🚀 go-pugleaf

HELP!  I AM BEING HELP PRISONER BY THE HPUX OPERATING SYSTEM!

Here's the story...

I  wanted to install  gcc on this  HPUX v7  system  because: a) I
heard it  is  a  very good (and  efficient)  compiler, b) gcc  is
"fully"  ANSI compatable  (even more so   that  the ANSI standard
is...if you know  what   I   mean), and  c) I've  always  thought
educational institutions were perfect places to have GNU software
installed.

So there I  was, getting ready to  teach a senior-level compilers
class and wanting to use  gcc.  The textbook ("Compiler Design in
C", by  A. Holub) looked to be   an excelent text for  the class.
The  software for the  text,   basically a compiler compiler, was
originally designed for PC's (Microsoft C) and was written with a
mix  of the ANSI and oldK&R  flavors of   C.  Moreover, while the
code for the  software was made close  enough to compile and work
on our system, much of the support code  was still more ANSI than
not.  Anyway, the decision  was  made to  start over and  use GCC
instead.

The point is that I installed  GCC  on our  HPUX system, and that
went VERY smoothly.  I compiled all of the code for the software;
after a little patching  up (mostly fixing prototypes), it worked
quite well (better than the HPCC version).

But,   alas, we really   should have  a  debugger  along with the
compiler, and everyone  KNOWS  that   HPUX need  GAS to  make   a
debugger, GDB, available (it seems  that GCC and the HP debugger,
XDB, are ioncompatable).... So, ...

I  thought I  should  install GAS.   This has   proved to  be   a
nightmare!  I ftp'ed gas-1.37 from prep.ai.mit.edu and set out to
compile just as I had originally for GCC (version 1.37.1).  To my
surprise, there was   precious  little documentation  on   how to
install the package (unlike GCC, which was very well documented).
I  thought there would at least  be a texinfo file.  Undaunted, I
followed  the  instructions,  and    proceeded to get    to work,
compiling with the GCC I had.

One problem was with the write.c stuff, which I assume was due to
the fact that we have hp9000s300,  and the  write.c has some sort
of conditional  based on it being hp9000s200.   I guess the a.out
headers changed a little with the new system.  What it  meant was
that the  EXEC_MACHINE_TYPE  was no longer  define'd, so one line
was gone.  But I don't think that's the problem.

As an example,  here is my test case.

% cat foo.c
main(){
int i;
i += 1;
}

% gcc -S foo.c
% cat foo.s
text
	lalign 2
	global _main
_main:
	link.w %a6,&-4
	addq.l &1,-4(%fp)
L1:
	unlk %a6
	rts
% a68 foo.s
foo.s:1:"Unknown operator" -- Statement 'text' ignored
foo.s:2:"Unknown operator" -- Statement 'lalign 2' ignored
foo.s:3:"Unknown operator" -- Statement 'global _main' ignored
foo.s:5:"Unknown operator" -- Statement 'link.w %a6,&-4' ignored
foo.s:6:"Unknown operator" -- Statement 'addq.l &1,-4(%fp)' ignored
foo.s:8:"instruction/operands mismatch" -- Statement 'unlk %a6' ignored

Has anyone  else seen this  problem?  Have I made  some incorrect
assumptions  while compiling?   Can I  use GDB without  compiling
GAS?

Please direct replies to me via e-mail.  I am in a bit of a
pickle, so any help would be greatly appreciated.

		Thanks in advance,

			Andy.


--

Andy Edwards * Computer Science Dept. * Iowa State University * Ames, IA 50011
INTERNET: jae@iastate.edu               PHONE: (515) 294-4377
------------------------------------------------------------------------------