1151299 - If the first frame decoding failed with WAITING_FOR_DATA, it will never re-attempt to get the first frame.

Status: RESOLVED FIXED

(Core :: Audio/Video, defect)

Description

[Jean-Yves Avenard [:jya]]

The problem appears with patch from bug 1149278 applied.

Following bug 1149278, we never do blocking read with MSE. If a decode error occurs for the first frame, it seems that it never attempts to decode the first frame again, leading to stalls.

Comment 1

[Jean-Yves Avenard [:jya]]

Attachment: Part1. Only attempt to decode first frame when available

Only attempt to decode the first frame when we know we have data available, and don't assume that the first decoder in our array is the one with the first sample.

Comment 2

[Jean-Yves Avenard [:jya]]

Attachment: Part2. Clear EOS flag when new data is received

Once a demuxer is marked as EOS it will be forever (or until the MP4Reader is flushed). More data may be added following a EOS. This prevent a stall under some circumstances (partial segment received)

Comment 3

[Jean-Yves Avenard [:jya]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=2acf9a7407f7

Comment 4

[Jean-Yves Avenard [:jya]]

remote:   https://hg.mozilla.org/integration/mozilla-inbound/rev/fdd90d58cb14
remote:   https://hg.mozilla.org/integration/mozilla-inbound/rev/86efbe39200a

Comment 5

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/fdd90d58cb14
https://hg.mozilla.org/mozilla-central/rev/86efbe39200a

Comment 6

[Chris Peterson [:cpeterson]]

Jean-Yves: should we uplift his fix to Beta 38 for Adobe's EME testing?

Comment 7

[Chris Pearce [:cpearce (Not reading bugmail)]]

We should uplift this for EME.

Comment 8

[Chris Pearce [:cpearce (Not reading bugmail)]]

Comment on attachment 8588823 [details] [diff] [review]
Part1. Only attempt to decode first frame when available

Approval Request Comment
[Feature/regressing bug #]: EME+MSE, bug 1149278.
[User impact if declined]: We need bug 1149278 to ship EME, and to make MSE safer, but the patch in there exposes a new bug where we can stall playback. Without the patches here, we can stall playback.
[Describe test coverage new/current, TreeHerder]: Lots of MSE mochitests
[Risks and why]: Seems low; fixes a stall.
[String/UUID change made/needed]: None

Comment 9

[Chris Pearce [:cpearce (Not reading bugmail)]]

Comment on attachment 8588824 [details] [diff] [review]
Part2. Clear EOS flag when new data is received

Approval Request Comment
[Feature/regressing bug #]: EME+MSE, bug 1149278.
[User impact if declined]: We need bug 1149278 to ship EME, and to make MSE safer, but the patch in there exposes a new bug where we can stall playback. Without the patches here, we can stall playback.
[Describe test coverage new/current, TreeHerder]: Lots of MSE mochitests
[Risks and why]: Seems low; fixes a stall.
[String/UUID change made/needed]: None

Comment 10

[Sylvestre Ledru [:Sylvestre]]

Comment on attachment 8588823 [details] [diff] [review]
Part1. Only attempt to decode first frame when available

Should be in 38 beta 3.

Comment 11

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/76f7ce361ada
https://hg.mozilla.org/releases/mozilla-aurora/rev/746b67d80253

Comment 12

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/releases/mozilla-beta/rev/68f61e9c41d2
https://hg.mozilla.org/releases/mozilla-beta/rev/01cf08a90d44