[vdr] Re: Exit state handling in runvdr

[Klaus Schmidinger <Klaus.Schmidinger@cadsoft.de>]

Jan Ekholm wrote:
> 
> On Thu, 28 Aug 2003, [ISO-8859-1] Jaakko Hyvätti wrote:
> 
> >  In my opinion, the exit line in runvdr is fine for testing, maybe saves
> >you from typing 'killall vdr runvdr' sometimes, but is there really any
> >reason not to make the loop infinite on production?  I do this.  If you
> >are away from the machine, for example 2 weeks on holiday, what else can
> >be done in any error situation that retry infinitely?  Maybe reboot the
> >whole machine if the loop iterates fast for say an hour?  It is a
> >computer, it does not get bored of repetitive tasks..
> 
> Well, I must agree. In the past I've missed a lot of recordings just
> because VDR died with buffer overflows sometimes during the day. So when
> the recording was supposed to start it was dead. I used to always check
> the machine before each recording to see that it was alive and well, but
> that defeated the whole idea of unattended recordings. I've also never
> 
> Actually, a ring buffer overflow doesn't exit VDR here, it just sits
> there. Shouldn't the watchdog restart VDR when it's died or shouldn't VDR
> itself exit when it has reached a situation (buffer overflow) from which
> it can't recover without a restart?

When VDR dies because of a segfault or similar events, runvdr should already
do a restart (at least it does here).

The watchdog makes VDR exit with a restart condition as soon as the main
program loop is no longer processed, maybe because the program is stuck
somewhere.

If you want additional "emergency exists" when buffer overflows or things
like that happen, you can just implement the necessary call at the right place.
But be prepared for lots of complaints from people who are already complaining
about the "emergency exists" VDR does as it is now ;-).

Klaus


-- 
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe vdr" as subject.