[Ada] fix in Ada 2005 interpretation

[Arnaud Charlet <charlet at adacore dot com>]

Tested on i686-linux, committed on trunk

In Ada 2005, a user-defined abstract operation is not a candidate
interpretation for an overloaded subprogram name. This complicates the
analysis of expressions whose operands are literals, when one of the
arithmetic operators has been declared abstract on some numeric type,
because the universal interpretation of the expression may still be
available. The expression may appear ambiguous during the second pass
of type resolution because of the spurious presence of the universal
interpretation, but this interpretation cannot be removed during the
first pass, because the context type is not yet known, and may itself
be universal. As a result, when disambiguating an operation that has
a universal interpretation and whose context is not universal, we must
examine whether it, or some subexpression of it, has an abstract
interpretation, in which case the ambiguity is resolved in favor of the
non-universal one.

gnat.dg/test_rational_arithmetic.adb must compile quietly.

2007-04-06  Ed Schonberg  <schonberg@adacore.com>
	    Javier Miranda  <miranda@adacore.com>

	* sem_type.ads, sem_type.adb (Has_Abstract_Interpretation): Make
	predicate recursive, to handle complex expressions on literals whose
	spurious ambiguity comes from the abstract interpretation of some
	subexpression.
	(Interface_Present_In_Ancestor): Add support to concurrent record
	types.
	(Add_One_Interp,Disambiguate): Split Is_Abstract flag into
	Is_Abstract_Subprogram and Is_Abstract_Type.

Attachment: difs
Description: Text document