To: J3                                                     J3/26-215r2
Subject: Edits to repair PROCEDURE_NAME
From: Dan Bonachea & Lorri Menard & JoR
Date: 2026-October-27
References: 26-007r2, 26-184r1, 26-215r1

1. Introduction
---------------

The module procedure PROCEDURE_NAME is new in Fortran 2028, and there
are a few typos in the description of the procedure.

Paper 26-215r1 (an earlier revision of this paper) was discussed and passed by
J3 on 2026-10-26, after which defects were found in the same subclause whose
resolution involves overlapping edits.  This paper supersedes prior revisions.

2. Updates in Revision 2
------------------------

The support for adding a procedure's binding-name into the contextual
output has shown some potential technical problems.  It was originally
added by subgroup because it was considered "free", but further
examples, and in fact original edits promoted by this paper, have
shown that not to be true.

One area of potential confusion is with procedures contained within a
routine that has a binding name.  For example:

SUBROUTINE S() BIND(C, NAME="S")
CONTAINS
    SUBROUTINE INNER()
        USE ISO_FORTRAN_ENV
        PRINT *, PROCEDURE_NAME(.TRUE.)
    END SUBROUTINE
END SUBROUTINE

Should it return S=S.INNER, or simply S.INNER?  We suggest that it
should be subprogram-name.internal-subprogram-name, but it is not
unlikely there could still be questions.

The more egregious issue is with deferred parameters and templates,
as shown below:

MODULE EXAMPLE2
    TEMPLATE TMPL{C}
        DEFERRED CHARACTER(*), PARAMETER :: C
    CONTAINS
        SUBROUTINE S() BIND(NAME=C)
            PRINT *, PROCEDURE_NAME(.TRUE.)
        END SUBROUTINE
    END TEMPLATE
END MODULE

In this case the value for "C" is only known at template instantiation time,
not at compile-time of the template. This implies the result of a PROCEDURE_NAME
invocation within a template *might* be a deferred constant, but only in this
corner case. In all other cases the result of a PROCEDURE_NAME invocation within
a template is constant at template declaration time (which was the intent).

With this all in mind subgroup recommends that we remove the output
of binding-name from the CONTEXT=.TRUE. return value.

This simplifies the edits requested in original 26-215

3. Edits
--------
17.10.2.29 PROCEDURE_NAME ([CONTEXT])

[550:16] After the **Result Value."" and before "If CONTEXT is not ..." add
"The result value is in upper case."
making this line read:
"**Result Value.** The result value is in upper case.  If CONTEXT is not
present or has the value false, the result value is determined as follows:"

[551:10-12] Remove case(ii), and renumber the remaining cases.

[551:15] One bracket is slightly misplaced, as cannot have
internal-subprogram-name without a subprogram name
module-name[/submodule-name]:[subprogram-name][.internal-subprogram-name]
Move the right bracket after "subprogram-name" to the end of the line as:
module-name[/submodule-name]:[subprogram-name[.internal-subprogram-name]]

[551:21] Remove the "**or** =binding-name" from the syntax, resulting in:
subprogram-name[.internal-subprogram-name]

[551:25] Using "enclosing scope unit" seems too broad
Replace "enclosing scope unit" with "innermost scoping unit that contains
the templated subprogram or TEMPLATE construct".

[551:30-31] Remove these two lines; binding-name has been removed, and the
statement that the result value is in upper case is at the beginning, not
buried here at the end.

[552.4] Modify the comment to remove the Qq binding-name, resulting in:
PRINT *, procedure_name(.TRUE.) ! 'A:S2'

--END--