2010/11/23 Stefan Dösinger stefan@codeweavers.com:
Am Dienstag 23 November 2010, 16:52:09 schrieb Henri Verbeet:
Looks like a flaw in the d3dx9_36 test, it isn't taking the POW2 cap into account. I would have expected that to fail the other way around though, cards that can't do npot textures should be pretty rare.
If the test was broken I'd expect it to fail on Windows, but it doesn't:
http://test.winehq.org/data/892502479c3acf64257e19ec4aac113024642f4d/xp_Wyld... wxp-32b-ie6/d3dx9_36:surface.html
According to http://test.winehq.org/data/892502479c3acf64257e19ec4aac113024642f4d/xp_Wyld... wxp-32b-ie6/d3d9:visual.html
from the same machine this card has a Geforce 8600 and is properly set up. I'd be pretty surprised if this card doesn't support NP2 cubemaps. I don't have such a card, but my geforce9600 supports NP2 cubemaps(which is how I became aware of the caps difference)
The test failing for AJ on his machine actually fails on that same Windows machine you looked at: http://test.winehq.org/data/892502479c3acf64257e19ec4aac113024642f4d/xp_Wyld... So, yes, it seems like the test is broken. I added it recently, didn't notice the failures before. I'll take a look.