Re: Compile errors with gcc 4.8

[Keir Fraser <keir.xen@gmail.com>]

On 07/02/2013 09:57, "Jan Beulich" <JBeulich@suse.com> wrote:

> Actually, I'm of the opposite opinion - it should be in other places
> too that handles don't get needlessly defined by the public
> headers. They should get defined only when there actually is a
> use for them. Everything else can be defined where actually
> consumed, as in this case: For one, a handle of a compat_*
> type can't be defined in a public header anyway, as the compat
> types don't get defined there. And then, as you say, the oddity
> of this type makes it desirable to scope restrict it as much as
> possible.
> 
> Now, for the actual solution, I'd prefer the -Wno-... Option

+1. It's not a compiler warning that I care about.

 -- Keir

> suggested above, or as a second best choice generalizing the
> solution suggested by Michael, applying the attribute in
> DEFINE_XEN_GUEST_HANDLE() itself. This is the more that
> attaching it in to the use of the macro just happens to work,
> but isn't guaranteed to in the future: Switching around the
> order of the two lines of the expansion
> 
> #define __DEFINE_XEN_GUEST_HANDLE(name, type) \
>     ___DEFINE_XEN_GUEST_HANDLE(name, type);   \
>     ___DEFINE_XEN_GUEST_HANDLE(const_##name, const type)
> 
> or adding a third (say, volatile) one would re-expose the
> problem.
> 
> Jan