Bug 240830 – REGRESSION (ANGLE update to 2022-04-26?): [ Monterey wk2 ] webgl/2.0.0/conformance/attribs/gl-vertexattribpointer-offsets.html is a flaky timeout

• Status: NEW
• Component: WebGL

[Karl Rackler]

Description:
webgl/2.0.0/conformance/attribs/gl-vertexattribpointer-offsets.html

The first failure that I saw on the dashboard for Monterey was on 4/27/2022 at r293519.

History:
https://results.webkit.org/?suite=layout-tests&test=webgl%2F2.0.0%2Fconformance%2Fattribs%2Fgl-vertexattribpointer-offsets.html&platform=mac&architecture=x86_64&flavor=wk2&limit=50000&version_name=Monterey

Diff:
No Diff.
https://build.webkit.org/results/Apple-Monterey-Debug-WK2-Tests/250050@main%20(1596)/webgl/2.0.0/conformance/attribs/gl-vertexattribpointer-offsets-diff.txt

[Radar WebKit Bug Importer]

<rdar://problem/93786598>

[Karl Rackler]

I have marked this test as a flaky timeout while this issue is investigated.

[Karl Rackler]

Pull request: https://github.com/WebKit/WebKit/pull/946

[EWS]

Test gardening commit r294687 (250893@main): <https://commits.webkit.org/250893@main>

Reviewed commits have been landed. Closing PR #946 and removing active labels.

[Ryan Haddad]

The test had been running in 19 seconds or fewer up until the result at 250045@main, then it started taking 

https://commits.webkit.org/250044@main was a update of ANGLE, so it is likely the cause of the slowdown.

[Kenneth Russell]

This test runs pretty much instantaneously on macOS 12.4 on an M1 MBP in both Safari Tech Preview and Chrome Canary, both with ANGLE's Metal backend.

What could be the essential difference between WebKit's layout test harness and the browser causing this test to be so slow?

[Ryan Haddad]

(In reply to Ryan Haddad from comment #5)
> The test had been running in 19 seconds or fewer up until the result at
> 250045@main, then it started taking 
(sorry, saved before finishing my sentence)
... then it started taking 20+ seconds and intermittently timing out.

(In reply to Kenneth Russell from comment #6)
at could be the essential difference between WebKit's layout test harness
> and the browser causing this test to be so slow?
This I'm not sure, maybe Kimmo or others can comment.