On OpenVMS targets, a configuration pragma Float_Representation
indicates that the specified representation (IEEE, VAX) should
be the default for predefined floating point types and for
new floating point definitions.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-12-23 Geert Bosch <bo...@adacore.com>
* sem_ch3.adb (Can_Derive_From): Check matching Float_Rep on VMS.
Index: sem_ch3.adb
===================================================================
--- sem_ch3.adb (revision 182654)
+++ sem_ch3.adb (working copy)
@@ -15333,10 +15333,23 @@
Spec : constant Entity_Id := Real_Range_Specification (Def);
begin
+ -- Check specified "digits" constraint
+
if Digs_Val > Digits_Value (E) then
return False;
end if;
+ -- Avoid types not matching pragma Float_Representation, if present
+
+ if (Opt.Float_Format = 'I' and then Float_Rep (E) /= IEEE_Binary)
+ or else
+ (Opt.Float_Format = 'V' and then Float_Rep (E) /= VAX_Native)
+ then
+ return False;
+ end if;
+
+ -- Check for matching range, if specified
+
if Present (Spec) then
if Expr_Value_R (Type_Low_Bound (E)) >
Expr_Value_R (Low_Bound (Spec))
