two comments:
 1) there is a way to kick the PCIC and then make sure it can kick you
back to verify that a given interrupt line works (though you have to
unmask interrupts to let this happen at autoconf time in the SCSI
controller case that was discussed).  This won't be perfect but it
will probably work automatically most of the time..

 2) it *may* be possible to share an interrupt between the PCIC itself
and one or more of the devices attached to it, which reduces the
number of possible interrupts you need...

					- Bill