1309138 - Rendering issues/artifacts on transitions

Status: NEW

(Core :: Layout, defect, P3)

Description

[MarkBauer]

Attachment: renderingBug.png

User Agent: Mozilla/5.0 (Windows NT 6.3; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/53.0.2785.116 Safari/537.36

Steps to reproduce:

Browser: Firefox Developer Edition (Version 51.0a2)

Opening the following reveal.js presentation and pressing the "right" and then, immediately after the first transition, "down" key, I could reproduce the rendering issue as mentioned below.

https://happy-ferret.github.io/FirefoxPres/


Actual results:

Torn frames and artifacts (see screenshot).

Turning off hardware acceleration did, unfortunately, not resolve the issue.


Expected results:

As is the case in Firefox 49, the page should've been properly rendered without any such issues.

Comment 1

[MarkBauer]

Please ignore the user agent.
I didn't realise that info was recorded by Bugzilla.

At the time of writing I'm using Chrome, as this is browser most commonly supported by my place of employment.

Comment 2

[Alice0775 White]

I can reproduce the problem on Beta50b5 and Aurora51.0a2 and latest Nightly52.0a1[1], but not on 49.0.1.

[1]https://hg.mozilla.org/mozilla-central/rev/7be6b348c431d69f96f0765af3a0c0a0fe56d4bf
Mozilla/5.0 (Windows NT 10.0; WOW64; rv:52.0) Gecko/20100101 Firefox/52.0 ID:20161010030204


Regression window:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=067521ccd17907843e1db644ca0f02a1f5a6520a&tochange=eac0c056235ee1174fc092fe159d2d5710331fbd

Regressed by:
eac0c056235e	Matt Woodrow — Bug 1283827 - Flatten transform-style:preserve-3d when combined with opacity. r=dbaron

Comment 3

[Alice0775 White]

[Tracking Requested - why for this release]: Visual problem on the Web page that Reveal.js is used

Comment 4

[Alice0775 White]

Attachment: reduced html (for the record)

Comment 5

[MarkBauer]

Just a heads-up that this bug seems to only affect Windows clients.
After I got home last night I did some extensive testing on my Ubuntu workstation (using the latest Firefox Nightly) and have not run into this problem once.

The same is true for a Mac OS device I tested it on just this morning.

Comment 6

[Marcia Knous [:marcia]]

Tracking 52+ - rendering regression that affects Windows on a popular js site.

Comment 7

[Astley Chen (inactive)]

Hi Alice, do you have any additional comments or info for setting "status-firefox50: --- → fix-optional" here ?

Hi Matt, could you help to take a look ?

Comment 8

[Alice0775 White]

(In reply to Astley Chen [:astley] (UTC+8) from comment #7)
> Hi Alice, do you have any additional comments or info for setting
> "status-firefox50: --- → fix-optional" here ?

I cannot find real world web site yet. So, I marked as optional.

Comment 9

[Gerry Chang [:gchang]]

Mark 51 fix-optional for now as no real world web site is found yet.

Comment 10

[Julien Cristau [:jcristau]]

David, both this and bug 1319825 are listed as regressions from bug 1283827.  Since Matt doesn't seem available, and you reviewed that patch, would you be able to help investigating those regressions, or advise on who else could do that?

Comment 11

[Jet Villegas (inactive)]

Thinker: WDYT?

Comment 12

[Thinker Li [:sinker]]

I will look into it.

Comment 13

[Peter Chang[:pchang]]

I can't reproduce this on my windows 7 with FF 53.
Alice, could you still reproduce this issue?
If yes, could you share your about:support?

Comment 14

[Alice0775 White]

Attachment: about:support

I can reproduce very easily(80% probability) with the reduced testcase with Nightly53.0a1.

Steps to reproduce.
1. Quickly click as much as possible [Downarrow] [Rightarrow] button in that order
2. if not reproduce, reload the page and then repeat Step1

Comment 15

[Astley Chen (inactive)]

Remove ni.

Alice,
I can't reproduce on macOS, Windows 10 either with my Firefox 50~53 against test cases in comment 0 and reduced one.
Can you also reproduce on comment 0 website?
Wondering what firefox/system configurations can help us to reproduce this issue easily...

Comment 16

[Alice0775 White]

(In reply to Astley Chen [:astley] UTC+8 from comment #15)

> Can you also reproduce on comment 0 website?

Yes, I can reproduce.

> Wondering what firefox/system configurations can help us to reproduce this
> issue easily...

I can also reproduce the issue on the following VMs.
guest os: ubuntu16.04 32bit, host os: win10 64bit
guest os: ubuntu16.04 64bit, host os: win10 64bit
guest os: win7 64bit, host os: win10 64bit

Comment 17

[Astley Chen (inactive)]

Thanks, Alice.

So far, I can easily reproduce on below conditions:

Firefox Nightly 53.0a1, running in
(Parallel VM)Host OS: macOS 10.12, Guest OS: Windows 10, Ubuntu 16.04

Not reproducible on my Windows 10, macOS host OS.

I'm wondering if it's related to HW simulation on guest OS.
Peter, thoughts ?

Comment 18

[Astley Chen (inactive)]

Attachment: Looks-good-on-my-windows10-desktop-firefox-beta-51.0b10

I can't reproduce on my Windows10 Firefox Beta 51.0b10.
Upload about:support for reference.

Comment 19

[Andrew Overholt [:overholt]]

I can reproduce comment 0 on a Surface Book with Windows 10 in today's nightly.

Comment 20

[Ryan VanderMeulen [:RyanVM]]

Jet, can you help find an owner for this?

Comment 21

[Julien Cristau [:jcristau]]

Looks like a wontfix for 52 at this point.  FWIW I can still reproduce glitches in nightly 54 when navigating quickly with arrow keys on the website from comment 0.

Comment 22

[Nathan Froyd [:froydnj]]

[Tracking Requested - why for this release]: maybe if we track for a few more releases, this regression for Windows clients can actually get fixed.

Comment 23

[Marcia Knous [:marcia]]

(In reply to Nathan Froyd [:froydnj] from comment #22)
> [Tracking Requested - why for this release]: maybe if we track for a few
> more releases, this regression for Windows clients can actually get fixed.

I tracked for all three, but I think we need to find someone to own this bug. This can be discussed at our next triage - and there is a ni on Jet.

Comment 24

[Liz Henry (:lizzard) (relman/hg->git project)]

Too late for a fix for 53. I will email Jet.

Comment 25

[Jet Villegas (inactive)]

Sorry for the late reply. I finally have a machine that reproduces this and showed it to Thinker. 

This one's odd in that I would have expected this to be all-platforms and 100% reproducible after inspecting the regressing patch. That this is Windows-only and intermittent points to underlying issues we have on Windows that likely existed before the regression range. 

Keeping the needinfo open until I can get a Windows developer on the GFx team to take a look.

Comment 26

[Gerry Chang [:gchang]]

Mark 54 fix-optional as there are no actions for the moment but still happy to have the fix in 54.

Comment 27

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

This has been tagged as a P3. I am doubtful whether this will get fixed anytime soon. Tracking "-" as it's not a very recent regression and a deved bug only I presume. When we fix it please feel free to nominate for uplift to Beta if it's deemed low risk.

Comment 28

[Miko Mynttinen]

I think the cause for this bug might be the same Windows only invalidation issue causing bug 1345891. I tried a quick hack that forces layer invalidation and was not able to reproduce this afterwards.

Comment 29

[Astley Chen (inactive)]

(In reply to Miko Mynttinen [:miko] from comment #28)
> I think the cause for this bug might be the same Windows only invalidation
> issue causing bug 1345891. I tried a quick hack that forces layer
> invalidation and was not able to reproduce this afterwards.

Thanks for the info.
Let's track on bug 1345891 and see whether the problem is fixed once it's fixed.