rtpdec: Calculate and report packet reception jitter

This brings back some code that was added originally in 4a6cc061
but never was used, and was removed as unused in 4cc843fa. The
code is updated to actually work and is tested to return sane
values.

Signed-off-by: Martin Storsjö <martin@martin.st>

libavformat/rtpdec.c

@@ -228,6 +228,24 @@ static int rtp_valid_packet_in_sequence(RTPStatistics *s, uint16_t seq)
     return 1;
 }
 
+static void rtcp_update_jitter(RTPStatistics *s, uint32_t sent_timestamp,
+                               uint32_t arrival_timestamp)
+{
+    // Most of this is pretty straight from RFC 3550 appendix A.8
+    uint32_t transit = arrival_timestamp - sent_timestamp;
+    uint32_t prev_transit = s->transit;
+    int32_t d = transit - prev_transit;
+    // Doing the FFABS() call directly on the "transit - prev_transit"
+    // expression doesn't work, since it's an unsigned expression. Doing the
+    // transit calculation in unsigned is desired though, since it most
+    // probably will need to wrap around.
+    d = FFABS(d);
+    s->transit = transit;
+    if (!prev_transit)
+        return;
+    s->jitter += d - (int32_t) ((s->jitter + 8) >> 4);
+}
+
 int ff_rtp_check_and_send_back_rr(RTPDemuxContext *s, URLContext *fd,
                                   AVIOContext *avio, int count)
 {
@@ -812,6 +830,16 @@ static int rtp_parse_one_packet(RTPDemuxContext *s, AVPacket *pkt,
         return rtcp_parse_packet(s, buf, len);
     }
 
+    if (s->st) {
+        int64_t received = av_gettime();
+        uint32_t arrival_ts = av_rescale_q(received, AV_TIME_BASE_Q,
+                                           s->st->time_base);
+        timestamp = AV_RB32(buf + 4);
+        // Calculate the jitter immediately, before queueing the packet
+        // into the reordering queue.
+        rtcp_update_jitter(&s->statistics, timestamp, arrival_ts);
+    }
+
     if ((s->seq == 0 && !s->queue) || s->queue_size <= 1) {
         /* First packet, or no reordering */
         return rtp_parse_packet_internal(s, pkt, buf, len);