Memory crash?

Hi! Ive got a strange problem. Ive tried the search the forums
but so far no luck...
My dedicated CS server (hlds_amd) on my debian box are eating memory....
Have anyone seen this...??

If I run this configuration my memory slowly grows.
In the end the server will crash..
hlguard v1.50
amx v0.16
metamodx v1.7X3
chicken v1.0.5.1
ivpbot 1.1.1

in my debug.log I get:
CRASH: Tue May 4 18:36 CEST 2004
Start Line: ./hlds_amd -game cstrike ....etc.....
Using host libthread_db library "/lib/libthread_db.so.1".
Core was generated by `./hlds_amd -game cstrike ....etc.....
Program terminated with signal 11, Segmentation fault.
#0 0x438d544d in ?? ()
#0 0x438d544d in ?? ()
End of crash report


Ive tried different versions of metamod! The original one, valve´s and metamodx
valve´s and metamodx work best but the server still crash!! The same happends
with amxmod and amxmodx.

If I dont load ivpbot when starting the server the memory grows
slower and it takes longer time for the server to crash.. But it still crash..

Have anyone else noticed this problem?

[Whistler]

Perhaps you have to wait for kacat come back and release a new version if you still want to use ivpbot
and btw, what's that "valve's version" of Metamod ?

[Pierre-Marie Baty]

There are known memory leaks in the POD-bot 2.6 source code ; maybe they haven't been fixed in IvPBot. KaCaT can look in PB2.6mm for the fix.
There might be memory leaks within metamod as well...

[Whistler]

"There are known memory leaks in the POD-bot 2.6 source code"
...but I have tested my current code with borland's codeguard and found no memory leak ? I never fixed any memory leaks in purpose. So would you please tell me where the memory leak is ? Thanks.

[Pierre-Marie Baty]

ouch, this is gonna get tricky. There have been so many changes over here to the source code that I can't remember exactly what it was like in its original form.

There is one in the function that builds the stringnodes for the Eliza chatter bot.
There is one with the waypoint loading process
When the waypoints are loaded (for ex. for a new map), there is also at least one memory chunk that is freed twice.
That's all I can remember, but sorry I can't be more explicit, I would have to go and dig that myself in the genuine PB26 source code again...