[Bug c++/57175] NRVO and alignment

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57175

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-05-05
                 CC|                            |ebotcazou at gcc dot
                   |                            |gnu.org
     Ever Confirmed|0                           |1

--- Comment #2 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2013-05-05 18:34:47 UTC ---
> in typeck.c (check_return_expr), before applying the NRV optimization, there
> is a check: DECL_ALIGN (retval) >= DECL_ALIGN (result)
> 
> It seems to me that this check is backwards and should be <= instead (compare
> with tree_nrv in tree-nrv.c which seems correct).

FWIW the Ada compiler also has the NRVO and the check is as in tree-nrv.c:

/* Return true if RET_VAL can be used as a Named Return Value for the
   anonymous return object RET_OBJ.  */

static bool
return_value_ok_for_nrv_p (tree ret_obj, tree ret_val)
{
[...]
  if (ret_obj && DECL_ALIGN (ret_val) > DECL_ALIGN (ret_obj))
    return false;