On Thu, May 28, 2020 at 7:29 AM Florian Weimer <fweimer@redhat.com> wrote:
>
> * H. J. Lu:
>
> > On Thu, May 28, 2020 at 5:58 AM Florian Weimer <fweimer@redhat.com> wrote:
> >>
> >> * H. J. Lu via Libc-alpha:
> >>
> >> > diff --git a/sysdeps/generic/hp-timing.h b/sysdeps/generic/hp-timing.h
> >> > index af9d92f7f7..99b12995f1 100644
> >> > --- a/sysdeps/generic/hp-timing.h
> >> > +++ b/sysdeps/generic/hp-timing.h
> >> > @@ -24,6 +24,11 @@
> >> >  #include <stdint.h>
> >> >  #include <hp-timing-common.h>
> >> >
> >> > +#ifdef _ISOMAC
> >> > +# define __timespec64 timespec
> >> > +# define __clock_gettime64 clock_gettime
> >> > +#endif
> >> > +
> >> >  /* It should not be used for ld.so.  */
> >> >  #define HP_TIMING_INLINE     (0)
> >>
> >> To be honest, I would have expected different definitions of the timing
> >> macros.  This redirection looks a bit iffy for me for _ISOMAC.
> >
> > What do you have in mind?
>
> Something like this:
>
> /* The clock_gettime (CLOCK_MONOTONIC) has unspecified starting time,
>    nano-second accuracy, and for some architectues is implemented as
>    vDSO symbol.  */
> #ifdef _ISOMAC
> # define HP_TIMING_NOW(var) \
> ({                                                              \
>   struct timespec tv;                                           \
>   clock_gettime (CLOCK_MONOTONIC, &tv);                         \
>   (var) = (tv.tv_nsec + UINT64_C(1000000000) * tv.tv_sec);      \
> })
> #else
> # define HP_TIMING_NOW(var) \
> ({                                                              \
>   struct __timespec64 tv;                                       \
>   __clock_gettime64 (CLOCK_MONOTONIC, &tv);                     \
>   (var) = (tv.tv_nsec + UINT64_C(1000000000) * tv.tv_sec);      \
> })
> #endif
>
> in sysdeps/generic/hp-timing.h.
>

Here is the updated patch.  OK for master?

Thanks.

-- 
H.J.