Patch for bug 29832 fixed a crash in Safari 4 caused by not resetting a JSDOMWindow wrapper on a detached frame in some cases.
The test reproducing the crash can not be added as layout test because it takes ~10 seconds to run. Need to add it as a manual test.
Created attachment 44583 [details]
style-queue ran check-webkit-style on attachment 44583 [details] without any errors.
Comment on attachment 44583 [details]
I don't quite understand why this needs to take ~10 seconds (what's back/forward cache expiration?), but maybe a new DRT method can make this instant?
(In reply to comment #4)
> I don't quite understand why this needs to take ~10 seconds (what's
> back/forward cache expiration?), but maybe a new DRT method can make this
Indeed, there is no way today to cause immediate b/f cache expiration so the destruction of the cached pages can not be realistically tested in DRT today. I thought about something like layoutTestController.setBackForwardCacheExpiration(seconds) but wasn't sure the single test justifies it... Perhaps we should do it next time we have a need?
This case seems significant enough to warrant adding a DRT method.
Perhaps it would be easier to just force immediate expiration, rather than set a timeout?
(In reply to comment #6)
> This case seems significant enough to warrant adding a DRT method.
> Perhaps it would be easier to just force immediate expiration, rather than set
> a timeout?
Ok. I'll take a look at adding a method. Will delay landing the patch until implementation is ready, then we'll decide.
Ping? Curious what the status is here. This patch has been approved for landing for a little over 2 weeks.
I agree with Alexey that having manual test is somewhat useless since nobody ever runs them. I have this task (make it possibel to test the bf cache expiration from DRT in my list and will get to it at some point, in a few weeks.
So for now I guess I'll commit this test so it is there, but will address the DRT testing at some point.