Benny Morgan [Tue, 2 Jul 2013 21:19:32 +0000 (23:19 +0200)]
- Remove assert - don't see why that assert should be - it only makes tvheadend crash for me
(cherry picked from commit 360ca58ad1335629afb5dc80eca0330d6cc382e3)
Benny Morgan [Sun, 30 Jun 2013 19:55:24 +0000 (21:55 +0200)]
- Fix - In case that num parameter was NULL, epg_episode_num_deserialize funcion would return newly calloc'd
epg_episode_num_t, but it's never used like this, so better remove the calloc, and make it void return.
(cherry picked from commit 4f3cf8e695114ff70079ec92fdf81352525d0904)
Benny Morgan [Sun, 30 Jun 2013 19:49:08 +0000 (21:49 +0200)]
- Fix - buf is assigned to to cm->cm_data with is defined 'uint8_t cm_data[CWS_NETMSGSIZE];' part of the cwc_message_t, never malloc'd and should be free'd
(cherry picked from commit 61491f394a92bfe20b18b61d2a130bf3a670efdc)
Benny Morgan [Sun, 30 Jun 2013 19:35:43 +0000 (21:35 +0200)]
- Fix memory leak - After cwc_running is set to 0 in capmt_destroy, capmt and capmt->id was not free'd
(cherry picked from commit 19b39e415981240b8178d557879f7f1d9b78320c)
Benny Morgan [Sun, 30 Jun 2013 19:19:15 +0000 (21:19 +0200)]
- Fix memory leaks - Some fields in tda was not free'd. tda_rootpath, tda_fe_path, tda_demux_path and tda_dvr_path
(cherry picked from commit 21a622986fcf8ec28b0b3ea6bb37d0148e3cb417)
Benny Morgan [Sun, 30 Jun 2013 19:09:09 +0000 (21:09 +0200)]
- Fix memory leak - in case htmsg_binary_des0 return's something < 0, n and f was not free'd
(cherry picked from commit 8fd70d013fa2263b0b11f5f128ba4f94f5570a9c)
Benny Morgan [Sun, 30 Jun 2013 18:18:13 +0000 (20:18 +0200)]
- Fix memory leak - dtf in tda->tda_table_feed queue, malloc'd in dvb_adapter.c - line 1115 was not free'd
(cherry picked from commit 901473d7e711b82124b7215f62e758c05b31317d)
Benny Morgan [Sun, 30 Jun 2013 18:10:56 +0000 (20:10 +0200)]
- Fix memory leak - cwc->cwc_id and cwc->cwc_viaccess_emm.shared_emm may are not free'd on cleanup
(cherry picked from commit 6dbd6a3cf837c341618239a3863ade74214ab8ee)
Benny Morgan [Sun, 30 Jun 2013 18:07:38 +0000 (20:07 +0200)]
- Fix memory leak - free cwc->cwc_cryptoworks_emm.shared_emm before set to NULL, if not NULL it must be malloc'd on line 1839 (case 0x84)
(cherry picked from commit 7eab6185e9c8a3399969cc9d30d8a4fdf9c892ef)
Adam Sutton [Tue, 16 Apr 2013 20:50:52 +0000 (21:50 +0100)]
services: include info from SDT (for Other services) if they already exist.
Services can be added from a variety of other tables which probably indicates
that they are in fact valid TV/Radio services, but we just don't know the
type.
Adam Sutton [Tue, 16 Apr 2013 20:11:11 +0000 (21:11 +0100)]
tdt: ignore other network in NIT this causes havoc with DVB-S networks.
This is particularly true where the other network is only available
via another orbital position (we could check for this) but it also
causes problems with other transmission types so I think we'll just
ignore for now.
(cherry picked from commit 3581869d83831f80dd7564ea4bb2a25b68ecf087)
Adam Sutton [Fri, 22 Mar 2013 10:50:03 +0000 (10:50 +0000)]
gtimer: increased resolution of the gtimer system.
This is needed to resolve a dvb tuner deadlock issued caused by my
re-jigging of the tuning setup code.
(cherry picked from commit 4f2f68a4b65a9ab5ecce80c867b9d7813de665a3)
Adam Sutton [Wed, 13 Mar 2013 19:36:59 +0000 (19:36 +0000)]
Fix #1643 - dvb: alter the way we handle dvb tuning
This ensures that demux filters are not installed until after tuning
is locked. This should resolve most stale data issues.
(cherry picked from commit ddc466c1bfa7405563a68a662114f5a3659d7cb5)
Adam Sutton [Fri, 8 Mar 2013 21:56:11 +0000 (21:56 +0000)]
dvr: ensure unlinked entry updates are properly handled
previously it updated things that were already linked and also on
different channels entirely.
(cherry picked from commit 6c6e0541be21fc574241c3c244f2b28c24db2021)
Adam Sutton [Fri, 8 Mar 2013 17:26:05 +0000 (17:26 +0000)]
Fix #1444 - dvr: ensure that DVR entries are not removed when EPG is updated
previously an overlapping (replacement) event could result in the DVR
entry being completely removed. This will now ensure that the original
entry is left in place (with just info and times).
Dave Chapman [Thu, 7 Mar 2013 14:53:20 +0000 (14:53 +0000)]
Swap the order of the PAT and PMT packets injected by the passthrough muxer - it is more logical to write the PAT first, followed by the PMT, as that is the order parsers will need to process the packets.
(cherry picked from commit 5d8af1a209a2c40b60252942ad77ff1df4d92712)
Adam Sutton [Fri, 15 Feb 2013 10:31:59 +0000 (10:31 +0000)]
Fix #1605 - timeshift: mark file space as free at earliest opportunity.
Previously the space was only marked as free when the file was physically
removed (which happens async) which resulted in the buffer immediately
being seen as full.
Andrew Martin [Sun, 10 Feb 2013 09:39:51 +0000 (02:39 -0700)]
fix compile errors with gcc 4.1.x and 4.2.x
- tested with gcc 4.1.2, 4.2.1, 4.4.6, 4.6.3 and 4.7.2
- disable gcc ignore pragma for "-Warray-bounds" if gcc version < 4.3
- when I took out the pragma tvheadend compiled fine in all the above compilers, but I assume some version of gcc is issuing a false-positive.
- "-Warray-bounds" was added in gcc 4.3: http://gcc.gnu.org/gcc-4.3/changes.html
- fixes typo in cmdline help for -C
- clarify usage of -a in help description
Note: there is still an issue I've realised in that writer_flush() could
potentially result in a single packet being written out of order as two
threads have the potential to write. This isn't fatal and is probably
rare enough to ignore at this stage.