!standard 3.9.4(18/2)                                        07-10-24    AI05-0070-1/01
!class binding interpretation 07-10-24
!status ARG Approved  9-0-0  06-11-10
!status work item 07-10-24
!status received 07-10-03
!priority Low
!difficulty Easy
!qualifier Error
!subject Elaboration of interface types

!summary

The elaboration of an interface type creates the type and its first subtype.

!question

The dynamic semantics for interface types (3.9.4(18/2))
indicates that the elaboration "has no effect."  That seems
inconsistent with all other non-incomplete types,
which at a minimum say:

   The elaboration of a <blah>_type_definition creates
   the <blah> type and its first subtype...

Is there any reason for this inconsistency? (No.)

!recommendation

(See Summary.)

!wording

Modify 3.9.4(18/2):

The elaboration of an interface_type_definition creates
the interface type and its first subtype.

AARM Discussion: An interface_list is made up of subtype_marks,
which do not need to be elaborated, so an interface_list does not either.
This is consistent with the handling of discriminant_parts. That means that
the only effect is the creation of the type.


!discussion

It's possible that types are created at runtime (although this is rare).


!corrigendum 3.9.4(18/2)

@drepl
The elaboration of an @fa<interface_type_definition> has no effect. 
@dby
The elaboration of an @fa<interface_type_definition> creates
the interface type and its first subtype.

!ACATS Test

An ACATS test is not needed, as there still is no runtime effect here,

!appendix

From: Tucker Taft
Sent: Wednesday, October 3, 2007  9:47 AM

The dynamic semantics for interface types (3.9.4(18/2))
indicates that the elaboration "has no effect."  That seems
inconsistent with all other non-incomplete types,
which at a minimum say:

   The elaboration of a <blah>_type_definition creates
   the <blah> type and its first subtype...

The AARM note indicates that there is no particular work
to be done at run-time since the subtype_marks don't
require dynamic elaboration.  Be that as it may, at least
formally we think of the "creation" of a type as something
that happens dynamically.

****************************************************************

From: Robert Dewar
Sent: Wednesday, October 3, 2007  3:19 PM

In the SETL version it really *did* happen dynamically :-)

****************************************************************