[Bug 738363] jitterbuffer: lost-events are broken

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=738363

--- Comment #6 from Håvard Graff (hgr) <havard.graff at gmail.com> ---
The test-failure was a wrong timestamp (overflowed?) in the timestamp of the
lost-event in the large-gap test.

The patch was to reintroduce the correct behavior on large-gap handling of the
lost-event.

The correct behavior for large gaps is to think that when a packet finally
arrives (after some time, creating a large gap), you must allow for up to
jb-latency ms of buffers still arriving, and not "starting again" from the new
packet. Example is that you have received buffers 0 and 1, with 20ms of
duration, and then get buffer 500, you must allow for buffers 491-499 still
being able to arrive (with default 200ms latency), and hence potentially create
lost-events for them if they do not arrive. The test checks that all this is
true, and that both the "large-gap" lost-event as well as the following
"single" lost-events all have the right timestamp and duration. This was badly
broken, and is why I would love for these tests (and the improvements on the
existing ones) to make it upstream to avoid more breakage.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.