Completely Random crash, but pretty frequent.
by Dreamer · in Torque Game Engine · 04/22/2005 (9:36 am) · 8 replies
I am having an issue with pretty frequent crashing during lighting calcs at game start clientside only.
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 16384 (LWP 9399)]
0x082254c5 in SceneLighting::addInterior(ShadowVolumeBSP*, SceneLighting::InteriorProxy&, LightInfo*, int) ()Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 16384 (LWP 9399)]
0x082254c5 in SceneLighting::addInterior(ShadowVolumeBSP*, SceneLighting::InteriorProxy&, LightInfo*, int) ()
This appears to be completely random, but will happen roughly once or twice per day, exact same error, exact same place, but it doesn't matter if I have deleted my .garagegames directory (containing all my.dso files).
Funny thing is NO engine mods I have made even touch the lighting code, or anything directly associated. Also one would tend to think that an error like this would happen most if not all of the time, it doesn't. It only happens often enough to be a serious nusiance.
Any ideas, on what could be wrong here?
Almost forgot...
TGE 1.3 on Gentoo Linux running latest.
ATI Radeon 9200 SE Videocard 256MB
1 GB of Ram
AMD 2200 +XP
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 16384 (LWP 9399)]
0x082254c5 in SceneLighting::addInterior(ShadowVolumeBSP*, SceneLighting::InteriorProxy&, LightInfo*, int) ()Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 16384 (LWP 9399)]
0x082254c5 in SceneLighting::addInterior(ShadowVolumeBSP*, SceneLighting::InteriorProxy&, LightInfo*, int) ()
This appears to be completely random, but will happen roughly once or twice per day, exact same error, exact same place, but it doesn't matter if I have deleted my .garagegames directory (containing all my.dso files).
Funny thing is NO engine mods I have made even touch the lighting code, or anything directly associated. Also one would tend to think that an error like this would happen most if not all of the time, it doesn't. It only happens often enough to be a serious nusiance.
Any ideas, on what could be wrong here?
Almost forgot...
TGE 1.3 on Gentoo Linux running latest.
ATI Radeon 9200 SE Videocard 256MB
1 GB of Ram
AMD 2200 +XP
#2
BTW this is happening on a practically brand new TGE install with Starter.FPS
Also when run in client/server mode, this still happens, but again as I said clientside only.
04/22/2005 (10:03 am)
Well no, nothing is getting deleted. As for dropping into the debugger, that output is FROM my debugger, and doing a BackTrace isn't revealing much of anything. Next time it happens though I will run a backtrace on it and post the results, maybe you'll see something I'm not.BTW this is happening on a practically brand new TGE install with Starter.FPS
Also when run in client/server mode, this still happens, but again as I said clientside only.
#3
04/22/2005 (10:09 am)
In Windows when i crash i can drop into the debugger and see the file and exact line number where the crash occurred (not just the memory address). Usually this points you to a null or garbage pointer and can lead you to the real source of the error. I'm not a Linux user, but I assume there is a way to do the same thing under Linux.
#4
Also, my windows build isn't suffering this problem. Makes me think this might just be due to some wierd ATI Driver issue, I do have the latest Kernel mode driver for linux going.
04/22/2005 (10:59 am)
I could too if I were running the debug compiled version, but I'm actually just running RELEASE. However this problem does not occur on the debug compiled version (or at least I haven't gotten it to occur, after about 150 tries at replicating it.).Also, my windows build isn't suffering this problem. Makes me think this might just be due to some wierd ATI Driver issue, I do have the latest Kernel mode driver for linux going.
#5
In the past when i had semi-random crashes in a release build i would rebuild the release with debug info and run it with the debugger all day until it would occur.
04/22/2005 (11:03 am)
Again in MSVC++ you can compile a release version with enough debug info to get some line information... at least close enough to be able to tell what happened. I suspect GCC or whatever else in Linux has the same ability.In the past when i had semi-random crashes in a release build i would rebuild the release with debug info and run it with the debugger all day until it would occur.
#6
04/22/2005 (11:08 am)
Good point it's worth a shot I guess.
#7
04/24/2005 (6:05 pm)
Do you have any custom interiors in the engine?
#8
WinXP pro
GeforceFX 5900 w/128MB
2.6 Ghz P4 HT w800Mhz FSB
1 Gig DDR400 Samsung RAM
"Do you have any custom interiors in the engine?"
Yes and no. All buildings are for exterior view only, they are nothing but shells that are never entered or even touched. If it were a building, it would seem that the problem would be constant untill that buildin were removed. But this crash is totally random, completely unpredictable, and only during mission lighting when the .ml file is being generated. If the lighting file exists already then I have no problems.
04/24/2005 (7:23 pm)
I've had this exact same problem during mission lighting. Every now and then I will just crash, or after lighting some things do not work right(collisions, triggers, etc..). In those cases, I can exit, come right back, and then everything works. I haven't found any other way to reproduce the lighting crash other than to just keep loading missions and it eventually happens. I'm on Windows XPpro, and like dreamer I do not believe anything I have done is even remotely connected to the lighting code in the engine. I'll see if I can get any debug info also but now that Dreamer mentions it, I don't think I've ever had the problem at all in Debug so I'm not sure what luck I'll have.WinXP pro
GeforceFX 5900 w/128MB
2.6 Ghz P4 HT w800Mhz FSB
1 Gig DDR400 Samsung RAM
"Do you have any custom interiors in the engine?"
Yes and no. All buildings are for exterior view only, they are nothing but shells that are never entered or even touched. If it were a building, it would seem that the problem would be constant untill that buildin were removed. But this crash is totally random, completely unpredictable, and only during mission lighting when the .ml file is being generated. If the lighting file exists already then I have no problems.
Associate Tom Spilman
Sickhead Games