On Mon, Apr 20, 2015 at 12:46:20PM -0600, Theo de Raadt wrote:
> > On Thu, Apr 16, 2015 at 09:19:13AM -0400, William Orr wrote:
> > > This documents the error code when passing multiple cmsg structs. Let me
> > > know if the wording needs to be improved.
> > > 
> > > Index: lib/libc/sys/send.2
> > > ===================================================================
> > > RCS file: /cvs/src/lib/libc/sys/send.2,v
> > > retrieving revision 1.31
> > > diff -u -p -r1.31 send.2
> > > --- lib/libc/sys/send.2   9 Sep 2014 06:32:37 -0000       1.31
> > > +++ lib/libc/sys/send.2   16 Apr 2015 12:48:32 -0000
> > > @@ -223,6 +223,17 @@ values in the
> > >  .Fa msg_iov
> > >  array overflowed an
> > >  .Em ssize_t .
> > > +.It Bq Er EINVAL
> > > +The socket
> > > +.Fa s
> > > +is a
> > > +.Xr unix 4
> > > +socket, and
> > > +.Em controlmsg
> > > +is an invalid size or multiple
> > > +.Em controlmsg
> > > +structures were passed as part of
> > > +.Fa msg .
> > >  .It Bq Er EMSGSIZE
> > >  The
> > >  .Fa msg_iovlen
> > > 
> > 
> > anyone want to take this? it's fine by me, if someone can verify its
> > accuracy.
> 
> It seems we cannot exhaustively document the scenarios that lead to
> certain errnos being returned (in particular, EAGAIN and EINVAL).
> 
> Overdocumentation leads to reader fatigue; reader fatigue leads to
> the documentation being ignored.
> 
> I think it is clearly understood that controlmsg layout is complex,
> and there are many illegal layouts, even non-portable ones at that.
> This is a complex space, under-documented perhaps?  In any case all
> those details are missing from this manual page.  So it seems
> completely wrong to document just this one narrow detail.
> 
> So I don't think the proposed diff improves the manual page.
> 

fine, i'll drop it.
jmc

Reply via email to