementation is ypbind-mt. For ypbind 3.3, | |
647 | you need the patch from ftp.kernel.org (See the previous question). I don't | |
648 | know about other versions. | |
a788b6c2 | 649 | |
4d06461a | 650 | |
61952351 UD |
651 | ?? After installing glibc name resolving doesn't work properly. |
652 | ||
f12944ec UD |
653 | {AJ} You probably should read the manual section describing nsswitch.conf |
654 | (just type `info libc "NSS Configuration File"'). The NSS configuration | |
655 | file is usually the culprit. | |
61952351 | 656 | |
3dcf8ea6 UD |
657 | |
658 | ?? How do I create the databases for NSS? | |
659 | ||
660 | {AJ} If you have an entry "db" in /etc/nsswitch.conf you should also create | |
661 | the database files. The glibc sources contain a Makefile which does the | |
7fd18ea2 | 662 | necessary conversion and calls to create those files. The file is |
3dcf8ea6 UD |
663 | `db-Makefile' in the subdirectory `nss' and you can call it with `make -f |
664 | db-Makefile'. Please note that not all services are capable of using a | |
665 | database. Currently passwd, group, ethers, protocol, rpc, services shadow | |
666 | and netgroup are implemented. | |
667 | ||
61952351 UD |
668 | ?? I have /usr/include/net and /usr/include/scsi as symlinks |
669 | into my Linux source tree. Is that wrong? | |
670 | ||
f12944ec UD |
671 | {PB} This was necessary for libc5, but is not correct when using glibc. |
672 | Including the kernel header files directly in user programs usually does not | |
673 | work (see ?kerhdr). glibc provides its own <net/*> and <scsi/*> header | |
674 | files to replace them, and you may have to remove any symlink that you have | |
675 | in place before you install glibc. However, /usr/include/asm and | |
676 | /usr/include/linux should remain as they were. | |
61952351 UD |
677 | |
678 | ?? Programs like `logname', `top', `uptime' `users', `w' and | |
679 | `who', show incorrect information about the (number of) | |
680 | users on my system. Why? | |
681 | ||
682 | {MK} See ?getlog. | |
683 | ||
684 | ?? After upgrading to glibc 2.1 with symbol versioning I get | |
685 | errors about undefined symbols. What went wrong? | |
686 | ||
f12944ec UD |
687 | {AJ} The problem is caused either by wrong program code or tools. In the |
688 | versioned libc a lot of symbols are now local that were global symbols in | |
689 | previous versions. It seems that programs linked against older versions | |
690 | often accidentally used libc global variables -- something that should not | |
691 | happen. | |
61952351 | 692 | |
f12944ec UD |
693 | The only way to fix this is to recompile your program. Sorry, that's the |
694 | price you might have to pay once for quite a number of advantages with | |
695 | symbol versioning. | |
61952351 UD |
696 | |
697 | ?? When I start the program XXX after upgrading the library | |
698 | I get | |
699 | XXX: Symbol `_sys_errlist' has different size in shared | |
700 | object, consider re-linking | |
701 | Why? What should I do? | |
702 | ||
f12944ec UD |
703 | {UD} As the message says, relink the binary. The problem is that a few |
704 | symbols from the library can change in size and there is no way to avoid | |
705 | this. _sys_errlist is a good example. Occasionally there are new error | |
706 | numbers added to the kernel and this must be reflected at user level, | |
707 | breaking programs that refer to them directly. | |
61952351 | 708 | |
f12944ec UD |
709 | Such symbols should normally not be used at all. There are mechanisms to |
710 | avoid using them. In the case of _sys_errlist, there is the strerror() | |
711 | function which should _always_ be used instead. So the correct fix is to | |
712 | rewrite that part of the application. | |
61952351 | 713 | |
f12944ec UD |
714 | In some situations (especially when testing a new library release) it might |
715 | be possible that a symbol changed size when that should not have happened. | |
716 | So in case of doubt report such a warning message as a problem. | |
61952351 | 717 | |
da2d1bc5 UD |
718 | ?? What do I need for C++ development? |
719 | ||
d89e7a96 UD |
720 | {HJ,AJ} You need either egcs 1.1 which comes directly with libstdc++ or |
721 | gcc-2.8.1 together with libstdc++ 2.8.1.1. egcs 1.1 has the better C++ | |
722 | support and works directly with glibc 2.1. If you use gcc-2.8.1 with | |
723 | libstdc++ 2.8.1.1, you need to modify libstdc++ a bit. A patch is available | |
724 | as: | |
725 | ftp://alpha.gnu.org/gnu/libstdc++-2.8.1.1-glibc2.1-diff.gz | |
726 | ||
727 | Please note that libg++ 2.7.2 (and the Linux Versions 2.7.2.x) doesn't work | |
728 | very well with the GNU C library due to vtable thunks. If you're upgrading | |
729 | from glibc 2.0.x to 2.1 you have to recompile libstdc++ since the library | |
730 | compiled for 2.0 is not compatible due to the new Large File Support (LFS) | |
731 | in version 2.1. | |
fb98e2bf UD |
732 | |
733 | {UD} But since in the case of a shared libstdc++ the version numbers should | |
734 | be different existing programs will continue to work. | |
da2d1bc5 | 735 | |
6ca96fe2 UD |
736 | ?? Even statically linked programs need some shared libraries |
737 | which is not acceptable for me. What can I do? | |
738 | ||
f12944ec UD |
739 | {AJ} NSS (for details just type `info libc "Name Service Switch"') won't |
740 | work properly without shared libraries. NSS allows using different services | |
741 | (e.g. NIS, files, db, hesiod) by just changing one configuration file | |
742 | (/etc/nsswitch.conf) without relinking any programs. The only disadvantage | |
743 | is that now static libraries need to access shared libraries. This is | |
744 | handled transparently by the GNU C library. | |
6ca96fe2 | 745 | |
f12944ec UD |
746 | A solution is to configure glibc with --enable-static-nss. In this case you |
747 | can create a static binary that will use only the services dns and files | |
748 | (change /etc/nsswitch.conf for this). You need to link explicitly against | |
749 | all these services. For example: | |
6ca96fe2 UD |
750 | |
751 | gcc -static test-netdb.c -o test-netdb.c \ | |
752 | -lc -lnss_files -lnss_dns -lresolv | |
753 | ||
754 | The problem with this approach is that you've got to link every static | |
755 | program that uses NSS routines with all those libraries. | |
756 | ||
757 | {UD} In fact, one cannot say anymore that a libc compiled with this | |
758 | option is using NSS. There is no switch anymore. Therefore it is | |
759 | *highly* recommended *not* to use --enable-static-nss since this makes | |
760 | the behaviour of the programs on the system inconsistent. | |
761 | ||
bf47fa23 UD |
762 | ?? I just upgraded my Linux system to glibc and now I get |
763 | errors whenever I try to link any program. | |
764 | ||
765 | {ZW} This happens when you have installed glibc as the primary C library but | |
766 | have stray symbolic links pointing at your old C library. If the first | |
767 | `libc.so' the linker finds is libc 5, it will use that. Your program | |
768 | expects to be linked with glibc, so the link fails. | |
769 | ||
770 | The most common case is that glibc put its `libc.so' in /usr/lib, but there | |
771 | was a `libc.so' from libc 5 in /lib, which gets searched first. To fix the | |
772 | problem, just delete /lib/libc.so. You may also need to delete other | |
773 | symbolic links in /lib, such as /lib/libm.so if it points to libm.so.5. | |
774 | ||
0e0316f4 UD |
775 | {AJ} The perl script test-installation.pl which is run as last step during |
776 | an installation of glibc that is configured with --prefix=/usr should help | |
777 | detect these situations. If the script reports problems, something is | |
778 | really screwed up. | |
779 | ||
48244d09 UD |
780 | ?? When I use nscd the machine freezes. |
781 | ||
d89e7a96 UD |
782 | {UD} You cannot use nscd with Linux 2.0.*. There is functionality missing |
783 | in the kernel and work-arounds are not suitable. Besides, some parts of the | |
784 | kernel are too buggy when it comes to using threads. | |
48244d09 | 785 | |
440d13e2 | 786 | If you need nscd, you have to use at least a 2.1 kernel. |
48244d09 UD |
787 | |
788 | Note that I have at this point no information about any other platform. | |
789 | ||
0155a773 UD |
790 | ?? I need lots of open files. What do I have to do? |
791 | ||
792 | {AJ} This is at first a kernel issue. The kernel defines limits with | |
793 | OPEN_MAX the number of simultaneous open files and with FD_SETSIZE the | |
794 | number of used file descriptors. You need to change these values in your | |
795 | kernel and recompile the kernel so that the kernel allows to use more open | |
796 | files. You don't necessarily need to recompile the GNU C library since the | |
797 | only place where OPEN_MAX and FD_SETSIZE is really needed in the library | |
798 | itself is the size of fd_set which is used by select. | |
799 | ||
800 | The GNU C library is now (nearly) select free. This means it internally has | |
801 | no limits imposed by the `fd_set' type. Instead almost all places where the | |
802 | functionality is needed the `poll' function is used. | |
803 | ||
804 | If you increase the number of file descriptors in the kernel you don't need | |
805 | to recompile the C library. The remaining select calls are in the RPC code. | |
806 | If your RPC daemons don't need more than FD_SETSIZE file descriptors, you | |
807 | don't need to change anything at all. | |
808 | ||
809 | {UD} You can always get the maximum number of file descriptors a process is | |
810 | allowed to have open at any time using | |
811 | ||
812 | number = sysconf (_SC_OPEN_MAX); | |
813 | ||
814 | This will work even if the kernel limits change. | |
815 | ||
d8a167a5 UD |
816 | ?? How do I get the same behavior on parsing /etc/passwd and |
817 | /etc/group as I have with libc5 ? | |
818 | ||
819 | {TK} The name switch setup in /etc/nsswitch.conf selected by most Linux | |
820 | distributions does not support +/- and netgroup entries in the files like | |
821 | /etc/passwd. Though this is the preferred setup some people might have | |
822 | setups coming over from the libc5 days where it was the default to recognize | |
823 | lines like this. To get back to the old behaviour one simply has to change | |
824 | the rules for passwd, group, and shadow in the nsswitch.conf file as | |
825 | follows: | |
826 | ||
827 | passwd: compat | |
828 | group: compat | |
829 | shadow: compat | |
830 | ||
831 | passwd_compat: nis | |
832 | group_compat: nis | |
833 | shadow_compat: nis | |
834 | ||
0f6052a8 UD |
835 | ?? What needs to be recompiled when upgrading from glibc 2.0 to glibc |
836 | 2.1? | |
837 | ||
838 | {AJ,CG} If you just upgrade the glibc from 2.0.x (x <= 7) to 2.1, binaries | |
839 | that have been linked against glibc 2.0 will continue to work. | |
840 | ||
841 | If you compile your own binaries against glibc 2.1, you also need to | |
842 | recompile some other libraries. The problem is that libio had to be | |
843 | changed and therefore libraries that are based or depend on the libio | |
844 | of glibc, e.g. ncurses or slang, need to be recompiled. If you | |
845 | experience strange segmentation faults in your programs linked against | |
846 | glibc 2.1, you might need to recompile your libraries. | |
847 | ||
848 | Another problem is that older binaries that were linked statically against | |
849 | glibc 2.0 will reference the older nss modules (libnss_files.so.1 instead of | |
850 | libnss_files.so.2), so don't remove them. Also, the old glibc-2.0 compiled | |
851 | static libraries (libfoo.a) which happen to depend on the older libio | |
50b65db1 UD |
852 | behavior will be broken by the glibc 2.1 upgrade. We plan to produce a |
853 | compatibility library that people will be able to link in if they want | |
854 | to compile a static library generated against glibc 2.0 into a program | |
855 | on a glibc 2.1 system. You just add -lcompat and you should be fine. | |
856 | ||
857 | The glibc-compat add-on will provide the libcompat.a library, the older | |
858 | nss modules, and a few other files. Together, they should make it | |
859 | possible to do development with old static libraries on a glibc 2.1 | |
860 | system. This add-on is still in development. You can get it from <URL> | |
861 | but please keep in mind that it is experimental. | |
0155a773 | 862 | |
61952351 UD |
863 | ? Source and binary incompatibilities, and what to do about them |
864 | ||
865 | ?? I expect GNU libc to be 100% source code compatible with | |
866 | the old Linux based GNU libc. Why isn't it like this? | |
867 | ||
f12944ec UD |
868 | {DMT,UD} Not every extension in Linux libc's history was well thought-out. |
869 | In fact it had a lot of problems with standards compliance and with | |
870 | cleanliness. With the introduction of a new version number these errors can | |
871 | now be corrected. Here is a list of the known source code | |
61952351 UD |
872 | incompatibilities: |
873 | ||
874 | * _GNU_SOURCE: glibc does not make the GNU extensions available | |
875 | automatically. If a program depends on GNU extensions or some | |
876 | other non-standard functionality, it is necessary to compile it | |
877 | with the C compiler option -D_GNU_SOURCE, or better, to put | |
878 | `#define _GNU_SOURCE' at the beginning of your source files, before | |
879 | any C library header files are included. This difference normally | |
880 | manifests itself in the form of missing prototypes and/or data type | |
881 | definitions. Thus, if you get such errors, the first thing you | |
882 | should do is try defining _GNU_SOURCE and see if that makes the | |
883 | problem go away. | |
884 | ||
885 | For more information consult the file `NOTES' in the GNU C library | |
886 | sources. | |
887 | ||
888 | * reboot(): GNU libc sanitizes the interface of reboot() to be more | |
889 | compatible with the interface used on other OSes. reboot() as | |
890 | implemented in glibc takes just one argument. This argument | |
891 | corresponds to the third argument of the Linux reboot system call. | |
892 | That is, a call of the form reboot(a, b, c) needs to be changed into | |
893 | reboot(c). Beside this the header <sys/reboot.h> defines the needed | |
894 | constants for the argument. These RB_* constants should be used | |
895 | instead of the cryptic magic numbers. | |
896 | ||
897 | * swapon(): the interface of this function didn't change, but the | |
898 | prototype is in a separate header file <sys/swap.h>. This header | |
899 | file also provides the SWAP_* constants defined by <linux/swap.h>; | |
900 | you should use them for the second argument to swapon(). | |
901 | ||
902 | * errno: If a program uses the variable "errno", then it _must_ | |
903 | include <errno.h>. The old libc often (erroneously) declared this | |
904 | variable implicitly as a side-effect of including other libc header | |
905 | files. glibc is careful to avoid such namespace pollution, which, | |
906 | in turn, means that you really need to include the header files that | |
907 | you depend on. This difference normally manifests itself in the | |
908 | form of the compiler complaining about references to an undeclared | |
909 | symbol "errno". | |
910 | ||
911 | * Linux-specific syscalls: All Linux system calls now have appropriate | |
912 | library wrappers and corresponding declarations in various header files. | |
913 | This is because the syscall() macro that was traditionally used to | |
914 | work around missing syscall wrappers are inherently non-portable and | |
915 | error-prone. The following table lists all the new syscall stubs, | |
916 | the header-file declaring their interface and the system call name. | |
917 | ||
918 | syscall name: wrapper name: declaring header file: | |
919 | ------------- ------------- ---------------------- | |
920 | bdflush bdflush <sys/kdaemon.h> | |
921 | syslog ksyslog_ctl <sys/klog.h> | |
922 | ||
923 | * lpd: Older versions of lpd depend on a routine called _validuser(). | |
924 | The library does not provide this function, but instead provides | |
925 | __ivaliduser() which has a slightly different interface. Simply | |
926 | upgrading to a newer lpd should fix this problem (e.g., the 4.4BSD | |
927 | lpd is known to be working). | |
928 | ||
929 | * resolver functions/BIND: like on many other systems the functions of | |
930 | the resolver library are not included in libc itself. There is a | |
931 | separate library libresolv. If you get undefined symbol errors for | |
932 | symbols starting with `res_*' simply add -lresolv to your linker | |
933 | command line. | |
934 | ||
935 | * the `signal' function's behavior corresponds to the BSD semantic and | |
936 | not the SysV semantic as it was in libc-5. The interface on all GNU | |
937 | systems shall be the same and BSD is the semantic of choice. To use | |
938 | the SysV behavior simply use `sysv_signal', or define _XOPEN_SOURCE. | |
939 | See ?signal for details. | |
940 | ||
941 | ??getlog Why does getlogin() always return NULL on my Linux box? | |
942 | ||
f12944ec UD |
943 | {UD} The GNU C library has a format for the UTMP and WTMP file which differs |
944 | from what your system currently has. It was extended to fulfill the needs | |
945 | of the next years when IPv6 is introduced. The record size is different and | |
946 | some fields have different positions. The files written by functions from | |
947 | the one library cannot be read by functions from the other library. Sorry, | |
948 | but this is what a major release is for. It's better to have a cut now than | |
949 | having no means to support the new techniques later. | |
61952351 | 950 | |
f12944ec UD |
951 | {MK} There is however a (partial) solution for this problem. Please take a |
952 | look at the file `login/README.utmpd'. | |
61952351 UD |
953 | |
954 | ?? Where are the DST_* constants found in <sys/time.h> on many | |
955 | systems? | |
956 | ||
f12944ec UD |
957 | {UD} These constants come from the old BSD days and are not used anymore |
958 | (libc5 does not actually implement the handling although the constants are | |
959 | defined). | |
61952351 | 960 | |
f12944ec | 961 | Instead GNU libc contains zone database support and compatibility code for |
8b4a4715 UD |
962 | POSIX TZ environment variable handling. For former is very much preferred |
963 | (see ?tzdb). | |
61952351 UD |
964 | |
965 | ?? The prototypes for `connect', `accept', `getsockopt', | |
966 | `setsockopt', `getsockname', `getpeername', `send', | |
967 | `sendto', and `recvfrom' are different in GNU libc from | |
968 | any other system I saw. This is a bug, isn't it? | |
969 | ||
f12944ec UD |
970 | {UD} No, this is no bug. This version of GNU libc already follows the new |
971 | Single Unix specifications (and I think the POSIX.1g draft which adopted the | |
972 | solution). The type for a parameter describing a size is now `socklen_t', a | |
973 | new type. | |
61952351 UD |
974 | |
975 | ??kerhdr On Linux I've got problems with the declarations in Linux | |
976 | kernel headers. | |
977 | ||
f12944ec UD |
978 | {UD,AJ} On Linux, the use of kernel headers is reduced to the minimum. This |
979 | gives Linus the ability to change the headers more freely. Also, user | |
a9ddb793 | 980 | programs are now insulated from changes in the size of kernel data |
f12944ec | 981 | structures. |
61952351 | 982 | |
f12944ec UD |
983 | For example, the sigset_t type is 32 or 64 bits wide in the kernel. In |
984 | glibc it is 1024 bits wide. This guarantees that when the kernel gets a | |
985 | bigger sigset_t (for POSIX.1e realtime support, say) user programs will not | |
986 | have to be recompiled. Consult the header files for more information about | |
987 | the changes. | |
61952351 | 988 | |
f12944ec UD |
989 | Therefore you shouldn't include Linux kernel header files directly if glibc |
990 | has defined a replacement. Otherwise you might get undefined results because | |
991 | of type conflicts. | |
61952351 UD |
992 | |
993 | ?? I don't include any kernel headers myself but the compiler | |
994 | still complains about redeclarations of types in the kernel | |
995 | headers. | |
996 | ||
f12944ec UD |
997 | {UD} The kernel headers before Linux 2.1.61 and 2.0.32 don't work correctly |
998 | with glibc. Compiling C programs is possible in most cases but C++ programs | |
999 | have (due to the change of the name lookups for `struct's) problems. One | |
1000 | prominent example is `struct fd_set'. | |
61952351 | 1001 | |
f12944ec UD |
1002 | There might be some problems left but 2.1.61/2.0.32 fix most of the known |
1003 | ones. See the BUGS file for other known problems. | |
61952351 UD |
1004 | |
1005 | ??signal Why don't signals interrupt system calls anymore? | |
1006 | ||
f12944ec UD |
1007 | {ZW} By default GNU libc uses the BSD semantics for signal(), unlike Linux |
1008 | libc 5 which used System V semantics. This is partially for compatibility | |
1009 | with other systems and partially because the BSD semantics tend to make | |
1010 | programming with signals easier. | |
61952351 UD |
1011 | |
1012 | There are three differences: | |
1013 | ||
1014 | * BSD-style signals that occur in the middle of a system call do not | |
1015 | affect the system call; System V signals cause the system call to | |
1016 | fail and set errno to EINTR. | |
1017 | ||
1018 | * BSD signal handlers remain installed once triggered. System V signal | |
1019 | handlers work only once, so one must reinstall them each time. | |
1020 | ||
1021 | * A BSD signal is blocked during the execution of its handler. In other | |
1022 | words, a handler for SIGCHLD (for example) does not need to worry about | |
1023 | being interrupted by another SIGCHLD. It may, however, be interrupted | |
1024 | by other signals. | |
1025 | ||
1026 | There is general consensus that for `casual' programming with signals, the | |
1027 | BSD semantics are preferable. You don't need to worry about system calls | |
1028 | returning EINTR, and you don't need to worry about the race conditions | |
1029 | associated with one-shot signal handlers. | |
1030 | ||
1031 | If you are porting an old program that relies on the old semantics, you can | |
1032 | quickly fix the problem by changing signal() to sysv_signal() throughout. | |
1033 | Alternatively, define _XOPEN_SOURCE before including <signal.h>. | |
1034 | ||
1035 | For new programs, the sigaction() function allows you to specify precisely | |
1036 | how you want your signals to behave. All three differences listed above are | |
1037 | individually switchable on a per-signal basis with this function. | |
1038 | ||
f12944ec UD |
1039 | If all you want is for one specific signal to cause system calls to fail and |
1040 | return EINTR (for example, to implement a timeout) you can do this with | |
61952351 UD |
1041 | siginterrupt(). |
1042 | ||
1043 | ||
1044 | ??string I've got errors compiling code that uses certain string | |
1045 | functions. Why? | |
1046 | ||
f12944ec | 1047 | {AJ} glibc 2.1 has special string functions that are faster than the normal |
fdacb17d | 1048 | library functions. Some of the functions are additionally implemented as |
3dcf8ea6 | 1049 | inline functions and others as macros. |
61952351 UD |
1050 | |
1051 | The optimized string functions are only used when compiling with | |
fdacb17d | 1052 | optimizations (-O1 or higher). The behavior can be changed with two feature |
f12944ec | 1053 | macros: |
61952351 UD |
1054 | |
1055 | * __NO_STRING_INLINES: Don't do any string optimizations. | |
1056 | * __USE_STRING_INLINES: Use assembly language inline functions (might | |
1057 | increase code size dramatically). | |
1058 | ||
f12944ec UD |
1059 | Since some of these string functions are now additionally defined as macros, |
1060 | code like "char *strncpy();" doesn't work anymore (and is unnecessary, since | |
fdacb17d | 1061 | <string.h> has the necessary declarations). Either change your code or |
f12944ec | 1062 | define __NO_STRING_INLINES. |
61952351 | 1063 | |
f12944ec UD |
1064 | {UD} Another problem in this area is that gcc still has problems on machines |
1065 | with very few registers (e.g., ix86). The inline assembler code can require | |
1066 | almost all the registers and the register allocator cannot always handle | |
1067 | this situation. | |
61952351 UD |
1068 | |
1069 | One can disable the string optimizations selectively. Instead of writing | |
1070 | ||
1071 | cp = strcpy (foo, "lkj"); | |
1072 | ||
1073 | one can write | |
1074 | ||
1075 | cp = (strcpy) (foo, "lkj"); | |
1076 | ||
1077 | This disables the optimization for that specific call. | |
1078 | ||
4775243a UD |
1079 | ?? I get compiler messages "Initializer element not constant" with |
1080 | stdin/stdout/stderr. Why? | |
1081 | ||
1082 | {RM,AJ} Constructs like: | |
1083 | static FILE *InPtr = stdin; | |
1084 | ||
fdacb17d UD |
1085 | lead to this message. This is correct behaviour with glibc since stdin is |
1086 | not a constant expression. Please note that a strict reading of ISO C does | |
f12944ec | 1087 | not allow above constructs. |
4775243a | 1088 | |
f12944ec UD |
1089 | One of the advantages of this is that you can assign to stdin, stdout, and |
1090 | stderr just like any other global variable (e.g. `stdout = my_stream;'), | |
1091 | which can be very useful with custom streams that you can write with libio | |
fdacb17d | 1092 | (but beware this is not necessarily portable). The reason to implement it |
f12944ec | 1093 | this way were versioning problems with the size of the FILE structure. |
4775243a | 1094 | |
fdacb17d UD |
1095 | To fix those programs you've got to initialize the variable at run time. |
1096 | This can be done, e.g. in main, like: | |
1097 | ||
1098 | static FILE *InPtr; | |
bfcd44c3 | 1099 | int main(void) |
fdacb17d UD |
1100 | { |
1101 | InPtr = stdin; | |
1102 | } | |
1103 | ||
1104 | or by constructors (beware this is gcc specific): | |
1105 | ||
1106 | static FILE *InPtr; | |
1107 | static void inPtr_construct (void) __attribute__((constructor)); | |
1108 | static void inPtr_construct (void) { InPtr = stdin; } | |
1109 | ||
4775243a UD |
1110 | |
1111 | ?? I can't compile with gcc -traditional (or | |
1112 | -traditional-cpp). Why? | |
1113 | ||
1114 | {AJ} glibc2 does break -traditional and -traditonal-cpp - and will continue | |
fdacb17d | 1115 | to do so. For example constructs of the form: |
f12944ec | 1116 | |
4775243a UD |
1117 | enum {foo |
1118 | #define foo foo | |
1119 | } | |
f12944ec UD |
1120 | |
1121 | are useful for debugging purposes (you can use foo with your debugger that's | |
1122 | why we need the enum) and for compatibility (other systems use defines and | |
1123 | check with #ifdef). | |
4775243a UD |
1124 | |
1125 | ?? I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible? | |
1126 | ||
1127 | {AJ} The GNU C library is compatible with the ANSI/ISO C standard. If | |
f12944ec | 1128 | you're using `gcc -ansi', the glibc includes which are specified in the |
fdacb17d | 1129 | standard follow the standard. The ANSI/ISO C standard defines what has to be |
f12944ec UD |
1130 | in the include files - and also states that nothing else should be in the |
1131 | include files (btw. you can still enable additional standards with feature | |
1132 | flags). | |
4775243a | 1133 | |
f12944ec UD |
1134 | The GNU C library is conforming to ANSI/ISO C - if and only if you're only |
1135 | using the headers and library functions defined in the standard. | |
4775243a | 1136 | |
4d42000c UD |
1137 | ?? I can't access some functions anymore. nm shows that they do |
1138 | exist but linking fails nevertheless. | |
1139 | ||
f12944ec UD |
1140 | {AJ} With the introduction of versioning in glibc 2.1 it is possible to |
1141 | export only those identifiers (functions, variables) that are really needed | |
1142 | by application programs and by other parts of glibc. This way a lot of | |
1143 | internal interfaces are now hidden. nm will still show those identifiers | |
1144 | but marking them as internal. ISO C states that identifiers beginning with | |
1145 | an underscore are internal to the libc. An application program normally | |
1146 | shouldn't use those internal interfaces (there are exceptions, | |
1147 | e.g. __ivaliduser). If a program uses these interfaces, it's broken. These | |
1148 | internal interfaces might change between glibc releases or dropped | |
1149 | completely. | |
4d42000c | 1150 | |
a5f4e34a UD |
1151 | ?? When using the db-2 library which comes with glibc is used in |
1152 | the Perl db modules the testsuite is not passed. This did not | |
1153 | happen with db-1, gdbm, or ndbm. | |
1154 | ||
1155 | {UD} You are using an outdated copy of the DB_File Perl module. In fact db-2 | |
1156 | finally removed the handling of zero-sized keys which was one of the features | |
1157 | tested by the old Perl testsuite and therefore you see an error. But this | |
1158 | never was documented and guaranteed, only broken programs used this feature. | |
1159 | ||
1160 | Consequently db-2 does not need to support this feature and instead signals | |
1161 | an error which leads to easier debugging. The DB_File module maintainer | |
1162 | Paul Marquess <pmarquess@bfsec.bt.co.uk> acknowledged this change and fixed | |
1163 | the testsuite so that if you use DB_File v1.60 or later you should not have | |
1164 | any more problems with db-2. | |
1165 | ||
5148d49f UD |
1166 | ?? The pow() inline function I get when including <math.h> is broken. |
1167 | I get segmentation faults when I run the program. | |
1168 | ||
1169 | {UD} Nope, the implementation is correct. The problem is with egcs version | |
1170 | prior to 1.1. I.e., egcs 1.0 to 1.0.3 are all broken (at least on Intel). | |
1171 | If you have to use this compiler you must define __NO_MATH_INLINES before | |
1172 | including <math.h> to prevent the inline functions from being used. egcs 1.1 | |
1173 | fixes the problem. I don't know about gcc 2.8 and 2.8.1. | |
1174 | ||
05f732b3 UD |
1175 | ?? The sys/sem.h file lacks the definition of `union semun'. |
1176 | ||
1177 | {UD} Nope. This union has to be provided by the user program. Former glibc | |
1178 | versions defined this but it was an error since it does not make much sense | |
1179 | when thinking about it. The standards describing the System V IPC functions | |
1180 | define it this way and therefore programs must be adopted. | |
1181 | ||
a42134a7 UD |
1182 | ?? Why has <netinet/ip_fw.h> disappeared? |
1183 | ||
1184 | {AJ} The corresponding Linux kernel data structures and constants are | |
440d13e2 | 1185 | totally different in Linux 2.0 and Linux 2.2. This situation has to be |
a42134a7 UD |
1186 | taken care in user programs using the firewall structures and therefore |
1187 | those programs (ipfw is AFAIK the only one) should deal with this problem | |
1188 | themselves. | |
1189 | ||
ee586e0e UD |
1190 | ?? I get floods of warnings when I use -Wconversion and include |
1191 | <string.h> or <math.h>. | |
1192 | ||
1193 | {ZW} <string.h> and <math.h> intentionally use prototypes to override | |
1194 | argument promotion. -Wconversion warns about all these. You can safely | |
1195 | ignore the warnings. | |
1196 | ||
1197 | -Wconversion isn't really intended for production use, only for shakedown | |
1198 | compiles after converting an old program to standard C. | |
1199 | ||
4d42000c | 1200 | |
61952351 UD |
1201 | ? Miscellaneous |
1202 | ||
1203 | ?? After I changed configure.in I get `Autoconf version X.Y. | |
1204 | or higher is required for this script'. What can I do? | |
1205 | ||
1206 | {UD} You have to get the specified autoconf version (or a later one) | |
2eb45444 | 1207 | from your favorite mirror of ftp.gnu.org. |
61952351 UD |
1208 | |
1209 | ?? When I try to compile code which uses IPv6 headers and | |
1210 | definitions on my Linux 2.x.y system I am in trouble. | |
1211 | Nothing seems to work. | |
1212 | ||
f12944ec UD |
1213 | {UD} The problem is that IPv6 development still has not reached a point |
1214 | where the headers are stable. There are still lots of incompatible changes | |
1215 | made and the libc headers have to follow. | |
61952351 | 1216 | |
cb0509a8 UD |
1217 | {PB} The 2.1 release of GNU libc aims to comply with the current versions of |
1218 | all the relevant standards. The IPv6 support libraries for older Linux | |
1219 | systems used a different naming convention and so code written to work with | |
1220 | them may need to be modified. If the standards make incompatible changes in | |
1221 | the future then the libc may need to change again. | |
1222 | ||
1223 | IPv6 will not work with a 2.0.x kernel. When kernel 2.2 is released it | |
1224 | should contain all the necessary support; until then you should use the | |
3f7b3d9b | 1225 | latest 2.1.x release you can find. As of 98/11/26 the currently recommended |
cb0509a8 UD |
1226 | kernel for IPv6 is 2.1.129. |
1227 | ||
1228 | Also, as of the 2.1 release the IPv6 API provided by GNU libc is not | |
1229 | 100% complete. In particular the getipnodebyname and getipnodebyaddr | |
1230 | functions are not implemented. | |
61952351 | 1231 | |
8b4a4715 | 1232 | ??tzdb When I set the timezone by setting the TZ environment variable |
73237de3 UD |
1233 | to EST5EDT things go wrong since glibc computes the wrong time |
1234 | from this information. | |
1235 | ||
f12944ec UD |
1236 | {UD} The problem is that people still use the braindamaged POSIX method to |
1237 | select the timezone using the TZ environment variable with a format EST5EDT | |
8b4a4715 UD |
1238 | or whatever. People, if you insist on using TZ instead of the timezone |
1239 | database (see below), read the POSIX standard, the implemented behaviour is | |
f12944ec UD |
1240 | correct! What you see is in fact the result of the decisions made while |
1241 | POSIX.1 was created. We've only implemented the handling of TZ this way to | |
1242 | be POSIX compliant. It is not really meant to be used. | |
1243 | ||
1244 | The alternative approach to handle timezones which is implemented is the | |
1245 | correct one to use: use the timezone database. This avoids all the problems | |
1246 | the POSIX method has plus it is much easier to use. Simply run the tzselect | |
1247 | shell script, answer the question and use the name printed in the end by | |
8b4a4715 UD |
1248 | making a symlink /etc/localtime pointing to /usr/share/zoneinfo/NAME (NAME |
1249 | is the returned value from tzselect). That's all. You never again have to | |
1250 | worry. | |
f12944ec UD |
1251 | |
1252 | So, please avoid sending bug reports about time related problems if you use | |
1253 | the POSIX method and you have not verified something is really broken by | |
1254 | reading the POSIX standards. | |
73237de3 | 1255 | |
fdacb17d UD |
1256 | ?? What other sources of documentation about glibc are available? |
1257 | ||
1258 | {AJ} The FSF has a page about the GNU C library at | |
1259 | <http://www.gnu.org/software/libc/>. The problem data base of open and | |
1260 | solved bugs in GNU libc is available at | |
1261 | <http://www-gnats.gnu.org:8080/cgi-bin/wwwgnats.pl>. Eric Green has written | |
1262 | a HowTo for converting from Linux libc5 to glibc2. The HowTo is accessable | |
1263 | via the FSF page and at <http://www.imaxx.net/~thrytis/glibc>. Frodo | |
1264 | Looijaard describes a different way installing glibc2 as secondary libc at | |
1265 | <http://huizen.dds.nl/~frodol/glibc>. | |
1266 | ||
1267 | Please note that this is not a complete list. | |
1268 | ||
3f7b3d9b UD |
1269 | ?? The timezone string for Sydney/Australia is wrong since even when |
1270 | daylight saving time is in effect the timezone string is EST. | |
1271 | ||
1272 | {UD} The problem for some timezones is that the local authorities decided | |
1273 | to use the term "summer time" instead of "daylight saving time". In this | |
1274 | case the abbreviation character `S' is the same as the standard one. So, | |
1275 | for Sydney we have | |
1276 | ||
1277 | Eastern Standard Time = EST | |
1278 | Eastern Summer Time = EST | |
1279 | ||
1280 | Great! To get this bug fixed convince the authorities to change the laws | |
1281 | and regulations of the country this effects. glibc behaves correctly. | |
1282 | ||
eeabe877 UD |
1283 | ??make I've build make 3.77 against glibc 2.1 and now make gets |
1284 | segmentation faults. | |
1285 | ||
1286 | {AJ} GNU make 3.77 has support for 64 bit filesystems which is slightly | |
1287 | broken (and one of the new features in the GNU C library 2.1 is 64 bit | |
1288 | filesystem support :-( ). To get a working make you can use either make | |
1289 | 3.75 or patch 3.77. A working patch is available via RedHat's Rawhide server | |
1290 | (ftp://rawhide.redhat.com/SRPMS/SRPMS/make-3.77-*src.rpm). | |
1291 | ||
61952351 UD |
1292 | \f |
1293 | Answers were given by: | |
1294 | {UD} Ulrich Drepper, <drepper@cygnus.com> | |
1295 | {DMT} David Mosberger-Tang, <davidm@AZStarNet.com> | |
1296 | {RM} Roland McGrath, <roland@gnu.org> | |
1297 | {AJ} Andreas Jaeger, <aj@arthur.rhein-neckar.de> | |
1298 | {EY} Eric Youngdale, <eric@andante.jic.com> | |
1299 | {PB} Phil Blundell, <Philip.Blundell@pobox.com> | |
1300 | {MK} Mark Kettenis, <kettenis@phys.uva.nl> | |
1301 | {ZW} Zack Weinberg, <zack@rabi.phys.columbia.edu> | |
d71b808a | 1302 | {TK} Thorsten Kukuk, <kukuk@vt.uni-paderborn.de> |
8619129f | 1303 | {GK} Geoffrey Keating, <geoffk@ozemail.com.au> |
da2d1bc5 | 1304 | {HJ} H.J. Lu, <hjl@gnu.org> |
0f6052a8 | 1305 | {CG} Cristian Gafton, <gafton@redhat.com> |
61952351 UD |
1306 | \f |
1307 | Local Variables: | |
1308 | mode:outline | |
1309 | outline-regexp:"\\?" | |
f12944ec | 1310 | fill-column:76 |
61952351 | 1311 | End: |