Re: FYI, utimens was missing a dependency on timespec

[Jim Meyering]

Eric Blake <address@hidden> wrote:

> Jim Meyering <jim <at> meyering.net> writes:
>
>>
>> There is a declaration of struct timespec in time_.h,
>> so utimens.h, stat-time.h, and getdate.h should include <time.h>
>> rather than timespec.h.  Then timespec.h can be removed.
>>
>> I'll look into that over the weekend, if no one else does it first.
>
> I took a closer look at it this morning.  What should I do about timespec_cmp
> (), gettime(), and settime(), which currently appear only in timespec.h, but
> are in use by both coreutils and tar?  Until that is resolved, timespec.h (and
> the timespec module) can't be deleted.

Indeed.
I should have looked more carefully before proposing deletion :)

> Should we just move these function
> declarations into time_.h, even though they are not strictly part of time.h's
> namespace?

No.  Otherwise, gnulib's "time.h" replacement becomes mandatory,
and encourages gnulib-using applications to use functions
that are not standard.

> Or do we keep the timespec module, which both coreutils and tar
> explicitly request, as an add-on module not required by any other gnulib 
> module?

So, the timespec module provides something useful and nonstandard,
and we should keep it.

Just make stat-time depend on timespec, after all?