Again on gfortran for Mac OS-X

[Paul Thomas]

FX,
>>
> I just posted it as PR34681 - the problem arises because t.data is 
> nullified one each entry to 'mah'.
>
This is one that Erik and I should have seen in the first place.  
Derived types with allocatable components are excluded from the return 
from gfc_trans_deferred_array when TREE_STATIC (desc) = 1.  Clearly the 
nullification on entry and exit needs to be suppressed in this case.

This fixes it:

Index: gcc/fortran/trans-array.c
===================================================================
*** gcc/fortran/trans-array.c   (revision 131237)
--- gcc/fortran/trans-array.c   (working copy)
*************** gfc_trans_deferred_array (gfc_symbol * s
*** 5527,5533 ****
      }

    /* NULLIFY the data pointer.  */
!   if (GFC_DESCRIPTOR_TYPE_P (type))
      gfc_conv_descriptor_data_set (&fnblock, descriptor, 
null_pointer_node);

    gfc_add_expr_to_block (&fnblock, body);
--- 5527,5533 ----
      }

    /* NULLIFY the data pointer.  */
!   if (GFC_DESCRIPTOR_TYPE_P (type) && sym->attr.save != SAVE_EXPLICIT)
      gfc_conv_descriptor_data_set (&fnblock, descriptor, 
null_pointer_node);

    gfc_add_expr_to_block (&fnblock, body);
*************** gfc_trans_deferred_array (gfc_symbol * s
*** 5545,5551 ****
        gfc_add_expr_to_block (&fnblock, tmp);
      }

!   if (sym->attr.allocatable)
      {
        tmp = gfc_trans_dealloc_allocated (sym->backend_decl);
        gfc_add_expr_to_block (&fnblock, tmp);
--- 5545,5551 ----
        gfc_add_expr_to_block (&fnblock, tmp);
      }

!   if (sym->attr.allocatable && sym->attr.save != SAVE_EXPLICIT)
      {
        tmp = gfc_trans_dealloc_allocated (sym->backend_decl);
        gfc_add_expr_to_block (&fnblock, tmp);


I'll commit it as obvious tomorrow night, together with testcase and 
ChangeLog.

Cheers

Paul