improving g++ alias analysis (was: Re: g++ and aliasing bools)

[Dan Nicolaescu <dann at godzilla dot ICS dot UCI dot EDU>]

Now that 3.1 has branched we can go back to this discussion.

Here is a refinement of my initial patch (using the name/structure
proposed by Joe).

It does not deal with inheritance and unions yet.
bootstrapped on sparc-sun-solaris2.8, make check-c++ shows no new
regressions. 

It solves the problem shown here: http://gcc.gnu.org/ml/gcc/2002-01/msg01478.html


static bool
cxx_simple_enough_type (t)
     tree t;
{
  if (AGGREGATE_TYPE_P (t))
    {
      if (TREE_CODE (t) == RECORD_TYPE)
        {
          tree fields;
          /* PODs are safe */
          if (! CLASSTYPE_NON_POD_P(t))
            return true;
          /* classes with virtual baseclasses are not */
          if (TYPE_USES_VIRTUAL_BASECLASSES (t))
            return false;
          /* don't deal with inheritance for now */
          if (CLASSTYPE_N_BASECLASSES(t))
            return false;
          for (fields = TYPE_FIELDS (t); fields; fields = TREE_CHAIN (fields))
            {
              if (TREE_CODE (fields) != FIELD_DECL)
                continue;
              if (! cxx_simple_enough_type (TREE_TYPE (fields)))
                return false;
            }
          return true;
        }
      else if (TREE_CODE (t) == ARRAY_TYPE)
        return cxx_simple_enough_type (TREE_TYPE (t));
      /* FIXME: do something about UNION_TYPE */
      else
        return false;
    }
  else
    return true;
}


static HOST_WIDE_INT
cxx_get_alias_set (t)
     tree t;
{
  if (!cxx_simple_enough_type(t))
    return 0;
  else
    {
      if (AGGREGATE_TYPE_P (t))
        return 0;
    }

  return c_common_get_alias_set (t);
}


Mark, is this code acceptable for 3.2? If so I can submit a patch to
gcc-patches. 

Next in line is adding support for inheritance using Dan Berlin's 
idea: http://gcc.gnu.org/ml/gcc/2002-01/msg01849.html