From f33877bfe3135f5dbf4340e07a9dccd7f6ee680d Mon Sep 17 00:00:00 2001
From: Richard Fuchs <rfuchs@sipwise.com>
Date: Thu, 28 Jan 2021 13:43:46 -0500
Subject: [PATCH] TT#102450 fix some timestamping issues

Make sure G.722 rate multiplies is applied in the right places

Don't trust encoders to return proper timestamps, but instead track them
explicitly based on frame duration

Change-Id: Ia9fa96cf662da97159fa170c3a3f37516889e1bd
---
 daemon/codec.c | 23 ++++++++++++-----------
 lib/codeclib.c |  6 ++++++
 lib/codeclib.h |  1 +
 3 files changed, 19 insertions(+), 11 deletions(-)

diff --git a/daemon/codec.c b/daemon/codec.c
index fb69b2f30..17a1de118 100644
--- a/daemon/codec.c
+++ b/daemon/codec.c
@@ -1938,10 +1938,11 @@ static int packet_dtmf_fwd(struct codec_ssrc_handler *ch, struct transcode_packe
 		struct media_packet *mp)
 {
 	int payload_type = -1; // take from handler's output config
-	struct codec_ssrc_handler *output_ch = NULL;
-	struct codec_ssrc_handler *decoder_ch = NULL;
 
 	if (ch->handler->dtmf_scaler) {
+		struct codec_ssrc_handler *output_ch = NULL;
+		struct codec_ssrc_handler *decoder_ch = NULL;
+
 		// this is actually a DTMF -> PCM handler
 		// grab our underlying PCM transcoder
 		struct codec_handler *decoder_handler = __decoder_handler(ch->handler, mp);
@@ -1964,14 +1965,14 @@ static int packet_dtmf_fwd(struct codec_ssrc_handler *ch, struct transcode_packe
 		if (ch->dtmf_ts != packet->ts) {
 			// this is a new event
 			ch->dtmf_ts = packet->ts; // start TS
-			ch->last_dtmf_event_ts = 0; // last shifted FIFO PTS
+			ch->last_dtmf_event_ts = 0; // last DTMF event duration
 		}
 
-		unsigned long ts = output_ch->encoder->fifo_pts;
+		unsigned long ts = output_ch->encoder->next_pts / output_ch->encoder->def->clockrate_mult;
 		// roll back TS to start of event
 		ts -= ch->last_dtmf_event_ts;
 		// adjust to output RTP TS
-		unsigned long packet_ts = ts / output_ch->encoder->def->clockrate_mult + output_ch->first_ts;
+		unsigned long packet_ts = ts + output_ch->first_ts;
 
 		ilogs(transcoding, LOG_DEBUG, "Scaling DTMF packet timestamp and duration: TS %lu -> %lu "
 				"(%u -> %u)",
@@ -1986,18 +1987,18 @@ static int packet_dtmf_fwd(struct codec_ssrc_handler *ch, struct transcode_packe
 			dtmf->duration = htons(duration);
 
 			// shift forward our output RTP TS
-			output_ch->encoder->fifo_pts = ts + duration;
+			output_ch->encoder->next_pts = (ts + duration) * output_ch->encoder->def->clockrate_mult;
 			output_ch->encoder->packet_pts += (duration - ch->last_dtmf_event_ts) * output_ch->encoder->def->clockrate_mult;
 			ch->last_dtmf_event_ts = duration;
 		}
 		payload_type = ch->handler->dtmf_payload_type;
-	}
 
 skip:
-	if (output_ch)
-		obj_put(&output_ch->h);
-	if (decoder_ch)
-		obj_put(&decoder_ch->h);
+		if (output_ch)
+			obj_put(&output_ch->h);
+		if (decoder_ch)
+			obj_put(&decoder_ch->h);
+	}
 
 	char *buf = malloc(packet->payload->len + sizeof(struct rtp_header) + RTP_BUFFER_TAIL_ROOM);
 	memcpy(buf + sizeof(struct rtp_header), packet->payload->s, packet->payload->len);
diff --git a/lib/codeclib.c b/lib/codeclib.c
index 79e1e5dbc..e5c684f40 100644
--- a/lib/codeclib.c
+++ b/lib/codeclib.c
@@ -1285,11 +1285,17 @@ int encoder_input_data(encoder_t *enc, AVFrame *frame,
 			return -1;
 
 		if (enc->avpkt.size) {
+			// don't rely on the encoder producing steady timestamps,
+			// instead keep track of them ourselves based on the returned
+			// frame duration
+			enc->avpkt.pts = enc->next_pts;
+
 			if (enc->def->codec_type->encoder_got_packet)
 				enc->def->codec_type->encoder_got_packet(enc);
 
 			callback(enc, u1, u2);
 
+			enc->next_pts += enc->avpkt.duration;
 			enc->mux_dts = enc->avpkt.dts + 1; // min next expected dts
 
 			av_packet_unref(&enc->avpkt);
diff --git a/lib/codeclib.h b/lib/codeclib.h
index 7af12fdce..cda735ce9 100644
--- a/lib/codeclib.h
+++ b/lib/codeclib.h
@@ -240,6 +240,7 @@ struct encoder_s {
 	AVAudioFifo *fifo;
 	int64_t fifo_pts; // pts of first data in fifo
 	int64_t packet_pts; // first pts of data in packetizer buffer
+	int64_t next_pts; // next pts expected from the encoder
 	int ptime;
 	int bitrate;
 	int samples_per_frame; // for encoding