| 1 |
|
Known Bugs worthy of a mention: |
| 2 |
|
-------------------------------------------------------------------------------- |
| 3 |
|
|
| 4 |
< |
1. /MODUNLOAD causes cores: |
| 5 |
< |
- If a module is modified before being unloaded, /MODUNLOAD (and |
| 6 |
< |
therefore /MODRELOAD) may cause a core. |
| 7 |
< |
|
| 8 |
< |
This problem is caused by the behaviour of the OS, which treats |
| 9 |
< |
shared libraries differently to executables (modifying the ircd |
| 10 |
< |
binary whilst it is running would also cause a core, but is denied |
| 11 |
< |
by the OS). There is no way to fix this at the application level, |
| 12 |
< |
and fixing the OS to do the right thing is also difficult. |
| 13 |
< |
|
| 14 |
< |
A workaround to avoid coring is possible however. To install new |
| 15 |
< |
modules, first remove or rename the old module, then copy/move the |
| 16 |
< |
new file into place. /MODUNLOAD will then work successfully. |
| 17 |
< |
|
| 18 |
< |
Alternatively, running ./configure with the --disable-shared-modules |
| 19 |
< |
argument will link all the commands statically, losing the advantages |
| 20 |
< |
of upgrading at runtime, but reducing the chances of accidentally |
| 21 |
< |
coring your server. |
| 22 |
< |
|
| 23 |
< |
2. Solaris stock 32bit stdio is limited to 256 fds. This breaks when using |
| 24 |
< |
cryptlinks. |
| 25 |
< |
|
| 26 |
< |
3. New bug reported in some 64 bit and powerpc (yellowdog linux) code. |
| 27 |
< |
This bug is actually in gcc itself, the spec says you don't modify |
| 28 |
< |
a va_list in libc, but gcc does. The bug manifests itself as |
| 29 |
< |
immediately coring when anyone types anything on a channel. |
| 30 |
< |
You will find a quick work around in contrib until this bug in gcc |
| 31 |
< |
is fixed. |
| 32 |
< |
|
| 33 |
< |
This is a quick test program |
| 34 |
< |
libc_vprintf |
| 35 |
< |
cc -o libc_vprintf libc_vprintf.c |
| 36 |
< |
./libc_vprintf |
| 37 |
< |
|
| 38 |
< |
If the program doesn't core, you are fine. |
| 39 |
< |
|
| 40 |
< |
The work around is the vprintf_fix.patch in contrib. |
| 41 |
< |
(Reported by several people, fix and test by David) |
| 42 |
< |
|
| 4 |
> |
* None |
| 5 |
> |
|
| 6 |
> |
|
| 7 |
|
BUG REPORTS: If you run this code and encounter problems, you should report |
| 8 |
|
the bug on our SourceForge.net bug tracker, which you can find at this URL: |
| 9 |
|
|
| 10 |
|
https://sourceforge.net/tracker/?atid=409046&group_id=33573&func=browse |
| 11 |
|
|
| 12 |
< |
Please include a gdb backtrace and a copy of your defaults.h (if changed), |
| 13 |
< |
setup.h, and ircd.conf with any report (with passwords and other sensitive |
| 50 |
< |
information masked). |
| 12 |
> |
Please include a gdb backtrace and a copy of your config.h and ircd.conf with |
| 13 |
> |
any report (with passwords and other sensitive information masked). |
| 14 |
|
|
| 15 |
|
For information how to get a gdb backtrace, see INSTALL (near the end of file). |
| 16 |
|
|
