[Webkit-unassigned] [Bug 45627] REGRESSION: Infinite redirect on developer.apple.com

Thu Nov 18 17:21:22 PST 2010

https://bugs.webkit.org/show_bug.cgi?id=45627

--- Comment #17 from Tony Gentilcore <tonyg at chromium.org>  2010-11-18 17:21:22 PST ---
> I don't really know what commitData does, so I could be wrong here.  I would try removing it and see what breaks. :)

That didn't turn out to be fruitful. Too many other tests fail to call notifyDone without it.

Like I mentioned in http://code.google.com/p/chromium/issues/detail?id=61958, I'll be on vacation for a while so I'm releasing this in case someone is able to get to it in the mean time.

Here's a quick brain dump:
When we schedule a navigation that will change the location, we should hit the current loader with the hardest hammer (cancel all loads, tasks, subframes, etc). window.stop() uses FrameLoader.stopForUserCancel(), so it is probably safe and appropriate. Then we should stop the parser if the hammer didn't already do so. Finally, we need to make sure the unload event fires and the new page loads.

The current patch does all of that, but hits an assert that indicates it isn't canceling the loads hard enough (I believe it only cancels subresources and not the main resource). The flows I've found that cancel the main resource all have unintended side effects that either cancel the redirect, fire errors, or don't fire the unload event.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.