In message <970117144300.1b025@rmcs.cranfield.ac.uk>
          Neil <CARSON@rmcs.cranfield.ac.uk> wrote:

> 
> > Should setjmp save any floating point context for restoration by longjmp?
> > In the current implementation the only code to handle floating point 
> > registers in these routines is commented out.

That's because it came from the UnixLib implementation and it was used
before the ARM FPE was available and the floating point save/restore
tended not to work. If you need to get your context switching working,
then you can always write your own versions of setjmp/longjmp until the
fixed versions are available.
 
> Yep, it's been fixed but not tested yet.

I thought that it had been tested since it fixed a couple of problems in
X (or that's what I remember you or Mark telling me a week or so ago).
BTW, I presume it is using the SFM instruction and not needlessly saving
r1-r3 anymore so to get that little bit of extra speed ?

regards,
Pete