J3/10-249
To: J3
From: Malcolm Cohen
Subject: Editorial corrections for 10-165r2
Date: 2010 October 13


1. Introduction

10-165r2 suffers from a number of minor and fatal editorial flaws.
This paper attempts to improve these.

Note that other papers might also make changes in these areas.
Care needs to be taken in merging these changes.


2. Some philosophy

In previous Technical Reports we gave the actual edits to the Fortran
standard that were needed to implement the features.  That is in my view
unnecessary, however it is *absolutely essential* to specify precisely
which bits of the standard are replaced by the Technical Report.

When giving replacement syntax or constraints to existing syntax rules or
constraints, the new version *must* use exactly the same numbering as in
the Fortran standard.

In my view additional constraints and rules should be numbered in
accordance with the existing standard.  We already do this in edits made in
corrigenda, where we use {R/C}nnnn{a/b/c/...} where nnnn is the existing
rule/constraint number in the standard; to avoid confusion, I suggest using
a different scheme such as {R/C}nnnn{x}m, where m indicates the number of
additional rules/constraints at that point.  (If m is always ==1, we could
omit it.)


3. Edits to 10-165r2

[throughout] Stop using the confusing page footers with the clause numbers
in them, as it is all too easy to confuse these with the page numbers.
(Furthermore they add no real info and are forbidden by ISO.)

[1:27,29] "intrinsic procedures" -> "an intrinsic procedure".
[1:29] delete "one of".
{Correct number.}

[3:1] "2 Data Attributes" -> "2 Type specifiers and attributes"
{Improvement.}

[3:2+] Insert new text as follows

  "The syntax rule R403 \si{declaration-type-spec} in subclause 4.3.1.1 of
   ISO/IEC 1539-1:2010 is replaced by

   \bnfx{403}{declaration-type-spec}{\si{intrinsic-type-spec}}  \\
   \bnfo{TYPE ( \si{intrinsic-type-spec} )} \\
   \bnfo{TYPE ( \si{derived-type-spec} )} \\
   \bnfo{CLASS ( \si{derived-type-spec} )} \\
   \bnfo{CLASS ( * )} \\
   \bnfo{TYPE ( * )}"
{Fatal.}

[3:6-7] Merge into one constraint
  "\dcons[C407x1]{An assumed-type entity shall be a dummy variable that
   does not have the CODIMENSION or VALUE attribute."
{Improvement.}

[3:11+] Insert new text as follows
  "The syntax rule R515 \snref{array-spec} in subclause 5.3.8.1 of IS
   1539-1:2010 is replaced by

   \bnfx{515}{array-spec}{\si{explicit-shape-spec}\st{-list}}  \\
   \bnfo{\si{assumed-shape-spec}\st{-list}}  \\
   \bnfo{\si{deferred-shape-spec}\st{-list}}  \\
   \bnfo{\si{assumed-size-spec}} \\
   \bnfo{\si{implied-shape-spec}\st{-list}} \\
   \bnfo{\si{assumed-rank-spec}}"
{Fatal.}

[3:14-16] Replace with
  "\bnfx{522x1}{assumed-rank-spec}{..}

   \dcons[535x1] An assumed-rank entity shall be a dummy variable that does
   not have the CODIMENSION or VALUE attribute."
{Improvement.}

[3:20] "array"->"array or scalar object".
{Misleading.}

[3:25-26] Replace with
  "The constraint C1255 of subclause 12.6.2.2 is replaced by

   \dcons[1255]{(\snref{proc-language-binding-spec})
   If \si{proc-language-binding-spec} is specified for a procedure,
   each dummy argument of the procedure shall be an interoperable procedure
   (15.3.7) or an interoperable variable (15.3.5, 15.3.6) that does not
   have both the OPTIONAL and VALUE attributes.
   If \si{proc-language-binding-spec} is specified for a function, the
   function result shall be an \termi{interoperable} scalar variable.}"
{Fatal.}

[5:3] "Whether" ->
  "Additionally to the characteristics listed in subclause 12.3.2.2 of IS
   1539-1:2010, whether".
{Contradiction.}

[5:5-6] Replace with
  "Additionally to the rules of subclause 12.4.2.2 of ISO/IEC 1539-1:2010,
   a procedure is also required to have an explicit interface if it has a
   dummy argument that is assumed-type or assumed-rank."
{Correct requirement.}

[5:7+] I don't have an edit here, but careful reading of clause 12 could
       turn up a contradiction that needs to be addressed.  That should be
       checked later.

[5:20] Append
  "The value of the lower and upper bound of dimension \textit{N} of the
   dummy argument are equal to the result of applying the LBOUND and UBOUND
   intrinsic inquiry functions to the actual argument with DIM=\textit{N}
   specified."
{Somewhat convoluted to handle assumed-size, even though we don't have it
 yet.  Wording could be further improved, perhaps the previous sentence
 should also be replaced.  Note that array sections and array expressions
 don't have "bounds" in themselves... just shape.}

[12:39+] Insert text
  "The rules in this subclause replace the contents of subclause 15.3.7 of
   ISO/IEC 1539-1:2010 entirely."
{Fatal.}

[12:later] I'm convinced these rules are wrong, but it is probably better
           to wait until the dust settles on the technical changes before
           trying to get them right (since they might be changing anyway).

[19] Do we really need an index for this document?  I don't think so.
     In any case, the index is ***NOT*** an annex so will be rejected
     by ISO.

     Delete Annex B.

===END===