[Dri-users] mga driver crashing in glTranslatef()

[Frank B. <fb...@fo...>]

Hi,

I'm a user of Gem[1], an OpenGL development framework built on top of
another software called Pure Data[2]. Since several weeks, I guess
since when I updated my xorg packages on my Debian "testing" machine,
Gem is crashing quite often, because somehow it is triggering a failed
assertion reported as: 

  ../../../../../src/mesa/tnl_dd/t_dd_vb.c:298: 
  mga_interp_extras: Assertion `VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat)' failed.

A gdb backtrace looks like this: 

  Program received signal SIGABRT, Aborted.
  [Switching to Thread -1210751296 (LWP 27280)]
  0xb7d80947 in raise () from /lib/tls/libc.so.6
  (gdb) bt
  #0  0xb7d80947 in raise () from /lib/tls/libc.so.6
  #1  0xb7d820c9 in abort () from /lib/tls/libc.so.6
  #2  0xb7d7a05f in __assert_fail () from /lib/tls/libc.so.6
  #3  0xb55cb3aa in mgaChooseVertexState () from /usr/lib/dri/mga_dri.so
  #4  0xb5655247 in _tnl_disassem_vba_insn () from /usr/lib/dri/mga_dri.so
  #5  0xb565635f in _tnl_disassem_vba_insn () from /usr/lib/dri/mga_dri.so
  #6  0xb56593b6 in _tnl_RenderClippedLine () from /usr/lib/dri/mga_dri.so
  #7  0xb564699f in _tnl_run_pipeline () from /usr/lib/dri/mga_dri.so
  #8  0xb558bf93 in mgaEmitHwStateLocked () from /usr/lib/dri/mga_dri.so
  #9  0xb567463e in _tnl_flush_vtx () from /usr/lib/dri/mga_dri.so
  #10 0xb566cffc in _tnl_FlushVertices () from /usr/lib/dri/mga_dri.so
  #11 0xb56062f2 in _mesa_Translatef () from /usr/lib/dri/mga_dri.so
  #12 0xb79f6f98 in glTranslatef () from /usr/lib/libGL.so.1
  #13 0xb7bb54a7 in translateXYZ::render () from /usr/lib/pd/extra/Gem.pd_linux
  #14 0xb7b4d495 in GemBase::gem_renderMess () from /usr/lib/pd/extra/Gem.pd_linux
  #15 0xb7b4d610 in GemBase::gem_MessCallback () from /usr/lib/pd/extra/Gem.pd_linux
  #16 0x0809fb68 in pd_typedmess (x=0xbf9da790, s=0x2, argc=2, argv=0xbf9da790) at m_class.c:665
  Previous frame inner to this frame (corrupt stack?)
  (gdb) quit
  The program is running.  Exit anyway? (y or n) y
  socket receive error: Connection reset by peer (104)


The Gem developers told me that this is a crash in the driver
happening inside glTranslatef(), one of the most basic OpenGL
functions. (See [3] for the respective gem-dev thread), so I guess,
this list is the place to ask about this, right? Does anyone here have
an idea, how to get rid of the crashes? 

Some more info: I'm using Debian testing with these packages: 

ii  libgl1-mesa-dev     6.5.1-0.2           A free implementation of
the OpenGL API -- GLX develop
ii  libgl1-mesa-dri     6.5.1-0.2           A free implementation of
the OpenGL API -- DRI modules
ii  libgl1-mesa-glx     6.5.1-0.2           A free implementation of
the OpenGL API -- GLX runtime
ii  libc6               2.3.6.ds1-7         GNU C Library: Shared
libraries

My kernel is self-built and a bit older I admit:

Linux fliwatut.scifi 2.6.16-vanilla #1 PREEMPT Wed Mar 29 23:20:54 CEST 2006 i686 GNU/Linux

Although it's an older kernel, this kernel was running perfectly with
Gem from about March until, hm, September. ;)

[1] http://gem.iem.at/
[2] http://puredata.info/
[3] http://lists.puredata.info/pipermail/gem-dev/2006-11/002425.html

Ciao
-- 
 Frank Barknecht                 _ ______footils.org_ __goto10.org__