[kaffe] ARM detection fix

[Dalibor Topic]

Hi David,

David Marston wrote:
> In message <40697CF2.80907@kaffe.org>
>           Dalibor Topic <robilad@kaffe.org> wrote:
> 
> 
>>Peter Naulls wrote:
>>
>>>Believe it or not, there are other ARM processors in the world besides
>>>StrongARMs, so we shouldn't be looking at fixed architecture revisions
>>>(in most cases in any Unix code)
>>
>>Thanks for the patch, I've checked it in. You may want to give 
>>--enable-pxa a try to adapt SP_OFFSET to Xscale. :) I've also ripped out 
>>some things in the CVS in config/arm and replaced them with code from 
>>GNU libc (COMPARE_AND_EXCHANGE, mostly). I hope these work better for 
>>you, too.
>>
>>Is there going to be an arm-RISCOS port of 1.1.4? I've read the news on 
>>the gcc 3.3.3 port on drobe, and hoped that someone would pick 1.1.4 (or 
>>even 1.1.4+arm-CVS-changes) up and give it another go on RISC OS.
> 
> 
> Good choice of reading :) I have been looking at the RISC OS port again
> recently after a fairly long break being distracted with other things.

Yeah, drobe is very nice. It gets linked to regularly from osnews :)

> I did build kaffe from cvs about 2 weeks ago using the current RISC OS
> gcc cvs (3.3.3), but up until a few minutes ago I didn't think it was
> working. It does seem to be working now though :)

That's good news then, I guess :)

> I know what the difference is in the way I'm running it now to then, so
> I just need to work out why it makes a difference.

CVS is in quite heavy development atm. Right now, we're cleaning up all 
the ugly, messy warnings from kaffe's core, in order to stabilize things 
a little more. Feel free to join in :)

> So, there is some progress in that I have a version that seems to be
> working on the same things as the last released RISC OS port, but still
> failing on the same things too. Hopefully I'll have more time to look
> into the issues now.

I don't remember how far we were last time around. Does it run the 
regression tests?

> BTW, the change to COMPARE_AND_EXCHANGE doesn't actually work with the
> RISC OS assembler, but that's easy enough to sort out.

Sorry about that. I just copied the code over from GNU libc hoping that 
it would improve things for everyone on arm. A fix would be most welcome.

cheers,
dalibor topic