Re: tree_ssa_useless_type_conversion is too optimistic

[Richard Guenther <richard dot guenther at gmail dot com>]

On 5/15/05, Richard Guenther <rguenth@tat.physik.uni-tuebingen.de> wrote:
> 
> In fact, tree_ssa_useless_type_conversion_1 contradicts itself if given
> the two (C language) pointer types "char *" and "const char *".  Asking
> the langhook if they are compatible, it says no.  But then in the
> section that reads
> 
>   /* Pointers and references are equivalent once we get to GENERIC,
>      so strip conversions that just switch between them.  */
>   else if (POINTER_TYPE_P (inner_type)
>            && POINTER_TYPE_P (outer_type)
>            && TYPE_REF_CAN_ALIAS_ALL (inner_type)
>               == TYPE_REF_CAN_ALIAS_ALL (outer_type)
>            && lang_hooks.types_compatible_p (TREE_TYPE (inner_type),
>                                              TREE_TYPE (outer_type)))
>     return true;

Note that if the above only would apply to the C++ FE (suppose no
other frontends use REFERENCE_TYPE), the C++ langhook contains

  /* Once we get to the middle-end, references and pointers are
     interchangeable.  FIXME should we try to replace all references with
     pointers?  */
  if (POINTER_TYPE_P (x) && POINTER_TYPE_P (y)
      && TYPE_MODE (x) == TYPE_MODE (y)
      && TYPE_REF_CAN_ALIAS_ALL (x) == TYPE_REF_CAN_ALIAS_ALL (y)
      && same_type_p (TREE_TYPE (x), TREE_TYPE (y)))
    return 1;

which is _not_ the same as the test above - it uses same_type_p, while
the code in tree_ssa_useless_type_conversion_1 ends up using
same_type_ignoring_top_level_qualifiers_p from the C++ langhook:

int
cxx_types_compatible_p (tree x, tree y)
{
  if (same_type_ignoring_top_level_qualifiers_p (x, y))
    return 1;

  /* Once we get to the middle-end, references and pointers are

So maybe the C++ FE is supposed to handle the reference vs. pointer
stuff itself via the second check in tree_ssa_useless_type_conversion_1,
but deleting the (broader) check from tree_ssa_useless_type_conversion_1
makes temp1.C and the sra test fail exactly because this broader check.
(I wonder what happens if we change same_type_p to 
same_type_ignoring_top_level_qualifiers_p in the langhook).

Richard.