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

[Jason Merrill <jason at redhat dot com>]

Russell Yanofsky wrote:
> On Fri, 2007-03-09 at 16:26 -0500, Jason Merrill wrote:
> > The test in compare_ics seems to check for cases where one conversion 
> > has matching rvalueness and the other doesn't, but that isn't all the 
> > rule in the working paper talks about: it also requires that for one ICS 
> > to be better than the other, the two references must differ in 
> > rvalueness, one converting to lvalue ref and the other to rvalue ref or 
> > vice versa.
>
> I don't understand what the difference is. Can you give an example of a
> case compare_ics might not handle correctly?

A case where one ICS binds an rvalue reference to an rvalue and the 
other binds an rvalue reference to an lvalue.  I believe this example is 
ambiguous under the rules in the standard, but would be resolved to the 
first f with your patch.

struct A {};
struct B {};
struct C
{
 operator A();
 operator B&();
};

C c;

void f (A&&);
void f (B&&);

int main()
{
  f (c);
}

> I'll go ahead and do this. Rvalue references have also been implemented
> in the Metrowerks Codewarrior compiler, so it might be worth looking at
> how that compiler does the mangling also.

Definitely, but does CodeWarrior implement the ia64 C++ ABI?

> In the meantime do you want me
> to change the patch? Could using a new character code interfere with
> debugging tools? I don't even know what character code would make sense
> to use for rvalue references..

No, leave it as it is until we have an official mangling.

> > About the deduce_ref stuff: why not just add the REFERENCE_TYPE to arg 
> > in type_unification_real rather than add another argument to unify?
>
> I think you were looking at the Feb 21 patch instead of the Mar 9 patch.

Yes, I was, I'll look over the latest patch shortly.

Jason