[tree-ssa] Gimplifying Java

[Daniel Berlin]

>     a = a;   <--- how the hell is this true?

*actually*, this *does* occur.

I know because i short-circuited it just a week or so ago in the alias 
code so that we don't add pointless constraints (uncommitted yet), 
after noticing it actually occurring while debugging (ie lhs and rhs 
trees were the same tree, not just going by printout):

/* Inference for simple assignment (lhs = rhs) */
static void
andersen_simple_assign (ops, lhs, rhs)
      struct tree_alias_ops *ops ATTRIBUTE_UNUSED;
      alias_typevar lhs;
      alias_typevar rhs;
{
   if (dump_file && (dump_flags & TDF_DETAILS))
     fprintf (dump_file, "Simple assignment %s = %s\n",
              alias_get_name (ALIAS_TVAR_DECL (lhs)),
              alias_get_name (ALIAS_TVAR_DECL (rhs)));
   if (lhs == rhs)
     return;
^^^^^^^^^^^^^^^^^^^^^^