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,
17 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
21 1.1. What systems does the GNU C Library run on?
22 1.2. What compiler do I need to build GNU libc?
23 1.3. When I try to compile glibc I get only error messages.
25 1.4. Do I need a special linker or assembler?
26 1.5. Which compiler should I use for powerpc?
27 1.6. Which tools should I use for ARM?
28 1.7. Do I need some more things to compile the GNU C Library?
29 1.8. What version of the Linux kernel headers should be used?
30 1.9. The compiler hangs while building iconvdata modules. What's
32 1.10. When I run `nm -u libc.so' on the produced library I still
33 find unresolved symbols. Can this be ok?
34 1.11. What are these `add-ons'?
35 1.12. My XXX kernel emulates a floating-point coprocessor for me.
36 Should I enable --with-fp?
37 1.13. When compiling GNU libc I get lots of errors saying functions
38 in glibc are duplicated in libgcc.
39 1.14. Why do I get messages about missing thread functions when I use
40 librt? I don't even use threads.
41 1.15. What's the problem with configure --enable-omitfp?
42 1.16. I get failures during `make check'. What should I do?
43 1.17. What is symbol versioning good for? Do I need it?
44 1.18. How can I compile on my fast ix86 machine a working libc for my slow
45 i386? After installing libc, programs abort with "Illegal
48 2. Installation and configuration issues
50 2.1. Can I replace the libc on my Linux system with GNU libc?
51 2.2. How do I configure GNU libc so that the essential libraries
52 like libc.so go into /lib and the other into /usr/lib?
53 2.3. How should I avoid damaging my system when I install GNU libc?
54 2.4. Do I need to use GNU CC to compile programs that will use the
56 2.5. When linking with the new libc I get unresolved symbols
57 `crypt' and `setkey'. Why aren't these functions in the
59 2.6. When I use GNU libc on my Linux system by linking against
60 the libc.so which comes with glibc all I get is a core dump.
61 2.7. Looking through the shared libc file I haven't found the
62 functions `stat', `lstat', `fstat', and `mknod' and while
63 linking on my Linux system I get error messages. How is
64 this supposed to work?
65 2.8. When I run an executable on one system which I compiled on
66 another, I get dynamic linker errors. Both systems have the same
67 version of glibc installed. What's wrong?
68 2.9. How can I compile gcc 2.7.2.1 from the gcc source code using
70 2.10. The `gencat' utility cannot process the catalog sources which
71 were used on my Linux libc5 based system. Why?
72 2.11. Programs using libc have their messages translated, but other
73 behavior is not localized (e.g. collating order); why?
74 2.12. I have set up /etc/nis.conf, and the Linux libc 5 with NYS
75 works great. But the glibc NIS+ doesn't seem to work.
76 2.13. I have killed ypbind to stop using NIS, but glibc
78 2.14. Under Linux/Alpha, I always get "do_ypcall: clnt_call:
79 RPC: Unable to receive; errno = Connection refused" when using NIS.
80 2.15. After installing glibc name resolving doesn't work properly.
81 2.16. How do I create the databases for NSS?
82 2.17. I have /usr/include/net and /usr/include/scsi as symlinks
83 into my Linux source tree. Is that wrong?
84 2.18. Programs like `logname', `top', `uptime' `users', `w' and
85 `who', show incorrect information about the (number of)
86 users on my system. Why?
87 2.19. After upgrading to glibc 2.1 with symbol versioning I get
88 errors about undefined symbols. What went wrong?
89 2.20. When I start the program XXX after upgrading the library
91 XXX: Symbol `_sys_errlist' has different size in shared
92 object, consider re-linking
93 Why? What should I do?
94 2.21. What do I need for C++ development?
95 2.22. Even statically linked programs need some shared libraries
96 which is not acceptable for me. What can I do?
97 2.23. I just upgraded my Linux system to glibc and now I get
98 errors whenever I try to link any program.
99 2.24. When I use nscd the machine freezes.
100 2.25. I need lots of open files. What do I have to do?
101 2.26. How do I get the same behavior on parsing /etc/passwd and
102 /etc/group as I have with libc5 ?
103 2.27. What needs to be recompiled when upgrading from glibc 2.0 to glibc
105 2.28. Why is extracting files via tar so slow?
106 2.29. Compiling programs I get parse errors in libio.h (e.g. "parse error
107 before `_IO_seekoff'"). How should I fix this?
108 2.30. After upgrading to glibc 2.1, libraries that were compiled against
109 glibc 2.0.x don't work anymore.
110 2.31. What happened to the Berkeley DB libraries? Can I still use db
111 in /etc/nsswitch.conf?
113 3. Source and binary incompatibilities, and what to do about them
115 3.1. I expect GNU libc to be 100% source code compatible with
116 the old Linux based GNU libc. Why isn't it like this?
117 3.2. Why does getlogin() always return NULL on my Linux box?
118 3.3. Where are the DST_* constants found in <sys/time.h> on many
120 3.4. The prototypes for `connect', `accept', `getsockopt',
121 `setsockopt', `getsockname', `getpeername', `send',
122 `sendto', and `recvfrom' are different in GNU libc from
123 any other system I saw. This is a bug, isn't it?
124 3.5. On Linux I've got problems with the declarations in Linux
126 3.6. I don't include any kernel headers myself but the compiler
127 still complains about redeclarations of types in the kernel
129 3.7. Why don't signals interrupt system calls anymore?
130 3.8. I've got errors compiling code that uses certain string
132 3.9. I get compiler messages "Initializer element not constant" with
133 stdin/stdout/stderr. Why?
134 3.10. I can't compile with gcc -traditional (or
135 -traditional-cpp). Why?
136 3.11. I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible?
137 3.12. I can't access some functions anymore. nm shows that they do
138 exist but linking fails nevertheless.
139 3.13. When using the db-2 library which comes with glibc is used in
140 the Perl db modules the testsuite is not passed. This did not
141 happen with db-1, gdbm, or ndbm.
143 Removed. Does not apply anymore.
144 3.14. The pow() inline function I get when including <math.h> is broken.
145 I get segmentation faults when I run the program.
146 3.15. The sys/sem.h file lacks the definition of `union semun'.
147 3.16. Why has <netinet/ip_fw.h> disappeared?
148 3.17. I get floods of warnings when I use -Wconversion and include
149 <string.h> or <math.h>.
150 3.18. After upgrading to glibc 2.1, I receive errors about
151 unresolved symbols, like `_dl_initial_searchlist' and can not
152 execute any binaries. What went wrong?
153 3.19. bonnie reports that char i/o with glibc 2 is much slower than with
154 libc5. What can be done?
155 3.20. Programs compiled with glibc 2.1 can't read db files made with glibc
156 2.0. What has changed that programs like rpm break?
158 Removed. Does not apply anymore.
159 3.21. Autoconf's AC_CHECK_FUNC macro reports that a function exists, but
160 when I try to use it, it always returns -1 and sets errno to ENOSYS.
161 3.22. My program segfaults when I call fclose() on the FILE* returned
162 from setmntent(). Is this a glibc bug?
166 4.1. After I changed configure.in I get `Autoconf version X.Y.
167 or higher is required for this script'. What can I do?
168 4.2. When I try to compile code which uses IPv6 headers and
169 definitions on my Linux 2.x.y system I am in trouble.
170 Nothing seems to work.
171 4.3. When I set the timezone by setting the TZ environment variable
172 to EST5EDT things go wrong since glibc computes the wrong time
173 from this information.
174 4.4. What other sources of documentation about glibc are available?
175 4.5. The timezone string for Sydney/Australia is wrong since even when
176 daylight saving time is in effect the timezone string is EST.
177 4.6. I've build make 3.77 against glibc 2.1 and now make gets
179 4.7. Why do so many programs using math functions fail on my AlphaStation?
180 4.8. The conversion table for character set XX does not match with
184 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
188 1.1. What systems does the GNU C Library run on?
190 {UD} This is difficult to answer. The file `README' lists the architectures
191 GNU libc was known to run on *at some time*. This does not mean that it
192 still can be compiled and run on them now.
194 The systems glibc is known to work on as of this release, and most probably
198 i[3456]86-*-linux-gnu Linux-2.x on Intel
199 m68k-*-linux-gnu Linux-2.x on Motorola 680x0
200 alpha-*-linux-gnu Linux-2.x on DEC Alpha
201 powerpc-*-linux-gnu Linux and MkLinux on PowerPC systems
202 sparc-*-linux-gnu Linux-2.x on SPARC
203 sparc64-*-linux-gnu Linux-2.x on UltraSPARC
204 arm-*-none ARM standalone systems
205 arm-*-linux Linux-2.x on ARM
206 arm-*-linuxaout Linux-2.x on ARM using a.out binaries
208 Ports to other Linux platforms are in development, and may in fact work
209 already, but no one has sent us success reports for them. Currently no
210 ports to other operating systems are underway, although a few people have
213 If you have a system not listed above (or in the `README' file) and you are
214 really interested in porting it, contact
219 1.2. What compiler do I need to build GNU libc?
221 {UD} You must use GNU CC to compile GNU libc. A lot of extensions of GNU CC
222 are used to increase portability and speed.
224 GNU CC is found, like all other GNU packages, on
226 ftp://ftp.gnu.org/pub/gnu
228 and the many mirror sites. ftp.gnu.org is always overloaded, so try to find
229 a local mirror first.
231 You should always try to use the latest official release. Older versions
232 may not have all the features GNU libc requires. The current releases of
233 egcs (1.0.3 and 1.1.1) should work with the GNU C library (for powerpc see
234 question 1.5; for ARM see question 1.6).
236 While the GNU CC should be able to compile glibc it is nevertheless adviced
237 to use EGCS. Comparing the sizes of glibc on Intel compiled with a recent
238 EGCS and gcc 2.8.1 shows this:
240 text data bss dec hex filename
241 egcs-2.93.10 862897 15944 12824 891665 d9b11 libc.so
242 gcc-2.8.1 959965 16468 12152 988585 f15a9 libc.so
244 Make up your own decision.
246 GNU CC versions 2.95 and above are derived from egcs, and they may do even
249 Please note that gcc 2.95 and 2.95.x cannot compile glibc on Alpha due to
250 problems in the complex float support.
253 1.3. When I try to compile glibc I get only error messages.
256 {UD} You definitely need GNU make to build GNU libc. No other make
257 program has the needed functionality.
259 We recommend version GNU make version 3.75 or 3.77. Versions before 3.75
260 have bugs and/or are missing features. Version 3.76 has bugs which
261 appear when building big projects like GNU libc. 3.76.1 appears to work but
262 some people have reported problems. If you build GNU make 3.77 from source,
263 please read question 4.6 first.
266 1.4. Do I need a special linker or assembler?
268 {ZW} If you want a shared library, you need a linker and assembler that
269 understand all the features of ELF, including weak and versioned symbols.
270 The static library can be compiled with less featureful tools, but lacks key
271 features such as NSS.
273 For Linux or Hurd, you want binutils 2.8.1.0.23, 2.9.1, or 2.9.1.0.15 or
274 higher. These are the only versions we've tested and found reliable. Other
275 versions after 2.8.1.0.23 may work but we don't recommend them, especially
276 not when C++ is involved. Earlier versions do not work at all.
278 Other operating systems may come with system tools that have all the
279 necessary features, but this is moot because glibc hasn't been ported to
283 1.5. Which compiler should I use for powerpc?
285 {GK} You want to use at least gcc 2.95 (together with the right versions
286 of all the other tools, of course). See also question question 2.8.
289 1.6. Which tools should I use for ARM?
291 {PB} You should use egcs 1.1 or a later version. For ELF systems some
292 changes are needed to the compiler; a patch against egcs-1.1.x can be found
295 <ftp://ftp.netwinder.org/users/p/philb/egcs-1.1.1pre2-diff-981126>
297 Binutils 2.9.1.0.16 or later is also required.
300 1.7. Do I need some more things to compile the GNU C Library?
302 {UD} Yes, there are some more :-).
304 * GNU gettext. This package contains the tools needed to construct
305 `message catalog' files containing translated versions of system
306 messages. See ftp://ftp.gnu.org/pub/gnu or better any mirror
307 site. (We distribute compiled message catalogs, but they may not be
308 updated in patches.) Please note that the required minimal version
309 (0.10.35) of gettext is alpha software and available from
310 ftp://alpha.gnu.org/gnu .
312 * Some files are built with special tools. E.g., files ending in .gperf
313 need a `gperf' program. The GNU version (now available in a separate
314 package, formerly only as part of libg++) is known to work while some
315 vendor versions do not.
317 You should not need these tools unless you change the source files.
319 * Perl 5 is needed if you wish to test an installation of GNU libc
320 as the primary C library.
322 * When compiling for Linux, the header files of the Linux kernel must
323 be available to the compiler as <linux/*.h> and <asm/*.h>.
325 * lots of disk space (~400MB for i?86-linux; more for RISC platforms).
327 * plenty of time. Compiling just the shared and static libraries for
328 i?86-linux takes approximately 1h on an AMD-K6@225MHz w/ 96MB of RAM,
329 45mins on a Celeron@400MHz w/ 128MB, and 55mins on a Alpha@533MHz w/ 256MB.
330 Multiply this by 1.5 or 2.0 if you build profiling and/or the highly
331 optimized version as well. For Hurd systems times are much higher.
333 You should avoid compiling in a NFS mounted filesystem. This is
336 James Troup <J.J.Troup@comp.brad.ac.uk> reports a compile time of
337 45h34m for a full build (shared, static, and profiled) on Atari
338 Falcon (Motorola 68030 @ 16 Mhz, 14 Mb memory) and Jan Barte
339 <yann@plato.uni-paderborn.de> reports 22h48m on Atari TT030
340 (Motorola 68030 @ 32 Mhz, 34 Mb memory)
342 A full build of the PowerPC library took 1h on a PowerPC 750@400Mhz w/
343 64MB of RAM, and about 9h on a 601@60Mhz w/ 72Mb.
345 If you have some more measurements let me know.
348 1.8. What version of the Linux kernel headers should be used?
350 {AJ,UD} The headers from the most recent Linux kernel should be used. The
351 headers used while compiling the GNU C library and the kernel binary used
352 when using the library do not need to match. The GNU C library runs without
353 problems on kernels that are older than the kernel headers used. The other
354 way round (compiling the GNU C library with old kernel headers and running
355 on a recent kernel) does not necessarily work. For example you can't use
356 new kernel features if you used old kernel headers to compile the GNU C
359 {ZW} Even if you are using a 2.0 kernel on your machine, we recommend you
360 compile GNU libc with 2.2 kernel headers. That way you won't have to
361 recompile libc if you ever upgrade to kernel 2.2. To tell libc which
362 headers to use, give configure the --with-headers switch
363 (e.g. --with-headers=/usr/src/linux-2.2.0/include).
365 Note that you must configure the 2.2 kernel if you do this, otherwise libc
366 will be unable to find <linux/version.h>. Just change the current directory
367 to the root of the 2.2 tree and do `make include/linux/version.h'.
370 1.9. The compiler hangs while building iconvdata modules. What's
373 {ZW} This is a problem with old versions of GCC. Initialization of large
374 static arrays is very slow. The compiler will eventually finish; give it
377 The problem is fixed in egcs 1.1.
380 1.10. When I run `nm -u libc.so' on the produced library I still
381 find unresolved symbols. Can this be ok?
383 {UD} Yes, this is ok. There can be several kinds of unresolved symbols:
385 * magic symbols automatically generated by the linker. These have names
386 like __start_* and __stop_*
388 * symbols starting with _dl_* come from the dynamic linker
390 * weak symbols, which need not be resolved at all (fabs for example)
392 Generally, you should make sure you find a real program which produces
393 errors while linking before deciding there is a problem.
396 1.11. What are these `add-ons'?
398 {UD} To avoid complications with export rules or external source code some
399 optional parts of the libc are distributed as separate packages (e.g., the
400 crypt package, see question 2.5).
402 To use these packages as part of GNU libc, just unpack the tarfiles in the
403 libc source directory and tell the configuration script about them using the
404 --enable-add-ons option. If you give just --enable-add-ons configure tries
405 to find all the add-on packages in your source tree. This may not work. If
406 it doesn't, or if you want to select only a subset of the add-ons, give a
407 comma-separated list of the add-ons to enable:
409 configure --enable-add-ons=crypt,linuxthreads
413 Add-ons can add features (including entirely new shared libraries), override
414 files, provide support for additional architectures, and just about anything
415 else. The existing makefiles do most of the work; only some few stub rules
416 must be written to get everything running.
418 Most add-ons are tightly coupled to a specific GNU libc version. Please
419 check that the add-ons work with the GNU libc. For example the crypt and
420 linuxthreads add-ons have the same numbering scheme as the libc and will in
421 general only work with the corresponding libc.
424 1.12. My XXX kernel emulates a floating-point coprocessor for me.
425 Should I enable --with-fp?
427 {ZW} An emulated FPU is just as good as a real one, as far as the C library
428 is concerned. You only need to say --without-fp if your machine has no way
429 to execute floating-point instructions.
431 People who are interested in squeezing the last drop of performance
432 out of their machine may wish to avoid the trap overhead, but this is
433 far more trouble than it's worth: you then have to compile
434 *everything* this way, including the compiler's internal libraries
435 (libgcc.a for GNU C), because the calling conventions change.
438 1.13. When compiling GNU libc I get lots of errors saying functions
439 in glibc are duplicated in libgcc.
441 {EY} This is *exactly* the same problem that I was having. The problem was
442 due to the fact that configure didn't correctly detect that the linker flag
443 --no-whole-archive was supported in my linker. In my case it was because I
444 had run ./configure with bogus CFLAGS, and the test failed.
446 One thing that is particularly annoying about this problem is that once this
447 is misdetected, running configure again won't fix it unless you first delete
450 {UD} Starting with glibc-2.0.3 there should be a better test to avoid some
451 problems of this kind. The setting of CFLAGS is checked at the very
452 beginning and if it is not usable `configure' will bark.
455 1.14. Why do I get messages about missing thread functions when I use
456 librt? I don't even use threads.
458 {UD} In this case you probably mixed up your installation. librt uses
459 threads internally and has implicit references to the thread library.
460 Normally these references are satisfied automatically but if the thread
461 library is not in the expected place you must tell the linker where it is.
462 When using GNU ld it works like this:
464 gcc -o foo foo.c -Wl,-rpath-link=/some/other/dir -lrt
466 The `/some/other/dir' should contain the thread library. `ld' will use the
467 given path to find the implicitly referenced library while not disturbing
471 1.15. What's the problem with configure --enable-omitfp?
473 {AJ} When --enable-omitfp is set the libraries are built without frame
474 pointers. Some compilers produce buggy code for this model and therefore we
475 don't advise using it at the moment.
477 If you use --enable-omitfp, you're on your own. If you encounter problems
478 with a library that was build this way, we advise you to rebuild the library
479 without --enable-omitfp. If the problem vanishes consider tracking the
480 problem down and report it as compiler failure.
482 Since a library built with --enable-omitfp is undebuggable on most systems,
483 debuggable libraries are also built - you can use them by appending "_g" to
486 The compilation of these extra libraries and the compiler optimizations slow
487 down the build process and need more disk space.
490 1.16. I get failures during `make check'. What should I do?
492 {AJ} The testsuite should compile and run cleanly on your system; every
493 failure should be looked into. Depending on the failures, you probably
494 should not install the library at all.
496 You should consider using the `glibcbug' script to report the failure,
497 providing as much detail as possible. If you run a test directly, please
498 remember to set up the environment correctly. You want to test the compiled
499 library - and not your installed one. The best way is to copy the exact
500 command line which failed and run the test from the subdirectory for this
503 There are some failures which are not directly related to the GNU libc:
504 - Some compilers produce buggy code. No compiler gets single precision
505 complex numbers correct on Alpha. Otherwise, the egcs 1.1 release should be
506 ok; gcc 2.8.1 might cause some failures; gcc 2.7.2.x is so buggy that
507 explicit checks have been used so that you can't build with it.
508 - The kernel might have bugs. For example on Linux/Alpha 2.0.34 the
509 floating point handling has quite a number of bugs and therefore most of
510 the test cases in the math subdirectory will fail. Linux 2.2 has
511 fixes for the floating point support on Alpha. The Linux/SPARC kernel has
512 also some bugs in the FPU emulation code (as of Linux 2.2.0).
513 - Other tools might have problems. For example bash 2.03 gives a
514 segmentation fault running the tst-rpmatch.sh test script.
517 1.17. What is symbol versioning good for? Do I need it?
519 {AJ} Symbol versioning solves problems that are related to interface
520 changes. One version of an interface might have been introduced in a
521 previous version of the GNU C library but the interface or the semantics of
522 the function has been changed in the meantime. For binary compatibility
523 with the old library, a newer library needs to still have the old interface
524 for old programs. On the other hand, new programs should use the new
525 interface. Symbol versioning is the solution for this problem. The GNU
526 libc version 2.1 uses symbol versioning by default if the installed binutils
529 We don't advise building without symbol versioning, since you lose binary
530 compatibility - forever! The binary compatibility you lose is not only
531 against the previous version of the GNU libc (version 2.0) but also against
535 1.18. How can I compile on my fast ix86 machine a working libc for my slow
536 i386? After installing libc, programs abort with "Illegal
539 {AJ} glibc and gcc might generate some instructions on your machine that
540 aren't available on i386. You've got to tell glibc that you're configuring
541 for i386 with adding i386 as your machine, for example:
543 ../configure --prefix=/usr i386-pc-linux-gnu
545 And you need to tell gcc to only generate i386 code, just add `-mcpu=i386'
546 (just -m386 doesn't work) to your CFLAGS.
548 {UD} This applies not only to the i386. Compiling on a i686 for any older
549 model will also fail if the above methods are not used.
552 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
554 2. Installation and configuration issues
556 2.1. Can I replace the libc on my Linux system with GNU libc?
558 {UD} You cannot replace any existing libc for Linux with GNU libc. It is
559 binary incompatible and therefore has a different major version. You can,
560 however, install it alongside your existing libc.
562 For Linux there are three major libc versions:
564 libc-5 original ELF libc
567 You can have any combination of these three installed. For more information
568 consult documentation for shared library handling. The Makefiles of GNU
569 libc will automatically generate the needed symbolic links which the linker
573 2.2. How do I configure GNU libc so that the essential libraries
574 like libc.so go into /lib and the other into /usr/lib?
576 {UD,AJ} Like all other GNU packages GNU libc is designed to use a base
577 directory and install all files relative to this. The default is
578 /usr/local, because this is safe (it will not damage the system if installed
579 there). If you wish to install GNU libc as the primary C library on your
580 system, set the base directory to /usr (i.e. run configure --prefix=/usr
581 <other_options>). Note that this can damage your system; see question 2.3 for
584 Some systems like Linux have a filesystem standard which makes a difference
585 between essential libraries and others. Essential libraries are placed in
586 /lib because this directory is required to be located on the same disk
587 partition as /. The /usr subtree might be found on another
588 partition/disk. If you configure for Linux with --prefix=/usr, then this
589 will be done automatically.
591 To install the essential libraries which come with GNU libc in /lib on
592 systems other than Linux one must explicitly request it. Autoconf has no
593 option for this so you have to use a `configparms' file (see the `INSTALL'
594 file for details). It should contain:
599 The first line specifies the directory for the essential libraries, the
600 second line the directory for system configuration files.
603 2.3. How should I avoid damaging my system when I install GNU libc?
605 {ZW} If you wish to be cautious, do not configure with --prefix=/usr. If
606 you don't specify a prefix, glibc will be installed in /usr/local, where it
607 will probably not break anything. (If you wish to be certain, set the
608 prefix to something like /usr/local/glibc2 which is not used for anything.)
610 The dangers when installing glibc in /usr are twofold:
612 * glibc will overwrite the headers in /usr/include. Other C libraries
613 install a different but overlapping set of headers there, so the effect
614 will probably be that you can't compile anything. You need to rename
615 /usr/include out of the way before running `make install'. (Do not throw
616 it away; you will then lose the ability to compile programs against your
619 * None of your old libraries, static or shared, can be used with a
620 different C library major version. For shared libraries this is not a
621 problem, because the filenames are different and the dynamic linker
622 will enforce the restriction. But static libraries have no version
623 information. You have to evacuate all the static libraries in
624 /usr/lib to a safe location.
626 The situation is rather similar to the move from a.out to ELF which
627 long-time Linux users will remember.
630 2.4. Do I need to use GNU CC to compile programs that will use the
633 {ZW} In theory, no; the linker does not care, and the headers are supposed
634 to check for GNU CC before using its extensions to the C language.
636 However, there are currently no ports of glibc to systems where another
637 compiler is the default, so no one has tested the headers extensively
638 against another compiler. You may therefore encounter difficulties. If you
639 do, please report them as bugs.
641 Also, in several places GNU extensions provide large benefits in code
642 quality. For example, the library has hand-optimized, inline assembly
643 versions of some string functions. These can only be used with GCC. See
644 question 3.8 for details.
647 2.5. When linking with the new libc I get unresolved symbols
648 `crypt' and `setkey'. Why aren't these functions in the
651 {UD} The US places restrictions on exporting cryptographic programs and
652 source code. Until this law gets abolished we cannot ship the cryptographic
653 functions together with glibc.
655 The functions are available, as an add-on (see question 1.11). People in the US
656 may get it from the same place they got GNU libc from. People outside the
657 US should get the code from ftp.gwdg.de [134.76.11.100] in the directory
658 pub/linux/glibc, or another archive site outside the USA. The README explains
659 how to install the sources.
661 If you already have the crypt code on your system the reason for the failure
662 is probably that you did not link with -lcrypt. The crypto functions are in
663 a separate library to make it possible to export GNU libc binaries from the
667 2.6. When I use GNU libc on my Linux system by linking against
668 the libc.so which comes with glibc all I get is a core dump.
670 {UD} On Linux, gcc sets the dynamic linker to /lib/ld-linux.so.1 unless the
671 user specifies a --dynamic-linker argument. This is the name of the libc5
672 dynamic linker, which does not work with glibc.
674 For casual use of GNU libc you can just specify to the linker
675 --dynamic-linker=/lib/ld-linux.so.2
677 which is the glibc dynamic linker, on Linux systems. On other systems the
678 name is /lib/ld.so.1. When linking via gcc, you've got to add
679 -Wl,--dynamic-linker=/lib/ld-linux.so.2
681 to the gcc command line.
683 To change your environment to use GNU libc for compiling you need to change
684 the `specs' file of your gcc. This file is normally found at
686 /usr/lib/gcc-lib/<arch>/<version>/specs
688 In this file you have to change a few things:
690 - change `ld-linux.so.1' to `ld-linux.so.2'
692 - remove all expression `%{...:-lgmon}'; there is no libgmon in glibc
694 - fix a minor bug by changing %{pipe:-} to %|
696 Here is what the gcc-2.7.2 specs file should look like when GNU libc is
699 -----------------------------------------------------------------------
701 %{V} %{v:%{!V:-V}} %{Qy:} %{!Qn:-Qy} %{n} %{T} %{Ym,*} %{Yd,*} %{Wa,*:%*}
707 %{fPIC:-D__PIC__ -D__pic__} %{fpic:-D__PIC__ -D__pic__} %{!m386:-D__i486__} %{posix:-D_POSIX_SOURCE} %{pthread:-D_REENTRANT}
716 %{!shared:crtend.o%s} %{shared:crtendS.o%s} crtn.o%s
719 -m elf_i386 %{shared:-shared} %{!shared: %{!ibcs: %{!static: %{rdynamic:-export-dynamic} %{!dynamic-linker:-dynamic-linker /lib/ld-linux.so.2}} %{static:-static}}}
722 %{!shared: %{pthread:-lpthread} %{profile:-lc_p} %{!profile: -lc}}
728 %{!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}
730 *switches_need_spaces:
734 %{funsigned-char:-D__CHAR_UNSIGNED__}
737 -D__ELF__ -Dunix -Di386 -Dlinux -Asystem(unix) -Asystem(posix) -Acpu(i386) -Amachine(i386)
745 -----------------------------------------------------------------------
747 Things get a bit more complicated if you have GNU libc installed in some
748 other place than /usr, i.e., if you do not want to use it instead of the old
749 libc. In this case the needed startup files and libraries are not found in
750 the regular places. So the specs file must tell the compiler and linker
753 Version 2.7.2.3 does and future versions of GCC will automatically
754 provide the correct specs.
757 2.7. Looking through the shared libc file I haven't found the
758 functions `stat', `lstat', `fstat', and `mknod' and while
759 linking on my Linux system I get error messages. How is
760 this supposed to work?
762 {RM} Believe it or not, stat and lstat (and fstat, and mknod) are supposed
763 to be undefined references in libc.so.6! Your problem is probably a missing
764 or incorrect /usr/lib/libc.so file; note that this is a small text file now,
765 not a symlink to libc.so.6. It should look something like this:
767 GROUP ( libc.so.6 libc_nonshared.a )
770 2.8. When I run an executable on one system which I compiled on
771 another, I get dynamic linker errors. Both systems have the same
772 version of glibc installed. What's wrong?
774 {ZW} Glibc on one of these systems was compiled with gcc 2.7 or 2.8, the
775 other with egcs (any version). Egcs has functions in its internal
776 `libgcc.a' to support exception handling with C++. They are linked into
777 any program or dynamic library compiled with egcs, whether it needs them or
778 not. Dynamic libraries then turn around and export those functions again
779 unless special steps are taken to prevent them.
781 When you link your program, it resolves its references to the exception
782 functions to the ones exported accidentally by libc.so. That works fine as
783 long as libc has those functions. On the other system, libc doesn't have
784 those functions because it was compiled by gcc 2.8, and you get undefined
785 symbol errors. The symbols in question are named things like
786 `__register_frame_info'.
788 For glibc 2.0, the workaround is to not compile libc with egcs. We've also
789 incorporated a patch which should prevent the EH functions sneaking into
790 libc. It doesn't matter what compiler you use to compile your program.
792 For glibc 2.1, we've chosen to do it the other way around: libc.so
793 explicitly provides the EH functions. This is to prevent other shared
794 libraries from doing it.
796 {UD} Starting with glibc 2.1.1 you can compile glibc with gcc 2.8.1 or
797 newer since we have explicitly add references to the functions causing the
798 problem. But you nevertheless should use EGCS for other reasons
801 {GK} On some Linux distributions for PowerPC, you can see this when you have
802 built gcc or egcs from the Web sources (gcc versions 2.95 or earlier), then
803 re-built glibc. This happens because in these versions of gcc, exception
804 handling is implemented using an older method; the people making the
805 distributions are a little ahead of their time.
807 A quick solution to this is to find the libgcc.a file that came with the
808 distribution (it would have been installed under /usr/lib/gcc-lib), do
809 `ar x libgcc.a frame.o' to get the frame.o file out, and add a line saying
810 `LDLIBS-c.so += frame.o' to the file `configparms' in the directory you're
811 building in. You can check you've got the right `frame.o' file by running
812 `nm frame.o' and checking that it has the symbols defined that you're
815 This will let you build glibc with the C compiler. The C++ compiler
816 will still be binary incompatible with any C++ shared libraries that
817 you got with your distribution.
820 2.9. How can I compile gcc 2.7.2.1 from the gcc source code using
823 {AJ} There's only correct support for glibc 2.0.x in gcc 2.7.2.3 or later.
824 But you should get at least gcc 2.8.1 or egcs 1.1 (or later versions)
828 2.10. The `gencat' utility cannot process the catalog sources which
829 were used on my Linux libc5 based system. Why?
831 {UD} The `gencat' utility provided with glibc complies to the XPG standard.
832 The older Linux version did not obey the standard, so they are not
835 To ease the transition from the Linux version some of the non-standard
836 features are also present in the `gencat' program of GNU libc. This mainly
837 includes the use of symbols for the message number and the automatic
838 generation of header files which contain the needed #defines to map the
841 Here is a simple SED script to convert at least some Linux specific catalog
842 files to the XPG4 form:
844 -----------------------------------------------------------------------
845 # Change catalog source in Linux specific format to standard XPG format.
846 # Ulrich Drepper <drepper@cygnus.com>, 1996.
850 s/\$ #\([^ ]*\).*/\1/
852 s/\$ #[^ ]* *\(.*\)/\$ \1/
858 s/\(.*\)\n\(.*\)/\2 \1/
860 -----------------------------------------------------------------------
863 2.11. Programs using libc have their messages translated, but other
864 behavior is not localized (e.g. collating order); why?
866 {ZW} Translated messages are automatically installed, but the locale
867 database that controls other behaviors is not. You need to run localedef to
868 install this database, after you have run `make install'. For example, to
869 set up the French Canadian locale, simply issue the command
871 localedef -i fr_CA -f ISO-8859-1 fr_CA
873 Please see localedata/README in the source tree for further details.
876 2.12. I have set up /etc/nis.conf, and the Linux libc 5 with NYS
877 works great. But the glibc NIS+ doesn't seem to work.
879 {TK} The glibc NIS+ implementation uses a /var/nis/NIS_COLD_START file for
880 storing information about the NIS+ server and their public keys, because the
881 nis.conf file does not contain all the necessary information. You have to
882 copy a NIS_COLD_START file from a Solaris client (the NIS_COLD_START file is
883 byte order independent) or generate it with nisinit from the nis-tools
884 package; available at
886 http://www.suse.de/~kukuk/linux/nisplus.html
889 2.13. I have killed ypbind to stop using NIS, but glibc
892 {TK} For faster NIS lookups, glibc uses the /var/yp/binding/ files from
893 ypbind. ypbind 3.3 and older versions don't always remove these files, so
894 glibc will continue to use them. Other BSD versions seem to work correctly.
895 Until ypbind 3.4 is released, you can find a patch at
897 <ftp://ftp.kernel.org/pub/linux/utils/net/NIS/ypbind-3.3-glibc4.diff.gz>
900 2.14. Under Linux/Alpha, I always get "do_ypcall: clnt_call:
901 RPC: Unable to receive; errno = Connection refused" when using NIS.
903 {TK} You need a ypbind version which is 64bit clean. Some versions are not
904 64bit clean. A 64bit clean implementation is ypbind-mt. For ypbind 3.3,
905 you need the patch from ftp.kernel.org (See the previous question). I don't
906 know about other versions.
909 2.15. After installing glibc name resolving doesn't work properly.
911 {AJ} You probably should read the manual section describing nsswitch.conf
912 (just type `info libc "NSS Configuration File"'). The NSS configuration
913 file is usually the culprit.
916 2.16. How do I create the databases for NSS?
918 {AJ} If you have an entry "db" in /etc/nsswitch.conf you should also create
919 the database files. The glibc sources contain a Makefile which does the
920 necessary conversion and calls to create those files. The file is
921 `db-Makefile' in the subdirectory `nss' and you can call it with `make -f
922 db-Makefile'. Please note that not all services are capable of using a
923 database. Currently passwd, group, ethers, protocol, rpc, services shadow
924 and netgroup are implemented. See also question question 2.31.
927 2.17. I have /usr/include/net and /usr/include/scsi as symlinks
928 into my Linux source tree. Is that wrong?
930 {PB} This was necessary for libc5, but is not correct when using glibc.
931 Including the kernel header files directly in user programs usually does not
932 work (see question 3.5). glibc provides its own <net/*> and <scsi/*> header
933 files to replace them, and you may have to remove any symlink that you have
934 in place before you install glibc. However, /usr/include/asm and
935 /usr/include/linux should remain as they were.
938 2.18. Programs like `logname', `top', `uptime' `users', `w' and
939 `who', show incorrect information about the (number of)
940 users on my system. Why?
942 {MK} See question 3.2.
945 2.19. After upgrading to glibc 2.1 with symbol versioning I get
946 errors about undefined symbols. What went wrong?
948 {AJ} The problem is caused either by wrong program code or tools. In the
949 versioned libc a lot of symbols are now local that were global symbols in
950 previous versions. It seems that programs linked against older versions
951 often accidentally used libc global variables -- something that should not
954 The only way to fix this is to recompile your program. Sorry, that's the
955 price you might have to pay once for quite a number of advantages with
959 2.20. When I start the program XXX after upgrading the library
961 XXX: Symbol `_sys_errlist' has different size in shared
962 object, consider re-linking
963 Why? What should I do?
965 {UD} As the message says, relink the binary. The problem is that a few
966 symbols from the library can change in size and there is no way to avoid
967 this. _sys_errlist is a good example. Occasionally there are new error
968 numbers added to the kernel and this must be reflected at user level,
969 breaking programs that refer to them directly.
971 Such symbols should normally not be used at all. There are mechanisms to
972 avoid using them. In the case of _sys_errlist, there is the strerror()
973 function which should _always_ be used instead. So the correct fix is to
974 rewrite that part of the application.
976 In some situations (especially when testing a new library release) it might
977 be possible that a symbol changed size when that should not have happened.
978 So in case of doubt report such a warning message as a problem.
981 2.21. What do I need for C++ development?
983 {HJ,AJ} You need either egcs 1.1 which comes directly with libstdc++ or
984 gcc-2.8.1 together with libstdc++ 2.8.1.1. egcs 1.1 has the better C++
985 support and works directly with glibc 2.1. If you use gcc-2.8.1 with
986 libstdc++ 2.8.1.1, you need to modify libstdc++ a bit. A patch is available
988 <ftp://alpha.gnu.org/gnu/libstdc++-2.8.1.1-glibc2.1-diff.gz>
990 Please note that libg++ 2.7.2 (and the Linux Versions 2.7.2.x) doesn't work
991 very well with the GNU C library due to vtable thunks. If you're upgrading
992 from glibc 2.0.x to 2.1 you have to recompile libstdc++ since the library
993 compiled for 2.0 is not compatible due to the new Large File Support (LFS)
996 {UD} But since in the case of a shared libstdc++ the version numbers should
997 be different existing programs will continue to work.
1000 2.22. Even statically linked programs need some shared libraries
1001 which is not acceptable for me. What can I do?
1003 {AJ} NSS (for details just type `info libc "Name Service Switch"') won't
1004 work properly without shared libraries. NSS allows using different services
1005 (e.g. NIS, files, db, hesiod) by just changing one configuration file
1006 (/etc/nsswitch.conf) without relinking any programs. The only disadvantage
1007 is that now static libraries need to access shared libraries. This is
1008 handled transparently by the GNU C library.
1010 A solution is to configure glibc with --enable-static-nss. In this case you
1011 can create a static binary that will use only the services dns and files
1012 (change /etc/nsswitch.conf for this). You need to link explicitly against
1013 all these services. For example:
1015 gcc -static test-netdb.c -o test-netdb.c \
1016 -lc -lnss_files -lnss_dns -lresolv
1018 The problem with this approach is that you've got to link every static
1019 program that uses NSS routines with all those libraries.
1021 {UD} In fact, one cannot say anymore that a libc compiled with this
1022 option is using NSS. There is no switch anymore. Therefore it is
1023 *highly* recommended *not* to use --enable-static-nss since this makes
1024 the behaviour of the programs on the system inconsistent.
1027 2.23. I just upgraded my Linux system to glibc and now I get
1028 errors whenever I try to link any program.
1030 {ZW} This happens when you have installed glibc as the primary C library but
1031 have stray symbolic links pointing at your old C library. If the first
1032 `libc.so' the linker finds is libc 5, it will use that. Your program
1033 expects to be linked with glibc, so the link fails.
1035 The most common case is that glibc put its `libc.so' in /usr/lib, but there
1036 was a `libc.so' from libc 5 in /lib, which gets searched first. To fix the
1037 problem, just delete /lib/libc.so. You may also need to delete other
1038 symbolic links in /lib, such as /lib/libm.so if it points to libm.so.5.
1040 {AJ} The perl script test-installation.pl which is run as last step during
1041 an installation of glibc that is configured with --prefix=/usr should help
1042 detect these situations. If the script reports problems, something is
1046 2.24. When I use nscd the machine freezes.
1048 {UD} You cannot use nscd with Linux 2.0.*. There is functionality missing
1049 in the kernel and work-arounds are not suitable. Besides, some parts of the
1050 kernel are too buggy when it comes to using threads.
1052 If you need nscd, you have to use at least a 2.1 kernel.
1054 Note that I have at this point no information about any other platform.
1057 2.25. I need lots of open files. What do I have to do?
1059 {AJ} This is at first a kernel issue. The kernel defines limits with
1060 OPEN_MAX the number of simultaneous open files and with FD_SETSIZE the
1061 number of used file descriptors. You need to change these values in your
1062 kernel and recompile the kernel so that the kernel allows to use more open
1063 files. You don't necessarily need to recompile the GNU C library since the
1064 only place where OPEN_MAX and FD_SETSIZE is really needed in the library
1065 itself is the size of fd_set which is used by select.
1067 The GNU C library is now select free. This means it internally has no
1068 limits imposed by the `fd_set' type. Instead all places where the
1069 functionality is needed the `poll' function is used.
1071 If you increase the number of file descriptors in the kernel you don't need
1072 to recompile the C library.
1074 {UD} You can always get the maximum number of file descriptors a process is
1075 allowed to have open at any time using
1077 number = sysconf (_SC_OPEN_MAX);
1079 This will work even if the kernel limits change.
1082 2.26. How do I get the same behavior on parsing /etc/passwd and
1083 /etc/group as I have with libc5 ?
1085 {TK} The name switch setup in /etc/nsswitch.conf selected by most Linux
1086 distributions does not support +/- and netgroup entries in the files like
1087 /etc/passwd. Though this is the preferred setup some people might have
1088 setups coming over from the libc5 days where it was the default to recognize
1089 lines like this. To get back to the old behaviour one simply has to change
1090 the rules for passwd, group, and shadow in the nsswitch.conf file as
1102 2.27. What needs to be recompiled when upgrading from glibc 2.0 to glibc
1105 {AJ,CG} If you just upgrade the glibc from 2.0.x (x <= 7) to 2.1, binaries
1106 that have been linked against glibc 2.0 will continue to work.
1108 If you compile your own binaries against glibc 2.1, you also need to
1109 recompile some other libraries. The problem is that libio had to be changed
1110 and therefore libraries that are based or depend on the libio of glibc,
1111 e.g. ncurses, slang and most C++ libraries, need to be recompiled. If you
1112 experience strange segmentation faults in your programs linked against glibc
1113 2.1, you might need to recompile your libraries.
1115 Another problem is that older binaries that were linked statically against
1116 glibc 2.0 will reference the older nss modules (libnss_files.so.1 instead of
1117 libnss_files.so.2), so don't remove them. Also, the old glibc-2.0 compiled
1118 static libraries (libfoo.a) which happen to depend on the older libio
1119 behavior will be broken by the glibc 2.1 upgrade. We plan to produce a
1120 compatibility library that people will be able to link in if they want
1121 to compile a static library generated against glibc 2.0 into a program
1122 on a glibc 2.1 system. You just add -lcompat and you should be fine.
1124 The glibc-compat add-on will provide the libcompat.a library, the older
1125 nss modules, and a few other files. Together, they should make it
1126 possible to do development with old static libraries on a glibc 2.1
1127 system. This add-on is still in development. You can get it from
1128 <ftp://alpha.gnu.org/gnu/glibc/glibc-compat-2.1.tar.gz>
1129 but please keep in mind that it is experimental.
1132 2.28. Why is extracting files via tar so slow?
1134 {AJ} Extracting of tar archives might be quite slow since tar has to look up
1135 userid and groupids and doesn't cache negative results. If you have nis or
1136 nisplus in your /etc/nsswitch.conf for the passwd and/or group database,
1137 each file extractions needs a network connection. There are two possible
1140 - do you really need NIS/NIS+ (some Linux distributions add by default
1141 nis/nisplus even if it's not needed)? If not, just remove the entries.
1143 - if you need NIS/NIS+, use the Name Service Cache Daemon nscd that comes
1147 2.29. Compiling programs I get parse errors in libio.h (e.g. "parse error
1148 before `_IO_seekoff'"). How should I fix this?
1150 {AJ} You might get the following errors when upgrading to glibc 2.1:
1152 In file included from /usr/include/stdio.h:57,
1154 /usr/include/libio.h:335: parse error before `_IO_seekoff'
1155 /usr/include/libio.h:335: parse error before `_G_off64_t'
1156 /usr/include/libio.h:336: parse error before `_IO_seekpos'
1157 /usr/include/libio.h:336: parse error before `_G_fpos64_t'
1159 The problem is a wrong _G_config.h file in your include path. The
1160 _G_config.h file that comes with glibc 2.1 should be used and not one from
1161 libc5 or from a compiler directory. To check which _G_config.h file the
1162 compiler uses, compile your program with `gcc -E ...|grep G_config.h' and
1163 remove that file. Your compiler should pick up the file that has been
1164 installed by glibc 2.1 in your include directory.
1167 2.30. After upgrading to glibc 2.1, libraries that were compiled against
1168 glibc 2.0.x don't work anymore.
1170 {AJ} See question 2.27.
1173 2.31. What happened to the Berkeley DB libraries? Can I still use db
1174 in /etc/nsswitch.conf?
1176 {AJ} Due to too many incompatible changes in disk layout and API of Berkeley
1177 DB and a too tight coupling of libc and libdb, the db library has been
1178 removed completely from glibc 2.2. The only place that really used the
1179 Berkeley DB was the NSS db module.
1181 The NSS db module has been rewritten to support a number of different
1182 versions of Berkeley DB for the NSS db module. Currently the releases 2.x
1183 and 3.x of Berkeley DB are supported. The older db 1.85 library is not
1184 supported. You can use the version from glibc 2.1.x or download a version
1185 from Sleepycat Software (http://www.sleepycat.com). The library has to be
1186 compiled as shared library and installed in the system lib directory
1187 (normally /lib). The library needs to have a special soname to be found by
1190 If public structures change in a new Berkeley db release, this needs to be
1193 Currently the code searches for libraries with a soname of "libdb.so.3"
1194 (that's the name from db 2.4.14 which comes with glibc 2.1.x) and
1195 "libdb-3.0.so" (the name used by db 3.0.55 as default).
1198 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1200 3. Source and binary incompatibilities, and what to do about them
1202 3.1. I expect GNU libc to be 100% source code compatible with
1203 the old Linux based GNU libc. Why isn't it like this?
1205 {DMT,UD} Not every extension in Linux libc's history was well thought-out.
1206 In fact it had a lot of problems with standards compliance and with
1207 cleanliness. With the introduction of a new version number these errors can
1208 now be corrected. Here is a list of the known source code
1211 * _GNU_SOURCE: glibc does not make the GNU extensions available
1212 automatically. If a program depends on GNU extensions or some
1213 other non-standard functionality, it is necessary to compile it
1214 with the C compiler option -D_GNU_SOURCE, or better, to put
1215 `#define _GNU_SOURCE' at the beginning of your source files, before
1216 any C library header files are included. This difference normally
1217 manifests itself in the form of missing prototypes and/or data type
1218 definitions. Thus, if you get such errors, the first thing you
1219 should do is try defining _GNU_SOURCE and see if that makes the
1222 For more information consult the file `NOTES' in the GNU C library
1225 * reboot(): GNU libc sanitizes the interface of reboot() to be more
1226 compatible with the interface used on other OSes. reboot() as
1227 implemented in glibc takes just one argument. This argument
1228 corresponds to the third argument of the Linux reboot system call.
1229 That is, a call of the form reboot(a, b, c) needs to be changed into
1230 reboot(c). Beside this the header <sys/reboot.h> defines the needed
1231 constants for the argument. These RB_* constants should be used
1232 instead of the cryptic magic numbers.
1234 * swapon(): the interface of this function didn't change, but the
1235 prototype is in a separate header file <sys/swap.h>. This header
1236 file also provides the SWAP_* constants defined by <linux/swap.h>;
1237 you should use them for the second argument to swapon().
1239 * errno: If a program uses the variable "errno", then it _must_
1240 include <errno.h>. The old libc often (erroneously) declared this
1241 variable implicitly as a side-effect of including other libc header
1242 files. glibc is careful to avoid such namespace pollution, which,
1243 in turn, means that you really need to include the header files that
1244 you depend on. This difference normally manifests itself in the
1245 form of the compiler complaining about references to an undeclared
1248 * Linux-specific syscalls: All Linux system calls now have appropriate
1249 library wrappers and corresponding declarations in various header files.
1250 This is because the syscall() macro that was traditionally used to
1251 work around missing syscall wrappers are inherently non-portable and
1252 error-prone. The following table lists all the new syscall stubs,
1253 the header-file declaring their interface and the system call name.
1255 syscall name: wrapper name: declaring header file:
1256 ------------- ------------- ----------------------
1257 bdflush bdflush <sys/kdaemon.h>
1258 syslog ksyslog_ctl <sys/klog.h>
1260 * lpd: Older versions of lpd depend on a routine called _validuser().
1261 The library does not provide this function, but instead provides
1262 __ivaliduser() which has a slightly different interface. Simply
1263 upgrading to a newer lpd should fix this problem (e.g., the 4.4BSD
1264 lpd is known to be working).
1266 * resolver functions/BIND: like on many other systems the functions of
1267 the resolver library are not included in libc itself. There is a
1268 separate library libresolv. If you get undefined symbol errors for
1269 symbols starting with `res_*' simply add -lresolv to your linker
1272 * the `signal' function's behavior corresponds to the BSD semantic and
1273 not the SysV semantic as it was in libc-5. The interface on all GNU
1274 systems shall be the same and BSD is the semantic of choice. To use
1275 the SysV behavior simply use `sysv_signal', or define _XOPEN_SOURCE.
1276 See question 3.7 for details.
1279 3.2. Why does getlogin() always return NULL on my Linux box?
1281 {UD} The GNU C library has a format for the UTMP and WTMP file which differs
1282 from what your system currently has. It was extended to fulfill the needs
1283 of the next years when IPv6 is introduced. The record size is different and
1284 some fields have different positions. The files written by functions from
1285 the one library cannot be read by functions from the other library. Sorry,
1286 but this is what a major release is for. It's better to have a cut now than
1287 having no means to support the new techniques later.
1289 {MK} There is however a (partial) solution for this problem. Please take a
1290 look at the file `login/README.utmpd'.
1293 3.3. Where are the DST_* constants found in <sys/time.h> on many
1296 {UD} These constants come from the old BSD days and are not used anymore
1297 (libc5 does not actually implement the handling although the constants are
1300 Instead GNU libc contains zone database support and compatibility code for
1301 POSIX TZ environment variable handling. For former is very much preferred
1305 3.4. The prototypes for `connect', `accept', `getsockopt',
1306 `setsockopt', `getsockname', `getpeername', `send',
1307 `sendto', and `recvfrom' are different in GNU libc from
1308 any other system I saw. This is a bug, isn't it?
1310 {UD} No, this is no bug. This version of GNU libc already follows the new
1311 Single Unix specifications (and I think the POSIX.1g draft which adopted the
1312 solution). The type for a parameter describing a size is now `socklen_t', a
1316 3.5. On Linux I've got problems with the declarations in Linux
1319 {UD,AJ} On Linux, the use of kernel headers is reduced to the minimum. This
1320 gives Linus the ability to change the headers more freely. Also, user
1321 programs are now insulated from changes in the size of kernel data
1324 For example, the sigset_t type is 32 or 64 bits wide in the kernel. In
1325 glibc it is 1024 bits wide. This guarantees that when the kernel gets a
1326 bigger sigset_t (for POSIX.1e realtime support, say) user programs will not
1327 have to be recompiled. Consult the header files for more information about
1330 Therefore you shouldn't include Linux kernel header files directly if glibc
1331 has defined a replacement. Otherwise you might get undefined results because
1335 3.6. I don't include any kernel headers myself but the compiler
1336 still complains about redeclarations of types in the kernel
1339 {UD} The kernel headers before Linux 2.1.61 and 2.0.32 don't work correctly
1340 with glibc. Compiling C programs is possible in most cases but C++ programs
1341 have (due to the change of the name lookups for `struct's) problems. One
1342 prominent example is `struct fd_set'.
1344 There might be some problems left but 2.1.61/2.0.32 fix most of the known
1345 ones. See the BUGS file for other known problems.
1348 3.7. Why don't signals interrupt system calls anymore?
1350 {ZW} By default GNU libc uses the BSD semantics for signal(), unlike Linux
1351 libc 5 which used System V semantics. This is partially for compatibility
1352 with other systems and partially because the BSD semantics tend to make
1353 programming with signals easier.
1355 There are three differences:
1357 * BSD-style signals that occur in the middle of a system call do not
1358 affect the system call; System V signals cause the system call to
1359 fail and set errno to EINTR.
1361 * BSD signal handlers remain installed once triggered. System V signal
1362 handlers work only once, so one must reinstall them each time.
1364 * A BSD signal is blocked during the execution of its handler. In other
1365 words, a handler for SIGCHLD (for example) does not need to worry about
1366 being interrupted by another SIGCHLD. It may, however, be interrupted
1369 There is general consensus that for `casual' programming with signals, the
1370 BSD semantics are preferable. You don't need to worry about system calls
1371 returning EINTR, and you don't need to worry about the race conditions
1372 associated with one-shot signal handlers.
1374 If you are porting an old program that relies on the old semantics, you can
1375 quickly fix the problem by changing signal() to sysv_signal() throughout.
1376 Alternatively, define _XOPEN_SOURCE before including <signal.h>.
1378 For new programs, the sigaction() function allows you to specify precisely
1379 how you want your signals to behave. All three differences listed above are
1380 individually switchable on a per-signal basis with this function.
1382 If all you want is for one specific signal to cause system calls to fail and
1383 return EINTR (for example, to implement a timeout) you can do this with
1387 3.8. I've got errors compiling code that uses certain string
1390 {AJ} glibc 2.1 has special string functions that are faster than the normal
1391 library functions. Some of the functions are additionally implemented as
1392 inline functions and others as macros. This might lead to problems with
1393 existing codes but it is explicitly allowed by ISO C.
1395 The optimized string functions are only used when compiling with
1396 optimizations (-O1 or higher). The behavior can be changed with two feature
1399 * __NO_STRING_INLINES: Don't do any string optimizations.
1400 * __USE_STRING_INLINES: Use assembly language inline functions (might
1401 increase code size dramatically).
1403 Since some of these string functions are now additionally defined as macros,
1404 code like "char *strncpy();" doesn't work anymore (and is unnecessary, since
1405 <string.h> has the necessary declarations). Either change your code or
1406 define __NO_STRING_INLINES.
1408 {UD} Another problem in this area is that gcc still has problems on machines
1409 with very few registers (e.g., ix86). The inline assembler code can require
1410 almost all the registers and the register allocator cannot always handle
1413 One can disable the string optimizations selectively. Instead of writing
1415 cp = strcpy (foo, "lkj");
1419 cp = (strcpy) (foo, "lkj");
1421 This disables the optimization for that specific call.
1424 3.9. I get compiler messages "Initializer element not constant" with
1425 stdin/stdout/stderr. Why?
1427 {RM,AJ} Constructs like:
1428 static FILE *InPtr = stdin;
1430 lead to this message. This is correct behaviour with glibc since stdin is
1431 not a constant expression. Please note that a strict reading of ISO C does
1432 not allow above constructs.
1434 One of the advantages of this is that you can assign to stdin, stdout, and
1435 stderr just like any other global variable (e.g. `stdout = my_stream;'),
1436 which can be very useful with custom streams that you can write with libio
1437 (but beware this is not necessarily portable). The reason to implement it
1438 this way were versioning problems with the size of the FILE structure.
1440 To fix those programs you've got to initialize the variable at run time.
1441 This can be done, e.g. in main, like:
1449 or by constructors (beware this is gcc specific):
1452 static void inPtr_construct (void) __attribute__((constructor));
1453 static void inPtr_construct (void) { InPtr = stdin; }
1456 3.10. I can't compile with gcc -traditional (or
1457 -traditional-cpp). Why?
1459 {AJ} glibc2 does break -traditional and -traditonal-cpp - and will continue
1460 to do so. For example constructs of the form:
1466 are useful for debugging purposes (you can use foo with your debugger that's
1467 why we need the enum) and for compatibility (other systems use defines and
1471 3.11. I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible?
1473 {AJ} The GNU C library is compatible with the ANSI/ISO C standard. If
1474 you're using `gcc -ansi', the glibc includes which are specified in the
1475 standard follow the standard. The ANSI/ISO C standard defines what has to be
1476 in the include files - and also states that nothing else should be in the
1477 include files (btw. you can still enable additional standards with feature
1480 The GNU C library is conforming to ANSI/ISO C - if and only if you're only
1481 using the headers and library functions defined in the standard.
1484 3.12. I can't access some functions anymore. nm shows that they do
1485 exist but linking fails nevertheless.
1487 {AJ} With the introduction of versioning in glibc 2.1 it is possible to
1488 export only those identifiers (functions, variables) that are really needed
1489 by application programs and by other parts of glibc. This way a lot of
1490 internal interfaces are now hidden. nm will still show those identifiers
1491 but marking them as internal. ISO C states that identifiers beginning with
1492 an underscore are internal to the libc. An application program normally
1493 shouldn't use those internal interfaces (there are exceptions,
1494 e.g. __ivaliduser). If a program uses these interfaces, it's broken. These
1495 internal interfaces might change between glibc releases or dropped
1499 3.13. When using the db-2 library which comes with glibc is used in
1500 the Perl db modules the testsuite is not passed. This did not
1501 happen with db-1, gdbm, or ndbm.
1503 Removed. Does not apply anymore.
1508 3.14. The pow() inline function I get when including <math.h> is broken.
1509 I get segmentation faults when I run the program.
1511 {UD} Nope, the implementation is correct. The problem is with egcs version
1512 prior to 1.1. I.e., egcs 1.0 to 1.0.3 are all broken (at least on Intel).
1513 If you have to use this compiler you must define __NO_MATH_INLINES before
1514 including <math.h> to prevent the inline functions from being used. egcs 1.1
1515 fixes the problem. I don't know about gcc 2.8 and 2.8.1.
1518 3.15. The sys/sem.h file lacks the definition of `union semun'.
1520 {UD} Nope. This union has to be provided by the user program. Former glibc
1521 versions defined this but it was an error since it does not make much sense
1522 when thinking about it. The standards describing the System V IPC functions
1523 define it this way and therefore programs must be adopted.
1526 3.16. Why has <netinet/ip_fw.h> disappeared?
1528 {AJ} The corresponding Linux kernel data structures and constants are
1529 totally different in Linux 2.0 and Linux 2.2. This situation has to be
1530 taken care in user programs using the firewall structures and therefore
1531 those programs (ipfw is AFAIK the only one) should deal with this problem
1535 3.17. I get floods of warnings when I use -Wconversion and include
1536 <string.h> or <math.h>.
1538 {ZW} <string.h> and <math.h> intentionally use prototypes to override
1539 argument promotion. -Wconversion warns about all these. You can safely
1540 ignore the warnings.
1542 -Wconversion isn't really intended for production use, only for shakedown
1543 compiles after converting an old program to standard C.
1546 3.18. After upgrading to glibc 2.1, I receive errors about
1547 unresolved symbols, like `_dl_initial_searchlist' and can not
1548 execute any binaries. What went wrong?
1550 {AJ} This normally happens if your libc and ld (dynamic linker) are from
1551 different releases of glibc. For example, the dynamic linker
1552 /lib/ld-linux.so.2 comes from glibc 2.0.x, but the version of libc.so.6 is
1555 The path /lib/ld-linux.so.2 is hardcoded in every glibc2 binary but
1556 libc.so.6 is searched via /etc/ld.so.cache and in some special directories
1557 like /lib and /usr/lib. If you run configure with another prefix than /usr
1558 and put this prefix before /lib in /etc/ld.so.conf, your system will break.
1560 So what can you do? Either of the following should work:
1562 * Run `configure' with the same prefix argument you've used for glibc 2.0.x
1563 so that the same paths are used.
1564 * Replace /lib/ld-linux.so.2 with a link to the dynamic linker from glibc
1567 You can even call the dynamic linker by hand if everything fails. You've
1568 got to set LD_LIBRARY_PATH so that the corresponding libc is found and also
1569 need to provide an absolute path to your binary:
1571 LD_LIBRARY_PATH=<path-where-libc.so.6-lives> \
1572 <path-where-corresponding-dynamic-linker-lives>/ld-linux.so.2 \
1573 <path-to-binary>/binary
1575 For example `LD_LIBRARY_PATH=/libold /libold/ld-linux.so.2 /bin/mv ...'
1576 might be useful in fixing a broken system (if /libold contains dynamic
1577 linker and corresponding libc).
1579 With that command line no path is used. To further debug problems with the
1580 dynamic linker, use the LD_DEBUG environment variable, e.g.
1581 `LD_DEBUG=help echo' for the help text.
1583 If you just want to test this release, don't put the lib directory in
1584 /etc/ld.so.conf. You can call programs directly with full paths (as above).
1585 When compiling new programs against glibc 2.1, you've got to specify the
1586 correct paths to the compiler (option -I with gcc) and linker (options
1587 --dynamic-linker, -L and --rpath).
1590 3.19. bonnie reports that char i/o with glibc 2 is much slower than with
1591 libc5. What can be done?
1593 {AJ} The GNU C library uses thread safe functions by default and libc5 used
1594 non thread safe versions. The non thread safe functions have in glibc the
1595 suffix `_unlocked', for details check <stdio.h>. Using `putc_unlocked' etc.
1596 instead of `putc' should give nearly the same speed with bonnie (bonnie is a
1597 benchmark program for measuring disk access).
1600 3.20. Programs compiled with glibc 2.1 can't read db files made with glibc
1601 2.0. What has changed that programs like rpm break?
1603 Removed. Does not apply anymore.
1608 3.21. Autoconf's AC_CHECK_FUNC macro reports that a function exists, but
1609 when I try to use it, it always returns -1 and sets errno to ENOSYS.
1611 {ZW} You are using a 2.0 Linux kernel, and the function you are trying to
1612 use is only implemented in 2.1/2.2. Libc considers this to be a function
1613 which exists, because if you upgrade to a 2.2 kernel, it will work. One
1614 such function is sigaltstack.
1616 Your program should check at runtime whether the function works, and
1617 implement a fallback. Note that Autoconf cannot detect unimplemented
1618 functions in other systems' C libraries, so you need to do this anyway.
1621 3.22. My program segfaults when I call fclose() on the FILE* returned
1622 from setmntent(). Is this a glibc bug?
1624 {GK} No. Don't do this. Use endmntent(), that's what it's for.
1626 In general, you should use the correct deallocation routine. For instance,
1627 if you open a file using fopen(), you should deallocate the FILE * using
1628 fclose(), not free(), even though the FILE * is also a pointer.
1630 In the case of setmntent(), it may appear to work in most cases, but it
1631 won't always work. Unfortunately, for compatibility reasons, we can't
1632 change the return type of setmntent() to something other than FILE *.
1635 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1639 4.1. After I changed configure.in I get `Autoconf version X.Y.
1640 or higher is required for this script'. What can I do?
1642 {UD} You have to get the specified autoconf version (or a later one)
1643 from your favorite mirror of ftp.gnu.org.
1646 4.2. When I try to compile code which uses IPv6 headers and
1647 definitions on my Linux 2.x.y system I am in trouble.
1648 Nothing seems to work.
1650 {UD} The problem is that IPv6 development still has not reached a point
1651 where the headers are stable. There are still lots of incompatible changes
1652 made and the libc headers have to follow.
1654 {PB} The 2.1 release of GNU libc aims to comply with the current versions of
1655 all the relevant standards. The IPv6 support libraries for older Linux
1656 systems used a different naming convention and so code written to work with
1657 them may need to be modified. If the standards make incompatible changes in
1658 the future then the libc may need to change again.
1660 IPv6 will not work with a 2.0.x kernel. When kernel 2.2 is released it
1661 should contain all the necessary support; until then you should use the
1662 latest 2.1.x release you can find. As of 98/11/26 the currently recommended
1663 kernel for IPv6 is 2.1.129.
1665 Also, as of the 2.1 release the IPv6 API provided by GNU libc is not
1666 100% complete. In particular the getipnodebyname and getipnodebyaddr
1667 functions are not implemented.
1670 4.3. When I set the timezone by setting the TZ environment variable
1671 to EST5EDT things go wrong since glibc computes the wrong time
1672 from this information.
1674 {UD} The problem is that people still use the braindamaged POSIX method to
1675 select the timezone using the TZ environment variable with a format EST5EDT
1676 or whatever. People, if you insist on using TZ instead of the timezone
1677 database (see below), read the POSIX standard, the implemented behaviour is
1678 correct! What you see is in fact the result of the decisions made while
1679 POSIX.1 was created. We've only implemented the handling of TZ this way to
1680 be POSIX compliant. It is not really meant to be used.
1682 The alternative approach to handle timezones which is implemented is the
1683 correct one to use: use the timezone database. This avoids all the problems
1684 the POSIX method has plus it is much easier to use. Simply run the tzselect
1685 shell script, answer the question and use the name printed in the end by
1686 making a symlink /etc/localtime pointing to /usr/share/zoneinfo/NAME (NAME
1687 is the returned value from tzselect). That's all. You never again have to
1690 So, please avoid sending bug reports about time related problems if you use
1691 the POSIX method and you have not verified something is really broken by
1692 reading the POSIX standards.
1695 4.4. What other sources of documentation about glibc are available?
1697 {AJ} The FSF has a page about the GNU C library at
1698 <http://www.gnu.org/software/libc/>. The problem data base of open and
1699 solved bugs in GNU libc is available at
1700 <http://www-gnats.gnu.org:8080/cgi-bin/wwwgnats.pl>. Eric Green has written
1701 a HowTo for converting from Linux libc5 to glibc2. The HowTo is accessible
1702 via the FSF page and at <http://www.imaxx.net/~thrytis/glibc>. Frodo
1703 Looijaard describes a different way installing glibc2 as secondary libc at
1704 <http://huizen.dds.nl/~frodol/glibc>.
1706 Please note that this is not a complete list.
1709 4.5. The timezone string for Sydney/Australia is wrong since even when
1710 daylight saving time is in effect the timezone string is EST.
1712 {UD} The problem for some timezones is that the local authorities decided
1713 to use the term "summer time" instead of "daylight saving time". In this
1714 case the abbreviation character `S' is the same as the standard one. So,
1717 Eastern Standard Time = EST
1718 Eastern Summer Time = EST
1720 Great! To get this bug fixed convince the authorities to change the laws
1721 and regulations of the country this effects. glibc behaves correctly.
1724 4.6. I've build make 3.77 against glibc 2.1 and now make gets
1725 segmentation faults.
1727 {AJ} GNU make 3.77 has support for 64 bit filesystems which is slightly
1728 broken (and one of the new features in the GNU C library 2.1 is 64 bit
1729 filesystem support :-( ). To get a working make you can use either make
1730 3.75 or patch 3.77. A working patch is available via RedHat's Rawhide server
1731 (ftp://rawhide.redhat.com/SRPMS/SRPMS/make-3.77-*src.rpm).
1734 4.7. Why do so many programs using math functions fail on my AlphaStation?
1736 {AO} The functions floor() and floorf() use an instruction that is not
1737 implemented in some old PALcodes of AlphaStations. This may cause
1738 `Illegal Instruction' core dumps or endless loops in programs that
1739 catch these signals. Updating the firmware to a 1999 release has
1740 fixed the problem on an AlphaStation 200 4/166.
1743 4.8. The conversion table for character set XX does not match with
1746 {UD} I don't doubt for a minute that some of the conversion tables contain
1747 errors. We tried the best we can and relied on automatic generation of the
1748 data to prevent human-introduced errors but this still is no guarantee. If
1749 you think you found a problem please send a bug report describing it and
1750 give an authoritive reference. The latter is important since otherwise
1751 the current behaviour is as good as the proposed one.
1753 Before doing this look through the list of known problem first:
1755 - the GBK (simplified Chinese) encoding is based on Unicode tables. This
1756 is good. These tables, however, differ slightly from the tables used
1757 by the M$ people. The differences are these [+ Unicode, - M$]:
1764 In addition the Unicode tables contain mappings for the GBK characters
1765 0xA8BC, 0xA8BF, 0xA989 to 0xA995, and 0xFE50 to 0xFEA0.
1767 - when mapping from EUC-CN to GBK and vice versa we ignore the fact that
1768 the coded character at position 0xA1A4 maps to different Unicode
1769 characters. Since the iconv() implementation can do whatever it wants
1770 if it cannot directly map a character this is a perfectly good solution
1771 since the semantics and appearance of the character does not change.
1774 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
1776 Answers were given by:
1777 {UD} Ulrich Drepper, <drepper@cygnus.com>
1778 {DMT} David Mosberger-Tang, <davidm@AZStarNet.com>
1779 {RM} Roland McGrath, <roland@gnu.org>
1780 {AJ} Andreas Jaeger, <aj@suse.de>
1781 {EY} Eric Youngdale, <eric@andante.jic.com>
1782 {PB} Phil Blundell, <Philip.Blundell@pobox.com>
1783 {MK} Mark Kettenis, <kettenis@phys.uva.nl>
1784 {ZW} Zack Weinberg, <zack@rabi.phys.columbia.edu>
1785 {TK} Thorsten Kukuk, <kukuk@suse.de>
1786 {GK} Geoffrey Keating, <geoffk@ozemail.com.au>
1787 {HJ} H.J. Lu, <hjl@gnu.org>
1788 {CG} Cristian Gafton, <gafton@redhat.com>
1789 {AO} Alexandre Oliva, <oliva@lsd.ic.unicamp.br>
1793 outline-regexp:"\\?"