Meta with timestamp
Locked
 
|
Hello,
I'm using the code below to timestamp a pipeline and get a delay in the end for the buffers. It work fine usually, putting the but timestamp at the src pad and getting the difference in the sink. However, in a longer transcoding pipeline the transform function always get non writable buffers, and the meta is not replicated. Is there a way to now who is blocking the buffer, or a workaround? Any tips will be really appreciated. Regards Rafael /* meta functions */ #include <gst/gst.h> #include <unistd.h> #include <sys/time.h> #include <glib.h> #include <stdlib.h> #include <string.h> #include <stdio.h> #include <argp.h> typedef struct _GstMetaTS GstMetaTS; struct _GstMetaTS { GstMeta meta; GstClockTime in_timestamp; }; GType gst_meta_ts_api_get_type (void); const GstMetaInfo* gst_meta_ts_get_info (void); #define GST_META_TS_GET(buf) ((GstMetaTS *)gst_buffer_get_meta(buf,gst_meta_ts_api_get_type())) #define GST_META_TS_ADD(buf) ((GstMetaTS *)gst_buffer_add_meta(buf,gst_meta_ts_get_info(),(gpointer)NULL)) gboolean gst_meta_ts_transform (GstBuffer *dest_buf, GstMeta *src_meta, GstBuffer *src_buf, GQuark type, gpointer data) { GstBuffer * buf = NULL; if(! gst_buffer_is_writable(dest_buf)) { GST_DEBUG("NOT WRITABLE, NOT ABLE TO TRANSFORM\n"); return FALSE; } GstMeta* dest_meta = (GstMeta*)GST_META_TS_ADD(dest_buf); GstMetaTS* src_meta_ts = (GstMetaTS*)src_meta; GstMetaTS* dest_meta_ts = (GstMetaTS*)dest_meta; dest_meta_ts->in_timestamp = src_meta_ts->in_timestamp; return TRUE; } void gst_meta_ts_free (GstMeta *meta, GstBuffer *buffer) { } gboolean gst_meta_ts_init(GstMeta *meta, gpointer params, GstBuffer *buffer){ GstMetaTS* ts_meta = (GstMetaTS*)meta; ts_meta->in_timestamp = GST_CLOCK_TIME_NONE; return TRUE; } GType gst_meta_ts_api_get_type (void){ static volatile GType type; static const gchar *tags[] = { NULL }; if (g_once_init_enter (&type)) { GType _type = gst_meta_api_type_register ("GstMetaTSAPI", tags); g_once_init_leave (&type, _type); } return type; } /* transcoder code */ static GstPadProbeReturn probe_src (GstPad *pad, GstPadProbeInfo *info, gpointer user_data){ if (playing){ GstBuffer *b; b = GST_PAD_PROBE_INFO_BUFFER (info); if (! gst_buffer_is_writable(b)) { return GST_PAD_PROBE_OK; } GstMetaTS* meta = GST_META_TS_ADD(b); GstClock* clock = gst_system_clock_obtain (); meta->in_timestamp = gst_clock_get_time(clock); gst_object_unref(clock); } return GST_PAD_PROBE_OK; } static GstPadProbeReturn probe_sink (GstPad *pad, GstPadProbeInfo *info, gpointer user_data){ GstBuffer *b; b = GST_PAD_PROBE_INFO_BUFFER (info); if (! gst_buffer_is_writable(b)) { return GST_PAD_PROBE_OK; } GstMetaTS* meta = GST_META_TS_GET(b); if(meta == NULL) { return GST_PAD_PROBE_OK; } GstClock* clock = gst_system_clock_obtain (); GstClockTimeDiff diff = GST_CLOCK_DIFF(meta->in_timestamp, gst_clock_get_time(clock)); delay_avg = ((delay_avg*delay_count)+diff)/++delay_count; if (last_delay_print + DELAY_PRINT_INTERVAL < get_time_us()){ printf("DELAY_NOW:%3.4fms,DELAY_AVG:%3.4fms\n",((double)diff)/1000/1000,delay_avg/1000/1000); last_delay_print = get_time_us(); } gst_object_unref(clock); gst_buffer_remove_meta(b, (GstMeta*)meta); return GST_PAD_PROBE_OK; } (...) GstPad *pad = gst_element_get_static_pad (tra_src, "src"); gst_pad_add_probe (pad, GST_PAD_PROBE_TYPE_BUFFER, (GstPadProbeCallback) probe_src, NULL, NULL); gst_object_unref (pad); pad = gst_element_get_static_pad (tra_sink, "sink"); gst_pad_add_probe (pad, GST_PAD_PROBE_TYPE_BUFFER, (GstPadProbeCallback) probe_sink, NULL, NULL); gst_object_unref (pad); _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel |
|
|
Hi,
Just ensure the buffer is writable (it won't actually copy the underlying memory) with gst_buffer_make_writable() and set it back to the GST_PAD_PROBE_INFO_DATA so that it gets propagated. Edward On Tue, 2018-03-20 at 13:39 +0100, José Rafael wrote: > Hello, > > I'm using the code below to timestamp a pipeline and get a delay in > the end for the buffers. > It work fine usually, putting the but timestamp at the src pad and > getting the difference in the sink. > However, in a longer transcoding pipeline the transform function > always get non writable buffers, and the meta is not replicated. > > Is there a way to now who is blocking the buffer, or a workaround? > > Any tips will be really appreciated. gstreamer-devel mailing list [hidden email] https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel |
|
|
Hi Edward Thank you for the help. With gst_buffer_make_writable(), as below, I have the first meta propagated, and after a segmentation fault. It should make a copy if not writable, no? If I use gst_buffer_copy() it processes the buffer without the seg fault, but not writing back on the right buffer. This make me think dest_buff is being changes somewhere else. I don't get the funcionality of
GST_PAD_PROBE_INFO_DATA(), should be in the transformation function? Regards Rafael gboolean gst_meta_ts_transform (GstBuffer *dest_buf, GstMeta *src_meta, GstBuffer *src_buf, GQuark type, gpointer data) { dest_buf = gst_buffer_make_writable(dest_buf); if(! gst_buffer_is_writable(dest_buf)) { printf("gst_meta_ts_transform: not writable to be transformed\n"); return FALSE; } GstMeta* dest_meta = (GstMeta*)GST_META_TS_ADD(dest_buf); GstMetaTS* src_meta_ts = (GstMetaTS*)src_meta; GstMetaTS* dest_meta_ts = (GstMetaTS*)dest_meta; dest_meta_ts->in_timestamp = src_meta_ts->in_timestamp; return TRUE; } On 20 March 2018 at 15:24, Edward Hervey <[hidden email]> wrote: Hi, _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel |
|
|
Hi Rafeal,
did you find a solution for your problem ? I'm facing the same problem. -- Sent from: http://gstreamer-devel.966125.n4.nabble.com/ _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel |
|
|
Hi Frederik That time I found that ubuntu's gstreamer version had a bug, the metadata was not being copied in some cases. Compiling the latest from github solved it. Hope it works for you. If I find the bug URL I will send you. Best Regards Rafael On Sat, 22 Sep 2018 at 12:20, Frederik <[hidden email]> wrote: Hi Rafeal, _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel |
|
|
Hi Rafael,
Thanks for your reply. I believe mine was related to the 'avdec_h264' element. Setting the property 'direct-rendering' to FALSE seems to solve the problem. Best Regards Frederik -- Sent from: http://gstreamer-devel.966125.n4.nabble.com/ _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel |
|
|
Hi Frederik In my case it was the transformation, also related to the avdec_h264. Thanks! BR Rafael On Wed, 26 Sep 2018 at 13:14, Frederik <[hidden email]> wrote: Hi Rafael, _______________________________________________ gstreamer-devel mailing list [hidden email] https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel |
«
Return to GStreamer-devel
|
1 view|%1 views
| Free forum by Nabble | Edit this page |