From: Arjan.deVet@adv.IAEhv.nl (Arjan de Vet)
Message-Id: <199512302301.AAA05669@adv.IAEhv.nl>
Subject: Upcoming leap second 1996-01-01
To: questions@freebsd.org
Date: Sun, 31 Dec 1995 00:01:16 +0100 (MET)
Content-Type: text
Sender: owner-questions@freebsd.org
Precedence: bulk

Last year, on July 1st 1994, some accounting scripts of mine did strange
things because of a leap second the previous day which made that day 86401
seconds instead of 86400. I think I was using FreeBSD 1.1 at that time.

At January 1st, 1996 there will be another leap second. I started checking
some things in order to avoid the same problems this time, using FreeBSD
2.0.5 now.

I did some tests and it appears that FreeBSD now uses POSIX' idea of
time_t values, i.e., it does not support leap seconds and the internal
clock should be adjusted with -1 second on 1996-1-1 00:00:00 UTC. I
understand from RFC1305 (NTP) that this will be done automatically because
I'm using NTP.

Am I correct? Did FreeBSD 1.1 not use the POSIX idea of time_t values,
i.e., it was counting the real number of seconds since 1970-1-1 including
the leap seconds? That could explain why my scripts failed then because
they relied on a day being 86400 seconds and not 86401 seconds.

Happy New Year!

Arjan

--
Arjan de Vet                                           <devet@IAEhv.nl> (IAE)
Internet Access Eindhoven (IAE)             <Arjan.deVet@adv.IAEhv.nl> (home)
URL: http://www.IAEhv.nl/iae/people/devet/