Summary: | Crash in Connection::isValid when called in response to VisitedLinksProvider::pendingVisitedLinksTimerFired after WebProcessPRoxy::didClose was already called | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Jessie Berlin <jberlin> | ||||
Component: | WebKit2 | Assignee: | Jessie Berlin <jberlin> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | andersca, aroben, jberlin | ||||
Priority: | P2 | Keywords: | InRadar | ||||
Version: | 528+ (Nightly build) | ||||||
Hardware: | All | ||||||
OS: | All | ||||||
Attachments: |
|
Description
Jessie Berlin
2010-10-22 14:21:44 PDT
Created attachment 71594 [details]
Add WebContext::proccessDidClose and stop the visited links timer there
Comment on attachment 71594 [details] Add WebContext::proccessDidClose and stop the visited links timer there View in context: https://bugs.webkit.org/attachment.cgi?id=71594&action=review Looks great otherwise, r=me! > WebKit2/UIProcess/WebProcessProxy.cpp:420 > + m_context->processDidClose(this); You should call processDidClose before the call to WebProcessManager::processDidClose because calling it might delete the WebProcessProxy object. (In reply to comment #3) > (From update of attachment 71594 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=71594&action=review > > Looks great otherwise, r=me! > > > WebKit2/UIProcess/WebProcessProxy.cpp:420 > > + m_context->processDidClose(this); > > You should call processDidClose before the call to WebProcessManager::processDidClose because calling it might delete the WebProcessProxy object. Done. Thanks for the review! Comment on attachment 71594 [details] Add WebContext::proccessDidClose and stop the visited links timer there Committed in r70346 http://trac.webkit.org/changeset/70346 Is there any way to make a test for this? (In reply to comment #6) > Is there any way to make a test for this? I don't think so at the moment. The case in which we are currently seeing this happen is not necessarily caused by the WebProcess crashing - more has to do with the timing of the WebProcess ending which is a bit hard to control. |