On Sun, 17 Nov 1996, Perry E. Metzger wrote:

> 
> Dave Huang writes:
> > Which reminds me... could DST and TIMEZONE (or at least DST) be made into 
> > a sysctl?
> 
> I've thought about it, but its hard, partially because there aren't
> hooks in the time handling to deal with it. The offset is first used
> before even init(8) has started, and the *todr functions would get
> confused if things just jumped after boot time. I've been
> contemplating changing the design of this whole thing a bit to
> simplify the problem.

Actually it's a bit worse than that.  I looked into the problem a while
back and it seems that when the kernel mounts the root filesystem, if the
filesystem's time is greater than the current time, the kernel will assume
that the RTC is wrong and use the date from the rootfs.  If you want to
fixup TIMEZONE and DST, you have to do it *before* mounting root if you're
int the western hemisphere, or you will be setting it to some random value
between *now* and *TIMEZONE*.

=========================================================================
Eduardo Horvath				eeh@btr.com
"Cliffs are for climbing.  That's why God invented grappling hooks."
					- Benton Frasier