Fix flaky test on feature/v1.0.0

[itsjamie]

Flaky test on: should seek 5s from end and receive video ended event for Multiple non-alternate audio levels:

Getting video_error_3.

Originally posted by @itsjamie in #2828 (comment)

[robwalch]

I saw that too on another PR. I don't see any problems with the stream. And I'm not seeing much in the logs for that failure; It seems like decode error occurs before the seek, but it's unclear. Here's what stands out or could use improvement:

InitPTS for cc: 0 found from video track
What video track? I'm assuming this is what the audio-stream-controller logs when InitPTS comes from the stream-controller. In this scenario there is no video in the main track, but that was a former assumption so I could noticed and fixed these log statements. I see the same output when the stream plays so not sure this is related to the issue.

2020-06-23 04:46:08.415: "[log] > InitPTS for cc: 0 found from video track: 259680480"
2020-06-23 04:46:08.416: "[log] > creating sourceBuffer(audio/mp4;codecs=mp4a.40.5)"
2020-06-23 04:46:08.417: "[log] > main track:audio,container:audio/mp4,codecs[level/parsed]=[mp4a.40.5/mp4a.40.5]"

What was the currentTime at the time of the error? The onerror callback fails the test and looks pretty close to the log where we reachedcurrentTime:0.078,bufferEnd:12.800, but maybe some more details would help:

2020-06-23 04:46:08.551: "[log] > Loading 235776262 of [235776259 ,235776278],level 1, currentTime:0.078,bufferEnd:12.800"
2020-06-23 04:46:08.552: "[log] > main stream-controller: IDLE->FRAG_LOADING"
2020-06-23 04:46:08.564: "[test] > video error, code :3"

The worker logs are completely missing. For example locally I see these messages and it would be helpful to know if something odd happened in the worker.

[log] > manifest codec:mp4a.40.5,ADTS data:type:2,sampleingIndex:6[24000Hz],channelConfig:2
[log] > parsed codec:mp4a.40.5,rate:24000,nb channel:2 

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.

[robwalch]

Test runs have been more stable lately. This takes care of the issue mentioned above #2919

[robwalch]

@itsjamie Im not sure the test is flaky. "Multiple non-alternate audio levels" has two different audio sample rates, which I think present some edge cases when switching levels we don't see often.

We get a media error when appending segments out of order after a level switch and buffer flush. There are also cases where it looks like media is not appended in the correctly location - still looking into how that happens.

[itsjamie]

Good to know that it seems like it's illustrating a race. Thanks for looking into it.

[robwalch]

Troubleshooting details can be found here #2921

I merged those improvements for the next release.

We might want to change this ticket or replace it with one that describes the issue with buffering AAC of multiple sample rates.

[robwalch]

I'll close this out once the changes from #2943 are merged into feature/v1.0.0 and functional tests are passing (close enough).