]>
Commit | Line | Data |
---|---|---|
61952351 | 1 | Frequently Asked Questions about the GNU C Library |
f8cac037 | 2 | |
f12944ec UD |
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. | |
f8cac037 | 6 | |
f12944ec | 7 | The GNU C library is very complex. The installation process has not been |
fdacb17d | 8 | completely automated; there are too many variables. You can do substantial |
f12944ec UD |
9 | damage to your system by installing the library incorrectly. Make sure you |
10 | understand what you are undertaking before you begin. | |
f8cac037 | 11 | |
41f27456 RM |
12 | If you have any questions you think should be answered in this document, |
13 | please let me know. | |
f8cac037 RM |
14 | |
15 | --drepper@cygnus.com | |
16 | \f | |
61952351 UD |
17 | ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ |
18 | ||
19 | 1. Compiling glibc | |
20 | ||
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. | |
24 | What's wrong? | |
5edb9387 | 25 | 1.4. Do I need a special linker or assembler? |
8619129f | 26 | 1.5. Which compiler should I use for powerpc? |
348ed515 UD |
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 | |
f12944ec | 31 | wrong? |
348ed515 | 32 | 1.10. When I run `nm -u libc.so' on the produced library I still |
61952351 | 33 | find unresolved symbols. Can this be ok? |
348ed515 UD |
34 | 1.11. What are these `add-ons'? |
35 | 1.12. My XXX kernel emulates a floating-point coprocessor for me. | |
61952351 | 36 | Should I enable --with-fp? |
348ed515 | 37 | 1.13. When compiling GNU libc I get lots of errors saying functions |
61952351 | 38 | in glibc are duplicated in libgcc. |
348ed515 | 39 | 1.14. Why do I get messages about missing thread functions when I use |
a35cb74d | 40 | librt? I don't even use threads. |
348ed515 | 41 | 1.15. What's the problem with configure --enable-omitfp? |
b1418d8f | 42 | 1.16. I get failures during `make check'. What should I do? |
348ed515 | 43 | 1.17. What is symbol versioning good for? Do I need it? |
61952351 UD |
44 | |
45 | 2. Installation and configuration issues | |
46 | ||
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 | |
52 | GNU C Library? | |
53 | 2.5. When linking with the new libc I get unresolved symbols | |
54 | `crypt' and `setkey'. Why aren't these functions in the | |
55 | libc anymore? | |
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? | |
5edb9387 UD |
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 | |
61952351 | 66 | glibc 2.x? |
5edb9387 | 67 | 2.10. The `gencat' utility cannot process the catalog sources which |
61952351 | 68 | were used on my Linux libc5 based system. Why? |
5edb9387 | 69 | 2.11. Programs using libc have their messages translated, but other |
a35cb74d | 70 | behavior is not localized (e.g. collating order); why? |
5edb9387 | 71 | 2.12. I have set up /etc/nis.conf, and the Linux libc 5 with NYS |
61952351 | 72 | works great. But the glibc NIS+ doesn't seem to work. |
5edb9387 | 73 | 2.13. I have killed ypbind to stop using NIS, but glibc |
3dcf8ea6 | 74 | continues using NIS. |
5edb9387 | 75 | 2.14. Under Linux/Alpha, I always get "do_ypcall: clnt_call: |
3dcf8ea6 | 76 | RPC: Unable to receive; errno = Connection refused" when using NIS. |
5edb9387 UD |
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 | |
61952351 | 80 | into my Linux source tree. Is that wrong? |
5edb9387 | 81 | 2.18. Programs like `logname', `top', `uptime' `users', `w' and |
61952351 UD |
82 | `who', show incorrect information about the (number of) |
83 | users on my system. Why? | |
5edb9387 | 84 | 2.19. After upgrading to glibc 2.1 with symbol versioning I get |
61952351 | 85 | errors about undefined symbols. What went wrong? |
5edb9387 | 86 | 2.20. When I start the program XXX after upgrading the library |
61952351 UD |
87 | I get |
88 | XXX: Symbol `_sys_errlist' has different size in shared | |
89 | object, consider re-linking | |
90 | Why? What should I do? | |
5edb9387 UD |
91 | 2.21. What do I need for C++ development? |
92 | 2.22. Even statically linked programs need some shared libraries | |
ff44f2a5 | 93 | which is not acceptable for me. What can I do? |
5edb9387 | 94 | 2.23. I just upgraded my Linux system to glibc and now I get |
fdacb17d | 95 | errors whenever I try to link any program. |
5edb9387 | 96 | 2.24. When I use nscd the machine freezes. |
0155a773 | 97 | 2.25. I need lots of open files. What do I have to do? |
7db169c9 UD |
98 | 2.26. How do I get the same behavior on parsing /etc/passwd and |
99 | /etc/group as I have with libc5 ? | |
b710a6e2 UD |
100 | 2.27. What needs to be recompiled when upgrading from glibc 2.0 to glibc |
101 | 2.1? | |
b7398be5 | 102 | 2.28. Why is extracting files via tar so slow? |
2ee511d9 UD |
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? | |
4f7ea427 UD |
105 | 2.30. After upgrading to glibc 2.1, libraries that were compiled against |
106 | glibc 2.0.x don't work anymore. | |
61952351 UD |
107 | |
108 | 3. Source and binary incompatibilities, and what to do about them | |
109 | ||
110 | 3.1. I expect GNU libc to be 100% source code compatible with | |
111 | the old Linux based GNU libc. Why isn't it like this? | |
112 | 3.2. Why does getlogin() always return NULL on my Linux box? | |
113 | 3.3. Where are the DST_* constants found in <sys/time.h> on many | |
114 | systems? | |
115 | 3.4. The prototypes for `connect', `accept', `getsockopt', | |
116 | `setsockopt', `getsockname', `getpeername', `send', | |
117 | `sendto', and `recvfrom' are different in GNU libc from | |
118 | any other system I saw. This is a bug, isn't it? | |
119 | 3.5. On Linux I've got problems with the declarations in Linux | |
120 | kernel headers. | |
121 | 3.6. I don't include any kernel headers myself but the compiler | |
122 | still complains about redeclarations of types in the kernel | |
123 | headers. | |
124 | 3.7. Why don't signals interrupt system calls anymore? | |
125 | 3.8. I've got errors compiling code that uses certain string | |
126 | functions. Why? | |
4775243a UD |
127 | 3.9. I get compiler messages "Initializer element not constant" with |
128 | stdin/stdout/stderr. Why? | |
129 | 3.10. I can't compile with gcc -traditional (or | |
130 | -traditional-cpp). Why? | |
131 | 3.11. I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible? | |
a35cb74d UD |
132 | 3.12. I can't access some functions anymore. nm shows that they do |
133 | exist but linking fails nevertheless. | |
a5f4e34a UD |
134 | 3.13. When using the db-2 library which comes with glibc is used in |
135 | the Perl db modules the testsuite is not passed. This did not | |
136 | happen with db-1, gdbm, or ndbm. | |
5148d49f UD |
137 | 3.14. The pow() inline function I get when including <math.h> is broken. |
138 | I get segmentation faults when I run the program. | |
05f732b3 | 139 | 3.15. The sys/sem.h file lacks the definition of `union semun'. |
33127459 | 140 | 3.16. Why has <netinet/ip_fw.h> disappeared? |
28ab8526 UD |
141 | 3.17. I get floods of warnings when I use -Wconversion and include |
142 | <string.h> or <math.h>. | |
5ff1a70a UD |
143 | 3.18. After upgrading to glibc 2.1, I receive errors about |
144 | unresolved symbols, like `_dl_initial_searchlist' and can not | |
145 | execute any binaries. What went wrong? | |
7d1de115 UD |
146 | 3.19. bonnie reports that char i/o with glibc 2 is much slower than with |
147 | libc5. What can be done? | |
b93492aa UD |
148 | 3.20. Programs compiled with glibc 2.1 can't read db files made with glibc |
149 | 2.0. What has changed that programs like rpm break? | |
b5a9efcd UD |
150 | 3.21. Autoconf's AC_CHECK_FUNC macro reports that a function exists, but |
151 | when I try to use it, it always returns -1 and sets errno to ENOSYS. | |
152 | 3.22. My program segfaults when I call fclose() on the FILE* returned | |
153 | from setmntent(). Is this a glibc bug? | |
61952351 | 154 | |
7d1de115 | 155 | 4. Miscellaneous |
61952351 | 156 | |
7d1de115 | 157 | 4.1. After I changed configure.in I get `Autoconf version X.Y. |
61952351 | 158 | or higher is required for this script'. What can I do? |
7d1de115 | 159 | 4.2. When I try to compile code which uses IPv6 headers and |
61952351 UD |
160 | definitions on my Linux 2.x.y system I am in trouble. |
161 | Nothing seems to work. | |
7d1de115 | 162 | 4.3. When I set the timezone by setting the TZ environment variable |
ff44f2a5 UD |
163 | to EST5EDT things go wrong since glibc computes the wrong time |
164 | from this information. | |
7d1de115 UD |
165 | 4.4. What other sources of documentation about glibc are available? |
166 | 4.5. The timezone string for Sydney/Australia is wrong since even when | |
348ed515 | 167 | daylight saving time is in effect the timezone string is EST. |
7d1de115 | 168 | 4.6. I've build make 3.77 against glibc 2.1 and now make gets |
eeabe877 | 169 | segmentation faults. |
f8cac037 | 170 | |
61952351 UD |
171 | \f |
172 | ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ | |
f4017d20 | 173 | |
61952351 | 174 | 1. Compiling glibc |
04be94a8 | 175 | |
61952351 | 176 | 1.1. What systems does the GNU C Library run on? |
613a76ff | 177 | |
f12944ec UD |
178 | {UD} This is difficult to answer. The file `README' lists the architectures |
179 | GNU libc was known to run on *at some time*. This does not mean that it | |
180 | still can be compiled and run on them now. | |
f8cac037 | 181 | |
f12944ec UD |
182 | The systems glibc is known to work on as of this release, and most probably |
183 | in the future, are: | |
f8cac037 RM |
184 | |
185 | *-*-gnu GNU Hurd | |
4775243a UD |
186 | i[3456]86-*-linux-gnu Linux-2.x on Intel |
187 | m68k-*-linux-gnu Linux-2.x on Motorola 680x0 | |
188 | alpha-*-linux-gnu Linux-2.x on DEC Alpha | |
9a0a462c | 189 | powerpc-*-linux-gnu Linux and MkLinux on PowerPC systems |
4775243a UD |
190 | sparc-*-linux-gnu Linux-2.x on SPARC |
191 | sparc64-*-linux-gnu Linux-2.x on UltraSPARC | |
ff44f2a5 | 192 | arm-*-none ARM standalone systems |
348ed515 | 193 | arm-*-linux Linux-2.x on ARM |
ff44f2a5 | 194 | arm-*-linuxaout Linux-2.x on ARM using a.out binaries |
f8cac037 | 195 | |
f12944ec UD |
196 | Ports to other Linux platforms are in development, and may in fact work |
197 | already, but no one has sent us success reports for them. Currently no | |
198 | ports to other operating systems are underway, although a few people have | |
199 | expressed interest. | |
f8cac037 | 200 | |
f12944ec UD |
201 | If you have a system not listed above (or in the `README' file) and you are |
202 | really interested in porting it, contact | |
f8cac037 | 203 | |
4775243a | 204 | <bug-glibc@gnu.org> |
f8cac037 RM |
205 | |
206 | ||
61952351 | 207 | 1.2. What compiler do I need to build GNU libc? |
f8cac037 | 208 | |
f12944ec UD |
209 | {UD} You must use GNU CC to compile GNU libc. A lot of extensions of GNU CC |
210 | are used to increase portability and speed. | |
f8cac037 | 211 | |
61952351 | 212 | GNU CC is found, like all other GNU packages, on |
f12944ec | 213 | |
a35cb74d | 214 | ftp://ftp.gnu.org/pub/gnu |
f12944ec | 215 | |
a35cb74d | 216 | and the many mirror sites. ftp.gnu.org is always overloaded, so try to find |
61952351 | 217 | a local mirror first. |
f8cac037 | 218 | |
b0610668 | 219 | You should always try to use the latest official release. Older versions |
f12944ec | 220 | may not have all the features GNU libc requires. The current releases of |
b8f558b7 UD |
221 | egcs (1.0.3 and 1.1.1) should work with the GNU C library (for powerpc see |
222 | question 1.5; for ARM see question 1.6). | |
f8cac037 | 223 | |
95f7cecb UD |
224 | While the GNU CC should be able to compile glibc it is nevertheless adviced |
225 | to use EGCS. Comparing the sizes of glibc on Intel compiled with a recent | |
226 | EGCS and gcc 2.8.1 shows this: | |
227 | ||
228 | text data bss dec hex filename | |
229 | egcs-2.93.10 862897 15944 12824 891665 d9b11 libc.so | |
230 | gcc-2.8.1 959965 16468 12152 988585 f15a9 libc.so | |
231 | ||
232 | Make up your own decision. | |
5edb9387 | 233 | |
f8cac037 | 234 | |
61952351 UD |
235 | 1.3. When I try to compile glibc I get only error messages. |
236 | What's wrong? | |
f8cac037 | 237 | |
b1418d8f | 238 | {UD} You definitely need GNU make to build GNU libc. No other make |
f12944ec | 239 | program has the needed functionality. |
f8cac037 | 240 | |
5edb9387 UD |
241 | We recommend version GNU make version 3.75 or 3.77. Versions before 3.75 |
242 | have bugs and/or are missing features. Version 3.76 has bugs which | |
243 | appear when building big projects like GNU libc. 3.76.1 appears to work but | |
7db169c9 | 244 | some people have reported problems. If you build GNU make 3.77 from source, |
7d1de115 | 245 | please read question 4.6 first. |
f8cac037 | 246 | |
f8cac037 | 247 | |
5edb9387 | 248 | 1.4. Do I need a special linker or assembler? |
f8cac037 | 249 | |
5edb9387 UD |
250 | {ZW} If you want a shared library, you need a linker and assembler that |
251 | understand all the features of ELF, including weak and versioned symbols. | |
252 | The static library can be compiled with less featureful tools, but lacks key | |
253 | features such as NSS. | |
41f27456 | 254 | |
5edb9387 UD |
255 | For Linux or Hurd, you want binutils 2.8.1.0.23, 2.9.1, or 2.9.1.0.15 or |
256 | higher. These are the only versions we've tested and found reliable. Other | |
257 | versions after 2.8.1.0.23 may work but we don't recommend them, especially | |
258 | not when C++ is involved. Earlier versions do not work at all. | |
a379e56a | 259 | |
5edb9387 UD |
260 | Other operating systems may come with system tools that have all the |
261 | necessary features, but this is moot because glibc hasn't been ported to | |
262 | them. | |
f8cac037 | 263 | |
f8cac037 | 264 | |
8619129f | 265 | 1.5. Which compiler should I use for powerpc? |
4775243a | 266 | |
48244d09 | 267 | {GK} You want to use egcs 1.1 or later (together with the right versions |
f12944ec | 268 | of all the other tools, of course). |
4775243a | 269 | |
48244d09 UD |
270 | In fact, egcs 1.1 has a bug that causes linuxthreads to be |
271 | miscompiled, resulting in segmentation faults when using condition | |
272 | variables. There is a temporary patch at: | |
4775243a | 273 | |
48244d09 | 274 | <http://discus.anu.edu.au/~geoffk/egcs-3.diff> |
4775243a | 275 | |
48244d09 | 276 | Later versions of egcs may fix this problem. |
4775243a UD |
277 | |
278 | ||
348ed515 UD |
279 | 1.6. Which tools should I use for ARM? |
280 | ||
281 | {PB} You should use egcs 1.1 or a later version. For ELF systems some | |
282 | changes are needed to the compiler; a patch against egcs-1.1.x can be found | |
283 | at: | |
284 | ||
285 | <ftp://ftp.netwinder.org/users/p/philb/egcs-1.1.1pre2-diff-981126> | |
286 | ||
287 | Binutils 2.9.1.0.16 or later is also required. | |
288 | ||
289 | ||
290 | 1.7. Do I need some more things to compile the GNU C Library? | |
f8cac037 | 291 | |
61952351 | 292 | {UD} Yes, there are some more :-). |
78b5ba3e | 293 | |
61952351 UD |
294 | * GNU gettext. This package contains the tools needed to construct |
295 | `message catalog' files containing translated versions of system | |
a35cb74d | 296 | messages. See ftp://ftp.gnu.org/pub/gnu or better any mirror |
61952351 UD |
297 | site. (We distribute compiled message catalogs, but they may not be |
298 | updated in patches.) | |
f8cac037 | 299 | |
5edb9387 UD |
300 | * Some files are built with special tools. E.g., files ending in .gperf |
301 | need a `gperf' program. The GNU version (now available in a separate | |
302 | package, formerly only as part of libg++) is known to work while some | |
303 | vendor versions do not. | |
f8cac037 | 304 | |
61952351 | 305 | You should not need these tools unless you change the source files. |
1f205a47 | 306 | |
5edb9387 UD |
307 | * Perl 5 is needed if you wish to test an installation of GNU libc |
308 | as the primary C library. | |
4775243a | 309 | |
61952351 UD |
310 | * When compiling for Linux, the header files of the Linux kernel must |
311 | be available to the compiler as <linux/*.h> and <asm/*.h>. | |
f8cac037 | 312 | |
8619129f UD |
313 | * lots of disk space (~170MB for i?86-linux; more for RISC platforms, |
314 | as much as 400MB). | |
af6f3906 | 315 | |
61952351 | 316 | * plenty of time. Compiling just the shared and static libraries for |
b5a9efcd UD |
317 | i?86-linux takes approximately 1h on an AMD-K6@225MHz w/ 96MB of RAM, |
318 | 45mins on a Celeron@400MHz w/ 128MB, and 55mins on a Alpha@533MHz w/ 256MB. | |
319 | Multiply this by 1.5 or 2.0 if you build profiling and/or the highly | |
320 | optimized version as well. For Hurd systems times are much higher. | |
f8cac037 | 321 | |
61952351 UD |
322 | You should avoid compiling in a NFS mounted filesystem. This is |
323 | very slow. | |
0200214b | 324 | |
61952351 | 325 | James Troup <J.J.Troup@comp.brad.ac.uk> reports a compile time of |
4775243a UD |
326 | 45h34m for a full build (shared, static, and profiled) on Atari |
327 | Falcon (Motorola 68030 @ 16 Mhz, 14 Mb memory) and Jan Barte | |
328 | <yann@plato.uni-paderborn.de> reports 22h48m on Atari TT030 | |
329 | (Motorola 68030 @ 32 Mhz, 34 Mb memory) | |
0200214b | 330 | |
61952351 | 331 | If you have some more measurements let me know. |
0200214b | 332 | |
ba1ffaa1 | 333 | |
348ed515 | 334 | 1.8. What version of the Linux kernel headers should be used? |
a35cb74d | 335 | |
f12944ec UD |
336 | {AJ,UD} The headers from the most recent Linux kernel should be used. The |
337 | headers used while compiling the GNU C library and the kernel binary used | |
338 | when using the library do not need to match. The GNU C library runs without | |
339 | problems on kernels that are older than the kernel headers used. The other | |
340 | way round (compiling the GNU C library with old kernel headers and running | |
341 | on a recent kernel) does not necessarily work. For example you can't use | |
b1418d8f | 342 | new kernel features if you used old kernel headers to compile the GNU C |
f12944ec UD |
343 | library. |
344 | ||
b0610668 | 345 | {ZW} Even if you are using a 2.0 kernel on your machine, we recommend you |
b710a6e2 UD |
346 | compile GNU libc with 2.2 kernel headers. That way you won't have to |
347 | recompile libc if you ever upgrade to kernel 2.2. To tell libc which | |
b0610668 | 348 | headers to use, give configure the --with-headers switch |
b710a6e2 | 349 | (e.g. --with-headers=/usr/src/linux-2.2.0/include). |
b0610668 | 350 | |
b710a6e2 | 351 | Note that you must configure the 2.2 kernel if you do this, otherwise libc |
62595351 | 352 | will be unable to find <linux/version.h>. Just change the current directory |
b710a6e2 | 353 | to the root of the 2.2 tree and do `make include/linux/version.h'. |
b0610668 | 354 | |
f12944ec | 355 | |
348ed515 | 356 | 1.9. The compiler hangs while building iconvdata modules. What's |
f12944ec UD |
357 | wrong? |
358 | ||
5edb9387 UD |
359 | {ZW} This is a problem with old versions of GCC. Initialization of large |
360 | static arrays is very slow. The compiler will eventually finish; give it | |
361 | time. | |
a35cb74d | 362 | |
b8f558b7 | 363 | The problem is fixed in egcs 1.1. |
a35cb74d | 364 | |
f12944ec | 365 | |
348ed515 | 366 | 1.10. When I run `nm -u libc.so' on the produced library I still |
61952351 | 367 | find unresolved symbols. Can this be ok? |
f8cac037 | 368 | |
f12944ec | 369 | {UD} Yes, this is ok. There can be several kinds of unresolved symbols: |
f8cac037 | 370 | |
61952351 UD |
371 | * magic symbols automatically generated by the linker. These have names |
372 | like __start_* and __stop_* | |
f8cac037 | 373 | |
78b5ba3e RM |
374 | * symbols starting with _dl_* come from the dynamic linker |
375 | ||
61952351 | 376 | * weak symbols, which need not be resolved at all (fabs for example) |
f8cac037 RM |
377 | |
378 | Generally, you should make sure you find a real program which produces | |
41f27456 | 379 | errors while linking before deciding there is a problem. |
f8cac037 RM |
380 | |
381 | ||
348ed515 | 382 | 1.11. What are these `add-ons'? |
999493cb | 383 | |
f12944ec UD |
384 | {UD} To avoid complications with export rules or external source code some |
385 | optional parts of the libc are distributed as separate packages (e.g., the | |
386 | crypt package, see question 2.5). | |
999493cb | 387 | |
f12944ec UD |
388 | To use these packages as part of GNU libc, just unpack the tarfiles in the |
389 | libc source directory and tell the configuration script about them using the | |
390 | --enable-add-ons option. If you give just --enable-add-ons configure tries | |
391 | to find all the add-on packages in your source tree. This may not work. If | |
392 | it doesn't, or if you want to select only a subset of the add-ons, give a | |
393 | comma-separated list of the add-ons to enable: | |
613a76ff | 394 | |
61952351 | 395 | configure --enable-add-ons=crypt,linuxthreads |
41f27456 | 396 | |
61952351 | 397 | for example. |
0200214b | 398 | |
f12944ec UD |
399 | Add-ons can add features (including entirely new shared libraries), override |
400 | files, provide support for additional architectures, and just about anything | |
401 | else. The existing makefiles do most of the work; only some few stub rules | |
402 | must be written to get everything running. | |
613a76ff | 403 | |
613a76ff | 404 | |
348ed515 | 405 | 1.12. My XXX kernel emulates a floating-point coprocessor for me. |
61952351 | 406 | Should I enable --with-fp? |
613a76ff | 407 | |
f12944ec UD |
408 | {ZW} An emulated FPU is just as good as a real one, as far as the C library |
409 | is concerned. You only need to say --without-fp if your machine has no way | |
410 | to execute floating-point instructions. | |
f8cac037 | 411 | |
61952351 UD |
412 | People who are interested in squeezing the last drop of performance |
413 | out of their machine may wish to avoid the trap overhead, but this is | |
414 | far more trouble than it's worth: you then have to compile | |
415 | *everything* this way, including the compiler's internal libraries | |
416 | (libgcc.a for GNU C), because the calling conventions change. | |
a1470b6f | 417 | |
999493cb | 418 | |
348ed515 | 419 | 1.13. When compiling GNU libc I get lots of errors saying functions |
61952351 | 420 | in glibc are duplicated in libgcc. |
5290baf0 | 421 | |
f12944ec UD |
422 | {EY} This is *exactly* the same problem that I was having. The problem was |
423 | due to the fact that configure didn't correctly detect that the linker flag | |
424 | --no-whole-archive was supported in my linker. In my case it was because I | |
425 | had run ./configure with bogus CFLAGS, and the test failed. | |
78b5ba3e | 426 | |
f12944ec UD |
427 | One thing that is particularly annoying about this problem is that once this |
428 | is misdetected, running configure again won't fix it unless you first delete | |
429 | config.cache. | |
78b5ba3e | 430 | |
f12944ec UD |
431 | {UD} Starting with glibc-2.0.3 there should be a better test to avoid some |
432 | problems of this kind. The setting of CFLAGS is checked at the very | |
433 | beginning and if it is not usable `configure' will bark. | |
78b5ba3e | 434 | |
af6f3906 | 435 | |
348ed515 | 436 | 1.14. Why do I get messages about missing thread functions when I use |
a35cb74d | 437 | librt? I don't even use threads. |
4775243a | 438 | |
a35cb74d UD |
439 | {UD} In this case you probably mixed up your installation. librt uses |
440 | threads internally and has implicit references to the thread library. | |
f12944ec UD |
441 | Normally these references are satisfied automatically but if the thread |
442 | library is not in the expected place you must tell the linker where it is. | |
443 | When using GNU ld it works like this: | |
4775243a UD |
444 | |
445 | gcc -o foo foo.c -Wl,-rpath-link=/some/other/dir -lrt | |
446 | ||
f12944ec UD |
447 | The `/some/other/dir' should contain the thread library. `ld' will use the |
448 | given path to find the implicitly referenced library while not disturbing | |
449 | any other link path. | |
4775243a UD |
450 | |
451 | ||
348ed515 | 452 | 1.15. What's the problem with configure --enable-omitfp? |
78b5ba3e | 453 | |
61952351 | 454 | {AJ} When --enable-omitfp is set the libraries are built without frame |
fdacb17d | 455 | pointers. Some compilers produce buggy code for this model and therefore we |
f12944ec | 456 | don't advise using it at the moment. |
66219c07 | 457 | |
fdacb17d | 458 | If you use --enable-omitfp, you're on your own. If you encounter problems |
f12944ec UD |
459 | with a library that was build this way, we advise you to rebuild the library |
460 | without --enable-omitfp. If the problem vanishes consider tracking the | |
461 | problem down and report it as compiler failure. | |
66219c07 | 462 | |
b1418d8f UD |
463 | Since a library built with --enable-omitfp is undebuggable on most systems, |
464 | debuggable libraries are also built - you can use them by appending "_g" to | |
f12944ec | 465 | the library names. |
66219c07 | 466 | |
f12944ec UD |
467 | The compilation of these extra libraries and the compiler optimizations slow |
468 | down the build process and need more disk space. | |
66219c07 | 469 | |
b0610668 | 470 | |
b1418d8f | 471 | 1.16. I get failures during `make check'. What should I do? |
b0610668 | 472 | |
b1418d8f UD |
473 | {AJ} The testsuite should compile and run cleanly on your system; every |
474 | failure should be looked into. Depending on the failures, you probably | |
475 | should not install the library at all. | |
b0610668 UD |
476 | |
477 | You should consider using the `glibcbug' script to report the failure, | |
478 | providing as much detail as possible. If you run a test directly, please | |
479 | remember to set up the environment correctly. You want to test the compiled | |
480 | library - and not your installed one. The best way is to copy the exact | |
481 | command line which failed and run the test from the subdirectory for this | |
482 | test in the sources. | |
483 | ||
484 | There are some failures which are not directly related to the GNU libc: | |
b1418d8f UD |
485 | - Some compilers produce buggy code. No compiler gets single precision |
486 | complex numbers correct on Alpha. Otherwise, the egcs 1.1 release should be | |
487 | ok; gcc 2.8.1 might cause some failures; gcc 2.7.2.x is so buggy that | |
488 | explicit checks have been used so that you can't build with it. | |
b0610668 UD |
489 | - The kernel might have bugs. For example on Linux/Alpha 2.0.34 the |
490 | floating point handling has quite a number of bugs and therefore most of | |
b710a6e2 | 491 | the test cases in the math subdirectory will fail. Linux 2.2 has |
b1418d8f UD |
492 | fixes for the floating point support on Alpha. The Linux/SPARC kernel has |
493 | also some bugs in the FPU emulation code (as of Linux 2.2.0). | |
b0610668 | 494 | |
a379e56a | 495 | |
348ed515 | 496 | 1.17. What is symbol versioning good for? Do I need it? |
a379e56a UD |
497 | |
498 | {AJ} Symbol versioning solves problems that are related to interface | |
499 | changes. One version of an interface might have been introduced in a | |
500 | previous version of the GNU C library but the interface or the semantics of | |
501 | the function has been changed in the meantime. For binary compatibility | |
502 | with the old library, a newer library needs to still have the old interface | |
b1418d8f | 503 | for old programs. On the other hand, new programs should use the new |
a379e56a | 504 | interface. Symbol versioning is the solution for this problem. The GNU |
b1418d8f UD |
505 | libc version 2.1 uses symbol versioning by default if the installed binutils |
506 | supports it. | |
a379e56a | 507 | |
b1418d8f UD |
508 | We don't advise building without symbol versioning, since you lose binary |
509 | compatibility - forever! The binary compatibility you lose is not only | |
510 | against the previous version of the GNU libc (version 2.0) but also against | |
511 | all future versions. | |
a379e56a | 512 | |
61952351 UD |
513 | \f |
514 | . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . | |
e6c9a67a | 515 | |
61952351 | 516 | 2. Installation and configuration issues |
e6c9a67a | 517 | |
61952351 | 518 | 2.1. Can I replace the libc on my Linux system with GNU libc? |
e6c9a67a | 519 | |
f12944ec UD |
520 | {UD} You cannot replace any existing libc for Linux with GNU libc. It is |
521 | binary incompatible and therefore has a different major version. You can, | |
522 | however, install it alongside your existing libc. | |
e6c9a67a | 523 | |
61952351 UD |
524 | For Linux there are three major libc versions: |
525 | libc-4 a.out libc | |
526 | libc-5 original ELF libc | |
527 | libc-6 GNU libc | |
e6c9a67a | 528 | |
f12944ec UD |
529 | You can have any combination of these three installed. For more information |
530 | consult documentation for shared library handling. The Makefiles of GNU | |
531 | libc will automatically generate the needed symbolic links which the linker | |
532 | will use. | |
e6c9a67a RM |
533 | |
534 | ||
61952351 UD |
535 | 2.2. How do I configure GNU libc so that the essential libraries |
536 | like libc.so go into /lib and the other into /usr/lib? | |
ec42724d | 537 | |
61952351 UD |
538 | {UD,AJ} Like all other GNU packages GNU libc is designed to use a base |
539 | directory and install all files relative to this. The default is | |
f12944ec UD |
540 | /usr/local, because this is safe (it will not damage the system if installed |
541 | there). If you wish to install GNU libc as the primary C library on your | |
542 | system, set the base directory to /usr (i.e. run configure --prefix=/usr | |
543 | <other_options>). Note that this can damage your system; see question 2.3 for | |
544 | details. | |
545 | ||
546 | Some systems like Linux have a filesystem standard which makes a difference | |
547 | between essential libraries and others. Essential libraries are placed in | |
548 | /lib because this directory is required to be located on the same disk | |
549 | partition as /. The /usr subtree might be found on another | |
550 | partition/disk. If you configure for Linux with --prefix=/usr, then this | |
551 | will be done automatically. | |
ec42724d | 552 | |
61952351 | 553 | To install the essential libraries which come with GNU libc in /lib on |
f12944ec UD |
554 | systems other than Linux one must explicitly request it. Autoconf has no |
555 | option for this so you have to use a `configparms' file (see the `INSTALL' | |
556 | file for details). It should contain: | |
ec42724d RM |
557 | |
558 | slibdir=/lib | |
559 | sysconfdir=/etc | |
560 | ||
f12944ec UD |
561 | The first line specifies the directory for the essential libraries, the |
562 | second line the directory for system configuration files. | |
ec42724d | 563 | |
5290baf0 | 564 | |
61952351 | 565 | 2.3. How should I avoid damaging my system when I install GNU libc? |
ec42724d | 566 | |
f12944ec UD |
567 | {ZW} If you wish to be cautious, do not configure with --prefix=/usr. If |
568 | you don't specify a prefix, glibc will be installed in /usr/local, where it | |
569 | will probably not break anything. (If you wish to be certain, set the | |
570 | prefix to something like /usr/local/glibc2 which is not used for anything.) | |
845dcb57 | 571 | |
61952351 | 572 | The dangers when installing glibc in /usr are twofold: |
845dcb57 | 573 | |
61952351 | 574 | * glibc will overwrite the headers in /usr/include. Other C libraries |
27e309c1 UD |
575 | install a different but overlapping set of headers there, so the effect |
576 | will probably be that you can't compile anything. You need to rename | |
577 | /usr/include out of the way before running `make install'. (Do not throw | |
578 | it away; you will then lose the ability to compile programs against your | |
579 | old libc.) | |
845dcb57 | 580 | |
61952351 UD |
581 | * None of your old libraries, static or shared, can be used with a |
582 | different C library major version. For shared libraries this is not a | |
583 | problem, because the filenames are different and the dynamic linker | |
584 | will enforce the restriction. But static libraries have no version | |
585 | information. You have to evacuate all the static libraries in | |
586 | /usr/lib to a safe location. | |
845dcb57 | 587 | |
61952351 UD |
588 | The situation is rather similar to the move from a.out to ELF which |
589 | long-time Linux users will remember. | |
845dcb57 | 590 | |
845dcb57 | 591 | |
61952351 UD |
592 | 2.4. Do I need to use GNU CC to compile programs that will use the |
593 | GNU C Library? | |
845dcb57 | 594 | |
f12944ec UD |
595 | {ZW} In theory, no; the linker does not care, and the headers are supposed |
596 | to check for GNU CC before using its extensions to the C language. | |
845dcb57 | 597 | |
f12944ec UD |
598 | However, there are currently no ports of glibc to systems where another |
599 | compiler is the default, so no one has tested the headers extensively | |
600 | against another compiler. You may therefore encounter difficulties. If you | |
601 | do, please report them as bugs. | |
845dcb57 | 602 | |
61952351 UD |
603 | Also, in several places GNU extensions provide large benefits in code |
604 | quality. For example, the library has hand-optimized, inline assembly | |
f12944ec UD |
605 | versions of some string functions. These can only be used with GCC. See |
606 | question 3.8 for details. | |
845dcb57 | 607 | |
845dcb57 | 608 | |
61952351 UD |
609 | 2.5. When linking with the new libc I get unresolved symbols |
610 | `crypt' and `setkey'. Why aren't these functions in the | |
611 | libc anymore? | |
845dcb57 | 612 | |
f12944ec UD |
613 | {UD} The US places restrictions on exporting cryptographic programs and |
614 | source code. Until this law gets abolished we cannot ship the cryptographic | |
615 | functions together with glibc. | |
845dcb57 | 616 | |
348ed515 | 617 | The functions are available, as an add-on (see question 1.11). People in the US |
f12944ec | 618 | may get it from the same place they got GNU libc from. People outside the |
9f2a9248 AS |
619 | US should get the code from ftp.funet.fi [128.214.248.6] in the directory |
620 | pub/gnu/funet, or another archive site outside the USA. The README explains | |
621 | how to install the sources. | |
c4029823 | 622 | |
f12944ec UD |
623 | If you already have the crypt code on your system the reason for the failure |
624 | is probably that you did not link with -lcrypt. The crypto functions are in | |
625 | a separate library to make it possible to export GNU libc binaries from the | |
626 | US. | |
c4029823 | 627 | |
c4029823 | 628 | |
61952351 UD |
629 | 2.6. When I use GNU libc on my Linux system by linking against |
630 | the libc.so which comes with glibc all I get is a core dump. | |
c4029823 | 631 | |
f12944ec UD |
632 | {UD} On Linux, gcc sets the dynamic linker to /lib/ld-linux.so.1 unless the |
633 | user specifies a -dynamic-linker argument. This is the name of the libc5 | |
634 | dynamic linker, which does not work with glibc. | |
61952351 | 635 | |
a379e56a UD |
636 | For casual use of GNU libc you can just specify to the linker |
637 | --dynamic-linker=/lib/ld-linux.so.2 | |
61952351 | 638 | |
f12944ec | 639 | which is the glibc dynamic linker, on Linux systems. On other systems the |
a379e56a UD |
640 | name is /lib/ld.so.1. When linking via gcc, you've got to add |
641 | -Wl,--dynamic-linker=/lib/ld-linux.so.2 | |
642 | ||
643 | to the gcc command line. | |
c4029823 | 644 | |
f12944ec UD |
645 | To change your environment to use GNU libc for compiling you need to change |
646 | the `specs' file of your gcc. This file is normally found at | |
c4029823 UD |
647 | |
648 | /usr/lib/gcc-lib/<arch>/<version>/specs | |
649 | ||
650 | In this file you have to change a few things: | |
651 | ||
61952351 | 652 | - change `ld-linux.so.1' to `ld-linux.so.2' |
c4029823 UD |
653 | |
654 | - remove all expression `%{...:-lgmon}'; there is no libgmon in glibc | |
655 | ||
f4017d20 UD |
656 | - fix a minor bug by changing %{pipe:-} to %| |
657 | ||
f12944ec UD |
658 | Here is what the gcc-2.7.2 specs file should look like when GNU libc is |
659 | installed at /usr: | |
c4029823 UD |
660 | |
661 | ----------------------------------------------------------------------- | |
662 | *asm: | |
663 | %{V} %{v:%{!V:-V}} %{Qy:} %{!Qn:-Qy} %{n} %{T} %{Ym,*} %{Yd,*} %{Wa,*:%*} | |
664 | ||
665 | *asm_final: | |
f4017d20 | 666 | %| |
c4029823 UD |
667 | |
668 | *cpp: | |
68dbb3a6 | 669 | %{fPIC:-D__PIC__ -D__pic__} %{fpic:-D__PIC__ -D__pic__} %{!m386:-D__i486__} %{posix:-D_POSIX_SOURCE} %{pthread:-D_REENTRANT} |
c4029823 UD |
670 | |
671 | *cc1: | |
68dbb3a6 | 672 | %{profile:-p} |
c4029823 UD |
673 | |
674 | *cc1plus: | |
675 | ||
676 | ||
677 | *endfile: | |
68dbb3a6 | 678 | %{!shared:crtend.o%s} %{shared:crtendS.o%s} crtn.o%s |
c4029823 UD |
679 | |
680 | *link: | |
68dbb3a6 | 681 | -m elf_i386 %{shared:-shared} %{!shared: %{!ibcs: %{!static: %{rdynamic:-export-dynamic} %{!dynamic-linker:-dynamic-linker /lib/ld-linux.so.2}} %{static:-static}}} |
c4029823 UD |
682 | |
683 | *lib: | |
68dbb3a6 | 684 | %{!shared: %{pthread:-lpthread} %{profile:-lc_p} %{!profile: -lc}} |
c4029823 UD |
685 | |
686 | *libgcc: | |
68dbb3a6 | 687 | -lgcc |
c4029823 UD |
688 | |
689 | *startfile: | |
61952351 | 690 | %{!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} |
c4029823 UD |
691 | |
692 | *switches_need_spaces: | |
693 | ||
694 | ||
695 | *signed_char: | |
696 | %{funsigned-char:-D__CHAR_UNSIGNED__} | |
697 | ||
698 | *predefines: | |
699 | -D__ELF__ -Dunix -Di386 -Dlinux -Asystem(unix) -Asystem(posix) -Acpu(i386) -Amachine(i386) | |
700 | ||
701 | *cross_compile: | |
702 | 0 | |
703 | ||
704 | *multilib: | |
705 | . ; | |
706 | ||
707 | ----------------------------------------------------------------------- | |
708 | ||
f12944ec UD |
709 | Things get a bit more complicated if you have GNU libc installed in some |
710 | other place than /usr, i.e., if you do not want to use it instead of the old | |
711 | libc. In this case the needed startup files and libraries are not found in | |
712 | the regular places. So the specs file must tell the compiler and linker | |
713 | exactly what to use. | |
0d204b0a | 714 | |
f41c8091 | 715 | Version 2.7.2.3 does and future versions of GCC will automatically |
0d8733c4 | 716 | provide the correct specs. |
c4029823 UD |
717 | |
718 | ||
61952351 UD |
719 | 2.7. Looking through the shared libc file I haven't found the |
720 | functions `stat', `lstat', `fstat', and `mknod' and while | |
721 | linking on my Linux system I get error messages. How is | |
722 | this supposed to work? | |
c4029823 | 723 | |
f12944ec UD |
724 | {RM} Believe it or not, stat and lstat (and fstat, and mknod) are supposed |
725 | to be undefined references in libc.so.6! Your problem is probably a missing | |
726 | or incorrect /usr/lib/libc.so file; note that this is a small text file now, | |
727 | not a symlink to libc.so.6. It should look something like this: | |
c4029823 | 728 | |
ff44f2a5 | 729 | GROUP ( libc.so.6 libc_nonshared.a ) |
1f205a47 | 730 | |
c4029823 | 731 | |
5edb9387 UD |
732 | 2.8. When I run an executable on one system which I compiled on |
733 | another, I get dynamic linker errors. Both systems have the same | |
734 | version of glibc installed. What's wrong? | |
735 | ||
736 | {ZW} Glibc on one of these systems was compiled with gcc 2.7 or 2.8, the | |
737 | other with egcs (any version). Egcs has functions in its internal | |
738 | `libgcc.a' to support exception handling with C++. They are linked into | |
739 | any program or dynamic library compiled with egcs, whether it needs them or | |
740 | not. Dynamic libraries then turn around and export those functions again | |
741 | unless special steps are taken to prevent them. | |
742 | ||
743 | When you link your program, it resolves its references to the exception | |
744 | functions to the ones exported accidentally by libc.so. That works fine as | |
745 | long as libc has those functions. On the other system, libc doesn't have | |
746 | those functions because it was compiled by gcc 2.8, and you get undefined | |
747 | symbol errors. The symbols in question are named things like | |
748 | `__register_frame_info'. | |
749 | ||
750 | For glibc 2.0, the workaround is to not compile libc with egcs. We've also | |
751 | incorporated a patch which should prevent the EH functions sneaking into | |
752 | libc. It doesn't matter what compiler you use to compile your program. | |
753 | ||
754 | For glibc 2.1, we've chosen to do it the other way around: libc.so | |
755 | explicitly provides the EH functions. This is to prevent other shared | |
95f7cecb UD |
756 | libraries from doing it. |
757 | ||
758 | {UD} Starting with glibc 2.1.1 you can compile glibc with gcc 2.8.1 or | |
759 | newer since we have explicitly add references to the functions causing the | |
760 | problem. But you nevertheless should use EGCS for other reasons | |
761 | (see question 1.2). | |
5edb9387 UD |
762 | |
763 | ||
764 | 2.9. How can I compile gcc 2.7.2.1 from the gcc source code using | |
61952351 | 765 | glibc 2.x? |
ba1ffaa1 | 766 | |
f12944ec | 767 | {AJ} There's only correct support for glibc 2.0.x in gcc 2.7.2.3 or later. |
95f7cecb | 768 | But you should get at least gcc 2.8.1 or egcs 1.1 (or later versions) |
f12944ec | 769 | instead. |
ba1ffaa1 UD |
770 | |
771 | ||
5edb9387 | 772 | 2.10. The `gencat' utility cannot process the catalog sources which |
61952351 | 773 | were used on my Linux libc5 based system. Why? |
47707456 | 774 | |
f12944ec UD |
775 | {UD} The `gencat' utility provided with glibc complies to the XPG standard. |
776 | The older Linux version did not obey the standard, so they are not | |
777 | compatible. | |
47707456 | 778 | |
61952351 | 779 | To ease the transition from the Linux version some of the non-standard |
f12944ec UD |
780 | features are also present in the `gencat' program of GNU libc. This mainly |
781 | includes the use of symbols for the message number and the automatic | |
61952351 UD |
782 | generation of header files which contain the needed #defines to map the |
783 | symbols to integers. | |
47707456 | 784 | |
f12944ec UD |
785 | Here is a simple SED script to convert at least some Linux specific catalog |
786 | files to the XPG4 form: | |
68dbb3a6 | 787 | |
61952351 UD |
788 | ----------------------------------------------------------------------- |
789 | # Change catalog source in Linux specific format to standard XPG format. | |
790 | # Ulrich Drepper <drepper@cygnus.com>, 1996. | |
791 | # | |
792 | /^\$ #/ { | |
793 | h | |
794 | s/\$ #\([^ ]*\).*/\1/ | |
795 | x | |
796 | s/\$ #[^ ]* *\(.*\)/\$ \1/ | |
797 | } | |
68dbb3a6 | 798 | |
61952351 UD |
799 | /^# / { |
800 | s/^# \(.*\)/\1/ | |
801 | G | |
802 | s/\(.*\)\n\(.*\)/\2 \1/ | |
803 | } | |
804 | ----------------------------------------------------------------------- | |
19361cb7 | 805 | |
19361cb7 | 806 | |
5edb9387 | 807 | 2.11. Programs using libc have their messages translated, but other |
a35cb74d UD |
808 | behavior is not localized (e.g. collating order); why? |
809 | ||
810 | {ZW} Translated messages are automatically installed, but the locale | |
f12944ec UD |
811 | database that controls other behaviors is not. You need to run localedef to |
812 | install this database, after you have run `make install'. For example, to | |
813 | set up the French Canadian locale, simply issue the command | |
a35cb74d UD |
814 | |
815 | localedef -i fr_CA -f ISO-8859-1 fr_CA | |
816 | ||
817 | Please see localedata/README in the source tree for further details. | |
818 | ||
819 | ||
5edb9387 | 820 | 2.12. I have set up /etc/nis.conf, and the Linux libc 5 with NYS |
61952351 | 821 | works great. But the glibc NIS+ doesn't seem to work. |
19361cb7 | 822 | |
f12944ec UD |
823 | {TK} The glibc NIS+ implementation uses a /var/nis/NIS_COLD_START file for |
824 | storing information about the NIS+ server and their public keys, because the | |
825 | nis.conf file does not contain all the necessary information. You have to | |
826 | copy a NIS_COLD_START file from a Solaris client (the NIS_COLD_START file is | |
827 | byte order independent) or generate it with nisinit from the nis-tools | |
828 | package; available at | |
829 | ||
612fdf25 | 830 | http://www.suse.de/~kukuk/linux/nisplus.html |
19361cb7 | 831 | |
68dbb3a6 | 832 | |
5edb9387 | 833 | 2.13. I have killed ypbind to stop using NIS, but glibc |
3dcf8ea6 | 834 | continues using NIS. |
4d06461a | 835 | |
f12944ec UD |
836 | {TK} For faster NIS lookups, glibc uses the /var/yp/binding/ files from |
837 | ypbind. ypbind 3.3 and older versions don't always remove these files, so | |
838 | glibc will continue to use them. Other BSD versions seem to work correctly. | |
839 | Until ypbind 3.4 is released, you can find a patch at | |
840 | ||
05f732b3 | 841 | ftp://ftp.kernel.org/pub/linux/utils/net/NIS/ypbind-3.3-glibc4.diff.gz |
a35cb74d | 842 | |
4d06461a | 843 | |
5edb9387 | 844 | 2.14. Under Linux/Alpha, I always get "do_ypcall: clnt_call: |
3dcf8ea6 | 845 | RPC: Unable to receive; errno = Connection refused" when using NIS. |
4d06461a | 846 | |
f12944ec UD |
847 | {TK} You need a ypbind version which is 64bit clean. Some versions are not |
848 | 64bit clean. A 64bit clean implementation is ypbind-mt. For ypbind 3.3, | |
849 | you need the patch from ftp.kernel.org (See the previous question). I don't | |
850 | know about other versions. | |
a35cb74d UD |
851 | |
852 | ||
5edb9387 | 853 | 2.15. After installing glibc name resolving doesn't work properly. |
68dbb3a6 | 854 | |
f12944ec UD |
855 | {AJ} You probably should read the manual section describing nsswitch.conf |
856 | (just type `info libc "NSS Configuration File"'). The NSS configuration | |
857 | file is usually the culprit. | |
22d57dd3 | 858 | |
22d57dd3 | 859 | |
5edb9387 | 860 | 2.16. How do I create the databases for NSS? |
3dcf8ea6 UD |
861 | |
862 | {AJ} If you have an entry "db" in /etc/nsswitch.conf you should also create | |
863 | the database files. The glibc sources contain a Makefile which does the | |
a379e56a | 864 | necessary conversion and calls to create those files. The file is |
3dcf8ea6 UD |
865 | `db-Makefile' in the subdirectory `nss' and you can call it with `make -f |
866 | db-Makefile'. Please note that not all services are capable of using a | |
867 | database. Currently passwd, group, ethers, protocol, rpc, services shadow | |
868 | and netgroup are implemented. | |
869 | ||
870 | ||
5edb9387 | 871 | 2.17. I have /usr/include/net and /usr/include/scsi as symlinks |
61952351 | 872 | into my Linux source tree. Is that wrong? |
22d57dd3 | 873 | |
f12944ec UD |
874 | {PB} This was necessary for libc5, but is not correct when using glibc. |
875 | Including the kernel header files directly in user programs usually does not | |
876 | work (see question 3.5). glibc provides its own <net/*> and <scsi/*> header | |
877 | files to replace them, and you may have to remove any symlink that you have | |
878 | in place before you install glibc. However, /usr/include/asm and | |
879 | /usr/include/linux should remain as they were. | |
22d57dd3 | 880 | |
22d57dd3 | 881 | |
5edb9387 | 882 | 2.18. Programs like `logname', `top', `uptime' `users', `w' and |
61952351 UD |
883 | `who', show incorrect information about the (number of) |
884 | users on my system. Why? | |
22d57dd3 | 885 | |
61952351 | 886 | {MK} See question 3.2. |
22d57dd3 | 887 | |
22d57dd3 | 888 | |
5edb9387 | 889 | 2.19. After upgrading to glibc 2.1 with symbol versioning I get |
61952351 | 890 | errors about undefined symbols. What went wrong? |
26dee9c4 | 891 | |
f12944ec UD |
892 | {AJ} The problem is caused either by wrong program code or tools. In the |
893 | versioned libc a lot of symbols are now local that were global symbols in | |
894 | previous versions. It seems that programs linked against older versions | |
895 | often accidentally used libc global variables -- something that should not | |
896 | happen. | |
26dee9c4 | 897 | |
f12944ec UD |
898 | The only way to fix this is to recompile your program. Sorry, that's the |
899 | price you might have to pay once for quite a number of advantages with | |
900 | symbol versioning. | |
26dee9c4 | 901 | |
26dee9c4 | 902 | |
5edb9387 | 903 | 2.20. When I start the program XXX after upgrading the library |
61952351 UD |
904 | I get |
905 | XXX: Symbol `_sys_errlist' has different size in shared | |
906 | object, consider re-linking | |
907 | Why? What should I do? | |
26dee9c4 | 908 | |
f12944ec UD |
909 | {UD} As the message says, relink the binary. The problem is that a few |
910 | symbols from the library can change in size and there is no way to avoid | |
911 | this. _sys_errlist is a good example. Occasionally there are new error | |
912 | numbers added to the kernel and this must be reflected at user level, | |
913 | breaking programs that refer to them directly. | |
a2b08ee5 | 914 | |
f12944ec UD |
915 | Such symbols should normally not be used at all. There are mechanisms to |
916 | avoid using them. In the case of _sys_errlist, there is the strerror() | |
917 | function which should _always_ be used instead. So the correct fix is to | |
918 | rewrite that part of the application. | |
a2b08ee5 | 919 | |
f12944ec UD |
920 | In some situations (especially when testing a new library release) it might |
921 | be possible that a symbol changed size when that should not have happened. | |
922 | So in case of doubt report such a warning message as a problem. | |
a2b08ee5 | 923 | |
a35cb74d | 924 | |
5edb9387 UD |
925 | 2.21. What do I need for C++ development? |
926 | ||
927 | {HJ,AJ} You need either egcs 1.1 which comes directly with libstdc++ or | |
928 | gcc-2.8.1 together with libstdc++ 2.8.1.1. egcs 1.1 has the better C++ | |
929 | support and works directly with glibc 2.1. If you use gcc-2.8.1 with | |
930 | libstdc++ 2.8.1.1, you need to modify libstdc++ a bit. A patch is available | |
931 | as: | |
932 | ftp://alpha.gnu.org/gnu/libstdc++-2.8.1.1-glibc2.1-diff.gz | |
a35cb74d | 933 | |
5edb9387 UD |
934 | Please note that libg++ 2.7.2 (and the Linux Versions 2.7.2.x) doesn't work |
935 | very well with the GNU C library due to vtable thunks. If you're upgrading | |
936 | from glibc 2.0.x to 2.1 you have to recompile libstdc++ since the library | |
937 | compiled for 2.0 is not compatible due to the new Large File Support (LFS) | |
938 | in version 2.1. | |
fb98e2bf UD |
939 | |
940 | {UD} But since in the case of a shared libstdc++ the version numbers should | |
941 | be different existing programs will continue to work. | |
a35cb74d | 942 | |
ff44f2a5 | 943 | |
5edb9387 | 944 | 2.22. Even statically linked programs need some shared libraries |
ff44f2a5 UD |
945 | which is not acceptable for me. What can I do? |
946 | ||
f12944ec UD |
947 | {AJ} NSS (for details just type `info libc "Name Service Switch"') won't |
948 | work properly without shared libraries. NSS allows using different services | |
949 | (e.g. NIS, files, db, hesiod) by just changing one configuration file | |
950 | (/etc/nsswitch.conf) without relinking any programs. The only disadvantage | |
951 | is that now static libraries need to access shared libraries. This is | |
952 | handled transparently by the GNU C library. | |
ff44f2a5 | 953 | |
f12944ec UD |
954 | A solution is to configure glibc with --enable-static-nss. In this case you |
955 | can create a static binary that will use only the services dns and files | |
956 | (change /etc/nsswitch.conf for this). You need to link explicitly against | |
957 | all these services. For example: | |
ff44f2a5 UD |
958 | |
959 | gcc -static test-netdb.c -o test-netdb.c \ | |
960 | -lc -lnss_files -lnss_dns -lresolv | |
961 | ||
962 | The problem with this approach is that you've got to link every static | |
963 | program that uses NSS routines with all those libraries. | |
964 | ||
965 | {UD} In fact, one cannot say anymore that a libc compiled with this | |
966 | option is using NSS. There is no switch anymore. Therefore it is | |
967 | *highly* recommended *not* to use --enable-static-nss since this makes | |
968 | the behaviour of the programs on the system inconsistent. | |
969 | ||
fdacb17d | 970 | |
5edb9387 | 971 | 2.23. I just upgraded my Linux system to glibc and now I get |
fdacb17d UD |
972 | errors whenever I try to link any program. |
973 | ||
974 | {ZW} This happens when you have installed glibc as the primary C library but | |
975 | have stray symbolic links pointing at your old C library. If the first | |
976 | `libc.so' the linker finds is libc 5, it will use that. Your program | |
977 | expects to be linked with glibc, so the link fails. | |
978 | ||
979 | The most common case is that glibc put its `libc.so' in /usr/lib, but there | |
980 | was a `libc.so' from libc 5 in /lib, which gets searched first. To fix the | |
981 | problem, just delete /lib/libc.so. You may also need to delete other | |
982 | symbolic links in /lib, such as /lib/libm.so if it points to libm.so.5. | |
983 | ||
984 | {AJ} The perl script test-installation.pl which is run as last step during | |
985 | an installation of glibc that is configured with --prefix=/usr should help | |
986 | detect these situations. If the script reports problems, something is | |
987 | really screwed up. | |
988 | ||
48244d09 | 989 | |
5edb9387 | 990 | 2.24. When I use nscd the machine freezes. |
48244d09 | 991 | |
5edb9387 UD |
992 | {UD} You cannot use nscd with Linux 2.0.*. There is functionality missing |
993 | in the kernel and work-arounds are not suitable. Besides, some parts of the | |
994 | kernel are too buggy when it comes to using threads. | |
48244d09 | 995 | |
b710a6e2 | 996 | If you need nscd, you have to use at least a 2.1 kernel. |
48244d09 UD |
997 | |
998 | Note that I have at this point no information about any other platform. | |
999 | ||
0155a773 UD |
1000 | |
1001 | 2.25. I need lots of open files. What do I have to do? | |
1002 | ||
1003 | {AJ} This is at first a kernel issue. The kernel defines limits with | |
1004 | OPEN_MAX the number of simultaneous open files and with FD_SETSIZE the | |
1005 | number of used file descriptors. You need to change these values in your | |
1006 | kernel and recompile the kernel so that the kernel allows to use more open | |
1007 | files. You don't necessarily need to recompile the GNU C library since the | |
1008 | only place where OPEN_MAX and FD_SETSIZE is really needed in the library | |
1009 | itself is the size of fd_set which is used by select. | |
1010 | ||
1011 | The GNU C library is now (nearly) select free. This means it internally has | |
1012 | no limits imposed by the `fd_set' type. Instead almost all places where the | |
1013 | functionality is needed the `poll' function is used. | |
1014 | ||
1015 | If you increase the number of file descriptors in the kernel you don't need | |
1016 | to recompile the C library. The remaining select calls are in the RPC code. | |
1017 | If your RPC daemons don't need more than FD_SETSIZE file descriptors, you | |
1018 | don't need to change anything at all. | |
1019 | ||
1020 | {UD} You can always get the maximum number of file descriptors a process is | |
1021 | allowed to have open at any time using | |
1022 | ||
1023 | number = sysconf (_SC_OPEN_MAX); | |
1024 | ||
1025 | This will work even if the kernel limits change. | |
1026 | ||
7db169c9 UD |
1027 | |
1028 | 2.26. How do I get the same behavior on parsing /etc/passwd and | |
1029 | /etc/group as I have with libc5 ? | |
1030 | ||
1031 | {TK} The name switch setup in /etc/nsswitch.conf selected by most Linux | |
1032 | distributions does not support +/- and netgroup entries in the files like | |
1033 | /etc/passwd. Though this is the preferred setup some people might have | |
1034 | setups coming over from the libc5 days where it was the default to recognize | |
1035 | lines like this. To get back to the old behaviour one simply has to change | |
1036 | the rules for passwd, group, and shadow in the nsswitch.conf file as | |
1037 | follows: | |
1038 | ||
1039 | passwd: compat | |
1040 | group: compat | |
1041 | shadow: compat | |
1042 | ||
1043 | passwd_compat: nis | |
1044 | group_compat: nis | |
1045 | shadow_compat: nis | |
1046 | ||
b710a6e2 UD |
1047 | |
1048 | 2.27. What needs to be recompiled when upgrading from glibc 2.0 to glibc | |
1049 | 2.1? | |
1050 | ||
1051 | {AJ,CG} If you just upgrade the glibc from 2.0.x (x <= 7) to 2.1, binaries | |
1052 | that have been linked against glibc 2.0 will continue to work. | |
1053 | ||
1054 | If you compile your own binaries against glibc 2.1, you also need to | |
1055 | recompile some other libraries. The problem is that libio had to be | |
1056 | changed and therefore libraries that are based or depend on the libio | |
1057 | of glibc, e.g. ncurses or slang, need to be recompiled. If you | |
1058 | experience strange segmentation faults in your programs linked against | |
1059 | glibc 2.1, you might need to recompile your libraries. | |
1060 | ||
1061 | Another problem is that older binaries that were linked statically against | |
1062 | glibc 2.0 will reference the older nss modules (libnss_files.so.1 instead of | |
1063 | libnss_files.so.2), so don't remove them. Also, the old glibc-2.0 compiled | |
1064 | static libraries (libfoo.a) which happen to depend on the older libio | |
1065 | behavior will be broken by the glibc 2.1 upgrade. We plan to produce a | |
1066 | compatibility library that people will be able to link in if they want | |
1067 | to compile a static library generated against glibc 2.0 into a program | |
1068 | on a glibc 2.1 system. You just add -lcompat and you should be fine. | |
1069 | ||
1070 | The glibc-compat add-on will provide the libcompat.a library, the older | |
1071 | nss modules, and a few other files. Together, they should make it | |
1072 | possible to do development with old static libraries on a glibc 2.1 | |
8d8c6efa UD |
1073 | system. This add-on is still in development. You can get it from |
1074 | ftp://alpha.gnu.org/gnu/glibc-compat-2.1.tar.gz | |
b710a6e2 UD |
1075 | but please keep in mind that it is experimental. |
1076 | ||
b7398be5 UD |
1077 | |
1078 | 2.28. Why is extracting files via tar so slow? | |
1079 | ||
1080 | {AJ} Extracting of tar archives might be quite slow since tar has to look up | |
1081 | userid and groupids and doesn't cache negative results. If you have nis or | |
1082 | nisplus in your /etc/nsswitch.conf for the passwd and/or group database, | |
1083 | each file extractions needs a network connection. There are two possible | |
1084 | solutions: | |
1085 | ||
1086 | - do you really need NIS/NIS+ (some Linux distributions add by default | |
1087 | nis/nisplus even if it's not needed)? If not, just remove the entries. | |
1088 | ||
1089 | - if you need NIS/NIS+, use the Name Service Cache Daemon nscd that comes | |
1090 | with glibc 2.1. | |
1091 | ||
2ee511d9 UD |
1092 | |
1093 | 2.29. Compiling programs I get parse errors in libio.h (e.g. "parse error | |
1094 | before `_IO_seekoff'"). How should I fix this? | |
1095 | ||
1096 | {AJ} You might get the following errors when upgrading to glibc 2.1: | |
1097 | ||
1098 | In file included from /usr/include/stdio.h:57, | |
1099 | from ... | |
1100 | /usr/include/libio.h:335: parse error before `_IO_seekoff' | |
1101 | /usr/include/libio.h:335: parse error before `_G_off64_t' | |
1102 | /usr/include/libio.h:336: parse error before `_IO_seekpos' | |
1103 | /usr/include/libio.h:336: parse error before `_G_fpos64_t' | |
1104 | ||
1105 | The problem is a wrong _G_config.h file in your include path. The | |
1106 | _G_config.h file that comes with glibc 2.1 should be used and not one from | |
1107 | libc5 or from a compiler directory. To check which _G_config.h file the | |
1108 | compiler uses, compile your program with `gcc -E ...|grep G_config.h' and | |
1109 | remove that file. Your compiler should pick up the file that has been | |
1110 | installed by glibc 2.1 in your include directory. | |
1111 | ||
4f7ea427 UD |
1112 | |
1113 | 2.30. After upgrading to glibc 2.1, libraries that were compiled against | |
1114 | glibc 2.0.x don't work anymore. | |
1115 | ||
1116 | {AJ} See question 2.27. | |
1117 | ||
61952351 UD |
1118 | \f |
1119 | . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . | |
a5a0310d | 1120 | |
61952351 | 1121 | 3. Source and binary incompatibilities, and what to do about them |
a5a0310d | 1122 | |
61952351 UD |
1123 | 3.1. I expect GNU libc to be 100% source code compatible with |
1124 | the old Linux based GNU libc. Why isn't it like this? | |
a5a0310d | 1125 | |
f12944ec UD |
1126 | {DMT,UD} Not every extension in Linux libc's history was well thought-out. |
1127 | In fact it had a lot of problems with standards compliance and with | |
1128 | cleanliness. With the introduction of a new version number these errors can | |
1129 | now be corrected. Here is a list of the known source code | |
61952351 | 1130 | incompatibilities: |
af6f3906 | 1131 | |
61952351 UD |
1132 | * _GNU_SOURCE: glibc does not make the GNU extensions available |
1133 | automatically. If a program depends on GNU extensions or some | |
1134 | other non-standard functionality, it is necessary to compile it | |
1135 | with the C compiler option -D_GNU_SOURCE, or better, to put | |
1136 | `#define _GNU_SOURCE' at the beginning of your source files, before | |
1137 | any C library header files are included. This difference normally | |
1138 | manifests itself in the form of missing prototypes and/or data type | |
1139 | definitions. Thus, if you get such errors, the first thing you | |
1140 | should do is try defining _GNU_SOURCE and see if that makes the | |
1141 | problem go away. | |
af6f3906 | 1142 | |
61952351 UD |
1143 | For more information consult the file `NOTES' in the GNU C library |
1144 | sources. | |
af6f3906 | 1145 | |
61952351 UD |
1146 | * reboot(): GNU libc sanitizes the interface of reboot() to be more |
1147 | compatible with the interface used on other OSes. reboot() as | |
1148 | implemented in glibc takes just one argument. This argument | |
1149 | corresponds to the third argument of the Linux reboot system call. | |
1150 | That is, a call of the form reboot(a, b, c) needs to be changed into | |
1151 | reboot(c). Beside this the header <sys/reboot.h> defines the needed | |
1152 | constants for the argument. These RB_* constants should be used | |
1153 | instead of the cryptic magic numbers. | |
1154 | ||
1155 | * swapon(): the interface of this function didn't change, but the | |
1156 | prototype is in a separate header file <sys/swap.h>. This header | |
1157 | file also provides the SWAP_* constants defined by <linux/swap.h>; | |
1158 | you should use them for the second argument to swapon(). | |
1159 | ||
1160 | * errno: If a program uses the variable "errno", then it _must_ | |
1161 | include <errno.h>. The old libc often (erroneously) declared this | |
1162 | variable implicitly as a side-effect of including other libc header | |
1163 | files. glibc is careful to avoid such namespace pollution, which, | |
1164 | in turn, means that you really need to include the header files that | |
1165 | you depend on. This difference normally manifests itself in the | |
1166 | form of the compiler complaining about references to an undeclared | |
1167 | symbol "errno". | |
dd7d45e8 | 1168 | |
61952351 UD |
1169 | * Linux-specific syscalls: All Linux system calls now have appropriate |
1170 | library wrappers and corresponding declarations in various header files. | |
1171 | This is because the syscall() macro that was traditionally used to | |
1172 | work around missing syscall wrappers are inherently non-portable and | |
1173 | error-prone. The following table lists all the new syscall stubs, | |
1174 | the header-file declaring their interface and the system call name. | |
dd7d45e8 | 1175 | |
61952351 UD |
1176 | syscall name: wrapper name: declaring header file: |
1177 | ------------- ------------- ---------------------- | |
1178 | bdflush bdflush <sys/kdaemon.h> | |
1179 | syslog ksyslog_ctl <sys/klog.h> | |
dd7d45e8 | 1180 | |
61952351 UD |
1181 | * lpd: Older versions of lpd depend on a routine called _validuser(). |
1182 | The library does not provide this function, but instead provides | |
1183 | __ivaliduser() which has a slightly different interface. Simply | |
1184 | upgrading to a newer lpd should fix this problem (e.g., the 4.4BSD | |
1185 | lpd is known to be working). | |
dd7d45e8 | 1186 | |
61952351 UD |
1187 | * resolver functions/BIND: like on many other systems the functions of |
1188 | the resolver library are not included in libc itself. There is a | |
1189 | separate library libresolv. If you get undefined symbol errors for | |
1190 | symbols starting with `res_*' simply add -lresolv to your linker | |
1191 | command line. | |
dd7d45e8 | 1192 | |
61952351 UD |
1193 | * the `signal' function's behavior corresponds to the BSD semantic and |
1194 | not the SysV semantic as it was in libc-5. The interface on all GNU | |
1195 | systems shall be the same and BSD is the semantic of choice. To use | |
1196 | the SysV behavior simply use `sysv_signal', or define _XOPEN_SOURCE. | |
1197 | See question 3.7 for details. | |
1cab5444 | 1198 | |
1cab5444 | 1199 | |
61952351 UD |
1200 | 3.2. Why does getlogin() always return NULL on my Linux box? |
1201 | ||
f12944ec UD |
1202 | {UD} The GNU C library has a format for the UTMP and WTMP file which differs |
1203 | from what your system currently has. It was extended to fulfill the needs | |
1204 | of the next years when IPv6 is introduced. The record size is different and | |
1205 | some fields have different positions. The files written by functions from | |
1206 | the one library cannot be read by functions from the other library. Sorry, | |
1207 | but this is what a major release is for. It's better to have a cut now than | |
1208 | having no means to support the new techniques later. | |
1cab5444 | 1209 | |
f12944ec UD |
1210 | {MK} There is however a (partial) solution for this problem. Please take a |
1211 | look at the file `login/README.utmpd'. | |
1cab5444 | 1212 | |
6973fc01 | 1213 | |
61952351 UD |
1214 | 3.3. Where are the DST_* constants found in <sys/time.h> on many |
1215 | systems? | |
6973fc01 | 1216 | |
f12944ec UD |
1217 | {UD} These constants come from the old BSD days and are not used anymore |
1218 | (libc5 does not actually implement the handling although the constants are | |
1219 | defined). | |
6973fc01 | 1220 | |
f12944ec | 1221 | Instead GNU libc contains zone database support and compatibility code for |
8b4a4715 | 1222 | POSIX TZ environment variable handling. For former is very much preferred |
7d1de115 | 1223 | (see question 4.3). |
6973fc01 UD |
1224 | |
1225 | ||
61952351 UD |
1226 | 3.4. The prototypes for `connect', `accept', `getsockopt', |
1227 | `setsockopt', `getsockname', `getpeername', `send', | |
1228 | `sendto', and `recvfrom' are different in GNU libc from | |
1229 | any other system I saw. This is a bug, isn't it? | |
f4017d20 | 1230 | |
f12944ec UD |
1231 | {UD} No, this is no bug. This version of GNU libc already follows the new |
1232 | Single Unix specifications (and I think the POSIX.1g draft which adopted the | |
1233 | solution). The type for a parameter describing a size is now `socklen_t', a | |
1234 | new type. | |
f4017d20 | 1235 | |
f4017d20 | 1236 | |
61952351 UD |
1237 | 3.5. On Linux I've got problems with the declarations in Linux |
1238 | kernel headers. | |
f4017d20 | 1239 | |
f12944ec UD |
1240 | {UD,AJ} On Linux, the use of kernel headers is reduced to the minimum. This |
1241 | gives Linus the ability to change the headers more freely. Also, user | |
8f1c9b09 | 1242 | programs are now insulated from changes in the size of kernel data |
f12944ec | 1243 | structures. |
f4017d20 | 1244 | |
f12944ec UD |
1245 | For example, the sigset_t type is 32 or 64 bits wide in the kernel. In |
1246 | glibc it is 1024 bits wide. This guarantees that when the kernel gets a | |
1247 | bigger sigset_t (for POSIX.1e realtime support, say) user programs will not | |
1248 | have to be recompiled. Consult the header files for more information about | |
1249 | the changes. | |
61952351 | 1250 | |
f12944ec UD |
1251 | Therefore you shouldn't include Linux kernel header files directly if glibc |
1252 | has defined a replacement. Otherwise you might get undefined results because | |
1253 | of type conflicts. | |
f4017d20 | 1254 | |
f4017d20 | 1255 | |
61952351 UD |
1256 | 3.6. I don't include any kernel headers myself but the compiler |
1257 | still complains about redeclarations of types in the kernel | |
1258 | headers. | |
1259 | ||
f12944ec UD |
1260 | {UD} The kernel headers before Linux 2.1.61 and 2.0.32 don't work correctly |
1261 | with glibc. Compiling C programs is possible in most cases but C++ programs | |
1262 | have (due to the change of the name lookups for `struct's) problems. One | |
1263 | prominent example is `struct fd_set'. | |
61952351 | 1264 | |
f12944ec UD |
1265 | There might be some problems left but 2.1.61/2.0.32 fix most of the known |
1266 | ones. See the BUGS file for other known problems. | |
61952351 UD |
1267 | |
1268 | ||
1269 | 3.7. Why don't signals interrupt system calls anymore? | |
1270 | ||
f12944ec UD |
1271 | {ZW} By default GNU libc uses the BSD semantics for signal(), unlike Linux |
1272 | libc 5 which used System V semantics. This is partially for compatibility | |
1273 | with other systems and partially because the BSD semantics tend to make | |
1274 | programming with signals easier. | |
f4017d20 UD |
1275 | |
1276 | There are three differences: | |
1277 | ||
1278 | * BSD-style signals that occur in the middle of a system call do not | |
1279 | affect the system call; System V signals cause the system call to | |
1280 | fail and set errno to EINTR. | |
1281 | ||
1282 | * BSD signal handlers remain installed once triggered. System V signal | |
1283 | handlers work only once, so one must reinstall them each time. | |
1284 | ||
1285 | * A BSD signal is blocked during the execution of its handler. In other | |
1286 | words, a handler for SIGCHLD (for example) does not need to worry about | |
61952351 | 1287 | being interrupted by another SIGCHLD. It may, however, be interrupted |
f4017d20 UD |
1288 | by other signals. |
1289 | ||
1290 | There is general consensus that for `casual' programming with signals, the | |
1291 | BSD semantics are preferable. You don't need to worry about system calls | |
1292 | returning EINTR, and you don't need to worry about the race conditions | |
1293 | associated with one-shot signal handlers. | |
1294 | ||
1295 | If you are porting an old program that relies on the old semantics, you can | |
1296 | quickly fix the problem by changing signal() to sysv_signal() throughout. | |
1297 | Alternatively, define _XOPEN_SOURCE before including <signal.h>. | |
1298 | ||
1299 | For new programs, the sigaction() function allows you to specify precisely | |
1300 | how you want your signals to behave. All three differences listed above are | |
1301 | individually switchable on a per-signal basis with this function. | |
1302 | ||
f12944ec UD |
1303 | If all you want is for one specific signal to cause system calls to fail and |
1304 | return EINTR (for example, to implement a timeout) you can do this with | |
f4017d20 UD |
1305 | siginterrupt(). |
1306 | ||
1307 | ||
61952351 UD |
1308 | 3.8. I've got errors compiling code that uses certain string |
1309 | functions. Why? | |
1310 | ||
f12944ec | 1311 | {AJ} glibc 2.1 has special string functions that are faster than the normal |
fdacb17d | 1312 | library functions. Some of the functions are additionally implemented as |
a25f2023 UD |
1313 | inline functions and others as macros. This might lead to problems with |
1314 | existing codes but it is explicitly allowed by ISO C. | |
04be94a8 | 1315 | |
04be94a8 | 1316 | The optimized string functions are only used when compiling with |
fdacb17d | 1317 | optimizations (-O1 or higher). The behavior can be changed with two feature |
f12944ec | 1318 | macros: |
61952351 UD |
1319 | |
1320 | * __NO_STRING_INLINES: Don't do any string optimizations. | |
1321 | * __USE_STRING_INLINES: Use assembly language inline functions (might | |
1322 | increase code size dramatically). | |
04be94a8 | 1323 | |
f12944ec UD |
1324 | Since some of these string functions are now additionally defined as macros, |
1325 | code like "char *strncpy();" doesn't work anymore (and is unnecessary, since | |
fdacb17d | 1326 | <string.h> has the necessary declarations). Either change your code or |
f12944ec | 1327 | define __NO_STRING_INLINES. |
04be94a8 | 1328 | |
f12944ec UD |
1329 | {UD} Another problem in this area is that gcc still has problems on machines |
1330 | with very few registers (e.g., ix86). The inline assembler code can require | |
1331 | almost all the registers and the register allocator cannot always handle | |
1332 | this situation. | |
04be94a8 | 1333 | |
61952351 | 1334 | One can disable the string optimizations selectively. Instead of writing |
04be94a8 UD |
1335 | |
1336 | cp = strcpy (foo, "lkj"); | |
1337 | ||
1338 | one can write | |
1339 | ||
1340 | cp = (strcpy) (foo, "lkj"); | |
1341 | ||
61952351 UD |
1342 | This disables the optimization for that specific call. |
1343 | ||
4775243a UD |
1344 | |
1345 | 3.9. I get compiler messages "Initializer element not constant" with | |
1346 | stdin/stdout/stderr. Why? | |
1347 | ||
1348 | {RM,AJ} Constructs like: | |
1349 | static FILE *InPtr = stdin; | |
1350 | ||
fdacb17d UD |
1351 | lead to this message. This is correct behaviour with glibc since stdin is |
1352 | not a constant expression. Please note that a strict reading of ISO C does | |
f12944ec | 1353 | not allow above constructs. |
4775243a | 1354 | |
f12944ec UD |
1355 | One of the advantages of this is that you can assign to stdin, stdout, and |
1356 | stderr just like any other global variable (e.g. `stdout = my_stream;'), | |
1357 | which can be very useful with custom streams that you can write with libio | |
fdacb17d | 1358 | (but beware this is not necessarily portable). The reason to implement it |
f12944ec | 1359 | this way were versioning problems with the size of the FILE structure. |
4775243a | 1360 | |
fdacb17d UD |
1361 | To fix those programs you've got to initialize the variable at run time. |
1362 | This can be done, e.g. in main, like: | |
1363 | ||
1364 | static FILE *InPtr; | |
bfcd44c3 | 1365 | int main(void) |
fdacb17d UD |
1366 | { |
1367 | InPtr = stdin; | |
1368 | } | |
1369 | ||
1370 | or by constructors (beware this is gcc specific): | |
1371 | ||
1372 | static FILE *InPtr; | |
1373 | static void inPtr_construct (void) __attribute__((constructor)); | |
1374 | static void inPtr_construct (void) { InPtr = stdin; } | |
1375 | ||
4775243a UD |
1376 | |
1377 | 3.10. I can't compile with gcc -traditional (or | |
1378 | -traditional-cpp). Why? | |
1379 | ||
1380 | {AJ} glibc2 does break -traditional and -traditonal-cpp - and will continue | |
fdacb17d | 1381 | to do so. For example constructs of the form: |
f12944ec | 1382 | |
4775243a UD |
1383 | enum {foo |
1384 | #define foo foo | |
1385 | } | |
f12944ec UD |
1386 | |
1387 | are useful for debugging purposes (you can use foo with your debugger that's | |
1388 | why we need the enum) and for compatibility (other systems use defines and | |
1389 | check with #ifdef). | |
4775243a UD |
1390 | |
1391 | ||
1392 | 3.11. I get some errors with `gcc -ansi'. Isn't glibc ANSI compatible? | |
1393 | ||
1394 | {AJ} The GNU C library is compatible with the ANSI/ISO C standard. If | |
f12944ec | 1395 | you're using `gcc -ansi', the glibc includes which are specified in the |
fdacb17d | 1396 | standard follow the standard. The ANSI/ISO C standard defines what has to be |
f12944ec UD |
1397 | in the include files - and also states that nothing else should be in the |
1398 | include files (btw. you can still enable additional standards with feature | |
1399 | flags). | |
4775243a | 1400 | |
f12944ec UD |
1401 | The GNU C library is conforming to ANSI/ISO C - if and only if you're only |
1402 | using the headers and library functions defined in the standard. | |
4775243a | 1403 | |
a35cb74d UD |
1404 | |
1405 | 3.12. I can't access some functions anymore. nm shows that they do | |
1406 | exist but linking fails nevertheless. | |
1407 | ||
f12944ec UD |
1408 | {AJ} With the introduction of versioning in glibc 2.1 it is possible to |
1409 | export only those identifiers (functions, variables) that are really needed | |
1410 | by application programs and by other parts of glibc. This way a lot of | |
1411 | internal interfaces are now hidden. nm will still show those identifiers | |
1412 | but marking them as internal. ISO C states that identifiers beginning with | |
1413 | an underscore are internal to the libc. An application program normally | |
1414 | shouldn't use those internal interfaces (there are exceptions, | |
1415 | e.g. __ivaliduser). If a program uses these interfaces, it's broken. These | |
1416 | internal interfaces might change between glibc releases or dropped | |
1417 | completely. | |
a35cb74d | 1418 | |
a5f4e34a UD |
1419 | |
1420 | 3.13. When using the db-2 library which comes with glibc is used in | |
1421 | the Perl db modules the testsuite is not passed. This did not | |
1422 | happen with db-1, gdbm, or ndbm. | |
1423 | ||
3877d9ea UD |
1424 | {MK} Db-2 does not support zero-sized keys. The Perl testsuite |
1425 | tests the support for zero-sized keys and therefore fails when db-2 is | |
1426 | used. The Perl folks are looking for a solution, but thus far have | |
1427 | not found a satisfactory one. | |
a5f4e34a | 1428 | |
5148d49f UD |
1429 | |
1430 | 3.14. The pow() inline function I get when including <math.h> is broken. | |
1431 | I get segmentation faults when I run the program. | |
1432 | ||
1433 | {UD} Nope, the implementation is correct. The problem is with egcs version | |
1434 | prior to 1.1. I.e., egcs 1.0 to 1.0.3 are all broken (at least on Intel). | |
1435 | If you have to use this compiler you must define __NO_MATH_INLINES before | |
1436 | including <math.h> to prevent the inline functions from being used. egcs 1.1 | |
1437 | fixes the problem. I don't know about gcc 2.8 and 2.8.1. | |
1438 | ||
05f732b3 UD |
1439 | |
1440 | 3.15. The sys/sem.h file lacks the definition of `union semun'. | |
1441 | ||
1442 | {UD} Nope. This union has to be provided by the user program. Former glibc | |
1443 | versions defined this but it was an error since it does not make much sense | |
1444 | when thinking about it. The standards describing the System V IPC functions | |
1445 | define it this way and therefore programs must be adopted. | |
1446 | ||
33127459 UD |
1447 | |
1448 | 3.16. Why has <netinet/ip_fw.h> disappeared? | |
1449 | ||
1450 | {AJ} The corresponding Linux kernel data structures and constants are | |
b710a6e2 | 1451 | totally different in Linux 2.0 and Linux 2.2. This situation has to be |
33127459 UD |
1452 | taken care in user programs using the firewall structures and therefore |
1453 | those programs (ipfw is AFAIK the only one) should deal with this problem | |
1454 | themselves. | |
1455 | ||
28ab8526 UD |
1456 | |
1457 | 3.17. I get floods of warnings when I use -Wconversion and include | |
1458 | <string.h> or <math.h>. | |
1459 | ||
1460 | {ZW} <string.h> and <math.h> intentionally use prototypes to override | |
1461 | argument promotion. -Wconversion warns about all these. You can safely | |
1462 | ignore the warnings. | |
1463 | ||
1464 | -Wconversion isn't really intended for production use, only for shakedown | |
1465 | compiles after converting an old program to standard C. | |
1466 | ||
5ff1a70a UD |
1467 | |
1468 | 3.18. After upgrading to glibc 2.1, I receive errors about | |
1469 | unresolved symbols, like `_dl_initial_searchlist' and can not | |
1470 | execute any binaries. What went wrong? | |
1471 | ||
1472 | {AJ} This normally happens if your libc and ld (dynamic linker) are from | |
1473 | different releases of glibc. For example, the dynamic linker | |
1474 | /lib/ld-linux.so.2 comes from glibc 2.0.x, but the version of libc.so.6 is | |
1475 | from glibc 2.1. | |
1476 | ||
1477 | The path /lib/ld-linux.so.2 is hardcoded in every glibc2 binary but | |
1478 | libc.so.6 is searched via /etc/ld.so.cache and in some special directories | |
1479 | like /lib and /usr/lib. If you run configure with another prefix than /usr | |
1480 | and put this prefix before /lib in /etc/ld.so.conf, your system will break. | |
1481 | ||
1482 | So what can you do? Either of the following should work: | |
1483 | ||
1484 | * Run `configure' with the same prefix argument you've used for glibc 2.0.x | |
1485 | so that the same paths are used. | |
1486 | * Replace /lib/ld-linux.so.2 with a link to the dynamic linker from glibc | |
1487 | 2.1. | |
1488 | ||
1489 | You can even call the dynamic linker by hand if everything fails. You've | |
1490 | got to set LD_LIBRARY_PATH so that the corresponding libc is found and also | |
1491 | need to provide an absolute path to your binary: | |
1492 | ||
1493 | LD_LIBRARY_PATH=<path-where-libc.so.6-lives> \ | |
1494 | <path-where-corresponding-dynamic-linker-lives>/ld-linux.so.2 \ | |
1495 | <path-to-binary>/binary | |
1496 | ||
1497 | For example `LD_LIBRARY_PATH=/libold /libold/ld-linux.so.2 /bin/mv ...' | |
1498 | might be useful in fixing a broken system (if /libold contains dynamic | |
1499 | linker and corresponding libc). | |
1500 | ||
1501 | With that command line no path is used. To further debug problems with the | |
1502 | dynamic linker, use the LD_DEBUG environment variable, e.g. | |
1503 | `LD_DEBUG=help echo' for the help text. | |
1504 | ||
1505 | If you just want to test this release, don't put the lib directory in | |
1506 | /etc/ld.so.conf. You can call programs directly with full paths (as above). | |
1507 | When compiling new programs against glibc 2.1, you've got to specify the | |
1508 | correct paths to the compiler (option -I with gcc) and linker (options | |
1509 | --dynamic-linker, -L and --rpath). | |
1510 | ||
61952351 | 1511 | |
7d1de115 UD |
1512 | 3.19. bonnie reports that char i/o with glibc 2 is much slower than with |
1513 | libc5. What can be done? | |
1514 | ||
1515 | {AJ} The GNU C library uses thread safe functions by default and libc5 used | |
1516 | non thread safe versions. The non thread safe functions have in glibc the | |
1517 | suffix `_unlocked', for details check <stdio.h>. Using `putc_unlocked' etc. | |
1518 | instead of `putc' should give nearly the same speed with bonnie (bonnie is a | |
1519 | benchmark program for measuring disk access). | |
9f6b6d8d | 1520 | |
b93492aa UD |
1521 | |
1522 | 3.20. Programs compiled with glibc 2.1 can't read db files made with glibc | |
1523 | 2.0. What has changed that programs like rpm break? | |
1524 | ||
1525 | {AJ} The GNU C library 2.1 uses db2 instead of db1 which was used in version | |
1526 | 2.0. The internal formats of the actual db files are different. To convert | |
1527 | the db files from db1 format to db2 format, you can use the programs | |
1528 | `db_dump185' and `db_load'. Alternativly programs can be linked with db1 | |
1529 | using `-ldb1' instead of linking with db2 which uses `-ldb'. Linking with | |
1530 | db1 might be preferable if older programs need to access the db file. | |
1531 | ||
1532 | db2 supports the old db1 programming interface and also a new programming | |
1533 | interface. For compilation with the old API, <db_185.h> has to be included | |
1534 | (and not <db.h>) and you can link with either `-ldb1' or `-ldb' for either | |
1535 | of the db formats. | |
1536 | ||
b5a9efcd UD |
1537 | |
1538 | 3.21. Autoconf's AC_CHECK_FUNC macro reports that a function exists, but | |
1539 | when I try to use it, it always returns -1 and sets errno to ENOSYS. | |
1540 | ||
1541 | {ZW} You are using a 2.0 Linux kernel, and the function you are trying to | |
1542 | use is only implemented in 2.1/2.2. Libc considers this to be a function | |
1543 | which exists, because if you upgrade to a 2.2 kernel, it will work. One | |
1544 | such function is sigaltstack. | |
1545 | ||
1546 | Your program should check at runtime whether the function works, and | |
1547 | implement a fallback. Note that Autoconf cannot detect unimplemented | |
1548 | functions in other systems' C libraries, so you need to do this anyway. | |
1549 | ||
1550 | ||
1551 | 3.22. My program segfaults when I call fclose() on the FILE* returned | |
1552 | from setmntent(). Is this a glibc bug? | |
1553 | ||
1554 | {GK} No. Don't do this. Use endmntent(), that's what it's for. | |
1555 | ||
1556 | In general, you should use the correct deallocation routine. For instance, | |
1557 | if you open a file using fopen(), you should deallocate the FILE * using | |
1558 | fclose(), not free(), even though the FILE * is also a pointer. | |
1559 | ||
1560 | In the case of setmntent(), it may appear to work in most cases, but it | |
1561 | won't always work. Unfortunately, for compatibility reasons, we can't | |
1562 | change the return type of setmntent() to something other than FILE *. | |
1563 | ||
9f6b6d8d UD |
1564 | \f |
1565 | . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . | |
1566 | ||
7d1de115 | 1567 | 4. Miscellaneous |
61952351 | 1568 | |
7d1de115 | 1569 | 4.1. After I changed configure.in I get `Autoconf version X.Y. |
61952351 UD |
1570 | or higher is required for this script'. What can I do? |
1571 | ||
1572 | {UD} You have to get the specified autoconf version (or a later one) | |
a35cb74d | 1573 | from your favorite mirror of ftp.gnu.org. |
61952351 | 1574 | |
04be94a8 | 1575 | |
7d1de115 | 1576 | 4.2. When I try to compile code which uses IPv6 headers and |
61952351 UD |
1577 | definitions on my Linux 2.x.y system I am in trouble. |
1578 | Nothing seems to work. | |
1579 | ||
f12944ec UD |
1580 | {UD} The problem is that IPv6 development still has not reached a point |
1581 | where the headers are stable. There are still lots of incompatible changes | |
1582 | made and the libc headers have to follow. | |
61952351 | 1583 | |
348ed515 UD |
1584 | {PB} The 2.1 release of GNU libc aims to comply with the current versions of |
1585 | all the relevant standards. The IPv6 support libraries for older Linux | |
1586 | systems used a different naming convention and so code written to work with | |
1587 | them may need to be modified. If the standards make incompatible changes in | |
1588 | the future then the libc may need to change again. | |
1589 | ||
1590 | IPv6 will not work with a 2.0.x kernel. When kernel 2.2 is released it | |
1591 | should contain all the necessary support; until then you should use the | |
1592 | latest 2.1.x release you can find. As of 98/11/26 the currently recommended | |
1593 | kernel for IPv6 is 2.1.129. | |
1594 | ||
1595 | Also, as of the 2.1 release the IPv6 API provided by GNU libc is not | |
1596 | 100% complete. In particular the getipnodebyname and getipnodebyaddr | |
1597 | functions are not implemented. | |
04be94a8 | 1598 | |
ff44f2a5 | 1599 | |
7d1de115 | 1600 | 4.3. When I set the timezone by setting the TZ environment variable |
ff44f2a5 UD |
1601 | to EST5EDT things go wrong since glibc computes the wrong time |
1602 | from this information. | |
1603 | ||
f12944ec UD |
1604 | {UD} The problem is that people still use the braindamaged POSIX method to |
1605 | select the timezone using the TZ environment variable with a format EST5EDT | |
8b4a4715 UD |
1606 | or whatever. People, if you insist on using TZ instead of the timezone |
1607 | database (see below), read the POSIX standard, the implemented behaviour is | |
f12944ec UD |
1608 | correct! What you see is in fact the result of the decisions made while |
1609 | POSIX.1 was created. We've only implemented the handling of TZ this way to | |
1610 | be POSIX compliant. It is not really meant to be used. | |
1611 | ||
1612 | The alternative approach to handle timezones which is implemented is the | |
1613 | correct one to use: use the timezone database. This avoids all the problems | |
1614 | the POSIX method has plus it is much easier to use. Simply run the tzselect | |
1615 | shell script, answer the question and use the name printed in the end by | |
8b4a4715 UD |
1616 | making a symlink /etc/localtime pointing to /usr/share/zoneinfo/NAME (NAME |
1617 | is the returned value from tzselect). That's all. You never again have to | |
1618 | worry. | |
f12944ec UD |
1619 | |
1620 | So, please avoid sending bug reports about time related problems if you use | |
1621 | the POSIX method and you have not verified something is really broken by | |
1622 | reading the POSIX standards. | |
ff44f2a5 | 1623 | |
fdacb17d | 1624 | |
7d1de115 | 1625 | 4.4. What other sources of documentation about glibc are available? |
fdacb17d UD |
1626 | |
1627 | {AJ} The FSF has a page about the GNU C library at | |
1628 | <http://www.gnu.org/software/libc/>. The problem data base of open and | |
1629 | solved bugs in GNU libc is available at | |
1630 | <http://www-gnats.gnu.org:8080/cgi-bin/wwwgnats.pl>. Eric Green has written | |
1631 | a HowTo for converting from Linux libc5 to glibc2. The HowTo is accessable | |
1632 | via the FSF page and at <http://www.imaxx.net/~thrytis/glibc>. Frodo | |
1633 | Looijaard describes a different way installing glibc2 as secondary libc at | |
1634 | <http://huizen.dds.nl/~frodol/glibc>. | |
1635 | ||
1636 | Please note that this is not a complete list. | |
1637 | ||
348ed515 | 1638 | |
7d1de115 | 1639 | 4.5. The timezone string for Sydney/Australia is wrong since even when |
348ed515 UD |
1640 | daylight saving time is in effect the timezone string is EST. |
1641 | ||
1642 | {UD} The problem for some timezones is that the local authorities decided | |
1643 | to use the term "summer time" instead of "daylight saving time". In this | |
1644 | case the abbreviation character `S' is the same as the standard one. So, | |
1645 | for Sydney we have | |
1646 | ||
1647 | Eastern Standard Time = EST | |
1648 | Eastern Summer Time = EST | |
1649 | ||
1650 | Great! To get this bug fixed convince the authorities to change the laws | |
1651 | and regulations of the country this effects. glibc behaves correctly. | |
1652 | ||
eeabe877 | 1653 | |
7d1de115 | 1654 | 4.6. I've build make 3.77 against glibc 2.1 and now make gets |
eeabe877 UD |
1655 | segmentation faults. |
1656 | ||
1657 | {AJ} GNU make 3.77 has support for 64 bit filesystems which is slightly | |
1658 | broken (and one of the new features in the GNU C library 2.1 is 64 bit | |
1659 | filesystem support :-( ). To get a working make you can use either make | |
1660 | 3.75 or patch 3.77. A working patch is available via RedHat's Rawhide server | |
1661 | (ftp://rawhide.redhat.com/SRPMS/SRPMS/make-3.77-*src.rpm). | |
1662 | ||
f8cac037 | 1663 | \f |
61952351 UD |
1664 | ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ |
1665 | ||
f8cac037 RM |
1666 | Answers were given by: |
1667 | {UD} Ulrich Drepper, <drepper@cygnus.com> | |
613a76ff | 1668 | {DMT} David Mosberger-Tang, <davidm@AZStarNet.com> |
dd7d45e8 | 1669 | {RM} Roland McGrath, <roland@gnu.org> |
1f205a47 | 1670 | {AJ} Andreas Jaeger, <aj@arthur.rhein-neckar.de> |
22d57dd3 | 1671 | {EY} Eric Youngdale, <eric@andante.jic.com> |
a5a0310d | 1672 | {PB} Phil Blundell, <Philip.Blundell@pobox.com> |
af6f3906 | 1673 | {MK} Mark Kettenis, <kettenis@phys.uva.nl> |
f4017d20 | 1674 | {ZW} Zack Weinberg, <zack@rabi.phys.columbia.edu> |
612fdf25 | 1675 | {TK} Thorsten Kukuk, <kukuk@suse.de> |
8619129f | 1676 | {GK} Geoffrey Keating, <geoffk@ozemail.com.au> |
a35cb74d | 1677 | {HJ} H.J. Lu, <hjl@gnu.org> |
b710a6e2 | 1678 | {CG} Cristian Gafton, <gafton@redhat.com> |
f8cac037 RM |
1679 | \f |
1680 | Local Variables: | |
61952351 UD |
1681 | mode:outline |
1682 | outline-regexp:"\\?" | |
f12944ec | 1683 | fill-column:76 |
f8cac037 | 1684 | End: |