http://bugs.winehq.org/show_bug.cgi?id=10423
--- Comment #26 from Stefan Dösinger stefandoesinger@gmx.at 2011-08-22 09:18:11 CDT --- I can reproduce this bug(OSX 10.6, Geforce 9600). The game uses a HAL device with HARDWAREVERTEXPROCESSING | PUREDEVICE as flags, so there's no software vertex processing involved.
Testing various versions of the change it looks like it doesn't matter what the 0.0^0.0 case returns, as long as it is not NaN. But apparently that's what at least my GPU returns for pow(0.0, 0.0). Even Inf is OK. The ARB backend works as well, but I haven't checked what its LIT returns in this case.
I'll write a test for LIT that makes sure that the result in this special case is anything but NaN and fix the GLSL LIT implementation.