J3/99-163r1

Date:    3rd June 1999
To:      J3
From:    Malcolm Cohen
Subject: Unresolved issue 130


1. Introduction

In several places there is some confusion between "specifications",
"declarations", and "specification statements".

2. Discussion

(i)
The problems have arisen because we do not have a term for a statement that
is
purely a declaration - <specification-stmt> is much too narrow, and so
(perhaps surprisingly) is <declaration-construct> (e.g. it misses out
PARAMETER).  There are examples in Annex C of long-standing mistakes
(calling
PARAMETER a specification statement).

Maybe we should make up a name for this?  declarative-construct?  I note
that
using "specification statement" for this instead of the narrow meaning
occurs
in several places through the standard!

(ii)
BTW, is "nonkind type parameter value" ok?  I presume so, though I don't see
anything like a definition of what this would mean (I know what *I* think
it means!)  I'll leave this one alone.

(iii) [261:42+] says
  "The local identifier of an entity made accessible by a USE statement may
   appear in no other specification statement that would cause any attribute
   (5.1.2) of the entity to be respecified in the scoping unit that contains
   the USE statement, except that it may appear in a PUBLIC or PRIVATE
   statement in the scoping unit of a module and it may be given the
   ASYNCHRONOUS or VOLATILE attribute."
So it is ok to re-specify its value in a PARAMETER statement then?
I think not - this ought to be fixed.  Also - if it does not possess the
POINTER attribute, putting it in a POINTER statement would not be
"respecifying" the attribute!

(iv)
In 5.5 we say that
  "A namelist group object shall either be accessed by use or host
association
   or shall have its type, type parameters, and shape specified by previous
   specification statements in the same scoping unit or by the implicit
rules
   in effect for the scoping unit."
Which means that
   REAL FUNCTION I()
     NAMELIST/NL/I
is not legal.  This should really be an F95 interp since the standard is
not,
I think, actually contradictory or ambiguous - it just says the wrong thing!
But since we are not doing interp processing I am proposing to fix it in
F2002.

3. Edits

{Fix incorrect Note:
  (i) <declaration-type-spec> is used in all of the "declarative
constructs",
      including procedure headings;
  (ii) <type-spec> is only used in array constructors, not structure
constr.}
[69:27] Change "specification" to "non-executable"
        Change "a constructor" to "an array constructor".

{Revert wording to be more like the previous version...}
[72:31]
Change
   "In a declaration, an array bound or nonkind type parameter value"
To
  "A <specification-expr> in an <array-spec> or a <type-param-value>
   corresponding to a nonkind type parameter in a <declaration-type-spec>"
[72:32]
Change
  "provided the expression"
to
  "provided it"

{Delete the J3 Note.}
[73:1-14] Delete.

{Fix NAMELIST}
[99:6] After "specification statements" insert "or the procedure heading".

{Try to fix USE association to prohibit what mere mortals mean by
"local declaration" of USE-associated names.}
[261:43] Change "specification" to "nonexecutable"
         Change "respecified" to "specified".
{Or should we just delete the word "specification" and change "respecified"
to "explicitly specified"?  I don't think so.}
{Alternative: list everything that is prohibited - the trouble with this is
that although it is then crystal-clear it is quite long - see the list in
14.6.1.3 - it has 15 items.}