To: J3                                                     J3/19-115
From: Steve Lionel
Subject: Making note of Statement Entities
Date: 2019-January-20

Reference: 18-007r1

Introduction
------------

A recent discussion in comp.lang.fortran involved confusion about
whether the appearance of a name as an ac-do-variable in a module
created a declaration of the variable in that module. It was reported
that more than one current compiler treated an ac-do-variable that way,
even though it is expressly prohibited by 19.4p5 (Statement and Construct
Entities):

"The appearance of a name as a data-i-do-variable of an implied DO in a
DATA statement or an ac-do-variable in an array constructor is not an
implicit declaration of a variable whose scope is the scoping unit that
contains the statement."

Unfortunately, someone reading 7.8 (Construction of array values), or
8.6.7 (DATA Statement), is given no clue in nearby text that an
ac-do-variable or data-i-do-variable is a statement entity, and might
reasonably assume that it follows the normal language rules for implicit
declaration and scope.

A similar situation could have occurred for index-name in a
concurrent-header, but 11.1.7.1p4 says "The scope and attributes of an
index-name in a concurrent-header (DO CONCURRENT) are described in
19.4.". No similar text appears in 7.8 nor 8.6.7. The omission is
confusing and has led to implementation errors in compilers. It would
be beneficial to add appropriate wording to 7.8 and 8.6.7.

Suggested Edits
---------------

In 7.8 (Construction of array values), add a new paragraph after p1
(88:12+)

"The scope and attributes of an ac-do-variable are described in 19.4."

In 8.6.7 (DATA statement), add a new paragraph after p4 (110:26+)

"The scope and attributes of a data-i-do-variable are described in 19.4"

===END===