| Summary: | [ Monterey Release ] webaudio/offlineaudiocontext-gc.html is a consistent failure | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | WebKit | Reporter: | Karl Rackler <rackler> | ||||||
| Component: | New Bugs | Assignee: | Nobody <webkit-unassigned> | ||||||
| Status: | NEW --- | ||||||||
| Severity: | Normal | CC: | ap, cdumez, marcus.stenbeck, rniwa, webkit-bot-watchers-bugzilla, webkit-bug-importer | ||||||
| Priority: | P2 | Keywords: | InRadar | ||||||
| Version: | WebKit Nightly Build | ||||||||
| Hardware: | Unspecified | ||||||||
| OS: | Unspecified | ||||||||
| Attachments: |
|
||||||||
|
Description
Karl Rackler
2022-09-12 13:05:14 PDT
I have marked this test as a flaky failure while this issue is investigated. Test gardening commit 254404@main (ce0723f31884): <https://commits.webkit.org/254404@main> Reviewed commits have been landed. Closing PR #4276 and removing active labels. 254173@main is "EventTarget should hold EventTargetData in WeakPtrImpl", and this particular queue confidently blames it. But this is not happening on other wk1 bots, nor on internal bots, so something is super fishy. Created attachment 463326 [details]
Sample code showing the issue
I'm running into what I believe is GC failure for OfflineAudioContext on macOS Monterey 12.6 (21G115) using Safari Version 16.0 (17614.1.25.9.10, 17614) on a MacBookPro18,3 with M1 processor.
As you click the "process" button in the sample Chrome will consistently stay at the same processing time, while Safari steadily increases over time.
Created attachment 463359 [details]
Minimal OfflineAudioContext memory leak sample
The new attachment is a minimal reproduction showing the memory leaking. It seems related to the `suspend()` and `resume()` function of OfflineAudioContext. Looking at the History: https://results.webkit.org/?suite=layout-tests&test=webaudio%2Fofflineaudiocontext-gc.html&version_name=Monterey It appears the test began consistently passing after landing 254766@main. |