[ntp:questions] refclock use causes core dump of ntpd

[Ronan Flood]

"wa6zvp" <wa6zvp at gmail.com> wrote:

> * Oh, it does pass through doevent a couple of times before the crash.
>  The above are the values on the call that will end up at abort().

Ah, OK.

> * I remembered yesterday as well that optimisation can mess up
> debugging so
>  I removed the -O2 from the cflags.  Made no change to the symptoms at
> all.

Huh.  (or indeed e_Huh ...)

> >             case t_unknown:
> > +               if (event == e_Poll)
> > +                   break;

> * Yea, that looks simple enough.  I'll try that today.

Failing that (!), change true_receive() around line 540 so that e_Poll
is only called for t_goes:

		if (up->type == t_goes)
			true_event(peer, e_Poll);

> I am also of the opinion that refclocks should not call abort (or
> exit).
> Simply not getting data for the clock should be indication enough that
> something is wrong.

Indeed -- for those who check with ntpq, which not everyone does.

> Thanks for your assist in this, Ronan.

No problem.

-- 
Ronan Flood <usenet at umbral.org.uk>