(j3.2006) (SC22WG5.5416) [ukfortran] RE: J3/14-279: J3 Fortran interp letter ballot #32 revised - due 9-Jan-2015

[N.M. Maclaren]

I am not a member of J3, but agree with all of the interpretations
except F08/0110(1) and (potentially) F08/0122(6).


F08/0105

The answer might be clearer if it added something like the following
"Arguments with the VALUE and ASYNCHRONOUS attributes can be used for
asynchronous transfers that both start and complete during that
procedure call."


F08/0110

(1): I agree with the intent, of course.

I cannot see that the new wording excludes code like the following:

    INTEGER :: n
    READ (UNIT = n+1, IOSTAT = n)

or:

    INTEGER FUNCTION Joe (INTEGER x)
        RETURN x
    END FUNCTION Joe

    INTEGER :: n = 1
    READ (UNIT = Joe(n), IOSTAT = n)


F08/0115

I still have a concern about the case when the procedure is in a
language other than Fortran.  The point is that it could be argued that
a C parameter is not a dummy argument, and the interface block is not
evaluated by the call, so this does not apply.  However, formally, that
becomes relevant only when TS 29113 is incorporated (i.e. for Fortran
2xxx).


F08/0120

In the answer, "emend" => "amend".


F08/0122

(6): The answer makes the intent clear, and I will try to write a J3
paper explaining why there is a problem.  It is not an easy one to
explain.


Regards,
Nick Maclaren.