Re: Bug#137321: [Bug#137321] executing of conffiles (init scripts) in postinst
Eduard,
We disagree with the severity of this bug. Vorlon and I, maintainers of
the Samba packages, have already told you why we think the bug should
be 'wishlist'. I personally thing that changing the severity back to
what you originally assigned is a very unpolite thing to do (given the
circumstances.) You are not respecting our authority and discretion
as maintainers of the package.
I checked section E of the policy and I could not find anything about
"changing the permissions or relying on their state". Where exactly is it?
I won't waste time changing back the severity to what we think it should
be. In my personal opinion users have to much power when it comes to
filing bugs - they can hold a release just because something doesn't
work just for them, even when they have a corner case configuration
and everything works for everybody else, users have more discretion
and authority than the package maintainers, etc. But this is part of
another story...
The problem is being dealt with and will be fixed in our next release.
I recommend you use rcconf or, as Anthony said, put an 'exit 0' at the
top of the script if you want to disable a service.
Cheers!
Eloy.-
On Fri, Mar 08, 2002 at 11:49:04PM +0100, Eduard Bloch wrote:
> severity 137321 important
> thanks
>
> > Um... no. It is an *init script*, and while it is a conffile, any
>
> The init scripts are marked as conffiles, and this is good so. Many
> packages even depend on manual configuration IN the init scripts.
>
> > change that renders this file something other than a valid init script
> > is unsupported. Just as changing a conffile that is sourced as a shell
>
> But you do not have to use conffiles as part of your postinst execution
> without checking them. You do not even check whether the file still does
> exist.
>
> > script and making it invalid POSIX sh is unsupported, and changing a
> > conffile so that it's unusable by a program that reads it is
>
> If I remove the execute bit from the script, I DO NOT WANT to make any
> upgrade to execute it. As said, you just assume that everyone has to use
> if she installed the package. Stop thinking so.
>
> > I agree that this would be a nice feature, but it's only wishlist; there
>
> No. It breaks the upgrade, so it is at least important.
>
> > are many other maintainer scripts I find on my system that don't honor
> > this convention, either.
>
> :/var/lib/dpkg/info> ls | xargs grep 'init.d/.*start' | wc -l
> 96
>
> Not good. Changing the executable bit of a conffile is one of the good
> rights of the local administrator, and changing the permissions or
> relying on their state is violation of Policy,E.2.
>
> > > Setting up samba (2.2.3a-2) ...
> > > /var/lib/dpkg/info/samba.postinst: /etc/init.d/samba: /bin/sh: bad interpreter: Permission denied
> > > dpkg: error processing samba (--configure):
> > > subprocess post-installation script returned error exit status 126
> > > dpkg: dependency problems prevent configuration of swat:
> > > swat depends on samba (= 2.2.3a-2); however:
> > > Package samba is not configured yet.
> > > dpkg: error processing swat (--configure):
> > > dependency problems - leaving unconfigured
> > > Errors were encountered while processing:
> > > samba
> > > swat
> > > E: Sub-process /usr/bin/dpkg returned an error code (1)
>
> Gruss/Regards,
> Eduard.
> --
> #include <welt_ist_doch_ein_dorf.h>
>
Reply to: