(j3.2006) Need an interp for this?

[Malcolm Cohen]

>Change allocate to deallocate, and assume d%l2p is allocated.  Then what 
>happens to the associate name?

Program is not conforming, but perhaps only because the standard doesn't 
establish an interpretation for it.  16.5.1.6p5 cannot be satisfied if the 
selector disappears in some way.

That's the kind of thing I had in mind when I wrote
>> Some of the more obscure ones might be harder to extract though.

BTW, you don't need the selector to have the ALLOCATABLE or POINTER attribute 
for this to happen; being a subobject of such a variable is sufficient.

BTW2, I think there is at least one other situation where this arises, and again 
we fall back on the last resort.  I can't quite bring it to mind though.

BTW3, I don't actually object to improving the wording here (we could call out 
some of the cases explicitly, even if we don't get all of them), but I don't 
give it a very high priority.

...
>I had in mind primarily the requirement that actions that affect the dummy 
>argument be taken through the dummy argument.

We didn't do that.  Other languages with similar constructs don't do that either 
(then again, other languages mostly don't have our argument association rules so 
that's not a strong indication).

However I do distinctly recall discussing this topic, maybe at one of the London 
meetings, and at that time we decided not to put such a requirement into the 
standard.

Cheers,
-- 
................................Malcolm Cohen, Nihon NAG, Tokyo.