> >The switch shouldn't be able to break the driver, even if it is
> >a cheap nasty switch.
> 
> Right.
> >
> >OTOH most drivers only see 'normal' events....
> >
> Which is why debugging them in the face of garbage is such a 
> challenge...

Indeed - the usual problem is getting the failure to happen on
the development system, and not just on a 'live' customer system.

I recall having great 'fun' stopping an Sbus lance driver
from locking up (and locking the system up) when confronted
with all the possible sbus timeouts.  Some (but not all) of
the failure modes are covered by errata for the chipsets
concerned.


	David

-- 
David Laight: david@l8s.co.uk