[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

clock-setup: should not change hardware clock if the system time is not updated

Package: clock-setup

From a thread in -boot:

> On Thursday 13 August 2009, Uwe Bugla wrote:
> > a. as default installation I always use expert install
> > b. the NTP question I always answer with "Yes". The installer tells me
> > that my local time zone is "Europe / Berlin"
> > c. The UTC question I always answer with no, without having some crap
> > like Windoze installed on my HD. The question of what makes sense and
> > what does not I do leave open.....
> Using local time for your hardware clock when you're only running Linux 
> really makes no sense at all.
> The most obvious example is a notebook that you travel around with. If the 
> hardware clock is set to UTC you can take the notebook anywhere around 
> the world and always have the correct local time simply by changing the 
> time zone.
> But even for systems that don't travel local time does not make any sense. 
> Why would you want to change the hardware clock when summer time 
> starts/ends? It does not improve the way the system is used in any way.
> > d. After the installation is finished the clock setting of my BIOS was
> > trimmed by 2 hours which I wanted to avoid.
> I agree that what you describe looks like a bug.
> I suspect that the cause is the following:
> - the installer itself has no knowledge of timezones, it just runs in
>   "current" time, taken either from the hardware clock or rdate
> - rdate always updates the system time to UTC
> - we update the hardware clock from the /target chroot *after* configuring
>   the chroot to reflect that the hardware clock is supposed to be in local
>   time, but the time that gets written is the system's current time, which
>   is UTC.
> So basically finish-install.d/10clock-setup is wrong and should run 
> hwclock --systohc --utc _before_ changing the local/utc setting in
> /etc/adjtime.
> But that should be done *only* if rdate was run successfully. In fact, I 
> don't think there is any reason to write the hardware clock at all if the 
> system time was not updated by rdate.
> Uwe: please file a bug report against clock-setup with the details of the 
> problem, including the description of the choices you make during 
> installation and my analysis.

Well, I think that all needed info is in the above quote, so I take
responsibility for opening that bug (will reassign it to Uwe as
submitter). Uwe, no need for you to do it.

I agree with Frans (who, btw, corrected me on one point: when Windows
is detected on the system, defaults installs do *not* prompt about the
HW clock being UTC or not....as Windows nearly enforces having the HW
clock set to local time). We apparently have a bug here, that was not
discovered because very few people have the local time for the HW

Attachment: signature.asc
Description: Digital signature

Reply to: