Update: Can't compile BIND 8.3.1 under Solaris 8

[Kevin Darcy]

John,
         You need to troubleshoot your gcc environment. Apparently it's not so much
that you're using the wrong assembler, as that gcc (the compiler part) is generating
the wrong kind of assembler code in this particular instance -- code that is
apparently incomprehensible to the sun "as". It's as if your gcc thinks it's
invoking "gas" (the GNU assembler) instead of the native assembler (????). I just
ran a short test and noticed that gcc is invoking "/usr/ccs/bin/as" in my build
environment, but I get no errors when I build BIND 8.3.1, so there must be something
different between our gcc environments. I'm not an expert in gcc internals,
unfortunately, so I'm not sure even where one would start troubleshooting such a
problem...


- Kevin

John Heiden wrote:

> Hi Kevin and the list,
>
> How can I use the "right" assembler with this then?  I'm at a loss of how I
> would do this.  I have never ever (not even once) used assembler on a SUN
> before.  I have done VAX assembler, Intel and Motorola assembler.
> I have written a lot of C code under Solaris.  I am at a total loss as to what
> I need to do to use "the right" assembler here.  Any ideas?  In fact, when
> I last installed BIND, I didn't have this problem.  Actually, I had installed
> it on a Tru64 server, and it was BIND 8.2.2 and 8.2.3, but still same ideas.
>
> I hate to sound totally pathetic and lost here.  (Not that it's much of a stretch
> I must say.)
>
> Input sought...  thank you in advance for any kind help or ideas you might
> have.
>
> John
>
> Kevin Darcy wrote:
>
> > Why are folks here so anti-Sun? I've compiled practically every version of
> > BIND in existence on a Sun (SPARC) using both the SparcWorks compiler and gcc,
> > and rarely had any problems. The original poster's problem, as pointed out by
> > another poster, is that he is mixing the Sun "as" with gcc, which doesn't work
> > any better than, say, mixing gcc with Microsoft's dev tools on an Intel-based
> > box. Which is to say, mixing and matching pieces of different development
> > environments is a bad idea, regardless of platform. Stop blaming Sun for this.
> >
> > - Kevin
> >
> > phn at icke-reklam.ipsec.nu wrote:
> >
> > > John Heiden <jheiden2 at pop3.utoledo.edu> wrote:
> > >
> > > > OK, to a degree I am an idiot.  But still, no dice.
> > >
> > > > First, having just set up the server, I had not setup any paths.
> > > > Second, I was trying to install OpenSSH and that would not
> > > > compile either.  Evidently in one of my e-mails I confused the
> > > > two, and I'm not sure which one.
> > >
> > > > I apologize for any inconvenience and mistakes on my part.
> > > > By the time I left work last night, I couldn't see straight any
> > > > more.
> > >
> > > > But here's what I get when trying to compile BIND 8.3.1 under
> > > > Solaris 8 now...
> > >
> > > I don't think gcc is the best way to build bind on solaris.
> > >
> > > Either switch to *BSD ( where gcc _is_ supported ) or purchase
> > > a sun compiler.
> > >
> > > ( yes, i know it's very strange from sun to refuse their customers
> > > a compiler, but market creatures seldom does The Right Thing )
> > >
> > > > When I do a make clean, fine.
> > > > Make depend, fine.
> > >
> > > > # make all
> > > > Using .systype
> > > > Using .settings
> > > > /home/jheiden/bind/src/include
> > > > /home/jheiden/bind/src/include/arpa
> > > > /home/jheiden/bind/src/include/isc
> > > > /home/jheiden/bind/src/port
> > > > /home/jheiden/bind/src/port/solaris
> > > > /home/jheiden/bind/src/port/solaris/include
> > > > /home/jheiden/bind/src/port/solaris/include/sys
> > > > choose_bitypes
> > > > sys/bitypes.h is up-to-date.
> > > > make_os_version
> > > > probe_ipv6
> > > > gcc  -g -O2 -I../../port/solaris/include -I../../include -c noop.c
> > > > : ld -x -r noop.o && : mv a.out noop.o
> > > > /usr/ccs/bin/ar cru ../../lib/libbind.a  noop.o
> > > > /usr/ccs/bin/ranlib ../../lib/libbind.a
> > > > /home/jheiden/bind/src/lib
> > > > /home/jheiden/bind/src/lib/resolv
> > > > mkdir threaded 2> /dev/null || test -d threaded -a -w threaded
> > > > gcc -I../../port/solaris/include -I../../include -g -O2  -D_REENTRANT -c
> > > > herror.c \
> > > >  -o threaded/herror.o
> > > > : ld -x -r threaded/herror.o \
> > > >  -o threaded/herror.out && \
> > > >  : mv threaded/herror.out threaded/herror.o
> > > > gcc -I../../port/solaris/include -I../../include -g -O2  -c herror.c
> > > > : ld -x -r herror.o -o herror.out && \
> > > >  : mv herror.out herror.o
> > > > mkdir threaded 2> /dev/null || test -d threaded -a -w threaded
> > > > gcc -I../../port/solaris/include -I../../include -g -O2  -D_REENTRANT -c
> > > > res_debug.c \
> > > >  -o threaded/res_debug.o
> > > > /usr/ccs/bin/as: "/var/tmp/ccKNi7he.s", line 4131: error: unknown opcode
> > > > ".subsection"
> > > > /usr/ccs/bin/as: "/var/tmp/ccKNi7he.s", line 4131: error: statement
> > > > syntax
> > > > /usr/ccs/bin/as: "/var/tmp/ccKNi7he.s", line 4185: error: unknown opcode
> > > > ".previous"
> > > > /usr/ccs/bin/as: "/var/tmp/ccKNi7he.s", line 4185: error: statement
> > > > syntax
> > > > /usr/ccs/bin/as: "/var/tmp/ccKNi7he.s", line 4186: error: unknown opcode
> > > > ".subsection"
> > > > /usr/ccs/bin/as: "/var/tmp/ccKNi7he.s", line 4186: error: statement
> > > > syntax
> > > > /usr/ccs/bin/as: "/var/tmp/ccKNi7he.s", line 4240: error: unknown opcode
> > > > ".previous"
> > > > /usr/ccs/bin/as: "/var/tmp/ccKNi7he.s", line 4240: error: statement
> > > > syntax
> > > > *** Error code 1
> > > > make: Fatal error: Command failed for target `res_debug.o'
> > > > Current working directory /home/jheiden/bind/src/lib/resolv
> > > > *** Error code 1
> > > > make: Fatal error: Command failed for target `all'
> > > > Current working directory /home/jheiden/bind/src/lib
> > > > *** Error code 1
> > > > make: Fatal error: Command failed for target `all'
> > > > #
> > >
> > > > Any ideas?
> > >
> > > > By the way, anyone trying to compile OpenSSH 3.1p1 under Solaris 8, I
> > > > now
> > > > know the "trick" to making it compile.  =-)  And it's not listed
> > > > anywhere I could find.
> > >
> > > > Your help is *GREATLY* appreciated.
> > >
> > > > Thanks!
> > >
> > > > John Heiden
> > > > Network Engineer
> > > > The University of Toledo
> > > > Toledo,  OH  43606
> > > > jheiden2 at pop3.utoledo.edu
> > >
> > >