1173404 - background-attachment: fixed causes pixels to show up around background image

Status: UNCONFIRMED

(Core :: Graphics, defect)

Description

[pthread1981]

Attachment: Simple example html, image, and screen shot.

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:38.0) Gecko/20100101 Firefox/38.0
Build ID: 20150525141253

Steps to reproduce:

FF 38.0.5 on OS X 10.10.3 (2013 MacBook Air) with "Use hardware acceleration when available" checked. Navigate to https://test-ff-problem.s3.amazonaws.com/index.html or see the attached html/image/screenshot.


Actual results:

Use of "background-attachment: fixed" causes occasional lines (not quite borders - they don't always occupy the length of an edge) along various sides of the background image.  Resizing the window can prompt the appearance/disappearance of the lines.


Expected results:

These lines don't appear.

Comment 1

[pthread1981]

Also, my apologies in advance for the choice of sample image. This issue report isn't trolling, it's just what I had around.  :)

Comment 2

[:Gijs (he/him)]

This seems to be GPU/driver/OS-specific, because I cannot reproduce with my 2013 rMBP (though I guess I'm testing on beta...). Over to graphics. CC'ing mstange and seth in case they are interested / have ideas.

Reporter, for completeness' sake, could you go to about:support (Help > Troubleshooting Information), click "copy text to clipboard", paste somewhere (TextEdit.app, whatever), and select your graphics information only, and put that in a comment here?

Finally, could you test with a newer version of Firefox like Nightly ( https://nightly.mozilla.org/ ) to see if this has been fixed already? I know we've done some work on background-attachment as well as graphics recently. :-)

Thanks!

Comment 3

[:Gijs (he/him)]

... actually doing the CC'ing this time around...

Comment 4

[pthread1981]

Thanks for looking!

Here's the requested graphics information:

Graphics
--------

Device ID: 0x a26
GPU Accelerated Windows: 1/1 OpenGL (OMTC)
Vendor ID: 0x8086
WebGL Renderer: Intel Inc. -- Intel HD Graphics 5000 OpenGL Engine
windowLayerManagerRemote: true
AzureCanvasBackend: quartz
AzureContentBackend: quartz
AzureFallbackCanvasBackend: none
AzureSkiaAccelerated: 0

I also have a coworker who has seen the issue (a Mac also, but not an Air) so I'll see if I can get his graphics information as well and add that.

Comment 5

[pthread1981]

Oh - and Nightly (41.0a1 (2015-06-11)) exhibits the same problem.

Comment 6

[pthread1981]

Coworker's Graphics config (from a mid-2014 15" Retina MacBook Pro running 10.9.5):

Graphics
--------

Device ID: 0x fe9
GPU Accelerated Windows: 1/1 OpenGL (OMTC)
Vendor ID: 0x10de
WebGL Renderer: NVIDIA Corporation -- NVIDIA GeForce GT 750M OpenGL Engine
windowLayerManagerRemote: true
AzureCanvasBackend: quartz
AzureContentBackend: quartz
AzureFallbackCanvasBackend: none
AzureSkiaAccelerated: 0

Comment 7

[:Gijs (he/him)]

(In reply to pthread1981 from comment #6)
> Coworker's Graphics config (from a mid-2014 15" Retina MacBook Pro running
> 10.9.5):
> 
> Graphics
> --------
> 
> Device ID: 0x fe9
> GPU Accelerated Windows: 1/1 OpenGL (OMTC)
> Vendor ID: 0x10de
> WebGL Renderer: NVIDIA Corporation -- NVIDIA GeForce GT 750M OpenGL Engine
> windowLayerManagerRemote: true
> AzureCanvasBackend: quartz
> AzureContentBackend: quartz
> AzureFallbackCanvasBackend: none
> AzureSkiaAccelerated: 0

Huh. I'm stumped. I've been trying to repro this on 10.10, and by turning  my GPU switching back on (turned off because it kept causing system freezes). No luck so far.

Markus, do you have ideas (can you repro?)

Comment 8

[Markus Stange [:mstange]]

I can reproduce it sometimes when I resize the window quickly. I've tried to change the testcase to make reproduction easier but failed.

Comment 9

[pthread1981]

To show it in the wild (outside my initial test case) - this is my client's page that prompted the initial investigation: https://eon.phase5group.com/.  It tends to show up much more consistently here than in my test case. At most window sizes, even if resizing has made it disappear, reloading at that current window size will make it return (for me).

Comment 10

[pthread1981]

Hi - this seems much improved as of 40.0.2 - I still see a bit of flickering on the side during resizing in the linked test case above, but a Window at rest no longer exhibits this behavior. 

I'd be comfortable marking this as resolved if you are.