The implementation of Rectangle.intersection and 
javax.swing.SwingUtilities.computeIntersection was changed in JDK1.2.2C to return
a (0,0,0,0) rectangle instead of a bogus result if the intersection is empty - 
a bugfix for 4118972 (which was surprisingly closed as not a bug).
However, the spec was not updated to mention this.
This change caused a JCK1.2a regression:
  api/java_awt/Rectangle/descriptions.html#intersection16 
The spec should be updated to clarify the handling of non-intersecting rectangles.
The JCK test cannot be updated until the complete specification is provided.

======================================================================

N/A

I changed the SwingUtilities doc for computeIntersection
to fix my half of this bug.  Now we just need to fix the
Rectangle intersection doc.  

=================

The Rectangle docs will not be changed. See the discussion below:

Date: Wed, 20 Oct 1999 15:16:31 -0700 (PDT)



Subject: Re: Bug 4205002

Hi Scott,

Unfortunately, I don't like the idea of that method returning a
zero rectangle for empty.  For one thing, it is an inaccurate
answer for the problem.  That rectangle encloses no area, but it
exists.  The answer to some intersections is "there *is* no
rectangle that can represent the answer".  This is like the 
difference between returning 0 or NaN from a floating point
operation.

Do not spec this method to return a zero rectangle.

                                ...jim

  xxxxx@xxxxx   1999-10-20