1757261 - https://codepen.io/danwilson/full/VwzNLYd is broken after bug1749380

Status: VERIFIED FIXED

(Core :: Graphics: WebRender, defect)

Description

[Gregory Pappas [:gregp]]

Attachment: regression.png

Steps to reproduce:

Visit https://codepen.io/danwilson/full/VwzNLYd

Actual results:

Animation is broken

Expected results:

Animation should not be broken

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1749380

Comment 2

[Glenn Watson [:gw]]

Doesn't reproduce for me locally, but it might be related to a couple of the other remaining regressions for that bug. Could you check if enabling gfx.webrender.debug.force-picture-invalidation in about:config makes it behave any differently for you?

Comment 3

[Gregory Pappas [:gregp]]

(In reply to Glenn Watson [:gw] from comment #2)

Doesn't reproduce for me locally, but it might be related to a couple of the other remaining regressions for that bug. Could you check if enabling gfx.webrender.debug.force-picture-invalidation in about:config makes it behave any differently for you?

Setting gfx.webrender.debug.force-picture-invalidation to true in about:config and restarting does not result in any change. I also tried a new Firefox profile and am getting the same brokenness. My exact build ID is 20220228090129 if that helps.

Comment 4

[Gregory Pappas [:gregp]]

Update: Turning off the animation named and-the-news on the element main seems to fix the issue. Some sort of validation bug?

Comment 5

[Glenn Watson [:gw]]

It's possible. Are you able to check if the bug disappears if you remove any mix-blend-mode (or similar, e.g. background-blend-mode) from the demo? I'm currently working on a bug involving incorrect coordinate mapping with nested mix-blend, wondering if this might actually be the same problem.

Comment 6

[Gregory Pappas [:gregp]]

Attachment: mix-blend-mode-removed.png

(In reply to Glenn Watson [:gw] from comment #5)

It's possible. Are you able to check if the bug disappears if you remove any mix-blend-mode (or similar, e.g. background-blend-mode) from the demo? I'm currently working on a bug involving incorrect coordinate mapping with nested mix-blend, wondering if this might actually be the same problem.

Sure! mix-blend-mode: overlay is found on main > div which seems to match every element in the tree. This seems to have fixed it but exposes a new bug (or maybe chrome is wrong)

Comment 7

[Glenn Watson [:gw]]

Great, thanks for checking. OK, so I'll work on this mix-blend-mode issue first, then once that is sorted we'll come back to this bug and see if it resolves the glitching (and then consider the chrome/ff difference above).

Comment 8

[Glenn Watson [:gw]]

I added a fix for a similar problem in https://bugzilla.mozilla.org/show_bug.cgi?id=1757201. Try build in https://treeherder.mozilla.org/jobs?repo=try&revision=422897f59ab4bb22f71da5d2c68426ebb86ab467 will have build artifacts if you want to test locally. Otherwise, that should land in nightly within a day or so and we'll see if it also resolves this bug.

Comment 9

[Glenn Watson [:gw]]

The fix from the other bug for mix-blend mode has landed in nightly now - would you mind testing the latest nightly to see if this issue is resolved for you?

Comment 10

[Gregory Pappas [:gregp]]

(In reply to Glenn Watson [:gw] from comment #9)

The fix from the other bug for mix-blend mode has landed in nightly now - would you mind testing the latest nightly to see if this issue is resolved for you?

Hi Glenn, thanks for following up! I can confirm the mix-blend-mode issue is now fixed in the latest Nightly. The odd wireframe effect mentioned earlier is still present, should I file a separate bug for that?

Comment 11

[Glenn Watson [:gw]]

Yep, let's file a separate bug for that and try to get a regression range, if possible. Have you used mozregression before? It's an amazing tool that makes it very quick to find which patch caused a regression (see https://mozilla.github.io/mozregression/).

Comment 12

[Gregory Pappas [:gregp]]

(In reply to Glenn Watson [:gw] from comment #11)

Yep, let's file a separate bug for that and try to get a regression range, if possible. Have you used mozregression before? It's an amazing tool that makes it very quick to find which patch caused a regression (see https://mozilla.github.io/mozregression/).

Tried mozregression but all the Firefox builds from before late 2021 have a weird crash in TLS code that I can't figure out. Filed bug1757851 and will try mozregression on a different operating system later.

Comment 13

[Giorgia Nichita, Release Desktop QA]

Reproduced this issue on an affected Nightly build (2022-02-25) on WIn 10 x64.
Verified as fixed on Firefox 99.0 (20220328190900) across the following platforms: WIn 10 x64, macOS 10.15 and Ubuntu 20.04.