Black screen then crash


I posted a test over on JGO to get some performance figures, and one responce came back about the test just crashing. I asked them to run the webstart tests from JME here and it does the same thing. I was wondering if anyone else had seen this issue or has an idea what it might be?

The responce is here, and the post a little furthur up lists the machine specs.

Not sure he has any stack trace, but we might be able to sort some out, but is this card a known issue?



works here…

athlon x2 3800

2gb ram

gf 7800 gt

arch linux current/java 1.5.0_06

300fps@default settings

all the tests (i have run) work fine here, too, the only thing that doesn't work is furballz.

maybe you should ask me, the guy at jgo, not me here :slight_smile: to send in his log-files.

Ouch, an S3 card. That's going to be very finicky about what GL calls are made. But yeah, as Pavel suggested, if he can provide more information on the crash we might be able to track down the issue and see what can be done. He does mention that Kevin Glass's Tiltilation game did the same until recently, so maybe there are some hints there.

He’s posted a stack trace here.

Also, Kev mentioned that lwjgl did some updates to not just crash on horrid cards, whats the latest version in cvs?


s3 has allways had serious problems with their drivers…even with that prosavage card he has(with a whopping 16Mb mem) it's easy to google up things like "Via S3 SuperSavage and ProSavage DDR when used with OpenGL renders incorrectly and/or crashes due to driver problems. Please use DirectX as it works well" :slight_smile:

Invalid enum would imply that some GL function is getting sent a constant it believes is incorrect. Could his card only support a low OpenGL version 1.1 or so, where we send a 1.3 constant through?

yeah that sounds like a good explanation.

So… does anyone know how to find the offending call from what little information we have? Invalid Enum implies it can be anything, bad constant sent to glEnable, bad constant to glMatrixMode, etc, etc, etc.

a strange thing is the exception comes from the swapBuffers call

That's because that is when LWJGL checks the errors.

i must have an old lwjgl source or something

Or maybe I'm wrong… more likely.

nope, i was looking at the 0.99 version that didnt even have a swapbuffers method in :slight_smile: sigh

one thing we use above gl 1.1 that are not checked with OpenGL12 is GL12.GL_RESCALE_NORMAL … the gl13 stuff i find is checked before used through the capabilities…

Actually, we do test that, see where it says "&& capabilities.OpenGL12" on line 1371?  That's the only place the renderer can be put into a position to use GL_RESCALE_NORMAL.

ah you are right, that's the only place the enabling is done…

yeah, and when disabling, it will only happen if previously enabled.