Bug#954306: sysvinit-utils: consider lowering the Priority of sysvinit-utils to important or even optional

Andreas Henriksson andreas at fatal.se
Thu Mar 19 21:40:03 GMT 2020

Package: sysvinit-utils
Version: 2.96-2
Severity: wishlist

Dear Maintainer,

Please consider lowering the 'Priority: required' on the sysvinit-utils
to something lower, eg. important or even optional.

To be able to lower to important without breakage there are likely a
number of issues that has to be resolved first. These has previously
been discussed in the bug report that was opened about making
sysvinit-utils non-Essential (#851747) which has alot of discussion also
related to a potential future priority decrease which I'm now opening
this bug report about.

A summary of the discussion:

* Packages which ships an init scripts not masked by a systemd unit that
  uses /lib/init/init-d-script or /lib/init/vars.sh
  - preferable solution would be to ship a masking native systemd unit
  - Note: watch out for service files wrapping the init script!
    (lintian tag 'systemd-service-file-wraps-init-script')

* Users of /sbin/fstab-decode (if any)

* Users of /bin/pidof
  - consider potentially moving/using the procps pidof implementation.
    See #810018

* Users of killall5 (if any)

Note that (some of the) bug reports that should possibly be marked as
blockers for this has been filed, see:

I think to be able to accomplish lowering to Priority: optional likely
pidof should be moved to procps package (which is Priority: important).
This is the only widely used tool which probably deserves to stay in the
'important' set. (Also using the procps implementation of pidof would
mean we get rid of a pointless difference, as it seems basically every
other major linux distribution is using the procps implementation.)

I think if users of pidof start adding dependencies on sysvinit-utils
and it later gets moved to a different package there will be much
useless churn. A possible workaround for this could be to add a virtual
package 'pidof' that is provided by sysvinit-utils until pidof is
actually moved to procps and then procps provides the virtual pidof
package, while asking users to depend on pidof virtual package. This
might be overkill though.

Someone might have to figure out if entangling the pidof transition with
the Priority lowering is beneficial or if it's better to keep these
issues separate.

Andreas Henriksson

More information about the Debian-init-diversity mailing list