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?
45 2. Installation and configuration issues
47 2.1. Can I replace the libc on my Linux system with GNU libc?
48 2.2. How do I configure GNU libc so that the essential libraries
49 like libc.so go into /lib and the other into /usr/lib?
50 2.3. How should I avoid damaging my system when I install GNU libc?
51 2.4. Do I need to use GNU CC to compile programs that will use the
53 2.5. When linking with the new libc I get unresolved symbols
54 `crypt' and `setkey'. Why aren't these functions in the
56 2.6. When I use GNU libc on my Linux system by linking against
57 the libc.so which comes with glibc all I get is a core dump.
58 2.7. Looking through the shared libc file I haven't found the
59 functions `stat', `lstat', `fstat', and `mknod' and while
60 linking on my Linux system I get error messages. How is
61 this supposed to work?
62 2.8. When I run an executable on one system which I compiled on
63 another, I get dynamic linker errors. Both systems have the same
64 version of glibc installed. What's wrong?
65 2.9. How can I compile gcc 2.7.2.1 from the gcc source code using
67 2.10. The `gencat' utility cannot process the catalog sources which
68 were used on my Linux libc5 based system. Why?
69 2.11. Programs using libc have their messages translated, but other
70 behavior is not localized (e.g. collating order); why?
71 2.12. I have set up /etc/nis.conf, and the Linux libc 5 with NYS
72 works great. But the glibc NIS+ doesn't seem to work.
73 2.13. I have killed ypbind to stop using NIS, but glibc
75 2.14. Under Linux/Alpha, I always get "do_ypcall: clnt_call:
76 RPC: Unable to receive; errno = Connection refused" when using NIS.
77 2.15. After installing glibc name resolving doesn't work properly.
78 2.16. How do I create the databases for NSS?
79 2.17. I have /usr/include/net and /usr/include/scsi as symlinks
80 into my Linux source tree. Is that wrong?
81 2.18. Programs like `logname', `top', `uptime' `users', `w' and
82 `who', show incorrect information about the (number of)
83 users on my system. Why?
84 2.19. After upgrading to glibc 2.1 with symbol versioning I get
85 errors about undefined symbols. What went wrong?
86 2.20. When I start the program XXX after upgrading the library
88 XXX: Symbol `_sys_errlist' has different size in shared
89 object, consider re-linking
90 Why? What should I do?
91 2.21. What do I need for C++ development?
92 2.22. Even statically linked programs need some shared libraries
93 which is not acceptable for me. What can I do?
94 2.23. I just upgraded my Linux system to glibc and now I get
95 errors whenever I try to link any program.
96 2.24. When I use nscd the machine freezes.
97 2.25. I need lots of open files. What do I have to do?
98 2.26. How do I get the same behavior on parsing /etc/passwd and
99 /etc/group as I have with libc5 ?
100 2.27. What needs to be recompiled when upgrading from glibc 2.0 to glibc
102 2.28. Why is extracting files via tar so slow?
103 2.29. Compiling programs I get parse errors in libio.h (e.g. "parse error
104 before `_IO_seekoff'"). How should I fix this?
106 3. Source and binary incompatibilities, and what to do about them
108 3.1. I expect GNU libc to be 100% source code compatible with
109 the old Linux based GNU libc. Why isn't it like this?
110 3.2. Why does getlogin() always return NULL on my Linux box?
111 3.3. Where are the DST_* constants found in <sys/time.h> on many
113 3.4. The prototypes for `connect', `accept', `getsockopt',
114 `setsockopt', `getsockname', `getpeername', `send',
115 `sendto', and `recvfrom' are different in GNU libc from
116 any other system I saw. This is a bug, isn't it?
117 3.5. On Linux I've got problems with the declarations in Linux
119 3.6. I don't include any kernel headers myself but the compiler
120 still complains about redeclarations of types in the kernel
122 3.7. Why don't signals interrupt system calls anymore?
123 3.8. I've got errors compiling code that uses certain string
125 3.9. I get compiler messages "Initializer element not constant" with
126 stdin/stdout/stderr. Why?
127 3.10. I can't compile with gcc -traditional (or
128 -traditional-cpp). Why?
129 3.11. I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible?
130 3.12. I can't access some functions anymore. nm shows that they do
131 exist but linking fails nevertheless.
132 3.13. When using the db-2 library which comes with glibc is used in
133 the Perl db modules the testsuite is not passed. This did not
134 happen with db-1, gdbm, or ndbm.
135 3.14. The pow() inline function I get when including <math.h> is broken.
136 I get segmentation faults when I run the program.
137 3.15. The sys/sem.h file lacks the definition of `union semun'.
138 3.16. Why has <netinet/ip_fw.h> disappeared?
139 3.17. I get floods of warnings when I use -Wconversion and include
140 <string.h> or <math.h>.
141 3.18. After upgrading to glibc 2.1, I receive errors about
142 unresolved symbols, like `_dl_initial_searchlist' and can not
143 execute any binaries. What went wrong?
144 3.19. bonnie reports that char i/o with glibc 2 is much slower than with
145 libc5. What can be done?
146 3.20. Programs compiled with glibc 2.1 can't read db files made with glibc
147 2.0. What has changed that programs like rpm break?
151 4.1. After I changed configure.in I get `Autoconf version X.Y.
152 or higher is required for this script'. What can I do?
153 4.2. When I try to compile code which uses IPv6 headers and
154 definitions on my Linux 2.x.y system I am in trouble.
155 Nothing seems to work.
156 4.3. When I set the timezone by setting the TZ environment variable
157 to EST5EDT things go wrong since glibc computes the wrong time
158 from this information.
159 4.4. What other sources of documentation about glibc are available?
160 4.5. The timezone string for Sydney/Australia is wrong since even when
161 daylight saving time is in effect the timezone string is EST.
162 4.6. I've build make 3.77 against glibc 2.1 and now make gets
166 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
170 1.1. What systems does the GNU C Library run on?
172 {UD} This is difficult to answer. The file `README' lists the architectures
173 GNU libc was known to run on *at some time*. This does not mean that it
174 still can be compiled and run on them now.
176 The systems glibc is known to work on as of this release, and most probably
180 i[3456]86-*-linux-gnu Linux-2.x on Intel
181 m68k-*-linux-gnu Linux-2.x on Motorola 680x0
182 alpha-*-linux-gnu Linux-2.x on DEC Alpha
183 powerpc-*-linux-gnu Linux and MkLinux on PowerPC systems
184 sparc-*-linux-gnu Linux-2.x on SPARC
185 sparc64-*-linux-gnu Linux-2.x on UltraSPARC
186 arm-*-none ARM standalone systems
187 arm-*-linux Linux-2.x on ARM
188 arm-*-linuxaout Linux-2.x on ARM using a.out binaries
190 Ports to other Linux platforms are in development, and may in fact work
191 already, but no one has sent us success reports for them. Currently no
192 ports to other operating systems are underway, although a few people have
195 If you have a system not listed above (or in the `README' file) and you are
196 really interested in porting it, contact
201 1.2. What compiler do I need to build GNU libc?
203 {UD} You must use GNU CC to compile GNU libc. A lot of extensions of GNU CC
204 are used to increase portability and speed.
206 GNU CC is found, like all other GNU packages, on
208 ftp://ftp.gnu.org/pub/gnu
210 and the many mirror sites. ftp.gnu.org is always overloaded, so try to find
211 a local mirror first.
213 You should always try to use the latest official release. Older versions
214 may not have all the features GNU libc requires. The current releases of
215 egcs (1.0.3 and 1.1.1) should work with the GNU C library (for powerpc see
216 question 1.5; for ARM see question 1.6).
218 {ZW} Due to problems with C++ exception handling, you must use EGCS (any
219 version) to compile version 2.1 of GNU libc. See question 2.8 for details.
222 1.3. When I try to compile glibc I get only error messages.
225 {UD} You definitely need GNU make to build GNU libc. No other make
226 program has the needed functionality.
228 We recommend version GNU make version 3.75 or 3.77. Versions before 3.75
229 have bugs and/or are missing features. Version 3.76 has bugs which
230 appear when building big projects like GNU libc. 3.76.1 appears to work but
231 some people have reported problems. If you build GNU make 3.77 from source,
232 please read question 4.6 first.
235 1.4. Do I need a special linker or assembler?
237 {ZW} If you want a shared library, you need a linker and assembler that
238 understand all the features of ELF, including weak and versioned symbols.
239 The static library can be compiled with less featureful tools, but lacks key
240 features such as NSS.
242 For Linux or Hurd, you want binutils 2.8.1.0.23, 2.9.1, or 2.9.1.0.15 or
243 higher. These are the only versions we've tested and found reliable. Other
244 versions after 2.8.1.0.23 may work but we don't recommend them, especially
245 not when C++ is involved. Earlier versions do not work at all.
247 Other operating systems may come with system tools that have all the
248 necessary features, but this is moot because glibc hasn't been ported to
252 1.5. Which compiler should I use for powerpc?
254 {GK} You want to use egcs 1.1 or later (together with the right versions
255 of all the other tools, of course).
257 In fact, egcs 1.1 has a bug that causes linuxthreads to be
258 miscompiled, resulting in segmentation faults when using condition
259 variables. There is a temporary patch at:
261 <http://discus.anu.edu.au/~geoffk/egcs-3.diff>
263 Later versions of egcs may fix this problem.
266 1.6. Which tools should I use for ARM?
268 {PB} You should use egcs 1.1 or a later version. For ELF systems some
269 changes are needed to the compiler; a patch against egcs-1.1.x can be found
272 <ftp://ftp.netwinder.org/users/p/philb/egcs-1.1.1pre2-diff-981126>
274 Binutils 2.9.1.0.16 or later is also required.
277 1.7. Do I need some more things to compile the GNU C Library?
279 {UD} Yes, there are some more :-).
281 * GNU gettext. This package contains the tools needed to construct
282 `message catalog' files containing translated versions of system
283 messages. See ftp://ftp.gnu.org/pub/gnu or better any mirror
284 site. (We distribute compiled message catalogs, but they may not be
287 * Some files are built with special tools. E.g., files ending in .gperf
288 need a `gperf' program. The GNU version (now available in a separate
289 package, formerly only as part of libg++) is known to work while some
290 vendor versions do not.
292 You should not need these tools unless you change the source files.
294 * Perl 5 is needed if you wish to test an installation of GNU libc
295 as the primary C library.
297 * When compiling for Linux, the header files of the Linux kernel must
298 be available to the compiler as <linux/*.h> and <asm/*.h>.
300 * lots of disk space (~170MB for i?86-linux; more for RISC platforms,
303 * plenty of time. Compiling just the shared and static libraries for
304 i?86-linux takes approximately 1h on an i586@133, or 2.5h on
305 i486@66, or 4.5h on i486@33. Multiply this by 1.5 or 2.0 if you
306 build profiling and/or the highly optimized version as well. For
307 Hurd systems times are much higher.
309 You should avoid compiling in a NFS mounted filesystem. This is
312 James Troup <J.J.Troup@comp.brad.ac.uk> reports a compile time of
313 45h34m for a full build (shared, static, and profiled) on Atari
314 Falcon (Motorola 68030 @ 16 Mhz, 14 Mb memory) and Jan Barte
315 <yann@plato.uni-paderborn.de> reports 22h48m on Atari TT030
316 (Motorola 68030 @ 32 Mhz, 34 Mb memory)
318 If you have some more measurements let me know.
321 1.8. What version of the Linux kernel headers should be used?
323 {AJ,UD} The headers from the most recent Linux kernel should be used. The
324 headers used while compiling the GNU C library and the kernel binary used
325 when using the library do not need to match. The GNU C library runs without
326 problems on kernels that are older than the kernel headers used. The other
327 way round (compiling the GNU C library with old kernel headers and running
328 on a recent kernel) does not necessarily work. For example you can't use
329 new kernel features if you used old kernel headers to compile the GNU C
332 {ZW} Even if you are using a 2.0 kernel on your machine, we recommend you
333 compile GNU libc with 2.2 kernel headers. That way you won't have to
334 recompile libc if you ever upgrade to kernel 2.2. To tell libc which
335 headers to use, give configure the --with-headers switch
336 (e.g. --with-headers=/usr/src/linux-2.2.0/include).
338 Note that you must configure the 2.2 kernel if you do this, otherwise libc
339 will be unable to find <linux/version.h>. Just change the current directory
340 to the root of the 2.2 tree and do `make include/linux/version.h'.
343 1.9. The compiler hangs while building iconvdata modules. What's
346 {ZW} This is a problem with old versions of GCC. Initialization of large
347 static arrays is very slow. The compiler will eventually finish; give it
350 The problem is fixed in egcs 1.1.
353 1.10. When I run `nm -u libc.so' on the produced library I still
354 find unresolved symbols. Can this be ok?
356 {UD} Yes, this is ok. There can be several kinds of unresolved symbols:
358 * magic symbols automatically generated by the linker. These have names
359 like __start_* and __stop_*
361 * symbols starting with _dl_* come from the dynamic linker
363 * weak symbols, which need not be resolved at all (fabs for example)
365 Generally, you should make sure you find a real program which produces
366 errors while linking before deciding there is a problem.
369 1.11. What are these `add-ons'?
371 {UD} To avoid complications with export rules or external source code some
372 optional parts of the libc are distributed as separate packages (e.g., the
373 crypt package, see question 2.5).
375 To use these packages as part of GNU libc, just unpack the tarfiles in the
376 libc source directory and tell the configuration script about them using the
377 --enable-add-ons option. If you give just --enable-add-ons configure tries
378 to find all the add-on packages in your source tree. This may not work. If
379 it doesn't, or if you want to select only a subset of the add-ons, give a
380 comma-separated list of the add-ons to enable:
382 configure --enable-add-ons=crypt,linuxthreads
386 Add-ons can add features (including entirely new shared libraries), override
387 files, provide support for additional architectures, and just about anything
388 else. The existing makefiles do most of the work; only some few stub rules
389 must be written to get everything running.
392 1.12. My XXX kernel emulates a floating-point coprocessor for me.
393 Should I enable --with-fp?
395 {ZW} An emulated FPU is just as good as a real one, as far as the C library
396 is concerned. You only need to say --without-fp if your machine has no way
397 to execute floating-point instructions.
399 People who are interested in squeezing the last drop of performance
400 out of their machine may wish to avoid the trap overhead, but this is
401 far more trouble than it's worth: you then have to compile
402 *everything* this way, including the compiler's internal libraries
403 (libgcc.a for GNU C), because the calling conventions change.
406 1.13. When compiling GNU libc I get lots of errors saying functions
407 in glibc are duplicated in libgcc.
409 {EY} This is *exactly* the same problem that I was having. The problem was
410 due to the fact that configure didn't correctly detect that the linker flag
411 --no-whole-archive was supported in my linker. In my case it was because I
412 had run ./configure with bogus CFLAGS, and the test failed.
414 One thing that is particularly annoying about this problem is that once this
415 is misdetected, running configure again won't fix it unless you first delete
418 {UD} Starting with glibc-2.0.3 there should be a better test to avoid some
419 problems of this kind. The setting of CFLAGS is checked at the very
420 beginning and if it is not usable `configure' will bark.
423 1.14. Why do I get messages about missing thread functions when I use
424 librt? I don't even use threads.
426 {UD} In this case you probably mixed up your installation. librt uses
427 threads internally and has implicit references to the thread library.
428 Normally these references are satisfied automatically but if the thread
429 library is not in the expected place you must tell the linker where it is.
430 When using GNU ld it works like this:
432 gcc -o foo foo.c -Wl,-rpath-link=/some/other/dir -lrt
434 The `/some/other/dir' should contain the thread library. `ld' will use the
435 given path to find the implicitly referenced library while not disturbing
439 1.15. What's the problem with configure --enable-omitfp?
441 {AJ} When --enable-omitfp is set the libraries are built without frame
442 pointers. Some compilers produce buggy code for this model and therefore we
443 don't advise using it at the moment.
445 If you use --enable-omitfp, you're on your own. If you encounter problems
446 with a library that was build this way, we advise you to rebuild the library
447 without --enable-omitfp. If the problem vanishes consider tracking the
448 problem down and report it as compiler failure.
450 Since a library built with --enable-omitfp is undebuggable on most systems,
451 debuggable libraries are also built - you can use them by appending "_g" to
454 The compilation of these extra libraries and the compiler optimizations slow
455 down the build process and need more disk space.
458 1.16. I get failures during `make check'. What should I do?
460 {AJ} The testsuite should compile and run cleanly on your system; every
461 failure should be looked into. Depending on the failures, you probably
462 should not install the library at all.
464 You should consider using the `glibcbug' script to report the failure,
465 providing as much detail as possible. If you run a test directly, please
466 remember to set up the environment correctly. You want to test the compiled
467 library - and not your installed one. The best way is to copy the exact
468 command line which failed and run the test from the subdirectory for this
471 There are some failures which are not directly related to the GNU libc:
472 - Some compilers produce buggy code. No compiler gets single precision
473 complex numbers correct on Alpha. Otherwise, the egcs 1.1 release should be
474 ok; gcc 2.8.1 might cause some failures; gcc 2.7.2.x is so buggy that
475 explicit checks have been used so that you can't build with it.
476 - The kernel might have bugs. For example on Linux/Alpha 2.0.34 the
477 floating point handling has quite a number of bugs and therefore most of
478 the test cases in the math subdirectory will fail. Linux 2.2 has
479 fixes for the floating point support on Alpha. The Linux/SPARC kernel has
480 also some bugs in the FPU emulation code (as of Linux 2.2.0).
483 1.17. What is symbol versioning good for? Do I need it?
485 {AJ} Symbol versioning solves problems that are related to interface
486 changes. One version of an interface might have been introduced in a
487 previous version of the GNU C library but the interface or the semantics of
488 the function has been changed in the meantime. For binary compatibility
489 with the old library, a newer library needs to still have the old interface
490 for old programs. On the other hand, new programs should use the new
491 interface. Symbol versioning is the solution for this problem. The GNU
492 libc version 2.1 uses symbol versioning by default if the installed binutils
495 We don't advise building without symbol versioning, since you lose binary
496 compatibility - forever! The binary compatibility you lose is not only
497 against the previous version of the GNU libc (version 2.0) but also against
501 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
503 2. Installation and configuration issues
505 2.1. Can I replace the libc on my Linux system with GNU libc?
507 {UD} You cannot replace any existing libc for Linux with GNU libc. It is
508 binary incompatible and therefore has a different major version. You can,
509 however, install it alongside your existing libc.
511 For Linux there are three major libc versions:
513 libc-5 original ELF libc
516 You can have any combination of these three installed. For more information
517 consult documentation for shared library handling. The Makefiles of GNU
518 libc will automatically generate the needed symbolic links which the linker
522 2.2. How do I configure GNU libc so that the essential libraries
523 like libc.so go into /lib and the other into /usr/lib?
525 {UD,AJ} Like all other GNU packages GNU libc is designed to use a base
526 directory and install all files relative to this. The default is
527 /usr/local, because this is safe (it will not damage the system if installed
528 there). If you wish to install GNU libc as the primary C library on your
529 system, set the base directory to /usr (i.e. run configure --prefix=/usr
530 <other_options>). Note that this can damage your system; see question 2.3 for
533 Some systems like Linux have a filesystem standard which makes a difference
534 between essential libraries and others. Essential libraries are placed in
535 /lib because this directory is required to be located on the same disk
536 partition as /. The /usr subtree might be found on another
537 partition/disk. If you configure for Linux with --prefix=/usr, then this
538 will be done automatically.
540 To install the essential libraries which come with GNU libc in /lib on
541 systems other than Linux one must explicitly request it. Autoconf has no
542 option for this so you have to use a `configparms' file (see the `INSTALL'
543 file for details). It should contain:
548 The first line specifies the directory for the essential libraries, the
549 second line the directory for system configuration files.
552 2.3. How should I avoid damaging my system when I install GNU libc?
554 {ZW} If you wish to be cautious, do not configure with --prefix=/usr. If
555 you don't specify a prefix, glibc will be installed in /usr/local, where it
556 will probably not break anything. (If you wish to be certain, set the
557 prefix to something like /usr/local/glibc2 which is not used for anything.)
559 The dangers when installing glibc in /usr are twofold:
561 * glibc will overwrite the headers in /usr/include. Other C libraries
562 install a different but overlapping set of headers there, so the
563 effect will probably be that you can't compile anything. You need to
564 rename /usr/include out of the way first. (Do not throw it away; you
565 will then lose the ability to compile programs against your old libc.)
567 * None of your old libraries, static or shared, can be used with a
568 different C library major version. For shared libraries this is not a
569 problem, because the filenames are different and the dynamic linker
570 will enforce the restriction. But static libraries have no version
571 information. You have to evacuate all the static libraries in
572 /usr/lib to a safe location.
574 The situation is rather similar to the move from a.out to ELF which
575 long-time Linux users will remember.
578 2.4. Do I need to use GNU CC to compile programs that will use the
581 {ZW} In theory, no; the linker does not care, and the headers are supposed
582 to check for GNU CC before using its extensions to the C language.
584 However, there are currently no ports of glibc to systems where another
585 compiler is the default, so no one has tested the headers extensively
586 against another compiler. You may therefore encounter difficulties. If you
587 do, please report them as bugs.
589 Also, in several places GNU extensions provide large benefits in code
590 quality. For example, the library has hand-optimized, inline assembly
591 versions of some string functions. These can only be used with GCC. See
592 question 3.8 for details.
595 2.5. When linking with the new libc I get unresolved symbols
596 `crypt' and `setkey'. Why aren't these functions in the
599 {UD} The US places restrictions on exporting cryptographic programs and
600 source code. Until this law gets abolished we cannot ship the cryptographic
601 functions together with glibc.
603 The functions are available, as an add-on (see question 1.11). People in the US
604 may get it from the same place they got GNU libc from. People outside the
605 US should get the code from ftp://ftp.ifi.uio.no/pub/gnu, or another archive
606 site outside the USA. The README explains how to install the sources.
608 If you already have the crypt code on your system the reason for the failure
609 is probably that you did not link with -lcrypt. The crypto functions are in
610 a separate library to make it possible to export GNU libc binaries from the
614 2.6. When I use GNU libc on my Linux system by linking against
615 the libc.so which comes with glibc all I get is a core dump.
617 {UD} On Linux, gcc sets the dynamic linker to /lib/ld-linux.so.1 unless the
618 user specifies a -dynamic-linker argument. This is the name of the libc5
619 dynamic linker, which does not work with glibc.
621 For casual use of GNU libc you can just specify to the linker
622 --dynamic-linker=/lib/ld-linux.so.2
624 which is the glibc dynamic linker, on Linux systems. On other systems the
625 name is /lib/ld.so.1. When linking via gcc, you've got to add
626 -Wl,--dynamic-linker=/lib/ld-linux.so.2
628 to the gcc command line.
630 To change your environment to use GNU libc for compiling you need to change
631 the `specs' file of your gcc. This file is normally found at
633 /usr/lib/gcc-lib/<arch>/<version>/specs
635 In this file you have to change a few things:
637 - change `ld-linux.so.1' to `ld-linux.so.2'
639 - remove all expression `%{...:-lgmon}'; there is no libgmon in glibc
641 - fix a minor bug by changing %{pipe:-} to %|
643 Here is what the gcc-2.7.2 specs file should look like when GNU libc is
646 -----------------------------------------------------------------------
648 %{V} %{v:%{!V:-V}} %{Qy:} %{!Qn:-Qy} %{n} %{T} %{Ym,*} %{Yd,*} %{Wa,*:%*}
654 %{fPIC:-D__PIC__ -D__pic__} %{fpic:-D__PIC__ -D__pic__} %{!m386:-D__i486__} %{posix:-D_POSIX_SOURCE} %{pthread:-D_REENTRANT}
663 %{!shared:crtend.o%s} %{shared:crtendS.o%s} crtn.o%s
666 -m elf_i386 %{shared:-shared} %{!shared: %{!ibcs: %{!static: %{rdynamic:-export-dynamic} %{!dynamic-linker:-dynamic-linker /lib/ld-linux.so.2}} %{static:-static}}}
669 %{!shared: %{pthread:-lpthread} %{profile:-lc_p} %{!profile: -lc}}
675 %{!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}
677 *switches_need_spaces:
681 %{funsigned-char:-D__CHAR_UNSIGNED__}
684 -D__ELF__ -Dunix -Di386 -Dlinux -Asystem(unix) -Asystem(posix) -Acpu(i386) -Amachine(i386)
692 -----------------------------------------------------------------------
694 Things get a bit more complicated if you have GNU libc installed in some
695 other place than /usr, i.e., if you do not want to use it instead of the old
696 libc. In this case the needed startup files and libraries are not found in
697 the regular places. So the specs file must tell the compiler and linker
700 Version 2.7.2.3 does and future versions of GCC will automatically
701 provide the correct specs.
704 2.7. Looking through the shared libc file I haven't found the
705 functions `stat', `lstat', `fstat', and `mknod' and while
706 linking on my Linux system I get error messages. How is
707 this supposed to work?
709 {RM} Believe it or not, stat and lstat (and fstat, and mknod) are supposed
710 to be undefined references in libc.so.6! Your problem is probably a missing
711 or incorrect /usr/lib/libc.so file; note that this is a small text file now,
712 not a symlink to libc.so.6. It should look something like this:
714 GROUP ( libc.so.6 libc_nonshared.a )
717 2.8. When I run an executable on one system which I compiled on
718 another, I get dynamic linker errors. Both systems have the same
719 version of glibc installed. What's wrong?
721 {ZW} Glibc on one of these systems was compiled with gcc 2.7 or 2.8, the
722 other with egcs (any version). Egcs has functions in its internal
723 `libgcc.a' to support exception handling with C++. They are linked into
724 any program or dynamic library compiled with egcs, whether it needs them or
725 not. Dynamic libraries then turn around and export those functions again
726 unless special steps are taken to prevent them.
728 When you link your program, it resolves its references to the exception
729 functions to the ones exported accidentally by libc.so. That works fine as
730 long as libc has those functions. On the other system, libc doesn't have
731 those functions because it was compiled by gcc 2.8, and you get undefined
732 symbol errors. The symbols in question are named things like
733 `__register_frame_info'.
735 For glibc 2.0, the workaround is to not compile libc with egcs. We've also
736 incorporated a patch which should prevent the EH functions sneaking into
737 libc. It doesn't matter what compiler you use to compile your program.
739 For glibc 2.1, we've chosen to do it the other way around: libc.so
740 explicitly provides the EH functions. This is to prevent other shared
741 libraries from doing it. You must therefore compile glibc 2.1 with EGCS.
742 Again, it doesn't matter what compiler you use for your programs.
745 2.9. How can I compile gcc 2.7.2.1 from the gcc source code using
748 {AJ} There's only correct support for glibc 2.0.x in gcc 2.7.2.3 or later.
749 But you should get at least gcc 2.8.1 or egcs 1.0.2 (or later versions)
753 2.10. The `gencat' utility cannot process the catalog sources which
754 were used on my Linux libc5 based system. Why?
756 {UD} The `gencat' utility provided with glibc complies to the XPG standard.
757 The older Linux version did not obey the standard, so they are not
760 To ease the transition from the Linux version some of the non-standard
761 features are also present in the `gencat' program of GNU libc. This mainly
762 includes the use of symbols for the message number and the automatic
763 generation of header files which contain the needed #defines to map the
766 Here is a simple SED script to convert at least some Linux specific catalog
767 files to the XPG4 form:
769 -----------------------------------------------------------------------
770 # Change catalog source in Linux specific format to standard XPG format.
771 # Ulrich Drepper <drepper@cygnus.com>, 1996.
775 s/\$ #\([^ ]*\).*/\1/
777 s/\$ #[^ ]* *\(.*\)/\$ \1/
783 s/\(.*\)\n\(.*\)/\2 \1/
785 -----------------------------------------------------------------------
788 2.11. Programs using libc have their messages translated, but other
789 behavior is not localized (e.g. collating order); why?
791 {ZW} Translated messages are automatically installed, but the locale
792 database that controls other behaviors is not. You need to run localedef to
793 install this database, after you have run `make install'. For example, to
794 set up the French Canadian locale, simply issue the command
796 localedef -i fr_CA -f ISO-8859-1 fr_CA
798 Please see localedata/README in the source tree for further details.
801 2.12. I have set up /etc/nis.conf, and the Linux libc 5 with NYS
802 works great. But the glibc NIS+ doesn't seem to work.
804 {TK} The glibc NIS+ implementation uses a /var/nis/NIS_COLD_START file for
805 storing information about the NIS+ server and their public keys, because the
806 nis.conf file does not contain all the necessary information. You have to
807 copy a NIS_COLD_START file from a Solaris client (the NIS_COLD_START file is
808 byte order independent) or generate it with nisinit from the nis-tools
809 package; available at
811 http://www-vt.uni-paderborn.de/~kukuk/linux/nisplus.html
814 2.13. I have killed ypbind to stop using NIS, but glibc
817 {TK} For faster NIS lookups, glibc uses the /var/yp/binding/ files from
818 ypbind. ypbind 3.3 and older versions don't always remove these files, so
819 glibc will continue to use them. Other BSD versions seem to work correctly.
820 Until ypbind 3.4 is released, you can find a patch at
822 ftp://ftp.kernel.org/pub/linux/utils/net/NIS/ypbind-3.3-glibc4.diff.gz
825 2.14. Under Linux/Alpha, I always get "do_ypcall: clnt_call:
826 RPC: Unable to receive; errno = Connection refused" when using NIS.
828 {TK} You need a ypbind version which is 64bit clean. Some versions are not
829 64bit clean. A 64bit clean implementation is ypbind-mt. For ypbind 3.3,
830 you need the patch from ftp.kernel.org (See the previous question). I don't
831 know about other versions.
834 2.15. After installing glibc name resolving doesn't work properly.
836 {AJ} You probably should read the manual section describing nsswitch.conf
837 (just type `info libc "NSS Configuration File"'). The NSS configuration
838 file is usually the culprit.
841 2.16. How do I create the databases for NSS?
843 {AJ} If you have an entry "db" in /etc/nsswitch.conf you should also create
844 the database files. The glibc sources contain a Makefile which does the
845 necessary conversion and calls to create those files. The file is
846 `db-Makefile' in the subdirectory `nss' and you can call it with `make -f
847 db-Makefile'. Please note that not all services are capable of using a
848 database. Currently passwd, group, ethers, protocol, rpc, services shadow
849 and netgroup are implemented.
852 2.17. I have /usr/include/net and /usr/include/scsi as symlinks
853 into my Linux source tree. Is that wrong?
855 {PB} This was necessary for libc5, but is not correct when using glibc.
856 Including the kernel header files directly in user programs usually does not
857 work (see question 3.5). glibc provides its own <net/*> and <scsi/*> header
858 files to replace them, and you may have to remove any symlink that you have
859 in place before you install glibc. However, /usr/include/asm and
860 /usr/include/linux should remain as they were.
863 2.18. Programs like `logname', `top', `uptime' `users', `w' and
864 `who', show incorrect information about the (number of)
865 users on my system. Why?
867 {MK} See question 3.2.
870 2.19. After upgrading to glibc 2.1 with symbol versioning I get
871 errors about undefined symbols. What went wrong?
873 {AJ} The problem is caused either by wrong program code or tools. In the
874 versioned libc a lot of symbols are now local that were global symbols in
875 previous versions. It seems that programs linked against older versions
876 often accidentally used libc global variables -- something that should not
879 The only way to fix this is to recompile your program. Sorry, that's the
880 price you might have to pay once for quite a number of advantages with
884 2.20. When I start the program XXX after upgrading the library
886 XXX: Symbol `_sys_errlist' has different size in shared
887 object, consider re-linking
888 Why? What should I do?
890 {UD} As the message says, relink the binary. The problem is that a few
891 symbols from the library can change in size and there is no way to avoid
892 this. _sys_errlist is a good example. Occasionally there are new error
893 numbers added to the kernel and this must be reflected at user level,
894 breaking programs that refer to them directly.
896 Such symbols should normally not be used at all. There are mechanisms to
897 avoid using them. In the case of _sys_errlist, there is the strerror()
898 function which should _always_ be used instead. So the correct fix is to
899 rewrite that part of the application.
901 In some situations (especially when testing a new library release) it might
902 be possible that a symbol changed size when that should not have happened.
903 So in case of doubt report such a warning message as a problem.
906 2.21. What do I need for C++ development?
908 {HJ,AJ} You need either egcs 1.1 which comes directly with libstdc++ or
909 gcc-2.8.1 together with libstdc++ 2.8.1.1. egcs 1.1 has the better C++
910 support and works directly with glibc 2.1. If you use gcc-2.8.1 with
911 libstdc++ 2.8.1.1, you need to modify libstdc++ a bit. A patch is available
913 ftp://alpha.gnu.org/gnu/libstdc++-2.8.1.1-glibc2.1-diff.gz
915 Please note that libg++ 2.7.2 (and the Linux Versions 2.7.2.x) doesn't work
916 very well with the GNU C library due to vtable thunks. If you're upgrading
917 from glibc 2.0.x to 2.1 you have to recompile libstdc++ since the library
918 compiled for 2.0 is not compatible due to the new Large File Support (LFS)
921 {UD} But since in the case of a shared libstdc++ the version numbers should
922 be different existing programs will continue to work.
925 2.22. Even statically linked programs need some shared libraries
926 which is not acceptable for me. What can I do?
928 {AJ} NSS (for details just type `info libc "Name Service Switch"') won't
929 work properly without shared libraries. NSS allows using different services
930 (e.g. NIS, files, db, hesiod) by just changing one configuration file
931 (/etc/nsswitch.conf) without relinking any programs. The only disadvantage
932 is that now static libraries need to access shared libraries. This is
933 handled transparently by the GNU C library.
935 A solution is to configure glibc with --enable-static-nss. In this case you
936 can create a static binary that will use only the services dns and files
937 (change /etc/nsswitch.conf for this). You need to link explicitly against
938 all these services. For example:
940 gcc -static test-netdb.c -o test-netdb.c \
941 -lc -lnss_files -lnss_dns -lresolv
943 The problem with this approach is that you've got to link every static
944 program that uses NSS routines with all those libraries.
946 {UD} In fact, one cannot say anymore that a libc compiled with this
947 option is using NSS. There is no switch anymore. Therefore it is
948 *highly* recommended *not* to use --enable-static-nss since this makes
949 the behaviour of the programs on the system inconsistent.
952 2.23. I just upgraded my Linux system to glibc and now I get
953 errors whenever I try to link any program.
955 {ZW} This happens when you have installed glibc as the primary C library but
956 have stray symbolic links pointing at your old C library. If the first
957 `libc.so' the linker finds is libc 5, it will use that. Your program
958 expects to be linked with glibc, so the link fails.
960 The most common case is that glibc put its `libc.so' in /usr/lib, but there
961 was a `libc.so' from libc 5 in /lib, which gets searched first. To fix the
962 problem, just delete /lib/libc.so. You may also need to delete other
963 symbolic links in /lib, such as /lib/libm.so if it points to libm.so.5.
965 {AJ} The perl script test-installation.pl which is run as last step during
966 an installation of glibc that is configured with --prefix=/usr should help
967 detect these situations. If the script reports problems, something is
971 2.24. When I use nscd the machine freezes.
973 {UD} You cannot use nscd with Linux 2.0.*. There is functionality missing
974 in the kernel and work-arounds are not suitable. Besides, some parts of the
975 kernel are too buggy when it comes to using threads.
977 If you need nscd, you have to use at least a 2.1 kernel.
979 Note that I have at this point no information about any other platform.
982 2.25. I need lots of open files. What do I have to do?
984 {AJ} This is at first a kernel issue. The kernel defines limits with
985 OPEN_MAX the number of simultaneous open files and with FD_SETSIZE the
986 number of used file descriptors. You need to change these values in your
987 kernel and recompile the kernel so that the kernel allows to use more open
988 files. You don't necessarily need to recompile the GNU C library since the
989 only place where OPEN_MAX and FD_SETSIZE is really needed in the library
990 itself is the size of fd_set which is used by select.
992 The GNU C library is now (nearly) select free. This means it internally has
993 no limits imposed by the `fd_set' type. Instead almost all places where the
994 functionality is needed the `poll' function is used.
996 If you increase the number of file descriptors in the kernel you don't need
997 to recompile the C library. The remaining select calls are in the RPC code.
998 If your RPC daemons don't need more than FD_SETSIZE file descriptors, you
999 don't need to change anything at all.
1001 {UD} You can always get the maximum number of file descriptors a process is
1002 allowed to have open at any time using
1004 number = sysconf (_SC_OPEN_MAX);
1006 This will work even if the kernel limits change.
1009 2.26. How do I get the same behavior on parsing /etc/passwd and
1010 /etc/group as I have with libc5 ?
1012 {TK} The name switch setup in /etc/nsswitch.conf selected by most Linux
1013 distributions does not support +/- and netgroup entries in the files like
1014 /etc/passwd. Though this is the preferred setup some people might have
1015 setups coming over from the libc5 days where it was the default to recognize
1016 lines like this. To get back to the old behaviour one simply has to change
1017 the rules for passwd, group, and shadow in the nsswitch.conf file as
1029 2.27. What needs to be recompiled when upgrading from glibc 2.0 to glibc
1032 {AJ,CG} If you just upgrade the glibc from 2.0.x (x <= 7) to 2.1, binaries
1033 that have been linked against glibc 2.0 will continue to work.
1035 If you compile your own binaries against glibc 2.1, you also need to
1036 recompile some other libraries. The problem is that libio had to be
1037 changed and therefore libraries that are based or depend on the libio
1038 of glibc, e.g. ncurses or slang, need to be recompiled. If you
1039 experience strange segmentation faults in your programs linked against
1040 glibc 2.1, you might need to recompile your libraries.
1042 Another problem is that older binaries that were linked statically against
1043 glibc 2.0 will reference the older nss modules (libnss_files.so.1 instead of
1044 libnss_files.so.2), so don't remove them. Also, the old glibc-2.0 compiled
1045 static libraries (libfoo.a) which happen to depend on the older libio
1046 behavior will be broken by the glibc 2.1 upgrade. We plan to produce a
1047 compatibility library that people will be able to link in if they want
1048 to compile a static library generated against glibc 2.0 into a program
1049 on a glibc 2.1 system. You just add -lcompat and you should be fine.
1051 The glibc-compat add-on will provide the libcompat.a library, the older
1052 nss modules, and a few other files. Together, they should make it
1053 possible to do development with old static libraries on a glibc 2.1
1054 system. This add-on is still in development. You can get it from
1055 ftp://alpha.gnu.org/gnu/glibc-compat-2.1.tar.gz
1056 but please keep in mind that it is experimental.
1059 2.28. Why is extracting files via tar so slow?
1061 {AJ} Extracting of tar archives might be quite slow since tar has to look up
1062 userid and groupids and doesn't cache negative results. If you have nis or
1063 nisplus in your /etc/nsswitch.conf for the passwd and/or group database,
1064 each file extractions needs a network connection. There are two possible
1067 - do you really need NIS/NIS+ (some Linux distributions add by default
1068 nis/nisplus even if it's not needed)? If not, just remove the entries.
1070 - if you need NIS/NIS+, use the Name Service Cache Daemon nscd that comes
1074 2.29. Compiling programs I get parse errors in libio.h (e.g. "parse error
1075 before `_IO_seekoff'"). How should I fix this?
1077 {AJ} You might get the following errors when upgrading to glibc 2.1:
1079 In file included from /usr/include/stdio.h:57,
1081 /usr/include/libio.h:335: parse error before `_IO_seekoff'
1082 /usr/include/libio.h:335: parse error before `_G_off64_t'
1083 /usr/include/libio.h:336: parse error before `_IO_seekpos'
1084 /usr/include/libio.h:336: parse error before `_G_fpos64_t'
1086 The problem is a wrong _G_config.h file in your include path. The
1087 _G_config.h file that comes with glibc 2.1 should be used and not one from
1088 libc5 or from a compiler directory. To check which _G_config.h file the
1089 compiler uses, compile your program with `gcc -E ...|grep G_config.h' and
1090 remove that file. Your compiler should pick up the file that has been
1091 installed by glibc 2.1 in your include directory.
1094 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1096 3. Source and binary incompatibilities, and what to do about them
1098 3.1. I expect GNU libc to be 100% source code compatible with
1099 the old Linux based GNU libc. Why isn't it like this?
1101 {DMT,UD} Not every extension in Linux libc's history was well thought-out.
1102 In fact it had a lot of problems with standards compliance and with
1103 cleanliness. With the introduction of a new version number these errors can
1104 now be corrected. Here is a list of the known source code
1107 * _GNU_SOURCE: glibc does not make the GNU extensions available
1108 automatically. If a program depends on GNU extensions or some
1109 other non-standard functionality, it is necessary to compile it
1110 with the C compiler option -D_GNU_SOURCE, or better, to put
1111 `#define _GNU_SOURCE' at the beginning of your source files, before
1112 any C library header files are included. This difference normally
1113 manifests itself in the form of missing prototypes and/or data type
1114 definitions. Thus, if you get such errors, the first thing you
1115 should do is try defining _GNU_SOURCE and see if that makes the
1118 For more information consult the file `NOTES' in the GNU C library
1121 * reboot(): GNU libc sanitizes the interface of reboot() to be more
1122 compatible with the interface used on other OSes. reboot() as
1123 implemented in glibc takes just one argument. This argument
1124 corresponds to the third argument of the Linux reboot system call.
1125 That is, a call of the form reboot(a, b, c) needs to be changed into
1126 reboot(c). Beside this the header <sys/reboot.h> defines the needed
1127 constants for the argument. These RB_* constants should be used
1128 instead of the cryptic magic numbers.
1130 * swapon(): the interface of this function didn't change, but the
1131 prototype is in a separate header file <sys/swap.h>. This header
1132 file also provides the SWAP_* constants defined by <linux/swap.h>;
1133 you should use them for the second argument to swapon().
1135 * errno: If a program uses the variable "errno", then it _must_
1136 include <errno.h>. The old libc often (erroneously) declared this
1137 variable implicitly as a side-effect of including other libc header
1138 files. glibc is careful to avoid such namespace pollution, which,
1139 in turn, means that you really need to include the header files that
1140 you depend on. This difference normally manifests itself in the
1141 form of the compiler complaining about references to an undeclared
1144 * Linux-specific syscalls: All Linux system calls now have appropriate
1145 library wrappers and corresponding declarations in various header files.
1146 This is because the syscall() macro that was traditionally used to
1147 work around missing syscall wrappers are inherently non-portable and
1148 error-prone. The following table lists all the new syscall stubs,
1149 the header-file declaring their interface and the system call name.
1151 syscall name: wrapper name: declaring header file:
1152 ------------- ------------- ----------------------
1153 bdflush bdflush <sys/kdaemon.h>
1154 syslog ksyslog_ctl <sys/klog.h>
1156 * lpd: Older versions of lpd depend on a routine called _validuser().
1157 The library does not provide this function, but instead provides
1158 __ivaliduser() which has a slightly different interface. Simply
1159 upgrading to a newer lpd should fix this problem (e.g., the 4.4BSD
1160 lpd is known to be working).
1162 * resolver functions/BIND: like on many other systems the functions of
1163 the resolver library are not included in libc itself. There is a
1164 separate library libresolv. If you get undefined symbol errors for
1165 symbols starting with `res_*' simply add -lresolv to your linker
1168 * the `signal' function's behavior corresponds to the BSD semantic and
1169 not the SysV semantic as it was in libc-5. The interface on all GNU
1170 systems shall be the same and BSD is the semantic of choice. To use
1171 the SysV behavior simply use `sysv_signal', or define _XOPEN_SOURCE.
1172 See question 3.7 for details.
1175 3.2. Why does getlogin() always return NULL on my Linux box?
1177 {UD} The GNU C library has a format for the UTMP and WTMP file which differs
1178 from what your system currently has. It was extended to fulfill the needs
1179 of the next years when IPv6 is introduced. The record size is different and
1180 some fields have different positions. The files written by functions from
1181 the one library cannot be read by functions from the other library. Sorry,
1182 but this is what a major release is for. It's better to have a cut now than
1183 having no means to support the new techniques later.
1185 {MK} There is however a (partial) solution for this problem. Please take a
1186 look at the file `login/README.utmpd'.
1189 3.3. Where are the DST_* constants found in <sys/time.h> on many
1192 {UD} These constants come from the old BSD days and are not used anymore
1193 (libc5 does not actually implement the handling although the constants are
1196 Instead GNU libc contains zone database support and compatibility code for
1197 POSIX TZ environment variable handling. For former is very much preferred
1201 3.4. The prototypes for `connect', `accept', `getsockopt',
1202 `setsockopt', `getsockname', `getpeername', `send',
1203 `sendto', and `recvfrom' are different in GNU libc from
1204 any other system I saw. This is a bug, isn't it?
1206 {UD} No, this is no bug. This version of GNU libc already follows the new
1207 Single Unix specifications (and I think the POSIX.1g draft which adopted the
1208 solution). The type for a parameter describing a size is now `socklen_t', a
1212 3.5. On Linux I've got problems with the declarations in Linux
1215 {UD,AJ} On Linux, the use of kernel headers is reduced to the minimum. This
1216 gives Linus the ability to change the headers more freely. Also, user
1217 programs are now insulated from changes in the size of kernel data
1220 For example, the sigset_t type is 32 or 64 bits wide in the kernel. In
1221 glibc it is 1024 bits wide. This guarantees that when the kernel gets a
1222 bigger sigset_t (for POSIX.1e realtime support, say) user programs will not
1223 have to be recompiled. Consult the header files for more information about
1226 Therefore you shouldn't include Linux kernel header files directly if glibc
1227 has defined a replacement. Otherwise you might get undefined results because
1231 3.6. I don't include any kernel headers myself but the compiler
1232 still complains about redeclarations of types in the kernel
1235 {UD} The kernel headers before Linux 2.1.61 and 2.0.32 don't work correctly
1236 with glibc. Compiling C programs is possible in most cases but C++ programs
1237 have (due to the change of the name lookups for `struct's) problems. One
1238 prominent example is `struct fd_set'.
1240 There might be some problems left but 2.1.61/2.0.32 fix most of the known
1241 ones. See the BUGS file for other known problems.
1244 3.7. Why don't signals interrupt system calls anymore?
1246 {ZW} By default GNU libc uses the BSD semantics for signal(), unlike Linux
1247 libc 5 which used System V semantics. This is partially for compatibility
1248 with other systems and partially because the BSD semantics tend to make
1249 programming with signals easier.
1251 There are three differences:
1253 * BSD-style signals that occur in the middle of a system call do not
1254 affect the system call; System V signals cause the system call to
1255 fail and set errno to EINTR.
1257 * BSD signal handlers remain installed once triggered. System V signal
1258 handlers work only once, so one must reinstall them each time.
1260 * A BSD signal is blocked during the execution of its handler. In other
1261 words, a handler for SIGCHLD (for example) does not need to worry about
1262 being interrupted by another SIGCHLD. It may, however, be interrupted
1265 There is general consensus that for `casual' programming with signals, the
1266 BSD semantics are preferable. You don't need to worry about system calls
1267 returning EINTR, and you don't need to worry about the race conditions
1268 associated with one-shot signal handlers.
1270 If you are porting an old program that relies on the old semantics, you can
1271 quickly fix the problem by changing signal() to sysv_signal() throughout.
1272 Alternatively, define _XOPEN_SOURCE before including <signal.h>.
1274 For new programs, the sigaction() function allows you to specify precisely
1275 how you want your signals to behave. All three differences listed above are
1276 individually switchable on a per-signal basis with this function.
1278 If all you want is for one specific signal to cause system calls to fail and
1279 return EINTR (for example, to implement a timeout) you can do this with
1283 3.8. I've got errors compiling code that uses certain string
1286 {AJ} glibc 2.1 has special string functions that are faster than the normal
1287 library functions. Some of the functions are additionally implemented as
1288 inline functions and others as macros. This might lead to problems with
1289 existing codes but it is explicitly allowed by ISO C.
1291 The optimized string functions are only used when compiling with
1292 optimizations (-O1 or higher). The behavior can be changed with two feature
1295 * __NO_STRING_INLINES: Don't do any string optimizations.
1296 * __USE_STRING_INLINES: Use assembly language inline functions (might
1297 increase code size dramatically).
1299 Since some of these string functions are now additionally defined as macros,
1300 code like "char *strncpy();" doesn't work anymore (and is unnecessary, since
1301 <string.h> has the necessary declarations). Either change your code or
1302 define __NO_STRING_INLINES.
1304 {UD} Another problem in this area is that gcc still has problems on machines
1305 with very few registers (e.g., ix86). The inline assembler code can require
1306 almost all the registers and the register allocator cannot always handle
1309 One can disable the string optimizations selectively. Instead of writing
1311 cp = strcpy (foo, "lkj");
1315 cp = (strcpy) (foo, "lkj");
1317 This disables the optimization for that specific call.
1320 3.9. I get compiler messages "Initializer element not constant" with
1321 stdin/stdout/stderr. Why?
1323 {RM,AJ} Constructs like:
1324 static FILE *InPtr = stdin;
1326 lead to this message. This is correct behaviour with glibc since stdin is
1327 not a constant expression. Please note that a strict reading of ISO C does
1328 not allow above constructs.
1330 One of the advantages of this is that you can assign to stdin, stdout, and
1331 stderr just like any other global variable (e.g. `stdout = my_stream;'),
1332 which can be very useful with custom streams that you can write with libio
1333 (but beware this is not necessarily portable). The reason to implement it
1334 this way were versioning problems with the size of the FILE structure.
1336 To fix those programs you've got to initialize the variable at run time.
1337 This can be done, e.g. in main, like:
1345 or by constructors (beware this is gcc specific):
1348 static void inPtr_construct (void) __attribute__((constructor));
1349 static void inPtr_construct (void) { InPtr = stdin; }
1352 3.10. I can't compile with gcc -traditional (or
1353 -traditional-cpp). Why?
1355 {AJ} glibc2 does break -traditional and -traditonal-cpp - and will continue
1356 to do so. For example constructs of the form:
1362 are useful for debugging purposes (you can use foo with your debugger that's
1363 why we need the enum) and for compatibility (other systems use defines and
1367 3.11. I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible?
1369 {AJ} The GNU C library is compatible with the ANSI/ISO C standard. If
1370 you're using `gcc -ansi', the glibc includes which are specified in the
1371 standard follow the standard. The ANSI/ISO C standard defines what has to be
1372 in the include files - and also states that nothing else should be in the
1373 include files (btw. you can still enable additional standards with feature
1376 The GNU C library is conforming to ANSI/ISO C - if and only if you're only
1377 using the headers and library functions defined in the standard.
1380 3.12. I can't access some functions anymore. nm shows that they do
1381 exist but linking fails nevertheless.
1383 {AJ} With the introduction of versioning in glibc 2.1 it is possible to
1384 export only those identifiers (functions, variables) that are really needed
1385 by application programs and by other parts of glibc. This way a lot of
1386 internal interfaces are now hidden. nm will still show those identifiers
1387 but marking them as internal. ISO C states that identifiers beginning with
1388 an underscore are internal to the libc. An application program normally
1389 shouldn't use those internal interfaces (there are exceptions,
1390 e.g. __ivaliduser). If a program uses these interfaces, it's broken. These
1391 internal interfaces might change between glibc releases or dropped
1395 3.13. When using the db-2 library which comes with glibc is used in
1396 the Perl db modules the testsuite is not passed. This did not
1397 happen with db-1, gdbm, or ndbm.
1399 {UD} You are using an outdated copy of the DB_File Perl module. In fact db-2
1400 finally removed the handling of zero-sized keys which was one of the features
1401 tested by the old Perl testsuite and therefore you see an error. But this
1402 never was documented and guaranteed, only broken programs used this feature.
1404 Consequently db-2 does not need to support this feature and instead signals
1405 an error which leads to easier debugging. The DB_File module maintainer
1406 Paul Marquess <pmarquess@bfsec.bt.co.uk> acknowledged this change and fixed
1407 the testsuite so that if you use DB_File v1.60 or later you should not have
1408 any more problems with db-2.
1411 3.14. The pow() inline function I get when including <math.h> is broken.
1412 I get segmentation faults when I run the program.
1414 {UD} Nope, the implementation is correct. The problem is with egcs version
1415 prior to 1.1. I.e., egcs 1.0 to 1.0.3 are all broken (at least on Intel).
1416 If you have to use this compiler you must define __NO_MATH_INLINES before
1417 including <math.h> to prevent the inline functions from being used. egcs 1.1
1418 fixes the problem. I don't know about gcc 2.8 and 2.8.1.
1421 3.15. The sys/sem.h file lacks the definition of `union semun'.
1423 {UD} Nope. This union has to be provided by the user program. Former glibc
1424 versions defined this but it was an error since it does not make much sense
1425 when thinking about it. The standards describing the System V IPC functions
1426 define it this way and therefore programs must be adopted.
1429 3.16. Why has <netinet/ip_fw.h> disappeared?
1431 {AJ} The corresponding Linux kernel data structures and constants are
1432 totally different in Linux 2.0 and Linux 2.2. This situation has to be
1433 taken care in user programs using the firewall structures and therefore
1434 those programs (ipfw is AFAIK the only one) should deal with this problem
1438 3.17. I get floods of warnings when I use -Wconversion and include
1439 <string.h> or <math.h>.
1441 {ZW} <string.h> and <math.h> intentionally use prototypes to override
1442 argument promotion. -Wconversion warns about all these. You can safely
1443 ignore the warnings.
1445 -Wconversion isn't really intended for production use, only for shakedown
1446 compiles after converting an old program to standard C.
1449 3.18. After upgrading to glibc 2.1, I receive errors about
1450 unresolved symbols, like `_dl_initial_searchlist' and can not
1451 execute any binaries. What went wrong?
1453 {AJ} This normally happens if your libc and ld (dynamic linker) are from
1454 different releases of glibc. For example, the dynamic linker
1455 /lib/ld-linux.so.2 comes from glibc 2.0.x, but the version of libc.so.6 is
1458 The path /lib/ld-linux.so.2 is hardcoded in every glibc2 binary but
1459 libc.so.6 is searched via /etc/ld.so.cache and in some special directories
1460 like /lib and /usr/lib. If you run configure with another prefix than /usr
1461 and put this prefix before /lib in /etc/ld.so.conf, your system will break.
1463 So what can you do? Either of the following should work:
1465 * Run `configure' with the same prefix argument you've used for glibc 2.0.x
1466 so that the same paths are used.
1467 * Replace /lib/ld-linux.so.2 with a link to the dynamic linker from glibc
1470 You can even call the dynamic linker by hand if everything fails. You've
1471 got to set LD_LIBRARY_PATH so that the corresponding libc is found and also
1472 need to provide an absolute path to your binary:
1474 LD_LIBRARY_PATH=<path-where-libc.so.6-lives> \
1475 <path-where-corresponding-dynamic-linker-lives>/ld-linux.so.2 \
1476 <path-to-binary>/binary
1478 For example `LD_LIBRARY_PATH=/libold /libold/ld-linux.so.2 /bin/mv ...'
1479 might be useful in fixing a broken system (if /libold contains dynamic
1480 linker and corresponding libc).
1482 With that command line no path is used. To further debug problems with the
1483 dynamic linker, use the LD_DEBUG environment variable, e.g.
1484 `LD_DEBUG=help echo' for the help text.
1486 If you just want to test this release, don't put the lib directory in
1487 /etc/ld.so.conf. You can call programs directly with full paths (as above).
1488 When compiling new programs against glibc 2.1, you've got to specify the
1489 correct paths to the compiler (option -I with gcc) and linker (options
1490 --dynamic-linker, -L and --rpath).
1493 3.19. bonnie reports that char i/o with glibc 2 is much slower than with
1494 libc5. What can be done?
1496 {AJ} The GNU C library uses thread safe functions by default and libc5 used
1497 non thread safe versions. The non thread safe functions have in glibc the
1498 suffix `_unlocked', for details check <stdio.h>. Using `putc_unlocked' etc.
1499 instead of `putc' should give nearly the same speed with bonnie (bonnie is a
1500 benchmark program for measuring disk access).
1503 3.20. Programs compiled with glibc 2.1 can't read db files made with glibc
1504 2.0. What has changed that programs like rpm break?
1506 {AJ} The GNU C library 2.1 uses db2 instead of db1 which was used in version
1507 2.0. The internal formats of the actual db files are different. To convert
1508 the db files from db1 format to db2 format, you can use the programs
1509 `db_dump185' and `db_load'. Alternativly programs can be linked with db1
1510 using `-ldb1' instead of linking with db2 which uses `-ldb'. Linking with
1511 db1 might be preferable if older programs need to access the db file.
1513 db2 supports the old db1 programming interface and also a new programming
1514 interface. For compilation with the old API, <db_185.h> has to be included
1515 (and not <db.h>) and you can link with either `-ldb1' or `-ldb' for either
1519 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1523 4.1. After I changed configure.in I get `Autoconf version X.Y.
1524 or higher is required for this script'. What can I do?
1526 {UD} You have to get the specified autoconf version (or a later one)
1527 from your favorite mirror of ftp.gnu.org.
1530 4.2. When I try to compile code which uses IPv6 headers and
1531 definitions on my Linux 2.x.y system I am in trouble.
1532 Nothing seems to work.
1534 {UD} The problem is that IPv6 development still has not reached a point
1535 where the headers are stable. There are still lots of incompatible changes
1536 made and the libc headers have to follow.
1538 {PB} The 2.1 release of GNU libc aims to comply with the current versions of
1539 all the relevant standards. The IPv6 support libraries for older Linux
1540 systems used a different naming convention and so code written to work with
1541 them may need to be modified. If the standards make incompatible changes in
1542 the future then the libc may need to change again.
1544 IPv6 will not work with a 2.0.x kernel. When kernel 2.2 is released it
1545 should contain all the necessary support; until then you should use the
1546 latest 2.1.x release you can find. As of 98/11/26 the currently recommended
1547 kernel for IPv6 is 2.1.129.
1549 Also, as of the 2.1 release the IPv6 API provided by GNU libc is not
1550 100% complete. In particular the getipnodebyname and getipnodebyaddr
1551 functions are not implemented.
1554 4.3. When I set the timezone by setting the TZ environment variable
1555 to EST5EDT things go wrong since glibc computes the wrong time
1556 from this information.
1558 {UD} The problem is that people still use the braindamaged POSIX method to
1559 select the timezone using the TZ environment variable with a format EST5EDT
1560 or whatever. People, if you insist on using TZ instead of the timezone
1561 database (see below), read the POSIX standard, the implemented behaviour is
1562 correct! What you see is in fact the result of the decisions made while
1563 POSIX.1 was created. We've only implemented the handling of TZ this way to
1564 be POSIX compliant. It is not really meant to be used.
1566 The alternative approach to handle timezones which is implemented is the
1567 correct one to use: use the timezone database. This avoids all the problems
1568 the POSIX method has plus it is much easier to use. Simply run the tzselect
1569 shell script, answer the question and use the name printed in the end by
1570 making a symlink /etc/localtime pointing to /usr/share/zoneinfo/NAME (NAME
1571 is the returned value from tzselect). That's all. You never again have to
1574 So, please avoid sending bug reports about time related problems if you use
1575 the POSIX method and you have not verified something is really broken by
1576 reading the POSIX standards.
1579 4.4. What other sources of documentation about glibc are available?
1581 {AJ} The FSF has a page about the GNU C library at
1582 <http://www.gnu.org/software/libc/>. The problem data base of open and
1583 solved bugs in GNU libc is available at
1584 <http://www-gnats.gnu.org:8080/cgi-bin/wwwgnats.pl>. Eric Green has written
1585 a HowTo for converting from Linux libc5 to glibc2. The HowTo is accessable
1586 via the FSF page and at <http://www.imaxx.net/~thrytis/glibc>. Frodo
1587 Looijaard describes a different way installing glibc2 as secondary libc at
1588 <http://huizen.dds.nl/~frodol/glibc>.
1590 Please note that this is not a complete list.
1593 4.5. The timezone string for Sydney/Australia is wrong since even when
1594 daylight saving time is in effect the timezone string is EST.
1596 {UD} The problem for some timezones is that the local authorities decided
1597 to use the term "summer time" instead of "daylight saving time". In this
1598 case the abbreviation character `S' is the same as the standard one. So,
1601 Eastern Standard Time = EST
1602 Eastern Summer Time = EST
1604 Great! To get this bug fixed convince the authorities to change the laws
1605 and regulations of the country this effects. glibc behaves correctly.
1608 4.6. I've build make 3.77 against glibc 2.1 and now make gets
1609 segmentation faults.
1611 {AJ} GNU make 3.77 has support for 64 bit filesystems which is slightly
1612 broken (and one of the new features in the GNU C library 2.1 is 64 bit
1613 filesystem support :-( ). To get a working make you can use either make
1614 3.75 or patch 3.77. A working patch is available via RedHat's Rawhide server
1615 (ftp://rawhide.redhat.com/SRPMS/SRPMS/make-3.77-*src.rpm).
1618 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
1620 Answers were given by:
1621 {UD} Ulrich Drepper, <drepper@cygnus.com>
1622 {DMT} David Mosberger-Tang, <davidm@AZStarNet.com>
1623 {RM} Roland McGrath, <roland@gnu.org>
1624 {AJ} Andreas Jaeger, <aj@arthur.rhein-neckar.de>
1625 {EY} Eric Youngdale, <eric@andante.jic.com>
1626 {PB} Phil Blundell, <Philip.Blundell@pobox.com>
1627 {MK} Mark Kettenis, <kettenis@phys.uva.nl>
1628 {ZW} Zack Weinberg, <zack@rabi.phys.columbia.edu>
1629 {TK} Thorsten Kukuk, <kukuk@vt.uni-paderborn.de>
1630 {GK} Geoffrey Keating, <geoffk@ozemail.com.au>
1631 {HJ} H.J. Lu, <hjl@gnu.org>
1632 {CG} Cristian Gafton, <gafton@redhat.com>
1636 outline-regexp:"\\?"