Subject: Re: should "make update" check for new version first?
To: Alistair Crooks <>
From: Douglas Wade Needham <>
List: tech-pkg
Date: 10/19/2004 14:13:25

As I pointed out some time ago, I have a sandbox method of my own,
which uses union mounts, and from which I use rdist to push things to
the target machines, and have even done upgrades from one version to
another (such as two so far from 1.6.2 to 2.0).  It has even worked in
the past for my firewall while handling traffic going from 1.5 to 1.6.
The only outage besides the reboot was a few minutes to realize I
forgot to turn forwarding on in the new install. 8) Anyways, you can
see the scripts via cvsweb at the URL: may want to hold off for a couple of hours after I send this
message before you try to see it.  I have an rdist going to my web
server to upgrade it to 2.0RC4, so things are a bit slow and possibly
not 100% stable. ;)

- Doug

Quoting Alistair Crooks (
> On Mon, Oct 18, 2004 at 03:50:29PM +0100, Dick Davies wrote:
> > * Ronald van der Pol <> [1047 15:47]:
> > 
> > [snip discussion - though a big 'me too' on the 'build new then remove old'
> > issue ]
> >  
> > > I have heard horror stories about that one too.
> > 
> > It bites your hand off in a few edge cases, but it's still a lot
> > better than make update.
> > 
> > I'm not trolling, just try them both and see.
> Another viewpoint is to set up a sandbox, and to build all the
> packages you want in that sandbox.  Then, when you know that
> everything has been built to your satisfaction, you can cut over your
> trees.
> There are a number of ways to do builds in sandboxes - pkg_comp, the
> bulk build mechanism, mksandbox, libkver and other ways also spring to
> mind.
> Regards,
> Alistair

Douglas Wade Needham - KA8ZRT        UN*X Consultant & UW/BSD kernel programmer
Email:  cinnion @ ka8zrt . com
Disclaimer: My opinions are my own.  Since I don't want them, why
            should my employer, or anybody else for that matter!