Samp03svr with debugging symbols? - Printable Version
+- SA-MP Forums Archive (https://sampforum.blast.hk)
+-- Forum: SA-MP Server (https://sampforum.blast.hk/forumdisplay.php?fid=6)
+--- Forum: Server Support (https://sampforum.blast.hk/forumdisplay.php?fid=19)
+--- Thread: Samp03svr with debugging symbols? (/showthread.php?tid=168419)
+- SA-MP Forums Archive (https://sampforum.blast.hk)
+-- Forum: SA-MP Server (https://sampforum.blast.hk/forumdisplay.php?fid=6)
+--- Forum: Server Support (https://sampforum.blast.hk/forumdisplay.php?fid=19)
+--- Thread: Samp03svr with debugging symbols? (/showthread.php?tid=168419)
Samp03svr with debugging symbols? - nemesis- - 16.08.2010
Would it be possible to get a samp03svr linux executable with debugging symbols enabled? The reason is to help track down root causes of issues like this:
Quote:
|
Aug 11 20:49:35 ks307566 kernel: [69836.711847] samp03svr[1387]: segfault at 42c40010 ip 00000000f7578042 sp 00000000ff8fc6d4 error 6 in libc-2.9.so[f7505000+15c000] |
Re: Samp03svr with debugging symbols? - Kalcor - 16.08.2010
Run the program in gdb. When it segfaults, type> bt
Post the backtrace in this thread and I'll try and tell you what it is. If I don't notice this thread again you can PM me.
Re: Samp03svr with debugging symbols? - nemesis- - 16.08.2010
Thanks. I'll give it a try.
Re: Samp03svr with debugging symbols? - nemesis- - 19.09.2010
Kal,
An update on this:
I ran samp03svr under gdb with no success. The output from a 'bt' showed an array of memory addresses with nothing attached to them, just question marks. I decided to compile all the plugins that I use (8 in total) with debug symbols (compiled with -g -O0 -rdynamic) and run the server under gdb.
These 8 plugins are:
G-Stylez MySQL R5
Incognito Plugins (Streamer v2.5, DNS v2.4, IRC v1.3.6 SSL, GVars v1.3)
****** Plugins (Sscanf 2.0, Whirlpool 64-bit version)
Kalcor's MapAndreas v1.0 beta
Additionally, I added logging to every function, every callback, and every command (in-game and irc, both Zcmd-driven command systems). When the server does the libc segfault, it is always on a different function. This would lead me to believe the issue stems from something outside the script.
My server's specs are:
Kimsufi/OVH dedicated server
Intel Q6600 Quad-core CPU
4GB DDR2 Ram
1TB SATA2 (non-RAID)
100MB up/down
OS: Ubuntu Linux v9.04 64-bit
This is the output I got from the 'bt' output. With all plugins under debug, this is what I got after 2 successful 'bt' captures under gdb:
http://pawn.pastebin.com/N675w5ss
I initially assumed it was the streamer plugin though under debug I would have expected more output from gdb so I ran the server again under gdb.
My understanding is that Blacklite from SACNR has this exact same issue with libc segfaults under completely random functions. We've both run similar tests with little useful output from gdb/bt.
UPDATE: I found output that I had from an strace during one of the crashes: http://pawn.pastebin.com/w4UA7y1r. Here is one where it would commonly happen when a connection was made (prior to OnPlayerConnect firing), a known IP from a regular - not a malicious attacker: http://pawn.pastebin.com/EE2wuTcF. You can see my function execution logging in there.
Re: Samp03svr with debugging symbols? - Blacklite - 20.09.2010
Quote:
|
Originally Posted by nemesis-
 Kal,
An update on this: I ran samp03svr under gdb with no success. The output from a 'bt' showed an array of memory addresses with nothing attached to them, just question marks. I decided to compile all the plugins that I use (8 in total) with debug symbols (compiled with -g -O0 -rdynamic) and run the server under gdb. These 8 plugins are: G-Stylez MySQL R5 Incognito Plugins (Streamer v2.5, DNS v2.4, IRC v1.3.6 SSL, GVars v1.3) ****** Plugins (Sscanf 2.0, Whirlpool 64-bit version) Kalcor's MapAndreas v1.0 beta |
Adrenaline's MySQL plugin (that has been working for over a year)
A custom checkpoint streamer (been working for months)
Incognito's IRC plugin
All the plugins are compiled with -g -O0 -rdynamic.
Quote:
|
Originally Posted by nemesis-
Additionally, I added logging to every function, every callback, and every command (in-game and irc, both Zcmd-driven command systems). When the server does the libc segfault, it is always on a different function. This would lead me to believe the issue stems from something outside the script.
|
Quote:
|
Originally Posted by nemesis-
My server's specs are:
Kimsufi/OVH dedicated server Intel Q6600 Quad-core CPU 4GB DDR2 Ram 1TB SATA2 (non-RAID) 100MB up/down OS: Ubuntu Linux v9.04 64-bit |
Intel i7 Quad-core CPU
8GB DDR2 Ram
1TB SATA2 (non-RAID)
100MB up/down
OS: Debian Lenny v5.06 32-bit
Quote:
|
Originally Posted by nemesis-
This is the output I got from the 'bt' output. With all plugins under debug, this is what I got after 2 successful 'bt' captures under gdb:
http://pawn.pastebin.com/N675w5ss |
Quote:
|
Originally Posted by nemesis-
I initially assumed it was the streamer plugin though under debug I would have expected more output from gdb so I ran the server again under gdb.
My understanding is that Blacklite from SACNR has this exact same issue with libc segfaults under completely random functions. We've both run similar tests with little useful output from gdb/bt. |
Re: Samp03svr with debugging symbols? - nemesis- - 23.09.2010
Please?
Re: Samp03svr with debugging symbols? - Blacklite - 27.09.2010
Quick update: I fixed the getarg bug, but I'm still having the libc one... I'll post another backtrace if I can get it.
Re: Samp03svr with debugging symbols? - nemesis- - 25.11.2010
It looks like we(Blacklite and I) may have solved this LIBC bug. It appears using Kick(playerid) in OnPlayerUpdate is still an issue under 0.3b-R2. If you use Kick() in OnPlayerUpdate and someone shortly connects in and inherits the playerid of the person who was kicked, the server segfaults.