[time-nuts] OpenBSD / ntpd / gpsd / PPS problems

Rich Wales richw at richw.org
Sun Oct 25 07:42:26 UTC 2009


Chris Kuethe wrote:

> Hmf. Try this patch to ldattach

Thanks.  This does make ldattach stay around, and I'm seeing data
reported by gpsd, and the uncorrected NMEA timestamps (first shared
memory area) are being seen and reported by ntpd.

However, the PPS-corrected timestamps (second shared memory area)
are still *not* being generated, as best I can tell.

I tried running gpsd with the -D5 option (high debugging level),
and there's no indication in the system logs of *any* PPS events.
In particular, I'm not seeing any log messages in ntpshm.c starting
with "ntpshm_pps".

Even when I removed the "-t" option on the ldattach command -- which
is supposed to cause the NMEA line discipline to base timestamps on
the leading "$" of each block of sentences from the GPS (a sure way
to generate pseudo-PPS events) -- I still didn't see any system log
messages about PPS events.

Note to Hal Murray:  A "line discipline" is an additional processing
layer invoked on a serial line.  I've been trying several different
OS'es and combinations of applications, and OpenBSD appears to have
the most promise for my needs, mainly because PPS support is claimed
to be built into the standard release's kernel (through the NMEA line
discipline code).  For ease of maintainability, I would prefer not to
have to build a custom kernel to get PPS support -- which is what I
would need to do if I use Linux or FreeBSD.

Rich Wales
richw at richw.org



More information about the time-nuts mailing list