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

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

On 3/19/07, Russell Yanofsky <russ@yanofsky.org> wrote:
> On Sun, 2007-03-11 at 17:24 -0400, Jason Merrill wrote:
> > I think the boolean field is necessary, too, for distinguishing between
> > lvalue and rvalue types.  But it can be internal to the front end,
> > perhaps using TYPE_LANG_FLAG_2.
>
> I can switch to that flag, but the problem is that if you want to use
> _any_ flag to distinguish between REFERENCE_TYPE nodes, you have to
> modify the code which creates REFERENCE_TYPE nodes to check that flag.
> Otherwise, nodes which are supposed to have different flag values will
> wind up occupying the same spot on the TYPE_NEXT_REF_TO chain. There's
> no way to avoid modifying the middle end (even though those
> modifications can be done in a language-neutral way) if we want to keep
> on using REFERENCE_TYPE nodes to represent l-value and r-value
> references in the front end.

What if (a new) cp_build_reference_type keeps its own hash table of
rvalue REFERENCE_TYPE references, but forwards to build_reference_type
for lvalue references?

I still think Jason's trick might just work... the front end only
builds REFERNCE_TYPEs with mode==ptr_mode. So cp_build_reference_type
would always make sure to build the lvalue REFERENCE_TYPE (mode ==
ptr_mode) first, then put the rvalue REFERENCE_TYPE (also mode ==
ptr_mode) later in the list. The common build_reference_type_for_mode
will always find the first (lvalue) reference type, whereas
cp_build_reference_type knows better and will also use the appropriate
TYPE_LANG_FLAG bit.

 Cheers,
 Doug