gxquickly / angleproject

Automatically exported from code.google.com/p/angleproject
Other
0 stars 0 forks source link

Win7 ATI FYI bot failing d3d9-backed conformance tests #839

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago
ATI and Intel bots were recently added to the FYI waterfall to give d3d9 
coverage on a greater range of hardware. The ATI bot is failing WebGL 
conformance tests for ext_sRGB and webgl_depth_texture when run using d3d9. 

Original issue reported on code.google.com by shannonw...@chromium.org on 21 Nov 2014 at 10:01

GoogleCodeExporter commented 9 years ago
Suppressing these failures in https://codereview.chromium.org/751703002 . 
Unfortunately it's not currently possible to suppress them only with the D3D9 
backend, so this suppression will introduce the possibility of regressions on 
AMD GPUs even with the default D3D11 backend. It would be better to either fix 
the underlying failures or stop advertising EXT_sRGB and EXT_depth_texture in 
the D3D9 backend when on AMD GPUs.

Original comment by kbr@chromium.org on 21 Nov 2014 at 10:38

GoogleCodeExporter commented 9 years ago
Snagging, the plan is to stop advertising the extensions on AMD D3D9.

Original comment by jmad...@chromium.org on 26 Nov 2014 at 4:47

GoogleCodeExporter commented 9 years ago

Original comment by jmad...@chromium.org on 26 Nov 2014 at 4:48

GoogleCodeExporter commented 9 years ago
https://chromium-review.googlesource.com/#/c/232062/

Original comment by jmad...@chromium.org on 26 Nov 2014 at 7:11

GoogleCodeExporter commented 9 years ago
Handing off to Geoff for the depth textures part.

Original comment by jmad...@chromium.org on 26 Nov 2014 at 7:11

GoogleCodeExporter commented 9 years ago
The texparameter-test always fails for me on AMD D3D9 with ToT ANGLE. This was 
the test that started failing with the fix I posted above. Unsure how to 
proceed.

Original comment by jmad...@chromium.org on 3 Dec 2014 at 7:47

GoogleCodeExporter commented 9 years ago
(Note: only fails if you run the entire CTS from start to end, skipping the 
sRGB test which crashes. If ran individually, the texparameter-test passes.)

Original comment by jmad...@chromium.org on 3 Dec 2014 at 7:48

GoogleCodeExporter commented 9 years ago
Something's weird with that texparameter-test failure. It looks like maybe 
dither state is not being preserved correctly across contexts in some 
situations? See the attached output. Once it starts to fail it keeps failing in 
all new renderer processes, indicating there's some global state in ANGLE that 
has been changed.

I'm not sure it's worth debugging this if it works in D3D11 on all GPUs. I 
think we should improve the GPU test expectations so we can ignore D3D9 
specific failures.

Original comment by kbr@chromium.org on 4 Dec 2014 at 1:15

Attachments:

GoogleCodeExporter commented 9 years ago
Ken, the test in question checks the texture "wrap" mode, as in "Repeat", 
"Mirror", "Clamp to Edge". I'm not sure exactly if ANGLE's messing something up 
or if the driver is, because I'm pretty sure the test passes on every other 
configuration. I don't recall it failing in any of my other testing.

If it turns out to be a hard-to-pin thing that only happens in AMD D3D9, I 
might recommend we fix the sRGB bug and disable the texparameter test instead, 
since at least we know we'd be fixing one bug.

Original comment by jmad...@chromium.org on 4 Dec 2014 at 3:37

GoogleCodeExporter commented 9 years ago
Stealing back!

Original comment by jmad...@chromium.org on 4 Dec 2014 at 9:27

GoogleCodeExporter commented 9 years ago
Suppressing the texparameter-test.html test for D3D9 only is under review in 
https://codereview.chromium.org/751763006 . Let's just leave that suppression 
in place after the sRGB and depth-texture test failures are gone.

Original comment by kbr@chromium.org on 6 Dec 2014 at 1:03

GoogleCodeExporter commented 9 years ago
https://codereview.chromium.org/751763006 just landed. Is ANGLE at a point 
where we can remove the suppressions for the EXT_sRGB and WEBGL_depth_texture 
failures?

Original comment by kbr@chromium.org on 9 Dec 2014 at 12:35

GoogleCodeExporter commented 9 years ago
[deleted comment]
GoogleCodeExporter commented 9 years ago
Yes Ken, we should be able to remove the suppressions. I'll make a CL for that 
if you'd like.

Original comment by jmad...@chromium.org on 9 Dec 2014 at 3:04

GoogleCodeExporter commented 9 years ago
Jamie, yes, please do. Thanks.

Original comment by kbr@chromium.org on 11 Dec 2014 at 1:26

GoogleCodeExporter commented 9 years ago
Jamie removed the suppressions in https://codereview.chromium.org/792863002/ . 
The remaining one is intended to be left in place for the time being so closing 
this as fixed. Jamie, please reopen if I've closed this in error.

Original comment by kbr@chromium.org on 15 Dec 2014 at 11:30