Hello to everyone, I'm new to the list and would like to ask about displaying the upcoming leap second on a linux workstation. I use a PC running linux 2.6.24 and the djb clockspeed daemon periodically synchronized with a couple of different NTP servers. I usually have a correct reading within ~10ms (I know, it's not atomic time 8-) I also use the "right/America/Los_Angeles" time zoneinfo file for correct synchronization with TAI, and installed the new leapsecond.dat file for 2008 in /etc. (and I restarted clockspeed) I'm on the pacific coast of the US and as such the leapsecond occurs here just before 4pm. I've used the date command to set the time of my computer to Dec 31 2008 15:58 PDT and watched as the hour rolled over: No LeapSecond! It goes from 15:59:59 to 16:00:00 I've been using XDaliClock as the display, which uses the unix localtime() call to retrieve the time from the system. I've done quite a bit of scroogling in an attempt to determine how the time is kept in the kernel and how it makes it's way through to userspace for applications like date and xdaliclock to access. I didn't determine 8-( Can anyone advise me as to the feasibility of using this configuration to display the leapsecond to a crowd of eager viewers at the end of this year? Thank you so much for your consideration of my noobish question, and for maintaining an active interest in something just because you like it! ciao bella... johnea

> I'm new to the list and would like to ask about displaying the > upcoming leap second on a linux workstation. Stick around. There will be a big party. You might check the archives to see the discussion from the last one. > I've done quite a bit of scroogling in an attempt to determine how > the time is kept in the kernel and how it makes it's way through > to userspace for applications like date and xdaliclock to access. The kernel works in UTC. Conversion to/from local time is done in userspace, usually in the c library. Daylight time is a nightmare because politicians keep changing the rules. There is a project that tries to keep track of things. Essentially, all of the open source projects use it. http://en.wikipedia.org/wiki/Zoneinfo The default conversion info is stored in /etc/localtime There is an environment variable, TZ (I think) so you can use your preference rather than the default. This might be handy if you dial in to a system in another zone. -- These are my opinions, not necessarily my employer's. I hate spam.

 I ask you this question because before the previous leap second, 

 I tried to get more help from our time researchers, but they simply 

Hello! I am sorry that I cannot give you any help. Instead I have a question to you: How and where did you get to know about the "leapsecond.dat" file? I ask you this question because before the previous leap second, January 1, 2006, I had a set-up similar to yours. I did not use or know about any "leapsecond.dat" file - that seems to be the reason why the clock in my computer was off by about (I am not sure today) 33 seconds when I used the ".../right/..." time zoneinfo file. As far as I remember, the clock _did_ show seconds 58, 59, 60, 00, 01 (but was off by half a minute). I tested this in advance, like you have done, but I could not test it at the actual event: because of social reasons, I had to spend the time at a New Year party :-) I tried to get more help from our time researchers, but they simply recommended that we turn off our Linux-based NTP servers a short time before the leap second, and then turn them on afterwards. That way, they would get the correct time initialisation from the Motorola Oncore GPS receiver, and the seconds would be in 'exact' phase because they are created by 1-PPS signals from four rubidium clocks and a cesium clock which were on continuously. Such a solution works, of course, but I would like to have a proper leap second handling in the kernel and the ntp protocol. Later, I had a discussion with a specialist at a university about paying him for an investigation of how the ntp and Linux kernel handle leap seconds, but that work was never initialised. There was also a discussion about the Linux kernel, ntp and leap seconds in the Linux kernel mailing list, but as far as I remember, there were radically diverging opinions on how these things should work together. I have not followed the kernel mailing list since then. Regards Uno ----- Original Message ----- From: <time@johnea.net> To: <time-nuts@febo.com> Sent: Thursday, July 10, 2008 8:15 AM Subject: [time-nuts] viewing the upcoming leapsecond with XDaliClock > Hello to everyone, > > I'm new to the list and would like to ask about displaying the upcoming > leap second on a linux workstation. > > I use a PC running linux 2.6.24 and the djb clockspeed daemon periodically > synchronized with a couple of different NTP servers. I usually have a > correct reading within ~10ms (I know, it's not atomic time 8-) > > I also use the "right/America/Los_Angeles" time zoneinfo file for correct > synchronization with TAI, and installed the new leapsecond.dat file for > 2008 in /etc. (and I restarted clockspeed) > > I'm on the pacific coast of the US and as such the leapsecond occurs here > just before 4pm. > > I've used the date command to set the time of my computer to Dec 31 2008 > 15:58 PDT and watched as the hour rolled over: No LeapSecond! > > It goes from 15:59:59 to 16:00:00 > > I've been using XDaliClock as the display, which uses the unix localtime() > call to retrieve the time from the system. > > I've done quite a bit of scroogling in an attempt to determine how the > time > is kept in the kernel and how it makes it's way through to userspace for > applications like date and xdaliclock to access. I didn't determine 8-( > > Can anyone advise me as to the feasibility of using this configuration to > display the leapsecond to a crowd of eager viewers at the end of this > year? > > Thank you so much for your consideration of my noobish question, and for > maintaining an active interest in something just because you like it! > > ciao bella... > > johnea > > > _______________________________________________ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. >

Hello all, I'm playing around with an NTP server using a Blackfin processor and uClinux, and a Oncore M12M-Timing. The board is our H8606, http://www.hvsistemas.com/en/prod/H8606.html , with the standard crystal oscillator replaced with a FOX924B TCXO, http://www.foxonline.com/pdfs/fox924.pdf . I've modified the kernel to support PPS from the Oncore, with timestamping done through the processor NMI and the cycles registers, yielding a 2.5ns resolution, and the preliminary observed jitter in the timestamping seems better than ±25ns (previous to applying Oncore sawtooth correction). So until now, it seems quite good :). I was thinking if the Allan intercept parameter of the ntp daemon could be optimized for this situation, since it has a default of 1500s that I suppose that is the adequate value for general usage (synchronization of a rather crappy computer oscillator from a WAN with quite spaced pollings and relatively high delays), and I presume that could be reduced for this system. Any thoughs would be appreciated :) Best regards, Javier -- ------------------------------------------------------------------------ Javier Herrero EMAIL: jherrero@hvsistemas.com HV Sistemas S.L. PHONE: +34 949 336 806 Los Charcones, 17A FAX: +34 949 336 792 19170 El Casar - Guadalajara - Spain WEB: http://www.hvsistemas.com