[Bug 783754] splitmuxsink: Added new async-finalize mode

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=783754

--- Comment #12 from Vivia Nikolaidou <vivia at ahiru.eu> ---
(In reply to Jan Schmidt from comment #10)
> Review of attachment 354845 [details] [review]:
> 
> ::: gst/multifile/gstsplitmuxsink.c
> @@ +1485,3 @@
> +                NULL);
> +            gst_element_call_async (sink, _lock_and_set_to_null, splitmux,
> +                NULL);
> 
> I don't think these calls should be async. At this point the muxer and sink
> should already be EOS, so it shouldn't take long to reach NULL
> synchronously. Otherwise, there's no guarantee the elements still exist by
> the time the callback happens since there's no ref held.

They should be async, at least for the sink - because we're at the callback of
the EOS message held by the sink. Added a comment and a ref.

Also fixed some small races (most notably, there was no guarantee anymore that
the reference context would have a buffer at the start of a new fragment, so
format-location-full might not have a buffer attached to it - fixed that), and
added automated tests.

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