1372766 - visual pauses playing video on twitter

Status: RESOLVED FIXED

(Core :: Audio/Video: Playback, defect)

Description

[Ben Kelly [:bkelly, not reviewing]]

STR:

1. Open nightly.  I used 56.0a1 (2017-06-13) (64-bit).
2. Go to https://twitter.com/firefox/status/874758953187418112
3. Play the video.
4. Observe that the audio is continuous, but the video has visual pauses of about 1 second in duration.

I took a profile, but don't see anything on the main threads that would explain this:

https://perfht.ml/2tjA99a

Is there a video decode thread or something that we can profile?

This reproduces on nightly for me, but not release FF54.

Comment 1

[Ryan VanderMeulen [:RyanVM]]

[Tracking Requested - why for this release]: ugly video stuttering

INFO: Last good revision: 3ad1b460540d2a6eda0011805457d9701d113f26
INFO: First bad revision: d0fea6559ec29e78eeb2d353975720d18a592b25
INFO: Pushlog:
https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=3ad1b460540d2a6eda0011805457d9701d113f26&tochange=d0fea6559ec29e78eeb2d353975720d18a592b25

30:36.22 INFO: Looks like the following bug has the  changes which introduced the regression:
https://bugzilla.mozilla.org/show_bug.cgi?id=1313398

I'm on Win10 FWIW.

Comment 2

[Ryan VanderMeulen [:RyanVM]]

Forgot to mention that I've reproduced this on Beta tip as well.

Comment 3

[Ben Kelly [:bkelly, not reviewing]]

Seems like a functional regression and not a qf perf issue.

Comment 4

[Julien Cristau [:jcristau]]

tracking this regression for 55/56

Comment 5

[Jean-Yves Avenard [:jya]]

I could reproduce it yesterday, but no longer... it's smooth here (windows 10)

Comment 6

[Jean-Yves Avenard [:jya]]

The issue appears to be with the low res version of the video.

I put a copy there:
https://people-mozilla.org/~jyavenard/mediatest/Firefox%20%F0%9F%A6%8A%F0%9F%8C%8D%20-%20What's%20multi-process%20Firefox-1.mp4

There's something wrong with this video. It plays just as badly in chrome. You can even see the frames being displayed out of order there.

I don't believe the issue is with what we're doing in. the video is poorly encoded.

It used to work out of sheer luck..

Comment 7

[Jean-Yves Avenard [:jya]]

In Edge the video shows small pauses too.

Comment 8

[Jean-Yves Avenard [:jya]]

I'm assuming when the first issue was reported the video was actually even more broken.

Comment 9

[Jean-Yves Avenard [:jya]]

The videos have changed SPS/PPS nal in-band and every few frames... this is why bug 1313398 shows a problem... I do believe this file is invalid just based on how other browsers are playing it.

Comment 10

[Ben Kelly [:bkelly, not reviewing]]

I'm sorry, but I only see pauses in firefox nightly and beta on this video.  It plays smoothly in chrome and edge on my machine.  It also plays smoothly in firefox release.

Are you opening the twitter link in each browser or did you download the video to play it in each browser?  Maybe we are getting served a different version of the video?

Comment 11

[Jean-Yves Avenard [:jya]]

(In reply to Ben Kelly [reviewing, but slowly][:bkelly] from comment #10)
> I'm sorry, but I only see pauses in firefox nightly and beta on this video. 
> It plays smoothly in chrome and edge on my machine.  It also plays smoothly
> in firefox release.
> 
> Are you opening the twitter link in each browser or did you download the
> video to play it in each browser?  Maybe we are getting served a different
> version of the video?

those will be fixed following bug 1374068

BTW, the videos you said give you problems, play awfully in both Edge and Chrome. The only videos that would show problems are invalid to start with.

Comment 12

[Jean-Yves Avenard [:jya]]

(In reply to Ben Kelly [reviewing, but slowly][:bkelly] from comment #10)
> It plays smoothly in chrome and edge on my machine.  It also plays smoothly
> in firefox release.

I responded too quickly to this.

Twitter can serve different content, they have a HD and low-res ones. It's much worse in the low-res one.

Comment 13

[Ben Kelly [:bkelly, not reviewing]]

Thanks!  Can we make sure to get bug 1347068 uplifted to beta?

Comment 14

[Ben Kelly [:bkelly, not reviewing]]

I mean bug 1374068 uplifted to beta.

Comment 15

[Liz Henry (:lizzard) (relman/hg->git project)]

Looks like this was fixed in 55 in bug 1348657.  Ben, can you still reproduce the issue in beta? I can ask QE to verify the fix there if you're busy.

Comment 16

[Ben Kelly [:bkelly, not reviewing]]

Sorry, I think you want bug 1374068 which is only in FF56 so far.

Comment 17

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Bug 1374068 which has the actual fix (waiting to be uplifted to 55) is tracked. Tracking "-" on this one.

Comment 18

[Hsin-Yi Tsai (she/her) [:hsinyi]]

(In reply to Ritu Kothari (:ritu) from comment #17)
> Bug 1374068 which has the actual fix (waiting to be uplifted to 55) is
> tracked. Tracking "-" on this one.

As bug 1374068 has been fixed in 55, change the 55 status of this one to "fixed" accordingly. Feel free to reset if I got something wrong.

Comment 19

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Hi Ben, could you please verify whether this issue is fixed as expected on a latest Nightly build? This will help us decide whether the fix ought to be uplifted to Beta55 or not. Thanks!

Comment 20

[Ben Kelly [:bkelly, not reviewing]]

WFM now.