(j3.2006) (SC22WG5.3540) [ukfortran] N1718: co_lbound and co_ubound

[Bill Long]

Malcolm Cohen wrote:
> On Sat, 01 Mar 2008 04:59:51 +0900, John Reid <j.k.reid at rl.ac.uk> wrote:
>   
>> N1718 is without co_lbound and co_ubound. I think this is a mistake.
>>     
>
> I agree.
>
> In retrospect it is a pity that 08-131 didn't have a reminder NOT to
> delete those (they are mixed in with the collectives), as it was a
> pretty easy mistake to make...
>
> ...but the mistake and the responsibility are certainly mine alone.
> My apologies.
>
> But does this mean we need a new document before proceeding to CD?
> The missing routines are certainly convenient (and should be reinstated
> in the next revision whenever that is) but they are I think not
> essential to coarray programming (unlike THIS_IMAGE et al), so
> it doesn't appear to be a fatal flaw.
>   

Cobounds for allocatable coarrays are not generally available to the 
user without these functions, unless the user manually keeps track of 
them in separate variables. I think that not having them rises above the 
level of inconvenience, and is certainly inconsistent with the way we 
treat ordinary bounds.  Especially in the case of procedure interfaces, 
having to carry along cobound information as separate variables would be 
out of character with modern Fortran coding forms.

> (Producing a new document is neither cost-free nor risk-free...)
>
> Opinions?
>   

I would prefer a new CD document.   I believe that public comments would 
prompt the addition of these before we get to the FCD stage.  Adding 
them into the CD will reduce the number of changes between the CD and 
FCD, which reduces the risk of problems in the arguably more important 
FCD document.

Cheers,
Bill

> Cheers,
>   

-- 
Bill Long                                   longb at cray.com
Fortran Technical Support    &              voice: 651-605-9024
Bioinformatics Software Development         fax:   651-605-9142
Cray Inc., 1340 Mendota Heights Rd., Mendota Heights, MN, 55120