Reynaldo H. Verdejo Pinochet [Fri, 27 Mar 2015 22:28:05 +0000 (19:28 -0300)]
aggregator: fix typo in test suite
Reynaldo H. Verdejo Pinochet [Fri, 27 Mar 2015 21:32:27 +0000 (18:32 -0300)]
aggregator: add gap event handling unit test
https://bugzilla.gnome.org/show_bug.cgi?id=746249
Reynaldo H. Verdejo Pinochet [Wed, 18 Mar 2015 01:13:06 +0000 (22:13 -0300)]
aggregator: implement gap handling
https://bugzilla.gnome.org/show_bug.cgi?id=746249
Olivier Crête [Thu, 2 Apr 2015 02:10:11 +0000 (22:10 -0400)]
aggregator: Unify downstream flow return and flushing
Also means that having a non-OK downstream flow return
wakes up the chain functions.
https://bugzilla.gnome.org/show_bug.cgi?id=747220
Olivier Crête [Thu, 2 Apr 2015 01:45:01 +0000 (21:45 -0400)]
aggregator: Flushing is always in pad lock, no need to atomics
The usage of atomics was always doubtful as it was used to release a
GCond
https://bugzilla.gnome.org/show_bug.cgi?id=747220
Olivier Crête [Thu, 2 Apr 2015 01:38:11 +0000 (21:38 -0400)]
aggregator: Reset pending_eos on pad flush
https://bugzilla.gnome.org/show_bug.cgi?id=747220
Olivier Crête [Thu, 2 Apr 2015 01:37:25 +0000 (21:37 -0400)]
aggregator: Unify code to set a pad flushing
https://bugzilla.gnome.org/show_bug.cgi?id=747220
Olivier Crête [Sat, 7 Mar 2015 02:12:52 +0000 (21:12 -0500)]
aggregator: Query latency on first incoming buffer.
And keep on querying upstream until we get a reply.
Also, the _get_latency_unlocked() method required being calld
with a private lock, so removed the _unlocked() variant from the API.
And it now returns GST_CLOCK_TIME_NONE when the element is not live as
we think that 0 upstream latency is possible.
https://bugzilla.gnome.org/show_bug.cgi?id=745768
Olivier Crête [Sat, 7 Mar 2015 02:12:13 +0000 (21:12 -0500)]
aggregator: Be more aggressive with invalid replies to our latency query
https://bugzilla.gnome.org/show_bug.cgi?id=745768
Matthew Waters [Sat, 7 Mar 2015 15:04:11 +0000 (02:04 +1100)]
aggregatory: don't redefine GST_FLOW_CUSTOM_SUCCESS
Arun Raghavan [Thu, 26 Feb 2015 18:56:00 +0000 (00:26 +0530)]
aggregator: Use standard upstream latency querying logic
The same functionality is duplicated in the default latency querying
now.
Olivier Crete [Fri, 20 Feb 2015 02:21:56 +0000 (21:21 -0500)]
aggregator: Use src_lock to protect latency related members
One has to use the src_lock anyway to protect the min/max/live so they
can be notified atomically to the src thread to wake it up on changes,
such as property changes. So no point in having a second lock.
Also, the object lock was being held across a call to
GST_ELEMENT_WARNING, guaranteeing a deadlock.
Olivier Crête [Thu, 19 Feb 2015 23:53:32 +0000 (18:53 -0500)]
aggregator: Remove untrue comment
Olivier Crête [Thu, 19 Feb 2015 23:30:35 +0000 (18:30 -0500)]
aggregator: Don't try to push tags while flush seeking
The downstream segment could have been flushed already, so
need to re-send the segment event before re-sending the tags.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Sebastian Dröge [Thu, 19 Feb 2015 09:04:28 +0000 (11:04 +0200)]
aggregator: Use the sinkpads iterator directly to query upstream latencies
While gst_aggregator_iterate_sinkpads() makes sure that every pad is only
visited once, even when the iterator has to resync, this is not all we have
to do for querying the latency. When the iterator resyncs we actually have
to query all pads for the latency again and forget our previous results. It
might have happened that a pad was removed, which influenced the result of
the latency query.
Sebastian Dröge [Thu, 19 Feb 2015 08:57:09 +0000 (10:57 +0200)]
aggregator: Move gst_aggregator_get_latency_unlocked() a bit
It was between another function and its helper function before, which was
confusing when reading the code as it had nothing to do with the other
functions.
Sebastian Dröge [Wed, 18 Feb 2015 23:28:06 +0000 (01:28 +0200)]
aggregator: Fail the latency query if one of the upstream queries fails
Olivier Crête [Wed, 18 Feb 2015 20:53:53 +0000 (15:53 -0500)]
aggregator: Document locking order
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Wed, 18 Feb 2015 20:11:14 +0000 (15:11 -0500)]
aggregator: Rename confusinly named SRC_STREAM_LOCK macros to SRC_LOCK
This will match the name of the lock itself. It is also not a stream
lock as it not recursive and not held while pushing.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Wed, 18 Feb 2015 20:06:01 +0000 (15:06 -0500)]
aggregator: Rename confusingly named stream lock to flush lock
This lock is not what is commonly known as a "stream lock" in GStremer,
it's not recursive and it's taken from the non-serialized FLUSH_START event.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Wed, 18 Feb 2015 20:04:04 +0000 (15:04 -0500)]
aggregator: Fix macro indendation
Changes no code
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Tim-Philipp Müller [Fri, 13 Feb 2015 23:45:20 +0000 (23:45 +0000)]
aggregator: drop GAP events until we handle them properly
Tim-Philipp Müller [Fri, 13 Feb 2015 15:53:19 +0000 (15:53 +0000)]
aggregator: use new gst_aggregator_pad_drop_buffer()
Tim-Philipp Müller [Fri, 13 Feb 2015 15:49:50 +0000 (15:49 +0000)]
aggregator: add gst_aggregator_pad_drop_buffer()
steal_buffer() + unref seems to be a wide-spread idiom
(which perhaps indicates that something is not quite
right with the way aggregator pad works currently).
Tim-Philipp Müller [Thu, 12 Feb 2015 13:32:39 +0000 (13:32 +0000)]
aggregator: only post latency message if anything changed
Perhaps we should check for element state as well and
only post it if in PLAYING state.
Sebastian Dröge [Wed, 11 Feb 2015 13:16:21 +0000 (14:16 +0100)]
Improve and fix LATENCY query handling
This now follows the design docs everywhere, especially the maximum latency
handling.
https://bugzilla.gnome.org/show_bug.cgi?id=744106
Sebastian Dröge [Tue, 10 Feb 2015 09:49:16 +0000 (10:49 +0100)]
aggregator: Pause srcpad task on flow errors
Otherwise we will call the task function over and over again until
upstream finally handled the flow return and shuts us down.
Sebastian Dröge [Fri, 6 Feb 2015 09:59:27 +0000 (10:59 +0100)]
aggregator: Streamline latency calculations
Min latency can never be invalid, latency property can never be invalid
either. So no need to check for all these things in various places.
Sebastian Dröge [Fri, 6 Feb 2015 09:36:28 +0000 (10:36 +0100)]
aggregator: If upstream has no max latency but the subclass has, take the subclass max latency
Sebastian Dröge [Fri, 6 Feb 2015 09:33:59 +0000 (10:33 +0100)]
aggregator: Fix min>max latency error check
We have to include the upstream latency, our own latency and the subclass
latency in the calculations.
FIXME: This is still not entirely correct
Sebastian Dröge [Fri, 6 Feb 2015 09:30:59 +0000 (10:30 +0100)]
aggregator: Don't add the latency property to the max latency
It has no meaning for the max latency and is only used to increase the min
latency.
Thibault Saunier [Mon, 26 Jan 2015 16:06:29 +0000 (17:06 +0100)]
aggregator: Cleanup locking around AggregatorPad flush related fields
And document the locking
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Mathieu Duponchelle [Mon, 26 Jan 2015 12:11:05 +0000 (13:11 +0100)]
aggregator: keep chain functions as dumb as possible.
+ A pad chain function has no business checking other pads,
that's what the aggregate thread is for.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Thibault Saunier [Mon, 26 Jan 2015 10:32:47 +0000 (11:32 +0100)]
aggregator: More fixes around locking when accessing protected private fields
In some more places we were accessing GstAggregator->segment
and GstAggregator->seqnum without holding the GST_OBJECT_LOCK
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Thibault Saunier [Mon, 26 Jan 2015 10:29:08 +0000 (11:29 +0100)]
aggregator: Make the PAD_LOCK private
Instead of using the GST_OBJECT_LOCK we should have
a dedicated mutex for the pad as it is also associated
with the mutex on the EVENT_MUTEX on which we wait
in the _chain function of the pad.
The GstAggregatorPad.segment is still protected with the
GST_OBJECT_LOCK.
Remove the gst_aggregator_pad_peak_unlocked method as it does not make
sense anymore with a private lock.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Thibault Saunier [Mon, 26 Jan 2015 10:25:54 +0000 (11:25 +0100)]
aggregator: Hide GstAggregatorPad buffer and EOS fileds
And add a getter for the EOS.
The user should always use the various getters to access
those fields
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Wed, 21 Jan 2015 23:41:43 +0000 (18:41 -0500)]
aggregator: Document locking of GstAggregatorPrivate members
Most of them are protected by the object lock, specify
which ones use a different lock.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Wed, 21 Jan 2015 23:47:09 +0000 (18:47 -0500)]
aggregator: Document how the segment is protected
Document that it can only be accessed with the object lock.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Thu, 22 Jan 2015 00:44:57 +0000 (19:44 -0500)]
aggregator: Protect all latency related members with the object lock
The locking was not consistent, now consistently use the object lock.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Thu, 22 Jan 2015 00:43:12 +0000 (19:43 -0500)]
aggregator: Document locking for gst_aggregator_get_latency_unlocked()
Renamed it to _unlocked() to make it clear.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Thu, 22 Jan 2015 00:35:25 +0000 (19:35 -0500)]
aggregator: Protect the srcpad caps negotiation with the stream lock
Instead of adding another lock, use the srcpad stream lock, which is already
taken anyway to push out the new caps if needed.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Thu, 22 Jan 2015 00:33:18 +0000 (19:33 -0500)]
aggregator: Protect the tags with the object lock
The tags related variables were sometimes protected, sometimes not and
sometimes atomic. Put them all under the object lock.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Wed, 21 Jan 2015 23:53:20 +0000 (18:53 -0500)]
aggregator: Consistenly lock the flow_return state
Use the object's lock to protect it.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Wed, 21 Jan 2015 23:45:36 +0000 (18:45 -0500)]
aggregator: Consistently lock some members
Some members sometimes used atomic access, sometimes where not locked at
all. Instead consistently use a mutex to protect them, also document
that.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Wed, 14 Jan 2015 19:38:09 +0000 (14:38 -0500)]
aggregator: Protect exported pad members with the pad's object lock
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Wed, 14 Jan 2015 19:35:15 +0000 (14:35 -0500)]
aggregator: Replace event lock with pad's object lock
Reduce the number of locks simplify code, what is protects
is exposed, but the lock was not.
Also means adding an _unlocked version of gst_aggregator_pad_steal_buffer().
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Olivier Crête [Sat, 10 Jan 2015 03:01:00 +0000 (22:01 -0500)]
aggregator: Protect data with the same mutex as GCond
Whenever a GCond is used, the safest paradigm is to protect
the variable which change is signalled by the GCond with the same
mutex that the GCond depends on.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Nirbheek Chauhan [Wed, 14 Jan 2015 18:17:19 +0000 (23:47 +0530)]
aggregator: Nitpick spacing/punctuation in debug logging
Olivier Crête [Sat, 10 Jan 2015 02:51:40 +0000 (21:51 -0500)]
aggregator: Remove pointless atomic
It is only modified from the streaming thread
Olivier Crête [Sat, 10 Jan 2015 02:30:36 +0000 (21:30 -0500)]
aggregator: Fix query leak
Sebastian Dröge [Fri, 9 Jan 2015 15:43:39 +0000 (16:43 +0100)]
aggregator: Print jitter from clock waiting in the debug logs
Tim-Philipp Müller [Sun, 4 Jan 2015 17:15:37 +0000 (17:15 +0000)]
aggregator: don't use iterator when setting flush pending on pads
Tim-Philipp Müller [Sun, 4 Jan 2015 16:57:05 +0000 (16:57 +0000)]
aggregator: check if pads are ready more efficiently
No need to use an iterator for this which creates a temporary
structure every time and also involves taking and releasing the
object lock many times in the course of iterating. Not to mention
all that GList handling in gst_aggregator_iterate_sinkpads().
Tim-Philipp Müller [Sun, 4 Jan 2015 12:59:19 +0000 (12:59 +0000)]
aggregator: name vfunc arguments consistently
Tim-Philipp Müller [Thu, 1 Jan 2015 15:46:00 +0000 (15:46 +0000)]
aggregator: add g-i transfer and scope annotations
Tim-Philipp Müller [Thu, 1 Jan 2015 14:10:05 +0000 (14:10 +0000)]
aggregator: register names of iterate_sinkpads functions with debug system
Tim-Philipp Müller [Thu, 1 Jan 2015 14:03:02 +0000 (14:03 +0000)]
aggregator: reduce debug messages for taking/releasing logs to TRACE level
Don't spam debug log with this stuff.
Tim-Philipp Müller [Wed, 31 Dec 2014 18:16:21 +0000 (18:16 +0000)]
aggregator: move property member into private structure
Our locking (or lack thereof) while accessing this also
looks generally quite dodgy.
Tim-Philipp Müller [Wed, 31 Dec 2014 14:50:58 +0000 (14:50 +0000)]
aggregator: remove empty dispose function
Tim-Philipp Müller [Tue, 30 Dec 2014 23:58:34 +0000 (23:58 +0000)]
aggregator: give private functions namespace prefix
Especially the GST_DEBUG_FUNCPTR ones.
Tim-Philipp Müller [Wed, 31 Dec 2014 12:35:06 +0000 (12:35 +0000)]
aggregator: fix up some docs comments in header
Tim-Philipp Müller [Tue, 30 Dec 2014 23:44:46 +0000 (23:44 +0000)]
aggregator: remove now-unused system clock member
Tim-Philipp Müller [Tue, 30 Dec 2014 19:22:01 +0000 (19:22 +0000)]
aggregator: make GstAggregatorPadForeachFunc take an GstAggregatorPad
Tim-Philipp Müller [Tue, 30 Dec 2014 17:50:17 +0000 (17:50 +0000)]
aggregator: bring start/stop vfunc semantics in line with other baseclasses
Sub-class should not have to chain up to GstAggregator's start/stop
vfuncs, same as in GstBaseSrc, GstBaseSink, GstBaseTransform etc.
Tim-Philipp Müller [Sun, 28 Dec 2014 18:26:49 +0000 (18:26 +0000)]
aggregator: remove pointless GST_DEBUG_FUNCPTR
Not useful for GObject vfuncs.
Tim-Philipp Müller [Sun, 28 Dec 2014 18:24:21 +0000 (18:24 +0000)]
aggregator: remove duplicate pad parent_class variable
G_DEFINE_TYPE already provides one, just need to use it.
Tim-Philipp Müller [Sun, 28 Dec 2014 18:22:57 +0000 (18:22 +0000)]
aggregator: add _CAST() variants for cast macros
Tim-Philipp Müller [Sun, 28 Dec 2014 01:13:33 +0000 (01:13 +0000)]
aggregator: make padding larger
Esp. the class structures, can't have enough
spare space for virtual functions.
Nirbheek Chauhan [Sat, 27 Dec 2014 10:45:41 +0000 (16:15 +0530)]
aggregator: Log to the pad instead of the element
More correct way of doing the same thing as before
Sebastian Dröge [Sat, 27 Dec 2014 08:49:43 +0000 (09:49 +0100)]
aggregator: Make sure that the minimum latencies are never GST_CLOCK_TIME_NONE
Sebastian Dröge [Sat, 27 Dec 2014 08:42:57 +0000 (09:42 +0100)]
aggregator: Wait for the minimum latency, not the maximum
The minimum latency is the latency we have to wait at least
to guarantee that all upstreams have produced data. The maximum
latency has no meaning like that and shouldn't be used for waiting.
Nirbheek Chauhan [Fri, 26 Dec 2014 22:51:36 +0000 (04:21 +0530)]
aggregator: Clamp the min latency at the max if it's greater
Nirbheek Chauhan [Fri, 26 Dec 2014 22:51:26 +0000 (04:21 +0530)]
aggregator: Print the sinkpad name while logging latency queries
Very useful while debugging.
Nirbheek Chauhan [Fri, 26 Dec 2014 22:49:52 +0000 (04:19 +0530)]
aggregator: Take the stream lock when iterating sink pads
When iterating sink pads to collect some data, we should take the stream lock so
we don't get stale data and possibly deadlock because of that. This fixes
a definitive deadlock in _wait_and_check() that manifests with high max
latencies in a live pipeline, and fixes other possible race conditions.
Sebastian Dröge [Tue, 23 Dec 2014 10:45:05 +0000 (11:45 +0100)]
aggregator: Don't leak flush-start events
Sebastian Dröge [Tue, 23 Dec 2014 09:24:27 +0000 (10:24 +0100)]
aggregator: Also change the default latency to 0, not just the minimum
Sebastian Dröge [Tue, 23 Dec 2014 08:52:20 +0000 (09:52 +0100)]
aggregator: Fix docs and default value of the latency property
Sebastian Dröge [Mon, 22 Dec 2014 21:19:52 +0000 (22:19 +0100)]
aggregator: Also include the subclass latency in the result of the latency query
Sebastian Dröge [Mon, 22 Dec 2014 14:26:37 +0000 (15:26 +0100)]
aggregator: Post a latency message if the value of the latency property changes
Sebastian Dröge [Mon, 22 Dec 2014 14:03:59 +0000 (15:03 +0100)]
aggregator: Wake up the src thread after handling a latency query
Due to changed latencies or changed live-ness we might have to
adjust if we wait on a deadline at all and how long.
Sebastian Dröge [Mon, 22 Dec 2014 14:00:36 +0000 (15:00 +0100)]
aggregator: Don't count the number of times we need to wake up but instead check all conditions for waiting again
This simplifies the code and also makes sure that we don't forget to check all
conditions for waiting.
Also fix a potential deadlock caused by not checking if we're actually still
running before starting to wait.
Sebastian Dröge [Wed, 17 Dec 2014 18:51:32 +0000 (19:51 +0100)]
aggregator: Add function to allow subclasses to set their own latency
For audiomixer this is one blocksize, for videoaggregator this should
be the duration of one output frame.
Sebastian Dröge [Wed, 17 Dec 2014 16:54:09 +0000 (17:54 +0100)]
aggregator: Add a timeout parameter to ::aggregate()
When this is TRUE, we really have to produce output. This happens
in live mixing mode when we have to output something for the current
time, no matter if we have enough input or not.
Tim-Philipp Müller [Tue, 16 Dec 2014 19:49:35 +0000 (19:49 +0000)]
tests: fix aggregator unit test after property renaming
Sebastian Dröge [Tue, 16 Dec 2014 16:33:01 +0000 (17:33 +0100)]
aggregator: Some minor cleanup
Matthew Waters [Fri, 5 Dec 2014 07:19:54 +0000 (18:19 +1100)]
aggregator: make the src pad task drive the pipeline for live pipelines
This removes the uses of GAsyncQueue and replaces it with explicit
GMutex, GCond and wakeup count which is used for the non-live case.
For live pipelines, the aggregator waits on the clock until either
data arrives on all sink pads or the expected output buffer time
arrives plus the timeout/latency at which time, the subclass
produces a buffer.
https://bugzilla.gnome.org/show_bug.cgi?id=741146
Sebastian Rasmussen [Sun, 14 Dec 2014 00:29:26 +0000 (01:29 +0100)]
tests/aggregator: Use correct type when setting property
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=741495
Sebastian Dröge [Wed, 19 Nov 2014 16:17:06 +0000 (17:17 +0100)]
aggregator: Unblock events/queries immediately if the pad is flushing
https://bugzilla.gnome.org/show_bug.cgi?id=740376
Sebastian Dröge [Wed, 19 Nov 2014 16:15:02 +0000 (17:15 +0100)]
aggregator: Drop serialized events/queries if the pad is flushing
https://bugzilla.gnome.org/show_bug.cgi?id=740376
Sebastian Dröge [Wed, 19 Nov 2014 16:03:41 +0000 (17:03 +0100)]
aggregator: Block serialized events/queries until the pad has consumed all buffers
Otherwise the caps of the pad might change while the subclass still works with
a buffer of the old caps, assuming the the current pad caps apply to that
buffer. Which then leads to crashes and other nice effects.
https://bugzilla.gnome.org/show_bug.cgi?id=740376
Sebastian Dröge [Wed, 19 Nov 2014 16:03:33 +0000 (17:03 +0100)]
aggregator: Fix typo in debug output
Matthew Waters [Mon, 17 Nov 2014 03:00:10 +0000 (14:00 +1100)]
aggregator: add _get_latency() for subclass usage
API: gst_aggregator_get_latency
https://bugzilla.gnome.org/show_bug.cgi?id=739996
Vineeth T M [Mon, 20 Oct 2014 12:55:08 +0000 (18:25 +0530)]
audiomixer: critical error for blocksize, timeout min/max values
Audiomixer blocksize, cant be 0, hence adjusting the minimum value to 1
timeout value of aggregator is defined with MAX of MAXINT64,
but it cannot cross G_MAXLONG * GST_SECOND - 1
Hence changed the max value of the same
https://bugzilla.gnome.org/show_bug.cgi?id=738845
Matthew Waters [Tue, 7 Oct 2014 05:57:27 +0000 (16:57 +1100)]
tests/aggregator: add timeout handling test for the timeout parameter
Matthew Waters [Mon, 6 Oct 2014 10:46:24 +0000 (21:46 +1100)]
aggregator: add latency query handling
Matthew Waters [Mon, 6 Oct 2014 07:23:03 +0000 (18:23 +1100)]
aggregator: add a timeout property determining buffer wait time
Determines the amount of time that a pad will wait for a buffer before
being marked unresponsive.
Network sources may fail to produce buffers for an extended period of time,
currently causing the pipeline to stall possibly indefinitely, waiting for
these buffers to appear.
Subclasses should render unresponsive pads with either silence (audio), the
last (video) frame or what makes the most sense in the given context.
Jan Alexander Steffens (heftig) [Wed, 17 Sep 2014 14:48:02 +0000 (16:48 +0200)]
aggregator: Replace GMainContext with GAsyncQueue (v2)
The previous implementation kept accumulating GSources,
slowing down the iteration and leaking memory.
Instead of trying to fix the main context flushing, replace
it with a GAsyncQueue which is simple to flush and has
less overhead.
https://bugzilla.gnome.org/show_bug.cgi?id=736782
Mathieu Duponchelle [Tue, 5 Aug 2014 13:36:30 +0000 (15:36 +0200)]
aggregator: Set seqnum only when segments are received.
Thibault Saunier [Sat, 2 Aug 2014 16:25:01 +0000 (18:25 +0200)]
aggregator: Add a streaming lock so to secure flush start action
Without a lock that is taken in FLUSH_START we had a rare race where we
end up aggregating a buffer that was before the whole FLUSH_START/STOP
dance. That could lead to very wrong behaviour in subclasses.
Thibault Saunier [Fri, 18 Jul 2014 11:58:55 +0000 (13:58 +0200)]
aggregator: Query seeking when a seek failed to see if it was expected
And do not worry if seeking failed on a stream that is not seekable