I just thought I would give you an update. In December, I uncovered a rather serious problem in the kernel loop as used by the NTP v4 software. It turned out to be a bug in the original micro-kernel code as released from UDEL. The v3 software did not exercise the bug, and the nano-kernel code that is the current object of NTP work did not have it, so it remained dormant.
I got Dr. Mills to code a workaround into the v4 software, and I was able to fix the bug in Nevada, so we have a belt and braces solution. Unfortunately because of the holidays, the new v4 software wasn't released until last week. This uncovered another bug in the v4 code, which has now been fixed. Even more unfortunately, the new software seems to have another bug. We are still in testing to determine if the problem is in the Solaris kernel or the v4 code. However, the new bug is only evident if the kernel loop is enabled; it does not affect the daemon loop, so there is an easy workaround. All you have to do is put "disable kernel" into the ntp.conf file. So, I am in a quandary. Should I release the bits I have for testing, with the caevaet that the kernel loop should be disabled, or should I wait until this current problem is resolved? Any ideas? What do you all think? Brian Utterback
