Update on 4/3X0 clock problem

[Keith S. Pickens]

An update on the problem with clock discontinuities on the new 4/3X0
machines.

The bug is listed under BugID#1029022.  Sun engineering has seen the
problem but there is no fix available for it.  In my opinion the problem
breaks the adjtime() syscall and can be expected to impact time dependent
user code by causing difficult to isolate bugs (the user program only
breaks when a critical operation occurs during a system time jump).  The
explanation which I have gotten for the problem does not match the facts
but some bits may have been dropped in "engineer->support_center->me"
chain.  I asked for a schedule for resolution and got a rather arrogant
"go away and stop bothering us, don't call us, we will call you".

>From software support:

" ... we cannot offer you any schedule for when (or wether [sic] ) this
problem will be addressed.  Bugs are addressed in relation to their
relative severity to the customer base as a whole.  This particular
problem has not had a big effect, and hasnt demonstrated itself outside of
the test you provided that checks for it."

So ... if you are trying to run NTP or just want a stable clock call Sun
and complain, loud and long.

An aside: In my recent experience, software support has been doing much
better with routine problems which have already been identified.  However,
I found the level of effort required on my part to get a response on this
new problem was very high.  An all around unpleasant time.

That's all for now on clock problems, I'm retiring from this one.

	-keith
	 ksp at maxwell.nde.swri.edu