Tvheadend

Provide '--trace nit,sdt,bat' for the mux without LCNs...

Jaroslav Kysela wrote:

Provide '--trace nit,sdt,bat' for the mux without LCNs...

I tried this but failed. I can't seem to stop tvh - it keeps restarting immediately.

OK. stopped it but now what?
tvheadend --trace nit,sdt,bat > logfile doesn't produce anything

You should compile tvh with --enable-trace . It's not available in packages.

... subscription: 000F: "epggrab" subscribing to mux, weight: 3, adapter: "Silicon Labs Si2168 : DVB-T #0", network: "Dover", mux: "730MHz", hostname: "<N/A>", username: "<N/A>", client: "<N/A>" 
2015-01-07 17:28:36.978 [  DEBUG]:sdt: onid 233A (9018) tsid 104C (4172)
2015-01-07 17:28:36.978 [  DEBUG]:sdt:   sid 104C (4172) running 4 free_ca 0
2015-01-07 17:28:36.978 [  TRACE]:sdt:     dtag 48 dlen 17
2015-01-07 17:28:36.978 [  TRACE]:sdt:     dtag 73 dlen 12
2015-01-07 17:28:36.978 [  TRACE]:sdt:   type 01 (1) name [BBC ONE S East] provider [] def_auth [fp.bbc.co.uk]
2015-01-07 17:28:39.952 [  TRACE]:nit:     dtag 5F dlen 4
2015-01-07 17:28:39.952 [  TRACE]:nit:       private 0000233A
2015-01-07 17:28:39.952 [  TRACE]:nit:     dtag 83 dlen 84
2015-01-07 17:28:39.952 [  DEBUG]:nit:     sid 4172 lcn 1

So you should have LCN for "BBC ONE S East" set to 1 . I see LCNs for all received channels.

Graham Horner wrote:

Yes the LCNs are there in the stream but they are not in the Channel/EPG table (see attached). BTW you have missed out the line directly above dtag 5F dlen 4:

2015-01-07 17:28:39.952 [ DEBUG]:nit: DVBT freq 674000000 bw 8MHz cons QAM/64 hier NONE code_rate 2/3:AUTO guard 1/32 trans 8k

It's not relevant. These mux parameters are for other muxes (not for the current mux). It's for the fast auto-discovery. The subscription line shows which mux is tuned.

Do you see LCNs in the service grid (configuration) ?

The LCN should be associated to service in src/input/mpegts/dvb_psi.c line 457 .

Jaroslav Kysela wrote:

It's not relevant. These mux parameters are for other muxes (not for the current mux). It's for the fast auto-discovery. The subscription line shows which mux is tuned.

I'm sorry, my c is a bit rusty, and I don't have an IDE set up, but I think it is relevant. When you get to the dsd referring to 674MHz, mux is pointed to that at line 1461 of dvb_psi.c. It then finds the services, but presumably doesn't store them because Network discovery is turned off. When it gets to the LCNs, seems to me it can't find a matching service at line 449.

This morning I turned Network discovery on, and sure enough LCNs were found for all the Bluebell Hill services, but the Dover service LCNs remain blank (0).

I see. You're right. TSID 4192 is in both SDT and NIT tables, although NIT table shows another mux on different frequency. Perhaps, there some wrong tsid check.

Could you add mpegts to trace (--trace nit,sdt,bat,mpegts) to monitor TSID changes for a mux ?

I don't see any problem. All muxes are allocated except one for tsid 0xA000 (40960). OK, basically, I don't know where to look, because I don't know which LCNs are missing.

Could we start again with one mux where the services have missing LCNs ? Just start with an empty config, configure only one mux (one which have services with completely missing LCNs) and try to scan it multiple times (if you edit the mux, you can set the scan status to pending multiple times to force new scan).

OK. We have SDT (tableid 42 = current mux) for onid 9018 tsid 4172. There is also service is service with id 4172 (BBC ONE S East). In nit table, there is also service 4172 with lcn 1 for onid 9018 tsid 4172. So everything should be straight and the code in dvb_psi.c - function dvb_desc_local_channel() line 457 should be called, but appearently it isn't.

2015-01-08 22:21:11.005 [   INFO]:subscription: 0001: "scan" subscribing to mux, weight: 4, adapter: "Silicon Labs Si2168 : DVB-T #0", network: "Dover one mux", mux: "706MHz", hostname: "<N/A>", username: "<N/A>", client: "<N/A>" 
2015-01-08 22:21:12.690 [  TRACE]:sdt: pid 11 tableid 42 extraid 00000000233a104c len 724
2015-01-08 22:21:12.694 [  DEBUG]:sdt: onid 233A (9018) tsid 104C (4172)
2015-01-08 22:21:12.695 [  TRACE]:mpegts: 706MHz in Dover one mux - set onid 233A (9018)
2015-01-08 22:21:12.695 [  DEBUG]:sdt:   sid 104C (4172) running 4 free_ca 0
2015-01-08 22:21:12.695 [  TRACE]:sdt:     dtag 48 dlen 17
2015-01-08 22:21:12.695 [  TRACE]:sdt:     dtag 73 dlen 12
2015-01-08 22:21:12.695 [  TRACE]:sdt:   type 01 (1) name [BBC ONE S East] provider [] def_auth [fp.bbc.co.uk]
2015-01-08 22:21:12.695 [  TRACE]:sdt:     type changed / old 01 (1)
2015-01-08 22:21:12.695 [  TRACE]:sdt:     cridauth changed
2015-01-08 22:21:12.695 [  TRACE]:sdt:     name changed
2015-01-08 22:21:12.695 [  DEBUG]:sdt:   nicename Dover one mux/706MHz/BBC ONE S East
2015-01-08 22:21:13.010 [  DEBUG]:nit: network 3032 (12338) [South East]
2015-01-08 22:21:13.010 [  DEBUG]:nit:   onid 233A (9018) tsid 104C (4172) mux 0xb6604df0
2015-01-08 22:21:13.010 [  TRACE]:nit:     dtag 41 dlen 63
2015-01-08 22:21:13.010 [  DEBUG]:nit:     service 104C (4172) type 01 (1)
2015-01-08 22:21:13.010 [  DEBUG]:nit:     sid 4172 lcn 1

If you don't want to dig to it, could you give me cca 60 seconds of the this whole mux dump (use the link in the mux grid - configuration - and wget or curl to get the contents).

I've grabbed it - 170MB which would take all afternoon to send on my connection. However, I now have a hack which works. I just tell tvh to ignore delivery system descriptors. I'm assuming this has no side effects except disabling network discovery but I'm testing longer term now to see what happens. The hack is at line 1455 in dvb_psi.c:

case DVB_DESC_TERR_DEL:
if (dtag == DVB_DESC_SAT_DEL)
mux = dvb_desc_sat_del(mm, onid, tsid, dptr, dlen);
else if (dtag == DVB_DESC_CABLE_DEL)
mux = dvb_desc_cable_del(mm, onid, tsid, dptr, dlen);
else if (onid != 0x233A) // ignore if UK DVB-T
mux = dvb_desc_terr_del(mm, onid, tsid, dptr, dlen);
if (mux && onid != 0x233A) { // ignore if UK DBV-T - this may not be necessary - don't know what tsid is at this point
mpegts_mux_set_onid(mux, onid);
mpegts_mux_set_tsid(mux, tsid, 0);
}
break;

This stops mux being pointed at nothing (line 612 of logfile7). Let me know if you still want the stream dump.

Or indeed, more elegantly:

        case DVB_DESC_TERR_DEL:
          if (onid == 0x233A) break;
        case DVB_DESC_SAT_DEL:
        case DVB_DESC_CABLE_DEL:
          if (dtag == DVB_DESC_SAT_DEL)
            mux = dvb_desc_sat_del(mm, onid, tsid, dptr, dlen);
          else if (dtag == DVB_DESC_CABLE_DEL)
            mux = dvb_desc_cable_del(mm, onid, tsid, dptr, dlen);
          else
            mux = dvb_desc_terr_del(mm, onid, tsid, dptr, dlen);
          if (mux) {
            mpegts_mux_set_onid(mux, onid);
            mpegts_mux_set_tsid(mux, tsid, 0);
          }
          break;

Thinking about this issue, the problem is probably that you have 2 muxes with same TSID but on different frequency. The current TVH code matches only TSID, so if the wrong mux is matched, the service operations are handled for this wrong mux. The loop in the NIT callback should iterate through all muxes with matching TSID to assign service details like LCNs..

I think it's like this. Within the UK South East Region, there are about 40 transmitter masts. They all broadcast the same services with the same transport stream IDs, serviceIDs and the same LCNs but on different frequencies. It seems they therefore all carry the same service information, which includes the transport delivery system descriptors for Bluebell Hill. That is, Freeview/DigitalUK don't bother to re-encode the tdsds for each transmitter. When tvh finds a tdsd it doesn't already know it tries to create a new one, which succeeds if network discovery is on, but fails otherwise. I think it must therefore try to match both the tsid and the frequency.

The problem is that when the subsequent LCN entries are parsed, they are assigned to services from this new mux (if it exists) or discarded (if it doesn't). So yes, you could go through all the muxes matching tsids (and that was my first thought) but all it needs is to get it to ignore the tdsds completely. I can't really see they are any use, anyway, are they mandatory in the DVB standard?

Correct me if I'm wrong - I'm new to tvh and DVB standards - but I'm pretty sure my hack will work across the UK. The only thing I don't understand is why the issue hasn't been raised before. Maybe most people don't use the pre-set LCNs and just make up their own.

Could you retest with v3.9-2346-geed4d8c ?

In case I was not clear, this is not fixed. Please revert status.

OK. The idea was correct, but the implementation missed one thing. It should be fixed in v3.9-2373-g30cc0a7 .