On, Mon Apr 07, 2008, Lenard Lindstrom wrote: > Forget that last observation. The traceback doesn't show the arguments > correctly. I have traced the problem to clipaaline. When pts[0] is 0 and > pts[2] is 1 or pts[1] is 0 and pts[3] is 1 it copies pts[2] and pts[3] to > pts[0] and pts[1]. This causes a divided by 0.0 later in drawaaline which > leads to a memory access problem. It doesn't matter if the surface is 24 or > 32 bit. [...] That's the same I figured out half an hour ago :-). It's the Cohen-Sutherland clipping line implementation with slight adaptions and works well for normal pixel lines. I'm currently looking into the aaline clipping implementation and try to find a matching fix that takes the AA into account _and_ still provides a more or less reasonable result for those cases. Regards Marcus
Attachment:
pgplVXbIhME9U.pgp
Description: PGP signature