459883 - Crash [@ nsSVGFilterProperty::GetFilterFrame] with filter, MathML

Status: VERIFIED FIXED

(Core :: Layout, defect)

Description

[Jesse Ruderman]

Attachment: testcase

Loading the testcase triggers three assertions and a scary crash.

###!!! ASSERTION: frame was not removed from primary frame map before destruction or was readded to map after being removed: 'Not Reached', file /Users/jruderman/central/layout/base/nsFrameManager.cpp, line 720

###!!! ASSERTION: GetPrimaryFrameFor() called while frames are being destroyed!: '!mIsDestroyingFrames', file /Users/jruderman/central/layout/base/nsFrameManager.cpp, line 328

###!!! ASSERTION: Cached frame is incorrect!: 'mElement.get() && static_cast<nsGenericElement*>(mElement.get())->GetPrimaryFrame() == mReferencedFrame', file /Users/jruderman/central/layout/svg/base/src/nsSVGEffects.cpp, line 78

Crash with nsSVGFilterProperty::GetFilterFrame calling 0xdddddddd.

Comment 1

[Robert Longson [:longsonr]]

Probably the same issue as bug 458493 and bug 455314.

Comment 2

[Robert Longson [:longsonr]]

See bug 455314 comment 9 for analysis.

Comment 3

[Robert Longson [:longsonr]]

Attachment: patch

Comment 4

[Robert Longson [:longsonr]]

This patch does not fix bug 455314 nor does it fix bug 454945 perhaps the patch in bug 458453 would fix those.

Comment 5

[Robert Longson [:longsonr]]

I'll do s/Ensure that the filter's is/Ensure that the filter is/ on check in.

Comment 6

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

Comment on attachment 344482 [details] [diff] [review]
patch

+  // Ensure that the filter's is repainted correctly

s/filters's/filter/

Comment 7

[Robert Longson [:longsonr]]

Checked in http://hg.mozilla.org/mozilla-central/rev/678883a079ba

Comment 8

[Henrik Skupin [:whimboo][⌚️UTC+2]]

Robert, how can I verify this fix? I cannot get crashed an older build with the attached testcase on OS X nor Windows.

Comment 9

[Robert Longson [:longsonr]]

It certainly used to crash for me without this fix on Windows.

Comment 10

[Henrik Skupin [:whimboo][⌚️UTC+2]]

Ok, I was able to crash Firefox with a debug build without this patch. The crashes don't happen anymore with builds after the patch went in.

Jesse, the assertions still happen. Shall we handle those in a separate bug?

Comment 11

[Jesse Ruderman]

I only see these assertions now:

###!!! ASSERTION: GetPrimaryFrameFor() called while frames are being destroyed!: '!mIsDestroyingFrames', file /Users/jruderman/central/layout/base/nsFrameManager.cpp, line 336

###!!! ASSERTION: frame was not removed from primary frame map before destruction or was readded to map after being removed: 'Not Reached', file /Users/jruderman/central/layout/base/nsFrameManager.cpp, line 734

The first could be bug 459666, but the second one is only covered by -moz-column bugs.  I don't think it's worth filing a new bug now, especially since it would have to be security-sensitive.  I'll retest when bug 459666 is fixed.

Comment 12

[Henrik Skupin [:whimboo][⌚️UTC+2]]

Attachment: Stack of assertion GetPrimaryFrameFor

Jesse, the stack looks completely different from the one on bug 459666. See the attachment. Are you sure that I should not file a new bug on that?

Comment 13

[Jesse Ruderman]

Attachment: Stack of assertion GetPrimaryFrameFor (using fix-macosx-stack.pl)

You have to run tools/rb/fix-macosx-stack.pl to turn the XPCOM_DEBUG_BREAK=trap output into a useful stack trace.  (I don't know why XPCOM_DEBUG_BREAK bothers trying to print functions names at all, given how wrong they are.)

Comment 14

[Jesse Ruderman]

But you're right, the stacks do look different.  I filed bug 477569.

Comment 15

[Mats Palmgren (inactive)]

Landed the crashtest:
https://hg.mozilla.org/integration/mozilla-inbound/rev/cc86e0fcb91b

Comment 16

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/cc86e0fcb91b