Major Clock Drift

[David Kramer]

On Mon, 30 Aug 2004, Jerry Feldman wrote:
> There are two issues involved. The RTC clock in your computer starts the
> time when booted, but the OS maintains its own time based on a clock
> tick interval that drives interrupts. Linux sometimes can store its
> concept of time on shutdown. 

Here's a crazy idea.  Maybe I'm talking out of my /dev/stdout, but could 
the CPU and/or bus speeds be set wrong in the BIOS setup?  Doesn't Linux 
use those numbers for the clock math?

> Take a look at the hwclock(8) command. 
> 
> One definitive method to really check the clock is to leave your
> computer in the CMOS setup mode or boot an OS that does not muck with
> the clock, like (ugh) MS-DOS. The CMOS setup will display the hardware
> clock. So, figure out what your drift is over a fixed period of time,
> bring it up under setup or DOS, record the time, and take a sample at
> some later time. It could be that the RTC is reasonable, but the clock
> tick interval is off causing the drift.

Excellent test.  Just leaving up the CMOS setup screen is easy and 
definitive.


> and Make(1) have trouble.  

Yup.  At Polaroid, this guy couldn't get an app to compile at all, and was 
getting strange error messages, so he called me over.  Duh, his machine's 
date was set for sometime in 1971.


----------------------------------------------------------------------------
DDDD   David Kramer         david at thekramers.net       http://thekramers.net
DK KD  
DKK D        You hear that? That's the sound of a thousand horrible 
DK KD        things headed this way.
DDDD