(j3.2006) (SC22WG5.5110) [ukfortran] WG5 letter ballot 7 on Fortran 2008interpretations
Thu Oct 24 22:06:00 EDT 2013
Malcolm Cohen writes:
> -----Original Message-----
> From: j3-bounces at mailman.j3-fortran.org [mailto:j3-bounces at mailman.j3-fortran.org]
> On Behalf Of Malcolm Cohen
> Sent: Thursday, October 24, 2013 19:43
> To: sc22wg5 at open-std.org
> Subject: (j3.2006) (SC22WG5.5110) [ukfortran] WG5 letter ballot 7 on Fortran
> David Muxworthy writes:
> >Should new constraint C852a have a reference to R864?
> The short answer is "no".
> The long answer is long...
> People are frequently confused by rule references in constraints, variously
> (a) it limits the application to the rule referenced [most common]
> (b) it applies the constraint everywhere that rule appears,
> (c) they just ignore it and read what they think the constraint is saying.
> Actually maybe (c) is most common.
I strongly disagree with Malcolm's interpretation. Unfortunately for Malcolm,
the Fortran standard directly contradicts his interpretation of constraints that are
associated with syntax rules. See section 1.4.2p2. It expressly states that constraints
that are associated with syntax rules are part of those syntax rules wherever they appear.
> Anyway, in my opinion constraints should be written identically to other normative
> requirements, and not use some special notation to limit/broadcast/whatever the
> constraint. If the constraint is written unambiguously in that fashion, then
> people don't get so confused by the rule reference because it is totally
> superfluous. And then we can and should drop the rule reference entirely.
This was the case in Fortran 90 and Fortran 95. This previous practice caused
ambiguity in interpreting the applicability of some constraints. Did they apply to only
one syntax rule or to several? In the development of Fortran 2003, J3 explicitly debated
whether to associate certain constraints with particular syntax rules. J3 explicitly
decided to adopt the current convention specifically to eliminate this kind of ambiguity.
I was at the meeting at which this practice was adopted.
I believe that the current practice has served us well.
> C852a is indeed written just like any normative requirement and does not require
> any rule reference for its understanding or application.
And, it is explicitly authorized by section 1.4.2p3, the paragraph immediately
after the paragraph that limits rule-associated constraints specifically to those rules.
Constraints not explicitly associated with specific rules are general in applicability.
> ................................Malcolm Cohen, Nihon NAG, Tokyo.
Craig T. Dedo
17130 W. Burleigh Place
P. O. Box 423 Mobile Phone: (414) 412-5869
Brookfield, WI 53008-0423 E-mail: < <mailto:craig at ctdedo.com> craig at ctdedo.com>
Linked-In: <http://www.linkedin.com/in/craigdedo> http://www.linkedin.com/in/craigdedo
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the J3