Issue of association of unconstrained array parameters

John,

I have a note to myself to follow up your comment about an issue in
definition of association of the actual with an out-mode formal parameter in
a subprogram call.

Do I understand the issue correctly, that the formal's index ranges come
from the associated actual, but the actual is not yet associated when the
formal is elaborated and created? If that is the issue, I guess it stems
from a somewhat imprecise definition of what "association" actually
involves.

According to the LRM, an association list "establishes correspondences"
between formals and actuals, and for an unconstrained formal, the index
ranges are taken from the "corresponding association element." So presumably
the correpondence needs to be established before the index ranges can be
inferred.

The specification of dynamic elaboration involves elaborating the formal
parameter list, after which "Actual parameters are then associated with
formal parameters." The problem seems to be that, though the association
list establishes the correspondence, there is no explicit statement that
this is done by *associating* the actuals with the formals.

One possible interpretation is that "associating the actuals with the
formals" means interpreting the association list to determine the
correpondence. This interpretation gives rise to the problem you've
identified.

Another possible interpretation is that "associating the actuals with the
formals" means invoking the parameter passing mechanism to copy values or
pass references for the parameters. In this case, the use of the word
"associate" is really inappropriate, and some other wording should be found.
Given that the meaning of an association list is elsewhere specified, my
suspicion is that this second interpretation was intended. It would avoid
the issue you raised.

Either way, there is an issue to be addressed. In the case of the first
interpretation, we need to relax the sequential order of parameter
elaboration and association, to allow index ranges to be taken from the
corresponding actual during elaboration. In the case of the second
interpretation, we need to change the wording to avoid inappropriate use of
"asociate."

Let me know what you think, so we can raise the appropriate IR for ISAC.
Meanwhile, we can proceed with addressing FT-12 in the assumption ISAC will
fix the issue. (In other words, us with different hats on!)

Cheers,

PA

--
Dr. Peter J. Ashenden                peter@ashenden.com.au
Ashenden Designs Pty. Ltd.           www.ashenden.com.au
PO Box 640                           VoIP: 0871270078@sip.internode.on.net
Stirling, SA 5152                    Phone (mobile):  +61 414 709 106
Australia