[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ubuntu-devel
Subject:    Accepted debootstrap 0.3.3.3ubuntu4 (source)
From:       cjwatson () ubuntu ! com (Colin Watson)
Date:       2007-05-30 16:49:14
Message-ID: 20070530164913.GA17368 () riva ! ucam ! org
[Download RAW message or body]

On Wed, May 30, 2007 at 08:02:19AM +0200, Martin Pitt wrote:
> Colin Watson [2007-05-29 18:43 +0100]:
> > On Sat, May 26, 2007 at 08:20:15PM -0000, Ubuntu Installer wrote:
> > > Changes: 
> > >  debootstrap (0.3.3.3ubuntu4) gutsy; urgency=low
> > >  .
> > >    * {feisty,gutsy}.fakechroot: Remove libdevmapper1 postinst before
> > >      configuring packages. It tries to create /dev/mapper which fails with
> > >      fakechroot (since /dev/ is a symlink to the real /dev).
> > 
> > This feels like the wrong fix. Shouldn't libdevmapper1 be changed
> > instead to succeed if /dev/mapper already exists? (I appreciate that you
> > may not be able to fix feisty at this point.)
> 
> The postinst itself is correct, at least in a normal installation. It
> does that check already.
> 
> The problem is that e. g. on the DC hosts, /dev/mapper does *not*
> exist, and neither does /dev/.static/dev. In a chroot used by
> fakechroot, /dev is a symlink to the real outside /dev, since you
> cannot create any devices as non-root.

Isn't the point of fakeroot/fakechroot to pretend to do things that you
couldn't ordinarily do as root? :-) For instance, it could proxy
operations on the fakechrooted /dev through to an external /dev, if
device nodes with the corresponding major/minor numbers are available.

One thing I'm concerned about is that removing maintainer scripts is
necessarily rather subject to package names changing under you.

-- 
Colin Watson                                       [cjwatson at ubuntu.com]


[prev in list] [next in list] [prev in thread] [next in thread]