>I'd suggest having a KASSERT2(cond, complaint) and then KASSERT could be
>#define KASSERT(cond)  KASSERT2(cond, ("assertion failure: " #cond))

I've worked with both styles (netipsec and opencrypto started with the
FreeBSD two-argument version).  I do see why you suggest this ;
its exactly what i'd have suggested a year or so ago.

Now, I'd favour using KASSERT() for the two-argument version, and
mechanically changing all extant uses to KASSERT1().  If we do this,
the one-argument version should be considered obsolescent.