To: J3                                                     J3/26-152
From: Brad Richardson & Generics
Subject: Response to UTI 030
Date: 2026-May-26

Reference: 26-007r1

UTI 030 states:

    "This makes interface bodies even more inconsistent than before.

    A module procedure interface body has host association but
    inconsistently has the "default" implicit mapping table.
    Unfortunately this was not noticed when we did submodules, so we
    cannot fix it now without an incompatibility. But that is water
    under the bridge.

    Now we have, in a template or templated subprogram, interface
    bodies for deferred procedures having host association and the host
    (null) implicit mapping, but interface bodies for non-deferred
    procedures do not have host association even though they have the
    host (null) implicit mapping. This is the reverse of the
    unfortunate mpib inconsistency, but in any case, it is a new
    inconsistency.

    Surely, in a template or templated subprogram, an interface body
    for a non-deferred procedure is just as likely to need host
    entities like the deferred types and constants?

    This needs careful consideration. It would be better either for a
    non-deferred interface body to have default mapping and not host
    association, or to have host mapping and host association. Or at
    least follow the mpib precedent and give it host assoc but the
    default mapping! The current situation seems to be the worst
    possible."

To clarify the inconsistency here, we have effectively the following
different situations.

  INTERFACE
    SUBROUTINE EXT_SUB()
      ! No host association
      ! implicit typing
    END SUBROUTINE

    MODULE SUBROUTINE MOD_SUB()
      ! Host association
      ! implicit typing
    END MODULE
  END INTERFACE

  TEMPLATE TMPL(DEF_SUB)
    DEFERRED INTERFACE
      SUBROUTINE DEF_SUB()
        ! Host association
        ! No implicit typing
      END SUBROUTINE
    END INTERFACE

    INTERFACE
      SUBROUTINE EXT_SUB()
        ! No host association
        ! No implicit typing
      END SUBROUTINE
    END INTERFACE
  END TEMPLATE

I.e. we have different implicit typing behavior for the interface for
EXT_SUB depending on where it appears. This was unintentional, and is
undesirable. It would indeed be better if the behavior of the interface
for EXT_SUB did not depend on its surrounding context. The following
edits achieve this.

[147:23-24] Change "contained in a TEMPLATE construct or templated
            subprogram (16.2)" to "in a deferred interface block",
            such that the bullet now reads:

            "the default for a program unit, or an interface body that
            is not in a deferred interface block, is default integer
            if the letter is I, J, ..., or N and default real otherwise,"

[147:27] Change "contained therein" to "in a deferred interface block",
         such that the bullet now reads:

         "the default for a TEMPLATE construct, templated subprogram,
         or an interface body in a deferred interface block, is the
         null mapping."

--END--