Odd culling behavior under Linux

[fbi]

Hi people, this is my first post...sorry, it's cause of a bug  
I found out that running two examples in LWJGL 0.5 (under RedHat 8), namely Blending and Fog, the faces that should point outward point inward and viceversa. Running the same example in LWJGL 0.4 all is fine  
My system is a 800 Mhz Athlon, GeForce 2 Titanium, 384 Mb, latest nVidia drivers (GLX 1.0-4191 and kernel 2.4.18-18.8.0).
Did anyone else experiment this strange behavior?
Were the examples sources modified in the 0.5 release (but I don't think so)?!?

[elias]

It works here on a similar configuration, rh8, geforce. It sounds like a z buffer problem. Could you take a snapshot of the demos?

- elias

[EgonOlsen]

It sounds like a z buffer problem.

Sounds more like a reversed backface-culling to me than a zbuffer problem...

Games published by our own members! Check 'em out!

[elias]

Maybe, but that's even moere weird when it's the same demo.

- elias

[fbi]

Maybe, but that's even moere weird when it's the same demo.

- elias

Yeah...it sounded more like a reversed backface-culling problem to me, too  
Anyway I'm gonna a take a snapshot of the "weird" demo as soon as I come back home...I'm currently at work.
I could also try to test what happens on another GeForce 2 in my laboratory...I'll give you more details as soon as possible.

[fbi]

Hi elias, I took two screenshots from one of the "weird" linux demos All the demos where you get a rotating cube with some textures on, behave basically the same!
It seems like we both are right: some faces seem to disappear cause of uncorrect normals but even some others seem to have wrong z-values in the depth buffer.  
You can see the shots at:

http://ant1.di.uniba.it/lwjgl/04 (the correct one from 0.4 version)
http://ant1.di.uniba.it/lwjgl/05 (the wrong one from 0.5 version)

Hope I gave some help...  

[elias]

That helped, kind of. There's definitely something wrong, but I still can't figure out exactly what. So I added a printf to the debug build in cvs that prints out the selected pixel format. CVS won't help you much I guess, so I also compiled it for rh 8 and uploaded the debug library to

http://odense.kollegienet.dk/~naur/liblwjgl_d.so.tar.bz2

Look for a line saying something like:

Pixel format chosen sizes: r = 8, g = 8, b = 8, a = 8, depth = 24, stencil = 8

We're especially interested in the depth size.

- elias