Oh, thanks heaps. I was trying to eliminate one of the variables
but I chose the wrong one :).

I think the interrupt-code test should actually  be

	if (ISSET(delta, sc->sc_ppsmask)) && (msr & sc_ppmask) == ppsmatch)

the first clause finds if the signal we're interested in changed, the
second checks it's the right edge.

Jeremy Cooper also pointed out a potential race with setting
`sc_dopps' flag before setting the edgemask.  And I guess the DCD_
names should really be PPS_NONE and PPS_DCD_XXX_EDGE?

I've fixed that, and I'm folding in Chris Torek's suggestion
(thanks again).

I'll put the fixed patch in 
     ftp://ftp.netbsd.org/ftp/pub/incoming/jonathan/pps-com-patch