planned failure

Matthias Benkmann matthias at
Wed Feb 27 04:12:17 PST 2002

On Tue, 26 Feb 2002 22:43:41 -0700 Brendan Taylor
<craftian at> wrote:

> On Sun, Feb 24, 2002 at 09:42:24PM -0800, Jesse Tie-Ten-Quee wrote:
> > Why not?  sure it matters.  Just because it's automated doesn't mean
> > it wont happen.  
> That's the beauty of an automated system - you can work the majority of
> the bugs out before you run it, and you can't make typos as it installs.

If you tell me you have squashed all bugs in the building system, that
doesn't mean it is true. I'm sure you did the best you could but I would
still prefer if the system used root privileges as little as possible.
There is a reason why most setuid root programs drop root privileges as
soon as they've finished doing the stuff that can only be done as root. A
good developer never trusts himself, because he knows he makes mistakes. 
> > It really is not a good idea to compile software as root.
> > Everybody does it, that doesn't make it right.
> At this time though, there isn't any other option for ALFS... unless you
> want to have a seperate profile for each package, su-ing between users
> for the pre-build/post-build.

Where's the problem with that? My personal automated building system does
just that (it deals with package users). It is invoked as root and
executes commands with

su <user> -c command

A building system where the controlling program is written in C can just
use the proper set*id() calls. I see absolutely no problems with this
approach, especially since a normal LFS build only has to distinguish
between chapter 5 user and chapter 6 user.


Depression is merely anger without enthusiasm.

Unsubscribe: send email to listar at
and put 'unsubscribe alfs-discuss' in the subject header of the message

More information about the alfs-discuss mailing list