Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 10 May 2009 21:54:42 -0700
From:      Doug Barton <dougb@FreeBSD.org>
To:        Jeremie Le Hen <jeremie@le-hen.org>
Cc:        freebsd-current@FreeBSD.org
Subject:   Re: New mergemaster option -I, failsafe install files
Message-ID:  <4A07AF92.1040309@FreeBSD.org>
In-Reply-To: <20090510165737.GB88857@obiwan.tataz.chchile.org>

index | next in thread | previous in thread | raw e-mail

Jeremie Le Hen wrote:
> Hi Doug,
> 
> As you may guess from my multiple emails, I'm in the process of
> upgrading my jails :-).
> 
> Since I have one jail per service, a very few number of configuration
> files are modified on each jail.  As most of user of FreeBSD I think,
> I'm used to run "mergemaster -iU" to automate the process as much as
> possible.  The problem with service jails (chapter 15.6.1 of the
> handbook) is that / is read-only mounted on all jails, /etc /var /root
> and a few other places being symlinks to /s, the private read-write
> space of each jail.  Thus when mergemaster tries to update
> /boot/devices.hints it fails and abort.

I think the way to solve this problem would be with an
MM_PRE_COMPARE_SCRIPT that deletes /boot/device.hints (and any other
relevant files) from the temproot. If they are not present in that
directory when the comparison starts then it's a non-issue.

> Therefore I've implemented a new -I option that does the same thing as
> -i except that it will proceed on failure. 

ewwww, scary. :)  Seriously though, I have very strong feelings about
not blasting through errors since I have no way of determining which
errors are/should be show stoppers, and which are merely annoying. And
even if I thought I could write code to do that, the real answer would
depend heavily on local policy in any case.


Thanks for thinking about this issue in any case,

Doug


home | help

Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4A07AF92.1040309>