This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Don't document fclean
On Mon, 27 Feb 2012, Roland McGrath wrote:
> I don't really have any reliable recollection about this. But
> the manual text makes the point of how it differs from fflush.
> This was never implemented anywhere, but was one of those cases
> from the early days of designing GNU systems where we thought the
> usual behavior was especially unhelpful to applications and
> figured we'd do something nicer. Here the issue is that fflush
> on an input stream drops data that was read from the system but
> never seen by the application, so it's lost entirely. The claim
> for fclean was that it would push the unread data back into the
> system buffer so that the next application to read from the same
> descriptor would see it. But even for the Hurd, we never
> implemented the underlying facility to make that possible.
Thanks for the history. I don't think this changes the basic argument for
my patch (that the manual should document facilities that are actually
implemented, and if someone implements fclean in future then the
documentation can be added back at that point).
--
Joseph S. Myers
joseph@codesourcery.com