1275538 - Crash in mozilla::MediaFormatReader::RequestVideoData

Status: RESOLVED FIXED

(Core :: Audio/Video: Playback, defect, P1)

Description

[Marcia Knous [:marcia]]

This bug was filed from the Socorro interface and is 
report bp-51fb7abe-4cc5-4723-9411-4fd6d2160525.
=============================================================

29 crashes reported on Nightly Build 20160524073714 - https://crash-stats.mozilla.com/signature/?build_id=20160524073714&product=Firefox&release_channel=nightly&platform=Windows&signature=mozilla%3A%3AMediaFormatReader%3A%3ARequestVideoData&date=%3E%3D2016-05-24

Bug 1273018 had some similiar signatures in MediaFormatReader. Adding a few folks from that bug.

Comment 1

[Jean-Yves Avenard [:jya]]

this is a consequence of bug 1272964 (P4).

We continue to service request for video data until we now the skip will fail... maybe we should just drop all pending frames at least (we've already stopped the decoder, but it will continue to return already decoded frames)

Comment 2

[Jean-Yves Avenard [:jya]]

Attachment: MozReview Request: Bug 1275538: P1. Abort early if a skip request is in progress. r?kamidphish

Review commit: https://reviewboard.mozilla.org/r/55346/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/55346/

Comment 3

[Jean-Yves Avenard [:jya]]

Attachment: MozReview Request: Bug 1275538: P2. Drop decoded frames that we know are already too late. r?kamidphish

Review commit: https://reviewboard.mozilla.org/r/55348/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/55348/

Comment 4

[Jean-Yves Avenard [:jya]]

Comment on attachment 8756709 [details]
MozReview Request: Bug 1275538: P1. Abort early if a skip request is in progress. r?kamidphish

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/55346/diff/1-2/

Comment 5

[Jean-Yves Avenard [:jya]]

Comment on attachment 8756710 [details]
MozReview Request: Bug 1275538: P2. Drop decoded frames that we know are already too late. r?kamidphish

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/55348/diff/1-2/

Comment 6

[Gerald Squelart (he/him) (not at Mozilla since 2022-09-15)]

Comment on attachment 8756709 [details]
MozReview Request: Bug 1275538: P1. Abort early if a skip request is in progress. r?kamidphish

https://reviewboard.mozilla.org/r/55346/#review52402

Comment 7

[Jean-Yves Avenard [:jya]]

sorry, Dan indicated he wanted to review this, so I'll wait for his feedback.

Comment 8

[u480271]

Comment on attachment 8756710 [details]
MozReview Request: Bug 1275538: P2. Drop decoded frames that we know are already too late. r?kamidphish

https://reviewboard.mozilla.org/r/55348/#review52412

Comment 9

[u480271]

Comment on attachment 8756709 [details]
MozReview Request: Bug 1275538: P1. Abort early if a skip request is in progress. r?kamidphish

https://reviewboard.mozilla.org/r/55346/#review52414

Comment 10

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/dc73bf8a692f
https://hg.mozilla.org/integration/mozilla-inbound/rev/d3428c4506b5

Comment 11

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Will this fix also fix the crashes in RequestAudioData that also spiked starting on May 24 builds?

https://crash-stats.mozilla.com/signature/?product=Firefox&release_channel=nightly&platform=Windows&date=%3E%3D2016-05-15&signature=mozilla%3A%3AMediaFormatReader%3A%3ARequestAudioData

Comment 12

[Jean-Yves Avenard [:jya]]

No it wouldn't. However, all those in those reports in build earlier than 24th have been fixed already 

The problem with those pgo build is that we often can't trust the backtrace, and it doesn't make sense.

The crashes are always because there's a pending audio request promise. Dan any ideas?

Comment 13

[Jean-Yves Avenard [:jya]]

One of the crash in RequestAudioData has a build id of 20160521030227 and another 20160518030234 ; this is right after bug 1273018, but before bug 1272964 which is what is causing this bug.

So obviously bug 1273018 didn't fix the problem of that assertion. So maybe the crash spiking after May 24th is just a coincidence and bug 1272964 has nothing to do with the audio assert. will open a new bug about it

Comment 14

[u480271]

jya, I've found an issue. I'll upload a patch soon.

Comment 15

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/dc73bf8a692f
https://hg.mozilla.org/mozilla-central/rev/d3428c4506b5