Tvheadend

Any updates on this? I'm having the same errors...

Nabil
sketchgrowl.com (for contact details, I'm a pet portrait artist when I'm not watching TV lol)

note from mpmc: link "removed" due to spam issues.

+1 - Seen with both H264 and H265 streams

Apply the patch bellow and provide new traces, please:

diff --git a/src/plumbing/tsfix.c b/src/plumbing/tsfix.c
index 4b4843e5d..6584597f5 100644
--- a/src/plumbing/tsfix.c
+++ b/src/plumbing/tsfix.c
@@ -552,6 +552,8 @@ tsfix_input_packet(tsfix_t *tf, streaming_message_t *sm)
   tfs = tfs_find(tf, pkt);
   streaming_msg_free(sm);

+  pkt_trace(LS_TSFIX, pkt, "*input*");
+
   if (tfs == NULL || mclk() < tf->tf_start_time) {
     tsfix_packet_drop(tfs, pkt, "start time");
     return;

Thanks. It seems that the first I-frame is not passed to tsfix for a reason, so everything gets confused:

Sep  2 16:33:37 nas tvheadend[53654]: parser: deliver (pkt stream 1 HEVC type I pcr 2371666945 dts 2371729945 pts 2371729945 dur 7200 len 59897 err 0 meta 101)
Sep  2 16:33:37 nas tvheadend[53654]: parser: deliver (pkt stream 2 AAC-LATM pcr 2371681345 dts 2371730125 pts 2371730125 dur 2089 len 346 err 0)
Sep  2 16:33:37 nas tvheadend[53654]: tsfix: *input* (pkt stream 2 AAC-LATM pcr 2371681345 dts 2371730125 pts 2371730125 dur 2089 len 346 err 0)
Sep  2 16:33:37 nas tvheadend[53654]: parser: deliver (pkt stream 2 AAC-LATM pcr 2371681345 dts 2371732214 pts 2371732214 dur 2089 len 368 err 0)
Sep  2 16:33:37 nas tvheadend[53654]: tsfix: *input* (pkt stream 2 AAC-LATM pcr 2371681345 dts 2371732214 pts 2371732214 dur 2089 len 368 err 0)
Sep  2 16:33:37 nas tvheadend[53654]: parser: deliver (pkt stream 2 AAC-LATM pcr 2371681345 dts 2371734303 pts 2371734303 dur 2089 len 378 err 0)

Could you also save the MPEG-TS stream (use the URL in the service grid and the pass streaming profile)? Cca 30 seconds. Attach the file to this bug.

I see the problem now. It is in the upper level of code. The first packet is lost, because the stream restart notification is not in sync with the packet delivery, thus the first packets might be lost. We need to queue packets when the streaming restart is asked from the parser code. Give me some time to look into this. I am unfortunately very busy with the "real" work and the personal life.

Hints, if someone other will look into: service_update_elementary_stream() is called, t->s_pending_restart flag is set. From this time, the parser should queue packets before they are delivered - parser_deliver() function - until the streaming is restarted.

I've tried this, but doesnt seems to work :/

diff --git a/src/parsers/parsers.c b/src/parsers/parsers.c
index a274a094d..a7952ebfa 100644
--- a/src/parsers/parsers.c
+++ b/src/parsers/parsers.c
@@ -125,21 +125,24 @@ parser_deliver(parser_t *t, parser_es_t *st, th_pkt_t *pkt)
   }

 deliver:
+  //ici
diff --git a/src/parsers/parsers.c b/src/parsers/parsers.c
index a274a094d..a7952ebfa 100644
--- a/src/parsers/parsers.c
+++ b/src/parsers/parsers.c
@@ -125,21 +125,24 @@ parser_deliver(parser_t *t, parser_es_t *st, th_pkt_t *pkt)
   }

 deliver:
+  //ici
   pkt->pkt_componentindex = st->es_index;

-  pkt_trace(LS_PARSER, pkt, "deliver");
-
   if (SCT_ISVIDEO(pkt->pkt_type)) {
     pkt->v.pkt_aspect_num = st->es_aspect_num;
     pkt->v.pkt_aspect_den = st->es_aspect_den;
   }

-  /* Forward packet */
-  streaming_target_deliver2(t->prs_output, streaming_msg_create_pkt(pkt));
-
-  /* Decrease our own reference to the packet */
-  pkt_ref_dec(pkt);
-
+  if(!st->es_service->s_pending_restart) {
+    pkt_trace(LS_PARSER, pkt, "deliver");
+    /* Forward packet */
+    streaming_target_deliver2(t->prs_output, streaming_msg_create_pkt(pkt));
+    
+    /* Decrease our own reference to the packet */
+    pkt_ref_dec(pkt);
+  } else {
+    parser_backlog(t, st, pkt);
+  }
 }

 /**
diff --git a/src/parsers/parsers.h b/src/parsers/parsers.h
index f37a56c57..b1d074599 100644
--- a/src/parsers/parsers.h
+++ b/src/parsers/parsers.h
@@ -134,6 +134,8 @@ void parser_enqueue_packet(struct service *t, parser_es_t *st, th_pkt_t *pkt);

 void parser_set_stream_vparam(parser_es_t *st, int width, int height, int duration);

+static void parser_backlog(parser_t *t, parser_es_t *st, th_pkt_t *pkt);
+
 extern const unsigned int mpeg2video_framedurations[16];

 #endif /* PARSERS_H */

I'm also having this issue, I'm attempting to stream a VOD video in .mkv format. The logs will look like above but if I keep attempting to view the content I'll eventually be able to stream it from tvheadend. I'm piping the following command when trying to stream the content:

pipe:///usr/bin/ffmpeg -loglevel 0 -re -i http://example.tld/tmp.mkv -c copy -f mpegts pipe:1

Same issue here. It happens every 5 or 6 attempts to play an IPTV stream.

Jaroslav Kysela wrote:

I see the problem now. It is in the upper level of code. The first packet is lost, because the stream restart notification is not in sync with the packet delivery, thus the first packets might be lost. We need to queue packets when the streaming restart is asked from the parser code. Give me some time to look into this. I am unfortunately very busy with the "real" work and the personal life.

Hints, if someone other will look into: service_update_elementary_stream() is called, t->s_pending_restart flag is set. From this time, the parser should queue packets before they are delivered - parser_deliver() function - until the streaming is restarted.

Jaroslav, could you give me more information about how to solve this ? Not very experienced with this part of the code but I can try. Otherwise, if you find time to fix this, that would be fantastic :)

Hi i compiled latest master with your patch and it fixed the issue where iptv start with audio only! Thanks.. This patch should be added to tvheadend master.
Laurent H. wrote:

Hi,

I moved the queue to parser struct, see attached rstlog-ok.patch

Seems to works for me in most cases, but i'm not sure if function parser_create is the best place to send pkt from rstlog.

Can someone test and review this patch ?

Thank you.

Hi,

I've made a Pull Request : https://github.com/tvheadend/tvheadend/pull/1339

Pull request was merged https://github.com/tvheadend/tvheadend/pull/1339/commits/4c742c1c019924bb2d7aef4b8a7d3f51523d2f87

Ticket can be closed.