> On Fri, 23 Mar 2001, Richard Earnshaw wrote:
> 
> > Unless I've mis-understood the consenus here, I think we have decided on:
> 
> FWIW, I think I agree with you on all of them.

Good; then at least we are moving towards closure on this.

> 
> > - ATPCS variant for shared libs is still not determined.
> 
> FWIW, shared libraries are something I don't currently understand, so
> someone else needs to deal with them.
> 
> > - Traditional model for register saving (r0-r3,ip,lr) clobbered over a 
> > call (if using fpa then f0-f3 are call clobbered; vfp still to be 
> > determined).
> 
> VFP doesn't matter till the compiler generates code for it.

Not quite true.  We need to leave enough space in a jmp_buf to hold all 
the call-saved VFP regs.

R.