Bug 244134 – REGRESSION(253549@main): [ iOS ] 2X animations/(Layout tests) are constant failures

• Component: New Bugs
• Status: RESOLVED FIXED

[Hercules Hjalmarsson]

animations/no-style-recalc-during-accelerated-animation.html

Is a constant text failure on iOS ToT and since 253549@main.

HISTORY:
https://results.webkit.org/?suite=layout-tests&test=animations/no-style-recalc-during-accelerated-animation.html

DIFF:
@@ -1,2 +1,2 @@
 Got iteration event.
-PASS: saw five or fewer style recalcs during the animation.
+FAIL: saw 8 style recalcs during the animation, should only see five.

[Radar WebKit Bug Importer]

<rdar://problem/98894287>

[Hercules Hjalmarsson]

DIFF URL:
https://build.webkit.org/results/Apple-iOS-15-Simulator-Release-WK2-Tests/253549@main%20(4145)/animations/no-style-recalc-during-accelerated-animation-pretty-diff.html

[Hercules Hjalmarsson]

This issue can be bisected to 253549@main using command:

run-webkit-tests --ios-simulator --iterations=2 animations/no-style-recalc-during-accelerated-animation.html

[EWS]

Test gardening commit 253601@main (886100c335f4): <https://commits.webkit.org/253601@main>

Reviewed commits have been landed. Closing PR #3481 and removing active labels.

[Hercules Hjalmarsson]

animations/steps-transform-rendering-updates.html

Is a constant text failure on iOS ToT with the same regression point of 253549@main. However, with a different DIFF. They both appear to be related.

HISTORY:
https://results.webkit.org/?suite=layout-tests&suite=layout-tests&test=animations%2Fsteps-transform-rendering-updates.html&test=animations%2Fno-style-recalc-during-accelerated-animation.html

DIFF:
@@ -1,5 +1,5 @@
 PASS count is 0
-PASS count < 7 is true
+FAIL count < 7 should be true. Was false.
 PASS successfullyParsed is true

 TEST COMPLETE

[EWS]

Test gardening commit 253602@main (7199d3824d94): <https://commits.webkit.org/253602@main>

Reviewed commits have been landed. Closing PR #3482 and removing active labels.

[Alexey Proskuryakov]

The commit message claims that a few more rendering updates "should be benign". Not sure if the count is flaky now, or what the impact on power use is though.

[Simon Fraser (smfr)]

I've been investigating the causes of the extra updates on iOS; haven't nailed it down yet, but it's benign (a few more per animation, not the "update every frame" that the test was designed to detect).

[Simon Fraser (smfr)]

After lots of debugging, I think the iOS difference is due to differences in timer slop, and how we drive rendering updates with UI-side compositing. I did file bug 244276, and I'll adjust the test to call `setSpeculativeTilingDelayDisabledForTesting(true)` to make sure that doesn't introduce flakiness.

[Simon Fraser (smfr)]

Pull request: https://github.com/WebKit/WebKit/pull/3482

[Simon Fraser (smfr)]

Fixed via https://github.com/WebKit/WebKit/pull/3602