Created attachment 8685541 [details] [diff] [review] Randall's patch While working on getting Fennec passing tests with the C++ APZ Randall found what appears to be a bug in WrapTexCoord. Specifically when called with a negative whole number it returns 1.0 instead of 0.0. Patch attached, but after mxr'ing I see that there's another copy of this function in gfx/gl/DecomposeIntoNoRepeatTriangles.cpp which probably needs updating as well. Also I'm not sure why the code uses fmodf when it could be doing |return v - floorf(v)| which I think returns the same thing.
Created attachment 8686746 [details] [diff] [review] 0001-Bug-1223486-WrapTexCoord-1.0-returns-1.0-instead-of-0.0-15111212-0c2648e.patch
On IRC Bas gave an r+ to the floorf version that I suggested in comment 0, I'll steal this bug and update the patch.
Created attachment 8689562 [details] [diff] [review] Patch v2 Updated patch, carrying r+ from Bas on IRC