Re: [rfc] rint() rounds to even

[Ben Collins <bcollins at debian dot org>]

On Mon, Jan 29, 2001 at 03:18:36PM -0800, Ulrich Drepper wrote:
> Roland McGrath <frob@debian.org> writes:
> 
> > What I meant by my suggestion is to say something rather than nothing, even
> > if to explicitly say "unspecified and the behavior might change in the
> > future".  It is useful to be explicitly made aware that you should not be
> > relying on something that you might not have realized you were depending on
> > before the manual made you think about it.
> 
> If you start doing this half of the document would consist of comments
> like this.  There are more cases not covered by standards than vice
> versa.  To avoid this unnecessary bloat standards are written the way
> I said, declare things not specified as officially unspecified.

But libc docs are not standards, they are documentation on an
implementation. There are plenty of places in the libc manual where it
notes where libc either differentiates from a spec, or implements
certain things in addition to the spec. I don't think adding
clarification on which way GNU libc went is going to hurt, especially
when the GNU libc docs currently do not even say "the implementation of
the rounding is unspecified".

-- 
 -----------=======-=-======-=========-----------=====------------=-=------
/  Ben Collins  --  ...on that fantastic voyage...  --  Debian GNU/Linux   \
`  bcollins@debian.org  --  bcollins@openldap.org  --  bcollins@linux.com  '
 `---=========------=======-------------=-=-----=-===-======-------=--=---'