Bug 112334 – Reset all clients on WebPage close

• Component: WebKit2
• Status: RESOLVED FIXED

[Xan Lopez]

There's a small chance that someone will call one of the clients WebPage holds when the page has already been destructed, since the pointers are valid after ::close(). This won't work and will (at least in the GTK+ port) cause a bunch of warnings. To avoid this situation just reset all clients to null, like the WebPageProxy class does.

[Xan Lopez]

Created attachment 193094 [details]
0001-Reset-all-clients-on-WebPage-close.patch

[Xan Lopez]

Ping about this?

[Sergio Villar Senin]

*** Bug 115509 has been marked as a duplicate of this bug. ***

[Sergio Villar Senin]

(In reply to comment #0)
> There's a small chance that someone will call one of the clients WebPage holds when the page has already been destructed, since the pointers are valid after ::close(). This won't work and will (at least in the GTK+ port) cause a bunch of warnings. To avoid this situation just reset all clients to null, like the WebPageProxy class does.

Actually there is not only a small chance but I found an scenario where not reseting the clients causes a crash 100% of the times.

So imagine that a webpage has some javascript that executes a XmlHttpRequest in the onunload() event handler. The ::close() method of the WebPage calls injectedBundle->willDestroyPage(). That will likely be used by the WKBundlePageResourceLoadClients to free their resources, including the clientInfo they have passed to the WKBundlePageSetResourceLoadClient.

In the same ::close() method, we find a call to  m_mainFrame->coreFrame()->loader()->detachFromParent() which will trigger the execution of the XMLHttpRequest, which will issue a willSendRequest among other notifications. As the client is still set, the method will be called but it will likely try to use client's already freed resources leading to crashes.

So yeah I big +1 for this change :)

[Xan Lopez]

If it's possible to make a layout test for this crasher I bet this patch would land a lot faster ;)

[Alejandro G. Castro]

Fixes the issue for me, it was easy to reproduce just closing a tab while loading. We should add this to the release today.

[Xan Lopez]

Comment on attachment 193094 [details]
0001-Reset-all-clients-on-WebPage-close.patch

Thanks. Let's see if cq can still land it.

[WebKit Commit Bot]

Comment on attachment 193094 [details]
0001-Reset-all-clients-on-WebPage-close.patch

Clearing flags on attachment: 193094

Committed r150758: <http://trac.webkit.org/changeset/150758>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.