[ntp:questions] Windows ntpd April Fool's bug when using leapfile

Dave Hart hart at ntp.org
Tue Mar 13 10:10:42 UTC 2012


There is a bug in scheduling a leap second more than 3 months in
advance specific to the Windows port of ntpd, introduced sometime
after 4.2.4.  It results in scheduling a spurious leap second
insertion for midnight April 1, 2012 if a correct leapfile is loaded
(scheduling the July 1, 2012 leap second insertion).  The event log or
configured ntpd.log shows:

10 Mar 05:18:33 ntpd[15836]: 0.0.0.0 c01e 0e TAI 35 leap 201207010000
expire 201212280000
10 Mar 05:18:40 ntpd[15836]: 0.0.0.0 0619 09 leap_armed
10 Mar 05:18:41 ntpd[15836]: Detected positive leap second
announcement for 2012-04-01 00:00:00 UTC

This is corrected in 4.2.7p263.  Since very few people use ntpd on
Windows compared to other systems, and of those very few load a
leapfile manually, there are currently no plans to fix this for 4.2.6.
 A workaround is available:  comment out leapfile in ntp.conf until
April 1.

So far, all leap insertions have been scheduled on the preferred
January 1 or July 1 dates, but the IERS notices could in the future
schedule a leap second insertion or deletion on the less-preferred
April 1 or October 1 dates.

Thanks for your time,
Dave Hart


More information about the questions mailing list