Delivery-Date: Mon, 19 Jun 1995 18:58:27 -0700
Date: Mon, 19 Jun 1995 18:53:53 -0700
From: jhanley@us.oracle.com (John Hanley)
Message-Id: <9506200153.AA12459@borogove.us.oracle.com>
To: bind-workers@vix.com
Phone: +1 415 506 2360
Subject: BIND on ptx V4.0.2 (svr4)
Content-Length: 3248

This is not so much a cry for help as a comment on what I've seen, in case
it is useful to anyone.  Otherwise, just ignore it, as Sequent will eventually
support 4.9.3 so the libraries will agree on the the size of _res.  I only
have ptx 2.x primary nameservers putting protocol violations out on the wire,
and ptx 4.x hasn't become widely enough deployed to be a support issue for me.

The length mismatch caused by adding new fields to _res produces the
following on ptx 4.x:

  ld: libinet.so: fatal error: attempt to override defined size of symbol `_res`
  from file ../res/libresolv.a(res_init.o) with size of tentative definition

To get around this, make sure that the shared library reference, "-linet",
precedes the resolver library.  For example, compile a vanilla 4.9.3b17
with "-DSVR4" and use a link command like:

  cc -g -o nslookup main.o getinfo.o debug.o send.o skip.o list.o subr.o commands.o -linet ../../res/libresolv.a ../../compat/lib/lib44bsd.a -ll -lsocket -lnsl -lseq

This links cleanly against /usr/lib/libinet.so and friends.

Unfortunately, we easily core dump:

% debug ./nslookup
New program nslookup (process p1) created
debug> run       
Default Server:  dcsun2.us.oracle.com
Address:  139.185.20.52

> dcsun1.us.oracle.com.
Server:  dcsun2.us.oracle.com
Address:  139.185.20.52

Name:    dcsun1.us.oracle.com
Address:  139.185.20.51

> dcsun1.us.oracle.com
Server:  dcsun2.us.oracle.com
Address:  139.185.20.52

SIGNALED 11 (segv) in p1 [_doprnt()]
        0xbffd6a78 (_doprnt+13368:)      movb   (%ecx),%al
debug>     
debug> stack
Stack Trace for p1, Program nslookup
[8] _doprnt(0x8055b65,0x80469cc,0x8046994)      [0xbffd6a78]
[7] sprintf(0x80469d2,0x8055b5c,0x100,0x8047460,0x100,0x61726f2e)       [0xbffe7510]
[6] GetHostDomain(nsAddrPtr=0x805c54c,queryClass=1,queryType=1,name="dcsun1.us.oracle.com",domain=Read at address 0x61726f2e failed
,hostPtr=0x805b4f8,isServer=0)  [getinfo.c@676]
[5] GetHostInfoByName(nsAddrPtr=0x805c54c,queryClass=1,queryType=1,name="dcsun1.us.oracle.com",hostPtr=0x805b4f8,isServer=0)    [getinfo.c@610]
[4] DoLookup(host="dcsun1.us.oracle.com",servPtr=0x805c48c,serverName="dcsun2.us.oracle.com")   [main.c@594]
[3] LookupHost(string="dcsun1.us.oracle.com",putToFile=0)       [main.c@688]
[2] yylex()     [commands.c@0x8050d7f]
[1] main(argc=0,argv=0x80475c0,0x80475c4)       [main.c@349]
[0] _start()    [0x8049315]
debug>  


The problem, as has been noted on this list before, is that dnsrch[]
and defdname[] were inadvertently swapped.  This is Bad Karma for
shared libs.  Applying the following yields a robust ``nslookup''.


--- include/resolv.h~	Wed Dec 14 22:24:07 1994
+++ include/resolv.h	Mon Jun 19 18:38:52 1995
@@ -111,8 +111,8 @@
 		nsaddr_list[MAXNS];	/* address of name server */
 #define	nsaddr	nsaddr_list[0]		/* for backward compatibility */
 	u_short	id;			/* current packet id */
-	char	*dnsrch[MAXDNSRCH+1];	/* components of domain to search */
 	char	defdname[MAXDNAME];	/* default domain */
+	char	*dnsrch[MAXDNSRCH+1];	/* components of domain to search */
 	u_long	pfcode;			/* RES_PRF_ flags - see below. */
 	unsigned ndots:4;		/* threshold for initial abs. query */
 	unsigned nsort:4;		/* number of elements in sort_list[] */


		Cheers,
		JH