> --- ntp-4.1.1/ntpd/ntp_proto.c.ORIG   Wed May 29 09:39:46 2002
> +++ ntp-4.1.1/ntpd/ntp_proto.c        Wed May 29 11:07:39 2002
> @@ -348,11 +348,10 @@
>       }
> 
>       /*
> -      * Discard broadcast packets received on the wildcard interface
> -      * or if not enabled as broadcast client.
> +      * Discard broadcast packets if not enabled as broadcast client.
>        */
>       if (PKT_MODE(pkt->li_vn_mode) == MODE_BROADCAST &&
> -         (rbufp->dstadr == any_interface || !sys_bclient))
> +         !sys_bclient)
>               return;
> 
>       /*
> 
> For all of our arriving broadcast packets the equality
> "rbufp->dstadr == any_interface" was true.  I do not see the necessity
for
> this
> test as there is no way in ntp-4.1.1 to set which interface listens to
> broadcast packets.
[...]
> I received this reply (condensed) from David L. Mills, the original
author
> of ntp.
> 
> > The extra test was put in as defense when multicast breaks and the
I/O
> > code is unable to determine which outbound interface will be used.
This
> > works in all architectures known to me, but breaks in Linux. I have
> > since removed the paranoid test so Linux does work with broadcast,
but
> > of course flames out with multicast. I don't know which incremental
> > release version has the removed test, which was made some time ago.
> 
> > Dave
> 
> The following was my reply to his note:
> 
> It is still present in 4.1.1a, the latest "stable" version listed on
the ntp
> web site.  Mandrake 8.2 picked up ntp-4.1.1 which, unfortunately,
broke
> things.
> 

well, at this time 4.1.1 has been "the latest stable". As I understand
the same problem is still present in 4.1.1a? 

It would be nice if anybody could figure out why this happens. As I
understand Dave's reply, it is still going to fail on multicast which
makes it just a workaround for a particular problem not a general fix.

-andrej

Reply via email to