Brian Paul wrote:
>
> I've got the Quake3 IHV test for Linux and tried running it with
> the latest Mesa code.
>
> It dies with a SEGV in do_texture_0() in pipeline.c, line 479.
>
> I've determined that the gl_transform_tab has a NULL pointer.
>
> gl_transform_tab[VB->CullFlag[0]!=0][(VB->TexCoordPtr[0])->size] = 0
>
> VB->CullFlag[0] = 0
> VB->TexCoordPtr[0]->size = 0
> VB->TexCoordPtr[0]->stride = 8
> ctx->TextureMatrix[0].type = 1 (MATRIX_IDENTITY)
>
> Keith, is that enough info to narrow down the bug? I'm unable to
> use gdb because when the crash occurs the X server is grabbed by
> Quake so mouse & keyboard are dead. If I don't use gdb the
> segfault terminates Quake and X server is unlocked. I've been
> using asserts and printf's to track it this far.
Size zero texcoords certainly sounds like a bug to me. The fact that
stride is 8 would indicate we're pulling data from a vertex array, so
yes this does narrow things down. Do you know if CVA is active at this
point? Can you try re-running the demo using a MESA_CONFIG which
disables the CVA extension, or a quake flag which does the same?
If Q3 is like Q1 and Q2, there should be a console command something
like '_windowed_mouse 0', or a command line arg '+set _windowed_mouse
0'... Invaluable for debugging.
> Keith, if you'd like I could try to arrange for you to get a
> copy of the Q3 IHV test. I can't redistribute the copy I got from
> zoid@id.
I think this is a good idea. What's the procedure?
> Another bug: user clip-planes are broke. Run the demos/logo program.
> Press the up-cursor key a few times until you see the problem.
>
OK, I'll look at this too. Incidentally I'm coding up fast-ish support
for polygon offset - I don't know if this is still a quake issue or not.
Keith
_______________________________________________
Mesa-dev maillist - [EMAIL PROTECTED]
http://lists.mesa3d.org/mailman/listinfo/mesa-dev
