2019-12-16  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

	* scng.adb (Scan): In Ada2020, a left-bracket indicates the
	start of an homogenous aggregate.
	* par-ch4.adb (P_Reduction_Attribute_Reference): New procedure.
	(P_Aggregate): Recognize Ada2020 bracket-delimited aggregates.
	(P_Primary): Ditto.
	* par-util.adb (Comma_Present): Return false on a right bracket
	in Ada2020, indicating the end of an aggregate.
	* snames.ads-tmpl: Introduce Name_Reduce and Attribute Reduce.
	* sinfo.ads, sinfo.adb (Is_Homogeneous_Aggregate): New flag on
	aggregates, to designate an Ada2020 array or container aggregate
	that is bracket-delimited in the source.
	* sem_attr.adb (Analyze_Attribute): For attribute Reduce, verify
	that two arguments are present, and verify that the prefix is a
	stream or an object that is iterable (array or contrainer).
	(Resolve_Attribute): For attribute Reduce, resolve initial value
	with the type of the context. Type-checking of element type of
	prefix is performed after expansion.
	* exp_attr.adb (Expand_N_Attribute_Reference): For attribute
	Reduce, expand into a loop: a) If prefix is an aggregate with a
	single iterated component association, use its iterator
	specification to construct a loop, b) If prefix is a name, build
	a loop using an element iterator loop.
	* scans.ads: Add brackets tokens.

From-SVN: r279431