Closed Bug 1417286 Opened 7 years ago Closed 6 years ago

black background when scrolling in extensions

Categories

(Core :: Graphics, defect, P3)

57 Branch
defect

Tracking

()

RESOLVED DUPLICATE of bug 1425455

People

(Reporter: apps, Unassigned)

Details

(Whiteboard: [gfx-noted])

Attachments

(2 files)

Attached image Screenshot (13).jpg
User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/62.0.3202.94 Safari/537.36

Steps to reproduce:

Make sure you have at least one unread email then install
the extension Checker Plus for Gmail https://addons.mozilla.org/firefox/addon/checker-plus-gmail/
View the email and use the scrollbar.


Actual results:

Background goes black


Expected results:

Background should not be affected.
Please make sure you have the latest graphics driver: https://support.mozilla.org/kb/upgrade-graphics-drivers-use-hardware-acceleration
Component: Untriaged → Graphics
Product: Firefox → Core
Disabling hardware acceleration and webgl did not work.

The issue is only showing up in Firefox v57.

It worked correctly in all previous versions of Firefox and the same extension code is working in Chrome also.

Other users of my extension are reporting the issue with the latest update to Firefox v57.
I sometimes encounter the same bug in with extension WhatRuns (https://addons.mozilla.org/en-US/firefox/addon/whatruns/), when the popup have a scroll bar (just like Checker Plus for Gmail).

But it's very fast, and cannot be consistently repeated so I can't take a screenshot.

But you can try it yourself on complex websites like https://www.reddit.com which will have a scroll bar.
(In reply to fireattack from comment #3)
> I sometimes encounter the same bug in with extension WhatRuns
> (https://addons.mozilla.org/en-US/firefox/addon/whatruns/), when the popup
> have a scroll bar (just like Checker Plus for Gmail).
> 
> But it's very fast, and cannot be consistently repeated so I can't take a
> screenshot.
> 
> But you can try it yourself on complex websites like https://www.reddit.com
> which will have a scroll bar.

This thread is not for that app, please open another ticket if you need to report it, thanks....are there no mods on here to enforce this...?
I originally was the one to report the black screen happening with Jason's app, I hope something can be done to fix it, thanks.
Same problem with the release of Firefox 57
(In reply to Michael Livote from comment #4)
> (In reply to fireattack from comment #3)
> > I sometimes encounter the same bug in with extension WhatRuns
> > (https://addons.mozilla.org/en-US/firefox/addon/whatruns/), when the popup
> > have a scroll bar (just like Checker Plus for Gmail).
> > 
> > But it's very fast, and cannot be consistently repeated so I can't take a
> > screenshot.
> > 
> > But you can try it yourself on complex websites like https://www.reddit.com
> > which will have a scroll bar.
> 
> This thread is not for that app, please open another ticket if you need to
> report it, thanks....are there no mods on here to enforce this...?

That's not how Bugzilla works.
Priority: -- → P3
Whiteboard: [gfx-noted]
Same problem in latest Developer branch v 58b7. Just updated the graphic drivers, did the restart, nothing changes. Windows 10 v. 1709, x64.
Found a workaround by explicitly placing a white background with css ie. *{background:white}
... perhaps a clue for the investigating team.
We’ve observed the same issue in the Amazon Assistant extension on our Windows machines. After testing Firefox 55, 56, and 57, we found the issue to only be in Firefox 57. We do not observe this issue on MacOS. We’ve observed that some of the artifacts disappear when they are hovered over while others do not. Some come and go based on how far the user has scrolled down (they appear and disappear consistently).
 
We experimented with our CSS to see if there was any effect and found that changing the color property on the body caused the artifacts to flicker. We also tried the reporter’s suggestion of applying a solid background everywhere which, for us, caused more artifacts which disappeared upon scrolling. Once we scrolled the old artifacts reappeared.
 
Adjusting the hardware acceleration settings had no effect nor did pulling down all of the latest updates from Microsoft.
 
We’ve attached a screenshot. Please let us know if there’s anything we can do to help with getting this issue resolved.
(In reply to bit-shared-platform from comment #11)
> We’ve observed the same issue in the Amazon Assistant extension on our
> Windows machines. After testing Firefox 55, 56, and 57, we found the issue
> to only be in Firefox 57. We do not observe this issue on MacOS. We’ve
> observed that some of the artifacts disappear when they are hovered over
> while others do not. Some come and go based on how far the user has scrolled
> down (they appear and disappear consistently).
>  
> We experimented with our CSS to see if there was any effect and found that
> changing the color property on the body caused the artifacts to flicker. We
> also tried the reporter’s suggestion of applying a solid background
> everywhere which, for us, caused more artifacts which disappeared upon
> scrolling. Once we scrolled the old artifacts reappeared.
>  
> Adjusting the hardware acceleration settings had no effect nor did pulling
> down all of the latest updates from Microsoft.
>  
> We’ve attached a screenshot. Please let us know if there’s anything we can
> do to help with getting this issue resolved.

If you guys have steps to easily and consistently reproduce it, doing a Bisection using mozregression (http://mozilla.github.io/mozregression/) will be very helpful to identify the culprit.
Did a bisect myself, using Amazon Assistant for test.

2017-12-13T21:37:16: INFO : Narrowed inbound regression window from [6fa4468d, 66ef885f] (3 builds) to [a4fb7838, 66ef885f] (2 builds) (~1 steps left)
2017-12-13T21:37:16: DEBUG : Starting merge handling...
2017-12-13T21:37:16: DEBUG : Using url: https://hg.mozilla.org/integration/autoland/json-pushes?changeset=66ef885f0b66cb680a3f0734bfbfd8fd385fa104&full=1
2017-12-13T21:37:17: DEBUG : Found commit message:
Bug 1400259 - Add will-change:transform to the panel in order to work around a compositor bug. r=mconley

The bug has something to do with ContainerLayer nesting changes being mishandled:
a new ContainerLayer for the transform is being inserted around the container
layer for the opacity, which has an intermediate surface.
This change makes the outer ContainerLayer permanent so that the dynamic
insertion case is not hit.

MozReview-Commit-ID: lETpsr4YJi

2017-12-13T21:37:17: INFO : The bisection is done.
2017-12-13T21:37:17: INFO : Stopped

Last good: a4fb78381ea0fe9fddd36a8ea7b74a63062a1bf3
First bad: 66ef885f0b66cb680a3f0734bfbfd8fd385fa104

Diff: 
https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=a4fb78381ea0fe9fddd36a8ea7b74a63062a1bf3&tochange=66ef885f0b66cb680a3f0734bfbfd8fd385fa104
Status: UNCONFIRMED → NEW
Ever confirmed: true
Thanks for also running a bisection fireattack! 

We ran two separate bisections and did not see the issue pop up in any version the tool had us test against, but could consistently see and reproduce the issue repeatedly on Firefox 57.0, 57.1, and 57.2 production versions.

Is there anything else we can do at this time to help out?
Correction, we meant 57.0, 57.0.1, & 57.0.2 in the previous comment.
Actually, I have a similar issue. When I was investigating this problem on a different machine (Win 10) earlier, I can't seem to reproduce the bug at all on multiple version of Nightly either (but *can* reproduce on stable v57.0.2).

The above bisection was done on a different, Win 7 machine, which can reliably reproduce the bug on any Nightly after 17-11-02. I am going to retest it again tomorrow when I have my Win 10 laptop on hand.
OK, just tested again with my Win 10 machine. This time, I can reliably reproduce the bug within the same range as I reported above: after changeset 66ef885f0b66cb680a3f0734bfbfd8fd385fa104.

Could someone help add regression window and tracking flags? Thanks!
bug 1425455 indicates that this bug is due to usage of @-moz-document, in which case the bug regressing this should be bug 1035091, though not sure that matches the time this was reported...
(In reply to Emilio Cobos Álvarez [:emilio] from comment #18)
> bug 1425455 indicates that this bug is due to usage of @-moz-document, in
> which case the bug regressing this should be bug 1035091, though not sure
> that matches the time this was reported...

Err, nvm, I see, so this is the bug that forced the work around to use @-moz-document...

Jason, would it be possible for you to insert that stylesheet as a user-sheet in the meantime? @-moz-document still works on user stylesheets.
Flags: needinfo?(apps)
Not sure I understand what you mean by "insert that stylesheet as a user-sheet" but regardless the extension has been working since the workaround so I'm hesitant "fix it" if it ain't broken now.
Flags: needinfo?(apps)
I'm duping this bug to bug #1425455 per more info there like regression range etc.
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → DUPLICATE
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: