1562218 - [Intermittent] PiP window displays only a white border instead of a video

Status: RESOLVED WORKSFORME

(Toolkit :: Video/Audio Controls, defect, P2)

Description

[Timea Cernea [:tbabos][inactive]]

Attachment: 2019-06-27_14h24_37.png

Affected versions
Nightly 69.0a1 (2019-06-27) (64-bit)

Affected platforms
Windows 10 x64

Steps to reproduce
have 2 monitors

• Launch Firefox
• Go to any page with video content, such as Youtube
• Move the Firefox window on the second monitor
• Open anything on the main monitor, such as a different browser or etc.
• Toggle PiP for the video

Expected Result:
PiP is correctly displayed for the video

Actual Result:
Very Intermittent: there will be only a white border displayed on the main monitor where the PiP window should be. Check attached screenshot.

• The sound of the video is played back and the controls are working but not visible.

Comment 2

[Timea Cernea [:tbabos][inactive]]

Attachment: Nightly71

Screenshot in Nightly71

Comment 3

[Timea Cernea [:tbabos][inactive]]

Reproduced on latest Nightly 71.0a1 (2019-09-16) (64-bit) on Windows 10 x64 on the first PiP toggle click. Investigating steps to repro, still very intermittent.

Comment 4

[Mike Conley (:mconley) (:⚙️)]

Just making sure you're aware of this one, astevenson.

Comment 5

[Mike Conley (:mconley) (:⚙️)]

Attachment: Video frames in top left

I got into this state, and then I resized the player window, and then noticed the video playing in the top left corner! (The resized window doesn't draw correctly - the transparent white rectangle in the corner was what the window looked like before I resized it - it didn't redraw after the resize was complete).

Comment 6

[Mike Conley (:mconley) (:⚙️)]

I can reproduce this often enough, and it's a disappointing enough result, that I'm upgrading its importance.

Comment 7

[BugBot [:suhaib / :marco/ :calixte]]

Change the status for beta to have the same as nightly and release.
For more information, please visit auto_nag documentation.

Comment 8

[Mike Conley (:mconley) (:⚙️)]

Barret is able to reproduce this quite often, and has done so a few days in a row, so cc'ing. Also cc'ing mstange who helped poke at Barret's machine while in the broken state.

Hey sotaro, mstange suggested you as someone who might have a sense of what's going wrong here. Specifically, the screenshot in comment 5 got mstange's attention: the video frames are scaled to the right size, but their position is offset as if the window (before resizing) was originally at (0,0), and then the rectangle where the video becomes visible is the intersection between the actual window and the window had it been at (0,0) before resizing.

We also noticed that this bug also occurs for the chrome of the window that Gecko draws in the parent process (we confirmed this by adding a border around the remote <browser> which renders in the parent process).

Is it possible that the compositor is confused about where the window is that it's supposed to be rendering into? Do you have any ideas of what we could do to debug this?

Comment 9

[Sotaro Ikeda [:sotaro]]

We also noticed that this bug also occurs for the chrome of the window that Gecko draws in the parent process (we confirmed this by adding a border around the remote <browser> which renders in the parent process).

:mconley, can you explain more about it?

Is it possible that the compositor is confused about where the window is that it's supposed to be rendering into? Do you have any ideas of what we could do to debug this?

If layout provides wrong information, it could happen.

Comment 10

[Mike Conley (:mconley) (:⚙️)]

I believe Gankra has been tackling a related issue in bug 1573710, so hopefully when that's sorted out, this one will get sorted at the same time.

Comment 11

[Mike Conley (:mconley) (:⚙️)]

Looks like Gankra has a bead on the issue - see bug 1573710 comment 28.

Comment 12

[Mike Conley (:mconley) (:⚙️)]

A fix for bug 1573710 is on autoland. When that closes, barret, do you think you could test to see if you can reproduce? If not, we can probably close this one out as a dupe.

Comment 13

[Beth Rennie [:beth] (she/her)]

The relevant patch got backed out so I can't test this yet.

Comment 14

[Sotaro Ikeda [:sotaro]]

:mconley, can you check if the problem is addressed on latest nightly?

Comment 15

[Mike Conley (:mconley) (:⚙️)]

Over to barret, who was able to reproduce this pretty regularly.

Comment 16

[Beth Rennie [:beth] (she/her)]

Great news, I cannot get this to reproduce any longer! I opened about 30 PIP windows in a long-running Fx process and none of them exhibited the bug. Previously, I could get it to reproduce in a long-running process within 5 or so attempts.

Comment 17

[Mike Conley (:mconley) (:⚙️)]

Hey sotaro,

Looks like the fix works! Can I assume that it's unlikely that bug 1573710 will be uplifted to beta (71)?

Comment 18

[Sotaro Ikeda [:sotaro]]

Great! I am going to nominate it to beta today.