[ntp:hackers] What happened to old sample checking?

David Mills mills at udel.edu
Tue Apr 28 02:31:37 UTC 2009


Looks like a glop of code got lost in edit. I'll check further.

The reason the check was there is to make sure the time constant doesn't 
get screwed up on a clock hop that might result in old data. There is 
another check that brackets the time constant should a clockhop occur to 
a source with a different range of minpoll/maxpoll.

Thanks for the sharp eye. Keep at it.


Brian Utterback wrote:

> Some time ago, Dave added a check in clock_update such that is the 
> sample being used to update the clock was older than the last sample 
> used to update the clock, the older sample would be ignored. This 
> ensured that the samples applied to the clock would be in 
> chronological order.
> Looking at clock_update recently, I see that this check is no longer 
> in place. It was there in p161 and is not there in p167.
> So, I guess this is really a question for Dave. What happened? Was the 
> check moved somewhere else, or did you decide that it wasn't working 
> right, or that it was not the right thing to do?  I am just trying to 
> understand what happened.

More information about the hackers mailing list