1382003 - Viewport got black

Status: UNCONFIRMED

(Core :: Graphics, defect, P3)

Description

[Eugene]

Attachment: Browser window screenshot

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/59.0.3071.115 Safari/537.36

Steps to reproduce:

Possible preconditions:
* Use a laptop with switchable graphics (Intel HD Graphics 6000 + NVIDIA GeForce 920M)
* Using NVidia Control Panel -- set Firefox to use NVidia GPU.
* PC was turning asleep several times with Firefox launched

Environment:
* Windows 10
* Firefox 54.0.1 (32-bit)

STR:
1. Watching Youtube video at 720p


Actual results:

In the middle of Youtube video, the viewport (Note: top bar (including bookmarks) and bottom bars (search) are visible) turned black.
Cursor type is changing properly depending on content
New tabs and windows have the problem

Firefox `Most used sites` page is displayed properly

Restarting Firefox solved the problem.

Text presentation of Troubleshooting information can be provided if necessary


Expected results:

Content is properly rendered

Comment 1

[Kohei Yoshino [:kohei]]

Please make sure your graphics driver is up to date: https://support.mozilla.org/kb/upgrade-graphics-drivers-use-hardware-acceleration

Comment 2

[Eugene]

(In reply to Kohei Yoshino [:kohei] from comment #1)
> Please make sure your graphics driver is up to date:
> https://support.mozilla.org/kb/upgrade-graphics-drivers-use-hardware-
> acceleration

Active Driver version was: 375.70: 10/28/2016
Updated. Will notify if issue is not resolved

Comment 3

[Eugene]

Problem was not resolved after GPU driver update

Comment 4

[Eugene]

Attachment: Hover behavior

Comment 5

[Eugene]

PC sleep during Firefox launched may not be a precondition.
Black screen occurred again 2 times within 10 minutes.
Notice: Both times there was a same set of resources within tabs (Might be correlated to content)
Notice 2: One tab was 'infinitely loading'

Comment 6

[Sotaro Ikeda [:sotaro]]

Eugene, can you go to about:support, click "Copy text to clipboard" and paste the result into here? Thanks!

Comment 7

[Eugene]

Attachment: Troubleshhot info during problem

Comment 8

[Eugene]

Attachment: Troubleshooting info during third problem occurence (Text)

Comment 9

[Eugene]

Attachment: Troubleshooting info during third problem occurence (Raw)

Comment 10

[Eugene]

Attachment: Latest troubleshooting info (no problem occured during its capture)

Comment 11

[Sotaro Ikeda [:sotaro]]

Thanks for the info!

Comment 12

[Sotaro Ikeda [:sotaro]]

Failure during problem occurred has the failure log like the following. From it, it seems related to device reset. Bug 1364563 might be related. :kechen, can you comment to this bug?

Failure Log
(#0): CP+[GFX1-]: (gfxWindowsPlatform) Detected device reset: 1
(#4139): CP+[GFX1-]: LayerManager::EndTransaction skip RenderLayer().
(#4140): CP+[GFX1-]: (gfxWindowsPlatform) Detected device reset: 1
(#4141): CP+[GFX1-]: (gfxWindowsPlatform) Finished device reset.
(#4142): CP+[GFX1-]: (gfxWindowsPlatform) Detected device reset: 1
(#4143): CP+[GFX1-]: (gfxWindowsPlatform) Finished device reset.
(#4144): CP+[GFX1-]: LayerManager::EndTransaction skip RenderLayer().
(#4145): CP+[GFX1-]: (gfxWindowsPlatform) Detected device reset: 1
(#4146): CP+[GFX1-]: (gfxWindowsPlatform) Finished device reset.
(#4147): CP+[GFX1-]: LayerManager::EndTransaction skip RenderLayer().
(#4148): CP+[GFX1-]: (gfxWindowsPlatform) Detected device reset: 1
(#4149): CP+[GFX1-]: (gfxWindowsPlatform) Finished device reset.
(#4150): CP+[GFX1-]: LayerManager::EndTransaction skip RenderLayer().
(#4151): CP+[GFX1-]: (gfxWindowsPlatform) Detected device reset: 1
(#4152): CP+[GFX1-]: (gfxWindowsPlatform) Finished device reset.
(#4153): CP+[GFX1-]: LayerManager::EndTransaction skip RenderLayer()

Comment 13

[Kevin Chen[:kechen] (UTC + 8)]

I think it's a duplicate of Bug 1364563
According to the information from about:support, there is no GPU process exists, and the backend of the compositor is still d3d11 when this problem occurred which is the same as Bug 1364563.
This problem is caused by unsynced backend between content and compositor process in the fallback process (UI process is fine since it uses skia at first place).

But the symptom of the bugs are a little different(this shows the black screen, bug 1364563 shows broken content). It would be nice to confirm the problem before we mark this as a duplicate.

Eugene could you reproduce this problem easily? If so could you help us to test it with our nightly build firefox[1]?
Thank you!

[1] nightly.mozilla.org

Comment 14

[Eugene]

Problem does not occur on Nightly (Specifically: 57.0a1 (2017-08-11) (64-bit))