truncated-playback-and-restart__splice_main_splice_ad.html failed in London plugfest

[jpiesing]

truncated-playback-and-restart__splice_main_splice_ad.html failed on all 8 TVs in the London plugfest.

• On 5 of 8 it timed out without running
• On the 3 of 8 where it ran, the following observations all failed.
  • [OF] Every video frame S[k,s] shall be rendered and the video frames shall be rendered in increasing presentation time order.
  • [OF] Video: The playback duration shall match the duration of the CMAF Track
  • [OF] Video start-up delay: The start-up delay for second presentation should be sufficiently low.

[yanj-github]

I have noticed the duplication happens at the truncating. The playback displayed from frame number 1 to frame number 128, then 'representation_change' detected from the status. It displays frame 1 to 37 of the 1st presentation which is duplicated. Then it changed to 2nd presentation and started from frame 49.

[jpiesing]

Here is an analysis of this test on another device.

Basically the end of the second presentation is not output. Some possible reasons I can think why this might happen are;

• The data never appended;
• The buffer underflows and decoding stops because the TV doesn't know that the end of the sequence has been reached.

I assume the second of these is more likely than the first.

Looking at the source code, test.done() is called in response to the "ended" event being fired. If the buffer underflowed and decoding stopped with frames still in the buffer because the TV didn't know the end of the content had been reached then this would be consistent.

MediaSource.endOfStream appears to be called from closeBuffer which in turn is called from closeStream. I can see the call to closeStream for the first presentation but not for the second.

@FritzHeiden Can you investigate when MediaSource.endOfStream is called for the second stream? What am I missing?

[FritzHeiden]

Calling endOfStream is now reported as an event in the QR code and logged to the console which is saved along with the results. #174 #175

[jpiesing]

2024-05-14: Keeping this one open to track the resolution of the issue.