[Bug fortran/50960] [OOP] vtables not marked as constant

["rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50960

--- Comment #17 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-11-03 14:34:02 UTC ---
(In reply to comment #16)
> (In reply to comment #11)
> > (In reply to comment #9)
> > > FAIL: gfortran.dg/extends_type_of_1.f03  -O0  (internal compiler error)
> > > FAIL: gfortran.dg/extends_type_of_3.f90  -O  (internal compiler error)
> > 
> > These two fail with:
> > 
> > internal compiler error: in fold_convert_loc, at fold-const.c:1894
> > 
> > Not quite sure what goes wrong there. Some typing problem?
> > 
> > 
> > Reduced test case:
> > 
> >  type :: t1
> >  end type
> > 
> >  type, extends(t1) :: t2
> >  end type
> > 
> >  class(t1), pointer :: c1
> >  type(t2) :: y
> > 
> >  if (.not. extends_type_of (y, c1)) call abort()
> > 
> > end
> 
> Looks like you are converting struct __vtype_MAIN___T1 to
> struct __vtype_MAIN___T1 *.  Thus probably too many TREE_TYPE ()
> wrappers somewhere or a forgotten address-taking.
> 
> Called from
> 
> 4608            {
> 4609              /* Scalar pointers.  */
> 4610              se.want_pointer = 1;
> 4611              gfc_conv_expr (&se, expr);
> 4612              gfc_add_block_to_block (&block, &se.pre);
> 4613              gfc_add_modify (&block, dest,
> 4614                                   fold_convert (TREE_TYPE (dest),
> se.expr))
> 
> where se.want_pointer (whatever it means) is not honored and se.expr
> is a variable of type struct __vtype_MAIN___T1.

Missing handling of se.want_pointer at least here:

void
gfc_conv_structure (gfc_se * se, gfc_expr * expr, int init)
{
...
  if (!init)
    {
      /* Create a temporary variable and fill it in.  */
      se->expr = gfc_create_var (type, expr->ts.u.derived->name);
      tmp = gfc_trans_structure_assign (se->expr, expr);
      gfc_add_expr_to_block (&se->pre, tmp);
      return;

but what's the desire of the caller?  Is it to get &expr?  Something
seems to be seriously wrong.