[PATCH] Implement DR61: Address of ambiguous bound methods should be disallowed

[David Majnemer]

On Mon, Jun 10, 2013 at 3:51 PM, Richard Smith <richard at metafoo.co.uk>wrote:

> On Mon, Jun 10, 2013 at 3:25 PM, David Majnemer
> <david.majnemer at gmail.com> wrote:
> > On Mon, Jun 10, 2013 at 12:00 PM, Richard Smith <richard at metafoo.co.uk>
> > wrote:
> >>
> >> On Sat, Jun 8, 2013 at 1:14 AM, David Majnemer <
> david.majnemer at gmail.com>
> >> wrote:
> >> > On Fri, Jun 7, 2013 at 10:12 PM, Richard Smith <richard at metafoo.co.uk
> >
> >> > wrote:
> >> >>
> >> >> Please remove the FIXME from dr0xx.cpp.
> >> >
> >> >
> >> > Done.
> >> >
> >> >>
> >> >>
> >> >> I'm not sure the changes to p2-resolve-single-template-id.cpp are
> >> >> right -- we should have resolved the name to a single static member
> >> >> function in those cases. EDG thinks those cases are valid.
> >> >
> >> >
> >> > You are totally correct. I *wrongly* tried to rationalize GCC's
> behavior
> >> > as
> >> > it seemed believable. That's the last time I do that...
> >> >
> >> > I've attached a patch that corrects for this.
> >>
> >> Thanks!
> >>
> >> I think we should actually be fixing up the & operand to point to the
> >> resolved function in CheckAddressOfOperand, rather than just checking
> >> it. Does (for instance) decltype(&a.static_fn) work with the current
> >> approach?
> >
> >
> > Neither my patch nor EDG allows decltype(&a.static_fn).
>
> Sorry, the case I meant was this one:
>
> struct A { template<typename T> static void f(); } a;
> decltype(&a.f<int>) p;
>

Both my patch and EDG allow this.


>
> >> >> On Fri, Jun 7, 2013 at 9:20 PM, David Majnemer
> >> >> <david.majnemer at gmail.com>
> >> >> wrote:
> >> >> > Sorry, I accidentally forgot to mark the test as passing in
> >> >> > dr0xx.cpp.
> >> >> >
> >> >> > An updated patch is attached.
> >> >> >
> >> >> > --
> >> >> > David Majnemer
> >> >> >
> >> >> >
> >> >> > On Fri, Jun 7, 2013 at 9:15 PM, David Majnemer
> >> >> > <david.majnemer at gmail.com>
> >> >> > wrote:
> >> >> >>
> >> >> >> clang currently disallows bound methods from having their address
> >> >> >> taken
> >> >> >> but sometimes allows them if the method is overloaded in some way.
> >> >> >>
> >> >> >> The attached patch implements DR61 [*] which affirms that
> >> >> >> expressions
> >> >> >> containing unresolved member access should be disallowed when
> >> >> >> performing
> >> >> >> "address of" operations.
> >> >> >>
> >> >> >>
> >> >> >> [*]
> http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#61
> >> >> >>
> >> >> >> Thanks
> >> >> >> --
> >> >> >> David Majnemer
> >> >> >
> >> >> >
> >> >
> >> >
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20130610/0214f94d/attachment.html>