Problems with renegotiation when using queues

[Jan Schmidt]

On 07/06/14 01:10, Jose Antonio Santos Cadenas wrote:
> Hi Sebastian,

Hi Jose,

> The problem is not with the content of the buffers, thats something that
> can be solved easily, the problem is that the source raises a not
> negotiated error to the bus and stops the streaming thread. It seems
> that the source pad of the queue has a problem sending the sticky events
> to its peer (the one that sends the reconfigure event):
>
> 0:00:00.031095032  3140      0x2547450 WARN                GST_PADS
> gstpad.c:3726:gst_pad_peer_query:<queue1:src> could not send sticky events
> .......
> 0:00:00.032475711  3140      0x2613b50 ERROR       negotiation_test
> negotiation_test/main.c:130:bus_message: Error: error message:
> 0x7fd114002280, time 99:99:99.999999999, seq-num 91, element
> 'audiotestsrc1', GstMessageError, gerror=(GError)NULL,
> debug=(string)"gstbasesrc.c\(2933\):\ gst_base_src_loop\ \(\):\
> /GstPipeline:negotiation_test_1/GstAudioTestSrc:audiotestsrc1:\012streaming\
> task\ paused\,\ reason\ not-negotiated\ \(-4\)";
>

I've attached a patch for your test case that makes it run successfully 
to completion. The change is to ensure that when you renegotiate, the 
sink still accepts buffers with the old caps until the caps change. 
Because the new caps are first in the list, they will be preferred over 
the old caps as the chosen format by upstream.

Regards,
Jan.

> ...
>
>
> 2014-06-06 11:42 GMT+02:00 Sebastian Dröge <sebastian at centricular.com
> <mailto:sebastian at centricular.com>>:
>
>     On Fr, 2014-06-06 at 10:54 +0200, Jose Antonio Santos Cadenas wrote:
>      > Hi,
>      >
>      > I've been investigating about reconfiguration of pipelines
>     dynamically and
>      > I run into negotiation problems while changing the caps of a
>     stream while
>      > playing.
>      >
>      > I've developed  a program to test the issue, you can download it
>     here:
>      >
>      > https://github.com/jcaden/negotiation_test
>      >
>      > Let me explain the test a little bit (copied from the repo README):
>      >
>      >   This test creates a simple pipeline and forces caps
>     renegotiation once the
>      >   GST_MESSAGE_STREAM_START is received on bus. It terminates
>     correctly if
>      > the
>      >   appsink receives a buffer with the renegotiated format.
>      >
>      >   This program runs a test continuously until it fails or it is
>     executed a
>      >   number of times selected by the used with -n option (1000000 by
>     default).
>      >
>      >   If option -q is given, the pipeline is this:
>      >
>      >     --------------      -------      ---------
>      >    | audiotestsrc | -> | queue | -> | appsink |
>      >     --------------      -------      ---------
>      >
>      >   This test frequently fails  (1 on 15?), because audiotestsrc
>     receive a
>      >   not negotiated error while pushing a buffer or because no
>     buffer with
>      >   the new format is received.
>      >
>      >    If -q option is not present the pipeline is this:
>      >
>      >     --------------      ---------
>      >    | audiotestsrc | -> | appsink |
>      >     --------------      ---------
>      >
>      >   This second pipeline works properly and renegotiates correctly.
>      >
>      > I'd like to know if I am suppose to do something else to perform the
>      > renegotiation, currently I am blocking the appsink:sink pad to
>     change the
>      > appsink caps and send a reconfigure event, then I unblock the pad
>     again.
>
>     Renegotiation and the RECONFIGURE event in GStreamer are only "hints",
>     not something that makes sure you immediately only get buffers in the
>     new format. And especially if a queue is used it can be possible that
>     there are still buffers with the old format inside the queue.
>
>     Your sink must be able to handle (or at least silently drop) buffers in
>     the old format after a reconfigure event.
>
>     --
>     Sebastian Dröge, Centricular Ltd - http://www.centricular.com
>     Expertise, Straight from the Source
>
>     _______________________________________________
>     gstreamer-devel mailing list
>     gstreamer-devel at lists.freedesktop.org
>     <mailto:gstreamer-devel at lists.freedesktop.org>
>     http://lists.freedesktop.org/mailman/listinfo/gstreamer-devel
>
>
>
>
> _______________________________________________
> gstreamer-devel mailing list
> gstreamer-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/gstreamer-devel
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: test-case-patch.diff
Type: text/x-patch
Size: 770 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/gstreamer-devel/attachments/20140614/fbaa4187/attachment.bin>