[mythtv] [PATCH] tuning failure and leak fix for dvb_on_demand mode

[Julian Scheel]

I tested dvb_on_demand with and without your patch.
It doesn't work at all for me.
If I start myth, it opens the card, tunes to a channel and closes the channel 
immediatelly. So far everything seems ok.
Now I start watching TV. The Backend spots the following:

2004-07-23 22:34:23 Changing from None to WatchingLiveTV
2004-07-23 22:34:23 DVB#0 Recorder: Card opened successfully.
2004-07-23 22:34:24 DVB#0 WARNING - No data from card in 1 second.
2004-07-23 22:34:25 DVB#0 WARNING - No data from card in 1 second.
2004-07-23 22:34:26 DVB#0 WARNING - No data from card in 1 second.
2004-07-23 22:34:26 Changing from WatchingLiveTV to None
2004-07-23 22:34:27 DVB#0 WARNING - No data from card in 1 second.
2004-07-23 22:34:27 Closing DVB recorder
2004-07-23 22:34:27 Closing DVB channel
Killed

That's it.
Any ideas what this could be caused by?

I use a WinTV Nova-s card, but had the same issues already with a Pinnalce 
PCTV Sat card.

Without dvb_on_demand myth works absolutely stable.

Julian

Am Samstag, 10. Juli 2004 14:01 schrieb Christopher Pascoe:
> There are a few problems in dvbchannel.cpp that are exposed when using
> dvb_on_demand mode, that I have addressed in the attached patch:
>
> Firstly, myth does not retune after reopening a frontend device if it
> still thinks it is on the same channel as it was when it closed the
> device.  This causes problems as the frontend is always initialised by the
> dvb-kernel drivers on reopen (which, depending on your frontend may cause
> a loss of tuning), or it may have been retuned along the way.  Setting
> first_tune at device open time rather than DVBChannel instantation time
> fixes this.
>
> Secondly, when closing the frontend connection, myth doesn't close its
> connection to the demultiplexer, etc, and generates a new one every time
> the frontend is reopened.  Moving the deletes for dvbsct and dvbcam into
> CloseDVB fixes this and stops them from being leaked.  Also add a delete
> for the diseqc at destruction time.
>
> Finally a minor nitpick - there are a few tests that assume a fd of 0 is
> invalid for the frontend connection.  While this may be the case in Myth
> on Linux at the moment, it mightn't always be.  Fix this and use -1 for
> the invalid fd.
>
> Regards
> Chris

-- 
Grüße,
Julian