735943 - Crash @ nsCSSFrameConstructor::ProcessPendingRestyles

Status: VERIFIED FIXED

(Core :: Layout, defect)

Description

[Jesse Ruderman]

Attachment: testcase (crashes Firefox when loaded)

About 50% of the time, this testcase causes a crash in nsCSSFrameConstructor::ProcessPendingRestyles.

Comment 1

[Jesse Ruderman]

Attachment: stack trace with popups allowed

Comment 2

[Jesse Ruderman]

Attachment: stack trace with popups blocked

Comment 3

[Mats Palmgren (inactive)]

Attachment: stack

Comment 4

[Mats Palmgren (inactive)]

Attachment: fix+test

FYI, the document has a strong pointer to the nsSMILAnimationController
object, which keeps a raw pointer to it (mDocument).

Try results pending:
https://tbpl.mozilla.org/?usebuildbot=1&tree=Try&rev=0281d2f4f580

Filed bug 735966 on the assertion when running the crash test.

Comment 5

[Mats Palmgren (inactive)]

Again, with just this change this time (there's a jsreftest crash in the one above):
https://tbpl.mozilla.org/?usebuildbot=1&tree=Try&rev=8316f137ca46

Comment 6

[Jesse Ruderman]

My testcase (reduced from what the fuzzer created) is an insane tangle of setTimeouts. It might be possible to create a more robust regression test ;)

Comment 7

[Scoobidiver (away)]

I hit bp-d0df45dc-63db-44d8-a6d8-f76722120315 (popups blocked) and bp-3d521fe3-cc02-44fc-8058-d85972120315 (popups allowed).

Comment 8

[Mats Palmgren (inactive)]

Hmm, there's a couple of "ASSERTION: This is unsafe! Fix the caller!: 'Error', content/events/src/nsEventDispatcher.cpp, line 558" on the new crash test.
https://tbpl.mozilla.org/php/getParsedLog.php?id=10090225&tree=Try&full=1#error0

Comment 9

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 606076 [details] [diff] [review]
fix+test

Document (probably in the header and at the presshell check for mIsDestroying) that FlushResampleRequests can trigger style flushes, and r=me

Comment 10

[Mats Palmgren (inactive)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/75f461e05221

Comment 11

[Mats Palmgren (inactive)]

This crash fix seems worth taking on branches after baking it for a while...

Comment 12

[Mounir Lamouri (:mounir)]

https://hg.mozilla.org/mozilla-central/rev/75f461e05221

Comment 13

[Alex Keybl [:akeybl]]

(In reply to Mats Palmgren [:mats] from comment #11)
> This crash fix seems worth taking on branches after baking it for a while...

Do we know what the regressing bug is or whether this is a pain point for users? If not, this can ride the trains.

Comment 14

[Mats Palmgren (inactive)]

I would guess the addition of the FlushResampleRequests() call in
http://hg.mozilla.org/mozilla-central/rev/d6dbb4231c11
is what caused the regression (bug 547333).

The reason I recommend merging this fix to branches is that without it we
run script from inside a destroyed pres shell, which I think is unsafe.
The fix is very low-risk.

Comment 15

[Mats Palmgren (inactive)]

Comment on attachment 606076 [details] [diff] [review]
fix+test

See comment 14.

Comment 17

[Alex Keybl [:akeybl]]

Comment on attachment 606076 [details] [diff] [review]
fix+test

[Triage Comment]
Low risk fix, and both Mats and the security team agree that this is something we should fix all the way up to FF12. Please land asap, as we'll go to build with Beta 4 tomorrow (4/3) at 5PM PT.

Please make sure to prepare a patch for the ESR branch as well.

Comment 18

[Mats Palmgren (inactive)]

Attachment: for esr10

The fix applies also to esr10, but without the test since the pref()
reftest feature doesn't exist there, afaict.

Comment 19

[Al Billings [:abillings - ex-MoCo]]

Verified fixed in nightly Mozilla/5.0 (Macintosh; Intel Mac OS X 10.7; rv:14.0) Gecko/20120402 Firefox/14.0a1. The testcase pretty reliably crashes my Aurora build.

Comment 20

[Mats Palmgren (inactive)]

https://hg.mozilla.org/releases/mozilla-aurora/rev/18753a4ee7cd
https://hg.mozilla.org/releases/mozilla-beta/rev/a958bd24a213

Comment 21

[Lukas Blakk [:lsblakk] use ?needinfo]

Comment on attachment 611648 [details] [diff] [review]
for esr10

Thanks, please go ahead and land as per https://wiki.mozilla.org/Release_Management/ESR_Landing_Process

Comment 22

[Mats Palmgren (inactive)]

https://hg.mozilla.org/releases/mozilla-esr10/rev/f66c6b1c5ee6