>The man pages fail to say that you must apply a kernel patch. Bela posted
>details for both 3.2v4.2 and SCO 5 about two months ago. Use "fight" in the
>subject and this newsgroup to find it on dejanews. The patch can be applied
>in source form to SCO 5, but needs to be applied with /etc/_fst (= adb)
>to SCO 3. If you have difficulties, I can dig it out from may mail archive.
>(If you don't apply the patch, the time will wander around the true time,
>and exceed the maximum error and be forced to abrubtly step back to the
>correct time, possibly moving backwards, rather than being gradually
>corrected and never going backwards.)
Here's the referenced article:
Subject: Re: Does SCO 3.2v4.2 fight NTP for control of clock?
. . .
Organization: work for SCO, speak for myself
. . .
rate, then settles to the drift rate of the CMOS clock.
Correct. The kernel's once-a-second processing checks the RTC and
attempts to synchronize system time to it by stretching or shrinking
the number of ticks in a second (by no more than one extra-or-missing
Under OpenServer Release 5, this is controlled by a set of variables
/etc/conf/pack.d/kernel/space.c. The main variable is "track_rtc",
set it to 0 to disable this behavior.
There is no user control under ODT 3.0. However, it is trivial to
patch the kernel to disable the behavior. Put a 0xc3 byte ("RET") at
the beginning of function check_rtc():
# echo 'check_rtc?h c3' | _fst -w /unix -
You can apply the same patch to clock.o (in
once you are satisfied that it works.
Take notes so that 1) you'll know what you did so you can repeat it
and 2) when seeking support, you can tell the support person what
done, and can undo it if necessary.