Re: Need advice on PR 19929

[Tobias Schlüter <tobias dot schlueter at physik dot uni-muenchen dot de>]

Steve Kargl wrote:
> Well, I think I have determined at least a working hacking.
> I don't know if there is a better way to this, but if I do
> 
>    if (expr->symtree->n.sym->attr.dimension 
>        && expr->symtree->n.sym->ts.derived == NULL)
> 
> then I execute the else{} branch.

The correct check for a non-derived type is
   expr->symtree->n.sym->ts.type != BT_DERIVED

Looking over the code in trans_deallocate, I'm wondering why we don't look
through the gfc_ref-chain, IOW how do we make sure that we're dealing with the
component when building the call?  What doesn't look right to me is that the
component being deallocated is not of a derived type, so if we were doing the
right thing, a check for a derived type wouldn't make a difference.

- Tobi