Menu
▾
▴
Re: [gst-devel] Timestamping problem with gdppay element
|
From: Marco B. <gib...@gm...> - 2010-11-22 21:58:46
|
Hi,
On Tue, Nov 16, 2010 at 5:19 PM, 4ernov <4e...@gm...> wrote:
> Hello,
>
> I continue working at some complex pipeline which I wrote about
> previously ("Problems with video, tee and queues"). Solution proposed
> by Marco works just fine for the pipeline I noticed there:
>
> queue -- xvimagesink
> /
> playbin -- tee
> \
> queue -- fakesink
>
> converted dynamically to:
>
> queue -- xvimagesink
> /
> playbin -- tee
> \
> queue -- xvimagesink
>
> But the final destination of my experiments is the following case:
>
>
> queue -- xvimagesink
> /
> playbin -- tee
> \
> queue -- fakesink
>
> converted dynamically to:
>
> queue -- xvimagesink
> /
> playbin -- tee
> \
> queue -- gdppay -- filesink
>
> with remote end like this:
>
> filesrc -- gdpdepay -- decodebin -- ffmpegcolorspace -- xvimagesink
>
> And for this case it doesn't work.
>
> Well, what is the result for now:
> 1. Everything works fine if the final pipeline is hardcoded from the
> beginning (i.e. we have gdppay right from start).
> 2. Everything works fine without gdppay (with xvimagesink or smth. instead).
>
> But when I start the desired configuration I have the following:
> 1. Xv window of the remote end appears and stalls the playing but
> start to refresh slowly (several frames are shown rapidly in intervals
> of 4-5 secs)
Are you transferring this somehow over a network? Using gdppay and
filesink/filesource for this is not really an optimal setup. Maybe you
could have better luck with (de)payloaders and udp elements + jitter
buffer.
> 2. Xv window of the main pipeline stops to play properly, too, and
> behaves just the same way like the remote one.
> 3. The main pipeline prints the following warnings to console:
>
> 0:00:07.134552825 19490 0x95be9c8 WARN gdppay
> gstgdppay.c:594:gst_gdp_pay_chain:<gdppay0> did not receive
> new-segment before first buffer
> 0:00:14.095700940 19490 0x948d2b0 WARN ffmpeg
> gstffmpegdec.c:2002:gst_ffmpegdec_video_frame:<ffdec_h2640> Dropping
> non-keyframe (seek/init)
> 0:00:14.691007420 19490 0x95b4978 WARN basesink
> gstbasesink.c:2686:gst_base_sink_is_too_late:<xvimagesink0> warning: A
> lot of buffers are being dropped.
> 0:00:14.691055331 19490 0x95b4978 WARN basesink
> gstbasesink.c:2686:gst_base_sink_is_too_late:<xvimagesink0> warning:
> There may be a timestamping problem, or this computer is too slow.
This message is pretty self-explanatory: xvimagesink is not rendering
anything because buffers are too late. Just to try, what's the effect
if you set sync=false on that element? What if you set the
max-lateness property to something like e.g. a few seconds?
Regards
>
> Here's the source code of the main app:
>
> #include <gst/gst.h>
> #include <glib.h>
>
> GstElement *_pipeline, *_fakesink, *_remotesink, *_valve, *_mpoint,
> *_videobin, *_pipe;
>
> void create_pipeline_bin()
> {
> GstElement *tee, *queue1, *queue2, *xvimagesink;
>
> _pipeline = gst_element_factory_make("playbin", NULL);
>
> tee = gst_element_factory_make("tee", NULL);
> xvimagesink = gst_element_factory_make("xvimagesink", NULL);
>
> GstPad* pad;
>
> _videobin = gst_bin_new("videobin");
> queue1 = gst_element_factory_make("queue", NULL);
> queue2 = gst_element_factory_make("queue", NULL);
> _valve = gst_element_factory_make("valve", NULL);
>
> _fakesink = gst_element_factory_make("fakesink", NULL);
>
> gst_object_ref(_fakesink);
>
> _mpoint = _valve;
>
> gst_bin_add_many(GST_BIN(_videobin), tee, queue1, queue2, _valve,
> xvimagesink, _fakesink, NULL);
>
> gst_element_link_many(tee, queue1, xvimagesink, NULL);
> gst_element_link_many(tee, queue2, _valve, _fakesink, NULL);
> pad = gst_element_get_static_pad (tee, "sink");
> gst_element_add_pad (_videobin, gst_ghost_pad_new ("sink", pad));
> gst_object_unref (GST_OBJECT (pad));
>
> g_object_set(G_OBJECT (_pipeline), "uri", "file:///home/alex/test.mkv", NULL);
> g_object_set(G_OBJECT (_pipeline), "video-sink", _videobin, NULL);
> g_object_set(G_OBJECT (_pipeline), "audio-sink", NULL, NULL);
>
> gst_element_set_state (_pipeline, GST_STATE_PLAYING);
> }
>
> void create_remote_bin()
> {
> GstElement *gdppay;
> GstPad* pad;
>
> gdppay = gst_element_factory_make("gdppay", NULL);
> _pipe = gst_element_factory_make("filesink", NULL);
>
> g_object_set(G_OBJECT (_pipe), "location",
> "/home/alex/work/playground/test.gdp", NULL);
>
> _remotesink = gst_bin_new("gdpbin");
> gst_bin_add_many(GST_BIN(_remotesink), gdppay, _pipe, NULL);
> gst_element_link_many(gdppay, _pipe, NULL);
>
> pad = gst_element_get_static_pad (gdppay, "sink");
> gst_element_add_pad (_remotesink, gst_ghost_pad_new ("sink", pad));
> gst_object_unref (GST_OBJECT (pad));
> }
>
> void connect_remote_client()
> {
> if (_remotesink && _mpoint && _videobin)
> {
> g_object_set(G_OBJECT (_valve), "drop", TRUE, NULL);
>
> gst_element_unlink(_mpoint, _fakesink);
> gst_bin_remove(GST_BIN(_videobin), _fakesink);
> gst_element_set_state (_fakesink, GST_STATE_NULL);
>
> gst_bin_add(GST_BIN(_videobin), _remotesink);
> gst_element_link(_mpoint, _remotesink);
>
> gst_element_sync_state_with_parent(_remotesink);
>
> gst_element_set_state (_pipeline, GST_STATE_PLAYING);
> g_object_set(G_OBJECT (_valve), "drop", FALSE, NULL);
> }
> }
>
> gboolean connect_callback(gpointer)
> {
> connect_remote_client();
> return FALSE;
> }
>
> int main(int argc, char *argv[])
> {
> gst_init(&argc, &argv);
>
> gst_debug_set_active(true);
> gst_debug_set_default_threshold(GST_LEVEL_WARNING);
>
> create_remote_bin();
> create_pipeline_bin();
>
> g_timeout_add(7000, connect_callback, NULL);
>
> GMainLoop *loop;
>
> loop = g_main_loop_new (NULL, FALSE);
>
> g_print ("Running...\n");
> g_main_loop_run (loop);
>
> return 0;
> }
>
> End on the remote end there's gst-launch command:
>
> gst-launch -v filesrc location=/home/alex/work/playground/test.gdp !
> gdpdepay ! decodebin ! ffmpegcolorspace ! xvimagesink
>
> What is the essential problem? Is it possible to keep the "main"
> branch (i.e. with xvimagesink in main pipeline) playing while the
> second establishes connection via GDP protocol and then keep them
> synced? I hoped on valve element very much and it works perfectly for
> a couple xvimagesink but with GDP element it doesn't work, everything
> is almost stalled.
>
> ------------------------------------------------------------------------------
> Beautiful is writing same markup. Internet Explorer 9 supports
> standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 & L3.
> Spend less time writing and rewriting code and more time creating great
> experiences on the web. Be a part of the beta today
> http://p.sf.net/sfu/msIE9-sfdev2dev
> _______________________________________________
> gstreamer-devel mailing list
> gst...@li...
> https://lists.sourceforge.net/lists/listinfo/gstreamer-devel
>
|