Chris Ross wrote:
>   I'm building the current ntp (et al) in usr.sbin/ntp.  I built
> it last week, and when I noticed [while trying to run it] today
> that it failed, I rebuilt from today's sources (for the binaries,
> I guess I should rebuild the kernel next, but I don't remember
> seeing any changes that I think would effect this)
> 
>   ntpd will SEGV shortly after starting.  Both the older and current
> binaries fail in the same way.  I have "ca.pool.ntp.org" and
> two "us.pool.ntp.org" "server" lines in my ntp.conf, which is
> the default after changing commented lines in and out.

   I've done a little more "printf-style debugging" of this problem.
The problem appears to be in ntpd/ntp_config.c.  In the CONFIG_SERVER
case of the main loop in the getconfig() function, it calls memset()
on the address of a local (stack) structure.  It's declared:

         struct sockaddr_storage peeraddr;

   And memset is called:

         istart = 1;
         memset((char *)&peeraddr, 0, sizeof(peeraddr));

   Immediately before those two lines (listed right above, here),
I put:

         if (debug)
              printf("Initializing, NULL'ing 0x%p\n", &peeraddr);

   And, when it runs, right before it SEGV's, I see:

Initializing, NULL'ing 0x0xffffffffffffc8d0
zsh: segmentation fault (core dumped)  ./ntpd -d

   So, is this a compiler/toolchain bug?  Looks like it to me.
Should I submit a pr on it?  I guess I'd have to make a
better test case.  But, this sort of thing can't *always*
be broken like this, or many more programs would be busted,
no?

   Thanks.  Any input welcomed...

                             - Chris