Mantis Bugtracker
  

Viewing Issue Simple Details Jump to Notes ] View Advanced ] Issue History ] Print ]
ID Category Severity Reproducibility Date Submitted Last Update
0000933 [Croquet] Jasmine crash random 03-09-05 02:45 04-18-05 22:12
Reporter elenzil View Status public  
Assigned To
Priority normal Resolution open  
Status new  
Summary 0000933: VM Crash during B3DHardwareEngine>swapBuffers
Description I did a stack dump on the unpredictable* crashes i get sometimes
in resource-heavy scenes, and it seems to be crashing in
B3DHardwareEngine>swapBuffers.

Anyone have a similar experience ?

Happens most often on XP laptops,
but also on full machine,
ATI Radeon 9XXX cards in both cases.
Additional Information Crash dump:

---------------------------------------------------------------------
Mon Jan 31 15:18:34 2005

Exception code: C0000005
Exception addr: 690E44C7
Access violation (read access) at 00006636
EAX:00000001 EBX:00000000 ECX:69630C84 EDX:00000000
ESI:4AE5D900 EDI:00000000 EBP:00E7AE60 ESP:0006FA48
EIP:690E44C7 EFL:00010246
FP Control: FFFF027F
FP Status: FFFF7869
FP Tag: FFFFFFFF
VM Version: Squeak 3.6.2 / Tea 1.9 VM (release) from Nov 12 2003
Compiler: gcc 2.95.2 19991024 (release)

Current byte code: 225
Primitive index: 117

Loaded plugins:
    Matrix2x3Plugin 12 November 2003 (i)
    SocketPlugin 12 November 2003 (i)
    B2DPlugin 12 November 2003 (i)
    SoundGenerationPlugin 12 November 2003 (i)
    SoundPlugin 12 November 2003 (i)
    opengl32.dll
    SqueakFFIPrims 12 November 2003 (e)
    B3DAcceleratorPlugin 12 November 2003 (i)
    BMPReadWriterPlugin 12 November 2003 (i)
    JPEGReadWriter2Plugin 12 November 2003 (i)
    ZipPlugin 12 November 2003 (i)
    Squeak3D 12 November 2003 (i)
    FloatArrayPlugin 12 November 2003 (i)
    LargeIntegers v1.2 12 November 2003 (i)
    BitBltPlugin 12 November 2003 (i)
    SecurityPlugin 12 November 2003 (i)
    FilePlugin 12 November 2003 (i)
    MiscPrimitivePlugin 12 November 2003 (i)


Stack dump:

278170004 B3DHardwareEngine>swapBuffers
278169756 TeapotMorph>forceToScreen
278170376 [] in DisplayScreen>forceDamageToScreen:
278169664 SequenceableCollection>do:
278168692 DisplayScreen>forceDamageToScreen:
278168600 WorldState>forceDamageToScreen:
278104784 WorldState>displayWorld:submorphs:
278104692 PasteUpMorph>privateOuterDisplayWorld
278104600 PasteUpMorph>displayWorld
278104324 [] in WorldState>displayWorldSafely:
278104232 BlockContext>on:do:
278104140 BlockContext>ifError:
278104048 WorldState>displayWorldSafely:
278091148 WorldState>doOneCycleNowFor:
278090504 WorldState>doOneCycleFor:
278090412 PasteUpMorph>doOneCycle
186958832 [] in Project class>spawnNewProcess
186959016 [] in BlockContext>newProcess

Attached Files

- Relationships

- Notes
(0001337 - 56 - 56 - 56 - 56 - 56 - 56)
elenzil
04-01-05 21:50

this appears to be due to using too much texture memory.
 
(0001360 - 58 - 58 - 58 - 58 - 58 - 58)
elenzil
04-13-05 02:28

this bug remains an absolute show stopper for our project.
 
(0001362 - 518 - 524 - 524 - 524 - 524 - 524)
andreas
04-13-05 07:44

I have been completely unable to reproduce this problem - is this possibly a problem with the ATI drivers? One thing to try is to run TTextureStress (which was specifically designed for testing boundary conditions) and see if this allows you to make the behavior more reliable. When I have tried this, the effect was always the same: Heavy slowdown, heavy paging, but no crashes. Try increasing the numbers inside TTextureStress>>setupStressTest to see if you can make the crash happen at least a little more reliable.
 
(0001371 - 136 - 160 - 160 - 160 - 160 - 160)
elenzil
04-13-05 12:05

thanks, i will try that,
and try to reproduce on other cards.

we've seen it in three flavours of ATI cards,
but untested on others.
 
(0001376 - 311 - 311 - 311 - 311 - 311 - 311)
elenzil
04-14-05 01:57

well with some modification i got TTextureStress to add 30 1024x1024x32 textures without crashing, and via conventional means got the texture usage up to about 330MB, which is much larger than my graphics card's 128MB, so this may unfortunately be hard to reproduce. Will continue looking for a sure-fire crash.
 
(0001377 - 1193 - 1205 - 1205 - 1205 - 1205 - 1205)
andreas
04-14-05 08:35

Yes, that's what I expected (and what should happen). In theory, due to having AGP all around these days the cards should be very happy with almost any amount of textures. As far as I can tell this leaves us with two possibilities:
a) It's an obscure bug which happens only when you have the right pattern of use (e.g., sizes and order involved). I will say I find this somewhat hard to believe but you may want to Google to see if OpenGL crashes have been reported with your graphics chip (if there's a problem I'm almost certain someone would have found and reported this with a game in some forum - if there's a specific ATI problem I would expect that Google will find you a few hundred of really pissed off users).
b) The crashes has *nothing* to do with textures. That would be my guess. That it crashes in SwapBuffers() only means that the GPU is at work but the cause of the problem could be anywhere before. Notably, I have heard rumors of a bug in the ASE importer which might screw up indexed meshes (talk to David about this if you can). But the only way to find the problem is to track it down by having some (at least semi-reliable) way of actually forcing the system to crash.
 
(0001383 - 181 - 217 - 217 - 217 - 217 - 217)
elenzil
04-18-05 22:12

Um,

this is no longer reproducible on 3 of 3 systems who's drivers i updated.

the previous drivers were only 6 months old, so, yeah. update drivers!

sorry and thanks again.
 

- Issue History
Date Modified Username Field Change
03-09-05 02:45 elenzil New Issue
04-01-05 21:50 elenzil Note Added: 0001337
04-13-05 02:28 elenzil Note Added: 0001360
04-13-05 07:44 andreas Note Added: 0001362
04-13-05 12:05 elenzil Note Added: 0001371
04-14-05 01:57 elenzil Note Added: 0001376
04-14-05 08:35 andreas Note Added: 0001377
04-18-05 22:12 elenzil Note Added: 0001383


Mantis 1.0.8[^]
Copyright © 2000 - 2007 Mantis Group
56 total queries executed.
39 unique queries executed.
Powered by Mantis Bugtracker