WebglConformance.conformance2_glsl3_array_complex_indexing failing with NVIDIA 364.19 driver |
||
Issue description
WebglConformance.conformance2_glsl3_array_complex_indexing is failing newly on NVIDIA's 364.19 driver. The test passed under the previous driver version on Chrome's GPU bots, which was 352.63.
The failure output is as follows:
[ RUN ] WebglConformance.conformance2_glsl3_array_complex_indexing
[16855:16855:0425/121453:INFO:CONSOLE(11)] "at (0, 0) expected: 0,255,0,255 was 255,0,0,255", source: (11)
[16855:16855:0425/121453:INFO:CONSOLE(11)] "FAIL at (0, 0) expected: 0,255,0,255 was 255,0,0,255", source: (11)
[16855:16855:0425/121453:INFO:CONSOLE(11)] "at (0, 0) expected: 0,255,0,255 was 255,0,0,255", source: (11)
[16855:16855:0425/121453:INFO:CONSOLE(11)] "FAIL at (0, 0) expected: 0,255,0,255 was 255,0,0,255", source: (11)
Traceback (most recent call last):
File "/tmp/isolated_runGR6pSa/third_party/catapult/telemetry/telemetry/internal/story_runner.py", line 84, in _RunStoryAndProcessErrorIfNeeded
state.RunStory(results)
File "/tmp/isolated_runGR6pSa/content/test/gpu/gpu_tests/gpu_test_base.py", line 122, in RunStory
RunStoryWithRetries(DesktopGpuSharedPageState, self, results)
File "/tmp/isolated_runGR6pSa/content/test/gpu/gpu_tests/gpu_test_base.py", line 72, in RunStoryWithRetries
super(cls, shared_page_state).RunStory(results)
File "/tmp/isolated_runGR6pSa/third_party/catapult/telemetry/telemetry/page/shared_page_state.py", line 326, in RunStory
self._current_page, self._current_tab, results)
File "/tmp/isolated_runGR6pSa/content/test/gpu/gpu_tests/webgl_conformance.py", line 87, in ValidateAndMeasurePage
raise page_test.Failure(_WebGLTestMessages(tab))
Failure: at (0, 0) expected: 0,255,0,255 was 255,0,0,255
FAIL at (0, 0) expected: 0,255,0,255 was 255,0,0,255
at (0, 0) expected: 0,255,0,255 was 255,0,0,255
FAIL at (0, 0) expected: 0,255,0,255 was 255,0,0,255
[ FAILED ] WebglConformance.conformance2_glsl3_array_complex_indexing (269 ms)
Failures can be seen on this machine:
https://build.chromium.org/p/chromium.gpu.fyi/builders/Linux%20Release%20(NVIDIA%20GeForce%20730)
in particular, the first failing build (until the logs expire):
https://build.chromium.org/p/chromium.gpu.fyi/builders/Linux%20Release%20%28NVIDIA%20GeForce%20730%29/builds/933
Full stdout attached.
Olli, do you think you could look into why this test is failing with the new driver? In the meantime I'm going to suppress the failure so that the bots can be upgraded to the new driver, which will fix other bugs.
,
Apr 26 2016
I reproduced the regression on my end and reported it to the GLSL compiler team. I also experimented with a workaround in ANGLE using some of the AST transformations that were already built to support HLSL output, but found out that the underlying issue is more severe than what the test immediately reveals, so this won't work. Fixing the driver seems like the only possibility.
,
Apr 26 2016
Thanks Olli for raising this issue with the driver team. Piers and I were just talking about the WebGL testing in NVIDIA's automated testing system. Should we try to get the native dEQP tests integrated in that harness? It's a bit early to add the WebGL 2.0 conformance tests there.
,
Apr 26 2016
I think more testing is almost always a good idea, but not sure what's the best way to get there. This particular test is not found from dEQP. And since there's some testing with dEQP being done on the Android side either way, that should also be reflected on desktop in many cases where driver code is shared between the two. If you're discussing this with Piers over email I'd appreciate being kept in the loop.
,
Apr 28 2016
We have a fix coming for the issue in the driver. No information yet about which driver release the fix will be in.
,
Apr 28 2016
Thanks very much Olli for tracking this down. Looped you into the conversation with Piers about NVIDIA's in-house automated testing of the WebGL conformance suite.
,
Jul 7 2016
This passes in the 367.27 driver that is on the bots. Closing as fixed.
,
Jul 8 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/039da87fc8fcf5124a454e20bad6c2e86f52709a commit 039da87fc8fcf5124a454e20bad6c2e86f52709a Author: cwallez <cwallez@chromium.org> Date: Fri Jul 08 01:00:58 2016 Update the WebGL2 CTS expectations after driver upgrade TBR=kbr@chromium.org TBR=zmo@chromium.org BUG= 606498 BUG=618447 BUG= 621178 BUG=622492 CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel Review-Url: https://codereview.chromium.org/2128313002 Cr-Commit-Position: refs/heads/master@{#404284} [modify] https://crrev.com/039da87fc8fcf5124a454e20bad6c2e86f52709a/content/test/gpu/gpu_tests/webgl2_conformance_expectations.py |
||
►
Sign in to add a comment |
||
Comment 1 by bugdroid1@chromium.org
, Apr 25 2016