[patch, fortran] Handle BT_CLASS as well as BT_DERIVED in trans-array.c

[Tobias Burnus]

Hello Thomas,

> the attached patch fixes a few instances where we only handled
> BT_DERIVED without taking care of BT_CLASS too.  Other files should
> probably be audited for this, as well.

I concur that other files should be audit - we already had a couple of 
bugs due to this. Side remark: With the current scheme, we will get in 
trouble as soon as we have unlimited polymorphism as then the relation 
  BT_CLASS => BT_DERIVED no longer holds.

Additionally, one needs to be careful whether one has the class 
container or the actual derived type. For some checks, looking at the 
class container - instead of the derived type it wraps - will lead to 
wrong code. I think the current patch is OK as the check for allocatable 
components, which should also work for BT_CLASS. However, I hope that 
Janus can confirm it.

Other than that, the patch looks OK and I think it can be committed.

Tobias