🚀 go-pugleaf

I have applied the coff patches to gas, and have with some limited 
success gotten it to generate running code for a i386 SysV 3.2.2 system. 
 
I am using gas as part of attempt to generate COFF formated object code
for kcl lisp.  It seems in their infinite wisdom, AT&T no longer supports
COFF, instead they have a new object format call ELF. The kcl systems
needs access to COFF files to load up objects.  Since kcl can be compiled
with gcc I thought that a way around the coff problem was to build gcc and
coff-gas and use the pair of these to build and run gcc.  However a number
of obstacles stand in the way of this goal:

     1. The AT&T preprocessor has some new features that aren't handled 
        by gcc.  Most annoying is the #machine preprocessor directive,
        which the gcc preprocessor does not like at all.  This is a minor
        problem easily over come by preprocessing with the AT&T preprocessor.

     2. The AT&T compiler no longer prepends an _ for function names. gcc
        does.  All the runtime library and the start up code function names
        don't have the underscores. When you go to link you get errors
        like, "Symbol main undefined". I have gotten around this with the 
        small programs I written by generating assembler code, then editing
        the function names by hand.  Does anyone know if there is an easy 
        way to stop gcc from generating underscores for function names?

     3. AT&T does not support the feature that most other unix cc compilers
        have of letting you see exactly what the calls to the various 
        parts of the compiler passes, assembler, and linker are. In my case
        this hurts because while I don't care how the compiler passes
        and assembler is getting called, I would like to know exactly what
        objects and libraries are getting link, sinces the names of these
        have changed.  

      
I had alot of trouble getting the AT&T as to except assembler generated by
gcc, however, since I was more interested in passing the stuff through
coff-gas, this was no big deal.  Like the original poster I would be
interested in hearing experinces of other i386 sysV 3.2.2 users.


Paul Puglia
Civil Engineering
Columbia University

puglia@cucevx.civil.columbia.edu