1 Frequently Asked Questions about the GNU C Library
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.
7 The GNU C library is very complex. The installation process has not been
8 completely automated; there are too many variables. You can do substantial
9 damage to your system by installing the library incorrectly. Make sure you
10 understand what you are undertaking before you begin.
12 If you have any questions you think should be answered in this document,
19 ?? What systems does the GNU C Library run on?
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.
25 The systems glibc is known to work on as of this release, and most probably
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
32 powerpc-*-linux-gnu Linux and MkLinux on PowerPC systems
33 sparc-*-linux-gnu Linux-2.x on SPARC
34 sparc64-*-linux-gnu Linux-2.x on UltraSPARC
35 arm-*-none ARM standalone systems
36 arm-*-linuxaout Linux-2.x on ARM using a.out binaries
38 Ports to other Linux platforms are in development, and may in fact work
39 already, but no one has sent us success reports for them. Currently no
40 ports to other operating systems are underway, although a few people have
43 If you have a system not listed above (or in the `README' file) and you are
44 really interested in porting it, contact
48 ?? What compiler do I need to build GNU libc?
50 {UD} You must use GNU CC to compile GNU libc. A lot of extensions of GNU CC
51 are used to increase portability and speed.
53 GNU CC is found, like all other GNU packages, on
55 ftp://ftp.gnu.org/pub/gnu
57 and the many mirror sites. ftp.gnu.org is always overloaded, so try to find
60 You should always try to use the latest official release. Older versions
61 may not have all the features GNU libc requires. The current releases of
62 egcs (1.0.3 and 1.1) and GNU CC (2.8.1) should work with the GNU C library
63 (for powerpc see question ?powerpc).
65 ?? When I try to compile glibc I get only error messages.
68 {UD} You definitely need GNU make to translate GNU libc. No other make
69 program has the needed functionality.
71 We recommend version GNU make version 3.75. Versions 3.76 and 3.76.1 have
72 bugs which appear when building big projects like GNU libc. Versions before
73 3.74 have bugs and/or are missing features.
75 ?? Do I need a special linker or archiver?
77 {UD} You may be able to use your system linker, but GNU libc works best with
80 On systems where the native linker does not support weak symbols you will
81 not get a fully ISO C compliant C library. Generally speaking you should
82 use the GNU binutils if they provide at least the same functionality as your
85 Always get the newest release of GNU binutils available. Older releases are
86 known to have bugs that prevent a successful compilation.
88 {AJ} Please don't use binutils 2.7. That release contains some bugs which
89 might make it necessary that you've got to recompile all your glibc2
90 binaries when upgrading the GNU C library.
92 {ZW} As of release 2.1 a linker supporting symbol versions is required. For
93 Linux, get binutils-2.8.1.0.23 or later. Other systems may have native
94 linker support, but it's moot right now, because glibc has not been ported
97 ??powerpc Which compiler should I use for powerpc?
99 {GK} You want to use egcs 1.0.1 or later (together with the right versions
100 of all the other tools, of course).
102 In fact, egcs 1.0.1 has a serious bug that prevents a clean make, relating
103 to switch statement folding. It also causes the resulting shared libraries
104 to use more memory than they should. There is a patch at:
106 <http://discus.anu.edu.au/~geoffk/egcs-1.0.1-geoffk.diff>
108 Later versions of egcs may fix these problems.
111 ?? Do I need some more things to compile GNU C Library?
113 {UD} Yes, there are some more :-).
115 * GNU gettext. This package contains the tools needed to construct
116 `message catalog' files containing translated versions of system
117 messages. See ftp://ftp.gnu.org/pub/gnu or better any mirror
118 site. (We distribute compiled message catalogs, but they may not be
121 * Some files depend on special tools. E.g., files ending in .gperf
122 need a `gperf' program. The GNU version (part of libg++) is known
123 to work while some vendor versions do not.
125 You should not need these tools unless you change the source files.
127 * Some scripts need perl5 - but at the moment those scripts are not
128 vital for building and installing GNU libc (some data files will not
131 * When compiling for Linux, the header files of the Linux kernel must
132 be available to the compiler as <linux/*.h> and <asm/*.h>.
134 * lots of disk space (~170MB for i?86-linux; more for RISC platforms,
137 * plenty of time. Compiling just the shared and static libraries for
138 i?86-linux takes approximately 1h on an i586@133, or 2.5h on
139 i486@66, or 4.5h on i486@33. Multiply this by 1.5 or 2.0 if you
140 build profiling and/or the highly optimized version as well. For
141 Hurd systems times are much higher.
143 You should avoid compiling in a NFS mounted filesystem. This is
146 James Troup <J.J.Troup@comp.brad.ac.uk> reports a compile time of
147 45h34m for a full build (shared, static, and profiled) on Atari
148 Falcon (Motorola 68030 @ 16 Mhz, 14 Mb memory) and Jan Barte
149 <yann@plato.uni-paderborn.de> reports 22h48m on Atari TT030
150 (Motorola 68030 @ 32 Mhz, 34 Mb memory)
152 If you have some more measurements let me know.
154 ?? What version of the Linux kernel headers should be used?
156 {AJ,UD} The headers from the most recent Linux kernel should be used. The
157 headers used while compiling the GNU C library and the kernel binary used
158 when using the library do not need to match. The GNU C library runs without
159 problems on kernels that are older than the kernel headers used. The other
160 way round (compiling the GNU C library with old kernel headers and running
161 on a recent kernel) does not necessarily work. For example you can't use
162 new kernel features when using old kernel headers for compiling the GNU C
165 {ZW} Even if you are using a 2.0 kernel on your machine, we recommend you
166 compile GNU libc with 2.1 kernel headers. That way you won't have to
167 recompile libc if you ever upgrade to kernel 2.1 or 2.2. To tell libc which
168 headers to use, give configure the --with-headers switch
169 (e.g. --with-headers=/usr/src/linux-2.1.107/include).
171 Note that you must configure the 2.1 kernel if you do this; otherwise libc
172 will be unable to find <linux/version.h>. Just copy .config from your 2.0
173 kernel sources to the 2.1 tree, do `make oldconfig', and say no to all the
176 ?? The compiler hangs while building iconvdata modules. What's
179 {ZW} This is a problem of older GCC. Initialization of large static arrays
180 is very slow. The compiler will eventually finish; give it time.
182 The problem is fixed in egcs 1.1 but not in earlier releases.
184 ?? When I run `nm -u libc.so' on the produced library I still
185 find unresolved symbols. Can this be ok?
187 {UD} Yes, this is ok. There can be several kinds of unresolved symbols:
189 * magic symbols automatically generated by the linker. These have names
190 like __start_* and __stop_*
192 * symbols starting with _dl_* come from the dynamic linker
194 * weak symbols, which need not be resolved at all (fabs for example)
196 Generally, you should make sure you find a real program which produces
197 errors while linking before deciding there is a problem.
199 ??addon What are these `add-ons'?
201 {UD} To avoid complications with export rules or external source code some
202 optional parts of the libc are distributed as separate packages (e.g., the
203 crypt package, see ?crypt).
205 To use these packages as part of GNU libc, just unpack the tarfiles in the
206 libc source directory and tell the configuration script about them using the
207 --enable-add-ons option. If you give just --enable-add-ons configure tries
208 to find all the add-on packages in your source tree. This may not work. If
209 it doesn't, or if you want to select only a subset of the add-ons, give a
210 comma-separated list of the add-ons to enable:
212 configure --enable-add-ons=crypt,linuxthreads
216 Add-ons can add features (including entirely new shared libraries), override
217 files, provide support for additional architectures, and just about anything
218 else. The existing makefiles do most of the work; only some few stub rules
219 must be written to get everything running.
221 ?? My XXX kernel emulates a floating-point coprocessor for me.
222 Should I enable --with-fp?
224 {ZW} An emulated FPU is just as good as a real one, as far as the C library
225 is concerned. You only need to say --without-fp if your machine has no way
226 to execute floating-point instructions.
228 People who are interested in squeezing the last drop of performance
229 out of their machine may wish to avoid the trap overhead, but this is
230 far more trouble than it's worth: you then have to compile
231 *everything* this way, including the compiler's internal libraries
232 (libgcc.a for GNU C), because the calling conventions change.
234 ?? When compiling GNU libc I get lots of errors saying functions
235 in glibc are duplicated in libgcc.
237 {EY} This is *exactly* the same problem that I was having. The problem was
238 due to the fact that configure didn't correctly detect that the linker flag
239 --no-whole-archive was supported in my linker. In my case it was because I
240 had run ./configure with bogus CFLAGS, and the test failed.
242 One thing that is particularly annoying about this problem is that once this
243 is misdetected, running configure again won't fix it unless you first delete
246 {UD} Starting with glibc-2.0.3 there should be a better test to avoid some
247 problems of this kind. The setting of CFLAGS is checked at the very
248 beginning and if it is not usable `configure' will bark.
250 ?? Why do I get messages about missing thread functions when I use
251 librt? I don't even use threads.
253 {UD} In this case you probably mixed up your installation. librt uses
254 threads internally and has implicit references to the thread library.
255 Normally these references are satisfied automatically but if the thread
256 library is not in the expected place you must tell the linker where it is.
257 When using GNU ld it works like this:
259 gcc -o foo foo.c -Wl,-rpath-link=/some/other/dir -lrt
261 The `/some/other/dir' should contain the thread library. `ld' will use the
262 given path to find the implicitly referenced library while not disturbing
265 ?? What's the problem with configure --enable-omitfp?
267 {AJ} When --enable-omitfp is set the libraries are built without frame
268 pointers. Some compilers produce buggy code for this model and therefore we
269 don't advise using it at the moment.
271 If you use --enable-omitfp, you're on your own. If you encounter problems
272 with a library that was build this way, we advise you to rebuild the library
273 without --enable-omitfp. If the problem vanishes consider tracking the
274 problem down and report it as compiler failure.
276 Since a library build with --enable-omitfp is undebuggable on most systems,
277 debuggable libraries are also built - you can use it by appending "_g" to
280 The compilation of these extra libraries and the compiler optimizations slow
281 down the build process and need more disk space.
283 ?? I get failures during `make check'. What shall I do?
285 {AJ} The testsuite should compile and run cleanly on your system, every
286 failure should be looked into. Depending on the failure I wouldn't advise
287 installing the library at all.
289 You should consider using the `glibcbug' script to report the failure,
290 providing as much detail as possible. If you run a test directly, please
291 remember to set up the environment correctly. You want to test the compiled
292 library - and not your installed one. The best way is to copy the exact
293 command line which failed and run the test from the subdirectory for this
296 There are some failures which are not directly related to the GNU libc:
297 - Some compiler produce buggy code. The egcs 1.1 release should be ok. gcc
298 2.8.1 might cause some failures, gcc 2.7.2.x is so buggy, that explicit
299 checks have been used so that you can't build with it.
300 - The kernel might have bugs. For example on Linux/Alpha 2.0.34 the
301 floating point handling has quite a number of bugs and therefore most of
302 the test cases in the math subdirectory will fail. The current Linux 2.1
303 development kernels have fixes for the floating point support on Alpha.
305 ?? What is symbol versioning good for? Do I need it?
307 {AJ} Symbol versioning solves problems that are related to interface
308 changes. One version of an interface might have been introduced in a
309 previous version of the GNU C library but the interface or the semantics of
310 the function has been changed in the meantime. For binary compatibility
311 with the old library, a newer library needs to still have the old interface
312 for old programs. On the other hand new programs should use the new
313 interface. Symbol versioning is the solution for this problem. The GNU
314 libc version 2.1 uses by default symbol versioning if the binutils support
317 We don't advise to build without symbol versioning since you lose binary
318 compatibility if you do - for ever! The binary compatibility you lose is
319 not only against the previous version of the GNU libc (version 2.0) but also
320 against future versions.
323 ? Installation and configuration issues
325 ?? Can I replace the libc on my Linux system with GNU libc?
327 {UD} You cannot replace any existing libc for Linux with GNU libc. It is
328 binary incompatible and therefore has a different major version. You can,
329 however, install it alongside your existing libc.
331 For Linux there are three major libc versions:
333 libc-5 original ELF libc
336 You can have any combination of these three installed. For more information
337 consult documentation for shared library handling. The Makefiles of GNU
338 libc will automatically generate the needed symbolic links which the linker
341 ?? How do I configure GNU libc so that the essential libraries
342 like libc.so go into /lib and the other into /usr/lib?
344 {UD,AJ} Like all other GNU packages GNU libc is designed to use a base
345 directory and install all files relative to this. The default is
346 /usr/local, because this is safe (it will not damage the system if installed
347 there). If you wish to install GNU libc as the primary C library on your
348 system, set the base directory to /usr (i.e. run configure --prefix=/usr
349 <other_options>). Note that this can damage your system; see ?safety for
352 Some systems like Linux have a filesystem standard which makes a difference
353 between essential libraries and others. Essential libraries are placed in
354 /lib because this directory is required to be located on the same disk
355 partition as /. The /usr subtree might be found on another
356 partition/disk. If you configure for Linux with --prefix=/usr, then this
357 will be done automatically.
359 To install the essential libraries which come with GNU libc in /lib on
360 systems other than Linux one must explicitly request it. Autoconf has no
361 option for this so you have to use a `configparms' file (see the `INSTALL'
362 file for details). It should contain:
367 The first line specifies the directory for the essential libraries, the
368 second line the directory for system configuration files.
370 ??safety How should I avoid damaging my system when I install GNU libc?
372 {ZW} If you wish to be cautious, do not configure with --prefix=/usr. If
373 you don't specify a prefix, glibc will be installed in /usr/local, where it
374 will probably not break anything. (If you wish to be certain, set the
375 prefix to something like /usr/local/glibc2 which is not used for anything.)
377 The dangers when installing glibc in /usr are twofold:
379 * glibc will overwrite the headers in /usr/include. Other C libraries
380 install a different but overlapping set of headers there, so the
381 effect will probably be that you can't compile anything. You need to
382 rename /usr/include out of the way first. (Do not throw it away; you
383 will then lose the ability to compile programs against your old libc.)
385 * None of your old libraries, static or shared, can be used with a
386 different C library major version. For shared libraries this is not a
387 problem, because the filenames are different and the dynamic linker
388 will enforce the restriction. But static libraries have no version
389 information. You have to evacuate all the static libraries in
390 /usr/lib to a safe location.
392 The situation is rather similar to the move from a.out to ELF which
393 long-time Linux users will remember.
395 ?? Do I need to use GNU CC to compile programs that will use the
398 {ZW} In theory, no; the linker does not care, and the headers are supposed
399 to check for GNU CC before using its extensions to the C language.
401 However, there are currently no ports of glibc to systems where another
402 compiler is the default, so no one has tested the headers extensively
403 against another compiler. You may therefore encounter difficulties. If you
404 do, please report them as bugs.
406 Also, in several places GNU extensions provide large benefits in code
407 quality. For example, the library has hand-optimized, inline assembly
408 versions of some string functions. These can only be used with GCC. See
411 ??crypt When linking with the new libc I get unresolved symbols
412 `crypt' and `setkey'. Why aren't these functions in the
415 {UD} The US places restrictions on exporting cryptographic programs and
416 source code. Until this law gets abolished we cannot ship the cryptographic
417 functions together with glibc.
419 The functions are available, as an add-on (see ?addon). People in the US
420 may get it from the same place they got GNU libc from. People outside the
421 US should get the code from ftp://ftp.ifi.uio.no/pub/gnu, or another archive
422 site outside the USA. The README explains how to install the sources.
424 If you already have the crypt code on your system the reason for the failure
425 is probably that you did not link with -lcrypt. The crypto functions are in
426 a separate library to make it possible to export GNU libc binaries from the
429 ?? When I use GNU libc on my Linux system by linking against
430 the libc.so which comes with glibc all I get is a core dump.
432 {UD} On Linux, gcc sets the dynamic linker to /lib/ld-linux.so.1 unless the
433 user specifies a -dynamic-linker argument. This is the name of the libc5
434 dynamic linker, which does not work with glibc.
436 For casual use of GNU libc you can just specify to the linker
437 --dynamic-linker=/lib/ld-linux.so.2
439 which is the glibc dynamic linker, on Linux systems. On other systems the
440 name is /lib/ld.so.1. When linking via gcc, you've got to add
441 -Wl,--dynamic-linker=/lib/ld-linux.so.2
443 to the gcc command line.
445 To change your environment to use GNU libc for compiling you need to change
446 the `specs' file of your gcc. This file is normally found at
448 /usr/lib/gcc-lib/<arch>/<version>/specs
450 In this file you have to change a few things:
452 - change `ld-linux.so.1' to `ld-linux.so.2'
454 - remove all expression `%{...:-lgmon}'; there is no libgmon in glibc
456 - fix a minor bug by changing %{pipe:-} to %|
458 Here is what the gcc-2.7.2 specs file should look like when GNU libc is
461 -----------------------------------------------------------------------
463 %{V} %{v:%{!V:-V}} %{Qy:} %{!Qn:-Qy} %{n} %{T} %{Ym,*} %{Yd,*} %{Wa,*:%*}
469 %{fPIC:-D__PIC__ -D__pic__} %{fpic:-D__PIC__ -D__pic__} %{!m386:-D__i486__} %{posix:-D_POSIX_SOURCE} %{pthread:-D_REENTRANT}
478 %{!shared:crtend.o%s} %{shared:crtendS.o%s} crtn.o%s
481 -m elf_i386 %{shared:-shared} %{!shared: %{!ibcs: %{!static: %{rdynamic:-export-dynamic} %{!dynamic-linker:-dynamic-linker /lib/ld-linux.so.2}} %{static:-static}}}
484 %{!shared: %{pthread:-lpthread} %{profile:-lc_p} %{!profile: -lc}}
490 %{!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}
492 *switches_need_spaces:
496 %{funsigned-char:-D__CHAR_UNSIGNED__}
499 -D__ELF__ -Dunix -Di386 -Dlinux -Asystem(unix) -Asystem(posix) -Acpu(i386) -Amachine(i386)
507 -----------------------------------------------------------------------
509 Things get a bit more complicated if you have GNU libc installed in some
510 other place than /usr, i.e., if you do not want to use it instead of the old
511 libc. In this case the needed startup files and libraries are not found in
512 the regular places. So the specs file must tell the compiler and linker
515 Version 2.7.2.3 does and future versions of GCC will automatically
516 provide the correct specs.
518 ?? Looking through the shared libc file I haven't found the
519 functions `stat', `lstat', `fstat', and `mknod' and while
520 linking on my Linux system I get error messages. How is
521 this supposed to work?
523 {RM} Believe it or not, stat and lstat (and fstat, and mknod) are supposed
524 to be undefined references in libc.so.6! Your problem is probably a missing
525 or incorrect /usr/lib/libc.so file; note that this is a small text file now,
526 not a symlink to libc.so.6. It should look something like this:
528 GROUP ( libc.so.6 libc_nonshared.a )
530 ?? How can I compile gcc 2.7.2.1 from the gcc source code using
533 {AJ} There's only correct support for glibc 2.0.x in gcc 2.7.2.3 or later.
534 But you should get at least gcc 2.8.1 or egcs 1.0.2 (or later versions)
537 ?? The `gencat' utility cannot process the catalog sources which
538 were used on my Linux libc5 based system. Why?
540 {UD} The `gencat' utility provided with glibc complies to the XPG standard.
541 The older Linux version did not obey the standard, so they are not
544 To ease the transition from the Linux version some of the non-standard
545 features are also present in the `gencat' program of GNU libc. This mainly
546 includes the use of symbols for the message number and the automatic
547 generation of header files which contain the needed #defines to map the
550 Here is a simple SED script to convert at least some Linux specific catalog
551 files to the XPG4 form:
553 -----------------------------------------------------------------------
554 # Change catalog source in Linux specific format to standard XPG format.
555 # Ulrich Drepper <drepper@cygnus.com>, 1996.
559 s/\$ #\([^ ]*\).*/\1/
561 s/\$ #[^ ]* *\(.*\)/\$ \1/
567 s/\(.*\)\n\(.*\)/\2 \1/
569 -----------------------------------------------------------------------
571 ?? Programs using libc have their messages translated, but other
572 behavior is not localized (e.g. collating order); why?
574 {ZW} Translated messages are automatically installed, but the locale
575 database that controls other behaviors is not. You need to run localedef to
576 install this database, after you have run `make install'. For example, to
577 set up the French Canadian locale, simply issue the command
579 localedef -i fr_CA -f ISO-8859-1 fr_CA
581 Please see localedata/README in the source tree for further details.
583 ?? I have set up /etc/nis.conf, and the Linux libc 5 with NYS
584 works great. But the glibc NIS+ doesn't seem to work.
586 {TK} The glibc NIS+ implementation uses a /var/nis/NIS_COLD_START file for
587 storing information about the NIS+ server and their public keys, because the
588 nis.conf file does not contain all the necessary information. You have to
589 copy a NIS_COLD_START file from a Solaris client (the NIS_COLD_START file is
590 byte order independent) or generate it with nisinit from the nis-tools
591 package; available at
593 http://www-vt.uni-paderborn.de/~kukuk/linux/nisplus.html
595 ?? I have killed ypbind to stop using NIS, but glibc
598 {TK} For faster NIS lookups, glibc uses the /var/yp/binding/ files from
599 ypbind. ypbind 3.3 and older versions don't always remove these files, so
600 glibc will continue to use them. Other BSD versions seem to work correctly.
601 Until ypbind 3.4 is released, you can find a patch at
603 ftp://ftp.kernel.org/pub/linux/utils/net/NIS/ypbind-3.3-glibc3.diff.gz
605 ?? Under Linux/Alpha, I always get "do_ypcall: clnt_call:
606 RPC: Unable to receive; errno = Connection refused" when using NIS.
608 {TK} You need a ypbind version which is 64bit clean. Some versions are not
609 64bit clean. A 64bit clean implementation is ypbind-mt. For ypbind 3.3,
610 you need the patch from ftp.kernel.org (See the previous question). I don't
611 know about other versions.
614 ?? After installing glibc name resolving doesn't work properly.
616 {AJ} You probably should read the manual section describing nsswitch.conf
617 (just type `info libc "NSS Configuration File"'). The NSS configuration
618 file is usually the culprit.
621 ?? How do I create the databases for NSS?
623 {AJ} If you have an entry "db" in /etc/nsswitch.conf you should also create
624 the database files. The glibc sources contain a Makefile which does the
625 necessary conversion and calls to create those files. The file is
626 `db-Makefile' in the subdirectory `nss' and you can call it with `make -f
627 db-Makefile'. Please note that not all services are capable of using a
628 database. Currently passwd, group, ethers, protocol, rpc, services shadow
629 and netgroup are implemented.
632 ?? I have /usr/include/net and /usr/include/scsi as symlinks
633 into my Linux source tree. Is that wrong?
635 {PB} This was necessary for libc5, but is not correct when using glibc.
636 Including the kernel header files directly in user programs usually does not
637 work (see ?kerhdr). glibc provides its own <net/*> and <scsi/*> header
638 files to replace them, and you may have to remove any symlink that you have
639 in place before you install glibc. However, /usr/include/asm and
640 /usr/include/linux should remain as they were.
642 ?? Programs like `logname', `top', `uptime' `users', `w' and
643 `who', show incorrect information about the (number of)
644 users on my system. Why?
648 ?? After upgrading to glibc 2.1 with symbol versioning I get
649 errors about undefined symbols. What went wrong?
651 {AJ} The problem is caused either by wrong program code or tools. In the
652 versioned libc a lot of symbols are now local that were global symbols in
653 previous versions. It seems that programs linked against older versions
654 often accidentally used libc global variables -- something that should not
657 The only way to fix this is to recompile your program. Sorry, that's the
658 price you might have to pay once for quite a number of advantages with
661 ?? When I start the program XXX after upgrading the library
663 XXX: Symbol `_sys_errlist' has different size in shared
664 object, consider re-linking
665 Why? What should I do?
667 {UD} As the message says, relink the binary. The problem is that a few
668 symbols from the library can change in size and there is no way to avoid
669 this. _sys_errlist is a good example. Occasionally there are new error
670 numbers added to the kernel and this must be reflected at user level,
671 breaking programs that refer to them directly.
673 Such symbols should normally not be used at all. There are mechanisms to
674 avoid using them. In the case of _sys_errlist, there is the strerror()
675 function which should _always_ be used instead. So the correct fix is to
676 rewrite that part of the application.
678 In some situations (especially when testing a new library release) it might
679 be possible that a symbol changed size when that should not have happened.
680 So in case of doubt report such a warning message as a problem.
682 ?? What do I need for C++ development?
684 {HJ,AJ} You need either egcs 1.0.2 or gcc-2.8.1 with libstdc++ 2.8.1 (or
685 more recent versions). libg++ 2.7.2 (and the Linux Versions 2.7.2.x) doesn't
686 work very well with the GNU C library due to vtable thunks. If you're
687 upgrading from glibc 2.0.x to 2.1 you have to recompile libstdc++ since the
688 library compiled for 2.0 is not compatible due to the new Large File Support
689 (LFS) in version 2.1.
691 ?? Even statically linked programs need some shared libraries
692 which is not acceptable for me. What can I do?
694 {AJ} NSS (for details just type `info libc "Name Service Switch"') won't
695 work properly without shared libraries. NSS allows using different services
696 (e.g. NIS, files, db, hesiod) by just changing one configuration file
697 (/etc/nsswitch.conf) without relinking any programs. The only disadvantage
698 is that now static libraries need to access shared libraries. This is
699 handled transparently by the GNU C library.
701 A solution is to configure glibc with --enable-static-nss. In this case you
702 can create a static binary that will use only the services dns and files
703 (change /etc/nsswitch.conf for this). You need to link explicitly against
704 all these services. For example:
706 gcc -static test-netdb.c -o test-netdb.c \
707 -lc -lnss_files -lnss_dns -lresolv
709 The problem with this approach is that you've got to link every static
710 program that uses NSS routines with all those libraries.
712 {UD} In fact, one cannot say anymore that a libc compiled with this
713 option is using NSS. There is no switch anymore. Therefore it is
714 *highly* recommended *not* to use --enable-static-nss since this makes
715 the behaviour of the programs on the system inconsistent.
717 ?? I just upgraded my Linux system to glibc and now I get
718 errors whenever I try to link any program.
720 {ZW} This happens when you have installed glibc as the primary C library but
721 have stray symbolic links pointing at your old C library. If the first
722 `libc.so' the linker finds is libc 5, it will use that. Your program
723 expects to be linked with glibc, so the link fails.
725 The most common case is that glibc put its `libc.so' in /usr/lib, but there
726 was a `libc.so' from libc 5 in /lib, which gets searched first. To fix the
727 problem, just delete /lib/libc.so. You may also need to delete other
728 symbolic links in /lib, such as /lib/libm.so if it points to libm.so.5.
730 {AJ} The perl script test-installation.pl which is run as last step during
731 an installation of glibc that is configured with --prefix=/usr should help
732 detect these situations. If the script reports problems, something is
735 ? Source and binary incompatibilities, and what to do about them
737 ?? I expect GNU libc to be 100% source code compatible with
738 the old Linux based GNU libc. Why isn't it like this?
740 {DMT,UD} Not every extension in Linux libc's history was well thought-out.
741 In fact it had a lot of problems with standards compliance and with
742 cleanliness. With the introduction of a new version number these errors can
743 now be corrected. Here is a list of the known source code
746 * _GNU_SOURCE: glibc does not make the GNU extensions available
747 automatically. If a program depends on GNU extensions or some
748 other non-standard functionality, it is necessary to compile it
749 with the C compiler option -D_GNU_SOURCE, or better, to put
750 `#define _GNU_SOURCE' at the beginning of your source files, before
751 any C library header files are included. This difference normally
752 manifests itself in the form of missing prototypes and/or data type
753 definitions. Thus, if you get such errors, the first thing you
754 should do is try defining _GNU_SOURCE and see if that makes the
757 For more information consult the file `NOTES' in the GNU C library
760 * reboot(): GNU libc sanitizes the interface of reboot() to be more
761 compatible with the interface used on other OSes. reboot() as
762 implemented in glibc takes just one argument. This argument
763 corresponds to the third argument of the Linux reboot system call.
764 That is, a call of the form reboot(a, b, c) needs to be changed into
765 reboot(c). Beside this the header <sys/reboot.h> defines the needed
766 constants for the argument. These RB_* constants should be used
767 instead of the cryptic magic numbers.
769 * swapon(): the interface of this function didn't change, but the
770 prototype is in a separate header file <sys/swap.h>. This header
771 file also provides the SWAP_* constants defined by <linux/swap.h>;
772 you should use them for the second argument to swapon().
774 * errno: If a program uses the variable "errno", then it _must_
775 include <errno.h>. The old libc often (erroneously) declared this
776 variable implicitly as a side-effect of including other libc header
777 files. glibc is careful to avoid such namespace pollution, which,
778 in turn, means that you really need to include the header files that
779 you depend on. This difference normally manifests itself in the
780 form of the compiler complaining about references to an undeclared
783 * Linux-specific syscalls: All Linux system calls now have appropriate
784 library wrappers and corresponding declarations in various header files.
785 This is because the syscall() macro that was traditionally used to
786 work around missing syscall wrappers are inherently non-portable and
787 error-prone. The following table lists all the new syscall stubs,
788 the header-file declaring their interface and the system call name.
790 syscall name: wrapper name: declaring header file:
791 ------------- ------------- ----------------------
792 bdflush bdflush <sys/kdaemon.h>
793 syslog ksyslog_ctl <sys/klog.h>
795 * lpd: Older versions of lpd depend on a routine called _validuser().
796 The library does not provide this function, but instead provides
797 __ivaliduser() which has a slightly different interface. Simply
798 upgrading to a newer lpd should fix this problem (e.g., the 4.4BSD
799 lpd is known to be working).
801 * resolver functions/BIND: like on many other systems the functions of
802 the resolver library are not included in libc itself. There is a
803 separate library libresolv. If you get undefined symbol errors for
804 symbols starting with `res_*' simply add -lresolv to your linker
807 * the `signal' function's behavior corresponds to the BSD semantic and
808 not the SysV semantic as it was in libc-5. The interface on all GNU
809 systems shall be the same and BSD is the semantic of choice. To use
810 the SysV behavior simply use `sysv_signal', or define _XOPEN_SOURCE.
811 See ?signal for details.
813 ??getlog Why does getlogin() always return NULL on my Linux box?
815 {UD} The GNU C library has a format for the UTMP and WTMP file which differs
816 from what your system currently has. It was extended to fulfill the needs
817 of the next years when IPv6 is introduced. The record size is different and
818 some fields have different positions. The files written by functions from
819 the one library cannot be read by functions from the other library. Sorry,
820 but this is what a major release is for. It's better to have a cut now than
821 having no means to support the new techniques later.
823 {MK} There is however a (partial) solution for this problem. Please take a
824 look at the file `login/README.utmpd'.
826 ?? Where are the DST_* constants found in <sys/time.h> on many
829 {UD} These constants come from the old BSD days and are not used anymore
830 (libc5 does not actually implement the handling although the constants are
833 Instead GNU libc contains zone database support and compatibility code for
834 POSIX TZ environment variable handling.
836 ?? The prototypes for `connect', `accept', `getsockopt',
837 `setsockopt', `getsockname', `getpeername', `send',
838 `sendto', and `recvfrom' are different in GNU libc from
839 any other system I saw. This is a bug, isn't it?
841 {UD} No, this is no bug. This version of GNU libc already follows the new
842 Single Unix specifications (and I think the POSIX.1g draft which adopted the
843 solution). The type for a parameter describing a size is now `socklen_t', a
846 ??kerhdr On Linux I've got problems with the declarations in Linux
849 {UD,AJ} On Linux, the use of kernel headers is reduced to the minimum. This
850 gives Linus the ability to change the headers more freely. Also, user
851 programs are now insulated from changes in the size of kernel data
854 For example, the sigset_t type is 32 or 64 bits wide in the kernel. In
855 glibc it is 1024 bits wide. This guarantees that when the kernel gets a
856 bigger sigset_t (for POSIX.1e realtime support, say) user programs will not
857 have to be recompiled. Consult the header files for more information about
860 Therefore you shouldn't include Linux kernel header files directly if glibc
861 has defined a replacement. Otherwise you might get undefined results because
864 ?? I don't include any kernel headers myself but the compiler
865 still complains about redeclarations of types in the kernel
868 {UD} The kernel headers before Linux 2.1.61 and 2.0.32 don't work correctly
869 with glibc. Compiling C programs is possible in most cases but C++ programs
870 have (due to the change of the name lookups for `struct's) problems. One
871 prominent example is `struct fd_set'.
873 There might be some problems left but 2.1.61/2.0.32 fix most of the known
874 ones. See the BUGS file for other known problems.
876 ??signal Why don't signals interrupt system calls anymore?
878 {ZW} By default GNU libc uses the BSD semantics for signal(), unlike Linux
879 libc 5 which used System V semantics. This is partially for compatibility
880 with other systems and partially because the BSD semantics tend to make
881 programming with signals easier.
883 There are three differences:
885 * BSD-style signals that occur in the middle of a system call do not
886 affect the system call; System V signals cause the system call to
887 fail and set errno to EINTR.
889 * BSD signal handlers remain installed once triggered. System V signal
890 handlers work only once, so one must reinstall them each time.
892 * A BSD signal is blocked during the execution of its handler. In other
893 words, a handler for SIGCHLD (for example) does not need to worry about
894 being interrupted by another SIGCHLD. It may, however, be interrupted
897 There is general consensus that for `casual' programming with signals, the
898 BSD semantics are preferable. You don't need to worry about system calls
899 returning EINTR, and you don't need to worry about the race conditions
900 associated with one-shot signal handlers.
902 If you are porting an old program that relies on the old semantics, you can
903 quickly fix the problem by changing signal() to sysv_signal() throughout.
904 Alternatively, define _XOPEN_SOURCE before including <signal.h>.
906 For new programs, the sigaction() function allows you to specify precisely
907 how you want your signals to behave. All three differences listed above are
908 individually switchable on a per-signal basis with this function.
910 If all you want is for one specific signal to cause system calls to fail and
911 return EINTR (for example, to implement a timeout) you can do this with
915 ??string I've got errors compiling code that uses certain string
918 {AJ} glibc 2.1 has special string functions that are faster than the normal
919 library functions. Some of the functions are additionally implemented as
920 inline functions and others as macros.
922 The optimized string functions are only used when compiling with
923 optimizations (-O1 or higher). The behavior can be changed with two feature
926 * __NO_STRING_INLINES: Don't do any string optimizations.
927 * __USE_STRING_INLINES: Use assembly language inline functions (might
928 increase code size dramatically).
930 Since some of these string functions are now additionally defined as macros,
931 code like "char *strncpy();" doesn't work anymore (and is unnecessary, since
932 <string.h> has the necessary declarations). Either change your code or
933 define __NO_STRING_INLINES.
935 {UD} Another problem in this area is that gcc still has problems on machines
936 with very few registers (e.g., ix86). The inline assembler code can require
937 almost all the registers and the register allocator cannot always handle
940 One can disable the string optimizations selectively. Instead of writing
942 cp = strcpy (foo, "lkj");
946 cp = (strcpy) (foo, "lkj");
948 This disables the optimization for that specific call.
950 ?? I get compiler messages "Initializer element not constant" with
951 stdin/stdout/stderr. Why?
953 {RM,AJ} Constructs like:
954 static FILE *InPtr = stdin;
956 lead to this message. This is correct behaviour with glibc since stdin is
957 not a constant expression. Please note that a strict reading of ISO C does
958 not allow above constructs.
960 One of the advantages of this is that you can assign to stdin, stdout, and
961 stderr just like any other global variable (e.g. `stdout = my_stream;'),
962 which can be very useful with custom streams that you can write with libio
963 (but beware this is not necessarily portable). The reason to implement it
964 this way were versioning problems with the size of the FILE structure.
966 To fix those programs you've got to initialize the variable at run time.
967 This can be done, e.g. in main, like:
975 or by constructors (beware this is gcc specific):
978 static void inPtr_construct (void) __attribute__((constructor));
979 static void inPtr_construct (void) { InPtr = stdin; }
982 ?? I can't compile with gcc -traditional (or
983 -traditional-cpp). Why?
985 {AJ} glibc2 does break -traditional and -traditonal-cpp - and will continue
986 to do so. For example constructs of the form:
992 are useful for debugging purposes (you can use foo with your debugger that's
993 why we need the enum) and for compatibility (other systems use defines and
996 ?? I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible?
998 {AJ} The GNU C library is compatible with the ANSI/ISO C standard. If
999 you're using `gcc -ansi', the glibc includes which are specified in the
1000 standard follow the standard. The ANSI/ISO C standard defines what has to be
1001 in the include files - and also states that nothing else should be in the
1002 include files (btw. you can still enable additional standards with feature
1005 The GNU C library is conforming to ANSI/ISO C - if and only if you're only
1006 using the headers and library functions defined in the standard.
1008 ?? I can't access some functions anymore. nm shows that they do
1009 exist but linking fails nevertheless.
1011 {AJ} With the introduction of versioning in glibc 2.1 it is possible to
1012 export only those identifiers (functions, variables) that are really needed
1013 by application programs and by other parts of glibc. This way a lot of
1014 internal interfaces are now hidden. nm will still show those identifiers
1015 but marking them as internal. ISO C states that identifiers beginning with
1016 an underscore are internal to the libc. An application program normally
1017 shouldn't use those internal interfaces (there are exceptions,
1018 e.g. __ivaliduser). If a program uses these interfaces, it's broken. These
1019 internal interfaces might change between glibc releases or dropped
1022 ?? When using the db-2 library which comes with glibc is used in
1023 the Perl db modules the testsuite is not passed. This did not
1024 happen with db-1, gdbm, or ndbm.
1026 {UD} You are using an outdated copy of the DB_File Perl module. In fact db-2
1027 finally removed the handling of zero-sized keys which was one of the features
1028 tested by the old Perl testsuite and therefore you see an error. But this
1029 never was documented and guaranteed, only broken programs used this feature.
1031 Consequently db-2 does not need to support this feature and instead signals
1032 an error which leads to easier debugging. The DB_File module maintainer
1033 Paul Marquess <pmarquess@bfsec.bt.co.uk> acknowledged this change and fixed
1034 the testsuite so that if you use DB_File v1.60 or later you should not have
1035 any more problems with db-2.
1037 ?? The pow() inline function I get when including <math.h> is broken.
1038 I get segmentation faults when I run the program.
1040 {UD} Nope, the implementation is correct. The problem is with egcs version
1041 prior to 1.1. I.e., egcs 1.0 to 1.0.3 are all broken (at least on Intel).
1042 If you have to use this compiler you must define __NO_MATH_INLINES before
1043 including <math.h> to prevent the inline functions from being used. egcs 1.1
1044 fixes the problem. I don't know about gcc 2.8 and 2.8.1.
1049 ?? After I changed configure.in I get `Autoconf version X.Y.
1050 or higher is required for this script'. What can I do?
1052 {UD} You have to get the specified autoconf version (or a later one)
1053 from your favorite mirror of ftp.gnu.org.
1055 ?? When I try to compile code which uses IPv6 headers and
1056 definitions on my Linux 2.x.y system I am in trouble.
1057 Nothing seems to work.
1059 {UD} The problem is that IPv6 development still has not reached a point
1060 where the headers are stable. There are still lots of incompatible changes
1061 made and the libc headers have to follow.
1063 Also, make sure you have a suitably recent kernel. As of the 970401
1064 snapshot, according to Philip Blundell <Philip.Blundell@pobox.com>, the
1065 required kernel version is at least 2.1.30.
1067 ?? When I set the timezone by setting the TZ environment variable
1068 to EST5EDT things go wrong since glibc computes the wrong time
1069 from this information.
1071 {UD} The problem is that people still use the braindamaged POSIX method to
1072 select the timezone using the TZ environment variable with a format EST5EDT
1073 or whatever. People, read the POSIX standard, the implemented behaviour is
1074 correct! What you see is in fact the result of the decisions made while
1075 POSIX.1 was created. We've only implemented the handling of TZ this way to
1076 be POSIX compliant. It is not really meant to be used.
1078 The alternative approach to handle timezones which is implemented is the
1079 correct one to use: use the timezone database. This avoids all the problems
1080 the POSIX method has plus it is much easier to use. Simply run the tzselect
1081 shell script, answer the question and use the name printed in the end by
1082 making a symlink to /usr/share/zoneinfo/NAME (NAME is the returned value
1083 from tzselect) from the file /etc/localtime. That's all. You never again
1086 So, please avoid sending bug reports about time related problems if you use
1087 the POSIX method and you have not verified something is really broken by
1088 reading the POSIX standards.
1090 ?? What other sources of documentation about glibc are available?
1092 {AJ} The FSF has a page about the GNU C library at
1093 <http://www.gnu.org/software/libc/>. The problem data base of open and
1094 solved bugs in GNU libc is available at
1095 <http://www-gnats.gnu.org:8080/cgi-bin/wwwgnats.pl>. Eric Green has written
1096 a HowTo for converting from Linux libc5 to glibc2. The HowTo is accessable
1097 via the FSF page and at <http://www.imaxx.net/~thrytis/glibc>. Frodo
1098 Looijaard describes a different way installing glibc2 as secondary libc at
1099 <http://huizen.dds.nl/~frodol/glibc>.
1101 Please note that this is not a complete list.
1104 Answers were given by:
1105 {UD} Ulrich Drepper, <drepper@cygnus.com>
1106 {DMT} David Mosberger-Tang, <davidm@AZStarNet.com>
1107 {RM} Roland McGrath, <roland@gnu.org>
1108 {AJ} Andreas Jaeger, <aj@arthur.rhein-neckar.de>
1109 {EY} Eric Youngdale, <eric@andante.jic.com>
1110 {PB} Phil Blundell, <Philip.Blundell@pobox.com>
1111 {MK} Mark Kettenis, <kettenis@phys.uva.nl>
1112 {ZW} Zack Weinberg, <zack@rabi.phys.columbia.edu>
1113 {TK} Thorsten Kukuk, <kukuk@vt.uni-paderborn.de>
1114 {GK} Geoffrey Keating, <geoffk@ozemail.com.au>
1115 {HJ} H.J. Lu, <hjl@gnu.org>
1119 outline-regexp:"\\?"
projects / platform / upstream / glibc.git / blob