Open Bug 1521777 Opened 7 years ago Updated 3 years ago

Memory of decoded audio is not freed when a sound card / driver is missing

Categories

(Core :: Audio/Video: Playback, defect, P3)

64 Branch
defect

Tracking

()

UNCONFIRMED

People

(Reporter: mhoppner, Unassigned)

Details

User Agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Ubuntu Chromium/70.0.3538.110 Chrome/70.0.3538.110 Safari/537.36

Steps to reproduce:

Platform: Windows 10 1803 64bit and Window 7 64bit
Firefox: 64.0.2
No sound card / driver installed

Playback of a long video on demand or a live video stream (html5: hls, dash, ...). One can take for example https://video-dev.github.io/hls.js/stable/demo/

Actual results:

The memory footprint of firefox increases until the playback stops or the page is reloaded or the page is closed. This leads to crashes because of running out-of-memory in case of a continuous playback of a live stream. Looking at the output of about:memory reveals that

explicit allocations -- explicit - media - decoded - audio

increases continuously and is not freed. This only happens if no audio device is present (disable for example all audio devices in the device manager).

Expected results:

The audio buffer should not exceed a maximum size.

Summary: Memory of encoded audio is not freed when a sound card is missing → Memory of decoded audio is not freed when a sound card is missing
Summary: Memory of decoded audio is not freed when a sound card is missing → Memory of decoded audio is not freed when a sound card / driver is missing
Component: Untriaged → Audio/Video: cubeb
Product: Firefox → Core

Paul, I'm marking this a P3 since it seems like it is not a common case, but please adjust if you disagree.

Rank: 20
Flags: needinfo?(padenot)
Priority: -- → P3

Probably an easy fix, and probably quite common in enterprise situation. Alastor, do you think you'll have cycles for this at some point?

Component: Audio/Video: cubeb → Audio/Video: Playback
Flags: needinfo?(padenot) → needinfo?(alwu)

(In reply to Paul Adenot (:padenot) from comment #2)

Probably an easy fix, and probably quite common in enterprise situation. Alastor, do you think you'll have cycles for this at some point?

@Paul Adenot: agreed, especially when running in a virtual environment like vmWare ESXi, where we discovered the bug.

Sure, I'll take a look later, keep NI.

Flags: needinfo?(alwu)
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.