1312362 - Assertion failure: aChild->GetApzc() != parent, at HitTestingTreeNode.cpp:90

Status: RESOLVED DUPLICATE of bug 1298218

(Core :: Panning and Zooming, defect, P3)

Description

[Carsten Book [:Tomcat]]

Attachment: stack

Found via bughunter topsite tests and reproduced on latest windows tinderbox debug trunk build.

Assertion failure: aChild->GetApzc() != parent, at c:/builds/moz2_slave/m-cen-w32-d-000000000000000000/build/src/gfx/layers/apz/src/HitTestingTreeNode.cpp:90

Steps to reproduce:
-> Load http://www.socialtriggers.com
--> Assertion

affects trunk -> beta and all platforms like windows, mac and linux

Comment 1

[Carsten Book [:Tomcat]]

[Tracking Requested - why for this release]:

topsite tests results - affects trunk -> beta on debug builds

Comment 2

[Kartikaya Gupta (email:[e-mail])]

Attachment: Bad layer tree

This is the layer tree that the APZ ends up getting. There are multiple instances of layers with a metrics scrollId=3 having children which also have the same metrics. That's not allowed. So something on the layout side is not constructing a proper layer tree.

Comment 3

[Kartikaya Gupta (email:[e-mail])]

That being said, the page seems to load and scroll fine in an opt build, so we seem to have enough graceful handling that this is not an issue for users.

Comment 4

[Kartikaya Gupta (email:[e-mail])]

/cc mstange and matt since this is likely a bug or unhandled case in FrameLayerBuilder.

Comment 5

[Markus Stange [:mstange]]

Is there a blend mode on that page? I know that there is currently a bug about a double ScrollMetaData when blend modes are involved, and that bug 1298218 will fix that.

Comment 6

[Botond Ballo [:botond]]

(In reply to Markus Stange [:mstange] from comment #5)
> Is there a blend mode on that page?

It looks that way. [1] contains:

  .blend-overlay {
    mix-blend-mode: overlay; }

[1] http://socialtriggers.com/wp-content/themes/socialtriggers/style.css?v=0.6.1

Comment 7

[Kartikaya Gupta (email:[e-mail])]

Ok, let's mark it as a dependency and revisit this once bug 1298218 lands.

Comment 8

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Too late for 50, untracked.

Comment 9

[Gerry Chang [:gchang]]

Per comment #3, mark 51 fix-optional and track 51- as the issue is handled in opt build. But, still happy to take fix in 51 early beta.

Comment 10

[Julien Cristau [:jcristau]]

bug 1298218 looks like a massive refactoring, doesn't seem likely to be a candidate for uplift, so marking this fix-optional for 52

Comment 11

[Kartikaya Gupta (email:[e-mail])]

Verified that this still reproduces on a m-c debug OS X build just before markus's patches in bug 1298218, and that it is fixed after his patches landed.