Re: [PATCH] rvalue reference implementation for C++0x

["Doug Gregor" <doug dot gregor at gmail dot com>]

On 4/25/07, Russell Yanofsky <russ@yanofsky.org> wrote:
> This is an update of a patch first posted two months ago:
>
> http://gcc.gnu.org/ml/gcc-patches/2007-02/msg01760.html
>
> There was some discussion about it which ended here:
>
> http://gcc.gnu.org/ml/gcc-patches/2007-03/msg01283.html
>
> The updated patch implements rvalue references entirely in the C++ front
> end instead of modifying middle end code.

Thanks, Russell!

> +/* Return a reference type node referring to TO_TYPE.  If RVAL is
> +   true, return an rvalue reference type, otherwise return an lvalue
> +   reference type.  If a type node exists, reuse it, otherwise create
> +   a new one.  */
> +tree
> +cp_build_reference_type (tree to_type, bool rval)
> +{
> +  tree lvalue_ref, t;
> +  lvalue_ref = build_reference_type (to_type);
> +  if (!rval)
> +    return lvalue_ref;
> +
> +  /* This code to create rvalue reference types is based on and tied
> +     to the code creating lvalue reference types in the middle-end
> +     functions build_reference_type_for_mode and build_reference_type.
> +
> +     It works by putting the rvalue reference type nodes after the
> +     lvalue reference nodes in the TYPE_NEXT_REF_TO linked list, so
> +     they will effectively be ignored by the middle end.  */
> +
> +  for (t = lvalue_ref; (t = TYPE_NEXT_REF_TO (t)); )
> +    if (TYPE_REF_IS_RVALUE (t))
> +      return t;
> +
> +  t = make_node (REFERENCE_TYPE);
> +
> +  TREE_TYPE (t) = to_type;
> +  TYPE_MODE (t) = ptr_mode;
> +  TYPE_REF_CAN_ALIAS_ALL (t) = false;
> +  TYPE_REF_IS_RVALUE (t) = true;
> +  TYPE_NEXT_REF_TO (t) = TYPE_NEXT_REF_TO (lvalue_ref);
> +  TYPE_NEXT_REF_TO (lvalue_ref) = t;
> +
> +  if (TYPE_STRUCTURAL_EQUALITY_P (to_type))
> +    SET_TYPE_STRUCTURAL_EQUALITY (t);
> +  else if (TYPE_CANONICAL (to_type) != to_type)
> +    TYPE_CANONICAL (t)
> +      = cp_build_reference_type (TYPE_CANONICAL (to_type), rval);
> +
> +  layout_type (t);
> +
> +  return t;
> +
> +}

Yes, this is definitely the right approach to the rvalue reference
type... and you didn't have to touch the middle-end at all. Great!

Russell, when I applied this patch to the cxx0x-branch, I ran into
three failing test cases that I'd like you to take a look at. They
behave differently under -std=c++0x than under -std=c++98, and I'm not
sure if they're exploiting C++98-specific behavior or whether they are
exposing a bug in the rvalue references implementation. The test cases
are:

 g++.dg/init/copy7.C
 g++.dg/overload/arg1.C
 g++.dg/overload/arg4.C

Would you take a look at these?

 - Doug