After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 749315 - funnel: Add "forward-sticky-events" property
funnel: Add "forward-sticky-events" property
Status: RESOLVED FIXED
Product: GStreamer
Classification: Platform
Component: gstreamer (core)
git master
Other Linux
: Normal normal
: 1.5.90
Assigned To: GStreamer Maintainers
GStreamer Maintainers
Depends on:
Blocks:
 
 
Reported: 2015-05-13 13:07 UTC by Miguel París Díaz
Modified: 2015-09-22 07:51 UTC
See Also:
GNOME target: ---
GNOME version: ---


Attachments
0001-funnel-add-forward-sticky-events-property.patch (3.50 KB, patch)
2015-05-14 09:51 UTC, Miguel París Díaz
none Details | Review
0001-funnel-add-forward-sticky-events-property.patch (3.74 KB, patch)
2015-06-24 13:39 UTC, Miguel París Díaz
committed Details | Review

Description Miguel París Díaz 2015-05-13 13:07:43 UTC
I propose deleting the funnel from dtlssrtpenc. Reasons:
 - funnel sends all sticky events again on stream changes, which adds an high overhead.
 - If the user want to merge into a single stream, he can use a funnel connected to the output of the dtlssrtpenc, so he has the same current behaviour.
 - Having separated pads is useful for connecting different elements for each srcpad, testing, etc.

Related with https://bugzilla.gnome.org/show_bug.cgi?id=748723
Comment 1 Sebastian Dröge (slomo) 2015-05-13 13:09:36 UTC
Don't you always have to merge the output anyway to send it over the network?
Comment 2 Miguel París Díaz 2015-05-13 13:16:28 UTC
I am using in the way you say, but the important thing is the overhead added on stream changes.
So, we are planning to create a "litefunnel" to avoid sending all sticky events.
Comment 3 Tim-Philipp Müller 2015-05-13 13:18:47 UTC
Why not just add a property to funnel ?
Comment 4 Miguel París Díaz 2015-05-13 13:24:08 UTC
Yeah, it is a good idea ;).
It will be easier to use for other users without changing their current implementations.
How can we call this property?
And an important point: we should doc about possible inconvenients of not sending the sticky events.
Comment 5 Sebastian Dröge (slomo) 2015-05-13 13:26:05 UTC
Also a funnel like that would be the right thing to do for dtlssrtpenc anyway. It should output DTLS caps and a single stream, from the outside it shouldn't look like separate streams at all. It's just like a muxer here.
Comment 6 Sebastian Dröge (slomo) 2015-05-13 16:17:16 UTC
Maybe call that property forward-sticky-events and put it on the sinkpads of funnel. If set to false, no sticky events of that pad are forwarded when it becomes active.

What do you think?
Comment 7 Miguel París Díaz 2015-05-14 09:50:19 UTC
I prefer and easier way.
So I am uploading a patch that adds the property you suggested but is used for all pads.
Comment 8 Miguel París Díaz 2015-05-14 09:51:05 UTC
Created attachment 303356 [details] [review]
0001-funnel-add-forward-sticky-events-property.patch
Comment 9 Sebastian Dröge (slomo) 2015-05-18 07:40:26 UTC
I think a per-pad property would be more useful. Especially your patch now will stop funnel from sending any sticky events, making data flow wrong because of missing stream-start/segment events. It should somehow invent the events then. With the per-pad approach you could pass through the events of a single pad as needed.
Comment 10 Miguel París Díaz 2015-06-24 13:39:10 UTC
Created attachment 306001 [details] [review]
0001-funnel-add-forward-sticky-events-property.patch
Comment 11 Sebastian Dröge (slomo) 2015-06-25 08:55:23 UTC
Let's add the per-pad property at a later time when needed. Now only this property has to be used by elements.


commit c2e15651beca382245fee9935ef81c1b6620fd0f
Author: Miguel París Díaz <mparisdiaz@gmail.com>
Date:   Thu May 14 11:48:45 2015 +0200

    funnel: add "forward-sticky-events" property
    
    It is useful to avoid sending sticky event on stream changes.
    
    https://bugzilla.gnome.org/show_bug.cgi?id=749315
Comment 12 Miguel París Díaz 2015-06-25 14:21:58 UTC
I agree, my point of view was the same ;).