Bug 295599 – [ iOS Release ] http/tests/webcodecs/audio-decoder-callbacks-do-not-leak.html is a flakey text failure

• Component: Media
• Status: NEW

[Bryan Medina]

http/tests/webcodecs/audio-decoder-callbacks-do-not-leak.html

is a flakey text failure occurring on iOS Release

HISTORY:
https://results.webkit.org/?suite=layout-tests&test=http%2Ftests%2Fwebcodecs%2Faudio-decoder-callbacks-do-not-leak.html

TEXT DIFF:
--- /Volumes/Data/worker/Apple-iOS-18-Simulator-Release-WK2-Tests/build/layout-test-results/http/tests/webcodecs/audio-decoder-callbacks-do-not-leak-expected.txt
+++ /Volumes/Data/worker/Apple-iOS-18-Simulator-Release-WK2-Tests/build/layout-test-results/http/tests/webcodecs/audio-decoder-callbacks-do-not-leak-actual.txt
@@ -7,4 +7,12 @@
 PASS successfullyParsed is true
 
 TEST COMPLETE
+PASS The iframe document didn't leak.
+PASS successfullyParsed is true
 
+TEST COMPLETE
+PASS The iframe document didn't leak.
+PASS successfullyParsed is true
+
+TEST COMPLETE
+

TEXT DIFF URL:
https://build.webkit.org/results/Apple-iOS-18-Simulator-Release-WK2-Tests/297115@main%20(2510)/http/tests/webcodecs/audio-decoder-callbacks-do-not-leak-diff.txt

[Radar WebKit Bug Importer]

<rdar://problem/155367858>

[Bryan Medina]

Pull request: https://github.com/WebKit/WebKit/pull/47734

[Bryan Medina]

I was not able to reproduce the text failure locally on iOS 18 simulator using these following commands:

run-webkit-tests http/tests/webcodecs/audio-decoder-callbacks-do-not-leak.html --ios-simulator --iterations=10 -f

[Alexey Proskuryakov]

This looks like it must be a test bug, multiple frames are racing with test completion.

[Ryan Reno]

See also https://bugs.webkit.org/show_bug.cgi?id=295751

Frames racing is kind of the point. I found it makes false negatives less likely. We bail as soon as we get a passing frame. But we get two passing frames before the test runner gets a chance to clean up sometimes.

The test can probably be improved to prevent this kind of behavior but on the balance I'd rather see two passes than zero!

[Alexey Proskuryakov]

The gardening PR will effectively disable the test, so yes, it's important to fix it up.

[EWS]

Test gardening commit 298776@main (03f43d5b4f05): <https://commits.webkit.org/298776@main>

Reviewed commits have been landed. Closing PR #47734 and removing active labels.