[ntp:bugs] [Bug 740] The frequency in the kernel time discipline is initialized from the drift file, even when the kernel time discipline itself is not used

[David L.Mills via the NTP Bugzilla]

http://bugs.ntp.isc.org/740



----------------------------------------------------------------------------
Additional Comments From mills at udel.edu (David L. Mills)
Submitted on 2007-04-17 03:07

Subject: The frequency in the kernel time discipline is initialized
 from the drift file, even when the kernel time discipline itself is not used

Juergen,

This has been easily the most nasty patch of weeds in a long while. It's 
been worked over by several people who made nasty little workarounds. 
The problem is that the default, -x option and configuration file are 
handledifferenctly and step on each other.

Your suggestions won't work. They would result in the initial frequency 
to be in error until the first clock update. The object is to disable 
the kernel if either -x or tinker step 0 s or over 0.5 s.

I pulled out still more code, moved the default as an initializer, set 
the -x function consistent with the kern_enable function. This is done 
in the ntp-dev version, but might take awhile to escape to the web.

Unfortunately this might not work due to possible order dependency in 
the configuration file. This only works if the -x option or the tinker 
step option precede the frequency file processing in the LOOP_DRIFTFILE 
section in ntp_loopfilter.c. The command line option processing is 
beyond me; the call to the LOOP_DRIFTFILE section in ntpd.c looks wrong 
and very different from where I last placed it (in ntp_util.c).

Dave

Juergen Salm via the NTP Bugzilla wrote:

[---=| TOFU protection by t-prot: 37 lines snipped |=---]


-- 
David L. Mills <mills at udel.edu>



------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.