Subject: Re: pipe(2) and invalid fildes
To: =?ISO-8859-1?Q?Jarom=EDr?= Dolecek <jdolecek@netbsd.org>
From: Charles M. Hannum <abuse@spamalicious.com>
List: tech-userlevel
Date: 10/01/2001 07:37:57
On Mon, 2001-10-01 at 06:28, Jarom=EDr Dolecek wrote:
> Charles M. Hannum wrote:
> > > Well, EFAULT can't be returned by pipe(2) currently. So it seems
> > > wrong to document it as possible return value in manpage.
> >=20
> > The key word here is "currently".  It may in the future.
>=20
> I doubt it would. However, if it would, the manpage would be changed
> accordingly. Why should we document EFAULT in pipe(2), and not
> other errors never returned by pipe(2), like e.g.  ENXIO, ENOTDIR
> or EACCES there?
>=20
> The manpage states that EFAULT is returned by pipe(2) if fildes
> buffer is in an invalid area of the process's address space.  That
> is simply not true, and thus I think that sentense should be removed.

Let me translate:

"Rather than having a static interface definition independent of the
implementation, we could have a floating definition dependent on the
current implementation."

I fail to see how this is any friendlier.  In fact, it sounds downright
hostile to me.