Modern IDB: IDBTransaction::hasPendingActivity() was wrong, leading to premature GC of the wrapper Since the JS wrapper for an IDBTransaction might have been GC'ed before the transaction was done, javascript event callbacks to the transaction would fail. This would stall many tests, which would fail with a timeout. Because it's GC and therefore timing related, it showed up as flakiness. Known cases: https://bugs.webkit.org/show_bug.cgi?id=151219 https://bugs.webkit.org/show_bug.cgi?id=151364 https://bugs.webkit.org/show_bug.cgi?id=151389 https://bugs.webkit.org/show_bug.cgi?id=151448 However, the bug was fixed in https://trac.webkit.org/changeset/192687 So I suspect that all of these tests will un-flake going forward. I'll be closing all of those bugs as dupes of this one, and then once we're sure they've unflaked we can revert the TestExpectations changes made in each of those bugs.
*** Bug 151219 has been marked as a duplicate of this bug. ***
*** Bug 151364 has been marked as a duplicate of this bug. ***
*** Bug 151389 has been marked as a duplicate of this bug. ***
*** Bug 151448 has been marked as a duplicate of this bug. ***
(In reply to comment #0) > Known cases: > https://bugs.webkit.org/show_bug.cgi?id=151219 > https://bugs.webkit.org/show_bug.cgi?id=151364 > https://bugs.webkit.org/show_bug.cgi?id=151389 > https://bugs.webkit.org/show_bug.cgi?id=151448 > > However, the bug was fixed in https://trac.webkit.org/changeset/192687 All of the tests marked flaky in those 4 bugs have unflaked since 192687.
Created attachment 266950 [details] Patch v1
https://trac.webkit.org/changeset/193792