I've _just_ started testing the latest Mesa code against the OpenGL conformance tests. There is one test which caused a segfault. The "divide by zero test" attempts to draw geometry with glVertex4f() with W = 0. This led to trouble. Keith W., could you try this and see if you can fix the problem? The second problem I found involved the coverage* test and glDrawBuffer. A recent patch for stereo support broke it. Stereo has never been fully supported so now I'm digging into the core code to support it right. The main problem is dealing with glDrawBuffer set to draw to two or even four color buffers at once. Should have my changes checked in by the end of the weekend. A third problem I found is that a glPushAttrib(ALL) / glPopAttrib() sequence didn't restore all the texture object state. I'm fixing that too. Otherwise, there are quite a few conformance subtests which Mesa fails, but has been failing all along. Since I can't disclose the conformance test code I'll have to isolate the problems and convey the problems to the rest of you as best I can. -Brian *The coverage test basically checks that all API functions accept and reject all the right enum values.
