mozilla pre 1.7a w/ changes, but none to layout. we were shutting down > gklayout.dll!PresShell::DidDoReflow() Line 6349 + 0xe C++ gklayout.dll!PresShell::ProcessReflowCommands(int aInterruptible=1) Line 6493 C++ gklayout.dll!ReflowEvent::HandleEvent() Line 6256 C++ gklayout.dll!HandlePLEvent(ReflowEvent * aEvent=0x03ce8b10) Line 6270 C++ xpcom.dll!PL_HandleEvent(PLEvent * self=0x03ce8b10) Line 671 + 0xa C xpcom.dll!PL_ProcessPendingEvents(PLEventQueue * self=0x009bb2b0) Line 606 + 0x9 C xpcom.dll!_md_EventReceiverProc(HWND__ * hwnd=0x00660290, unsigned int uMsg=49702, unsigned int wParam=0, long lParam=10203824) Line 1412 + 0x9 C user32.dll!_InternalCallWinProc@20() + 0x1b user32.dll!_UserCallWinProcCheckWow@32() + 0xb7 user32.dll!_DispatchMessageWorker@8() + 0xd8 user32.dll!_DispatchMessageW@4() + 0xb gkwidget.dll!nsAppShell::Run() Line 135 C++ appshell.dll!nsAppShellService::Run() Line 484 C++ mozilla.exe!main1(int argc=1, char * * argv=0x002a4df8, nsISupports * nativeApp=0x0097af80) Line 1291 + 0x20 C++ mozilla.exe!main(int argc=1, char * * argv=0x002a4df8) Line 1678 + 0x25 C++ mozilla.exe!WinMain(HINSTANCE__ * __formal=0x00400000, HINSTANCE__ * __formal=0x00400000, char * args=0x00152300, HINSTANCE__ * __formal=0x00400000) Line 1702 + 0x17 C++ mozilla.exe!WinMainCRTStartup() Line 390 + 0x1b C kernel32.dll!_BaseProcessStart@4() + 0x23 void PresShell::DidDoReflow() { HandlePostedDOMEvents(); HandlePostedAttributeChanges(); HandlePostedReflowCallbacks(); if (!mPaintingSuppressed) mViewManager->SynthesizeMouseMove(PR_FALSE); // crashed } + mViewManager 0x00000000 nsIViewManager * I have a proposal

Comment on attachment 144681 [details] [diff] [review] try to suppress painting when the viewmanager goes away sorry about the tabs, i'm using a coworker's box and his non default editor wasn't tamed.

Comment on attachment 144681 [details] [diff] [review] try to suppress painting when the viewmanager goes away It's much clearer to null-check mViewManager at the crash site, which doesn't exist anymore, with a comment about happening during Destroy. However, this stack shouldn't be possible -- like some of the other crashes related to bug 20022, it seems like revoking events didn't actually revoke the events. This crash stack can't happen anymore, but adding as a bug 20022 dependency for when that is relanded. Any useful steps to reproduce?

I relanded bug *** This bug has been marked as a duplicate of 234435 ***

sorry, no steps, all i know is what i wrote in the bug which was that we were quiting. is there a cvs checkin i can peel off into our tree, or was this fixed by some large patch? we're shipping based on a 1.7bish thing... (which replaced the 1.7aish thing of this report)