Open Bug 1362558 Opened 7 years ago Updated 2 years ago

Non-starting video makes for constant CPU usage after autoplay has been disabled

Categories

(Core :: Audio/Video: Playback, defect, P2)

defect

Tracking

()

UNCONFIRMED

People

(Reporter: mirh, Unassigned)

References

()

Details

(Keywords: stale-bug)

Attachments

(1 file)

Assuming I shouldn't instead open a bug report because the video isn't playing in the first place (with media.autoplay.enabled=false).. 
I get constant 20-30% cpu usage, even though content never starts - https://perfht.ml/2pPjCcG

Moreover, once I have installed profiler Firefox seems to hang every now and then while browsing the website, but I couldn't reliably reproduce this.
Component: Untriaged → Audio/Video: Playback
Product: Firefox → Core
Can you give the link that demonstrates the issue?
Flags: needinfo?(mirh)
It's in the details.
Flags: needinfo?(mirh)
I can't repro the issue on Linux.

Can you
1. set media.autoplay.enabled to true
2. press 'pause' after playback starts for a few seconds
3. observe the CPU usage?

Thanks!
Flags: needinfo?(mirh)
I'm on Windows by the way. 
Anyway, CPU usage quite rapidly disappears once I pause it in this scenario
Flags: needinfo?(mirh)
I cannot repro this issue on Windows Nightly, Beta, and Release. 

@mirh, can you still repro it on your side? 
If it is reproducible, could you please record the performance of "MediaPDecoder" and "MediaPlayback" threads? Also, could you please install the "about:media" add-on and report the information?

Thanks.
Flags: needinfo?(mirh)
Attached file about:media
https://perfht.ml/2qTcXOe
Flags: needinfo?(mirh)
:mirh, thanks for recording the new profile, and I am sorry for my late response. 

I have a look into it today and here is an update.

1. Media related threads are not working at all. All 4 playback threads and 4 decoder threads are pending. From about:media, there's no decoder at all. 

2. The main thread of the content process, which loads the https://video.vice.com/, is somewhat busing. However, we're not able to figure out what is it focusing on because the address-function-name-mapping fails on the provided profile (https://perfht.ml/2qTcXOe).

A question, are you using Nightly? If not, could you try to reproduce it again on Nightly?
I was using latest nightly available on that day, yes.
Then, I have no idea why the address-function-name-mapping doesn't work......

:Blake, looks like the constant CPU usage does not come from media code (see comment 7). 
Could you please transfer this bug to a more generic component?
Flags: needinfo?(bwu)
Kanru, 
Can someone from your team help us to narrow down which component is more suspicious?
Flags: needinfo?(bwu) → needinfo?(kchen)
The profile shows that the content process (1 of 2) keeps repainting, probably layout a more suitable component. But I can't reproduce this on my Windows 10 with Nightly 56.0a1 20170621102301, better make sure this is reproducible before diving in.
Flags: needinfo?(kchen)
This is a P1 bug without an assignee. 

P1 are bugs which are being worked on for the current release cycle/iteration/sprint. 

If the bug is not assigned by Monday, 28 August, the bug's priority will be reset to '--'.
Keywords: stale-bug
Mass change P1->P2 to align with new Mozilla triage process
Priority: P1 → P2
Ok, then.. 58b4 still kind of exhibit the problem (though 10-20 constant cpu usage now). 
Latest nightly on the other hand is more like 2-3%, which albeit pretty close 0.. I dunno, means something could still be done? Or perhaps that should be expected with the player doing something when attempting to start. 

So I thought to just lower severity of bug for now. 

Also, in the meantime, bug 1382574 seems about fixing the "doesn't play in the first place" issue.
Severity: normal → minor
Severity: minor → S4
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: