[Patch,Fortran] PRs 45451 / 43018: Fixes for allocatables

[Tobias Burnus]

Hi Janus,

On 10/26/2010 08:37 AM, Janus Weil wrote:
> I think the patch is ok. Thanks for taking care of this.

Thanks for the review. (Committed as Rev. 165936.)

>> PS: This patch does not fix all problems with PR 45451.
> You mean comment #8/9?

Actually, I rather mean the issues I mentioned in comment 16. I am not 
sure whether comment 8 is a new bug of the consequence of a missing deep 
copy. And regarding comment 9: I am too lazy to debug an invalid program 
- at least until the known problems are fixed.

Frankly, I have no idea how to best solve the deep-copying problem for 
polymorphic types. One could think of creating hidden COPY and FREE 
type-bound procedures, which free the added allocatables of the type and 
then call the parent free function. That is:

if (a$vtab->free)
   free(a)

as deallocation function -- and in the deallocation function:
   if (a$vtab->parent->free)
     free(a)
if a type extension does not have a allocatable components, it simply 
directly points to the parent.

Tobias