Date: Wed, 1 Oct 2003 18:03:07 -0700 From: Kris Kennaway <kris@obsecurity.org> To: Pat Lashley <patl+freebsd@volant.org> Cc: freebsd-questions@freebsd.org Subject: Re: tar vs cp Message-ID: <20031002010307.GA15545@rot13.obsecurity.org> In-Reply-To: <2156421632.1065047127@mccaffrey.phoenix.volant.org> References: <20031001122603.B71418-100000@floyd.gnulife.org> <3F7B0D5C.7080009@mac.com> <2156421632.1065047127@mccaffrey.phoenix.volant.org>
next in thread | previous in thread | raw e-mail | index | archive | help
[-- Attachment #1 --] On Wed, Oct 01, 2003 at 03:25:27PM -0700, Pat Lashley wrote: > --On Wednesday, October 01, 2003 13:22:36 -0400 Chuck Swiger > <cswiger@mac.com> wrote: > > >Jamie wrote: > >[ ... ] > >> I don't know what the actual rationale is for this. Can anyone > >> explain why it is oftentimes better to tar something rather than > >>using cp when copying directories and their contents? > > > >tar handles symbolic links properly, whereas cp will "copy through" the > >contents of the link. > > Another technique is 'cd /source ; find . -print | cpio -pdmv /dest'. > > But none of the built in tools seem to preserve links, flags, and > sparseness. If you want as close to a true copy as possible, check > out the cpdup port. using tar | tar instead of cp -r is usually faster because it makes more efficient use of disk I/O, because reads and writes are queued up at the same time, from the two processes) whereas cp -r reads and writes chunks sequentially (it's actually implemented using mmap'ed memory, which gains some efficiency, but it's still a sequential process because there's only one single-threaded cp running). Kris [-- Attachment #2 --] -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.3 (FreeBSD) iD8DBQE/e3lLWry0BWjoQKURAiWTAKD2tj7NlWnSXpbPQ9+fTgXt4iKNUwCcDoBz VebHIYt94KCXXdqBd9za6iM= =UXjW -----END PGP SIGNATURE-----
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20031002010307.GA15545>