1276408 - black artefacts with layers.offmainthreadcomposition enabled

Status: RESOLVED INCOMPLETE

(Core :: Graphics, defect, P3)

Description

[bluppfisk]

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:46.0) Gecko/20100101 Firefox/46.0
Build ID: 20160502172042

Steps to reproduce:

When scrolling, resizing the browser window or otherwise interacting with the website. It seems random but is quite reproduceable. 

I installed the latest video drivers from Dell (I'm on the late 2015 XPS 13) and Intel. The issue has persisted throughout the various Windows 10 builds but it seems extremely present in the latest (build 14352).

Turning layers.offmainthreadcomposition off fixes the problem, BUT also makes the browser a lot more sluggish.


Actual results:

Black or white artefacts appear all over the browser window, not only in the viewport.


Expected results:

They should not.

Comment 1

[:Gijs (out for now; he/him)]

Can you please copy-paste the 'graphics' section from about:support (Help > Troubleshooting information) into a comment here, and link to a web page or explain a particular situation where this is particularly easy to notice? Thanks!

Comment 2

[bluppfisk]

Dear Gijs.

Problem seems not to occur now but I changed a few things in my configuration:

1. When actually checking for the setting, it showed that my layers.offmainthreadcomposition was automatically set to true again by Firefox (or another actor, but not me). I.e. the artefacts happened anyway; the reason I disabled the off-main thread composition was because I was getting them.

2. I disabled hardware acceleration (FF feels more sluggish)

3. I scaled back my WIn 10 UI scaling from the recommended 150% to 125%.

Below is my graphics troubleshooting info:


Graphics
Adapter Description	Intel(R) HD Graphics 520
Adapter Drivers	igdumdim64 igd10iumd64 igd10iumd64 igd12umd64 igdumdim32 igd10iumd32 igd10iumd32 igd12umd32
Adapter RAM	Unknown
Asynchronous Pan/Zoom	none
Device ID	0x1916
DirectWrite Enabled	false (10.0.14352.1002)
Driver Date	2-18-2016
Driver Version	20.19.15.4390
GPU #2 Active	false
GPU Accelerated Windows	0/2 Basic (OMTC)
Subsys ID	07041028
Supports Hardware H264 Decoding	No; Hardware video decoding disabled or blacklisted
Vendor ID	0x8086
WebGL Renderer	Google Inc. -- ANGLE (Intel(R) HD Graphics 520 Direct3D11 vs_5_0 ps_5_0)
windowLayerManagerRemote	true
AzureCanvasBackend	skia
AzureContentBackend	cairo
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0

Comment 3

[bluppfisk]

And then, they start happening again when I reactivate Hardware Acceleration. I checked and off-main thread composition is still enabled.

Graphics
Adapter Description	Intel(R) HD Graphics 520
Adapter Drivers	igdumdim64 igd10iumd64 igd10iumd64 igd12umd64 igdumdim32 igd10iumd32 igd10iumd32 igd12umd32
Adapter RAM	Unknown
Asynchronous Pan/Zoom	none
Device ID	0x1916
Direct2D Enabled	true
DirectWrite Enabled	true (10.0.14352.1002)
Driver Date	2-18-2016
Driver Version	20.19.15.4390
GPU #2 Active	false
GPU Accelerated Windows	2/2 Direct3D 11 (OMTC)
Subsys ID	07041028
Supports Hardware H264 Decoding	Yes
Vendor ID	0x8086
WebGL Renderer	Google Inc. -- ANGLE (Intel(R) HD Graphics 520 Direct3D11 vs_5_0 ps_5_0)
windowLayerManagerRemote	true
AzureCanvasBackend	direct2d 1.1
AzureContentBackend	direct2d 1.1
AzureFallbackCanvasBackend	cairo
AzureSkiaAccelerated	0

Comment 4

[:Gijs (out for now; he/him)]

Thanks for all the details!

Radar'ing this to the gfx team - Milan, looks like a very new driver to me, so a bit confused about how this is happening...

Comment 5

[bluppfisk]

Gijs,


as additional info, I can confirm I've been having these problems for a long time now, so also with previous versions of the driver. I _believe_, but I am not 100% sure, that this would also occur on the previous edition of the Dell XPS 13 (9343; the current one is 9350) which had Intel HD Graphics 5500 (the current one has Intel HD Graphics 520).

Reason I can't remember is because I switched to Firefox from Chrome relatively shortly before switching to a new XPS 13.

Hope this helps. If solved, this would eliminate all browser competition for me.

Comment 6

[Milan Sreckovic [:milan] (needinfo for best results)]

Yes, we have seen this even with the latest drivers, although the upcoming Windows update seems to improve things drastically, at least for some people.

Wasn't clear - resizing sometimes causes the issues - does it resolve them as well?  Or minimizing and reexpanding the window?

If there is a window behind Firefox when this happens, is the "black" area actually an absence of things, and can you partially see the window behind?

Comment 7

[bluppfisk]

> Wasn't clear - resizing sometimes causes the issues - does it resolve them
> as well?  Or minimizing and reexpanding the window?

Yes. Every pixel of window resizing creates a different pattern of black artefacts and every so often it will also clear all. Minimizing or reexpanding may also damage the GUI elements, such as the menu bar or the tabs. They may turn white and disappear altogether until a redraw is forced on them.

> 
> If there is a window behind Firefox when this happens, is the "black" area
> actually an absence of things, and can you partially see the window behind?


Seems to be on top/below other elements.

Comment 8

[Milan Sreckovic [:milan] (needinfo for best results)]

Bas, see previous comment.

Comment 9

[Bas Schouten (:bas.schouten)]

Could you post some screenshots? And does this device also have a secondary discrete GPU?

Comment 10

[bluppfisk]

Hi Bas, no discrete GPU, only the integrated one.

it seems to occur a lot less in Win10 build 160522-1930, but it's still there (attached GMail screenshot was taken just now). I have not been able to reproduce the glitches in the UI outside the viewport.

I suppose it happens when a viewport resize causes a redraw because of media queries, but sometimes also without resizing the viewport.

Comment 11

[bluppfisk]

Attachment: artefacts in GMail

Comment 12

[bluppfisk]

Attachment: Artefacts on this very page (Bugzilla)

Comment 13

[bluppfisk]

Attachment: Artefacts in the UI

Comment 14

[bluppfisk]

It started happening again in the UI and I shot a screenshot for you guys

Comment 15

[ekerazha]

As I reported here https://bugzilla.mozilla.org/show_bug.cgi?id=837489#c255 I've the same issue. Disabling "offmainthreadcomposition" seems to workaround the issue.

Comment 16

[bluppfisk]

(In reply to ekerazha from comment #15)
> As I reported here https://bugzilla.mozilla.org/show_bug.cgi?id=837489#c255
> I've the same issue. Disabling "offmainthreadcomposition" seems to
> workaround the issue.

Indeed this fixes it, but it also makes the browser quite sluggish.

Comment 17

[:Gijs (out for now; he/him)]

Bas, anything we can do here now that there are screenshots and some more info?

Comment 18

[Milan Sreckovic [:milan] (needinfo for best results)]

We are uncovering driver issues - there were changes in Windows 8 through 10 on how the graphics memory is managed between the system and the drivers, and our scenario of dealing with a lot of dynamic allocations is the case they didn't quite get to debugging.  We're working with vendors to improve the drivers, and perhaps tell us what we could avoid doing to trigger these problems.

Comment 19

[bluppfisk]

Hopefully but carefully adding that these artefacts issues _seem_ to be solved by today's driver update for the Intel Graphics 520. Will write if it comes back.

Comment 21

[[:philipp]]

there is an increase of reports about this issue on various support channels after the firefox 49 update (from intel hd 520 users with older drivers).

Comment 22

[Milan Sreckovic [:milan] (needinfo for best results)]

Possibly the same issue as bug 1304360?  If we have access to the users, it would be good if they can test if setting layers.allow-d3d9-fallback to false (in about:config) fixes the problem.

Comment 23

[Andrei Purice]

Hey bluppfisk,
Can you still reproduce this issue or should we close it?

Comment 24

[Andrei Purice]

Marking this as Resolved > Incomplete as per reporter's lack of response.
If anyone can still reproduce this issue re-open it or file a new bug.

Comment 25

[bluppfisk]

I was locked out of my account due to losing my 2FA so I didn't get back to you. That type of artefacts has indeed been solved. I have other things happening now (on Wayland/gnome3.36/Ubuntu 21.04 and an XPS 13 9300). More like blank triangles in the background colour of an element (also in the browser's chrome, e.g. on the tabs). These triangles go away when triggering a change there (e.g. by hovering the mouse over it).

I should probably open a new bug for that.