[ntp:hackers] Any chance of someone looking at bug 1189?

Dave Hart davehart at gmail.com
Wed Jun 10 15:16:09 UTC 2009


On Wed, Jun 10, 2009 at 10:05 AM, David J
Taylor<david-taylor at blueyonder.co.uk> wrote:
> Any chance of someone looking at bug 1189?
>
> 4.2.4 and 4.2.5 appear to have grossly different behaviours on Windows
> Vista and Windows-7 systems.  Some help in resolving this would be
> appreciated.

Vista and Win7 systems are different than older Windows releases
because they have a system clock precision around 1 msec (-10).  With
my private "QPC" binaries built on 4.2.4p6, and with 4.2.5p162 and
later, the Windows interpolation scheme is disabled on Vista and
later, and ntpd runs with a precision around -10.  With all official
4.2.4 releases on Windows, and with earlier 4.2.5, ntpd performance on
Vista and later is miserable due to failing interpolation.

This makes Windows a poor choice for a platform to track down the
apparent regresion sometime during 4.2.5 development, as 4.2.4 and
most 4.2.5 releases can't be tested on Windows Vista and later
meaningfully.

Assuming there is a generic issue with ntpd 4.2.5 on low-precision
systems compared to 4.2.4, it would be much easier to identify it on a
different platform, where a binary search of 4.2.5 releases could be
tried to hone in on the change responsible.  To that end, I've tried
to repro this on WRT54G routers running OpenWRT.  The first time I
tried to pursue that approach, the person with the OpenWRT box had a
specific problem I helped them with, and then they disappeared before
I had a chance to try to reproduce the regression.  Another IRC user
responded to my call for OpenWRTs and dusted theirs off and started
looking at cross-compiling ntpd for it, but has not made any progress
in the last few weeks, presumably due to other priorities.

If you have built custom OpenWRT images and are willing to help
investigate this apparent regression in ntpd performance on
low-precision systems, please give me a yell.

Cheers,
Dave Hart


More information about the hackers mailing list