Re: Adding -l option to cp

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Sat, Jan 22, 2011 at 05:44:19PM -0500, D'Arcy J.M. Cain wrote:
> Index: utils.c
> ===================================================================
> RCS file: /cvsroot/src/bin/cp/utils.c,v
> retrieving revision 1.38
> diff -u -r1.38 utils.c
> --- utils.c   4 Jan 2011 10:35:10 -0000       1.38
> +++ utils.c   22 Jan 2011 22:38:22 -0000
> @@ -145,11 +145,23 @@
>  
>       rval = 0;
>  
> +     /* if hard linking then simply close the open fds, link and return */
> +     if (lflag) {
> +             (void)close(from_fd);
> +             (void)close(to_fd);
> +             (void)unlink(to.p_path);
> +             if (link(entp->fts_path, to.p_path)) {
> +                     warn("%s", to.p_path);
> +                     return (1);
> +             }
> +             return (0);
> +     }
> +     /* NOTREACHED */
> +

(a) The lint comment is rather obviously misplaced.
(b) Why doesn't this fall-through if it can't create the hard-link?
In the current form it should be a form for ln and not for cp...

> @@ -192,6 +204,7 @@
>                                       remainder -= chunk;
>                               } while (remainder > 0);
>  
> +                             /* Some systems don't unmap on close(2). */
>                               if (munmap(p, fsize) < 0) {
>                                       warn("%s", entp->fts_path);
>                                       rval = 1;

This comment is misleading at the very least. mmap is not disestablished
by close(2). Doing it would break quite a few programs, including
dynamic linkage...

Joerg