(Class-) array finalization and initialization question

[Salvatore Filippone]

cray ftn version 8.2.1

Salvatore

On Fri, Feb 20, 2015 at 12:15 PM, Andre Vehreschild <vehre@gmx.de> wrote:
> Hi all,
>
> first of all, thank you for all your help.
>
> Next you are absolutely right: The most recent gfortran needs all elemental
> init()-subroutines to be impure. I was accidentally using an old version of the
> trunk gfortran. The mistake clearly was in between the ears in front of the
> monitor. Attached is the corrected test program for all interested.
>
> From the results I got so far:
>
> ifort (13.0.1 20121010, thanks Paul): No result.
> nagfort (6.0(Hibiya) Build 1034, thanks Harald): This looks reasonable and in
>         the way I interpret the F2008 standard. Does anyone object to this, so
>         that gfortran should do the same?
> pgfortran (14.10-0 64-bit, thanks Harald): The counts are IMHO to low, but the
>         resulting array is equal to nagfort's result, therefore I like the
>         array result.
> cray fortran (no version given, thanks Salvatore): I doubt we should go for
>         these counts/results. Looks like the implementation of f2008 features
>         in the compiler is still very basic.
>
> So I am willing to make gfortran behave like the nagfort one. Mostly because
> that is the way I think the standard is to be interpreted (just for reference:
> F2008, 5.3.10, §3). Any objections?
>
> On Thu, 19 Feb 2015 22:28:25 +0100 Tobias Burnus <burnus@net-b.de> wrote:
> <<< snip >>>
>> Hence, the procedure cannot be PURE - or the dummy argument not
>> polymorphic. If one has a nonpolymorphic dummy argument, one can check
>> the pureness of the finalizer at compile time. (Do we actually do this?)
>
> While waiting for any objections, I will check, if I can find anything in the
> gfortran code that looks like this check.
>
> Regards,
>         Andre
> --
> Andre Vehreschild * Email: vehre ad gmx dot de