CVS difference for ai05s/ai05-0153-3.txt

--- ai05s/ai05-0153-3.txt	2011/03/10 04:49:49	1.12
+++ ai05s/ai05-0153-3.txt	2011/03/17 06:13:16	1.13
@@ -1,4 +1,4 @@
-!standard  3.2.4(0)                                11-03-04    AI05-0153-3/06
+!standard  3.2.4(0)                                11-03-16    AI05-0153-3/07
 !class Amendment 09-05-27
 !status Amendment 2012 11-03-04
 !status ARG Approved  7-0-1  10-02-18
@@ -191,7 +191,7 @@
 subtype to which predicate clauses apply.
 
 The prefix of an attribute_reference whose attribute_designator is First, Last,
-or Range shall not denote a subtype to which predicate clauses apply.
+or Range shall not denote a scalar subtype to which predicate clauses apply.
 
 The discrete_subtype_definition of a loop_parameter_specification shall not
 denote a subtype to which Dynamic_Predicate clauses apply.
@@ -225,9 +225,10 @@
 12.3), this rule applies also in the private part of an instance of a generic
 unit.
 
-AARM Note: This is usual way around the contract model. The "In addition..."
-wording is included for consistency with similar rules, even though it's
-not really necessary, since Program_Error will be raised anyway.
+AARM Note: This is the usual way around the contract model; this applies even
+in generic bodies. The "In addition..." wording is included for consistency
+with similar rules, even though it's not really necessary, since Program_Error
+will be raised anyway.
 
 NOTE: A predicate clause does not cause a subtype to be considered
 "constrained".
@@ -2702,3 +2703,43 @@
 
 ****************************************************************
 
+From: Edmond Schonberg
+Sent: Wednesday, March 16, 2011  7:54 AM
+
+An enthusiastic user has detected a problem with a proposed  legality rule
+in the new section 3.2.4 :
+
+The prefix of an attribute_reference whose attribute_designator is First, Last,
+or Range shall not denote a subtype to which predicate clauses apply.
+
+
+We know the motivation for this rule (problem with enumerations with holes, etc.)
+but the rule only makes sense for scalar types. Otherwise
+it is hard to write useful predicates on array subtypes, for example:
+
+package Pred is
+   type A is array (Integer range <>) of Integer
+     with Predicate => (for all J in A'Range =>
+ (for all K in A'Range =>
+    (if J /= K then A (J) /= A (K))));
+end Pred;
+
+The rule should say  "scalar subtype".
+
+****************************************************************
+
+From: Bob Duff
+Sent: Wednesday, March 16, 2011  8:08 AM
+
+Please make the following editorial change to AI05-0153-3, Subtype
+predicates:
+
+The prefix of an attribute_reference whose attribute_designator is First, Last,
+or Range shall not denote a {scalar} subtype to which predicate clauses apply.
+
+I think we all meant "scalar" here.  Nothing wrong with doing 'Range an a
+predicated array type.
+
+Thanks.
+
+****************************************************************