Re: update commands / world file pollution Re: [gentoo-dev] Stable sys-devel/gcc USE flag changes WAS: [gentoo-commits] gentoo-x86 commit in eclass: ChangeLog toolchain.eclass

[Michael Weber <xmw@gentoo.org>]

On 01/18/2013 08:36 AM, Benedikt Böhm wrote:
> On Fri, Jan 18, 2013 at 8:27 AM, Michael Weber <xmw@gentoo.org
> <mailto:xmw@gentoo.org>> wrote:
>     I'd like to drop one strong suggestion about configuration management
>     that might be beneficial here: use version control software!
> or even /etc/.git ... it saved my life on numerous occasions

Sure, bit thats's the point were diversity (hostnames, ssh_host_keys)
kicks in (which has been eliminated in mentioned example) and
the repo carries confidential information.
(Well, if somebody places an compromised update in the
 local-overlay, i'd blindly install anything)

I even have / inside git for testing, with excludes on /opt/ /usr
/{s,}/bin /etc/ssl and so on.

It works and is handy to easily add apache config, web-app-config
installed roundcube, layman overlay list, but the maintenance of the
.gitignore raises and hardlink solutions like dirvish make more sense
for being complete backups (LD_LIBRRY_PATH=/backup/.../tree/usr/lib).

> for reference, here is my updateworld script, which also handles python,
> ruby, perl, revdep-rebuild and all that
> crap: https://github.com/zenops/cookbooks/blob/master/cookbooks/portage/files/default/scripts/updateworld
cool.

So basically everyone uses personal `apt-get update` (cvs co, porticron,
emerge+layman, eix-sync) strategies and even more
funny little scripts for `apt-get upgrade` (-avuND world, aliases,
scripts).

I wonder if anybody uses unattended [backup+]emerge as cron job.
I'm really temped to do so, but with users relying on these machines I'm
always chicken-out.

-- 
Michael Weber
Gentoo Developer
web: https://xmw.de/
mailto: Michael Weber <xmw@gentoo.org>