Re: Two bugs in kernel time code

Philip Gladstone (philip@raptor.com)
Thu, 18 Sep 1997 09:52:47 -0400

Rogier Wolff wrote:
>
> Philip Gladstone wrote:
> > My revolting approach to this is to have a sysconfig variable that gives
> > the offset of the CMOS clock from UTC. This defaults to 'unknown'. If it
> > is known, then the CMOS clock is updated with the correct hour (etc)
> > values.
>
> How about just calculating the difference between the unix clock and
> the CMOS clock. If that's within 10 minutes of a whole number of hours,
> then you can safely round it to "a whole number of hours".

My goal was to eliminate the 'clock -w' that you have to do
occasionally.

There are two steps to this:

1) Fix settimeofday to set the CMOS clock. I do this if I know
the cmos clock offset (in my systems, we run CMOS=GMT). Set all
the fields in the clock.

[I use a minute offset field, with -1 being the 'unknown' value]

2) Do something mysterious with the clock warping code for
backwards compatibility. [not done]

I could provide these patches if anybody is interested.

Philip

-- 
Philip Gladstone                           +1 617 487 7700
Raptor Systems, Waltham, MA         http://www.raptor.com/