New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 682753 link

Starred by 3 users

Issue metadata

Status: Fixed
Owner:
Closed: Mar 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 3
Type: Bug


Sign in to add a comment

Pass WebGL 2.0.0 conformance suite on Pixel

Project Member Reported by zmo@chromium.org, Jan 19 2017

Issue description

conformance/extensions/webgl-compressed-texture-atc.html
conformance/extensions/webgl-compressed-texture-etc.html
conformance/glsl/bugs/sampler-struct-function-arg.html
conformance/glsl/bugs/sequence-operator-evaluation-order.html
conformance/textures/misc/copy-tex-sub-image-2d-partial-texture.html
conformance/textures/misc/tex-image-and-sub-image-2d-with-array-buffer-view.html
conformance/textures/canvas/*
conformance/textures/image/*
conformance/textures/image_data/*
conformance/textures/svg_image/*
conformance/textures/webgl_canvas/*
conformance/textures/image_bitmap_from_image_data/*
conformance/textures/image_bitmap_from_image/*
conformance/textures/image_bitmap_from_video/* (TIMEOUT)
conformance/textures/image_bitmap_from_canvas/*
conformance/textures/image_bitmap_from_blob/*
conformance/textures/image_bitmap_from_image_bitmap/*

deqp/data/gles3/shaders/arrays.html
deqp/data/gles3/shaders/constant_expressions.html
deqp/functional/gles3/fragmentoutput/*
deqp/functional/gles3/fbocompleteness.html
deqp/functional/gles3/integerstatequery.html
deqp/functional/gles3/multisample.html
deqp/functional/gles3/negativetextureapi.html
deqp/functional/gles3/shadercommonfunction.html
deqp/functional/gles3/shaderderivate_dfdy.html
deqp/functional/gles3/shaderprecision_float.html

conformance2/glsl3/array-in-complex-expression.html
conformance2/glsl3/compare-structs-containing-arrays.html (CRASH)
conformance2/glsl3/short-circuiting-in-loop-condition.html
conformance2/glsl3/vector-dynamic-indexing.html
conformance2/programs/gl-get-frag-data-location.html
conformance2/reading/read-pixels-from-fbo-test.html
conformance2/renderbuffers/readbuffer.html
conformance2/rendering/blitframebuffer-filter-outofbounds.html
conformance2/rendering/blitframebuffer-test.html
conformance2/rendering/clear-func-buffer-type-match.html
conformance2/rendering/draw-buffers.html
conformance2/rendering/framebuffer-completeness-unaffected.html
conformance2/rendering/framebuffer-unsupported.html
conformance2/rendering/uniform-block-buffer-size.html
conformance2/state/gl-get-calls.html
conformance2/state/gl-object-get-calls.html
conformance2/textures/misc/tex-storage-2d.html
conformance2/textures/canvas/*
conformance2/textures/image/*
conformance2/textures/image_data/*
conformance2/textures/svg_image/*
conformance2/textures/webgl_canvas/*
conformance2/textures/image_bitmap_from_image_data/*
conformance2/textures/image_bitmap_from_image/*
conformance2/textures/image_bitmap_from_video/* (TIMEOUT)
conformance2/textures/image_bitmap_from_canvas/*
conformance2/textures/image_bitmap_from_blob/*
conformance2/textures/image_bitmap_from_image_bitmap/*

Running with Chrome ToT build with the following workarounds removed:
   disable_chromium_framebuffer_multisample
   disable_depth_texture
GPU: Qualcomm Adreno (TM) 530 ES 3.2 V@145.0
in about:flags, disable "Gesture requirement for media playback"
 

Comment 1 by zmo@chromium.org, Jan 27 2017

Blocking: 673426

Comment 2 by zmo@chromium.org, Feb 3 2017

Owner: kainino@chromium.org
Status: Assigned (was: Available)
Blockedon: 688632
The max_color_attachments=1 issue is not a workaround as I had expected. Instead, it's just a logical error in feature detection. See  Issue 688632 .
conformance/extensions/webgl-compressed-texture-atc.html is fixed in
https://github.com/KhronosGroup/WebGL/pull/2285
conformance/extensions/webgl-compressed-texture-etc.html looks like a probable driver bug.

compressedTexImage2D with COMPRESSED_R11_EAC throws INVALID_VALUE because imageSize is 0 instead of 8 (which is the correct size of a 4x4 block). However, if this is fixed in the test, the error value changes from INVALID_VALUE to NO_ERROR, instead of INVALID_OPERATION.
Blockedon: 689786
Mostly finished going through the bug list. Results are pretty encouraging.
Blockedon: 682075
Blocking: 695673
Blocking: 695677
Blocking: 695679
Blocking: 695680
Blocking: 695683
Blocking: 695685
Blocking: 695687
Blocking: 695689
Blocking: 695692
Blocking: 695695
Blocking: 695742
Project Member

Comment 22 by bugdroid1@chromium.org, Feb 24 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/5fa975c1785e2a47c96471691447846d1be8b51d

commit 5fa975c1785e2a47c96471691447846d1be8b51d
Author: kainino <kainino@chromium.org>
Date: Fri Feb 24 04:08:32 2017

Disable some old Qualcomm workarounds on new Android

Workarounds disable_depth_texture and disable_chromium_framebuffer_multisample were added for old Qualcomm devices but work correctly on newer devices and are needed for WebGL 2.

This also un-disables GL_OES_depth_texture on these platforms.

Test note: The change in availability of MSAA causes the tests to access a different (currently non-existent) set of reference images on Qualcomm bots (_msaa.png to _non_msaa.png). This CL effectively creates a new configuration - which is why the tests do not need revisions incremented.

BUG= 682075 , 682753 
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/2707623002
Cr-Commit-Position: refs/heads/master@{#452748}

[modify] https://crrev.com/5fa975c1785e2a47c96471691447846d1be8b51d/content/test/gpu/gpu_tests/pixel_expectations.py
[modify] https://crrev.com/5fa975c1785e2a47c96471691447846d1be8b51d/gpu/config/gpu_driver_bug_list_json.cc
[modify] https://crrev.com/5fa975c1785e2a47c96471691447846d1be8b51d/gpu/config/software_rendering_list_json.cc

Labels: -Pri-1 Pri-3
By fixing a few test bugs, disabling many more tests due to driver bugs, and disabling old gpu driver bug workarounds, I was able to complete a 2.0.0 conformance run on Pixel at 100%.

However, due to a copy-paste limitation on Android, the results I saved are truncated, so I'll need to re-run when I'm less busy.

Fix for copy-paste limitation:
https://github.com/KhronosGroup/WebGL/pull/2319

Test changes:
https://github.com/KhronosGroup/WebGL/pull/2304
https://github.com/KhronosGroup/WebGL/pull/2306
https://github.com/KhronosGroup/WebGL/pull/2318
https://github.com/KhronosGroup/WebGL/pull/2312

Disabled tests:
https://github.com/KhronosGroup/WebGL/pull/2311
https://github.com/KhronosGroup/WebGL/pull/2316
https://github.com/KhronosGroup/WebGL/pull/2305

Downgrading to P3 as re-capturing the results is the only task left.
Saw a failure of
deqp/functional/gles3/sync.html
> failed: fence_sync.client_wait_sync_finish: Fail
Maybe it's flaky? (Pretty sure this has passed before.)
Status: Fixed (was: Assigned)
I got a 100% run but sync.html is still flaky. Will require some investigation.
Blocking: 699763
Status: Assigned (was: Fixed)
I should actually capture our final results after M58 goes to beta in about 1 week.
> deqp/functional/gles3/sync.html
Believe this is a bug, fixed in
https://github.com/KhronosGroup/WebGL/pull/2333
Blocking: -699763
Blockedon: 699763
Renderer process crash (probably not reproducible)
deqp/functional/gles3/fbomultisample.2_samples.html

Failure (don't know why I'm seeing new failures)
deqp/functional/gles3/fboinvalidate/format_00.html: invalidate.format.r8ui
deqp/functional/gles3/fboinvalidate/format_01.html: invalidate.format.r16i
deqp/functional/gles3/fboinvalidate/format_00.html: several tests flaky
deqp/functional/gles3/fboinvalidate/format_01.html: several tests flaky

These are flaky and it seems like it's probably a V8 problem. It doesn't repro on Canary, so it has probably been fixed. It also doesn't repro on one device, which is a total mystery.

Comment 33 by kbr@chromium.org, Mar 21 2017

Cc: jkummerow@chromium.org hablich@chromium.org bmeu...@chromium.org
+hablich, bmeurer, jkummerow: FYI regarding these flaky failures.

Blockedon: v8:6136
Project Member

Comment 35 by bugdroid1@chromium.org, Mar 23 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/4ea0b4441fd5f3a083fe3d93716721e80603356e

commit 4ea0b4441fd5f3a083fe3d93716721e80603356e
Author: kainino <kainino@chromium.org>
Date: Thu Mar 23 21:36:20 2017

Clean up pixel test failure expectations

from http://crrev.com/2707623002

BUG= 682075 ,  682753 
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/2706243005
Cr-Commit-Position: refs/heads/master@{#459232}

[modify] https://crrev.com/4ea0b4441fd5f3a083fe3d93716721e80603356e/content/test/gpu/gpu_tests/pixel_expectations.py

Blockedon: v8:6151
Blockedon: -v8:6151
Blocking: v8:6151
Status: Fixed (was: Assigned)
Closing via https://github.com/KhronosGroup/WebGLConformanceSubmissions/pull/2

I worked around  issue v8:6151  locally.
The following tests were seen to flake on this configuration for unknown reasons:

- conformance2/rendering/uniform-block-buffer-size.html
despite the workaround in https://github.com/KhronosGroup/WebGL/pull/2311
- conformance2/samplers/samplers.html

Sign in to add a comment