53868 2011-02-06 02:42:03 -0800 fast/notifications/notifications-document-close-crash.html fails and flakes after r77738 2024-02-18 16:49:28 -0800 1 1 1 Unclassified WebKit Tools / Tests 528+ (Nightly build) All All RESOLVED CONFIGURATION CHANGED P2 Normal --- 1 robert webkit-unassigned abarth ahmad.saleem792 ap eric inferno japhet mihaip ossy webkit.review.bot zmo oldest_to_newest 346260 0 robert 2011-02-06 02:42:03 -0800 Fails after http://trac.webkit.org/changeset/77738 -CONSOLE MESSAGE: line 36: INVALID_STATE_ERR: DOM Exception 11: An attempt was made to use an object that is not, or is no longer, usable. +CONSOLE MESSAGE: line 36: TypeError: Type error PASS Failure is because of: JSValue JSNotificationCenter::requestPermission(ExecState* exec) { ScriptExecutionContext* context = impl()->scriptExecutionContext(); // Make sure that script execution context is valid. if (!context) { setDOMException(exec, INVALID_STATE_ERR); return jsUndefined(); } <..> if (!exec->argument(0).isObject()) return throwTypeError(exec); } 'context' above used to be a local copy of ActiveDomObject::scriptExecutionContext() but is now the scriptExectionContext() (an upcast of Document) for the page's Document object. On Qt, apparently unlike other platforms, this context object is not destroyed by the time the test calls requestPermission(), so instead Qt fails in the above function at 'if (!exec->argument(0).isObject())' instead. Not clear to me why this is happening only for Qt, so skipping for now. 346261 1 robert 2011-02-06 02:47:19 -0800 Committed r77764: <http://trac.webkit.org/changeset/77764> 346262 2 robert 2011-02-06 02:49:14 -0800 Reopening for fix. 346271 3 webkit.review.bot 2011-02-06 05:46:17 -0800 http://trac.webkit.org/changeset/77764 might have broken Leopard Intel Release (Tests) The following tests are not passing: platform/mac/fast/AppleScript/date.html 349371 4 robert 2011-02-10 14:14:19 -0800 This test is flaky on Chromium now. Until r77738 the test could depend on the ScriptExecutionContext being null because NotificationCenter held a local copy which got nulled on NotificationCenter::disconnectFrame(). Now NotificationCenter will return a valid ScriptExecutionContext until the Document/ScriptExecutionContext the DomWindow originally associated it with is destroyed. I believe the reason Qt is now failing on 'if (!exec->argument(0).isObject())' instead is because 'script()->clearScriptObjects();' has been called in 'void Frame::pageDestroyed()' on the way to calling NotificationCenter::disconnectFrame(). So any crash is still safely avoided. Unless there's a good reason to expect the Document/ScriptExecutionContext to be destroyed by the time notification.RequestPermission() is called in the test the new result is valid. Can anyone point to why it should be destroyed by that time? 357595 5 mihaip 2011-02-24 17:25:18 -0800 Re-titling bug since the crash affects Chromium too (and perhaps other platforms): http://test-results.appspot.com/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=fast%2Fnotifications%2Fnotifications-document-close-crash.html&showExpectations=true&group=%40ToT%20-%20chromium.org 2014981 6 ahmad.saleem792 2024-02-18 16:49:28 -0800 Test is gone with - https://github.com/WebKit/WebKit/commit/34d6c0d0c6a163321646de3e58e9e2dee5f9d739 Nothing to do it - it was for legacy Notifications.