On Thu, Feb 22, 2018 at 06:30:47AM -0700, Jan Beulich wrote:
> >>> On 22.02.18 at 13:28, <andrew.coop...@citrix.com> wrote:
> > On 22/02/18 12:24, Jan Beulich wrote:
> >>>>> On 22.02.18 at 13:03, <roger....@citrix.com> wrote:
> >>> On FreeBSD the build fails with:
> >>> /root/src/xtf/include/xtf/compiler.h:13:9: error: '__section' macro
> >>> redefined
> >>> [-Werror,-Wmacro-redefined]
> >>> #define __section(s) __attribute__((__section__(s)))
> >>> ^
> >>> /usr/include/sys/cdefs.h:229:9: note: previous definition is here
> >>> #define __section(x) __attribute__((__section__(x)))
> >>> ^
> >>> 1 error generated.
> >>> Only define __section if it's undefined in order to prevent conflicts.
> >> But the actual bug is that /usr/include/sys/cdefs.h is (magically?)
> >> being included. Just like the hypervisor build, the xtf one shouldn't
> >> pull in _any_ external headers.
> > XTF uses the following freestanding headers only.
> > #include <stdint.h>
> > #include <stdbool.h>
> > #include <stdarg.h>
> > #include <stddef.h>
> > #include <inttypes.h>
> > #include <limits.h>
> None of which call for __section() to be defined. Otoh __section()
> is a reserved name, so it could be used for _anything_. I therefore
> think it should be #undef-ed, instead of the #define being made
> conditional. (The comment Roger adds is pretty useless as well,
> considering there's no cdefs.h anywhere in the xtf tree.)
Hm, that's basically in-line with the other comments regarding
__noinline and __always_inline in the same file, that's why I didn't
bother much and just did the same for __section.
Xen-devel mailing list