[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ntp-hackers
Subject:    Re: [ntp:hackers] Refclock drivers update
From:       "David L. Mills" <mills () udel ! edu>
Date:       2001-12-21 21:43:56
[Download RAW message or body]

Sven Dietrich,

I hope you are reading this. I cranked up a Palisade GPS to check that I
didn't break anything and found some good and some evil. The good is
that I apparently didn't break anything. The bad is that there are
numerous places where an error message is conditionally compiled with
the DEBUG flag; however, in many of these there is no if (debug)
conditional. As is, debug messages go to the bitbucket at some cost
under normal operation when the daemon is disconnected from the
controlling terminal. I haven't checked all the drivers for this little
violation, but I did find yours.

A suggestion. You have lots of error messages that would be useful for
user eyeball with ntpq. What I have done in other drivers is use
pp->lastcode as eyeball assist. If the timecode is blatently invalid,
use that buffer in place of the timecode so that it is visible to ntpq.
Always send a valid timecode to clockstats, but also send the error
messages which wind up in the buffer, then tack a \n at the end and
expose for debug trace if the debug switch is lit. If it turns out
volumes of error messages would splurge the log, watch flag4 and send
only if that is lit.

Dave
[prev in list] [next in list] [prev in thread] [next in thread]