multiqueue: avoid signaling overrun on the first segment

When the first segment has position != 0 and position > max-size-time
it will immediatelly cause the multiqueue to signal overrun.

This can happen easily with adaptive streams when switching bitrates
and starting a new group. The segment for this new group will have
a position that is much greater than 0 and will lead to this issue.
This is particularly harmful when the adaptive stream uses mpegts
that doesn't emit no-more-pads and it might happen that only one
of the stream pads was added when the multiqueue overruns and gets
the group ready for exposing. So the user will only get audio or
video.

The solution is to fallback to the sink segment while the source pad
has no segment.

https://bugzilla.gnome.org/show_bug.cgi?id=729124
This commit is contained in:
Thiago Santos 2014-04-28 13:02:11 -03:00
parent 3b617f1fb2
commit 0a3bc6be1f

View file

@ -144,6 +144,9 @@ struct _GstSingleQueue
/* segments */
GstSegment sink_segment;
GstSegment src_segment;
gboolean has_src_segment; /* preferred over initializing the src_segment to
* UNDEFINED as this doesn't requires adding ifs
* in every segment usage */
/* position of src/sink */
GstClockTime sinktime, srctime;
@ -826,6 +829,7 @@ gst_single_queue_flush (GstMultiQueue * mq, GstSingleQueue * sq, gboolean flush,
gst_single_queue_flush_queue (sq, full);
gst_segment_init (&sq->sink_segment, GST_FORMAT_TIME);
gst_segment_init (&sq->src_segment, GST_FORMAT_TIME);
sq->has_src_segment = FALSE;
/* All pads start off not-linked for a smooth kick-off */
sq->srcresult = GST_FLOW_OK;
sq->pushed = FALSE;
@ -945,9 +949,27 @@ update_time_level (GstMultiQueue * mq, GstSingleQueue * sq)
sink_time = sq->sinktime;
if (sq->src_tainted) {
GstSegment *segment;
gint64 position;
if (sq->has_src_segment) {
segment = &sq->src_segment;
position = sq->src_segment.position;
} else {
/*
* If the src pad had no segment yet, use the sink segment
* to avoid signalling overrun if the received sink segment has a
* a position > max-size-time while the src pad time would be the default=0
*
* This can happen when switching pads on chained/adaptive streams and the
* new chain has a segment with a much larger position
*/
segment = &sq->sink_segment;
position = sq->sink_segment.position;
}
src_time = sq->srctime =
gst_segment_to_running_time (&sq->src_segment, GST_FORMAT_TIME,
sq->src_segment.position);
gst_segment_to_running_time (segment, GST_FORMAT_TIME, position);
/* if we have a time, we become untainted and use the time */
if (G_UNLIKELY (src_time != GST_CLOCK_TIME_NONE))
sq->src_tainted = FALSE;
@ -993,8 +1015,10 @@ apply_segment (GstMultiQueue * mq, GstSingleQueue * sq, GstEvent * event,
if (segment == &sq->sink_segment)
sq->sink_tainted = TRUE;
else
else {
sq->has_src_segment = TRUE;
sq->src_tainted = TRUE;
}
GST_DEBUG_OBJECT (mq,
"queue %d, configured SEGMENT %" GST_SEGMENT_FORMAT, sq->id, segment);