[Resolvconf-devel] lintian cleanup

[Thomas Hood]

Daniel Kahn Gillmor wrote:
>> I don't see anything in
>> /usr/share/doc/debian-policy/upgrading-checklist.txt.gz to suggest that
>> we shouldn't bump the policy.  does anyone else see a reason we
>> shouldn't do that?
>>     

No, we're good for 3.8.1

>> W: resolvconf: script-calls-init-script-directly ./etc/resolvconf/update.d/bind:100
>> W: resolvconf: script-calls-init-script-directly ./etc/resolvconf/update.d/libc:155
>>     
>
> I assume this warning suggests that we should be calling invoke-rc.d
> instead:
>
>   http://lintian.debian.org/tags/script-calls-init-script-directly.html
>
>   http://www.debian.org/doc/debian-policy/ch-opersys.html#s9.3.3.2
>
> any reason this shouldn't be done?
>   

The bind initscript only calls the "reload" method so invoke-rc.d won't
make any difference.

The libc initscript does some careful checking to see whether or not
nscd is running before it restarts nscd.  This is admittedly not completely
kosher but it actually works better than any other approach.

These are not instances of oversight or carelessness.  These are known
transgressions which I think we should not eliminate until someone
provides good practical reasons for making changes.

Keep in mind that the bind hook script is intended to be part of the
bind package and the libc hook script is intended to be part of the
libc package.  (The reasons these hook scripts are not actually part
of the respective packages is political: when I was introducing
resolvconf the maintainers of bind and libc weren't interested in
including these scripts in their packages, yet the scripts are obviously
essential for resolvconf.)  Thus the bind hook script can be considered
to be making an "internal" call to its own initscript and libc to be
making an "internal" call to its own initscript.

>> W: resolvconf: too-long-extended-description-in-templates resolvconf/link-tail-to-original
>> W: resolvconf: possible-debconf-note-abuse config:36 resolvconf/downup-interfaces
>>     
>
> we discussed these about a year ago, and decided to wait until
> "post-lenny", i think:
>   

There have been several discussions about this.  The reason for having
relatively long debconf notes is that resolvconf's maintainer scripts can't
install the package completely under all imaginable circumstances and
the consequences of incomplete installation are relatively grave, namely,
loss of name service until the administrator intervenes to do some
reconfiguration.

I am inclined to keep the debconf notes as they are until someone with
high standing in Debian advises us to do otherwise.
-- 
Thomas