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-*-linux Linux-2.x on ARM
37 arm-*-linuxaout Linux-2.x on ARM using a.out binaries
38 mips*-*-linux-gnu Linux-2.x on MIPS
39 ia64-*-linux-gnu Linux-2.x on ia64
40 s390-*-linux-gnu Linux-2.x on IBM S/390
42 Ports to other Linux platforms are in development, and may in fact work
43 already, but no one has sent us success reports for them. Currently no
44 ports to other operating systems are underway, although a few people have
47 If you have a system not listed above (or in the `README' file) and you are
48 really interested in porting it, contact
52 ??binsize What compiler do I need to build GNU libc?
54 {UD} You must use GNU CC to compile GNU libc. A lot of extensions of GNU CC
55 are used to increase portability and speed.
57 GNU CC is found, like all other GNU packages, on
59 ftp://ftp.gnu.org/pub/gnu
61 and the many mirror sites. ftp.gnu.org is always overloaded, so try to find
64 You should always try to use the latest official release. Older versions
65 may not have all the features GNU libc requires. The current releases of
66 egcs (1.0.3 and 1.1.1) should work with the GNU C library (for powerpc see
67 ?powerpc; for ARM see ?arm; for MIPS see ?mips).
69 While the GNU CC should be able to compile glibc it is nevertheless adviced
70 to use EGCS. Comparing the sizes of glibc on Intel compiled with a recent
71 EGCS and gcc 2.8.1 shows this:
73 text data bss dec hex filename
74 egcs-2.93.10 862897 15944 12824 891665 d9b11 libc.so
75 gcc-2.8.1 959965 16468 12152 988585 f15a9 libc.so
77 Make up your own decision.
79 GNU CC versions 2.95 and above are derived from egcs, and they may do even
82 Please note that gcc 2.95 and 2.95.x cannot compile glibc on Alpha due to
83 problems in the complex float support.
85 ?? When I try to compile glibc I get only error messages.
88 {UD} You definitely need GNU make to build GNU libc. No other make
89 program has the needed functionality.
91 We recommend version GNU make version 3.79 or newer. Older versions have
92 bugs and/or are missing features.
94 ?? Do I need a special linker or assembler?
96 {ZW} If you want a shared library, you need a linker and assembler that
97 understand all the features of ELF, including weak and versioned symbols.
98 The static library can be compiled with less featureful tools, but lacks key
101 For Linux or Hurd, you want binutils 2.8.1.0.23, 2.9.1, or 2.9.1.0.15 or
102 higher. These are the only versions we've tested and found reliable. Other
103 versions after 2.8.1.0.23 may work but we don't recommend them, especially
104 not when C++ is involved. Earlier versions do not work at all.
106 Other operating systems may come with system tools that have all the
107 necessary features, but this is moot because glibc hasn't been ported to
110 ??powerpc Which compiler should I use for powerpc?
112 {GK} You want to use at least gcc 2.95 (together with the right versions
113 of all the other tools, of course). See also question ?excpt.
115 ??arm Which tools should I use for ARM?
117 {PB} You should use egcs 1.1 or a later version. For ELF systems some
118 changes are needed to the compiler; a patch against egcs-1.1.x can be found
121 <ftp://ftp.netwinder.org/users/p/philb/egcs-1.1.1pre2-diff-981126>
123 Binutils 2.9.1.0.16 or later is also required.
125 ?? Do I need some more things to compile the GNU C Library?
127 {UD} Yes, there are some more :-).
129 * GNU gettext. This package contains the tools needed to construct
130 `message catalog' files containing translated versions of system
131 messages. See ftp://ftp.gnu.org/pub/gnu or better any mirror
132 site. (We distribute compiled message catalogs, but they may not be
133 updated in patches.) Please note that the required minimal version
134 (0.10.35) of gettext is alpha software and available from
135 ftp://alpha.gnu.org/gnu .
137 * Some files are built with special tools. E.g., files ending in .gperf
138 need a `gperf' program. The GNU version (now available in a separate
139 package, formerly only as part of libg++) is known to work while some
140 vendor versions do not.
142 You should not need these tools unless you change the source files.
144 * Perl 5 is needed if you wish to test an installation of GNU libc
145 as the primary C library.
147 * When compiling for Linux, the header files of the Linux kernel must
148 be available to the compiler as <linux/*.h> and <asm/*.h>.
150 * lots of disk space (~400MB for i?86-linux; more for RISC platforms).
152 * plenty of time. Compiling just the shared and static libraries for
153 i?86-linux takes approximately 1h on an AMD-K6@225MHz w/ 96MB of RAM,
154 45mins on a Celeron@400MHz w/ 128MB, and 55mins on a Alpha@533MHz w/ 256MB.
155 Multiply this by 1.5 or 2.0 if you build profiling and/or the highly
156 optimized version as well. For Hurd systems times are much higher.
158 You should avoid compiling in a NFS mounted filesystem. This is
161 James Troup <J.J.Troup@comp.brad.ac.uk> reports a compile time of
162 45h34m for a full build (shared, static, and profiled) on Atari
163 Falcon (Motorola 68030 @ 16 Mhz, 14 Mb memory) and Jan Barte
164 <yann@plato.uni-paderborn.de> reports 22h48m on Atari TT030
165 (Motorola 68030 @ 32 Mhz, 34 Mb memory)
167 A full build of the PowerPC library took 1h on a PowerPC 750@400Mhz w/
168 64MB of RAM, and about 9h on a 601@60Mhz w/ 72Mb.
170 If you have some more measurements let me know.
172 ?? What version of the Linux kernel headers should be used?
174 {AJ,UD} The headers from the most recent Linux kernel should be used. The
175 headers used while compiling the GNU C library and the kernel binary used
176 when using the library do not need to match. The GNU C library runs without
177 problems on kernels that are older than the kernel headers used. The other
178 way round (compiling the GNU C library with old kernel headers and running
179 on a recent kernel) does not necessarily work. For example you can't use
180 new kernel features if you used old kernel headers to compile the GNU C
183 {ZW} Even if you are using a 2.0 kernel on your machine, we recommend you
184 compile GNU libc with 2.2 kernel headers. That way you won't have to
185 recompile libc if you ever upgrade to kernel 2.2. To tell libc which
186 headers to use, give configure the --with-headers switch
187 (e.g. --with-headers=/usr/src/linux-2.2.0/include).
189 Note that you must configure the 2.2 kernel if you do this, otherwise libc
190 will be unable to find <linux/version.h>. Just change the current directory
191 to the root of the 2.2 tree and do `make include/linux/version.h'.
193 ?? The compiler hangs while building iconvdata modules. What's
196 {ZW} This is a problem with old versions of GCC. Initialization of large
197 static arrays is very slow. The compiler will eventually finish; give it
200 The problem is fixed in egcs 1.1.
202 ?? When I run `nm -u libc.so' on the produced library I still
203 find unresolved symbols. Can this be ok?
205 {UD} Yes, this is ok. There can be several kinds of unresolved symbols:
207 * magic symbols automatically generated by the linker. These have names
208 like __start_* and __stop_*
210 * symbols starting with _dl_* come from the dynamic linker
212 * weak symbols, which need not be resolved at all (fabs for example)
214 Generally, you should make sure you find a real program which produces
215 errors while linking before deciding there is a problem.
217 ??addon What are these `add-ons'?
219 {UD} To avoid complications with export rules or external source code some
220 optional parts of the libc are distributed as separate packages, e.g., the
221 linuxthreads package.
223 To use these packages as part of GNU libc, just unpack the tarfiles in the
224 libc source directory and tell the configuration script about them using the
225 --enable-add-ons option. If you give just --enable-add-ons configure tries
226 to find all the add-on packages in your source tree. This may not work. If
227 it doesn't, or if you want to select only a subset of the add-ons, give a
228 comma-separated list of the add-ons to enable:
230 configure --enable-add-ons=linuxthreads
234 Add-ons can add features (including entirely new shared libraries), override
235 files, provide support for additional architectures, and just about anything
236 else. The existing makefiles do most of the work; only some few stub rules
237 must be written to get everything running.
239 Most add-ons are tightly coupled to a specific GNU libc version. Please
240 check that the add-ons work with the GNU libc. For example the linuxthreads
241 add-on has the same numbering scheme as the libc and will in general only
242 work with the corresponding libc.
244 {AJ} With glibc 2.2 the crypt add-on and with glibc 2.1 the localedata
245 add-on have been integrated into the normal glibc distribution, crypt and
246 localedata are therefore not anymore add-ons.
248 ?? My XXX kernel emulates a floating-point coprocessor for me.
249 Should I enable --with-fp?
251 {ZW} An emulated FPU is just as good as a real one, as far as the C library
252 is concerned. You only need to say --without-fp if your machine has no way
253 to execute floating-point instructions.
255 People who are interested in squeezing the last drop of performance
256 out of their machine may wish to avoid the trap overhead, but this is
257 far more trouble than it's worth: you then have to compile
258 *everything* this way, including the compiler's internal libraries
259 (libgcc.a for GNU C), because the calling conventions change.
261 ?? When compiling GNU libc I get lots of errors saying functions
262 in glibc are duplicated in libgcc.
264 {EY} This is *exactly* the same problem that I was having. The problem was
265 due to the fact that configure didn't correctly detect that the linker flag
266 --no-whole-archive was supported in my linker. In my case it was because I
267 had run ./configure with bogus CFLAGS, and the test failed.
269 One thing that is particularly annoying about this problem is that once this
270 is misdetected, running configure again won't fix it unless you first delete
273 {UD} Starting with glibc-2.0.3 there should be a better test to avoid some
274 problems of this kind. The setting of CFLAGS is checked at the very
275 beginning and if it is not usable `configure' will bark.
277 ?? Why do I get messages about missing thread functions when I use
278 librt? I don't even use threads.
280 {UD} In this case you probably mixed up your installation. librt uses
281 threads internally and has implicit references to the thread library.
282 Normally these references are satisfied automatically but if the thread
283 library is not in the expected place you must tell the linker where it is.
284 When using GNU ld it works like this:
286 gcc -o foo foo.c -Wl,-rpath-link=/some/other/dir -lrt
288 The `/some/other/dir' should contain the thread library. `ld' will use the
289 given path to find the implicitly referenced library while not disturbing
292 ?? What's the problem with configure --enable-omitfp?
294 {AJ} When --enable-omitfp is set the libraries are built without frame
295 pointers. Some compilers produce buggy code for this model and therefore we
296 don't advise using it at the moment.
298 If you use --enable-omitfp, you're on your own. If you encounter problems
299 with a library that was build this way, we advise you to rebuild the library
300 without --enable-omitfp. If the problem vanishes consider tracking the
301 problem down and report it as compiler failure.
303 Since a library built with --enable-omitfp is undebuggable on most systems,
304 debuggable libraries are also built - you can use them by appending "_g" to
307 The compilation of these extra libraries and the compiler optimizations slow
308 down the build process and need more disk space.
310 ?? I get failures during `make check'. What should I do?
312 {AJ} The testsuite should compile and run cleanly on your system; every
313 failure should be looked into. Depending on the failures, you probably
314 should not install the library at all.
316 You should consider using the `glibcbug' script to report the failure,
317 providing as much detail as possible. If you run a test directly, please
318 remember to set up the environment correctly. You want to test the compiled
319 library - and not your installed one. The best way is to copy the exact
320 command line which failed and run the test from the subdirectory for this
323 There are some failures which are not directly related to the GNU libc:
324 - Some compilers produce buggy code. No compiler gets single precision
325 complex numbers correct on Alpha. Otherwise, the egcs 1.1 release should be
326 ok; gcc 2.8.1 might cause some failures; gcc 2.7.2.x is so buggy that
327 explicit checks have been used so that you can't build with it.
328 - The kernel might have bugs. For example on Linux/Alpha 2.0.34 the
329 floating point handling has quite a number of bugs and therefore most of
330 the test cases in the math subdirectory will fail. Linux 2.2 has
331 fixes for the floating point support on Alpha. The Linux/SPARC kernel has
332 also some bugs in the FPU emulation code (as of Linux 2.2.0).
333 - Other tools might have problems. For example bash 2.03 gives a
334 segmentation fault running the tst-rpmatch.sh test script.
336 ?? What is symbol versioning good for? Do I need it?
338 {AJ} Symbol versioning solves problems that are related to interface
339 changes. One version of an interface might have been introduced in a
340 previous version of the GNU C library but the interface or the semantics of
341 the function has been changed in the meantime. For binary compatibility
342 with the old library, a newer library needs to still have the old interface
343 for old programs. On the other hand, new programs should use the new
344 interface. Symbol versioning is the solution for this problem. The GNU
345 libc version 2.1 uses symbol versioning by default if the installed binutils
348 We don't advise building without symbol versioning, since you lose binary
349 compatibility - forever! The binary compatibility you lose is not only
350 against the previous version of the GNU libc (version 2.0) but also against
353 ?? How can I compile on my fast ix86 machine a working libc for my slow
354 i386? After installing libc, programs abort with "Illegal
357 {AJ} glibc and gcc might generate some instructions on your machine that
358 aren't available on i386. You've got to tell glibc that you're configuring
359 for i386 with adding i386 as your machine, for example:
361 ../configure --prefix=/usr i386-pc-linux-gnu
363 And you need to tell gcc to only generate i386 code, just add `-mcpu=i386'
364 (just -m386 doesn't work) to your CFLAGS.
366 {UD} This applies not only to the i386. Compiling on a i686 for any older
367 model will also fail if the above methods are not used.
369 ?? `make' complains about a missing dlfcn/libdl.so when building
370 malloc/libmemprof.so. How can I fix this?
372 {AJ} Older make version (<= 3.78.90) have a bug which was hidden by a bug in
373 glibc (<= 2.1.2). You need to upgrade make to a newer or fixed version.
375 After upgrading make, you should remove the file sysd-sorted in your build
376 directory. The problem is that the broken make creates a wrong order for
377 one list in that file. The list has to be recreated with the new make -
378 which happens if you remove the file.
380 You might encounter this bug also in other situations where make scans
381 directories. I strongly advise to upgrade your make version to 3.79 or
385 ??mips Which tools should I use for MIPS?
387 {AJ} Either use as compiler egcs 1.1.2 (with Ralf Baechle's MIPS patches) or
388 the current development version of gcc 2.96 from CVS. gcc 2.95.x does not
389 work correctly on mips-linux.
391 You need recent binutils, anything before and including 2.10 will not work
392 correctly. Either try the Linux binutils 2.10.0.33 from HJ Lu or the
393 current development version from CVS of binutils.
395 For details check also my page <http://www.suse.de/~aj/glibc-mips.html>.
397 ? Installation and configuration issues
399 ?? Can I replace the libc on my Linux system with GNU libc?
401 {UD} You cannot replace any existing libc for Linux with GNU libc. It is
402 binary incompatible and therefore has a different major version. You can,
403 however, install it alongside your existing libc.
405 For Linux there are three major libc versions:
407 libc-5 original ELF libc
410 You can have any combination of these three installed. For more information
411 consult documentation for shared library handling. The Makefiles of GNU
412 libc will automatically generate the needed symbolic links which the linker
415 ?? How do I configure GNU libc so that the essential libraries
416 like libc.so go into /lib and the other into /usr/lib?
418 {UD,AJ} Like all other GNU packages GNU libc is designed to use a base
419 directory and install all files relative to this. The default is
420 /usr/local, because this is safe (it will not damage the system if installed
421 there). If you wish to install GNU libc as the primary C library on your
422 system, set the base directory to /usr (i.e. run configure --prefix=/usr
423 <other_options>). Note that this can damage your system; see ?safety for
426 Some systems like Linux have a filesystem standard which makes a difference
427 between essential libraries and others. Essential libraries are placed in
428 /lib because this directory is required to be located on the same disk
429 partition as /. The /usr subtree might be found on another
430 partition/disk. If you configure for Linux with --prefix=/usr, then this
431 will be done automatically.
433 To install the essential libraries which come with GNU libc in /lib on
434 systems other than Linux one must explicitly request it. Autoconf has no
435 option for this so you have to use a `configparms' file (see the `INSTALL'
436 file for details). It should contain:
441 The first line specifies the directory for the essential libraries, the
442 second line the directory for system configuration files.
444 ??safety How should I avoid damaging my system when I install GNU libc?
446 {ZW} If you wish to be cautious, do not configure with --prefix=/usr. If
447 you don't specify a prefix, glibc will be installed in /usr/local, where it
448 will probably not break anything. (If you wish to be certain, set the
449 prefix to something like /usr/local/glibc2 which is not used for anything.)
451 The dangers when installing glibc in /usr are twofold:
453 * glibc will overwrite the headers in /usr/include. Other C libraries
454 install a different but overlapping set of headers there, so the effect
455 will probably be that you can't compile anything. You need to rename
456 /usr/include out of the way before running `make install'. (Do not throw
457 it away; you will then lose the ability to compile programs against your
460 * None of your old libraries, static or shared, can be used with a
461 different C library major version. For shared libraries this is not a
462 problem, because the filenames are different and the dynamic linker
463 will enforce the restriction. But static libraries have no version
464 information. You have to evacuate all the static libraries in
465 /usr/lib to a safe location.
467 The situation is rather similar to the move from a.out to ELF which
468 long-time Linux users will remember.
470 ?? Do I need to use GNU CC to compile programs that will use the
473 {ZW} In theory, no; the linker does not care, and the headers are supposed
474 to check for GNU CC before using its extensions to the C language.
476 However, there are currently no ports of glibc to systems where another
477 compiler is the default, so no one has tested the headers extensively
478 against another compiler. You may therefore encounter difficulties. If you
479 do, please report them as bugs.
481 Also, in several places GNU extensions provide large benefits in code
482 quality. For example, the library has hand-optimized, inline assembly
483 versions of some string functions. These can only be used with GCC. See
486 ??crypt When linking with the new libc I get unresolved symbols
487 `crypt' and `setkey'. Why aren't these functions in the
491 Removed. Does not apply anymore.
493 ?? When I use GNU libc on my Linux system by linking against
494 the libc.so which comes with glibc all I get is a core dump.
496 {UD} On Linux, gcc sets the dynamic linker to /lib/ld-linux.so.1 unless the
497 user specifies a --dynamic-linker argument. This is the name of the libc5
498 dynamic linker, which does not work with glibc.
500 For casual use of GNU libc you can just specify to the linker
501 --dynamic-linker=/lib/ld-linux.so.2
503 which is the glibc dynamic linker, on Linux systems. On other systems the
504 name is /lib/ld.so.1. When linking via gcc, you've got to add
505 -Wl,--dynamic-linker=/lib/ld-linux.so.2
507 to the gcc command line.
509 To change your environment to use GNU libc for compiling you need to change
510 the `specs' file of your gcc. This file is normally found at
512 /usr/lib/gcc-lib/<arch>/<version>/specs
514 In this file you have to change a few things:
516 - change `ld-linux.so.1' to `ld-linux.so.2'
518 - remove all expression `%{...:-lgmon}'; there is no libgmon in glibc
520 - fix a minor bug by changing %{pipe:-} to %|
522 Here is what the gcc-2.7.2 specs file should look like when GNU libc is
525 -----------------------------------------------------------------------
527 %{V} %{v:%{!V:-V}} %{Qy:} %{!Qn:-Qy} %{n} %{T} %{Ym,*} %{Yd,*} %{Wa,*:%*}
533 %{fPIC:-D__PIC__ -D__pic__} %{fpic:-D__PIC__ -D__pic__} %{!m386:-D__i486__} %{posix:-D_POSIX_SOURCE} %{pthread:-D_REENTRANT}
542 %{!shared:crtend.o%s} %{shared:crtendS.o%s} crtn.o%s
545 -m elf_i386 %{shared:-shared} %{!shared: %{!ibcs: %{!static: %{rdynamic:-export-dynamic} %{!dynamic-linker:-dynamic-linker /lib/ld-linux.so.2}} %{static:-static}}}
548 %{!shared: %{pthread:-lpthread} %{profile:-lc_p} %{!profile: -lc}}
554 %{!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}
556 *switches_need_spaces:
560 %{funsigned-char:-D__CHAR_UNSIGNED__}
563 -D__ELF__ -Dunix -Di386 -Dlinux -Asystem(unix) -Asystem(posix) -Acpu(i386) -Amachine(i386)
571 -----------------------------------------------------------------------
573 Things get a bit more complicated if you have GNU libc installed in some
574 other place than /usr, i.e., if you do not want to use it instead of the old
575 libc. In this case the needed startup files and libraries are not found in
576 the regular places. So the specs file must tell the compiler and linker
579 Version 2.7.2.3 does and future versions of GCC will automatically
580 provide the correct specs.
582 ?? Looking through the shared libc file I haven't found the
583 functions `stat', `lstat', `fstat', and `mknod' and while
584 linking on my Linux system I get error messages. How is
585 this supposed to work?
587 {RM} Believe it or not, stat and lstat (and fstat, and mknod) are supposed
588 to be undefined references in libc.so.6! Your problem is probably a missing
589 or incorrect /usr/lib/libc.so file; note that this is a small text file now,
590 not a symlink to libc.so.6. It should look something like this:
592 GROUP ( libc.so.6 libc_nonshared.a )
594 ??excpt When I run an executable on one system which I compiled on
595 another, I get dynamic linker errors. Both systems have the same
596 version of glibc installed. What's wrong?
598 {ZW} Glibc on one of these systems was compiled with gcc 2.7 or 2.8, the
599 other with egcs (any version). Egcs has functions in its internal
600 `libgcc.a' to support exception handling with C++. They are linked into
601 any program or dynamic library compiled with egcs, whether it needs them or
602 not. Dynamic libraries then turn around and export those functions again
603 unless special steps are taken to prevent them.
605 When you link your program, it resolves its references to the exception
606 functions to the ones exported accidentally by libc.so. That works fine as
607 long as libc has those functions. On the other system, libc doesn't have
608 those functions because it was compiled by gcc 2.8, and you get undefined
609 symbol errors. The symbols in question are named things like
610 `__register_frame_info'.
612 For glibc 2.0, the workaround is to not compile libc with egcs. We've also
613 incorporated a patch which should prevent the EH functions sneaking into
614 libc. It doesn't matter what compiler you use to compile your program.
616 For glibc 2.1, we've chosen to do it the other way around: libc.so
617 explicitly provides the EH functions. This is to prevent other shared
618 libraries from doing it.
620 {UD} Starting with glibc 2.1.1 you can compile glibc with gcc 2.8.1 or
621 newer since we have explicitly add references to the functions causing the
622 problem. But you nevertheless should use EGCS for other reasons
625 {GK} On some Linux distributions for PowerPC, you can see this when you have
626 built gcc or egcs from the Web sources (gcc versions 2.95 or earlier), then
627 re-built glibc. This happens because in these versions of gcc, exception
628 handling is implemented using an older method; the people making the
629 distributions are a little ahead of their time.
631 A quick solution to this is to find the libgcc.a file that came with the
632 distribution (it would have been installed under /usr/lib/gcc-lib), do
633 `ar x libgcc.a frame.o' to get the frame.o file out, and add a line saying
634 `LDLIBS-c.so += frame.o' to the file `configparms' in the directory you're
635 building in. You can check you've got the right `frame.o' file by running
636 `nm frame.o' and checking that it has the symbols defined that you're
639 This will let you build glibc with the C compiler. The C++ compiler
640 will still be binary incompatible with any C++ shared libraries that
641 you got with your distribution.
643 ?? How can I compile gcc 2.7.2.1 from the gcc source code using
646 {AJ} There's only correct support for glibc 2.0.x in gcc 2.7.2.3 or later.
647 But you should get at least gcc 2.8.1 or egcs 1.1 (or later versions)
650 ?? The `gencat' utility cannot process the catalog sources which
651 were used on my Linux libc5 based system. Why?
653 {UD} The `gencat' utility provided with glibc complies to the XPG standard.
654 The older Linux version did not obey the standard, so they are not
657 To ease the transition from the Linux version some of the non-standard
658 features are also present in the `gencat' program of GNU libc. This mainly
659 includes the use of symbols for the message number and the automatic
660 generation of header files which contain the needed #defines to map the
663 Here is a simple SED script to convert at least some Linux specific catalog
664 files to the XPG4 form:
666 -----------------------------------------------------------------------
667 # Change catalog source in Linux specific format to standard XPG format.
668 # Ulrich Drepper <drepper@cygnus.com>, 1996.
672 s/\$ #\([^ ]*\).*/\1/
674 s/\$ #[^ ]* *\(.*\)/\$ \1/
680 s/\(.*\)\n\(.*\)/\2 \1/
682 -----------------------------------------------------------------------
684 ?? Programs using libc have their messages translated, but other
685 behavior is not localized (e.g. collating order); why?
687 {ZW} Translated messages are automatically installed, but the locale
688 database that controls other behaviors is not. You need to run localedef to
689 install this database, after you have run `make install'. For example, to
690 set up the French Canadian locale, simply issue the command
692 localedef -i fr_CA -f ISO-8859-1 fr_CA
694 Please see localedata/README in the source tree for further details.
696 ?? I have set up /etc/nis.conf, and the Linux libc 5 with NYS
697 works great. But the glibc NIS+ doesn't seem to work.
699 {TK} The glibc NIS+ implementation uses a /var/nis/NIS_COLD_START file for
700 storing information about the NIS+ server and their public keys, because the
701 nis.conf file does not contain all the necessary information. You have to
702 copy a NIS_COLD_START file from a Solaris client (the NIS_COLD_START file is
703 byte order independent) or generate it with nisinit from the nis-tools
704 package; available at
706 http://www.suse.de/~kukuk/linux/nisplus.html
708 ?? I have killed ypbind to stop using NIS, but glibc
711 {TK} For faster NIS lookups, glibc uses the /var/yp/binding/ files from
712 ypbind. ypbind 3.3 and older versions don't always remove these files, so
713 glibc will continue to use them. Other BSD versions seem to work correctly.
714 Until ypbind 3.4 is released, you can find a patch at
716 <ftp://ftp.kernel.org/pub/linux/utils/net/NIS/ypbind-3.3-glibc4.diff.gz>
718 ?? Under Linux/Alpha, I always get "do_ypcall: clnt_call:
719 RPC: Unable to receive; errno = Connection refused" when using NIS.
721 {TK} You need a ypbind version which is 64bit clean. Some versions are not
722 64bit clean. A 64bit clean implementation is ypbind-mt. For ypbind 3.3,
723 you need the patch from ftp.kernel.org (See the previous question). I don't
724 know about other versions.
727 ?? After installing glibc name resolving doesn't work properly.
729 {AJ} You probably should read the manual section describing nsswitch.conf
730 (just type `info libc "NSS Configuration File"'). The NSS configuration
731 file is usually the culprit.
734 ?? How do I create the databases for NSS?
736 {AJ} If you have an entry "db" in /etc/nsswitch.conf you should also create
737 the database files. The glibc sources contain a Makefile which does the
738 necessary conversion and calls to create those files. The file is
739 `db-Makefile' in the subdirectory `nss' and you can call it with `make -f
740 db-Makefile'. Please note that not all services are capable of using a
741 database. Currently passwd, group, ethers, protocol, rpc, services shadow
742 and netgroup are implemented. See also question ?nssdb.
744 ?? I have /usr/include/net and /usr/include/scsi as symlinks
745 into my Linux source tree. Is that wrong?
747 {PB} This was necessary for libc5, but is not correct when using glibc.
748 Including the kernel header files directly in user programs usually does not
749 work (see ?kerhdr). glibc provides its own <net/*> and <scsi/*> header
750 files to replace them, and you may have to remove any symlink that you have
751 in place before you install glibc. However, /usr/include/asm and
752 /usr/include/linux should remain as they were.
754 ?? Programs like `logname', `top', `uptime' `users', `w' and
755 `who', show incorrect information about the (number of)
756 users on my system. Why?
760 ?? After upgrading to glibc 2.1 with symbol versioning I get
761 errors about undefined symbols. What went wrong?
763 {AJ} The problem is caused either by wrong program code or tools. In the
764 versioned libc a lot of symbols are now local that were global symbols in
765 previous versions. It seems that programs linked against older versions
766 often accidentally used libc global variables -- something that should not
769 The only way to fix this is to recompile your program. Sorry, that's the
770 price you might have to pay once for quite a number of advantages with
773 ?? When I start the program XXX after upgrading the library
775 XXX: Symbol `_sys_errlist' has different size in shared
776 object, consider re-linking
777 Why? What should I do?
779 {UD} As the message says, relink the binary. The problem is that a few
780 symbols from the library can change in size and there is no way to avoid
781 this. _sys_errlist is a good example. Occasionally there are new error
782 numbers added to the kernel and this must be reflected at user level,
783 breaking programs that refer to them directly.
785 Such symbols should normally not be used at all. There are mechanisms to
786 avoid using them. In the case of _sys_errlist, there is the strerror()
787 function which should _always_ be used instead. So the correct fix is to
788 rewrite that part of the application.
790 In some situations (especially when testing a new library release) it might
791 be possible that a symbol changed size when that should not have happened.
792 So in case of doubt report such a warning message as a problem.
794 ?? What do I need for C++ development?
796 {HJ,AJ} You need either egcs 1.1 which comes directly with libstdc++ or
797 gcc-2.8.1 together with libstdc++ 2.8.1.1. egcs 1.1 has the better C++
798 support and works directly with glibc 2.1. If you use gcc-2.8.1 with
799 libstdc++ 2.8.1.1, you need to modify libstdc++ a bit. A patch is available
801 <ftp://alpha.gnu.org/gnu/libstdc++-2.8.1.1-glibc2.1-diff.gz>
803 Please note that libg++ 2.7.2 (and the Linux Versions 2.7.2.x) doesn't work
804 very well with the GNU C library due to vtable thunks. If you're upgrading
805 from glibc 2.0.x to 2.1 you have to recompile libstdc++ since the library
806 compiled for 2.0 is not compatible due to the new Large File Support (LFS)
809 {UD} But since in the case of a shared libstdc++ the version numbers should
810 be different existing programs will continue to work.
812 ?? Even statically linked programs need some shared libraries
813 which is not acceptable for me. What can I do?
815 {AJ} NSS (for details just type `info libc "Name Service Switch"') won't
816 work properly without shared libraries. NSS allows using different services
817 (e.g. NIS, files, db, hesiod) by just changing one configuration file
818 (/etc/nsswitch.conf) without relinking any programs. The only disadvantage
819 is that now static libraries need to access shared libraries. This is
820 handled transparently by the GNU C library.
822 A solution is to configure glibc with --enable-static-nss. In this case you
823 can create a static binary that will use only the services dns and files
824 (change /etc/nsswitch.conf for this). You need to link explicitly against
825 all these services. For example:
827 gcc -static test-netdb.c -o test-netdb \
828 -Wl,--start-group -lc -lnss_files -lnss_dns -lresolv -Wl,--end-group
830 The problem with this approach is that you've got to link every static
831 program that uses NSS routines with all those libraries.
833 {UD} In fact, one cannot say anymore that a libc compiled with this
834 option is using NSS. There is no switch anymore. Therefore it is
835 *highly* recommended *not* to use --enable-static-nss since this makes
836 the behaviour of the programs on the system inconsistent.
838 ?? I just upgraded my Linux system to glibc and now I get
839 errors whenever I try to link any program.
841 {ZW} This happens when you have installed glibc as the primary C library but
842 have stray symbolic links pointing at your old C library. If the first
843 `libc.so' the linker finds is libc 5, it will use that. Your program
844 expects to be linked with glibc, so the link fails.
846 The most common case is that glibc put its `libc.so' in /usr/lib, but there
847 was a `libc.so' from libc 5 in /lib, which gets searched first. To fix the
848 problem, just delete /lib/libc.so. You may also need to delete other
849 symbolic links in /lib, such as /lib/libm.so if it points to libm.so.5.
851 {AJ} The perl script test-installation.pl which is run as last step during
852 an installation of glibc that is configured with --prefix=/usr should help
853 detect these situations. If the script reports problems, something is
856 ?? When I use nscd the machine freezes.
858 {UD} You cannot use nscd with Linux 2.0.*. There is functionality missing
859 in the kernel and work-arounds are not suitable. Besides, some parts of the
860 kernel are too buggy when it comes to using threads.
862 If you need nscd, you have to use at least a 2.1 kernel.
864 Note that I have at this point no information about any other platform.
866 ?? I need lots of open files. What do I have to do?
868 {AJ} This is at first a kernel issue. The kernel defines limits with
869 OPEN_MAX the number of simultaneous open files and with FD_SETSIZE the
870 number of used file descriptors. You need to change these values in your
871 kernel and recompile the kernel so that the kernel allows more open
872 files. You don't necessarily need to recompile the GNU C library since the
873 only place where OPEN_MAX and FD_SETSIZE is really needed in the library
874 itself is the size of fd_set which is used by select.
876 The GNU C library is now select free. This means it internally has no
877 limits imposed by the `fd_set' type. Instead all places where the
878 functionality is needed the `poll' function is used.
880 If you increase the number of file descriptors in the kernel you don't need
881 to recompile the C library.
883 {UD} You can always get the maximum number of file descriptors a process is
884 allowed to have open at any time using
886 number = sysconf (_SC_OPEN_MAX);
888 This will work even if the kernel limits change.
890 ?? How do I get the same behavior on parsing /etc/passwd and
891 /etc/group as I have with libc5 ?
893 {TK} The name switch setup in /etc/nsswitch.conf selected by most Linux
894 distributions does not support +/- and netgroup entries in the files like
895 /etc/passwd. Though this is the preferred setup some people might have
896 setups coming over from the libc5 days where it was the default to recognize
897 lines like this. To get back to the old behaviour one simply has to change
898 the rules for passwd, group, and shadow in the nsswitch.conf file as
909 ??libs What needs to be recompiled when upgrading from glibc 2.0 to glibc
912 {AJ,CG} If you just upgrade the glibc from 2.0.x (x <= 7) to 2.1, binaries
913 that have been linked against glibc 2.0 will continue to work.
915 If you compile your own binaries against glibc 2.1, you also need to
916 recompile some other libraries. The problem is that libio had to be changed
917 and therefore libraries that are based or depend on the libio of glibc,
918 e.g. ncurses, slang and most C++ libraries, need to be recompiled. If you
919 experience strange segmentation faults in your programs linked against glibc
920 2.1, you might need to recompile your libraries.
922 Another problem is that older binaries that were linked statically against
923 glibc 2.0 will reference the older nss modules (libnss_files.so.1 instead of
924 libnss_files.so.2), so don't remove them. Also, the old glibc-2.0 compiled
925 static libraries (libfoo.a) which happen to depend on the older libio
926 behavior will be broken by the glibc 2.1 upgrade. We plan to produce a
927 compatibility library that people will be able to link in if they want
928 to compile a static library generated against glibc 2.0 into a program
929 on a glibc 2.1 system. You just add -lcompat and you should be fine.
931 The glibc-compat add-on will provide the libcompat.a library, the older
932 nss modules, and a few other files. Together, they should make it
933 possible to do development with old static libraries on a glibc 2.1
934 system. This add-on is still in development. You can get it from
935 <ftp://alpha.gnu.org/gnu/glibc/glibc-compat-2.1.tar.gz>
936 but please keep in mind that it is experimental.
938 ?? Why is extracting files via tar so slow?
940 {AJ} Extracting of tar archives might be quite slow since tar has to look up
941 userid and groupids and doesn't cache negative results. If you have nis or
942 nisplus in your /etc/nsswitch.conf for the passwd and/or group database,
943 each file extractions needs a network connection. There are two possible
946 - do you really need NIS/NIS+ (some Linux distributions add by default
947 nis/nisplus even if it's not needed)? If not, just remove the entries.
949 - if you need NIS/NIS+, use the Name Service Cache Daemon nscd that comes
952 ?? Compiling programs I get parse errors in libio.h (e.g. "parse error
953 before `_IO_seekoff'"). How should I fix this?
955 {AJ} You might get the following errors when upgrading to glibc 2.1:
957 In file included from /usr/include/stdio.h:57,
959 /usr/include/libio.h:335: parse error before `_IO_seekoff'
960 /usr/include/libio.h:335: parse error before `_G_off64_t'
961 /usr/include/libio.h:336: parse error before `_IO_seekpos'
962 /usr/include/libio.h:336: parse error before `_G_fpos64_t'
964 The problem is a wrong _G_config.h file in your include path. The
965 _G_config.h file that comes with glibc 2.1 should be used and not one from
966 libc5 or from a compiler directory. To check which _G_config.h file the
967 compiler uses, compile your program with `gcc -E ...|grep G_config.h' and
968 remove that file. Your compiler should pick up the file that has been
969 installed by glibc 2.1 in your include directory.
971 ?? After upgrading to glibc 2.1, libraries that were compiled against
972 glibc 2.0.x don't work anymore.
976 ??nssdb What happened to the Berkeley DB libraries? Can I still use db
977 in /etc/nsswitch.conf?
979 {AJ} Due to too many incompatible changes in disk layout and API of Berkeley
980 DB and a too tight coupling of libc and libdb, the db library has been
981 removed completely from glibc 2.2. The only place that really used the
982 Berkeley DB was the NSS db module.
984 The NSS db module has been rewritten to support a number of different
985 versions of Berkeley DB for the NSS db module. Currently the releases 2.x
986 and 3.x of Berkeley DB are supported. The older db 1.85 library is not
987 supported. You can use the version from glibc 2.1.x or download a version
988 from Sleepycat Software (http://www.sleepycat.com). The library has to be
989 compiled as shared library and installed in the system lib directory
990 (normally /lib). The library needs to have a special soname to be found by
993 If public structures change in a new Berkeley db release, this needs to be
996 Currently the code searches for libraries with a soname of "libdb.so.3"
997 (that's the name from db 2.4.14 which comes with glibc 2.1.x) and
998 "libdb-3.0.so" (the name used by db 3.0.55 as default).
1001 ? Source and binary incompatibilities, and what to do about them
1003 ?? I expect GNU libc to be 100% source code compatible with
1004 the old Linux based GNU libc. Why isn't it like this?
1006 {DMT,UD} Not every extension in Linux libc's history was well thought-out.
1007 In fact it had a lot of problems with standards compliance and with
1008 cleanliness. With the introduction of a new version number these errors can
1009 now be corrected. Here is a list of the known source code
1012 * _GNU_SOURCE: glibc does not make the GNU extensions available
1013 automatically. If a program depends on GNU extensions or some
1014 other non-standard functionality, it is necessary to compile it
1015 with the C compiler option -D_GNU_SOURCE, or better, to put
1016 `#define _GNU_SOURCE' at the beginning of your source files, before
1017 any C library header files are included. This difference normally
1018 manifests itself in the form of missing prototypes and/or data type
1019 definitions. Thus, if you get such errors, the first thing you
1020 should do is try defining _GNU_SOURCE and see if that makes the
1023 For more information consult the file `NOTES' in the GNU C library
1026 * reboot(): GNU libc sanitizes the interface of reboot() to be more
1027 compatible with the interface used on other OSes. reboot() as
1028 implemented in glibc takes just one argument. This argument
1029 corresponds to the third argument of the Linux reboot system call.
1030 That is, a call of the form reboot(a, b, c) needs to be changed into
1031 reboot(c). Beside this the header <sys/reboot.h> defines the needed
1032 constants for the argument. These RB_* constants should be used
1033 instead of the cryptic magic numbers.
1035 * swapon(): the interface of this function didn't change, but the
1036 prototype is in a separate header file <sys/swap.h>. This header
1037 file also provides the SWAP_* constants defined by <linux/swap.h>;
1038 you should use them for the second argument to swapon().
1040 * errno: If a program uses the variable "errno", then it _must_
1041 include <errno.h>. The old libc often (erroneously) declared this
1042 variable implicitly as a side-effect of including other libc header
1043 files. glibc is careful to avoid such namespace pollution, which,
1044 in turn, means that you really need to include the header files that
1045 you depend on. This difference normally manifests itself in the
1046 form of the compiler complaining about references to an undeclared
1049 * Linux-specific syscalls: All Linux system calls now have appropriate
1050 library wrappers and corresponding declarations in various header files.
1051 This is because the syscall() macro that was traditionally used to
1052 work around missing syscall wrappers are inherently non-portable and
1053 error-prone. The following table lists all the new syscall stubs,
1054 the header-file declaring their interface and the system call name.
1056 syscall name: wrapper name: declaring header file:
1057 ------------- ------------- ----------------------
1058 bdflush bdflush <sys/kdaemon.h>
1059 syslog ksyslog_ctl <sys/klog.h>
1061 * lpd: Older versions of lpd depend on a routine called _validuser().
1062 The library does not provide this function, but instead provides
1063 __ivaliduser() which has a slightly different interface. Simply
1064 upgrading to a newer lpd should fix this problem (e.g., the 4.4BSD
1065 lpd is known to be working).
1067 * resolver functions/BIND: like on many other systems the functions of
1068 the resolver library are not included in libc itself. There is a
1069 separate library libresolv. If you get undefined symbol errors for
1070 symbols starting with `res_*' simply add -lresolv to your linker
1073 * the `signal' function's behavior corresponds to the BSD semantic and
1074 not the SysV semantic as it was in libc-5. The interface on all GNU
1075 systems shall be the same and BSD is the semantic of choice. To use
1076 the SysV behavior simply use `sysv_signal', or define _XOPEN_SOURCE.
1077 See ?signal for details.
1079 ??getlog Why does getlogin() always return NULL on my Linux box?
1081 {UD} The GNU C library has a format for the UTMP and WTMP file which differs
1082 from what your system currently has. It was extended to fulfill the needs
1083 of the next years when IPv6 is introduced. The record size is different and
1084 some fields have different positions. The files written by functions from
1085 the one library cannot be read by functions from the other library. Sorry,
1086 but this is what a major release is for. It's better to have a cut now than
1087 having no means to support the new techniques later.
1089 ?? Where are the DST_* constants found in <sys/time.h> on many
1092 {UD} These constants come from the old BSD days and are not used anymore
1093 (libc5 does not actually implement the handling although the constants are
1096 Instead GNU libc contains zone database support and compatibility code for
1097 POSIX TZ environment variable handling. For former is very much preferred
1100 ?? The prototypes for `connect', `accept', `getsockopt',
1101 `setsockopt', `getsockname', `getpeername', `send',
1102 `sendto', and `recvfrom' are different in GNU libc from
1103 any other system I saw. This is a bug, isn't it?
1105 {UD} No, this is no bug. This version of GNU libc already follows the new
1106 Single Unix specifications (and I think the POSIX.1g draft which adopted the
1107 solution). The type for a parameter describing a size is now `socklen_t', a
1110 ??kerhdr On Linux I've got problems with the declarations in Linux
1113 {UD,AJ} On Linux, the use of kernel headers is reduced to the minimum. This
1114 gives Linus the ability to change the headers more freely. Also, user
1115 programs are now insulated from changes in the size of kernel data
1118 For example, the sigset_t type is 32 or 64 bits wide in the kernel. In
1119 glibc it is 1024 bits wide. This guarantees that when the kernel gets a
1120 bigger sigset_t (for POSIX.1e realtime support, say) user programs will not
1121 have to be recompiled. Consult the header files for more information about
1124 Therefore you shouldn't include Linux kernel header files directly if glibc
1125 has defined a replacement. Otherwise you might get undefined results because
1128 ?? I don't include any kernel headers myself but the compiler
1129 still complains about redeclarations of types in the kernel
1132 {UD} The kernel headers before Linux 2.1.61 and 2.0.32 don't work correctly
1133 with glibc. Compiling C programs is possible in most cases but C++ programs
1134 have (due to the change of the name lookups for `struct's) problems. One
1135 prominent example is `struct fd_set'.
1137 There might be some problems left but 2.1.61/2.0.32 fix most of the known
1138 ones. See the BUGS file for other known problems.
1140 ??signal Why don't signals interrupt system calls anymore?
1142 {ZW} By default GNU libc uses the BSD semantics for signal(), unlike Linux
1143 libc 5 which used System V semantics. This is partially for compatibility
1144 with other systems and partially because the BSD semantics tend to make
1145 programming with signals easier.
1147 There are three differences:
1149 * BSD-style signals that occur in the middle of a system call do not
1150 affect the system call; System V signals cause the system call to
1151 fail and set errno to EINTR.
1153 * BSD signal handlers remain installed once triggered. System V signal
1154 handlers work only once, so one must reinstall them each time.
1156 * A BSD signal is blocked during the execution of its handler. In other
1157 words, a handler for SIGCHLD (for example) does not need to worry about
1158 being interrupted by another SIGCHLD. It may, however, be interrupted
1161 There is general consensus that for `casual' programming with signals, the
1162 BSD semantics are preferable. You don't need to worry about system calls
1163 returning EINTR, and you don't need to worry about the race conditions
1164 associated with one-shot signal handlers.
1166 If you are porting an old program that relies on the old semantics, you can
1167 quickly fix the problem by changing signal() to sysv_signal() throughout.
1168 Alternatively, define _XOPEN_SOURCE before including <signal.h>.
1170 For new programs, the sigaction() function allows you to specify precisely
1171 how you want your signals to behave. All three differences listed above are
1172 individually switchable on a per-signal basis with this function.
1174 If all you want is for one specific signal to cause system calls to fail and
1175 return EINTR (for example, to implement a timeout) you can do this with
1179 ??string I've got errors compiling code that uses certain string
1182 {AJ} glibc 2.1 has special string functions that are faster than the normal
1183 library functions. Some of the functions are additionally implemented as
1184 inline functions and others as macros. This might lead to problems with
1185 existing codes but it is explicitly allowed by ISO C.
1187 The optimized string functions are only used when compiling with
1188 optimizations (-O1 or higher). The behavior can be changed with two feature
1191 * __NO_STRING_INLINES: Don't do any string optimizations.
1192 * __USE_STRING_INLINES: Use assembly language inline functions (might
1193 increase code size dramatically).
1195 Since some of these string functions are now additionally defined as macros,
1196 code like "char *strncpy();" doesn't work anymore (and is unnecessary, since
1197 <string.h> has the necessary declarations). Either change your code or
1198 define __NO_STRING_INLINES.
1200 {UD} Another problem in this area is that gcc still has problems on machines
1201 with very few registers (e.g., ix86). The inline assembler code can require
1202 almost all the registers and the register allocator cannot always handle
1205 One can disable the string optimizations selectively. Instead of writing
1207 cp = strcpy (foo, "lkj");
1211 cp = (strcpy) (foo, "lkj");
1213 This disables the optimization for that specific call.
1215 ?? I get compiler messages "Initializer element not constant" with
1216 stdin/stdout/stderr. Why?
1218 {RM,AJ} Constructs like:
1219 static FILE *InPtr = stdin;
1221 lead to this message. This is correct behaviour with glibc since stdin is
1222 not a constant expression. Please note that a strict reading of ISO C does
1223 not allow above constructs.
1225 One of the advantages of this is that you can assign to stdin, stdout, and
1226 stderr just like any other global variable (e.g. `stdout = my_stream;'),
1227 which can be very useful with custom streams that you can write with libio
1228 (but beware this is not necessarily portable). The reason to implement it
1229 this way were versioning problems with the size of the FILE structure.
1231 To fix those programs you've got to initialize the variable at run time.
1232 This can be done, e.g. in main, like:
1240 or by constructors (beware this is gcc specific):
1243 static void inPtr_construct (void) __attribute__((constructor));
1244 static void inPtr_construct (void) { InPtr = stdin; }
1247 ?? I can't compile with gcc -traditional (or
1248 -traditional-cpp). Why?
1250 {AJ} glibc2 does break -traditional and -traditonal-cpp - and will continue
1251 to do so. For example constructs of the form:
1257 are useful for debugging purposes (you can use foo with your debugger that's
1258 why we need the enum) and for compatibility (other systems use defines and
1261 ?? I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible?
1263 {AJ} The GNU C library is compatible with the ANSI/ISO C standard. If
1264 you're using `gcc -ansi', the glibc includes which are specified in the
1265 standard follow the standard. The ANSI/ISO C standard defines what has to be
1266 in the include files - and also states that nothing else should be in the
1267 include files (btw. you can still enable additional standards with feature
1270 The GNU C library is conforming to ANSI/ISO C - if and only if you're only
1271 using the headers and library functions defined in the standard.
1273 ?? I can't access some functions anymore. nm shows that they do
1274 exist but linking fails nevertheless.
1276 {AJ} With the introduction of versioning in glibc 2.1 it is possible to
1277 export only those identifiers (functions, variables) that are really needed
1278 by application programs and by other parts of glibc. This way a lot of
1279 internal interfaces are now hidden. nm will still show those identifiers
1280 but marking them as internal. ISO C states that identifiers beginning with
1281 an underscore are internal to the libc. An application program normally
1282 shouldn't use those internal interfaces (there are exceptions,
1283 e.g. __ivaliduser). If a program uses these interfaces, it's broken. These
1284 internal interfaces might change between glibc releases or dropped
1287 ?? When using the db-2 library which comes with glibc is used in
1288 the Perl db modules the testsuite is not passed. This did not
1289 happen with db-1, gdbm, or ndbm.
1291 Removed. Does not apply anymore.
1293 ?? The pow() inline function I get when including <math.h> is broken.
1294 I get segmentation faults when I run the program.
1296 {UD} Nope, the implementation is correct. The problem is with egcs version
1297 prior to 1.1. I.e., egcs 1.0 to 1.0.3 are all broken (at least on Intel).
1298 If you have to use this compiler you must define __NO_MATH_INLINES before
1299 including <math.h> to prevent the inline functions from being used. egcs 1.1
1300 fixes the problem. I don't know about gcc 2.8 and 2.8.1.
1302 ?? The sys/sem.h file lacks the definition of `union semun'.
1304 {UD} Nope. This union has to be provided by the user program. Former glibc
1305 versions defined this but it was an error since it does not make much sense
1306 when thinking about it. The standards describing the System V IPC functions
1307 define it this way and therefore programs must be adopted.
1309 ?? Why has <netinet/ip_fw.h> disappeared?
1311 {AJ} The corresponding Linux kernel data structures and constants are
1312 totally different in Linux 2.0 and Linux 2.2. This situation has to be
1313 taken care in user programs using the firewall structures and therefore
1314 those programs (ipfw is AFAIK the only one) should deal with this problem
1317 ?? I get floods of warnings when I use -Wconversion and include
1318 <string.h> or <math.h>.
1320 {ZW} <string.h> and <math.h> intentionally use prototypes to override
1321 argument promotion. -Wconversion warns about all these. You can safely
1322 ignore the warnings.
1324 -Wconversion isn't really intended for production use, only for shakedown
1325 compiles after converting an old program to standard C.
1328 ?? After upgrading to glibc 2.1, I receive errors about
1329 unresolved symbols, like `_dl_initial_searchlist' and can not
1330 execute any binaries. What went wrong?
1332 {AJ} This normally happens if your libc and ld (dynamic linker) are from
1333 different releases of glibc. For example, the dynamic linker
1334 /lib/ld-linux.so.2 comes from glibc 2.0.x, but the version of libc.so.6 is
1337 The path /lib/ld-linux.so.2 is hardcoded in every glibc2 binary but
1338 libc.so.6 is searched via /etc/ld.so.cache and in some special directories
1339 like /lib and /usr/lib. If you run configure with another prefix than /usr
1340 and put this prefix before /lib in /etc/ld.so.conf, your system will break.
1342 So what can you do? Either of the following should work:
1344 * Run `configure' with the same prefix argument you've used for glibc 2.0.x
1345 so that the same paths are used.
1346 * Replace /lib/ld-linux.so.2 with a link to the dynamic linker from glibc
1349 You can even call the dynamic linker by hand if everything fails. You've
1350 got to set LD_LIBRARY_PATH so that the corresponding libc is found and also
1351 need to provide an absolute path to your binary:
1353 LD_LIBRARY_PATH=<path-where-libc.so.6-lives> \
1354 <path-where-corresponding-dynamic-linker-lives>/ld-linux.so.2 \
1355 <path-to-binary>/binary
1357 For example `LD_LIBRARY_PATH=/libold /libold/ld-linux.so.2 /bin/mv ...'
1358 might be useful in fixing a broken system (if /libold contains dynamic
1359 linker and corresponding libc).
1361 With that command line no path is used. To further debug problems with the
1362 dynamic linker, use the LD_DEBUG environment variable, e.g.
1363 `LD_DEBUG=help echo' for the help text.
1365 If you just want to test this release, don't put the lib directory in
1366 /etc/ld.so.conf. You can call programs directly with full paths (as above).
1367 When compiling new programs against glibc 2.1, you've got to specify the
1368 correct paths to the compiler (option -I with gcc) and linker (options
1369 --dynamic-linker, -L and --rpath).
1371 ?? bonnie reports that char i/o with glibc 2 is much slower than with
1372 libc5. What can be done?
1374 {AJ} The GNU C library uses thread safe functions by default and libc5 used
1375 non thread safe versions. The non thread safe functions have in glibc the
1376 suffix `_unlocked', for details check <stdio.h>. Using `putc_unlocked' etc.
1377 instead of `putc' should give nearly the same speed with bonnie (bonnie is a
1378 benchmark program for measuring disk access).
1380 ?? Programs compiled with glibc 2.1 can't read db files made with glibc
1381 2.0. What has changed that programs like rpm break?
1383 Removed. Does not apply anymore.
1385 ?? Autoconf's AC_CHECK_FUNC macro reports that a function exists, but
1386 when I try to use it, it always returns -1 and sets errno to ENOSYS.
1388 {ZW} You are using a 2.0 Linux kernel, and the function you are trying to
1389 use is only implemented in 2.1/2.2. Libc considers this to be a function
1390 which exists, because if you upgrade to a 2.2 kernel, it will work. One
1391 such function is sigaltstack.
1393 Your program should check at runtime whether the function works, and
1394 implement a fallback. Note that Autoconf cannot detect unimplemented
1395 functions in other systems' C libraries, so you need to do this anyway.
1397 ?? My program segfaults when I call fclose() on the FILE* returned
1398 from setmntent(). Is this a glibc bug?
1400 {GK} No. Don't do this. Use endmntent(), that's what it's for.
1402 In general, you should use the correct deallocation routine. For instance,
1403 if you open a file using fopen(), you should deallocate the FILE * using
1404 fclose(), not free(), even though the FILE * is also a pointer.
1406 In the case of setmntent(), it may appear to work in most cases, but it
1407 won't always work. Unfortunately, for compatibility reasons, we can't
1408 change the return type of setmntent() to something other than FILE *.
1413 ?? After I changed configure.in I get `Autoconf version X.Y.
1414 or higher is required for this script'. What can I do?
1416 {UD} You have to get the specified autoconf version (or a later one)
1417 from your favorite mirror of ftp.gnu.org.
1419 ?? When I try to compile code which uses IPv6 headers and
1420 definitions on my Linux 2.x.y system I am in trouble.
1421 Nothing seems to work.
1423 {UD} The problem is that IPv6 development still has not reached a point
1424 where the headers are stable. There are still lots of incompatible changes
1425 made and the libc headers have to follow.
1427 {PB} The 2.1 release of GNU libc aims to comply with the current versions of
1428 all the relevant standards. The IPv6 support libraries for older Linux
1429 systems used a different naming convention and so code written to work with
1430 them may need to be modified. If the standards make incompatible changes in
1431 the future then the libc may need to change again.
1433 IPv6 will not work with a 2.0.x kernel. When kernel 2.2 is released it
1434 should contain all the necessary support; until then you should use the
1435 latest 2.1.x release you can find. As of 98/11/26 the currently recommended
1436 kernel for IPv6 is 2.1.129.
1438 Also, as of the 2.1 release the IPv6 API provided by GNU libc is not
1441 ??tzdb When I set the timezone by setting the TZ environment variable
1442 to EST5EDT things go wrong since glibc computes the wrong time
1443 from this information.
1445 {UD} The problem is that people still use the braindamaged POSIX method to
1446 select the timezone using the TZ environment variable with a format EST5EDT
1447 or whatever. People, if you insist on using TZ instead of the timezone
1448 database (see below), read the POSIX standard, the implemented behaviour is
1449 correct! What you see is in fact the result of the decisions made while
1450 POSIX.1 was created. We've only implemented the handling of TZ this way to
1451 be POSIX compliant. It is not really meant to be used.
1453 The alternative approach to handle timezones which is implemented is the
1454 correct one to use: use the timezone database. This avoids all the problems
1455 the POSIX method has plus it is much easier to use. Simply run the tzselect
1456 shell script, answer the question and use the name printed in the end by
1457 making a symlink /etc/localtime pointing to /usr/share/zoneinfo/NAME (NAME
1458 is the returned value from tzselect). That's all. You never again have to
1461 So, please avoid sending bug reports about time related problems if you use
1462 the POSIX method and you have not verified something is really broken by
1463 reading the POSIX standards.
1465 ?? What other sources of documentation about glibc are available?
1467 {AJ} The FSF has a page about the GNU C library at
1468 <http://www.gnu.org/software/libc/>. The problem data base of open and
1469 solved bugs in GNU libc is available at
1470 <http://www-gnats.gnu.org:8080/cgi-bin/wwwgnats.pl>. Eric Green has written
1471 a HowTo for converting from Linux libc5 to glibc2. The HowTo is accessible
1472 via the FSF page and at <http://www.imaxx.net/~thrytis/glibc>. Frodo
1473 Looijaard describes a different way installing glibc2 as secondary libc at
1474 <http://huizen.dds.nl/~frodol/glibc>.
1476 Please note that this is not a complete list.
1478 ?? The timezone string for Sydney/Australia is wrong since even when
1479 daylight saving time is in effect the timezone string is EST.
1481 {UD} The problem for some timezones is that the local authorities decided
1482 to use the term "summer time" instead of "daylight saving time". In this
1483 case the abbreviation character `S' is the same as the standard one. So,
1486 Eastern Standard Time = EST
1487 Eastern Summer Time = EST
1489 Great! To get this bug fixed convince the authorities to change the laws
1490 and regulations of the country this effects. glibc behaves correctly.
1492 ??make I've build make 3.77 against glibc 2.1 and now make gets
1493 segmentation faults.
1495 Removed. Does not apply anymore, use make 3.79 or newer.
1497 ?? Why do so many programs using math functions fail on my AlphaStation?
1499 {AO} The functions floor() and floorf() use an instruction that is not
1500 implemented in some old PALcodes of AlphaStations. This may cause
1501 `Illegal Instruction' core dumps or endless loops in programs that
1502 catch these signals. Updating the firmware to a 1999 release has
1503 fixed the problem on an AlphaStation 200 4/166.
1505 ?? The conversion table for character set XX does not match with
1508 {UD} I don't doubt for a minute that some of the conversion tables contain
1509 errors. We tried the best we can and relied on automatic generation of the
1510 data to prevent human-introduced errors but this still is no guarantee. If
1511 you think you found a problem please send a bug report describing it and
1512 give an authoritive reference. The latter is important since otherwise
1513 the current behaviour is as good as the proposed one.
1515 Before doing this look through the list of known problem first:
1517 - the GBK (simplified Chinese) encoding is based on Unicode tables. This
1518 is good. These tables, however, differ slightly from the tables used
1519 by the M$ people. The differences are these [+ Unicode, - M$]:
1526 In addition the Unicode tables contain mappings for the GBK characters
1527 0xA8BC, 0xA8BF, 0xA989 to 0xA995, and 0xFE50 to 0xFEA0.
1529 - when mapping from EUC-CN to GBK and vice versa we ignore the fact that
1530 the coded character at position 0xA1A4 maps to different Unicode
1531 characters. Since the iconv() implementation can do whatever it wants
1532 if it cannot directly map a character this is a perfectly good solution
1533 since the semantics and appearance of the character does not change.
1536 Answers were given by:
1537 {UD} Ulrich Drepper, <drepper@cygnus.com>
1538 {DMT} David Mosberger-Tang, <davidm@AZStarNet.com>
1539 {RM} Roland McGrath, <roland@gnu.org>
1540 {AJ} Andreas Jaeger, <aj@suse.de>
1541 {EY} Eric Youngdale, <eric@andante.jic.com>
1542 {PB} Phil Blundell, <Philip.Blundell@pobox.com>
1543 {MK} Mark Kettenis, <kettenis@phys.uva.nl>
1544 {ZW} Zack Weinberg, <zack@rabi.phys.columbia.edu>
1545 {TK} Thorsten Kukuk, <kukuk@suse.de>
1546 {GK} Geoffrey Keating, <geoffk@ozemail.com.au>
1547 {HJ} H.J. Lu, <hjl@gnu.org>
1548 {CG} Cristian Gafton, <gafton@redhat.com>
1549 {AO} Alexandre Oliva, <oliva@lsd.ic.unicamp.br>
1553 outline-regexp:"\\?"