REAL type lossage

[William_Lott@ARDATH.SLISP.CS.CMU.EDU]

Rob,

I've figured out where the REAL type lossage is, but I'm not sure what to
do about fixing it.  This is a piece of the backtrace:

    (KERNEL::ROUND-NUMERIC-BOUND -6.283185307179586d0 NIL NIL T)
    (KERNEL::NUMBER-SIMPLE-INTERSECTION-TYPE-METHOD
	#<NUMERIC-TYPE (REAL -6.283185307179586d0 6.283185307179586d0)>
	#<NUMERIC-TYPE REAL>)

It's obvious that what it is trying to do is valid, but round-numeric-bound
does an:

	(assert (or (rationalp cx) (eq class 'float)))

In this case, class=NIL and CX is a float.  I think the thing to do is to
just remove this assertion.  Then it would return the bound as is.  That
would work for this case, but I'm not sure if it's a valid general fix.

-William