This file contains notes on problems in the ANSI CL spec found
during the construction of the tests.

1. When building a composite stream, what happens when the component
  streams have different element types?

2. Should there be an UPGRADED-STREAM-ELEMENT-TYPE function.

3. The spec requires that arrays specialized to type NIL exist.
  Was this intended?

4. If NIL specialized arrays exist, then NIL vectors are also strings.
  Was this intended?

5. The spec requires that (UPGRADED-COMPLEX-PART-TYPE NIL) be (type
  equivalent to) NIL.

6. The definition of UPGRADED-COMPLEX-PART-TYPE appears to require that
  it work on arbitrary typespecs, including SATISFIES, which is not
  possible.

7. Was it intended that values of 'smaller' float types be coercible
  to values of larger float types?  In CLISP, short-float has a larger
  range of exponents than single-float, so some shorts cannot be coerced
  to singles without over/underflow.

8. IMAGPART is defined as returning (* 0 number) on reals.  If the
  implementation supports negative zero and number is a negative float, this
  will be -0.0 (of the appropriate type).  Was this intended?