1266380 - Regression: 3D transformed elements disappear when parent have filter+perspective applied

Status: VERIFIED FIXED

(Core :: Layout, defect)

Description

[vincent]

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/50.0.2661.75 Safari/537.36

Steps to reproduce:

1. Open this JSBin: http://output.jsbin.com/jusoya in Firefox 48
2. The red square is not visible
3. Remove the `filter` declarations from `.perspective` selector
4. The red square become visible

This only occur when an element is 3D transformed AND
its parent have CSS `perspective` and `filter` set.

Even stranger, remove the `top` and `left` properties from the `.perspective`, and the red square appears, but seems to be transformed weirdly.


Actual results:

The `filter` property makes elements dissapear.
Bug is not present in Firefox 45.0.2 but is in Firefox 48. So it seems to be a regression.

Comment 1

[Logan Rosen [:Logan]]

> 11:00.29 INFO: Last good revision: 604a180b6cc0e101eef1f974356b613aa0a40146
> 11:00.29 INFO: First bad revision: 51102a2a44b51fa19ccb8f7504ea07c4a65ebf55
> 11:00.29 INFO: Pushlog:
> https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=604a180b6cc0e101eef1f974356b613aa0a40146&tochange=51102a2a44b51fa19ccb8f7504ea07c4a65ebf55

> 11:01.22 INFO: Looks like the following bug has the changes which introduced the regression:
> https://bugzilla.mozilla.org/show_bug.cgi?id=1252324

Comment 2

[Lee Salzman [:lsalzman]]

Attachment: fix offset of untransformed surface when 3d transforming in BasicLayerManager

The untransformed draw target is drawn with an offset at bounds.x/y, but this was not factored into the effective transform that is applied to move the untransformed DT snapshot from untransformed space to transformed space.

This fixes that and some incidental questionable uses of references on rvalues nearby.

Comment 3

[Lee Salzman [:lsalzman]]

So, my patch deals with one part of the regression, but I ran through mozregression twice myself, and the regression actually started with bug 1168263 in the following patch by Matt Woodrow: https://hg.mozilla.org/mozilla-central/rev/7e18014be68d

Comment 4

[Lee Salzman [:lsalzman]]

Attachment: simplified test case

A simplified version of the test case that can be run locally.

Comment 5

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/4f8e510a1b7f

Comment 6

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/4f8e510a1b7f

Comment 7

[Matt Woodrow (:mattwoodrow)]

Attachment: Don't try to untransform visible regions for inactive perspective layers

Comment 8

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/d3b2f4b2e016

Comment 9

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/d3b2f4b2e016

Comment 10

[Alice0775 White]

[Tracking Requested - why for this release]: Web page not shown correctly on Aurora48.0a2 with both e10s and HWA disabled.

See Bug 1275504


Could you up lift this to 48?

Comment 11

[Markus Stange [:mstange]]

And to 47, if possible.

Comment 12

[Matt Woodrow (:mattwoodrow)]

Comment on attachment 8744536 [details] [diff] [review]
fix offset of untransformed surface when 3d transforming in BasicLayerManager

Approval Request Comment
[Feature/regressing bug #]: Bug 1168263
[User impact if declined]:Broken rendering when combining perspective and css filters.
[Describe test coverage new/current, TreeHerder]: Manually tested
[Risks and why]: Low risk.
[String/UUID change made/needed]: None

Comment 13

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

Hi folks, I am leaning towards wontfixing this for Fx47. We enter RC week on Monday. I am not sure but this regression might already be in Fx46 and we might need to ship another release with it. At this point only issues that might lead to dot-releases are making the cut for Beta uplift. Please let me know if this is a release blocker. Thanks!

Comment 14

[Milan Sreckovic [:milan] (needinfo for best results)]

Based on comment 3, this is actually in 45, so I'd say keep it out of beta given the timing.

Comment 15

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

Thanks Milan. This is a regression since Fx45, it's too late to take a fix in Beta47.

Comment 16

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

Comment on attachment 8744536 [details] [diff] [review]
fix offset of untransformed surface when 3d transforming in BasicLayerManager

Let's uplift to Aurora48, Beta47-

Comment 17

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/releases/mozilla-aurora/rev/036020df2de8

Comment 19

[Petruta Horea [:phorea], Desktop QA]

Reproduced with 2016-04-21 Nightly.
Verified as fixed across platforms using latest Aurora 48.0a2, 49.0a2.