Re: threads in C++

[Dima Volodin <dvv at dvv dot org>]

Jonathan Larmour wrote:

> Dima Volodin wrote:
> >
> > Jifl wrote:
> > >C++ defines that if a static class member function is declared as e.g.
> > >static void *func(void *), then if you take the address of this function,
> > >it has the required type for pthread_create. This is well-defined behaviour
> > >in C++.
> >
> > Would you please give me the exact reference to where in the Standard
> > this behavior is defined? Or just save your time and see section 9.2.5
> > in _The C++ Programming Language_, Third Edition, by Bjarne Stroustrup.
>
> I've had a look at the standard, and in a sense we're both right ;-).

In what sense? Neither this paragraph, nor any other place in the Standard says
that a member function or a class member function could be of a C linkage. On the
contrary, the paragraph you refer to clearly states that "A C language linkage is
ignored for the names of class members and the member function type of class
member functions."

> There
> is a distinction between the linkage of the member's name and the linkage of
> the member's type. (see sect 7.5 para 4).
>
> So I tried the following example which comes pretty much straight from this
> para. According to the std, this defines a data member such that the name
> has C++ language linkage and the type is pointer to C function:
>
> extern "C" typedef void FUNC_c();
>
> class C {
> public:
>         static FUNC_c* q;
> };
>
> Compiling with 2.95.2, I get:
>
> $ gcc -c flibble.cxx
> flibble.cxx:1: multiple storage classes in declaration of `FUNC_c'
>
> g++ bug?

Yep, looks like that.

> Jifl

Dima