SUMMARY & more Q, Re: [HPADM] ORD_16/32 on HP-UX 10.20 for 8.2.3 bind

[Heather Ebey]

Thanks to Ralf Hildebrandt, Steve Holdoway, Joseph Yao,
Daniel Wittenberg and others, I have a little better understanding.

To summarize what I did:

Since yesterday, I have been running named 8.2.3 without problems
and confired that the clients, nslookup, host..., work too.

Q: A colleague downloaded the sources the moment he knew
   about the security problem. Our version reads:
      named 8.2.3-REL Tue Jan 30 09:06:05 PST 2001

   Is this the correct output, or was another posted after
   he downloaded?

-------
As soon as I uncommented #define ORD_32, it compiled without
a hitch. A message today said that I should be using ORD_16.
I still have no clue as to why this was necessary for me.
Another responder said that he too had to uncomment the
#define ORD_32 in port/hpux10/include/port_before.h.

Q: Was this the right thing to do???

Q: Did the people who did not have to uncomment it check
   this file and confirm that it is commented out?

Q: Did you use gcc or HP ansi/c?

Q: Do others verify that I should use ORD_16?

   [To answer a few people's questions.
     Yes, I have an ANSI C compiler--though I have gcc 2.95.1,
     I used HP's ansi/c:
        LINT G.10.32.05 CXREF  G.10.32.05
        HP92453-01 G.10.32.05 HP C Compiler
         /usr/lib/libc: $Revision: 76.3 $

      The compile system has patch bundle 51--the latest
      official release from HP.

      I compiled this on a PA-RISC 2.0 system, but used
      +DA1.1 so that it will work on the older HPs too. ]

Q: Do the people who had no problems have newer OS patches.

Q: What compile options did you use?
   Aside from changing DEST* paths in bind-8.2.3/port/hpux10/Makefile.set,
   I didn't do anything else and just compiled with "make", with no
   other compile options.

Q: Finally, is it true that 9.1 works fine without threads on 10.20?
   Ralf and Joseph, have you been running it since 1/30 and
   have you found any gotchas? Clients work fine?

--heather

P.S. Steve, Thank you for the offer of the compiled binaries.

The original message below:

> > The error:
> >         if test ! -d threaded ; then mkdir threaded ; else true ; fi
> >         cc -Ae -z -I../../port/hpux10/include -I../../include -g -Wl,+vnocompatwarnings  -D_REENTRANT -I../cylink -I../dnssafe -DCYLINK_DSS -DHMAC_MD5 -DUSE_MD5 -DDNSSAFE -c cylink_link.c -o threaded/cylink_link.o
> > cpp: "ctk_endian.h", line 135: error 4062: Not defined basic word type ORD_32 or ORD_16.
> >
> >   This is the specific line in ctk_endian.h
> >    #if  !defined( ORD_32 ) && !defined( ORD_16 )
> >    #error Not defined basic word type ORD_32 or ORD_16.
> >    #endif
> >
> > Prior to this in that header file, there is an attempt to
> > set either ORD_16 or ORD_32.
> >
> > I tracked down where ORD_32 or ORD_16 should be defined.
> > 8.2.2p7 port/hpux10/include/port_before.h:
> > #define ORD_32
> >
> > 8.2.3 port/hpux10/include/port_before.h:
> > /* The ORD_32 define breaks DNSSEC on HP-UX 10.20.  */
> > /* #define ORD_32 */
> >
> > QUESTION:
> > Can I define ORD_16 in 8.2.3 for HP-UX? Or, can I just
> > uncomment ORD_32? One or the other seems to need defining.
> > Or, do I just eliminate CYLINK_DSS?  I fear I have no idea
> > whether cylink is needed. Or, do I add something to
> > ctk_endian.h so that it knows how to determine if it
> > should set ORD_16 or ORD_32?
> >
> > _________________________________________________________
> > _ hebey at ucsd.edu
> > _ UC San Diego, Academic Computing Services, Systems Group
> > _ http://www-acs.ucsd.edu/home-pages/heather/