Re: C++ testcase question: pointers to member functions

[Mark Mitchell <mark at codesourcery dot com>]

Zack Weinberg wrote:
> On Thu, 2005-02-10 at 16:05 -0800, Janis Johnson wrote:
>
> > On Thu, Feb 10, 2005 at 04:00:49PM -0800, Zack Weinberg wrote:
> >
> > > On Thu, 2005-02-10 at 15:47 -0800, Janis Johnson wrote:
> > >
> > > > There are eight tests that check for that error message, although
> > > > probably not in quite the same way.
> > >
> > > I may be confused - this *is* g++.mike/p10769a.C we are talking about,
> > > yes?  If so, I'd like to point out that g++.mike/p10769b.C appears to be
> > > constructively the exact same test case except with dg-error markers on
> > > the offending conversions instead of dg-boguses.
> >
> > Except that p10769b.C doesn't use -Wno-pmf-conversions and gets different
> > error messages.
>
>
> My bad, I should read more carefully.  Um, is -Wno-pmf-conversions a
> useful feature anymore, given that we are issuing errors with or without
> it?

It looks like we do still use that flag in a couple of places, and I 
don't see a deprecation note anywhere.  Perhaps I was mistaken, but I 
distinctly thought we'd killed this off.

It looks like we do still accept:

  struct S {
    void f();
  };

  void g() {
    reinterpret_cast<void*>(&S::f);
  }

But, the form in the test case where we are not even starting with a 
pointer-to-member, but merely the name of a member function.  I think 
that's an intentional tightening; C++ doesn't allow you to do anything 
with the name of a member function, except call it.  And, the tightening 
goes back at least to G++ 3.2, as far as I can tell.  So, I think 
removing the test case is still OK.

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304