This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Skunk GCC and GAS in Sco 5.0.2


> I am successfully compiling the first part of GCC, but when the 
> makefile switches to compiling with XGCC, it is failing with 
> 
> as: unrecognized option `-b'
>
> I suspect that egcs is assuming that I am using the native assembler 
> (which isn't on the system at all).  Instead I am using GAS.

As the one that created the GDS package on all of the pre-97 
Skunkwares, I do understand what's going on.

The native compiler and assembler use "-b elf|coff" to swap between
the two object file formats.   I was able, via sleazy abuse of symlinks
and the "-b" option of the GCC front-end, to emulate this behavior by
providing two completely different back-ends (cc1, cc1plus, etc.) and
two completely different assemblers.

In development GCC and EGCS, the ELF/COFF stuff is handled quite 
differently, switching between the two formats at runtime.

Currently, the OpenServer part of GCC/EGCS requires the SCO assembler.
This is not provided with 5.0.0 or 5.0.2, but allegedly is provided in
5.0.4.    I haven't looked to confirm or deny that.

I've posted to this list a shell script and the funky build procedure
required to wrap the GNU assembler and other binutils to make them work.
Look in the web version of the archives for this group for that posting.
I sent it and the subject should have been something obvious enough to 
make it easy to spot.

I have a sinking feeling that the day either GCC 2.8.0 or EGCS ships, it 
will become more apparant that this is a bad situation...

-- 
Robert Lipe       http://www.dgii.com/people/robertl       robertl@dgii.com


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]