[ntp:questions] leap second (was: radioclk DCF doubling)

Serge Bets serge.bets at NOSPAM.laposte.invalid
Mon Jul 18 14:57:16 UTC 2005


 On Monday, July 18, 2005 at 11:04:04 AM +0100, Jonathan Buzzard wrote:

> The problem with leap seconds is that I have never had one to study
> before.

IIUC second 19 of DCF signal will be 1 during 60 minutes before leap
second, so radioclk could inform SHM() hoping ntpd will inform local
kernel and ntp clients. The kernel should then wait midnight to insert a
23:59:60 second.

Difficulty is to reject parasite false 1 in bit 19. Probably take it
only between 23:00 and 23:59 on June or December 31, and only if 5
minutes consistently have it set.

I wonder if using adjtimex to set STA_INS bit of kernel clock status,
either manually or thru cron at around 23:00, can work too.


Serge.
-- 
Serge point Bets arobase laposte point net




More information about the questions mailing list