Bernd Kreimeier wrote:
> Certainly true. I am not sure how far along the LSB efforts
> are, though. If some requirement is put into the OpenGL ABI
> specs for now with the disclaimer that this to be replaced by
> generic LSB requirements when available, that would do fine.
> Even a bare recommendation would be good.
I think we agree that GL libs shouldn't register gratuitous signal
and atexit handlers.
While you didn't make it explicit, I think you've specifically been
talking about the case of Mesa on Glide with Voodoo1/2 hardware.
There's a bit of ugliness (I didn't write it :) in the Mesa/Glide
driver meant to deal with video passthrough restoration upon exit
or abnormal termination.
If that's the scope of the problem, I think it's better discussed
on the Mesa list (where I've posted a follow-up to Michael Vance's
post) than here.
-Brian
