Re: [gpsd-dev] ntpd/gpsd tangle

[Eric S. Raymond]

Hal Murray <address@hidden>:
> 
> address@hidden said:
> > OK, I want to be sure I know what's going on here.  You were restarting ntpd
> >  *while gpsd was running*?
> 
> Sure.  I've been doing that for years.  It works with SHM.
> 
> I assume I can start/stop/restart them in any order.

If you're using the SHM driver restarting them in any order will work.  
In that case the shared-memory segment defines a fixed rendezvous; gpsd
doesn't actually care whether an ntpd instance is reading it, but just
dumps updates there.  After an ntpd restart ntpd can just go back to
the rendezvous.

The socket-export case is different.  For that to work, ntpd has to behave 
as a GPSD client - start a session on port 2947 and issue a WATCH.  It's
OK for ntpd to do this multiple times, but only if it does a new session
start and issues a new watch for each connect.  

Which, now that I think about it, it certainly ought to be doing.  Hmmm.
 
> Restarting ntpd is the same as control-Cing a random program JSONing to gpsd. 
>  (or having a program crash)  I expect gpsd to recover from that.

You're right.  I'm going to run some smoke tests and get back to you.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>