Re: new regression in g++.old-deja/g++.other/decl3.C

[Nathan Sidwell <nathan at codesourcery dot com>]

Richard Henderson wrote:
> This is BINFO related, Nathan, so I assume it's yours.
hm, it looks like my binfo stuff uncovered, gosh, a latent bug.  I need
some help.  here's the two slots of the call stack

#5  0x08100293 in cp_fold_obj_type_ref (ref=0x40287b18, known_type=0x402fb32c) at ../../gcc/gcc/cp/class.c:7911
#6  0x0826194e in fold_stmt (stmt_p=0x402f4554) at ../../gcc/gcc/tree-ssa-ccp.c:2072

and it's this code in fold_stmt

	  /* Check for resolvable OBJ_TYPE_REF.  The only sorts we can resolve
	     here are when we've propagated the address of a decl into the
	     object slot.  */
	  /* ??? Should perhaps do this in fold proper.  However, doing it
	     there requires that we create a new CALL_EXPR, and that requires
	     copying EH region info to the new node.  Easier to just do it
	     here where we can just smash the call operand.  */
	  callee = TREE_OPERAND (rhs, 0);
	  if (TREE_CODE (callee) == OBJ_TYPE_REF
	      && lang_hooks.fold_obj_type_ref
	      && TREE_CODE (OBJ_TYPE_REF_OBJECT (callee)) == ADDR_EXPR
	      && DECL_P (TREE_OPERAND (OBJ_TYPE_REF_OBJECT (callee), 0)))
	    {
	      tree t;

	      t = TREE_TYPE (TREE_OPERAND (OBJ_TYPE_REF_OBJECT (callee), 0));
	      t = lang_hooks.fold_obj_type_ref (callee, t); <-- here
	      if (t)
		{
		  TREE_OPERAND (rhs, 0) = t;
		  changed = true;
		}
	    }

at this point t is an ARRAY_TYPE (of cow_t), and so can't have any BINFO
stuff of course. Is the fault in fold_stmt passing an array type, or in the
lang hook for not dealing with it?

nathan
--
Nathan Sidwell    ::   http://www.codesourcery.com   ::     CodeSourcery LLC
nathan@codesourcery.com    ::     http://www.planetfall.pwp.blueyonder.co.uk

struct cow_t {
  virtual void f()=0;
};


int main()
{
  cow_t cow[2];
  cow[0].f();
  return 0;
}