1489337 - snap-perspective-1.html reftest sets preserve-3d when it (probably) shouldn't.

Status: RESOLVED FIXED

(Core :: Graphics: WebRender, defect, P2)

Description

[Glenn Watson [:gw]]

In the snap-perspective-1.html reftest, Gecko passes the stacking context style as preserve-3d, since there is a perspective matrix.

The WR code shouldn't need this, since it knows to propagate a perspective matrix down to child stacking contexts.

However, the presence of preserve-3d here confuses WR into creating a 3d rendering context, which is the incorrect behavior (and also a performance overhead, since it draws 3d rendering contexts to offscreen surfaces).

Comment 1

[Glenn Watson [:gw]]

I *think* the correct behavior here is for the Gecko code to only pass preserve-3d as a stacking context flag when it is explicitly specified in the CSS (that's the expectation of WR, anyway).

Comment 2

[Matt Woodrow (:mattwoodrow)]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=13832484b28224246d5f97c909c9c97020983281

Comment 3

[Matt Woodrow (:mattwoodrow)]

Attachment: Bug 1489337 - Don't set preserve-3d to true when creating WebRender commands for nsDisplayPerspective since it's not needed. r?jrmuizel

MozReview-Commit-ID: GiRda3WyTEu

Comment 4

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 9007088 [details]
Bug 1489337 - Don't set preserve-3d to true when creating WebRender commands for nsDisplayPerspective since it's not needed. r?jrmuizel

Jeff Muizelaar [:jrmuizel] has approved the revision.

Comment 5

[Maire Reavy [:mreavy]]

Hey Matt -- does P2 feel right? (I noticed Glenn filed this as a blocker, but I don't think that was intentional.)  Is it worth investigating the test failures and getting this in before the patch rots?

Comment 6

[Matt Woodrow (:mattwoodrow)]

Yeah it does. I'm just waiting for the current set of WebRender transform (and backface-visibility) changes to land, and then I'll rebase on top of that and investigate what failures remain.

Comment 7

[Glenn Watson [:gw]]

Yup, I initially flagged it as a blocker as I (incorrectly) thought it was causing some issues with the 3d transforms patches.

Feel free to ping me if you do get failures after rebasing.

Comment 8

[Matt Woodrow (:mattwoodrow)]

I had to mark a couple of web-platform-tests as failing since they had small fuzzy differences.

Setting a dependency on the wpt fuzzy support so that we can change these to fuzzy at some point.

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/368cda88b85f
Don't set preserve-3d to true when creating WebRender commands for nsDisplayPerspective since it's not needed. r=jrmuizel

Comment 10

[Cosmin Sabou [:CosminS]]

Backed out for reftest failures on transform-3d/1035611-1.html

Push with failures: https://treeherder.mozilla.org/#/jobs?repo=autoland&resultStatus=success,testfailed&classifiedState=unclassified&fromchange=368cda88b85f0259a0a43da733f027effe8131d0&tochange=17bd2cc6216a2305175425da0c3421f01fa4ae9d&searchStr=windows,10,x64,quantumrender,debug,reftests,with,e10s,test-windows10-64-qr%2Fdebug-reftest-e10s-1,r-e10s(r1)&selectedJob=199231132

Failure log: https://treeherder.mozilla.org/logviewer.html#?job_id=199231132&repo=autoland&lineNumber=45034

Backout link: https://hg.mozilla.org/integration/autoland/rev/17bd2cc6216a2305175425da0c3421f01fa4ae9d

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/76d218ea875e
Don't set preserve-3d to true when creating WebRender commands for nsDisplayPerspective since it's not needed. r=jrmuizel

Comment 12

[Bogdan Tara[:bogdan_tara | bogdant]]

https://hg.mozilla.org/mozilla-central/rev/76d218ea875e