Hi,
Just a little nitpick below...
On Tue, Dec 7, 2021 at 7:56 AM Alejandro Colomar (man-pages)
<alx.manpages@xxxxxxxxx> wrote:
>
> Hi наб,
>
> On 12/6/21 21:12, наб wrote:
> > diff --git a/man7/system_data_types.7 b/man7/system_data_types.7
> > index 1e6a3f74c..80679b180 100644
> > --- a/man7/system_data_types.7
> > +++ b/man7/system_data_types.7
> > @@ -1544,7 +1544,11 @@ or
> > .EX
> > struct timespec {
> > time_t tv_sec; /* Seconds */
> > +#if !(__x86_64__ && __ILP32__ /* == x32 */)
> > long tv_nsec; /* Nanoseconds [0 .. 999999999] */
> > +#else
> > + long long tv_nsec;
> > +#endif
>
> I prefer showing there the (simple) POSIX definition, and then in Notes
> add a message that glibc uses a different type on most systems.
> Otherwise, we're loosing the information that non-glibc systems use long.
>
> How about something like the following?:
>
> Notes: glibc uses long long instead of long for tv_nsec in some
> architectures:
>
> #if !(__x86_64__ && __ILP32__ /* == x32 */)
> long tv_nsec;
> #else
> long long tv_nsec;
> #endif
The #if condition is actually "we're not on x32", so maybe the comment
could say /* != x32 */? Otherwise it looks like the comment
contradicts the code; I know the comment is inside the paranthesis,
and the negation ('!') is outside, but the negation might be easy to
overlook.
Or maybe leave the x32 field size for a note? (E.g. something like
"This is true on almost all arches, but see NOTES", and then under
NOTES specify that "actually, on x32 tv_nsec is long long"). The #if
in the definition doesn't *exactly* match the code anyway.
Thanks,
Stefan.
>
>
> BTW, I don't understand why long long, if long can perfectly represent 0
> .. 999999999. If you know the rationale for such extension, it would be
> a great addition to Notes too.
>
> Thanks!
>
> Alex
>
> > };
> > .EE
> > .PP
> >
>
>
> --
> Alejandro Colomar
> Linux man-pages comaintainer; https://www.kernel.org/doc/man-pages/
> http://www.alejandro-colomar.es/
