537416 - mochitest-plain: intermittent "test_bug450930.xhtml | Didn't get subdoc invalidation while we were privileged (iframe2)"

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Paul O'Shannessy [:zpao] (not bugmail, email directly)]

Just started after http://hg.mozilla.org/mozilla-central/pushloghtml?changeset=ab3e23cefcce.

Note: same file, but different test from bug 501221.

http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1262294460.1262296106.4336.gz
WINNT 5.2 mozilla-central debug test mochitests-4/5 on 2009/12/31 13:21:00

http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1262293564.1262295614.30942.gz
WINNT 5.2 mozilla-central debug test mochitests-4/5 on 2009/12/31 13:06:04

Comment 1

[Boris Zbarsky [:bzbarsky]]

I just pushed http://hg.mozilla.org/mozilla-central/rev/228adabcc071 in an attempt to fix this.  It seems like the test was assuming that restyle processing in the two documents is related in a particular way, but I don't think we have such guarantees with the refresh driver timers...

That said, I'm suprised the invalidation for the outer document didn't flush out style/layout in the child.  Is that right?

Comment 2

[Boris Zbarsky [:bzbarsky]]

The push mentioned in comment 1 seems to have fixed this orange.  But I'd still like roc to check it over...

Comment 3

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Your patch is obviously safe, but what did it fix? Were we invalidating in the subdoc before it had been reflowed to a non-zero size in the parent?

Comment 4

[Boris Zbarsky [:bzbarsky]]

My best guess is that due to timer lag on the refresh timer under high-load conditions we were not actually processing those restyles before something else triggered the MozAfterPaint handler on the parent (not sure what).  Hence my question at the end of comment 1...

Comment 5

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Ah, I see. Thanks, that sounds right.

Comment 6

[Boris Zbarsky [:bzbarsky]]

As long as it's ok for the invalidate event in the parent to be happening without triggering restyle processing in the child, sure.  I thought that painting in a parent flushed out restyles and layout in kids, though...