🚀 go-pugleaf

I have a query about the way in which RISCOS accesses the IOC KART.

I need to communicate with a keyboard-like device via the KART, but without
the OS intercepting my data or inserting it's own data.  I tried this two
ways:

1.  Illegal way.  Mask TxE and RxF interrupts in the IOC IMR-B, then drive
the KART in a polling fashion using ISR-B to get the status of the device.

2.  Legal way.  OS_ClaimDeviceVector for devices 14 & 15.

The first way worked, but the KART intercepted my data and attempted to
interpret it.  The data happenned to correspond with keyboard mouse position
codes, so the pointer went berserk.

The second way worked 20% of the time.  A further 20% of the time the machine
locked solid.  The remainder of the time the data rx'd from the KART was
incorrect and a strange thing happened - the pointer axis reversed.  The
mouse step was NOT changed.

My conclusion is that the KART handler in RISCOS polls the KART (perhaps on
a timer IRQ) rather than relying purely on interrupts.  Both the Arthur &
the RO2 PRMs imply that this is not the case, however.  So..  anyone know
for sure?

Cheers

Bob
-- 
  Robert J. Voisey  :  r_voisey@csd.brispoly.ac.uk
         -- Probug Software Developments --
162 Staplehill Rd, Fishponds, Bristol, Avon, England