Question about finalization of abstract types

[Damian Rouson]

Submitted as PR 59694.

Damian 

On Jan 5, 2014, at 6:52 AM, Tobias Burnus <burnus@net-b.de> wrote:

> Hi Damian,
> 
> Damian Rouson wrote:
>> I don’t think the following code falls into one of the three categories of cases you listed for which finalization is not yet implemented, but gfortran doesn’t call the final subroutine:
> 
> The problem is that gfortran regards "bar" as unused - and never creates the variable:
> 
>>   block
>>     type(foo) :: bar
>>   end block
> 
> (See also -fdump-tree-original. The the variable isn't created is also problematic with regards to debugging as "pt bar" wouldn't work.)
> 
> Try:
> 
>  block
>    type(foo) :: bar
>    print *, bar
>  end block
> 
> which prints:
> ---------------cut--------
> 
> Hello from finalizer.
> ---------------cut--------
> 
> I think it deserves a PR - and we have to think about how to handle 'unused' variables in general. At least for nonallocatable nonpointer local (-> nonpolymophic) derived-type variables with finalizers, we have to generate the variable. For others (-> debugging) we may (at least for -O0).
> 
> Tobias