Re: coreutils patch to port to hosts lacking fsync

[Jim Meyering]

Paul Eggert <address@hidden> wrote:
> "Matthew Woehlke" <address@hidden> writes:
>> "shred" is sort-of in the same boat due to lacking fsync()
>
> I don't recall that problem but it is easy enough to fix.  Here's
> a patch.
>
> 2006-12-19  Paul Eggert  <address@hidden>
>
>       * m4/jm-macros.m4 (coreutils_MACROS): Check for fsync.
>       * src/dd.c (fsync) [! HAVE_FSYNC]: Define.
>       (dd_copy): Fall back on sync if fsync isn't available.
>       * src/shred.c (dosync) [! HAVE_FSYNC]: Likewise.
>       * src/dd.c (dd_copy): Check for EINTR from fdatasync, for
>       consistency with fsync.

Thanks, but I'd prefer to skip this change, unless there is a
reasonably modern portability target that lacks the fsync function.
I think of this as an opportunity to avoid adding infrastructure
(and #ifdefs) that may never be used :-)

We know how hard it is to remove such bits once they've been added.