[Patch, fortran] PR31692 - Wrong code when passing function name as result to procedures

[FX Coudert]

>> 2007-05-04  Paul Thomas  <pault@gcc.gnu.org>
>>
>>         PR fortran/31292
>>         * trans-array.c (gfc_conv_array_parameter): Convert full  
>> array
>>         references to the result of the procedure enclusing the call.

OK with the following whitespace change:

> +   if (this_array_result)
> +     {
> +       if (g77 == 1 && TREE_TYPE (TREE_TYPE (se->expr)) != NULL_TREE
> + 	    && GFC_DESCRIPTOR_TYPE_P (TREE_TYPE (TREE_TYPE (se->expr))))
> + 	se->expr = gfc_conv_array_data (build_fold_indirect_ref (se- 
> >expr));

Mis-indentation of this last line?

I think "g77 && ..." would be way better than "g77 == 1 && ...". E.g.  
I spent some time to look if g77 could have other integer values (the  
all too common magical values

> +   /* Is this the result of the enclosing procedure?  */
> +   this_array_result = (full_array_var && sym->attr.flavor ==  
> FL_PROCEDURE);
> +   if (this_array_result)
> +     {
> +       this_array_result = (sym->backend_decl ==  
> current_function_decl);
> +       this_array_result |= (sym->backend_decl == parent);
> +     }

Why not the following?

+   /* Is this the result of the enclosing procedure?  */
+   if (full_array_var && sym->attr.flavor == FL_PROCEDURE)
+     this_array_result = sym->backend_decl == current_function_decl
+                         || sym->backend_decl == parent;
+   else
+     this_array_result = false;


Other than that, it's OK for me. The preexisting logic of this code  
is not crystal-clear, but hopefully we won't have it to change it  
every other day! :)

FX