It happens when the plugin process crashes and Connection::postConnectionDidCloseOnConnectionWorkQueue() schedules a connectionDidClose but connection was already closed.
Created attachment 98335 [details] Patch
Comment on attachment 98335 [details] Patch Not sure if this fix is correct. is m_shouldWaitForSyncReplies protected by a mutex? Would it be better to check something in connectionDidClose, maybe modify the assertion?
Created attachment 98466 [details] New patch Right, I don't think calling connectionDidClose() more than once is a bug, since it can be triggered by different sources, so I've removed the assert and simply check if m_shouldWaitForSyncReplies is true.
Comment on attachment 98466 [details] New patch View in context: https://bugs.webkit.org/attachment.cgi?id=98466&action=review > Source/WebKit2/ChangeLog:10 > + (CoreIPC::Connection::connectionDidClose): Do not assert when > + connectionDidClose is called more than once. Why do we want to allow calling this more than once?
Sorry, I see your answer there above. Seems Anders should review this.
(In reply to comment #5) > Sorry, I see your answer there above. Seems Anders should review this. Anders can you please have a look at this patch?
Comment on attachment 98466 [details] New patch The ChangeLog needs an explanation on: -What this change fixes. -Why this is the correct fix.