From: jan p. springer (jsd++at++igroup.org)
Date: 12/17/2002 13:49:48
Jason Daly wrote:
>>is it possible that you see a multiprocess problem rather then a glx problem?
>>have you tried w/ APPCULLDRAW?
>>
>
>
> Good thought. I switched to APPCULLDRAW, and now when I force a stereo
> visual, it doesn't quit with SIGCHLD, but when I try to render in
> stereo, I seem to only get the right eye. I should mention that the
> same code (single- or multiprocessed) works fine on an Octane MXE.
some things off the top of my head:
- what nvidia driver version? i remember the 23.13/4 being able to supply
a non-null pointer for multi-sampling but locking up the whole system
when performer used it. since then we avoid them in our code, so i can't
say for sure if that still may a problem source.
- switching to APPCAULLDRAW and then not getting any signal usually would
mean there's sth. missing in the arena, e.g. a shared variable setup in
APP but also needed in CULL/DRAW. from my experience the linux
implementation is far more unforgiving on that then the irix one
- the same goes for doing gl related calls from a non-gfx context. under
APPCULLDRAW that's no problem, forking DRAW and doing gl calls in APP
will not work. unfortunately the only test for this is to run the whole
thing on at least two cpu's.
maybe you could strip down your app to the relevant parts (init, setup,
stereo draw) and maybe someone on the list has an idea where the dragons
are?
j.
-- +------------------------------------+--------------------------------------+ | jan p. springer | jsd++at++igroup.org | | computer science, buw.media.vrsys | jan.springer++at++medien.uni-weimar.de | +------------------------------------+--------------------------------------+
This archive was generated by hypermail 2b29 : Tue Dec 17 2002 - 13:51:26 PST