1319683 - Overlapping elements sometimes invisible/stacked below other elements in combination with translate3d and rotates

Status: RESOLVED DUPLICATE of bug 689498

(Core :: Layout, defect, P1)

Description

[Dennis Schubert [:denschub]]

Attachment: Screenshot illustrating the issue

It's better to have a look at the attached screenshot of the example in this bug than having me explaining this issue in words. It looks like calculating the Z position fails under some circumstances, causing some elements to be rendered "behind" other elements although they should be in front.

As far as mozregression and manual bisecting told me, this is a regression from bug 1097464, but it looks like it got backed out and I was not able to find the re-landing. Thinker, you landed that, can you help out and update the blocking bug if needed?

[Tracking Requested - why for this release]: As discussed with :miketaylr, requesting tracking as this is a regression that could affect a lot of sites using 3D transforms.

As a side note, even before landing bug 1097464, this already got regressed once in July 2015 between 8e5c888d0d89 and 1f77b78797d6 but it got fixed some time after that and broke again with 1097464.

Comment 1

[Thinker Li [:sinker]]

kip had worked on plane-splitting as an idea solution for this problem.  I am not sure how it is.

Comment 2

[Markus Stange [:mstange]]

Miko, does your plane splitting work fix this testcase?

Comment 3

[Miko Mynttinen]

(In reply to Markus Stange [:mstange] from comment #2)
> Miko, does your plane splitting work fix this testcase?

It does. https://www.dropbox.com/s/wr7toe4aiefx34v/rubik.mov?dl=0

Comment 4

[Marcia Knous [:marcia]]

Tracking 53- will move the tracking flags to the duplicate bug.

Comment 5

[Miko Mynttinen]

Apologies for closing this bug a bit hastily.

The layer sorting code is currently getting completely rewritten (and the old code is getting removed) along with plane splitting implementation.
The new sorting code should fix most issues like this. I'm hoping to get some of it landed within a next couple of weeks.

Comment 6

[:kip (Kearwood Gilbert)]

(In reply to Thinker Li [:sinker] from comment #1)
> kip had worked on plane-splitting as an idea solution for this problem.  I
> am not sure how it is.

Indeed, as Miko has confirmed the rendering bug is related to sorting and/or plane splitting.

Comment 7

[Dennis Schubert [:denschub]]

That's cool and nice to know. Thanks everyone! :)