ISO/IEC JTC1/SC22/WG5 N2120

         Result of the WG5 straw ballot on technical changes

                             John Reid

All four changes are approved by WG5. 

WG5 requests that J3 considers the comments at its next meeting. 


            16-   16-   16-  16-  
           277r1 280r2 285r2 289 
Bader        Y     Y     Y    Y    
Chen         Y     Y     Y    Y    
Clune        Y     Y     Y    Y
Cohen        Y     Y     Y    Y    
Corbett      C     Y     C    Y   
LeAir        Y     Y     Y    Y    
Long         C     Y     Y    Y    
Maclaren     C     Y     -    Y    
Muxworthy    Y     Y     Y    Y    
Nagle        Y     Y     Y    Y    
Reid         C     Y     Y    Y    
Shterenlikht Y     Y     Y    Y    
Sircombe     Y     Y     Y    Y    
Snyder       Y     Y     Y    Y   
Whitlock     Y     Y     Y    Y   

 

Comments

06-277r1 

Maclaren
Because assumed-size arrays can be converted to assumed-rank
objects, this change permits a usage that has no defined behaviour, and
where there are several possibilities compatible with the rest of the
standard.  C_SIZEOF (18.2.3.7 486:24) forbids assumed-size arrays and,
in lines 29-31 defines the result in terms of the element size and
number of elements (which is unspecified for assumed-size arrays).  Note
that I am not voting against the change in principle, but against the
lack of an explicit specification for this case.

Reid
In relation to Nick Maclaren's vote, the size of an assumed-size array 
is specified in 8.5.8.5 of J3/16-007r2.

Long
I'm afraid that does not resolve the issue Nick raised. In 9.5.2 we say 
"An assumed-size array (8.5.8.5) is permitted to appear as a whole array 
in an executable construct or specification expression only as an actual 
argument in a procedure reference that does not require the shape."  
This is mainly because, even though we define a size in the standard, 
that information is not normally available to a subprogram with an 
assumed-size dummy argument.   That is why the SIZE intrinsic has 
restrictions on assumed-size array arguments, and the existing C_SIZEOF 
function prohibits assumed-size arguments.  That limitation for C_SIZEOF() 
needs to be preserved for the case of an assumed-rank argument that is 
argument associated with an assumed-size array. By allowing assumed-rank 
arguments to SIZE, we need a similar limitation there as well.  A note 
about this is on my list of issues that need papers at m212. Thanks to 
Nick for exposing this oversight.

Corbett
While I share Nick Maclaren's concern regarding assumed-size arrays,
I recognize that that is a separate issue from the one being balloted.



06-285r2 

Maclaren
I am abstaining deliberately, because I cannot work out
whether the change clarifies the situation or not.

Corbett
I agree with the technical change.  I have a concern about the wording
apart from the technical change.

The text of Subclauses 7.5.6.2 and 7.5.6.3 seem to conflict.
Subclause 7.5.6.2 states "Only finalizable entities are finalized."
Subclause 7.5.6.3 says that entities are finalized without requiring
that they be finalizable.