[ntp:hackers] Time constant too large?

David L. Mills mills at udel.edu
Tue Jan 8 03:53:25 UTC 2008


As the author of the code, I declare it a bug, by bug. It was my intent 
that the time and frequency updates be atomic. But, the fix for the bug 
should be to upgrade to the nanokernel.

In any case, I modified the code to do the frequency initialization 
separate from the time adjustment. In the process, I see the frequency 
initialization is not as precise as I had intended. Workin' on it.

Dave

Brian Utterback wrote:

> First, do you assert that it is in fact a bug in the kernel? It is 
> following the behavior
> of the reference code for the microkernel.  If it is a bug, then it 
> should be easy
> enough to get fixed going forward, but I am concerned about people 
> building
> for currently released kernels.
>
> As for upgrading the kernel to incorporate the nanokernel, that is my 
> intention. The
> build process is not nearly as baroque (broke?) as it once was. Now 
> the it is open
> source, it can be built from freely available tools. However, the 
> weeds that have
> grown over the last ten years have roots the twine throughout the 
> code. Pulling them
> out will be rather tricky.
>
> Do you have any interest in running Nevada on any of the herd, 
> preferably something
> with a PPS capable refclock attached to test on?
>
> s n i p




More information about the hackers mailing list