$Id$ tags aren't required to have a file in CVS, just an extra. I prefer
to get specific with my keywords and use $File$, $Author$, $Revision$ on
separate lines I was intending to only add RCS keywords to files that I
edit, the first time i edit them. If I was going to do LFS I'd start
with all config files in source control. Worst case, if you screw
something up, you can use RCS from a boot disk to manually restore the
file to any revision you have made, including factory default. I see the
PITA involved in the extra steps to tweaking a config:
$ cd /etc/mail
$ cvs update
$ vi sendmail.cf #yes VI, MUWAHHHHHhahahahahah
$ cvs commit
But the advantages (to me) outweigh the pain.
--Christopher
On Wed, 2001-12-12 at 04:38, Jacob Meuser wrote:
> On Tue, Dec 11, 2001 at 07:51:39PM -0800, Christopher Maujean wrote:
> > I was wondering how many people kept their /etc under CVS control. I
> > haven't found any cons to it so far, but it's not mentioned in any of my
> > "sysadmin" books. Are there reasons (besides not wanting to take the
> > extra steps to change a config file, and the potential security hole
> > generated by not properly configuring the CVS pserver) not to?
>
> I don't keep /etc under CVS, but it does sound like a good idea ...
>
> Except ...
>
> It could be a PITA to put $Id$ tags in every file, and make sure they
> don't affect the workings of the file.
>
> (most) package managers (OpenBSD has a policy that forbids fiddling with
> /etc, instead, new versions of config files are installed to
> $PREFIX/share/examples/$name_of_pkg) keep track of files in /etc and
> modifying them could lead to more PITAes at update/upgrade time
>
> But, how many config files do you really need to keep tabs on? It might
> be easier to just write a sh or perl script, or Makefile for that matter,
> to copy the files you want to save to a safe place and then maybe even
> automatically produce diffs of your version vs the default version?
>
> I could like something like that ...
>
> /ec/Makefile.lo-tek
> *-----------------*
> DIR= /etc
>
> FILES= rc.conf.local rc.local rc.shutdown mail/sendmail.cf mail/aliases \
> hosts group passwd fstab sshd_config
>
> XDIRS= mail
>
> SAVE_DIR= /var/save/root
>
> DIFF_DIR= /var/save/diffs
>
> INSTALL_DATA= /usr/bin/install -c -m 444
>
> INSTALL_DIR= /usr/bin/install -d
>
> save-dir:
> test -d ${SAVE_DIR} || ${INSTALL_DIR} ${SAVE_DIR}
>
> xdirs:
> .for i in ${XDIRS}
> ${INSTALL_DIR} ${SAVE_DIR}/${i}
> .endfor
>
> save: save-dir xdirs
> .for i in ${FILES}
> ${INSTALL_DATA} ${DIR}/${i} ${SAVE_DIR}/${i}
> .endfor
>
> update: save
> echo "garbage" >> /etc/hosts
> echo "Eugene\nUnix\nand\nGNU\n/Linux\nUsers\nGroup\nwas\nhere!" \
> >> /etc/mail/sendmail.cf
> # as "for example"s, more interesting would be
> # cd / && tar zxpvf /pub/EUGLIX/etc01.tgz
> # apt-get update && apt-get upgrade
> # but then, these .for loops don't work
> # with GNU make (wouldn't be hard to "port"
> # though)
>
> diff-dir:
> ${INSTALL_DIR} ${DIFF_DIR}
>
> diff: update diff-dir
> .for i in ${FILES}
> cmp -s ${DIR}/${i} ${SAVE_DIR}/${i} || \
> diff -u ${DIR}/${i} ${SAVE_DIR}/${i} > \
> ${DIFF_DIR}/${i:S/\//-/g}.diff || /usr/bin/true
> .endfor
>
> recover: diff
> .for i in ${FILES}
> test -f ${DIFF_DIR}/${i:S/\//-/g}.diff && \
> patch -p 0 < ${DIFF_DIR}/${i:S/\//-/g}.diff || \
> /usr/bin/true
> .endfor
> *------------------*
>
> --
> <[EMAIL PROTECTED]>
> <[EMAIL PROTECTED]>
> <[EMAIL PROTECTED]>
--
Christopher Maujean
IT Director, Premierelink Communications
[EMAIL PROTECTED]
http://www.premierelink.com/
541-344-8575x305
PGP:
---------------------------------------------------------------
http://www.keyserver.net/
KeyID: EFAF4176
Fingerprint: 55E6 4DE1 D7D3 361E F265 C094 46F2 7B62 EFAF 4176
---------------------------------------------------------------
