[mythtv-users] Does Myth not realise when it looses signal ?

[f-myth-users at media.mit.edu]

    > Date: Wed, 16 Dec 2009 18:20:08 +0000
    > From: Paul <paul at paulhurley.co.uk>

    > I went away last weekend, and came back to four days of records with 
    > 'File Size B', indicating it didn't record anything.  I eventualy traced 
    > it to a user induced hardware error (I pulled out the aerial cable from 
    > the machine while looking for something in my garage !).

    > Does myth not realise when it can't get a lock, or when it records 
    > something with zero length ?  Ideally it would have logged it so I could 
    > see it easily...

You could check out http://svn.mythtv.org/trac/ticket/6899, which
solves a slightly different problem, and adapt it.

That tool looks for suspiciously silent sections.  If your recordings
are zero-length, it'll never fire, because a recording that short
can't -have- a silent section.  However, -if- the following are true
of these failures (I don't know, since I don't have your setup):
(a) the zero-length file is written at the -beginning- of the recording,
(b) its modification-date stays unmodified until it's closed at the -end-
    of the recording (e.g., nothing's touching it in between---even if
    no new data is being appended)
then you could replace scan-and-notify-on-eas in notice-new-files-dispatcher
with something that merely says, "If this file is zero-length, scream
by calling `nc -w 1 ds your-port-here'" (as scan-and-notify-on-eas does)
assuming you have a listener elsewhere on your network.  Or you could
call something that logs to syslog.  Or have it send you mail.  Whatever.

Anyway, you could start with (just) the two notice-new-files*
files (which are both really simple shell scripts), make a trivial
little script that says "this file is too short", and make that log
somewhere.  [And if zero-length files are the only problem you ever
see, you can ignore the rest of the tool.]

(If you decide to do that, post it, and I'll probably roll it into a
newer version of that someday.)

[Be aware that that tool is not storage-group aware; you need to run
one instantiation of notice-new-files per directory you'd like to
monitor, or you need to modify it to take more than a single directory
as an argument and scan 'em all.]

P.S.  But I thought that 0.22 (trunk?) had some code to specifically
detect zero-length recordings and mark them failed.  True?  False?
Unreliable?  Not yet implemented?