[Bug 754851] aggregator: Use the whole aggregator latency for the queue size, not just the latency property

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=754851

--- Comment #2 from Olivier Crête <olivier.crete at ocrete.ca> ---
Created attachment 311101
  --> https://bugzilla.gnome.org/attachment.cgi?id=311101&action=edit
That feels wrong, because the upstream latency is time where the upstream
elements have already queued the buffer, so we're effectively adding a queue
with max-size-time=upstream-latency. The attached patch basically updated the
max in our response to ref

I wonder if the problem is just about CPU scheduling, the "latency" from the
latency property is basically how much leeway the scheduler has before the
source thread has been scheduled enough to give us buffers and I think this is
kind of arbitrary (it's a deadline, not something inherent to the pipeline),
but I'd love to be proven wrong. But if I am right, then maybe it is
independant of the output buffer size, the subclass latency or the upstream
latency.

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