[patch, fortran] PR32432 SEGV/endless loop after: "ERROR: ... already is initialized"

[Janne Blomqvist]

:REVIEWMAIL:

On 7/3/07, Jerry DeLisle <jvdelisle@verizon.net> wrote:
> OK for trunk?

Yes. A few minor issues below though.

*************** gfc_assign_data_value (gfc_expr *lvalue,
*** 293,300 ****
  	      gfc_error ("'%s' at %L already is initialized at %L",
  			 lvalue->symtree->n.sym->name, &lvalue->where,
  			 &init->where);
! 	      gfc_free_expr (init);
! 	      init = NULL;
  	    }

  	  if (init == NULL)
--- 295,301 ----
  	      gfc_error ("'%s' at %L already is initialized at %L",
  			 lvalue->symtree->n.sym->name, &lvalue->where,
  			 &init->where);
! 	      return FAILURE;
  	    }

Won't this create a memory leak as you are no longer freeing init?

*************** gfc_assign_data_value (gfc_expr *lvalue,
*** 423,428 ****
--- 424,431 ----
      symbol->value = expr;
    else
      last_con->expr = expr;
+
+   return t;
  }

Why not get rid of the t variable and just "return SUCCESS;", since
you never actually use the t variable for anything beyond initially
setting it to SUCCESS?

Similarly in resolve.c,

! 	  t = gfc_assign_data_value (var->expr, values.vnode->expr, offset);
! 	  if (t == FAILURE)
! 	    break;

Why not just

if (gfc_assign_data_value (...) == FAILURE)
    break;

Finally, please use unified diffs in the future; I suspect I'm not the
only one who has difficulties following context diffs.

Thanks for the patch.

-- 
Janne Blomqvist