Bug 241027
Summary: | Flakes in WebGL conformance[2]/textures/image_bitmap_from_image_bitmap/ tests | ||
---|---|---|---|
Product: | WebKit | Reporter: | Kenneth Russell <kbr> |
Component: | WebGL | Assignee: | Kenneth Russell <kbr> |
Status: | RESOLVED FIXED | ||
Severity: | Normal | CC: | dino, Hironori.Fujii, kbr, kkinnunen, webkit-bug-importer |
Priority: | P2 | Keywords: | InRadar |
Version: | WebKit Nightly Build | ||
Hardware: | Unspecified | ||
OS: | Unspecified | ||
Bug Depends on: | |||
Bug Blocks: | 189641, 189672, 241145 |
Kenneth Russell
Flakes are seen in the conformance[2]/textures/image_bitmap_from_image_bitmap/ portion of the WebGL 1.0/2.0 conformance tests. Here's an example test build of a PR which didn't touch this code at all:
https://ews-build.webkit.org/#/builders/70/builds/8221
(from https://github.com/WebKit/WebKit/pull/1063 )
Per the test history from one of these tests, webgl/2.0.0/conformance/textures/image_bitmap_from_image_bitmap/tex-2d-rgba-rgba-unsigned_short_5_5_5_1.html :
https://results.webkit.org/?suite=layout-tests&test=webgl%2F2.0.0%2Fconformance%2Ftextures%2Fimage_bitmap_from_image_bitmap%2Ftex-2d-rgba-rgba-unsigned_short_5_5_5_1.html
flakes are seen on the "Big Sur E wk2 Release on Macmini9,1 with arm64" bot as far back as 5/20/2022:
https://build.webkit.org/#/builders/104/builds/8259
and on the "Big Sur E wk2 Release on Macmini8,1 with x86_64" bot also as far back as 5/20/2022:
https://build.webkit.org/#/builders/70/builds/9383
The symptom is that the first couple of ImageBitmaps created from ImageBitmaps, and then uploaded to textures, are black instead of solid red:
https://ews-build.webkit.org/#/builders/70/builds/8221
https://ews-build.s3-us-west-2.amazonaws.com/macOS-BigSur-Release-WK2-Tests-EWS/2e874539-8221/webgl/2.0.0/conformance/textures/image_bitmap_from_image_bitmap/tex-2d-rgba-rgba-unsigned_short_5_5_5_1-pretty-diff.html
This is more likely a bug in the ImageBitmap implementation rather than a bug in ANGLE.
Attachments | ||
---|---|---|
Add attachment proposed patch, testcase, etc. |
Kenneth Russell
Actually these all seem to only be happening in the webgl/2.0.0/ versions of the tests, which makes suppression simpler.
Kenneth Russell
Pull request: https://github.com/WebKit/WebKit/pull/1107
EWS
Test gardening commit r294961 (251065@main): <https://commits.webkit.org/251065@main>
Reviewed commits have been landed. Closing PR #1107 and removing active labels.
Fujii Hironori
Commit-Queue bot is frequently blocking due to this flakiness now.
https://ews-build.webkit.org/#/builders/28
Why not marking in the top TestExpectations rather than webgl/TestExpectations?
Kenneth Russell
(In reply to Fujii Hironori from comment #4)
> Commit-Queue bot is frequently blocking due to this flakiness now.
> https://ews-build.webkit.org/#/builders/28
>
> Why not marking in the top TestExpectations rather than
> webgl/TestExpectations?
How do these suppressions behave differently if they are in one or the other file? I'll move them to the top level in Bug 241145.
Fujii Hironori
I didn't know about webgl/TestExpectations. According to but#204327 comment#0, it should have just a single line. `webgl [ Pass Slow ]`
Fujii Hironori
See bug#204327 comment#0
Radar WebKit Bug Importer
<rdar://problem/94356218>
Kimmo Kinnunen
Further work on this particular bug description seems unactionable. Let's file specific bugs on the further issues.