How to dynamically add and remove branch to a "tee" element?
Locked
 
|
After a lot of search, I found this is not quite simple as I used to
think. The coding following only work when 2 sink is linked to the tee. When the 3rd comes, all 3 sink doesn't work. What is the correct way to add and remove branch to a "tee"? On IRC, I was told to try these 1. setting sink "asycn = FALSE" 2. add a queue between tee and sink 3. use pad instead of gst_element_link None of them works. GstElement* dd = gst_element_factory_make("directdrawsink", NULL); g_object_set(dd, "async", FALSE, NULL); gst_x_overlay_set_xwindow_id(GST_X_OVERLAY(dd), (gulong)handle); gst_element_set_state (dd, GST_STATE_PAUSED); gst_bin_add(GST_BIN (self->priv->pipeline), dd); if(!gst_element_link (self->priv->tee, dd)) { vivs_warning("Can not link tee to directdrawsink"); } if(self->priv->sink_num == 1) gst_element_set_state (self->priv->pipeline, GST_STATE_PLAYING); else gst_element_set_state (dd, GST_STATE_PLAYING); self->priv->sink_num += 1 -- look to the things around you,the immediate world around you, if you are alive,it will mean something to you ——Paul Strand ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
Re: How to dynamically add and remove branch to a "tee" element?
|
|
In my experience, the bin/elements/branch you're trying to attach to one of the src%d pads should be in the PLAYING state before you link it. This is because as the tee tries to push a buffer to its src pads, if any of the pads is linked to a non-PLAYING element, the buffer push method returrn a wrong-state value which propagates all the way to the source and pauses everything, GST_DEBUG=4 will show you this.
Also, like you mentioned, gst_pad_link is how you would link a tee's src%d to the branch. While unlinking, I first unlink the pad and then take care of the disposing the branch. Ideally, however, we should block data on the pad and then unlink it, but that hasn't work for me so far.
HTH
On Fri, Nov 6, 2009 at 8:37 AM, Peter Cai <[hidden email]> wrote: After a lot of search, I found this is not quite simple as I used to ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
|
|
Oh， it took me a whole day on this. It's not the problem of state any more.
The pipeline is like this : appsrc ! decodin ffmpegcsp ！ tee ! 3 sinks. Something wrong happens when decodebin notify new pad is added. In the callback of "new-decoded-pad", I tried to link decodebin and ffmpegcsp so that all the elements in the pipeline is linked. But it could never finish link appsrc and ffmpegcsp. The log shows after gst_base_transform_getcaps is called, it never returned. 0:00:09.803216000 2300 00A0F6E0 DEBUG basetransform gstbasetransform.c:620:gst_base_transform_getcaps:<ffmpegcsp0:sink> peer caps 0ABE4DE0 0:00:09.803216000 2300 00A0F6E0 DEBUG basetransform gstbasetransform.c:624:gst_base_transform_getcaps:<ffmpegcsp0:sink> our template 058598C0 Some debugging shows it because gst_capsIntersect is doing a huge loop that wouldn't be possible to finish in short time. That because caps2->structs->len is too large to be correct. The most strange thing is that when 2 sink is attached to tee, everything is OK. So, this is totally out of my ability and I really appreciate your help! GstCaps * gst_caps_intersect (const GstCaps * caps1, const GstCaps * caps2) { .... for (i = 0; i < caps1->structs->len + caps2->structs->len - 1; i++) { .... while (k < caps2->structs->len) { .... } } return dest; } On Fri, Nov 6, 2009 at 10:51 PM, Uday Verma <[hidden email]> wrote: > In my experience, the bin/elements/branch you're trying to attach to one of > the src%d pads should be in the PLAYING state before you link it. This is > because as the tee tries to push a buffer to its src pads, if any of the > pads is linked to a non-PLAYING element, the buffer push method returrn a > wrong-state value which propagates all the way to the source and pauses > everything, GST_DEBUG=4 will show you this. > Also, like you mentioned, gst_pad_link is how you would link a tee's src%d > to the branch. > While unlinking, I first unlink the pad and then take care of the disposing > the branch. Ideally, however, we should block data on the pad and then > unlink it, but that hasn't work for me so far. > HTH > > On Fri, Nov 6, 2009 at 8:37 AM, Peter Cai <[hidden email]> wrote: >> >> After a lot of search, I found this is not quite simple as I used to >> think. The coding following only work when 2 sink is linked to the >> tee. >> When the 3rd comes, all 3 sink doesn't work. >> >> What is the correct way to add and remove branch to a "tee"? >> On IRC, I was told to try these >> >> 1. setting sink "asycn = FALSE" >> 2. add a queue between tee and sink >> 3. use pad instead of gst_element_link >> >> None of them works. >> >> GstElement* dd = gst_element_factory_make("directdrawsink", NULL); >> g_object_set(dd, "async", FALSE, NULL); >> gst_x_overlay_set_xwindow_id(GST_X_OVERLAY(dd), (gulong)handle); >> >> gst_element_set_state (dd, GST_STATE_PAUSED); >> gst_bin_add(GST_BIN (self->priv->pipeline), dd); >> >> if(!gst_element_link (self->priv->tee, dd)) >> { >> vivs_warning("Can not link tee to directdrawsink"); >> } >> >> if(self->priv->sink_num == 1) >> gst_element_set_state (self->priv->pipeline, >> GST_STATE_PLAYING); >> else >> gst_element_set_state (dd, GST_STATE_PLAYING); >> self->priv->sink_num += 1 >> >> >> -- >> look to the things around you,the immediate world around you, if you >> are alive,it will mean something to you ——Paul Strand >> >> >> ------------------------------------------------------------------------------ >> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 >> 30-Day >> trial. Simplify your report design, integration and deployment - and focus >> on >> what you do best, core application coding. Discover what's new with >> Crystal Reports now. http://p.sf.net/sfu/bobj-july >> _______________________________________________ >> gstreamer-devel mailing list >> [hidden email] >> https://lists.sourceforge.net/lists/listinfo/gstreamer-devel > > > ------------------------------------------------------------------------------ > Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day > trial. Simplify your report design, integration and deployment - and focus > on > what you do best, core application coding. Discover what's new with > Crystal Reports now. http://p.sf.net/sfu/bobj-july > _______________________________________________ > gstreamer-devel mailing list > [hidden email] > https://lists.sourceforge.net/lists/listinfo/gstreamer-devel > > -- look to the things around you,the immediate world around you, if you are alive,it will mean something to you ——Paul Strand ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/gstreamer-devel |
«
Return to GStreamer-devel
|
1 view|%1 views
| Free forum by Nabble | Edit this page |