Replied: Mon, 05 Jan 1998 16:33:37 -0500
Replied: "Poul-Henning Kamp <phk@critter.freebsd.dk> Dave Mills <mills@huey.udel.edu>, mills@udel.edu"
Received: from huey2.ee.udel.edu by mail.eecis.udel.edu id aa26800;
          5 Jan 1998 12:46 EST
Received: from critter.freebsd.dk by huey.udel.edu id aa17104;
          5 Jan 1998 12:45 EST
Received: from critter.freebsd.dk (localhost.cybercity.dk [127.0.0.1])
	by critter.freebsd.dk (8.8.7/8.8.7) with ESMTP id SAA00396;
	Mon, 5 Jan 1998 18:42:38 +0100 (CET)
	(envelope-from phk@critter.freebsd.dk)
To: Dave Mills <mills@huey.udel.edu>
cc: mills@udel.edu
Subject: Re: probably a bug... 
In-reply-to: Your message of "Mon, 05 Jan 1998 09:03:55 EST."
             <199801050903.aa16175@huey.udel.edu> 
Date: Mon, 05 Jan 1998 18:42:38 +0100
Message-ID: <394.884022158@critter.freebsd.dk>
From: Poul-Henning Kamp <phk@critter.freebsd.dk>

In message <199801050903.aa16175@huey.udel.edu>, Dave Mills writes:
>P-H,
>
>This could be fixed by toodling the code where it is noticed the file
>does not exist;

This is what it does now.  Unfortunately that also only gets executed
after the "driftfile" instrux has been found...  No it would need to
be fixed so that if at the end of reading the config file we havn't
done a loop_config(), then we do one there.

>I would be happy to see this
>fixed; however, operating without a drift file would be pretty gruesome,
>since it takes up to a day to refine the loop frequency after every
>restart.

Ohh sure, I only did it because I wanted it to start from scratch since
this exposed some instability in my end.  I may look at a fix, I just
wanted to be sure this was not some hideous secret way to blow up the
world by accident.

Status btw is that I have rewritten the entire time-keeping mechanics
of the FreeBSD kernel (just to see if it is better this way) and am
testing it right now.  So far it looks about the same, but more modular,
and it gives us more knobs to twist so I think I like it.

Another couple of things I noticed:

  * If you have a prefer peer, and pps_sync, the maxpoll of that peer
    should be 2^6.  If goes higher you will loose pps_sync (after 120s)
    (I have no fix for this yet)

  * there is a mistake in the ntpdc code, in doconfig() which prevents
    you from adding a "prefer" peer:

diff -u -r ntp-4.0.70a-export/ntpdc/ntpdc_ops.c copy/ntpdc/ntpdc_ops.c
--- ntp-4.0.70a-export/ntpdc/ntpdc_ops.c	Thu Oct 30 19:13:49 1997
+++ copy/ntpdc/ntpdc_ops.c	Sun Jan  4 23:44:44 1998
@@ -1124,7 +1124,7 @@
 	while (pcmd->nargs > items) {
 		if (STREQ(pcmd->argval[items].string, "prefer"))
 		    flags |= CONF_FLAG_PREFER;
-		if (STREQ(pcmd->argval[items].string, "burst"))
+		else if (STREQ(pcmd->argval[items].string, "burst"))
 		    flags |= CONF_FLAG_BURST;
 		else {
 		        long val;

Give me a trigger a little before you roll the next set of tar-balls
of xntp so I can send you the refclock_oncore diff for inclusion.

Poul-Henning

PS: if your 97-98 year change problems were anything like mine, I'm
sure you're wondering how to get to be on desert island a little
less than two years from now :-/

--
Poul-Henning Kamp             FreeBSD coreteam member
phk@FreeBSD.ORG               "Real hackers run -current on their laptop."
