Aleksandr Chernokondratenko Reporter
	Description • 7 years ago

User Agent: Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:57.0) Gecko/20100101 Firefox/57.0
Build ID: 20171128222554

Steps to reproduce:

If I add transform scale to the inner element and transform rotate to the wrapper element, it'll rotate with artifacts on Mozilla Firefox
The issue occurs on Firefox for Windows.
I've tested it 47.0.2 and 57.0.1 versions.
Firefox 57.0.1 works good for Ubuntu
Here is the code https://plnkr.co/edit/uX0V8QYiJI7e6Rae1niS?p=preview



Actual results:

I see artifacts
The screenshot attached


Expected results:

It must animate without artifacts

	:Gijs (he/him)
	Comment 1 • 7 years ago

Not a security issue that needs to stay hidden.

	Alice0775 White
	Comment 2 • 7 years ago

I can reproduce.

	Alice0775 White
	Comment 3 • 7 years ago

Regression window:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=958d2a5d10091401fd5e900e8e063d21940c137e&tochange=7f894f791cdf170d788507d0eff30024ce699523


Via local build,
Last good: build from cset 1561522b1179 + patch 7f894f791cdf
First bad: build from cset 68932b1a5b71 + patch 7f894f791cdf


Regressed by: Matt Woodrow — Bug 1361970 - Make PostProcessLayers occlusion culling work against the surface we will draw to rather than the parent layer. r=mstange

	Matt Woodrow (:mattwoodrow) Assignee
	Comment 4 • 7 years ago

This is bug 1381753 again, except for Advanced Layers this time. Getting reftests for async animations would be nice so that we don't break this a third time.

	Pulsebot
	Comment 5 • 6 years ago

Pushed by mwoodrow@mozilla.com:
https://hg.mozilla.org/integration/mozilla-inbound/rev/32bf2817fb7a
Recompute shadow visible rect for ContainerLayerMLGPU in case it has been changed by async animations. r=mstange

	Raul Gurzau (:RaulG)
	Comment 6 • 6 years ago
bugherder

https://hg.mozilla.org/mozilla-central/rev/32bf2817fb7a

	Ryan VanderMeulen [:RyanVM]
	Comment 7 • 6 years ago

We're late in the cycle and the lack of tests for this makes me nervous. That said, it grafts cleanly to Beta as-landed. Should we consider this for Beta uplift or let it ride the trains?

	Matt Woodrow (:mattwoodrow) Assignee
	Comment 8 • 6 years ago

The lack of tests make me nervous too.

That said, this is just a conversion of the existing fix to be used on advanced layers, so the risk should be very low.

I think we should go for uplift!

	Matt Woodrow (:mattwoodrow) Assignee
	Comment 9 • 6 years ago

Comment on attachment 8951043 [details] [diff] [review]
Recompute visible rect for ContainerLayerMLGPU

Approval Request Comment
[Feature/Bug causing the regression]: Bug 1361970
[User impact if declined]: Incorrect invalidation (trail left behind) for some async transform animations on Windows. Affects google calendar.
[Is this code covered by automated tests?]: No.
[Has the fix been verified in Nightly?]: Yes.
[Needs manual test from QE? If yes, steps to reproduce]: No.
[List of other uplifts needed for the feature/fix]: None.
[Is the change risky?]: No.
[Why is the change risky/not risky?]: Just takes the existing fix, and enables it for advanced layers.
[String changes made/needed]: None.

	Ryan VanderMeulen [:RyanVM]
	Comment 12 • 6 years ago

Comment on attachment 8951043 [details] [diff] [review]
Recompute visible rect for ContainerLayerMLGPU

Let's get this into tomorrow's 59b12 build so we can get it tested as much as possible before we ship 59. Marking qe-verify+ as well to get some manual verification.

	Liz Henry (:lizzard Please n-i to RyanVM, jcristau, or pascal)
	Comment 13 • 6 years ago

Comment on attachment 8951043 [details] [diff] [review]
Recompute visible rect for ContainerLayerMLGPU

Not a new regression (it's from 56) but let's give this a try in 59. 

Matt, maybe you could file a followup bug for writing reftests ?

Aleksandr, would you mind verifying the fix in nightly (60) or in 59 beta 12 later this week?

	Matt Woodrow (:mattwoodrow) Assignee
	Comment 14 • 6 years ago

(In reply to Liz Henry (:lizzard) (needinfo? me) from comment #13)

> Matt, maybe you could file a followup bug for writing reftests ?

Unfortunately this codepath is in code that we can't hit using reftests. It's specific to async animations, and we don't have any infrastructure currently that can test the rendering of those.

We have bug 1248828 filed for getting that support, but it hasn't been finished yet.

	Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)
	Comment 15 • 6 years ago
bugherder uplift

https://hg.mozilla.org/releases/mozilla-beta/rev/5463350d58a7

	Ciprian Georgiu, Desktop QA
	Comment 16 • 6 years ago

I've managed to reproduce this bug using STR from comment 0, on an affected Nightly build (2017-12-06).

I can no longer see the issue on latest Nightly 60.0a1 under Win 10 x64, Win 8.1 x86 and Win 7 x64.

Ni? myself as a reminder to verify this on 59.0b12 as well.

	Ciprian Georgiu, Desktop QA
	Comment 17 • 6 years ago

This bug is also verified fixed on Beta 59.0b12 (20180222170353) across the same platforms mentioned in the above comment.