[Coco] FYI - 6809 GCC

[Alan Jones]

James,
I am assuming that the GCC for the 6809 is being compiled on a Linux 
machine?

Alan

James Dessart wrote:
> I've been experimenting with Brian Dominy's changes to the 6809 port 
> of gcc. All-in-all things seem to be good. 16-bit ints can be enabled 
> with the option -mint16, and the use of registers for passing 
> arguments can be avoided with -mnoreg-args, which allows the use of 
> the libc I had modified for RSDOS.
>
> It's a much smoother experience than I had things set up. His site 
> <http://www.oddchange.com/gcc6809/> contains all the info you need to 
> get set up. Apart from a Subversion client and a gcc tarball, and the 
> crt libraries available on my gcc page 
> <http://skwirl.ca/coco/gcc.html>, everything is available through 
> Brian's Subversion repository. I suggest the 2006-05-13 tag, since 
> it's stable, and gcc 3.4.6. This is what I'm working from.
>
> His scripts will extract gcc for you, patch the sources, and then set 
> the build in motion. It'll even install it for you.
>
> One caveat, my libc doesn't seem to work well when it comes to using 
> printf, et al. Varargs seems to break, and I'm not sure why. 
> Unfortunately I have no idea how to run gdb using MESS, so I can't 
> really debug the problem. But there's no reason why someone else 
> couldn't just write a replacement that doesn't assume the use of the 
> stack to pass arguments, and that is all-around better for RSDOS in 
> general. Might be possible to write it in C, without resorting to 
> assembler.
>
> In fact, with some support in the linker, OS-9 software could be 
> written easily. Or a modification of the assembler to generate 
> rlink-compatible object files. I'm not conversant in OS-9, so I might 
> just be leading you guys astray.
>
> -- 
> James Dessart
> <http://skwirl.net/>
>
>
>
> --Coco mailing list
> Coco at maltedmedia.com
> http://five.pairlist.net/mailman/listinfo/coco
>