This is the mail archive of the crossgcc@cygnus.com mailing list for the crossgcc project.


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

Re: Cross compile to DOS.


On Sun, 6 Jul 1997, Michael Beach wrote:

> I don't know of any support in gcc for real mode 80x86 code generation.
> I would have said it was not possible, as gcc is pretty much 32 bits and
> up only, but now that you mention it, there are MDs for Hitachi H8 etc.
I happen to have an .md for the 6809, for an older GCC, that can generate
code with either 16 or 32bit ints. GCC is not limited to specific integer
sizes. Even the 68000 md can have either 16 or 32 bit ints.

> If you don't want to run BCC under dosemu, you could look at the free C
> compiler that the Linux ELK project uses for the 8086. I'm not sure if
> it is full ISO though.
<BIG GRIN>

I can see you haven't actually looked at the ELKS project; I just happen
to be the person dealing with the compiler/assembler etc part of the
project. 

Bruce's C compiler is a full K&R compiler with a reasonable code
generator. But it doesn't understand ansi and Borland's code generator
is a bit better. There is an 'unprotoize' command integerated with the
front end but that is at best a hack.

On Sun, 6 Jul 1997, Johan Borkhuis wrote:

> I do not know about cross compilers but you could try to use DJGPP, the dos
> port of gcc. I don't know if it works within a dosemu-box. It includes a
> memory manager, so you don't have the 'nice' dos memory model.
DJGPP is i386+ and yes it make Dross look more like a real OS; I prefer
the real thing; that's why I want a good unix compiler for the 8086.
I think djgpp works inside Linux-dosemu; but that's not the point; if
I want to make djgpp executables under linux I can cross-compile them.

On 6 Jul 1997, Dave Love wrote:

> GCC requires a flat address space (16 bit isn't an issue).  Ergo no
> 8086.  GCC code for MS-DOG on 386+ is fine, of course.
This is irrelevent; 64k+64k is a flat address space and is fine for my
application, it'd be good for ELKS too. 

BTW in theroy huge model is also a flat memory model, nasty and slow
but flat nevertheless.

On Fri, 4 Jul 1997, Ken Greenberg wrote:

> Probably the silence only means those of us who have been on the list 
> for a long time are tired of answering this question, but that isn't 
> your fault. And yes, it should be in the FAQ.
> 
> The answer is, in general, no. The only x86 architecture supported by 
                 ^^^^^^^^^^ "in general" and in 'special' ?

Anyway, I don't suppose anyone is going to make an .md for GCC-8086 soon,
I can continue to use a dos compiler until it happens or perhaps the
compiler at:
   http://ourworld.compuserve.com/homepages/davewalker/homepage.htm
gets the bugs ironed out.

-- 
Rob.                           (Robert de Bath <http://www.cix.co.uk/~mayday>)