Fifth draft of the Y2038 design document

[Joseph Myers]

On Wed, 22 Feb 2017, Zack Weinberg wrote:

> On Wed, Feb 22, 2017 at 7:51 PM, Joseph Myers <joseph@codesourcery.com> wrote:
> > On Wed, 22 Feb 2017, Zack Weinberg wrote:
> >>
> >> I don't much care why the Austin Group doesn't understand that this is
> >> a bug.  It remains a bug and we should not delay fixing it just
> >> because they don't get it.
> >
> > I don't consider it a bug (on the contrary, I agree with Rich Felker that
> > the x32 choice of a different type is a bug and it would be preferable for
> > the kernel to handle the high part of nanoseconds as padding when coming
> > from a 32-bit process, but in the absence of the kernel doing this glibc
> > should do so).
> 
> Why do you see this as a better option, when it involves additional
> complexity on both ends?

Because it keeps compatibility with code that does e.g.

  printf ("%09ld", something.tv_nsec);

which is both a natural thing to do with tv_nsec, and fully legitimate 
according to current standards.

-- 
Joseph S. Myers
joseph@codesourcery.com