Tvheadend

Hi,

I've just set up a TVHeadend server with TBS Octa 6909X on Ubuntu using the linuxserver.io docker image (version 4.3-1792~g466a01431, so current master).

The adapters seem to be working fine, however TVHeadend is quite unstable, I'm getting a lot of:

tvheadend    | 2019-06-08 02:46:24.078 [WARNING] mpegts: too much queued table input data (over 2MB) for TurboSight TBS 6909x (Octa DVB-S/S2/S2X) #6 : DVB-S #0, discarding new
tvheadend    | 2019-06-08 02:46:29.250 [WARNING] mpegts: too much queued table input data (over 2MB) for TurboSight TBS 6909x (Octa DVB-S/S2/S2X) #3 : DVB-S #0, discarding new
tvheadend    | 2019-06-08 02:46:29.946 [WARNING] mpegts: too much queued table input data (over 2MB) for TurboSight TBS 6909x (Octa DVB-S/S2/S2X) #7 : DVB-S #0, discarding new
tvheadend    | 2019-06-08 02:46:34.459 [WARNING] mpegts: too much queued table input data (over 2MB) for TurboSight TBS 6909x (Octa DVB-S/S2/S2X) #6 : DVB-S #0, discarding new

When this happens, TVHeadend becomes unresponsive. It doesn't happen when only one subscription is active, but the more subscriptions (even EPG ones) activate, the more common the message above and TVHeadend becomes unusable.

As far I can tell, this is related to EPG grabbing. I was able to record 10 channels (as can be seen at the top of the log) in a stable manner. As soon as I stopped recording (that's when the attached trace starts) and epgrab kicked in, the problem came back.

I've seen this issue, but it's marked fixed some 10 commits before master.

Attached is a file with the traces enabled as asked in the above issue.

I'll be trying to bisect tvhdeadend to pinpoint the commit introducing this.