http://bugs.winehq.org/show_bug.cgi?id=20053
--- Comment #9 from Henri Verbeet hverbeet@gmail.com 2009-09-19 13:41:59 --- (In reply to comment #8)
If you can point me in the direction of what would need to be done to make this a proper patch, I can look into it.
A proper patch would generate +INF instead of NaN for rcp 0. My preferred way of fixing that would be by giving us some control over which floating point specials are generated, but I don't think the current GLSL spec has anything to allow that.