1137576 - We don't always play media to the end

Status: RESOLVED INVALID

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

Description

[Jean-Yves Avenard [:jya]]

The behaviour appears to be worse when the mediasource isn't in ended state.

For example:
http://people.mozilla.org/~jyavenard/tests/mse_mp4/bipbop.html?eos=1&eosat=-1&duration=-1&init=0

will play 296 frames.

while http://people.mozilla.org/~jyavenard/tests/mse_mp4/bipbop.html?eos=0&eosat=-1&duration=-1&init=0
will play 292 frames.

This is not an issue related to the decoder being used: both Apple VT, ffmpeg and WMF provide the same results (though WMF without EndOfStream being called will only display 269 frames).

In bug 1136138, the provided video will never play the last 4s, stopping at 141s even though we have 145s of buffered audio and video available.

Comment 1

[Jean-Yves Avenard [:jya]]

this is the output of about:media
http://www.dash-player.com/demo/4k/
	mediasource:http://www.dash-player.com/64929df4-937d-6f43-8064-763b94fb3c63
	currentTime: 141.546666
		SourceBuffer 0
			start=93.95833300000001 end=104.08333300000001
			start=110.625 end=124.541667
			start=124.58333300000001 end=144.666667
		SourceBuffer 1
			start=0 end=144.618666
	Internal Data:
	Dumping data for reader 1378a1400:
		Dumping Audio Track Decoders: - mLastAudioTime: 141.546666
			Reader 1: 12f84c800 ranges=[(134.976000, 144.618666)] active=true size=99836
			Reader 0: 120d89000 ranges=[(0.000000, 141.546666)] active=false size=2860472
		Dumping Video Track Decoders - mLastVideoTime: 141.958333
			Reader 5: 130240000 ranges=[(134.958333, 144.666667)] active=true size=1623722
			Reader 4: 12fd3c800 ranges=[(134.958333, 144.666667)] active=false size=3677905
			Reader 3: 1306cd000 ranges=[(124.583333, 134.958333)] active=false size=17087051
			Reader 2: 11ef90000 ranges=[(110.625000, 124.541667)] active=false size=23835477
			Reader 1: 12d0dd800 ranges=[(99.250000, 104.083333)] active=false size=8235381
			Reader 0: 1210a5000 ranges=[(93.958333, 99.291667)] active=false size=10070535
	

we have 144.66 worth of data, but we stop playback at 141.5466

Comment 2

[Jean-Yves Avenard [:jya]]

Ok for that 4K video, the problem is elsewhere. The moof indicates that we have the data, but mdat missing those.
Both chrome and IE stops at 141s too.

so back to bipbop

Comment 3

[Jean-Yves Avenard [:jya]]

With the 2nd stream; we don't call EOS ; as we don't call drain() (and we shouldn't) playback doesn't end nd instead is in "waiting" mode