On Fri, Mar 14, 2003 at 10:48:43PM +0000, Martin J. Laubach wrote:
> | > Yes - it wasn't my fault, the bug has been there for (probably) ever.
> 
>   Is bin/17609 the same thing? 

No!  This particular 'feature' always adds 2^32 usecs (just over
an hour) so won't be responsible for those values.

There are cases when microtime() will go backwards (if nothing else
when the clock is set back).

It is probably worth making both calcru() and mi_switch() ignore
-ve times.

Also microtime() need to be returning 'time since boot' so that it
is monatonic modulo hardware problems.

I think I decided that christos's system actually had a broken
timer chip.  My system reports -ve microtime deltas during the
boot process (with the cycle-counter based microtime).


	David

-- 
David Laight: david@l8s.co.uk