On Mon, Jul 31, 2000 at 11:33:41AM +0100, Ben Harris wrote:

 > I'd like some opinions on the following proposed change, which isn't quite
 > in my field.
 > 
 > The purpose of this change is to get rid of the stub kern_ntptime.o that's
 > built if you don't turn on "options NTP", since it appears to be
 > redundant.  The only semantic change here should be that processes calling
 > ntp_gettime() will now get a SIGSYS in addition to its returning ENOSYS.  
 > As far as I can see, sysctl_ntptime() isn't used unless NTP is defined
 > anyway, so losing that is safe.
 > 
 > Comments?

This seems correct to me.  I actually made similar changes some time
ago, but never got around to committing them.  Make sure to get all
the NTP-related syscall vectors in the emulations, too.

Looks like ntpd(8) catches SIGSYS to deal with the case that the
kernel does not support those system calls, so the `side-effect'
you describe isn't going to cause any problems.

Go for it!

-- 
        -- Jason R. Thorpe <thorpej@zembu.com>