Update GCC from 2.95.2: 2.95.3, 3.1.1, or 3.2?

[David A. Bandel]

On Sat, 21 Sep 2002 17:34:20 -0700
begin  "Net Llama!" <netllama at linux-sxs.org> spewed forth:

> Collins wrote:
> > On Fri, 20 Sep 2002 18:18:58 -0700 "Net Llama!"
> > <netllama at linux-sxs.org> wrote:
> >>Tim, i'm serious, don't do this.  chroot'd linux installs are not a
> >>good idea.  Things will kinda work, but over time, it will be a
> >>disaster of processes dying, poor performance, and screwed up
> >>networking.
> >>
> >>If you want to 'try before you buy', use User Mode Linux to do the 
> >>Gentoo install.
> > 
> > 
> > I would be curious to know the basis for this diatribe.  Both LFS and
> > gentoo use this method of install with no problems.  Would you care to
> > elaborate?
> 
> For installation, it should be fine.  For trying to run it, there will 
> be problems.  Not immediately, but long term.  For starters, any daemons
> that you run will have to configured to listen solely on the IP for the 
> chroot'd environment.  This will be a configuration hassle.  But the 
> bigger problems come with trying to run syslog within the chroot'd 
> environment.  Getting it to run properly will be incredibly difficult. 
> Also, you will be using the kernel from the host, which may lead to 
> problems if its not what the chroot'd environment expects (especially in
> the case of gentoo).  Additionally, if you mount proc, there will be 
> weird behavior, as processes running outside the chroot might get axed 
> accidentally (such as if an init script does something like 'killall 
> <whatever').  If you don't mount proc, then an assortment of other 
> weirdness will occur, when some apps look to proc for certain things.
> 

I can emphasize the above with some examples:
I often build the kernel in a chroot environment to avoid other problems
or test complex scripts.  I've found that even with /proc mounted inside
the chroot environment, some scripts, like the iptables patch scripts,
don't work properly.  A `make patch-o-matic` would normally produce a tmp
directory (with a time or process-based directory name), copy files to
this tmp directory, patch them, test them, and apply them to the kernel. 
This patch process fails miserably in a chroot environment, and makes a
mess of the toplevel+1 kernel build tree.

Ciao,

David A. Bandel
-- 
Focus on the dream, not the competition.
		-- Nemesis Racing Team motto