[webkit-dev] Random Failures

[Eric Seidel]

I seem to be getting random failures when running run-webkit-tests.

I'm not sure if they are related to any changes I've made or not...

I see this one most often:
running css1/conformance/forward_compatible_parsing test -> failed

which I originally attributed to using Bison 2.0...
http://bugzilla.opendarwin.org/show_bug.cgi?id=3885

but it doesn't always fail there.  I've seen others:
running editing/deleting/delete-3928305-fix test -> failed
running editing/deleting/delete-and-undo test -> failed
running editing/deleting/delete-3608445-fix test -> failed

each time it seems to be a crap shoot as to whether it will succeed  
or fail at some random test.  Cosmic rays?

Cryo, bdash and Altha on #webkit are seeing similar issues.

About half the time all tests succeed, half the time one of them fails.

I've tried rolling out:
http://bugzilla.opendarwin.org/show_bug.cgi?id=3882
http://bugzilla.opendarwin.org/show_bug.cgi?id=3829
neither helped.

Those are the only fixes I made today which I thought could relate to  
these crashes at all.  Again, I have determined (and bdash has  
confirmed) that neither is related.

Off to bed now.  I'm not much use to anyone at this hour.

-eric

p.s. it is remotely possible that either:
http://bugzilla.opendarwin.org/show_bug.cgi?id=3826
or
http://bugzilla.opendarwin.org/show_bug.cgi?id=3674
are at fault.  Neither seems likely.

One guess is that this could be an uncaught error in HashSet??  It  
seems like some sort of memory smasher since it's so random.