Date: Tue, 16 Jun 1998 20:22:33 +0800 From: Peter Wemm <peter@netplex.com.au> To: Hostas Red <kong@kong.spb.ru> Cc: current@FreeBSD.ORG Subject: Re: ssh 1.2.25: port broken on current Message-ID: <199806161222.UAA03749@spinner.netplex.com.au> In-Reply-To: Your message of "Tue, 16 Jun 1998 11:35:19 %2B0400." <Pine.BSF.3.96.980616112851.5743K-100000@kong.dorms.spbu.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
Hostas Red wrote:
> Hi!
>
> On Tue, 16 Jun 1998, Peter Wemm wrote:
>
> > > sshconnect.o: Undefined symbol `___inet_addr' referenced from text segmen
t
> > > sshconnect.o: Undefined symbol `___inet_ntoa' referenced from text segmen
t
> > > sshconnect.o: Undefined symbol `___inet_ntoa' referenced from text segmen
t
>
> [...]
>
> > This is an include files / libc version mismatch. It looks like you've
> > updated the system includes, but have not updated libc yet. Check the
> > timestamps on /usr/include/arpa/inet.h and /usr/lib/libc.so.3.1 - they
> > should be around the same.
>
> Yes, it is. But - /usr/lib/aout/libc.so.3.1 (jun,16) almost the same date
> as /usr/include/arpa/inet.h (jun 12), but /usr/lib/libc.so.3.1 is
> (may,26). But - /usr/lib/aout in LIBPATH, not /usr/lib. Whats the problem?
At a guess, your ldconfig path in /etc/rc is putting /usr/lib first so
you're getting the "old" library...
If you do a grep inet_ntoa /usr/include/arpa/inet.h, you'll get:
peter@overcee[8:15am]~src-133> grep inet_ntoa /usr/include/arpa/inet.h
#define inet_ntoa __inet_ntoa
[..]
And if you then look at the libraries:
peter@overcee[8:16pm]~src-135> nm /usr/lib/libc.so.3.1 | grep inet_ntoa
0003e1ac T _inet_ntoa
peter@overcee[8:16pm]~src-136> nm /usr/lib/aout/libc.so.3.1 | grep inet_ntoa
U ___inet_ntoa
00044064 T ___inet_ntoa
00000000 I _inet_ntoa
The 'I' flag is "indirect", kinda like a backwards compatability alias so
that you can run old programs that are linked against "_inet_addr". New
programs will be compiled to use "___inet_addr" because of the #define in
arpa/inet.h.
Why do this? It's part of the bind 8 resolver interface.. Paul Vixie has
done this in the name of posix linker namespace protection. If you
happened a program that was unrelated to inet/socket/nameserver/etc, and
it had it's own "inet_ntop()" function for some reason, under the old
system you'd get a conflict, but under the new system you should be OK.
I'm not sure that it's worth the hassle, but I asked around and the
consensis was to go with the flow.
> Adios,
> /KONG
Cheers,
-Peter
--
Peter Wemm <peter@netplex.com.au> Netplex Consulting
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199806161222.UAA03749>