Re: [PATCH v3] Merge parts of <sys/time.h> from FreeBSD

[Marco Atzeri <marco dot atzeri at gmail dot com>]

On 7/6/2015 10:08 AM, Corinna Vinschen wrote:
> Hi Marco,
>
> On Jul  5 16:33, Marco Atzeri wrote:
> > On 3/13/2015 1:24 PM, Corinna Vinschen wrote:
> > > Hi Sebastian,
> > >
> > > On Mar 13 11:00, Sebastian Huber wrote:
> > > > Merge parts of the latest <sys/time.h> from FreeBSD to Newlib.  This has
> > > > two reasons.
> > >
> > > This patch also allowed to remove a Cygwin specific piece of code.
> > > I now pushed your patch as well as the Cygwin ones.
> > >
> > > Thanks,
> > > Corinna
> >
> >
> > with the current settings, on cygwin with _GNU_SOURCE on
> > /usr/include/sys/time.h
> >
> > bintime_mul(struct bintime *_bt, u_int _x)
> > is included but
> >
> > u_int is not defined
>
> Thanks, but you forgot to paste this detail from your mail to the
> Cygwin list:
>
> > u_int is defined in
> >   /usr/include/sys/types.h
> > and
> >   /usr/include/sys/time.h
> > includes it, but it is coveread by a
> >
> >   # ifndef        _POSIX_SOURCE
>
> Also your project defines _POSIX_SOURCE.
>
> So while u_int is guarded !_POSIX_SOURCE, the bintime_mul inline function
> using it is guarded __BSD_VISIBLE.
>
> Sebastian, the easy way out would be to redefine bintime_mul to use
> and "unsigned int" as second parameter.

this is cheating, and not solving the root problem.

> OTOH, shouldn't the affected types in sys/types.h guarded __BSD_VISIBLE,
> too, instead?  They are on FreeBSD, afaics.

I triggered the problem in an unusual way, however
functions and type should be guarded in the same way,
to avoid further lurking issue along the road.


> Thanks,
> Corinna

Ciao
Marco

Marco