CVS difference for ais/ai-00304.txt

Differences between 1.6 and version 1.7
Log of other versions for file ais/ai-00304.txt

--- ais/ai-00304.txt	2004/04/06 19:57:09	1.6
+++ ais/ai-00304.txt	2004/06/10 05:39:57	1.7
@@ -1021,3 +1021,28 @@
+From: Vincent Marciante
+Sent: Saturday, May 15, 2004 11:59 AM
+Subject: Opposite "solution" for Reemergence of predefined equality in generics
+If it is the case that the "Goodenough Trick" to redefine
+equality for a non-limitted type was really an unintended
+loophole, then a restriction pragma that would disallow
+redefinition of predefined "=" for a type that is not limitted
+(either via the trick or via the Ada95 allowance to just do it)
+might be enough to solve the problem.  The equality that
+would appear in generics then could never be the "wrong" one
+because it would be the only one that was allowed - the predefined one!
+This would just formalize the approach that I (and maybe others)
+have used and have enforced by manual code review which is
+"Don't do that!".
+If the above does not really solve the problem for reasons
+that I have not know to consider, I hope that the concept of
+this "opposite" approach inspires a new idea that really would
+improve the situation instead of dropping the issue all together.

Questions? Ask the ACAA Technical Agent