> I need to wade through the source again to 100% verify this for all cases
> (Unless Neil know the answer already) but :

[ Lengthy explanation deleted for brevity ]

Thanks for the explanation. This is indeed quite odd and, to the best of my
knowledge, does contradict (or at least makes it hard to handle) the IEEE
standard, as I'm quite sure that the user must be allowed to change the
rounding/exception handling methods.

Since I'd prefer to let the FPE be handled like a black box, it looks like
the best option is the access to the FPCR through mcr/mrc. And this from
within the kernel :-(. Not really what I'd have expected from such code.
--
ws@TooLs.DE     (Wolfgang Solfrank, TooLs GmbH) 	+49-228-985800