> Do you have any idea what happened to the version of ATPCS that mandated
> this?  The ARM Web site still seems to be giving me A-05, which doesn't.

No.  I'll try to remember to find out tomorrow.

> 
> >  I think we
> > sort of have VFP endian floats (Ben sort of hacked that up), but I haven't
> > tried it against the full implementation I'm working on.
> 
> I'd be rather surprised if the compiler were getting anything wrong, since
> that was just a case of flipping a single #define.  There may be bits of
> libraries that assume FPA endianness, though.

I'd be surprised if there were any problems too.  But you've hard-coded 
the compiler that way, which is why I said "hacked it up".

R.