On Fri, May 28, 2010 at 5:03 PM, Brian Paul <[email protected]> wrote:
> Still, the glTexImage() call should not be effected by the value of > GL_TEXTURE_MAX_LEVEL. I think there's a bug in the driver which this patch > is just working around. Yeah... It seems to be related to the driver falling back to placing the texture in regular malloc() memory instead of in a BO-backed miptree. I had previously described this as "failing to allocate" the texture memory, but that is a misleading way of putting it. Memory for the level is allocated, and the texture validation process copies it into a miptree. However, the CS checker terminates the program with an error about "texture bo too small". So far, my efforts to debug the process of migrating a texture in malloc() memory to the miptree have not met success. I think I can teach the driver not to use malloc() memory in this situation, but I am concerned that this is still just working around the underlying problem. > I wouldn't be suprised to find a GL app which calls glTexImage2D(level=X) > when GL_TEXTURE_MAX_LEVEL < X. Can you look into fixing the issue in the > driver? Other drivers seem OK in this regard. I had hoped to find time to look into it more deeply before now. I haven't forgotten, and will keep working on it. -- Will Dyson _______________________________________________ mesa-dev mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/mesa-dev
