http://bugs.winehq.org/show_bug.cgi?id=18916
--- Comment #7 from Henri Verbeet hverbeet@gmail.com 2009-06-15 06:43:50 --- (In reply to comment #6)
(In reply to comment #5)
Could you add +d3d_surface to the trace?
Sure -- I'll do it this evening, and I'll report back on the effect of your revert patch. Do you want all four traces in the log? (It'll be big!)
Yes, but the last 200 lines or so should do. I think we're calling surface_upload_data() on the surface, but the data we're giving it isn't actually in GL_UNSIGNED_INT_24_8_EXT format, so it'll probably try to read past the end of the buffer.
Is it worth looking at the Thief 2 demo? I know that that isn't a regression (as far as I can tell my copy has never worked), but presumably it uses the same underlying technology ... so its failure might be quite instructive.
It might be worth looking at on its own, but it's probably not needed for this bug.