Date: Wed, 11 Jul 2007 15:41:34 -0700 From: Garrett Cooper <youshi10@u.washington.edu> To: Doug Barton <dougb@FreeBSD.org> Cc: Pieter de Goeje <pieter@degoeje.nl>, Gaspar Chilingarov <casper@web.am>, freebsd-current@freebsd.org Subject: Re: RFT: bin/106642: [patch] Allow excluding certain files from mergemaster (8) Message-ID: <46955C9E.6090101@u.washington.edu> In-Reply-To: <46951CC2.5050602@FreeBSD.org> References: <20070419133550.GA65054@tirith.brixandersen.dk> <20070419200929.GA70735@tirith.brixandersen.dk> <4693BC3E.1050605@web.am> <200707110115.42139.pieter@degoeje.nl> <469417CB.7010705@FreeBSD.org> <4694AAF1.6050302@web.am> <46951CC2.5050602@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Doug Barton wrote:
> Gaspar Chilingarov wrote:
>
>
>> Well another question -- is it possible teach somehow mergemaster, that
>> it should never touch passwd, groups, /etc/ttys or bunch of another files?
>>
>
> Sure, you just do the opposite of what I did in the last script. If
> the file doesn't exist in the temproot when it's time to start
> comparing, mergemaster will never do anything with it. So here is some
> more out of the script that I actually use:
>
> #!/bin/sh
> # NOTE: No PATH needed, because mm's PATH is already draconian enough
> case "${PRE_WORLD}" in
> '') rm -f /etc/defaults/pccard.conf
> rm -f ${TEMPROOT}/etc/motd
> rm -f ${TEMPROOT}/etc/printcap
> rm -f ${TEMPROOT}/etc/namedb/named.*
> ;;
> esac
>
>
What if a user or group was added to /etc/group or /etc/passwd that was
required for the updated version of the OS? I would see this behavior a
bit counterproductive in this case.
>> About pre-script which you suggested, there is another problem -- I MAY
>> have new scripts in /etc/rc.d, but NEVER change the system's scripts.
>>
>
> Simple answer, don't do that. Put your scripts in /usr/local/etc/rc.d
> instead, and as of 6.1 and newer they will be included in the base
> rcorder. Therefore there will be no difference in when they are run.
> There is a default assumption that stuff in /etc/rc.d is the systems,
> and the systems only. Now that local scripts (including ports) are in
> the base rcorder, there is no reason not to do it that way.
>
Agreed. Stuff should stay out of /etc/rc.d. Why not have mergemaster
overwrite those scripts by default?
>> In other hand running rm /etc/rc.d/* before mergermaster's start can
>> render system unusable, if mergemaster will fail for some reason.
>>
>
> That's a good reason to do the update in single user mode, and always
> keep good backups. :) Please understand that I am not advocating that
> anyone actually DO what I suggested, I'm just pointing out that there
> are alternatives that already exist that can accomplish what people
> want to do here.
>
> Having given it a little more thought, it would probably be safer to
> do something like:
> case "${PRE_WORLD}" in
> '') mv ${TEMPROOT}/etc/rc.d/* /etc/rc.d/
> ;;
> esac
>
> But my basic point is the same. I added MM_PRE_COMPARE_SCRIPT and
> MM_EXIT_SCRIPT to give users MORE flexibility than adding more (and
> more complex) CLI options would do. Please take advantage of it. :)
>
> Doug
Agreed, once again.
-Garrett
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?46955C9E.6090101>