Pixel test failures on Haswell GPUs on macOS with 10.12.4 |
|||||||
Issue descriptionAfter the upgrade from 10.12.2 to 10.12.4, the following pixel tests are observed to fail on the Mac Minis in the Swarming fleet: Pixel_OffscreenCanvas2DResizeOnWorker Pixel_OffscreenCanvasAccelerated2D Pixel_OffscreenCanvasAccelerated2DWorker Pixel_WebGLGreenTriangle_NonChromiumImage_NoAA_NoAlpha This is currently the affected machine: https://build.chromium.org/p/chromium.gpu.fyi/builders/Mac%20Experimental%20Release%20(Intel) Example build: https://luci-milo.appspot.com/buildbot/chromium.gpu.fyi/Mac%20Experimental%20Release%20%28Intel%29/8 Failure images: http://chromium-browser-gpu-tests.commondatastorage.googleapis.com/view_test_results.html?692751dc80943ee7094ee1027d9cdc8c8f3d5170_Mac_Experimental_Release_Intel__telemetry I was able to reproduce the Pixel_WebGLGreenTriangle_NonChromiumImage_NoAA_NoAlpha failure on a MacBook Air with Intel HD 5000 GPU with Chrome Canary by launching the browser with the flag --disable-webgl-image-chromium and navigating to content/test/data/gpu/pixel_webgl_noaa_noalpha.html . Since this seems to only affect the Haswell family of GPUs, marking these as failures and going to close this as WontFix.
,
May 3 2017
,
May 3 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/543cbcbb37d29652edb36adcb64e1dee09e4efe8 commit 543cbcbb37d29652edb36adcb64e1dee09e4efe8 Author: kbr <kbr@chromium.org> Date: Wed May 03 23:30:47 2017 Suppress regressions in Haswell graphics driver in macOS 10.12.4. BUG= 718183 , 718194 TBR=zmo@chromium.org CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel Review-Url: https://codereview.chromium.org/2858123002 Cr-Commit-Position: refs/heads/master@{#469200} [modify] https://crrev.com/543cbcbb37d29652edb36adcb64e1dee09e4efe8/content/test/gpu/gpu_tests/pixel_expectations.py [modify] https://crrev.com/543cbcbb37d29652edb36adcb64e1dee09e4efe8/content/test/gpu/gpu_tests/webgl2_conformance_expectations.py
,
May 3 2017
,
May 4 2017
,
Jun 19 2017
re-opening because there are still suppressions associated with this bug. The tests are passing on the affected bot: https://build.chromium.org/p/chromium.gpu.fyi/builders/Mac%20Experimental%20Release%20(Intel) Much to my surprise, the suppression are not affecting this bot, so I am not sure what is going on. The bot matches the suppression criteria, I checked. But for some reason the tests are all expected to pass acctording to the json output: https://luci-logdog.appspot.com/v/?s=chromium%2Fbb%2Fchromium.gpu.fyi%2FMac_Experimental_Release__Intel_%2F479%2F%2B%2Frecipes%2Fsteps%2Fpixel_test%2F0%2Flogs%2Fjson.output%2F0
,
Jun 19 2017
I think we could just remove the suppressions?
,
Jun 19 2017
The suppressions are taking effect. The JSON output, which is generated by typ, doesn't know about these test expectations, which are handled by the test harness. See for example this output from a recent run of the pixel tests: https://luci-logdog.appspot.com/v/?s=chromium%2Fbb%2Fchromium.gpu.fyi%2FMac_Experimental_Release__Intel_%2F481%2F%2B%2Frecipes%2Fsteps%2Fpixel_test%2F0%2Fstdout I double checked and the tests are passing now, but the reason is that the machines have been upgraded to 10.12.5 (as shown in the logs). This upgrade is happening under Issue 733089 and so we can't remove the suppressions until it's been completed for all of them.
,
Jul 6 2017
Now that Issue 733089 has been fixed, removing the unneeded suppressions in https://chromium-review.googlesource.com/560674 .
,
Jul 6 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/b4ac9f46202c9659c5b684856e847eaf10317bb9 commit b4ac9f46202c9659c5b684856e847eaf10317bb9 Author: Kenneth Russell <kbr@chromium.org> Date: Thu Jul 06 04:52:00 2017 Remove suppressions for macOS Intel graphics driver bugs. These were regressions in 10.12.4, fixed in 10.12.5. BUG= 718194 , 718183 TBR=junov@chromium.org Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel Change-Id: I19a0e30b87b7a72fde0a968ba2f91ecb2ec66d2f Reviewed-on: https://chromium-review.googlesource.com/560674 Reviewed-by: Kenneth Russell <kbr@chromium.org> Commit-Queue: Kenneth Russell <kbr@chromium.org> Cr-Commit-Position: refs/heads/master@{#484468} [modify] https://crrev.com/b4ac9f46202c9659c5b684856e847eaf10317bb9/content/test/gpu/gpu_tests/pixel_expectations.py [modify] https://crrev.com/b4ac9f46202c9659c5b684856e847eaf10317bb9/content/test/gpu/gpu_tests/webgl2_conformance_expectations.py
,
Jul 6 2017
This was a driver bug, so closing as WontFix. |
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by kbr@chromium.org
, May 3 2017