Regression in calculating live edge in 2.x vs 1.6.x #490
Labels
status: archived
Archived and locked; will not be updated
type: bug
Something isn't working correctly
Milestone
I have a Manifest that looks something like this:
I downloaded it at at 14:36:56, and looking at the segments available for the stream with id video=4500000, my last video segment ends at 14:36:38 (73567349900 / 50 + (200/50)*24400 = 1471444598).
In Shaka 1.6.x, this stream plays as expected, but in Shaka 2 it will calculate live edge as being after 14:36:38, meaning the stream does not start until a few seconds have passed and the encoder spits out a segment that is at the live edge.
On the demos for 1.6.3 and for 2.0.0-beta3, this is quite apparent.
My initial findings points to the live edge in 1.6 being calculated using computeStreamLimits_ in stream_video_source, which looks at the segments.
In 2.x, the live edge seems to be calculated by getLiveEdge_ which doesn't take the last segment end time into account.
This particular manifest has quite a big time shift buffer, but I can reproduce with a time shift buffer of less than an hour, so it's unrelated as far as I can tell.
I tried writing a unit test, but I did not get very far, sorry.
Is this a problematic manifest that just happened to work on 1.6, or a regression between 1.6 and 2.0?
I can provide links so that you can test this locally, just email me at [email protected]
The text was updated successfully, but these errors were encountered: