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
47 1.19. `make' complains about a missing dlfcn/libdl.so when building
48 malloc/libmemprof.so. How can I fix this?
49 1.20. Which tools should I use for MIPS?
51 2. Installation and configuration issues
53 2.1. Can I replace the libc on my Linux system with GNU libc?
54 2.2. How do I configure GNU libc so that the essential libraries
55 like libc.so go into /lib and the other into /usr/lib?
56 2.3. How should I avoid damaging my system when I install GNU libc?
57 2.4. Do I need to use GNU CC to compile programs that will use the
59 2.5. When linking with the new libc I get unresolved symbols
60 `crypt' and `setkey'. Why aren't these functions in the
62 2.6. When I use GNU libc on my Linux system by linking against
63 the libc.so which comes with glibc all I get is a core dump.
64 2.7. Looking through the shared libc file I haven't found the
65 functions `stat', `lstat', `fstat', and `mknod' and while
66 linking on my Linux system I get error messages. How is
67 this supposed to work?
68 2.8. When I run an executable on one system which I compiled on
69 another, I get dynamic linker errors. Both systems have the same
70 version of glibc installed. What's wrong?
71 2.9. How can I compile gcc 2.7.2.1 from the gcc source code using
73 2.10. The `gencat' utility cannot process the catalog sources which
74 were used on my Linux libc5 based system. Why?
75 2.11. Programs using libc have their messages translated, but other
76 behavior is not localized (e.g. collating order); why?
77 2.12. I have set up /etc/nis.conf, and the Linux libc 5 with NYS
78 works great. But the glibc NIS+ doesn't seem to work.
79 2.13. I have killed ypbind to stop using NIS, but glibc
81 2.14. Under Linux/Alpha, I always get "do_ypcall: clnt_call:
82 RPC: Unable to receive; errno = Connection refused" when using NIS.
83 2.15. After installing glibc name resolving doesn't work properly.
84 2.16. How do I create the databases for NSS?
85 2.17. I have /usr/include/net and /usr/include/scsi as symlinks
86 into my Linux source tree. Is that wrong?
87 2.18. Programs like `logname', `top', `uptime' `users', `w' and
88 `who', show incorrect information about the (number of)
89 users on my system. Why?
90 2.19. After upgrading to glibc 2.1 with symbol versioning I get
91 errors about undefined symbols. What went wrong?
92 2.20. When I start the program XXX after upgrading the library
94 XXX: Symbol `_sys_errlist' has different size in shared
95 object, consider re-linking
96 Why? What should I do?
97 2.21. What do I need for C++ development?
98 2.22. Even statically linked programs need some shared libraries
99 which is not acceptable for me. What can I do?
100 2.23. I just upgraded my Linux system to glibc and now I get
101 errors whenever I try to link any program.
102 2.24. When I use nscd the machine freezes.
103 2.25. I need lots of open files. What do I have to do?
104 2.26. How do I get the same behavior on parsing /etc/passwd and
105 /etc/group as I have with libc5 ?
106 2.27. What needs to be recompiled when upgrading from glibc 2.0 to glibc
108 2.28. Why is extracting files via tar so slow?
109 2.29. Compiling programs I get parse errors in libio.h (e.g. "parse error
110 before `_IO_seekoff'"). How should I fix this?
111 2.30. After upgrading to glibc 2.1, libraries that were compiled against
112 glibc 2.0.x don't work anymore.
113 2.31. What happened to the Berkeley DB libraries? Can I still use db
114 in /etc/nsswitch.conf?
115 2.32. What has do be done when upgrading to glibc 2.2?
116 2.33. The makefiles want to do a CVS commit.
117 2.34. When compiling C++ programs, I get a compilation error in streambuf.h.
118 2.35. When recompiling GCC, I get compilation errors in libio.
120 3. Source and binary incompatibilities, and what to do about them
122 3.1. I expect GNU libc to be 100% source code compatible with
123 the old Linux based GNU libc. Why isn't it like this?
124 3.2. Why does getlogin() always return NULL on my Linux box?
125 3.3. Where are the DST_* constants found in <sys/time.h> on many
127 3.4. The prototypes for `connect', `accept', `getsockopt',
128 `setsockopt', `getsockname', `getpeername', `send',
129 `sendto', and `recvfrom' are different in GNU libc from
130 any other system I saw. This is a bug, isn't it?
131 3.5. On Linux I've got problems with the declarations in Linux
133 3.6. I don't include any kernel headers myself but the compiler
134 still complains about redeclarations of types in the kernel
136 3.7. Why don't signals interrupt system calls anymore?
137 3.8. I've got errors compiling code that uses certain string
139 3.9. I get compiler messages "Initializer element not constant" with
140 stdin/stdout/stderr. Why?
141 3.10. I can't compile with gcc -traditional (or
142 -traditional-cpp). Why?
143 3.11. I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible?
144 3.12. I can't access some functions anymore. nm shows that they do
145 exist but linking fails nevertheless.
146 3.13. When using the db-2 library which comes with glibc is used in
147 the Perl db modules the testsuite is not passed. This did not
148 happen with db-1, gdbm, or ndbm.
149 3.14. The pow() inline function I get when including <math.h> is broken.
150 I get segmentation faults when I run the program.
151 3.15. The sys/sem.h file lacks the definition of `union semun'.
152 3.16. Why has <netinet/ip_fw.h> disappeared?
153 3.17. I get floods of warnings when I use -Wconversion and include
154 <string.h> or <math.h>.
155 3.18. After upgrading to glibc 2.1, I receive errors about
156 unresolved symbols, like `_dl_initial_searchlist' and can not
157 execute any binaries. What went wrong?
158 3.19. bonnie reports that char i/o with glibc 2 is much slower than with
159 libc5. What can be done?
160 3.20. Programs compiled with glibc 2.1 can't read db files made with glibc
161 2.0. What has changed that programs like rpm break?
162 3.21. Autoconf's AC_CHECK_FUNC macro reports that a function exists, but
163 when I try to use it, it always returns -1 and sets errno to ENOSYS.
164 3.22. My program segfaults when I call fclose() on the FILE* returned
165 from setmntent(). Is this a glibc bug?
169 4.1. After I changed configure.in I get `Autoconf version X.Y.
170 or higher is required for this script'. What can I do?
171 4.2. When I try to compile code which uses IPv6 headers and
172 definitions on my Linux 2.x.y system I am in trouble.
173 Nothing seems to work.
174 4.3. When I set the timezone by setting the TZ environment variable
175 to EST5EDT things go wrong since glibc computes the wrong time
176 from this information.
177 4.4. What other sources of documentation about glibc are available?
178 4.5. The timezone string for Sydney/Australia is wrong since even when
179 daylight saving time is in effect the timezone string is EST.
180 4.6. I've build make 3.77 against glibc 2.1 and now make gets
182 4.7. Why do so many programs using math functions fail on my AlphaStation?
183 4.8. The conversion table for character set XX does not match with
187 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
191 1.1. What systems does the GNU C Library run on?
193 {UD} This is difficult to answer. The file `README' lists the architectures
194 GNU libc was known to run on *at some time*. This does not mean that it
195 still can be compiled and run on them now.
197 The systems glibc is known to work on as of this release, and most probably
201 i[3456]86-*-linux-gnu Linux-2.x on Intel
202 m68k-*-linux-gnu Linux-2.x on Motorola 680x0
203 alpha*-*-linux-gnu Linux-2.x on DEC Alpha
204 powerpc-*-linux-gnu Linux and MkLinux on PowerPC systems
205 sparc-*-linux-gnu Linux-2.x on SPARC
206 sparc64-*-linux-gnu Linux-2.x on UltraSPARC
207 arm-*-none ARM standalone systems
208 arm-*-linux Linux-2.x on ARM
209 arm-*-linuxaout Linux-2.x on ARM using a.out binaries
210 mips*-*-linux-gnu Linux-2.x on MIPS
211 ia64-*-linux-gnu Linux-2.x on ia64
212 s390-*-linux-gnu Linux-2.x on IBM S/390
214 Ports to other Linux platforms are in development, and may in fact work
215 already, but no one has sent us success reports for them. Currently no
216 ports to other operating systems are underway, although a few people have
219 If you have a system not listed above (or in the `README' file) and you are
220 really interested in porting it, contact
225 1.2. What compiler do I need to build GNU libc?
227 {UD} You must use GNU CC to compile GNU libc. A lot of extensions of GNU CC
228 are used to increase portability and speed.
230 GNU CC is found, like all other GNU packages, on
232 ftp://ftp.gnu.org/pub/gnu
234 and the many mirror sites. ftp.gnu.org is always overloaded, so try to find
235 a local mirror first.
237 You should always try to use the latest official release. Older versions
238 may not have all the features GNU libc requires. The current releases of
239 egcs (1.0.3 and 1.1.1) should work with the GNU C library (for powerpc see
240 question 1.5; for ARM see question 1.6; for MIPS see question 1.20).
242 While the GNU CC should be able to compile glibc it is nevertheless adviced
243 to use EGCS. Comparing the sizes of glibc on Intel compiled with a recent
244 EGCS and gcc 2.8.1 shows this:
246 text data bss dec hex filename
247 egcs-2.93.10 862897 15944 12824 891665 d9b11 libc.so
248 gcc-2.8.1 959965 16468 12152 988585 f15a9 libc.so
250 Make up your own decision.
252 GNU CC versions 2.95 and above are derived from egcs, and they may do even
255 Please note that gcc 2.95 and 2.95.x cannot compile glibc on Alpha due to
256 problems in the complex float support.
259 1.3. When I try to compile glibc I get only error messages.
262 {UD} You definitely need GNU make to build GNU libc. No other make
263 program has the needed functionality.
265 We recommend version GNU make version 3.79 or newer. Older versions have
266 bugs and/or are missing features.
269 1.4. Do I need a special linker or assembler?
271 {ZW} If you want a shared library, you need a linker and assembler that
272 understand all the features of ELF, including weak and versioned symbols.
273 The static library can be compiled with less featureful tools, but lacks key
274 features such as NSS.
276 For Linux or Hurd, you want binutils 2.8.1.0.23, 2.9.1, or 2.9.1.0.15 or
277 higher. These are the only versions we've tested and found reliable. Other
278 versions after 2.8.1.0.23 may work but we don't recommend them, especially
279 not when C++ is involved. Earlier versions do not work at all.
281 Other operating systems may come with system tools that have all the
282 necessary features, but this is moot because glibc hasn't been ported to
286 1.5. Which compiler should I use for powerpc?
288 {GK} You want to use at least gcc 2.95 (together with the right versions
289 of all the other tools, of course). See also question question 2.8.
292 1.6. Which tools should I use for ARM?
294 {PB} You should use egcs 1.1 or a later version. For ELF systems some
295 changes are needed to the compiler; a patch against egcs-1.1.x can be found
298 <ftp://ftp.netwinder.org/users/p/philb/egcs-1.1.1pre2-diff-981126>
300 Binutils 2.9.1.0.16 or later is also required.
303 1.7. Do I need some more things to compile the GNU C Library?
305 {UD} Yes, there are some more :-).
307 * GNU gettext. This package contains the tools needed to construct
308 `message catalog' files containing translated versions of system
309 messages. See ftp://ftp.gnu.org/pub/gnu or better any mirror
310 site. (We distribute compiled message catalogs, but they may not be
311 updated in patches.) Please note that the required minimal version
312 (0.10.35) of gettext is alpha software and available from
313 ftp://alpha.gnu.org/gnu .
315 * Some files are built with special tools. E.g., files ending in .gperf
316 need a `gperf' program. The GNU version (now available in a separate
317 package, formerly only as part of libg++) is known to work while some
318 vendor versions do not.
320 You should not need these tools unless you change the source files.
322 * Perl 5 is needed if you wish to test an installation of GNU libc
323 as the primary C library.
325 * When compiling for Linux, the header files of the Linux kernel must
326 be available to the compiler as <linux/*.h> and <asm/*.h>.
328 * lots of disk space (~400MB for i?86-linux; more for RISC platforms).
330 * plenty of time. Compiling just the shared and static libraries for
331 i?86-linux takes approximately 1h on an AMD-K6@225MHz w/ 96MB of RAM,
332 45mins on a Celeron@400MHz w/ 128MB, and 55mins on a Alpha@533MHz w/ 256MB.
333 Multiply this by 1.5 or 2.0 if you build profiling and/or the highly
334 optimized version as well. For Hurd systems times are much higher.
336 You should avoid compiling in a NFS mounted filesystem. This is
339 James Troup <J.J.Troup@comp.brad.ac.uk> reports a compile time of
340 45h34m for a full build (shared, static, and profiled) on Atari
341 Falcon (Motorola 68030 @ 16 Mhz, 14 Mb memory) and Jan Barte
342 <yann@plato.uni-paderborn.de> reports 22h48m on Atari TT030
343 (Motorola 68030 @ 32 Mhz, 34 Mb memory)
345 A full build of the PowerPC library took 1h on a PowerPC 750@400Mhz w/
346 64MB of RAM, and about 9h on a 601@60Mhz w/ 72Mb.
348 If you have some more measurements let me know.
351 1.8. What version of the Linux kernel headers should be used?
353 {AJ,UD} The headers from the most recent Linux kernel should be used. The
354 headers used while compiling the GNU C library and the kernel binary used
355 when using the library do not need to match. The GNU C library runs without
356 problems on kernels that are older than the kernel headers used. The other
357 way round (compiling the GNU C library with old kernel headers and running
358 on a recent kernel) does not necessarily work. For example you can't use
359 new kernel features if you used old kernel headers to compile the GNU C
362 {ZW} Even if you are using a 2.0 kernel on your machine, we recommend you
363 compile GNU libc with 2.2 kernel headers. That way you won't have to
364 recompile libc if you ever upgrade to kernel 2.2. To tell libc which
365 headers to use, give configure the --with-headers switch
366 (e.g. --with-headers=/usr/src/linux-2.2.0/include).
368 Note that you must configure the 2.2 kernel if you do this, otherwise libc
369 will be unable to find <linux/version.h>. Just change the current directory
370 to the root of the 2.2 tree and do `make include/linux/version.h'.
373 1.9. The compiler hangs while building iconvdata modules. What's
376 {ZW} This is a problem with old versions of GCC. Initialization of large
377 static arrays is very slow. The compiler will eventually finish; give it
380 The problem is fixed in egcs 1.1.
383 1.10. When I run `nm -u libc.so' on the produced library I still
384 find unresolved symbols. Can this be ok?
386 {UD} Yes, this is ok. There can be several kinds of unresolved symbols:
388 * magic symbols automatically generated by the linker. These have names
389 like __start_* and __stop_*
391 * symbols starting with _dl_* come from the dynamic linker
393 * weak symbols, which need not be resolved at all (fabs for example)
395 Generally, you should make sure you find a real program which produces
396 errors while linking before deciding there is a problem.
399 1.11. What are these `add-ons'?
401 {UD} To avoid complications with export rules or external source code some
402 optional parts of the libc are distributed as separate packages, e.g., the
403 linuxthreads package.
405 To use these packages as part of GNU libc, just unpack the tarfiles in the
406 libc source directory and tell the configuration script about them using the
407 --enable-add-ons option. If you give just --enable-add-ons configure tries
408 to find all the add-on packages in your source tree. This may not work. If
409 it doesn't, or if you want to select only a subset of the add-ons, give a
410 comma-separated list of the add-ons to enable:
412 configure --enable-add-ons=linuxthreads
416 Add-ons can add features (including entirely new shared libraries), override
417 files, provide support for additional architectures, and just about anything
418 else. The existing makefiles do most of the work; only some few stub rules
419 must be written to get everything running.
421 Most add-ons are tightly coupled to a specific GNU libc version. Please
422 check that the add-ons work with the GNU libc. For example the linuxthreads
423 add-on has the same numbering scheme as the libc and will in general only
424 work with the corresponding libc.
426 {AJ} With glibc 2.2 the crypt add-on and with glibc 2.1 the localedata
427 add-on have been integrated into the normal glibc distribution, crypt and
428 localedata are therefore not anymore add-ons.
431 1.12. My XXX kernel emulates a floating-point coprocessor for me.
432 Should I enable --with-fp?
434 {ZW} An emulated FPU is just as good as a real one, as far as the C library
435 is concerned. You only need to say --without-fp if your machine has no way
436 to execute floating-point instructions.
438 People who are interested in squeezing the last drop of performance
439 out of their machine may wish to avoid the trap overhead, but this is
440 far more trouble than it's worth: you then have to compile
441 *everything* this way, including the compiler's internal libraries
442 (libgcc.a for GNU C), because the calling conventions change.
445 1.13. When compiling GNU libc I get lots of errors saying functions
446 in glibc are duplicated in libgcc.
448 {EY} This is *exactly* the same problem that I was having. The problem was
449 due to the fact that configure didn't correctly detect that the linker flag
450 --no-whole-archive was supported in my linker. In my case it was because I
451 had run ./configure with bogus CFLAGS, and the test failed.
453 One thing that is particularly annoying about this problem is that once this
454 is misdetected, running configure again won't fix it unless you first delete
457 {UD} Starting with glibc-2.0.3 there should be a better test to avoid some
458 problems of this kind. The setting of CFLAGS is checked at the very
459 beginning and if it is not usable `configure' will bark.
462 1.14. Why do I get messages about missing thread functions when I use
463 librt? I don't even use threads.
465 {UD} In this case you probably mixed up your installation. librt uses
466 threads internally and has implicit references to the thread library.
467 Normally these references are satisfied automatically but if the thread
468 library is not in the expected place you must tell the linker where it is.
469 When using GNU ld it works like this:
471 gcc -o foo foo.c -Wl,-rpath-link=/some/other/dir -lrt
473 The `/some/other/dir' should contain the thread library. `ld' will use the
474 given path to find the implicitly referenced library while not disturbing
478 1.15. What's the problem with configure --enable-omitfp?
480 {AJ} When --enable-omitfp is set the libraries are built without frame
481 pointers. Some compilers produce buggy code for this model and therefore we
482 don't advise using it at the moment.
484 If you use --enable-omitfp, you're on your own. If you encounter problems
485 with a library that was build this way, we advise you to rebuild the library
486 without --enable-omitfp. If the problem vanishes consider tracking the
487 problem down and report it as compiler failure.
489 Since a library built with --enable-omitfp is undebuggable on most systems,
490 debuggable libraries are also built - you can use them by appending "_g" to
493 The compilation of these extra libraries and the compiler optimizations slow
494 down the build process and need more disk space.
497 1.16. I get failures during `make check'. What should I do?
499 {AJ} The testsuite should compile and run cleanly on your system; every
500 failure should be looked into. Depending on the failures, you probably
501 should not install the library at all.
503 You should consider using the `glibcbug' script to report the failure,
504 providing as much detail as possible. If you run a test directly, please
505 remember to set up the environment correctly. You want to test the compiled
506 library - and not your installed one. The best way is to copy the exact
507 command line which failed and run the test from the subdirectory for this
510 There are some failures which are not directly related to the GNU libc:
511 - Some compilers produce buggy code. No compiler gets single precision
512 complex numbers correct on Alpha. Otherwise, the egcs 1.1 release should be
513 ok; gcc 2.8.1 might cause some failures; gcc 2.7.2.x is so buggy that
514 explicit checks have been used so that you can't build with it.
515 - The kernel might have bugs. For example on Linux/Alpha 2.0.34 the
516 floating point handling has quite a number of bugs and therefore most of
517 the test cases in the math subdirectory will fail. Linux 2.2 has
518 fixes for the floating point support on Alpha. The Linux/SPARC kernel has
519 also some bugs in the FPU emulation code (as of Linux 2.2.0).
520 - Other tools might have problems. For example bash 2.03 gives a
521 segmentation fault running the tst-rpmatch.sh test script.
524 1.17. What is symbol versioning good for? Do I need it?
526 {AJ} Symbol versioning solves problems that are related to interface
527 changes. One version of an interface might have been introduced in a
528 previous version of the GNU C library but the interface or the semantics of
529 the function has been changed in the meantime. For binary compatibility
530 with the old library, a newer library needs to still have the old interface
531 for old programs. On the other hand, new programs should use the new
532 interface. Symbol versioning is the solution for this problem. The GNU
533 libc version 2.1 uses symbol versioning by default if the installed binutils
536 We don't advise building without symbol versioning, since you lose binary
537 compatibility - forever! The binary compatibility you lose is not only
538 against the previous version of the GNU libc (version 2.0) but also against
542 1.18. How can I compile on my fast ix86 machine a working libc for my slow
543 i386? After installing libc, programs abort with "Illegal
546 {AJ} glibc and gcc might generate some instructions on your machine that
547 aren't available on i386. You've got to tell glibc that you're configuring
548 for i386 with adding i386 as your machine, for example:
550 ../configure --prefix=/usr i386-pc-linux-gnu
552 And you need to tell gcc to only generate i386 code, just add `-mcpu=i386'
553 (just -m386 doesn't work) to your CFLAGS.
555 {UD} This applies not only to the i386. Compiling on a i686 for any older
556 model will also fail if the above methods are not used.
559 1.19. `make' complains about a missing dlfcn/libdl.so when building
560 malloc/libmemprof.so. How can I fix this?
562 {AJ} Older make version (<= 3.78.90) have a bug which was hidden by a bug in
563 glibc (<= 2.1.2). You need to upgrade make to a newer or fixed version.
565 After upgrading make, you should remove the file sysd-sorted in your build
566 directory. The problem is that the broken make creates a wrong order for
567 one list in that file. The list has to be recreated with the new make -
568 which happens if you remove the file.
570 You might encounter this bug also in other situations where make scans
571 directories. I strongly advise to upgrade your make version to 3.79 or
575 1.20. Which tools should I use for MIPS?
577 {AJ} You should use the current development version of gcc 2.97 from CVS.
578 gcc 2.95.x does not work correctly on mips-linux.
580 You need also recent binutils, anything before and including 2.10 will not
581 work correctly. Either try the Linux binutils 2.10.0.33 from HJ Lu or the
582 current development version from CVS of binutils.
584 For details check also my page <http://www.suse.de/~aj/glibc-mips.html>.
587 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
589 2. Installation and configuration issues
591 2.1. Can I replace the libc on my Linux system with GNU libc?
593 {UD} You cannot replace any existing libc for Linux with GNU libc. It is
594 binary incompatible and therefore has a different major version. You can,
595 however, install it alongside your existing libc.
597 For Linux there are three major libc versions:
599 libc-5 original ELF libc
602 You can have any combination of these three installed. For more information
603 consult documentation for shared library handling. The Makefiles of GNU
604 libc will automatically generate the needed symbolic links which the linker
608 2.2. How do I configure GNU libc so that the essential libraries
609 like libc.so go into /lib and the other into /usr/lib?
611 {UD,AJ} Like all other GNU packages GNU libc is designed to use a base
612 directory and install all files relative to this. The default is
613 /usr/local, because this is safe (it will not damage the system if installed
614 there). If you wish to install GNU libc as the primary C library on your
615 system, set the base directory to /usr (i.e. run configure --prefix=/usr
616 <other_options>). Note that this can damage your system; see question 2.3 for
619 Some systems like Linux have a filesystem standard which makes a difference
620 between essential libraries and others. Essential libraries are placed in
621 /lib because this directory is required to be located on the same disk
622 partition as /. The /usr subtree might be found on another
623 partition/disk. If you configure for Linux with --prefix=/usr, then this
624 will be done automatically.
626 To install the essential libraries which come with GNU libc in /lib on
627 systems other than Linux one must explicitly request it. Autoconf has no
628 option for this so you have to use a `configparms' file (see the `INSTALL'
629 file for details). It should contain:
634 The first line specifies the directory for the essential libraries, the
635 second line the directory for system configuration files.
638 2.3. How should I avoid damaging my system when I install GNU libc?
640 {ZW} If you wish to be cautious, do not configure with --prefix=/usr. If
641 you don't specify a prefix, glibc will be installed in /usr/local, where it
642 will probably not break anything. (If you wish to be certain, set the
643 prefix to something like /usr/local/glibc2 which is not used for anything.)
645 The dangers when installing glibc in /usr are twofold:
647 * glibc will overwrite the headers in /usr/include. Other C libraries
648 install a different but overlapping set of headers there, so the effect
649 will probably be that you can't compile anything. You need to rename
650 /usr/include out of the way before running `make install'. (Do not throw
651 it away; you will then lose the ability to compile programs against your
654 * None of your old libraries, static or shared, can be used with a
655 different C library major version. For shared libraries this is not a
656 problem, because the filenames are different and the dynamic linker
657 will enforce the restriction. But static libraries have no version
658 information. You have to evacuate all the static libraries in
659 /usr/lib to a safe location.
661 The situation is rather similar to the move from a.out to ELF which
662 long-time Linux users will remember.
665 2.4. Do I need to use GNU CC to compile programs that will use the
668 {ZW} In theory, no; the linker does not care, and the headers are supposed
669 to check for GNU CC before using its extensions to the C language.
671 However, there are currently no ports of glibc to systems where another
672 compiler is the default, so no one has tested the headers extensively
673 against another compiler. You may therefore encounter difficulties. If you
674 do, please report them as bugs.
676 Also, in several places GNU extensions provide large benefits in code
677 quality. For example, the library has hand-optimized, inline assembly
678 versions of some string functions. These can only be used with GCC. See
679 question 3.8 for details.
682 2.5. When linking with the new libc I get unresolved symbols
683 `crypt' and `setkey'. Why aren't these functions in the
686 {} Removed. Does not apply anymore.
689 2.6. When I use GNU libc on my Linux system by linking against
690 the libc.so which comes with glibc all I get is a core dump.
692 {UD} On Linux, gcc sets the dynamic linker to /lib/ld-linux.so.1 unless the
693 user specifies a --dynamic-linker argument. This is the name of the libc5
694 dynamic linker, which does not work with glibc.
696 For casual use of GNU libc you can just specify to the linker
697 --dynamic-linker=/lib/ld-linux.so.2
699 which is the glibc dynamic linker, on Linux systems. On other systems the
700 name is /lib/ld.so.1. When linking via gcc, you've got to add
701 -Wl,--dynamic-linker=/lib/ld-linux.so.2
703 to the gcc command line.
705 To change your environment to use GNU libc for compiling you need to change
706 the `specs' file of your gcc. This file is normally found at
708 /usr/lib/gcc-lib/<arch>/<version>/specs
710 In this file you have to change a few things:
712 - change `ld-linux.so.1' to `ld-linux.so.2'
714 - remove all expression `%{...:-lgmon}'; there is no libgmon in glibc
716 - fix a minor bug by changing %{pipe:-} to %|
718 Here is what the gcc-2.7.2 specs file should look like when GNU libc is
721 -----------------------------------------------------------------------
723 %{V} %{v:%{!V:-V}} %{Qy:} %{!Qn:-Qy} %{n} %{T} %{Ym,*} %{Yd,*} %{Wa,*:%*}
729 %{fPIC:-D__PIC__ -D__pic__} %{fpic:-D__PIC__ -D__pic__} %{!m386:-D__i486__} %{posix:-D_POSIX_SOURCE} %{pthread:-D_REENTRANT}
738 %{!shared:crtend.o%s} %{shared:crtendS.o%s} crtn.o%s
741 -m elf_i386 %{shared:-shared} %{!shared: %{!ibcs: %{!static: %{rdynamic:-export-dynamic} %{!dynamic-linker:-dynamic-linker /lib/ld-linux.so.2}} %{static:-static}}}
744 %{!shared: %{pthread:-lpthread} %{profile:-lc_p} %{!profile: -lc}}
750 %{!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}
752 *switches_need_spaces:
756 %{funsigned-char:-D__CHAR_UNSIGNED__}
759 -D__ELF__ -Dunix -Di386 -Dlinux -Asystem(unix) -Asystem(posix) -Acpu(i386) -Amachine(i386)
767 -----------------------------------------------------------------------
769 Things get a bit more complicated if you have GNU libc installed in some
770 other place than /usr, i.e., if you do not want to use it instead of the old
771 libc. In this case the needed startup files and libraries are not found in
772 the regular places. So the specs file must tell the compiler and linker
775 Version 2.7.2.3 does and future versions of GCC will automatically
776 provide the correct specs.
779 2.7. Looking through the shared libc file I haven't found the
780 functions `stat', `lstat', `fstat', and `mknod' and while
781 linking on my Linux system I get error messages. How is
782 this supposed to work?
784 {RM} Believe it or not, stat and lstat (and fstat, and mknod) are supposed
785 to be undefined references in libc.so.6! Your problem is probably a missing
786 or incorrect /usr/lib/libc.so file; note that this is a small text file now,
787 not a symlink to libc.so.6. It should look something like this:
789 GROUP ( libc.so.6 libc_nonshared.a )
792 2.8. When I run an executable on one system which I compiled on
793 another, I get dynamic linker errors. Both systems have the same
794 version of glibc installed. What's wrong?
796 {ZW} Glibc on one of these systems was compiled with gcc 2.7 or 2.8, the
797 other with egcs (any version). Egcs has functions in its internal
798 `libgcc.a' to support exception handling with C++. They are linked into
799 any program or dynamic library compiled with egcs, whether it needs them or
800 not. Dynamic libraries then turn around and export those functions again
801 unless special steps are taken to prevent them.
803 When you link your program, it resolves its references to the exception
804 functions to the ones exported accidentally by libc.so. That works fine as
805 long as libc has those functions. On the other system, libc doesn't have
806 those functions because it was compiled by gcc 2.8, and you get undefined
807 symbol errors. The symbols in question are named things like
808 `__register_frame_info'.
810 For glibc 2.0, the workaround is to not compile libc with egcs. We've also
811 incorporated a patch which should prevent the EH functions sneaking into
812 libc. It doesn't matter what compiler you use to compile your program.
814 For glibc 2.1, we've chosen to do it the other way around: libc.so
815 explicitly provides the EH functions. This is to prevent other shared
816 libraries from doing it.
818 {UD} Starting with glibc 2.1.1 you can compile glibc with gcc 2.8.1 or
819 newer since we have explicitly add references to the functions causing the
820 problem. But you nevertheless should use EGCS for other reasons
823 {GK} On some Linux distributions for PowerPC, you can see this when you have
824 built gcc or egcs from the Web sources (gcc versions 2.95 or earlier), then
825 re-built glibc. This happens because in these versions of gcc, exception
826 handling is implemented using an older method; the people making the
827 distributions are a little ahead of their time.
829 A quick solution to this is to find the libgcc.a file that came with the
830 distribution (it would have been installed under /usr/lib/gcc-lib), do
831 `ar x libgcc.a frame.o' to get the frame.o file out, and add a line saying
832 `LDLIBS-c.so += frame.o' to the file `configparms' in the directory you're
833 building in. You can check you've got the right `frame.o' file by running
834 `nm frame.o' and checking that it has the symbols defined that you're
837 This will let you build glibc with the C compiler. The C++ compiler
838 will still be binary incompatible with any C++ shared libraries that
839 you got with your distribution.
842 2.9. How can I compile gcc 2.7.2.1 from the gcc source code using
845 {AJ} There's only correct support for glibc 2.0.x in gcc 2.7.2.3 or later.
846 But you should get at least gcc 2.8.1 or egcs 1.1 (or later versions)
850 2.10. The `gencat' utility cannot process the catalog sources which
851 were used on my Linux libc5 based system. Why?
853 {UD} The `gencat' utility provided with glibc complies to the XPG standard.
854 The older Linux version did not obey the standard, so they are not
857 To ease the transition from the Linux version some of the non-standard
858 features are also present in the `gencat' program of GNU libc. This mainly
859 includes the use of symbols for the message number and the automatic
860 generation of header files which contain the needed #defines to map the
863 Here is a simple SED script to convert at least some Linux specific catalog
864 files to the XPG4 form:
866 -----------------------------------------------------------------------
867 # Change catalog source in Linux specific format to standard XPG format.
868 # Ulrich Drepper <drepper@cygnus.com>, 1996.
872 s/\$ #\([^ ]*\).*/\1/
874 s/\$ #[^ ]* *\(.*\)/\$ \1/
880 s/\(.*\)\n\(.*\)/\2 \1/
882 -----------------------------------------------------------------------
885 2.11. Programs using libc have their messages translated, but other
886 behavior is not localized (e.g. collating order); why?
888 {ZW} Translated messages are automatically installed, but the locale
889 database that controls other behaviors is not. You need to run localedef to
890 install this database, after you have run `make install'. For example, to
891 set up the French Canadian locale, simply issue the command
893 localedef -i fr_CA -f ISO-8859-1 fr_CA
895 Please see localedata/README in the source tree for further details.
898 2.12. I have set up /etc/nis.conf, and the Linux libc 5 with NYS
899 works great. But the glibc NIS+ doesn't seem to work.
901 {TK} The glibc NIS+ implementation uses a /var/nis/NIS_COLD_START file for
902 storing information about the NIS+ server and their public keys, because the
903 nis.conf file does not contain all the necessary information. You have to
904 copy a NIS_COLD_START file from a Solaris client (the NIS_COLD_START file is
905 byte order independent) or generate it with nisinit from the nis-tools
906 package; available at
908 http://www.suse.de/~kukuk/linux/nisplus.html
911 2.13. I have killed ypbind to stop using NIS, but glibc
914 {TK} For faster NIS lookups, glibc uses the /var/yp/binding/ files from
915 ypbind. ypbind 3.3 and older versions don't always remove these files, so
916 glibc will continue to use them. Other BSD versions seem to work correctly.
917 Until ypbind 3.4 is released, you can find a patch at
919 <ftp://ftp.kernel.org/pub/linux/utils/net/NIS/ypbind-3.3-glibc4.diff.gz>
922 2.14. Under Linux/Alpha, I always get "do_ypcall: clnt_call:
923 RPC: Unable to receive; errno = Connection refused" when using NIS.
925 {TK} You need a ypbind version which is 64bit clean. Some versions are not
926 64bit clean. A 64bit clean implementation is ypbind-mt. For ypbind 3.3,
927 you need the patch from ftp.kernel.org (See the previous question). I don't
928 know about other versions.
931 2.15. After installing glibc name resolving doesn't work properly.
933 {AJ} You probably should read the manual section describing nsswitch.conf
934 (just type `info libc "NSS Configuration File"'). The NSS configuration
935 file is usually the culprit.
938 2.16. How do I create the databases for NSS?
940 {AJ} If you have an entry "db" in /etc/nsswitch.conf you should also create
941 the database files. The glibc sources contain a Makefile which does the
942 necessary conversion and calls to create those files. The file is
943 `db-Makefile' in the subdirectory `nss' and you can call it with `make -f
944 db-Makefile'. Please note that not all services are capable of using a
945 database. Currently passwd, group, ethers, protocol, rpc, services shadow
946 and netgroup are implemented. See also question question 2.31.
949 2.17. I have /usr/include/net and /usr/include/scsi as symlinks
950 into my Linux source tree. Is that wrong?
952 {PB} This was necessary for libc5, but is not correct when using glibc.
953 Including the kernel header files directly in user programs usually does not
954 work (see question 3.5). glibc provides its own <net/*> and <scsi/*> header
955 files to replace them, and you may have to remove any symlink that you have
956 in place before you install glibc. However, /usr/include/asm and
957 /usr/include/linux should remain as they were.
960 2.18. Programs like `logname', `top', `uptime' `users', `w' and
961 `who', show incorrect information about the (number of)
962 users on my system. Why?
964 {MK} See question 3.2.
967 2.19. After upgrading to glibc 2.1 with symbol versioning I get
968 errors about undefined symbols. What went wrong?
970 {AJ} The problem is caused either by wrong program code or tools. In the
971 versioned libc a lot of symbols are now local that were global symbols in
972 previous versions. It seems that programs linked against older versions
973 often accidentally used libc global variables -- something that should not
976 The only way to fix this is to recompile your program. Sorry, that's the
977 price you might have to pay once for quite a number of advantages with
981 2.20. When I start the program XXX after upgrading the library
983 XXX: Symbol `_sys_errlist' has different size in shared
984 object, consider re-linking
985 Why? What should I do?
987 {UD} As the message says, relink the binary. The problem is that a few
988 symbols from the library can change in size and there is no way to avoid
989 this. _sys_errlist is a good example. Occasionally there are new error
990 numbers added to the kernel and this must be reflected at user level,
991 breaking programs that refer to them directly.
993 Such symbols should normally not be used at all. There are mechanisms to
994 avoid using them. In the case of _sys_errlist, there is the strerror()
995 function which should _always_ be used instead. So the correct fix is to
996 rewrite that part of the application.
998 In some situations (especially when testing a new library release) it might
999 be possible that a symbol changed size when that should not have happened.
1000 So in case of doubt report such a warning message as a problem.
1003 2.21. What do I need for C++ development?
1005 {HJ,AJ} You need either egcs 1.1 which comes directly with libstdc++ or
1006 gcc-2.8.1 together with libstdc++ 2.8.1.1. egcs 1.1 has the better C++
1007 support and works directly with glibc 2.1. If you use gcc-2.8.1 with
1008 libstdc++ 2.8.1.1, you need to modify libstdc++ a bit. A patch is available
1010 <ftp://alpha.gnu.org/gnu/libstdc++-2.8.1.1-glibc2.1-diff.gz>
1012 Please note that libg++ 2.7.2 (and the Linux Versions 2.7.2.x) doesn't work
1013 very well with the GNU C library due to vtable thunks. If you're upgrading
1014 from glibc 2.0.x to 2.1 you have to recompile libstdc++ since the library
1015 compiled for 2.0 is not compatible due to the new Large File Support (LFS)
1018 {UD} But since in the case of a shared libstdc++ the version numbers should
1019 be different existing programs will continue to work.
1022 2.22. Even statically linked programs need some shared libraries
1023 which is not acceptable for me. What can I do?
1025 {AJ} NSS (for details just type `info libc "Name Service Switch"') won't
1026 work properly without shared libraries. NSS allows using different services
1027 (e.g. NIS, files, db, hesiod) by just changing one configuration file
1028 (/etc/nsswitch.conf) without relinking any programs. The only disadvantage
1029 is that now static libraries need to access shared libraries. This is
1030 handled transparently by the GNU C library.
1032 A solution is to configure glibc with --enable-static-nss. In this case you
1033 can create a static binary that will use only the services dns and files
1034 (change /etc/nsswitch.conf for this). You need to link explicitly against
1035 all these services. For example:
1037 gcc -static test-netdb.c -o test-netdb \
1038 -Wl,--start-group -lc -lnss_files -lnss_dns -lresolv -Wl,--end-group
1040 The problem with this approach is that you've got to link every static
1041 program that uses NSS routines with all those libraries.
1043 {UD} In fact, one cannot say anymore that a libc compiled with this
1044 option is using NSS. There is no switch anymore. Therefore it is
1045 *highly* recommended *not* to use --enable-static-nss since this makes
1046 the behaviour of the programs on the system inconsistent.
1049 2.23. I just upgraded my Linux system to glibc and now I get
1050 errors whenever I try to link any program.
1052 {ZW} This happens when you have installed glibc as the primary C library but
1053 have stray symbolic links pointing at your old C library. If the first
1054 `libc.so' the linker finds is libc 5, it will use that. Your program
1055 expects to be linked with glibc, so the link fails.
1057 The most common case is that glibc put its `libc.so' in /usr/lib, but there
1058 was a `libc.so' from libc 5 in /lib, which gets searched first. To fix the
1059 problem, just delete /lib/libc.so. You may also need to delete other
1060 symbolic links in /lib, such as /lib/libm.so if it points to libm.so.5.
1062 {AJ} The perl script test-installation.pl which is run as last step during
1063 an installation of glibc that is configured with --prefix=/usr should help
1064 detect these situations. If the script reports problems, something is
1068 2.24. When I use nscd the machine freezes.
1070 {UD} You cannot use nscd with Linux 2.0.*. There is functionality missing
1071 in the kernel and work-arounds are not suitable. Besides, some parts of the
1072 kernel are too buggy when it comes to using threads.
1074 If you need nscd, you have to use at least a 2.1 kernel.
1076 Note that I have at this point no information about any other platform.
1079 2.25. I need lots of open files. What do I have to do?
1081 {AJ} This is at first a kernel issue. The kernel defines limits with
1082 OPEN_MAX the number of simultaneous open files and with FD_SETSIZE the
1083 number of used file descriptors. You need to change these values in your
1084 kernel and recompile the kernel so that the kernel allows more open
1085 files. You don't necessarily need to recompile the GNU C library since the
1086 only place where OPEN_MAX and FD_SETSIZE is really needed in the library
1087 itself is the size of fd_set which is used by select.
1089 The GNU C library is now select free. This means it internally has no
1090 limits imposed by the `fd_set' type. Instead all places where the
1091 functionality is needed the `poll' function is used.
1093 If you increase the number of file descriptors in the kernel you don't need
1094 to recompile the C library.
1096 {UD} You can always get the maximum number of file descriptors a process is
1097 allowed to have open at any time using
1099 number = sysconf (_SC_OPEN_MAX);
1101 This will work even if the kernel limits change.
1104 2.26. How do I get the same behavior on parsing /etc/passwd and
1105 /etc/group as I have with libc5 ?
1107 {TK} The name switch setup in /etc/nsswitch.conf selected by most Linux
1108 distributions does not support +/- and netgroup entries in the files like
1109 /etc/passwd. Though this is the preferred setup some people might have
1110 setups coming over from the libc5 days where it was the default to recognize
1111 lines like this. To get back to the old behaviour one simply has to change
1112 the rules for passwd, group, and shadow in the nsswitch.conf file as
1124 2.27. What needs to be recompiled when upgrading from glibc 2.0 to glibc
1127 {AJ,CG} If you just upgrade the glibc from 2.0.x (x <= 7) to 2.1, binaries
1128 that have been linked against glibc 2.0 will continue to work.
1130 If you compile your own binaries against glibc 2.1, you also need to
1131 recompile some other libraries. The problem is that libio had to be changed
1132 and therefore libraries that are based or depend on the libio of glibc,
1133 e.g. ncurses, slang and most C++ libraries, need to be recompiled. If you
1134 experience strange segmentation faults in your programs linked against glibc
1135 2.1, you might need to recompile your libraries.
1137 Another problem is that older binaries that were linked statically against
1138 glibc 2.0 will reference the older nss modules (libnss_files.so.1 instead of
1139 libnss_files.so.2), so don't remove them. Also, the old glibc-2.0 compiled
1140 static libraries (libfoo.a) which happen to depend on the older libio
1141 behavior will be broken by the glibc 2.1 upgrade. We plan to produce a
1142 compatibility library that people will be able to link in if they want
1143 to compile a static library generated against glibc 2.0 into a program
1144 on a glibc 2.1 system. You just add -lcompat and you should be fine.
1146 The glibc-compat add-on will provide the libcompat.a library, the older
1147 nss modules, and a few other files. Together, they should make it
1148 possible to do development with old static libraries on a glibc 2.1
1149 system. This add-on is still in development. You can get it from
1150 <ftp://alpha.gnu.org/gnu/glibc/glibc-compat-2.1.tar.gz>
1151 but please keep in mind that it is experimental.
1154 2.28. Why is extracting files via tar so slow?
1156 {AJ} Extracting of tar archives might be quite slow since tar has to look up
1157 userid and groupids and doesn't cache negative results. If you have nis or
1158 nisplus in your /etc/nsswitch.conf for the passwd and/or group database,
1159 each file extractions needs a network connection. There are two possible
1162 - do you really need NIS/NIS+ (some Linux distributions add by default
1163 nis/nisplus even if it's not needed)? If not, just remove the entries.
1165 - if you need NIS/NIS+, use the Name Service Cache Daemon nscd that comes
1169 2.29. Compiling programs I get parse errors in libio.h (e.g. "parse error
1170 before `_IO_seekoff'"). How should I fix this?
1172 {AJ} You might get the following errors when upgrading to glibc 2.1:
1174 In file included from /usr/include/stdio.h:57,
1176 /usr/include/libio.h:335: parse error before `_IO_seekoff'
1177 /usr/include/libio.h:335: parse error before `_G_off64_t'
1178 /usr/include/libio.h:336: parse error before `_IO_seekpos'
1179 /usr/include/libio.h:336: parse error before `_G_fpos64_t'
1181 The problem is a wrong _G_config.h file in your include path. The
1182 _G_config.h file that comes with glibc 2.1 should be used and not one from
1183 libc5 or from a compiler directory. To check which _G_config.h file the
1184 compiler uses, compile your program with `gcc -E ...|grep G_config.h' and
1185 remove that file. Your compiler should pick up the file that has been
1186 installed by glibc 2.1 in your include directory.
1189 2.30. After upgrading to glibc 2.1, libraries that were compiled against
1190 glibc 2.0.x don't work anymore.
1192 {AJ} See question 2.27.
1195 2.31. What happened to the Berkeley DB libraries? Can I still use db
1196 in /etc/nsswitch.conf?
1198 {AJ} Due to too many incompatible changes in disk layout and API of Berkeley
1199 DB and a too tight coupling of libc and libdb, the db library has been
1200 removed completely from glibc 2.2. The only place that really used the
1201 Berkeley DB was the NSS db module.
1203 The NSS db module has been rewritten to support a number of different
1204 versions of Berkeley DB for the NSS db module. Currently the releases 2.x
1205 and 3.x of Berkeley DB are supported. The older db 1.85 library is not
1206 supported. You can use the version from glibc 2.1.x or download a version
1207 from Sleepycat Software (http://www.sleepycat.com). The library has to be
1208 compiled as shared library and installed in the system lib directory
1209 (normally /lib). The library needs to have a special soname to be found by
1212 If public structures change in a new Berkeley db release, this needs to be
1215 Currently the code searches for libraries with a soname of "libdb.so.3"
1216 (that's the name from db 2.4.14 which comes with glibc 2.1.x) and
1217 "libdb-3.0.so" (the name used by db 3.0.55 as default).
1219 The nss_db module is now in a separate package since it requires a database
1220 library being available.
1223 2.32. What has do be done when upgrading to glibc 2.2?
1225 {AJ} The upgrade to glibc 2.2 should run smoothly, there's in general no
1226 need to recompile programs or libraries. Nevertheless, some changes might
1227 be needed after upgrading:
1228 - The utmp daemon has been removed and is not supported by glibc anymore.
1229 If it has been in use, it should be switched off.
1230 - Programs using IPv6 have to be recompiled due to incompatible changes in
1231 sockaddr_in6 by the IPv6 working group.
1232 - The Berkeley db libraries have been removed (for details see ??nssdb).
1233 - The format of the locale files has changed, all locales should be
1234 regenerated with localedef. All statically linked applications which use
1235 i18n should be recompiled, otherwise they'll not be localized.
1236 - glibc comes with a number of new applications. For example ldconfig has
1237 been implemented for glibc, the libc5 version of ldconfig is not needed
1239 - There's no more K&R compatibility in the glibc headers. The GNU C library
1240 requires a C compiler that handles especially prototypes correctly.
1241 Especially gcc -traditional will not work with glibc headers.
1243 Please read also the NEWS file which is the authoritative source for this
1244 and gives more details for some topics.
1247 2.33. The makefiles want to do a CVS commit.
1249 {UD} Only if you are not specifying the --without-cvs flag at configure
1250 time. This is what you always have to use if you are checking sources
1251 directly out of the public CVS repository or you have your own private
1255 2.34. When compiling C++ programs, I get a compilation error in streambuf.h.
1257 {BH} You are using g++ 2.95.2? After upgrading to glibc 2.2, you need to
1258 apply a patch to the include files in /usr/include/g++, because the fpos_t
1259 type has changed in glibc 2.2. The patch is at
1260 http://clisp.cons.org/~haible/gccinclude-glibc-2.2-compat.diff
1263 2.35. When recompiling GCC, I get compilation errors in libio.
1265 {BH} You are trying to recompile gcc 2.95.2? After upgrading to glibc 2.2,
1266 you need to apply a patch to the gcc sources, because the fpos_t type and
1267 a few libio internals have changed in glibc 2.2. T he patch is at
1268 http://clisp.cons.org/~haible/gcc-glibc-2.2-compat.diff
1271 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1273 3. Source and binary incompatibilities, and what to do about them
1275 3.1. I expect GNU libc to be 100% source code compatible with
1276 the old Linux based GNU libc. Why isn't it like this?
1278 {DMT,UD} Not every extension in Linux libc's history was well thought-out.
1279 In fact it had a lot of problems with standards compliance and with
1280 cleanliness. With the introduction of a new version number these errors can
1281 now be corrected. Here is a list of the known source code
1284 * _GNU_SOURCE: glibc does not make the GNU extensions available
1285 automatically. If a program depends on GNU extensions or some
1286 other non-standard functionality, it is necessary to compile it
1287 with the C compiler option -D_GNU_SOURCE, or better, to put
1288 `#define _GNU_SOURCE' at the beginning of your source files, before
1289 any C library header files are included. This difference normally
1290 manifests itself in the form of missing prototypes and/or data type
1291 definitions. Thus, if you get such errors, the first thing you
1292 should do is try defining _GNU_SOURCE and see if that makes the
1295 For more information consult the file `NOTES' in the GNU C library
1298 * reboot(): GNU libc sanitizes the interface of reboot() to be more
1299 compatible with the interface used on other OSes. reboot() as
1300 implemented in glibc takes just one argument. This argument
1301 corresponds to the third argument of the Linux reboot system call.
1302 That is, a call of the form reboot(a, b, c) needs to be changed into
1303 reboot(c). Beside this the header <sys/reboot.h> defines the needed
1304 constants for the argument. These RB_* constants should be used
1305 instead of the cryptic magic numbers.
1307 * swapon(): the interface of this function didn't change, but the
1308 prototype is in a separate header file <sys/swap.h>. This header
1309 file also provides the SWAP_* constants defined by <linux/swap.h>;
1310 you should use them for the second argument to swapon().
1312 * errno: If a program uses the variable "errno", then it _must_
1313 include <errno.h>. The old libc often (erroneously) declared this
1314 variable implicitly as a side-effect of including other libc header
1315 files. glibc is careful to avoid such namespace pollution, which,
1316 in turn, means that you really need to include the header files that
1317 you depend on. This difference normally manifests itself in the
1318 form of the compiler complaining about references to an undeclared
1321 * Linux-specific syscalls: All Linux system calls now have appropriate
1322 library wrappers and corresponding declarations in various header files.
1323 This is because the syscall() macro that was traditionally used to
1324 work around missing syscall wrappers are inherently non-portable and
1325 error-prone. The following table lists all the new syscall stubs,
1326 the header-file declaring their interface and the system call name.
1328 syscall name: wrapper name: declaring header file:
1329 ------------- ------------- ----------------------
1330 bdflush bdflush <sys/kdaemon.h>
1331 syslog ksyslog_ctl <sys/klog.h>
1333 * lpd: Older versions of lpd depend on a routine called _validuser().
1334 The library does not provide this function, but instead provides
1335 __ivaliduser() which has a slightly different interface. Simply
1336 upgrading to a newer lpd should fix this problem (e.g., the 4.4BSD
1337 lpd is known to be working).
1339 * resolver functions/BIND: like on many other systems the functions of
1340 the resolver library are not included in libc itself. There is a
1341 separate library libresolv. If you get undefined symbol errors for
1342 symbols starting with `res_*' simply add -lresolv to your linker
1345 * the `signal' function's behavior corresponds to the BSD semantic and
1346 not the SysV semantic as it was in libc-5. The interface on all GNU
1347 systems shall be the same and BSD is the semantic of choice. To use
1348 the SysV behavior simply use `sysv_signal', or define _XOPEN_SOURCE.
1349 See question 3.7 for details.
1352 3.2. Why does getlogin() always return NULL on my Linux box?
1354 {UD} The GNU C library has a format for the UTMP and WTMP file which differs
1355 from what your system currently has. It was extended to fulfill the needs
1356 of the next years when IPv6 is introduced. The record size is different and
1357 some fields have different positions. The files written by functions from
1358 the one library cannot be read by functions from the other library. Sorry,
1359 but this is what a major release is for. It's better to have a cut now than
1360 having no means to support the new techniques later.
1363 3.3. Where are the DST_* constants found in <sys/time.h> on many
1366 {UD} These constants come from the old BSD days and are not used anymore
1367 (libc5 does not actually implement the handling although the constants are
1370 Instead GNU libc contains zone database support and compatibility code for
1371 POSIX TZ environment variable handling. For former is very much preferred
1375 3.4. The prototypes for `connect', `accept', `getsockopt',
1376 `setsockopt', `getsockname', `getpeername', `send',
1377 `sendto', and `recvfrom' are different in GNU libc from
1378 any other system I saw. This is a bug, isn't it?
1380 {UD} No, this is no bug. This version of GNU libc already follows the new
1381 Single Unix specifications (and I think the POSIX.1g draft which adopted the
1382 solution). The type for a parameter describing a size is now `socklen_t', a
1386 3.5. On Linux I've got problems with the declarations in Linux
1389 {UD,AJ} On Linux, the use of kernel headers is reduced to the minimum. This
1390 gives Linus the ability to change the headers more freely. Also, user
1391 programs are now insulated from changes in the size of kernel data
1394 For example, the sigset_t type is 32 or 64 bits wide in the kernel. In
1395 glibc it is 1024 bits wide. This guarantees that when the kernel gets a
1396 bigger sigset_t (for POSIX.1e realtime support, say) user programs will not
1397 have to be recompiled. Consult the header files for more information about
1400 Therefore you shouldn't include Linux kernel header files directly if glibc
1401 has defined a replacement. Otherwise you might get undefined results because
1405 3.6. I don't include any kernel headers myself but the compiler
1406 still complains about redeclarations of types in the kernel
1409 {UD} The kernel headers before Linux 2.1.61 and 2.0.32 don't work correctly
1410 with glibc. Compiling C programs is possible in most cases but C++ programs
1411 have (due to the change of the name lookups for `struct's) problems. One
1412 prominent example is `struct fd_set'.
1414 There might be some problems left but 2.1.61/2.0.32 fix most of the known
1415 ones. See the BUGS file for other known problems.
1418 3.7. Why don't signals interrupt system calls anymore?
1420 {ZW} By default GNU libc uses the BSD semantics for signal(), unlike Linux
1421 libc 5 which used System V semantics. This is partially for compatibility
1422 with other systems and partially because the BSD semantics tend to make
1423 programming with signals easier.
1425 There are three differences:
1427 * BSD-style signals that occur in the middle of a system call do not
1428 affect the system call; System V signals cause the system call to
1429 fail and set errno to EINTR.
1431 * BSD signal handlers remain installed once triggered. System V signal
1432 handlers work only once, so one must reinstall them each time.
1434 * A BSD signal is blocked during the execution of its handler. In other
1435 words, a handler for SIGCHLD (for example) does not need to worry about
1436 being interrupted by another SIGCHLD. It may, however, be interrupted
1439 There is general consensus that for `casual' programming with signals, the
1440 BSD semantics are preferable. You don't need to worry about system calls
1441 returning EINTR, and you don't need to worry about the race conditions
1442 associated with one-shot signal handlers.
1444 If you are porting an old program that relies on the old semantics, you can
1445 quickly fix the problem by changing signal() to sysv_signal() throughout.
1446 Alternatively, define _XOPEN_SOURCE before including <signal.h>.
1448 For new programs, the sigaction() function allows you to specify precisely
1449 how you want your signals to behave. All three differences listed above are
1450 individually switchable on a per-signal basis with this function.
1452 If all you want is for one specific signal to cause system calls to fail and
1453 return EINTR (for example, to implement a timeout) you can do this with
1457 3.8. I've got errors compiling code that uses certain string
1460 {AJ} glibc 2.1 has special string functions that are faster than the normal
1461 library functions. Some of the functions are additionally implemented as
1462 inline functions and others as macros. This might lead to problems with
1463 existing codes but it is explicitly allowed by ISO C.
1465 The optimized string functions are only used when compiling with
1466 optimizations (-O1 or higher). The behavior can be changed with two feature
1469 * __NO_STRING_INLINES: Don't do any string optimizations.
1470 * __USE_STRING_INLINES: Use assembly language inline functions (might
1471 increase code size dramatically).
1473 Since some of these string functions are now additionally defined as macros,
1474 code like "char *strncpy();" doesn't work anymore (and is unnecessary, since
1475 <string.h> has the necessary declarations). Either change your code or
1476 define __NO_STRING_INLINES.
1478 {UD} Another problem in this area is that gcc still has problems on machines
1479 with very few registers (e.g., ix86). The inline assembler code can require
1480 almost all the registers and the register allocator cannot always handle
1483 One can disable the string optimizations selectively. Instead of writing
1485 cp = strcpy (foo, "lkj");
1489 cp = (strcpy) (foo, "lkj");
1491 This disables the optimization for that specific call.
1494 3.9. I get compiler messages "Initializer element not constant" with
1495 stdin/stdout/stderr. Why?
1497 {RM,AJ} Constructs like:
1498 static FILE *InPtr = stdin;
1500 lead to this message. This is correct behaviour with glibc since stdin is
1501 not a constant expression. Please note that a strict reading of ISO C does
1502 not allow above constructs.
1504 One of the advantages of this is that you can assign to stdin, stdout, and
1505 stderr just like any other global variable (e.g. `stdout = my_stream;'),
1506 which can be very useful with custom streams that you can write with libio
1507 (but beware this is not necessarily portable). The reason to implement it
1508 this way were versioning problems with the size of the FILE structure.
1510 To fix those programs you've got to initialize the variable at run time.
1511 This can be done, e.g. in main, like:
1519 or by constructors (beware this is gcc specific):
1522 static void inPtr_construct (void) __attribute__((constructor));
1523 static void inPtr_construct (void) { InPtr = stdin; }
1526 3.10. I can't compile with gcc -traditional (or
1527 -traditional-cpp). Why?
1529 {AJ} glibc2 does break -traditional and -traditonal-cpp - and will continue
1530 to do so. For example constructs of the form:
1536 are useful for debugging purposes (you can use foo with your debugger that's
1537 why we need the enum) and for compatibility (other systems use defines and
1541 3.11. I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible?
1543 {AJ} The GNU C library is compatible with the ANSI/ISO C standard. If
1544 you're using `gcc -ansi', the glibc includes which are specified in the
1545 standard follow the standard. The ANSI/ISO C standard defines what has to be
1546 in the include files - and also states that nothing else should be in the
1547 include files (btw. you can still enable additional standards with feature
1550 The GNU C library is conforming to ANSI/ISO C - if and only if you're only
1551 using the headers and library functions defined in the standard.
1554 3.12. I can't access some functions anymore. nm shows that they do
1555 exist but linking fails nevertheless.
1557 {AJ} With the introduction of versioning in glibc 2.1 it is possible to
1558 export only those identifiers (functions, variables) that are really needed
1559 by application programs and by other parts of glibc. This way a lot of
1560 internal interfaces are now hidden. nm will still show those identifiers
1561 but marking them as internal. ISO C states that identifiers beginning with
1562 an underscore are internal to the libc. An application program normally
1563 shouldn't use those internal interfaces (there are exceptions,
1564 e.g. __ivaliduser). If a program uses these interfaces, it's broken. These
1565 internal interfaces might change between glibc releases or dropped
1569 3.13. When using the db-2 library which comes with glibc is used in
1570 the Perl db modules the testsuite is not passed. This did not
1571 happen with db-1, gdbm, or ndbm.
1573 {} Removed. Does not apply anymore.
1576 3.14. The pow() inline function I get when including <math.h> is broken.
1577 I get segmentation faults when I run the program.
1579 {UD} Nope, the implementation is correct. The problem is with egcs version
1580 prior to 1.1. I.e., egcs 1.0 to 1.0.3 are all broken (at least on Intel).
1581 If you have to use this compiler you must define __NO_MATH_INLINES before
1582 including <math.h> to prevent the inline functions from being used. egcs 1.1
1583 fixes the problem. I don't know about gcc 2.8 and 2.8.1.
1586 3.15. The sys/sem.h file lacks the definition of `union semun'.
1588 {UD} Nope. This union has to be provided by the user program. Former glibc
1589 versions defined this but it was an error since it does not make much sense
1590 when thinking about it. The standards describing the System V IPC functions
1591 define it this way and therefore programs must be adopted.
1594 3.16. Why has <netinet/ip_fw.h> disappeared?
1596 {AJ} The corresponding Linux kernel data structures and constants are
1597 totally different in Linux 2.0 and Linux 2.2. This situation has to be
1598 taken care in user programs using the firewall structures and therefore
1599 those programs (ipfw is AFAIK the only one) should deal with this problem
1603 3.17. I get floods of warnings when I use -Wconversion and include
1604 <string.h> or <math.h>.
1606 {ZW} <string.h> and <math.h> intentionally use prototypes to override
1607 argument promotion. -Wconversion warns about all these. You can safely
1608 ignore the warnings.
1610 -Wconversion isn't really intended for production use, only for shakedown
1611 compiles after converting an old program to standard C.
1614 3.18. After upgrading to glibc 2.1, I receive errors about
1615 unresolved symbols, like `_dl_initial_searchlist' and can not
1616 execute any binaries. What went wrong?
1618 {AJ} This normally happens if your libc and ld (dynamic linker) are from
1619 different releases of glibc. For example, the dynamic linker
1620 /lib/ld-linux.so.2 comes from glibc 2.0.x, but the version of libc.so.6 is
1623 The path /lib/ld-linux.so.2 is hardcoded in every glibc2 binary but
1624 libc.so.6 is searched via /etc/ld.so.cache and in some special directories
1625 like /lib and /usr/lib. If you run configure with another prefix than /usr
1626 and put this prefix before /lib in /etc/ld.so.conf, your system will break.
1628 So what can you do? Either of the following should work:
1630 * Run `configure' with the same prefix argument you've used for glibc 2.0.x
1631 so that the same paths are used.
1632 * Replace /lib/ld-linux.so.2 with a link to the dynamic linker from glibc
1635 You can even call the dynamic linker by hand if everything fails. You've
1636 got to set LD_LIBRARY_PATH so that the corresponding libc is found and also
1637 need to provide an absolute path to your binary:
1639 LD_LIBRARY_PATH=<path-where-libc.so.6-lives> \
1640 <path-where-corresponding-dynamic-linker-lives>/ld-linux.so.2 \
1641 <path-to-binary>/binary
1643 For example `LD_LIBRARY_PATH=/libold /libold/ld-linux.so.2 /bin/mv ...'
1644 might be useful in fixing a broken system (if /libold contains dynamic
1645 linker and corresponding libc).
1647 With that command line no path is used. To further debug problems with the
1648 dynamic linker, use the LD_DEBUG environment variable, e.g.
1649 `LD_DEBUG=help echo' for the help text.
1651 If you just want to test this release, don't put the lib directory in
1652 /etc/ld.so.conf. You can call programs directly with full paths (as above).
1653 When compiling new programs against glibc 2.1, you've got to specify the
1654 correct paths to the compiler (option -I with gcc) and linker (options
1655 --dynamic-linker, -L and --rpath).
1658 3.19. bonnie reports that char i/o with glibc 2 is much slower than with
1659 libc5. What can be done?
1661 {AJ} The GNU C library uses thread safe functions by default and libc5 used
1662 non thread safe versions. The non thread safe functions have in glibc the
1663 suffix `_unlocked', for details check <stdio.h>. Using `putc_unlocked' etc.
1664 instead of `putc' should give nearly the same speed with bonnie (bonnie is a
1665 benchmark program for measuring disk access).
1668 3.20. Programs compiled with glibc 2.1 can't read db files made with glibc
1669 2.0. What has changed that programs like rpm break?
1671 {} Removed. Does not apply anymore.
1674 3.21. Autoconf's AC_CHECK_FUNC macro reports that a function exists, but
1675 when I try to use it, it always returns -1 and sets errno to ENOSYS.
1677 {ZW} You are using a 2.0 Linux kernel, and the function you are trying to
1678 use is only implemented in 2.1/2.2. Libc considers this to be a function
1679 which exists, because if you upgrade to a 2.2 kernel, it will work. One
1680 such function is sigaltstack.
1682 Your program should check at runtime whether the function works, and
1683 implement a fallback. Note that Autoconf cannot detect unimplemented
1684 functions in other systems' C libraries, so you need to do this anyway.
1687 3.22. My program segfaults when I call fclose() on the FILE* returned
1688 from setmntent(). Is this a glibc bug?
1690 {GK} No. Don't do this. Use endmntent(), that's what it's for.
1692 In general, you should use the correct deallocation routine. For instance,
1693 if you open a file using fopen(), you should deallocate the FILE * using
1694 fclose(), not free(), even though the FILE * is also a pointer.
1696 In the case of setmntent(), it may appear to work in most cases, but it
1697 won't always work. Unfortunately, for compatibility reasons, we can't
1698 change the return type of setmntent() to something other than FILE *.
1701 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1705 4.1. After I changed configure.in I get `Autoconf version X.Y.
1706 or higher is required for this script'. What can I do?
1708 {UD} You have to get the specified autoconf version (or a later one)
1709 from your favorite mirror of ftp.gnu.org.
1712 4.2. When I try to compile code which uses IPv6 headers and
1713 definitions on my Linux 2.x.y system I am in trouble.
1714 Nothing seems to work.
1716 {UD} The problem is that IPv6 development still has not reached a point
1717 where the headers are stable. There are still lots of incompatible changes
1718 made and the libc headers have to follow.
1720 {PB} The 2.1 release of GNU libc aims to comply with the current versions of
1721 all the relevant standards. The IPv6 support libraries for older Linux
1722 systems used a different naming convention and so code written to work with
1723 them may need to be modified. If the standards make incompatible changes in
1724 the future then the libc may need to change again.
1726 IPv6 will not work with a 2.0.x kernel. When kernel 2.2 is released it
1727 should contain all the necessary support; until then you should use the
1728 latest 2.1.x release you can find. As of 98/11/26 the currently recommended
1729 kernel for IPv6 is 2.1.129.
1731 Also, as of the 2.1 release the IPv6 API provided by GNU libc is not
1735 4.3. When I set the timezone by setting the TZ environment variable
1736 to EST5EDT things go wrong since glibc computes the wrong time
1737 from this information.
1739 {UD} The problem is that people still use the braindamaged POSIX method to
1740 select the timezone using the TZ environment variable with a format EST5EDT
1741 or whatever. People, if you insist on using TZ instead of the timezone
1742 database (see below), read the POSIX standard, the implemented behaviour is
1743 correct! What you see is in fact the result of the decisions made while
1744 POSIX.1 was created. We've only implemented the handling of TZ this way to
1745 be POSIX compliant. It is not really meant to be used.
1747 The alternative approach to handle timezones which is implemented is the
1748 correct one to use: use the timezone database. This avoids all the problems
1749 the POSIX method has plus it is much easier to use. Simply run the tzselect
1750 shell script, answer the question and use the name printed in the end by
1751 making a symlink /etc/localtime pointing to /usr/share/zoneinfo/NAME (NAME
1752 is the returned value from tzselect). That's all. You never again have to
1755 So, please avoid sending bug reports about time related problems if you use
1756 the POSIX method and you have not verified something is really broken by
1757 reading the POSIX standards.
1760 4.4. What other sources of documentation about glibc are available?
1762 {AJ} The FSF has a page about the GNU C library at
1763 <http://www.gnu.org/software/libc/>. The problem data base of open and
1764 solved bugs in GNU libc is available at
1765 <http://www-gnats.gnu.org:8080/cgi-bin/wwwgnats.pl>. Eric Green has written
1766 a HowTo for converting from Linux libc5 to glibc2. The HowTo is accessible
1767 via the FSF page and at <http://www.imaxx.net/~thrytis/glibc>. Frodo
1768 Looijaard describes a different way installing glibc2 as secondary libc at
1769 <http://huizen.dds.nl/~frodol/glibc>.
1771 Please note that this is not a complete list.
1774 4.5. The timezone string for Sydney/Australia is wrong since even when
1775 daylight saving time is in effect the timezone string is EST.
1777 {UD} The problem for some timezones is that the local authorities decided
1778 to use the term "summer time" instead of "daylight saving time". In this
1779 case the abbreviation character `S' is the same as the standard one. So,
1782 Eastern Standard Time = EST
1783 Eastern Summer Time = EST
1785 Great! To get this bug fixed convince the authorities to change the laws
1786 and regulations of the country this effects. glibc behaves correctly.
1789 4.6. I've build make 3.77 against glibc 2.1 and now make gets
1790 segmentation faults.
1792 {} Removed. Does not apply anymore, use make 3.79 or newer.
1795 4.7. Why do so many programs using math functions fail on my AlphaStation?
1797 {AO} The functions floor() and floorf() use an instruction that is not
1798 implemented in some old PALcodes of AlphaStations. This may cause
1799 `Illegal Instruction' core dumps or endless loops in programs that
1800 catch these signals. Updating the firmware to a 1999 release has
1801 fixed the problem on an AlphaStation 200 4/166.
1804 4.8. The conversion table for character set XX does not match with
1807 {UD} I don't doubt for a minute that some of the conversion tables contain
1808 errors. We tried the best we can and relied on automatic generation of the
1809 data to prevent human-introduced errors but this still is no guarantee. If
1810 you think you found a problem please send a bug report describing it and
1811 give an authoritive reference. The latter is important since otherwise
1812 the current behaviour is as good as the proposed one.
1814 Before doing this look through the list of known problem first:
1816 - the GBK (simplified Chinese) encoding is based on Unicode tables. This
1817 is good. These tables, however, differ slightly from the tables used
1818 by the M$ people. The differences are these [+ Unicode, - M$]:
1825 In addition the Unicode tables contain mappings for the GBK characters
1826 0xA8BC, 0xA8BF, 0xA989 to 0xA995, and 0xFE50 to 0xFEA0.
1828 - when mapping from EUC-CN to GBK and vice versa we ignore the fact that
1829 the coded character at position 0xA1A4 maps to different Unicode
1830 characters. Since the iconv() implementation can do whatever it wants
1831 if it cannot directly map a character this is a perfectly good solution
1832 since the semantics and appearance of the character does not change.
1835 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
1837 Answers were given by:
1838 {UD} Ulrich Drepper, <drepper@cygnus.com>
1839 {DMT} David Mosberger-Tang, <davidm@AZStarNet.com>
1840 {RM} Roland McGrath, <roland@gnu.org>
1841 {AJ} Andreas Jaeger, <aj@suse.de>
1842 {EY} Eric Youngdale, <eric@andante.jic.com>
1843 {PB} Phil Blundell, <Philip.Blundell@pobox.com>
1844 {MK} Mark Kettenis, <kettenis@phys.uva.nl>
1845 {ZW} Zack Weinberg, <zack@rabi.phys.columbia.edu>
1846 {TK} Thorsten Kukuk, <kukuk@suse.de>
1847 {GK} Geoffrey Keating, <geoffk@ozemail.com.au>
1848 {HJ} H.J. Lu, <hjl@gnu.org>
1849 {CG} Cristian Gafton, <gafton@redhat.com>
1850 {AO} Alexandre Oliva, <oliva@lsd.ic.unicamp.br>
1851 {BH} Bruno Haible, <haible@clisp.cons.org>
1855 outline-regexp:"\\?"