949294 - "ASSERTION: Frame passed in not in reflow?" with very interruptible reflow

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Jesse Ruderman]

Attachment: stack

GECKO_REFLOW_INTERRUPT_MODE=counter
GECKO_REFLOW_INTERRUPT_FREQUENCY=1
GECKO_REFLOW_INTERRUPT_CHECKS_TO_SKIP=0
GECKO_REFLOW_MIN_NOINTERRUPT_DURATION=0 

layout/generic/crashtests/381786-1.html

###!!! ASSERTION: Frame passed in not in reflow?: 'aFrame->GetStateBits() & NS_FRAME_IN_REFLOW', file /Users/jruderman/trees/mozilla-central/layout/base/nsPresShell.cpp, line 2725

Comment 1

[Jesse Ruderman]

Several other crashtests hit this assertion when I use the same env vars, but many of them don't hit it reliably.  Am I doing something wrong that makes these tests less deterministic than they could be?

Comment 2

[Mats Palmgren (inactive)]

Attachment: wallpaper

Reflowing abs.pos. children of XUL frames currently happens after the
delegating frame is fully reflowed (WillReflow/Reflow/DidReflow)
so the IN_REFLOW bit is not set when nsAbsoluteContainingBlock checks
for interrupts:
http://mxr.mozilla.org/mozilla-central/source/layout/generic/nsAbsoluteContainingBlock.cpp#181

This sort of wallpapers the problem, but this stuff is pretty low
priority so I don't think it's worth spending more time on.

https://tbpl.mozilla.org/?tree=Try&rev=c7f5819c6206

Comment 3

[Mats Palmgren (inactive)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/c86422c589c2

Comment 4

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/c86422c589c2