http://bugs.winehq.org/show_bug.cgi?id=22366
--- Comment #12 from Henri Verbeet hverbeet@gmail.com 2010-04-17 17:14:54 --- (In reply to comment #11)
There are a lot of different codepaths in wined3d. You do test certain apps and run the wine d3d tests (and also extend them) but you don't always cover all situations.
The d3d test coverage should be improved to prevent more regressions. Writing tests can be boring though. When I rewrite a piece I try to add some tests. In this specific case I wrote one for the ColorFill behavior and didn't look too much at the Clear behavior (I expected the current test to be enough there). Especially for trickier areas, I'll try to write more tests when I touch such an area.
More tests would be good of course, but you can hardly blame this one on the test coverage. I have the strong impression that this patch was created by simply copying around some code from color_fill_fbo() (compare the original http://www.winehq.org/pipermail/wine-patches/2010-April/087103.html with http://source.winehq.org/git/wine.git/?a=blob;f=dlls/wined3d/device.c;h=91e5...), without much understanding of what the involved code does.