How to debug linux crashes

[[NvT]_KaszpiR_]

KWo got an idea to also run epodbot on de_cahetau and test if it crashes aswell.

[KWo]

Originally Posted by Whistler

if you have mingw, try running the drmingw.exe file in the MingW/bin directory with the parameter -i. Then when it crashed, just click the "Debug" button. You still need to compile with -g3 to know the exact place of the crash.

Do I need to click that button right after it crashes or I can do this after few hours to get the correct info? I mean - I start my HLDS for the night and I go to sleep - at morning I can see it crashed - then I would like to get that info drmingw can give me (by pressing that button You said).

[Whistler]

If you compiled with the -g3 it will likely to give you the correct info by clicking the "debug" button in that famous "This program has performed illegal operation, please contact your vendor" dialog box. You just need to run "drmingw -i" to install it.

Or in Windows NT/2000/XP the debug button should be "Cancel" (which is rather confusing).

[KWo]

Yeah - before reading your post I saw it might be "cancel", but on my XP it was just "Debug".
Here is the info (I think I should remove some flags and use g3 only, because the info still isn't full) :

Originally Posted by drmingw

hlds.exe caused an Access Violation at location 077f5db9 in module podbot_mm.dll Reading from location 00000174.

Registers:
eax=0012f284 ebx=00000095 ecx=00000174 edx=00000000 esi=07813454 edi=00000000
eip=077f5db9 esp=0012f214 ebp=0012f22c iopl=0 nv up ei pl nz na po nc
cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00000206

Call stack:
077F5DB9 podbot_mm.dll:077F5DB9 GetEngineFunctions
077D1FCE podbot_mm.dll:077D1FCE
077E2D84 podbot_mm.dll:077E2D84
077F11F6 podbot_mm.dll:077F11F6 GiveFnptrsToDll

Can You tell me now what is this? I was trying to find GiveFnptrsToDll and it's defined in dll.cpp, but... it's nowhere called (at least not clearly way)...
Note - I compiled podbot with mm1.18.p26, but the game crashes on this map also at splorygon's linux server and I believe he is compiling pod with standard metamod.

[Whistler]

this looks like the call stack is messed up by invalid memory access, something like this:

char a[5];
strcpy(a, "abcdefghijklmnopqrstuvwxyz"); // the string is way longer than 5 chars

That function is actually called by the hl engine, and only called once at the game start.

[KWo]

You mean really "game start" or "round start"? I'm playing this map all time with mp_timelimit=0, mp_maxrounds=0, mw_winlimit=0 - after enf od one round the new one is started and the map isn't changed.
Anyway - how to narrow better the part of code causing such crash situation?

[Whistler]

Only when the CS launches.

[KWo]

So - how is possible to get such error if CS isn't restarted?

[[NvT]_KaszpiR_]

GiveFpnrsTdoDLL (or sht liek that ;P) is onyl initiated when metamod loads the plugin , not round start or game start

he ment metamod intitalization (well game start we can say)

[KWo]

Well - if th emap is never ending I don;t see the reason why it shoud be called them. But I noticed metamod is trying to load ini file (with list of plugin) also at the end of map (before it's changed). But th emap shoudn't finish with my settings.
BTW - @Whistler - can You give me the exact link to the page which clearly says what are available flags for mingw and cygwin compilations? I couldn't find it in the main page for docs on www.mingw.org...