Re: [PATCH] time: Skip overflow itimer tests on 32-bit systems

[Aurelien Jarno <aurelien@aurel32.net>]

On 2021-07-06 16:58, Adhemerval Zanella via Libc-alpha wrote:
> > diff --git a/include/time.h b/include/time.h
> > index 4372bfbd96..377a4a45ea 100644
> > --- a/include/time.h
> > +++ b/include/time.h
> > @@ -342,6 +342,14 @@ in_time_t_range (__time64_t t)
> >    return s == t;
> >  }
> >  
> > +/* Check whether T fits in a timeval32 (__int32_t).  */
> > +static inline bool
> > +in_timeval32_range (__time64_t t)
> > +{
> > +  __int32_t s = t;
> > +  return s == t;
> > +}
> > +
> 
> The name is confusing, it is mixing timeval from 'struct timeval' and
> time_t.  And there is no need to use __int32_t, we need to use it only
> on installed headers to avoid namespace pollution. 
> 
> I fact I think it would be better to just change 'in_time_t_range' to
> use int32_t internally instead of time_t; I am pretty sure that all
> usages assume that sizeof(time_t) == 32.

Unfortunately this is not correct. MIPS N64 uses a 64-bit time_t,
however it does not define XSTAT_IS_XSTAT64. Therefore fstatat is
different than fstatat64, and uses the default
sysdeps/unix/sysv/linux/fstatat.c implementation, which uses
in_time_t_range() with a 64-bit time_t.

Therefore this change broke y2038 support in fstat/fstatat/lstat on MIPS
N64, which now returns EOVERFLOW for files which need such support. I
have opened BZ #29730.

-- 
Aurelien Jarno                          GPG: 4096R/1DDD8C9B
aurelien@aurel32.net                 http://www.aurel32.net