On March 19, 2015 11:40:40 PM Ilia Mirkin wrote: > On Thu, Mar 19, 2015 at 11:25 PM, Matthew Dawson <[email protected]> wrote: > > The expected value for draw-pixel-with-texture does not match what is > > actually written using the glDrawPixels function, causing the test to > > always fail. Change the expected value to what is wanted instead. > > > > This test fails regardless on my AMD cards, but if texturing is disabled > > in the test, it can now pass with this patch. The piglit results at > > http://people.freedesktop.org/~imirkin/nv10-comparison/nv17-2014-10-03-imi > > rkin/spec/!OpenGL%201.1/draw-pixel-with-texture.html suggest that this is > > a problem even with drivers that would pass with texturing enabled. > > Not sure that nouveau_vieux is quite the definition of perfection... > nv40, nv50 and nvc0 all fail with > > Observed: 0.039216 1.000000 0.000000 1.000000 > > so they'd keep on failing even with your change. No idea what's going > on with this test, but is that result reasonable? Sorry, I misread the spec/commit log over this code. I thought it was testing that the texture system doesn't affect glDrawPixels, but that is incorrect.
It appears the original code is correct, as I believe the test is assuming the texture value is multiplied component-wise with the value set for the given pixel. For some reason, nothing is doing this. Just saw your other email come in, so it seems we agree :). I'll have a look into mesa, see if I can sort this out. Thanks for the review, -- Matthew
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ Piglit mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/piglit
