Re: [XEN PATCH v3] xen/include: avoid using a compiler extension for BUILD_BUG_ON_ZERO.

[Nicola Vetrini <nicola.vetrini@xxxxxxxxxxx>]

On 22/06/23 10:56, Jan Beulich wrote:
> On 22.06.2023 10:24, Nicola Vetrini wrote:
> > --- a/xen/include/xen/lib.h
> > +++ b/xen/include/xen/lib.h
> > @@ -51,9 +51,10 @@
> >      e.g. in a structure initializer (or where-ever else comma expressions
> >      aren't permitted). */
> >   #define BUILD_BUG_ON_ZERO(cond) \
> > -    sizeof(struct { _Static_assert(!(cond), "!(" #cond ")"); })
> > +    (sizeof(struct { char c; _Static_assert(!(cond), "!(" #cond ")"); }) & 0U)
> >   #else
> > -#define BUILD_BUG_ON_ZERO(cond) sizeof(struct { int:-!!(cond); })
> > +#define BUILD_BUG_ON_ZERO(cond) \
> > +    (sizeof(struct { unsigned u : (cond) ? -1 : 1; }) & 0U)
> Given your remark on named bitfields not being allowed to be zero length
> (which hopefully holds universally, i.e. isn't subject to compiler
> extensions), how about
>
> #define BUILD_BUG_ON_ZERO(cond) (sizeof(struct { int _:!(cond); }) & 0)
Well, the reason for the change is to drop any assumption on 
compiler-specific quirks that may arise (it would be a surprise if this 
wasn't the case, though)
> ? (Implicitly, as said before, I question the value of the U suffixes here.
> Even better might be to make sure the expression is explicitly not of
> unsigned type, to avoid surprises if someone used ~BUILD_BUG_ON_ZERO()
> somewhere.)
>
> Jan
The documentation for the macro definition states that the expression 
must have value 0 and type size_t when cond is false. If I understand 
correctly what you're saying here, then this is not allowed by the 
documentation.
Regards,

--
Nicola Vetrini, BSc
Software Engineer, BUGSENG srl (https://bugseng.com)