svn commit: r187782 - in head: etc/rc.d share/man/man5

[Rong-en Fan]

On Wed, Jan 28, 2009 at 3:25 PM, Giorgos Keramidas <keramida at freebsd.org> wrote:
> On Wed, 28 Jan 2009 12:50:22 +0800, Rong-en Fan <grafan at gmail.com> wrote:
>> On Wed, Jan 28, 2009 at 4:13 AM, Giorgos Keramidas <keramida at freebsd.org> wrote:
>>> Modified: head/etc/rc.d/ntpd
>>> ==============================================================================
>>> --- head/etc/rc.d/ntpd  Tue Jan 27 19:56:38 2009        (r187781)
>>> +++ head/etc/rc.d/ntpd  Tue Jan 27 20:13:24 2009        (r187782)
>>> @@ -23,7 +23,7 @@ ntpd_precmd()
>>>        rc_flags="-c ${ntpd_config} ${ntpd_flags}"
>>>
>>>        if checkyesno ntpd_sync_on_start; then
>>> -               rc_flags="-g $rc_flags"
>>> +               rc_flags="-q -g $rc_flags"
>>>        fi
>>>
>>>        if [ -z "$ntpd_chrootdir" ]; then
>>>
>>
>> According to ntp(8), -q makes ntpd exits just after the first time the
>> clock is set.
>
> Yes, you are right :)
>
>> Doesn't this mean if ntpd_sync_on_start is on and the time difference
>> is large, then ntpd will exit after the first time sync? Is this the
>> behavior we really want?
>>
>> Shouldn't we keep ntp running after the clock is adjusted?
>
> This is correct too.  The effect of `ntpd_sync_on_start' is supposed to
> be the same as if we run `ntpdate' before the real ntpd starts, so this
> option only applies to the first sync-once instance of ntpd.  The real
> ntpd starts later, and finds the clock pre-synced.

Hmm... I think I'm confused. According to rc.d/ntpd, if ntpd_sync_on_start
is set to yes, it adds '-q -g' to rc_flags. By doing so, ntpd start makes
ntpd exists immediately after the first sync. Then, who is responsible
to start the "real ntpd" you said above?

Thanks,
Rong-En Fan