[mythtv] Re:DVB Todo

[Kenneth Aafløy]

Hi,

Martin Smith wrote:
>> + Fix A possible silent loop in StartRecording
> It's possible that we've both changed the code for this one.
>
>> + Fix polling before every read
>> + Larger buffer for ts packets
> I haven't looked at these as I couldn't see any overflows on
> the stream, as we discussed.

I did those three because the StartRecording function was getting
to large, so I move out the reading and packet handling into
a new function, which also handles reading as much as possible
into a user definable buffer. At the minimal it skips a couple
of extra system calls..

>> - Fix silent EOVERFLOW in recorder
> I have done this, it now gets logged at RECORD level. I didn't
> know whether it should exit recording on repeated errors.
> Any opinions?

I don't think exiting is an option (i would preffer my recordings
to have holes in them over beeing cut short), but a warning at a
higher level than record would be good, because the record level
dumps quite a lot of data, if i'm not mistaken. Did you see this
get logged at any time, and would the continuity counter warning
catch it? If so, the warning at record level is probably ok.

>> + Slightly cleaner first continuity counter fix
>
> It's possible that this overlaps with my changes. I've made it
> log the frontend stats on each stream continuity error eg:

There was two arrays beeing used for the job when 1 would suffice,
if their purpose was combined. And both of them where leaking.

> The biggest thing I did yesterday is add a database table called
> dvb_signal_quality. When enabled a thread in the DVB recorder
> writes data to this table every N seconds.

Would not this belong in the DVBChannel class, along with the
signal monitor that is in there already?

> This includes the frontend signal stats and the number of stream
> errors, bad packets and overflows at the application level. Data
> is retained for M days where N and M are configurable via a new
> settings screen.

I dunno what you plan on making this do beyond logging it? Is the
intention to make a legend plot so that the user could easily see
signal dips?

> It worked fine for the recordings I did overnight. My plan was to
> do more testing before submitting it but if it might conflict with
> anything anyone else is doing with the code I'm happy to send you
> what I have for review.

I'm sure it's okay, but mail it to me or the list.

Kenneth