Done items, moved from TODO list.


DATE	FILE(S)		DESCRIPTION
===============================================================================
1may97	where.c		Finished generic instantiation of unions.  Turned out
			to be trivial, given that unions are identical to
			tuples except for type tag name.  Hence, generic
			instantiation of unions just calls the existing
			InstantiateRecordType.

1may97	typechk.c	Fixed problem that caused inherited union not to type
			check properly, in the following circumstance:

			    obj Y is a:A or b:B or c:C;
			    obj Y1 < Y

			    var y:Y, y1:Y1;

			    y.a		(* OK *)
			    y1.a	(* Erroneously reported error that a
					   was not a tag of y1 *)

			The problem was just a flat-out logic bug that had
			never been tested because a subclassed union had never
			been in the test data.  Here are the specifics:

			The following logic in chkRecordOrUnionRef

			    if (onetuple) {
				lerror(t, "%s %s", fname, errmsg2);
				return null;
			    }

			was replaced by this logic

			    if (onetuple) and
				    ((not desig) or			/*NEW*/
					(not isTupleOrUnion(desig)))) {	/*NEW*/
				lerror(t, "%s %s", fname, errmsg2);
				return null;
			    }

			It appears to work now, but the logic is pretty
			convoluted.  It is doubtful that anything will ever be
			done about it.

16jun95	parser.y,	Allow an attr value to be an expr in addition to a
	OTHERS		comment and ident list.  Though ident is a subcase of
			expr, it should still be distinguished semantically in
			that it defines a relational attribute.

27oct94	parser.y	Eliminate the much unused feature that op sigs can
			contain (unnested) uses of 'and' and '*'.
			NOT!  We're leaning again towards sturctural
			equivalence, so we wont do this.
			NOT NOT!!  Screw struct equiv, we leave op sigs as is
			since unnested '*' means varargs, and 'and' can just be
			the same as ',', which with auto-implosion of actuals
			is actually truish.
			NOT NOT NOT!!  We should unify comp exprs and op sigs
			completely.

29nov94	typechk.c	Allow a single, bundled list value as a match for a
			single-list-arity op.