On 26 September 2018 at 16:59, Thomas Huth <th...@redhat.com> wrote: > +/* > + * Clang 3.4 claims to be compatible with GCC 4.2, but does not have the > + * "flatten" attribute, so we've got to handle Clang via __has_attribute here > + */ > +#if defined(__clang__) && defined(__has_attribute) > +# if __has_attribute(flatten) > +# define QEMU_FLATTEN __attribute__((flatten)) > +# endif > +#elif !defined(__clang__) && QEMU_GNUC_PREREQ(4, 1) > +# define QEMU_FLATTEN __attribute__((flatten)) > +#endif > +#ifndef QEMU_FLATTEN > +# define QEMU_FLATTEN > +#endif
I think it would be cleaner to say: if we have __has_attribute, trust it; otherwise fall back to version testing. Also, we already require at least GCC 4.1 so a version test against that is unnecessary. So something like: #ifndef __has_attribute #define __has_attribute(x) 0 / compatibility with older GCC */ #endif /* * gcc doesn't provide __has_attribute() until gcc 5, * but we know all the gcc versions we support have flatten. * clang may not have flatten but always has __has_attribute(). */ #if __has_attribute(flatten) || !defined(__clang__) # define QEMU_FLATTEN __attribute__((flatten)) #else # define QEMU_FLATTEN #endif > + > #ifndef __has_feature > #define __has_feature(x) 0 /* compatibility with non-clang compilers */ > #endif > -- > 1.8.3.1 thanks -- PMM
