On Sat, Jul 26, 2008 at 3:39 PM, H. Verbeet hverbeet@gmail.com wrote:
2008/7/26 Vitaliy Margolen wine-devel@kievinfo.com:
So if we need to create an fbo for each thread, does that mean that everything needs to be rebound to it on the context switch?
It would mostly mean that apply_fbo_state() would need to track things per-context rather than per-device. I guess that would make things a bit more like eg. pbuffers. We'd also need a per-context src_fbo and dst_fbo for things like stretch_rect_fbo(). Unless I misunderstood, this sounds more like a driver bug than an issue with the way we use FBOs though.
You could try the attached hack patch to see if it fixes the problem you're seeing. It completely ignores FBO cleanup, but it should avoid the driver bug.