[LinuxPPS] LinuxPPS - odd behaviour.

[Udo van den Heuvel]

See?

http://bugs.ntp.org/956



----------------------------------------------------------------------------
Additional Comments From kardel at ntp.org (Frank Kardel)
Submitted on 2007-11-10 12:59

This has *never* been a problem for *all other* PPSAPI implementations.
This is a linuxpps specialty. If you have standalone PPSAPI capable devices
like GPIO they should provide a PPSAPI capable device that can be opened
without opening a "master" device.

Your problem is (from the linuxpps  mailing list) a linux setup like
this (for the sake of other readers here):
/dev/ttyS0 for serial communication
/dev/lp0 for parallel port not used for communication but needed open
for opening
/dev/ppsX for the PPS from lp0.

Instead of adding changes to ntpd for a specific PPSAPI implementation I
recommend to try to figure out a way in linuxpps to open the pps-device even
if the "master" device hasn't been opened. Why shouldn't GPIOs not being
true
PPS devices?

The reason I accepted the patch for refclock_nmea is that it is useful
in other
environments too. I currently lack the imagination why atom needs to open
one device it doesn't need for communication just in order to open the
real PPS
device. This looks (currently) like a very specific implementation
restriction and
we *try* to keep amount a special code low.