Bug#934463: initscripts: consider taking over hwclock policy machinery

Dmitry Bogatov KAction at debian.org
Mon Aug 12 20:31:03 BST 2019

control: tags -1 +confirmed +help

[2019-08-11 12:46] Andreas Henriksson <andreas at fatal.se>
> Package: initscripts
> Version: 2.95-4
> Severity: wishlist
> Dear Maintainer,
> Please consider taking over the hwclock related init script (et.al.) from the
> util-linux package.
> These init scripts aren't really related to util-linux in any way, they
> are just the old (sysvinit) *policy* on how to sync the RTC.
> Unrelated to init systems other inits have chosen to push for another
> policy (since syncing system clock to hardware clock has been deemed to
> not be a good solution). This means several packages currently carry
> overrides to disable the policy shipped in util-linux when not using
> sysvinit. Basically, the situation is unnessecarily complex and things
> would be much simpler if the sysvinit policy lived in src:sysvinit.
> As a bonus this would also mean it's easy for sysvinit maintainers to
> work on any potential changes / modernizations of their RTC policy.
> (This would also bring the util-linux package closer to being a
> "mechanism, not policy" package as I think it should only provide tools
> and leave it up to others to decide how those tools are used.)
> I would thus propose that the src:sysvinit package picks up the policy
> related bits and could for example ship them in the initscripts package.
> This includes:
> * /etc/init.d/hwclock.sh
> * /etc/default/hwclock
> * the /lib/udev/hwclock-set script and the related
>   udev rule file /lib/udev/rules.d/85-hwclock.rules.

Sounds reasonable.

> BEWARE! That the files under /etc are considered 'conffiles' by dpkg
> and need special attention to be correctly handled (in a policy
> compliant way). Please use extra caution and make sure changes are
> thoroughly reviewed before uploading. (The src:sysvinit didn't have any
> conffiles in the package last I looked, but rather used a home-brew
> system to handle it without involving dpkg which is why I'm pointing
> this out for people who don't have previous experience with how many
> gotchas there can be related to conffiles.)

bin:initscripts uses dpkg's notion of conffiles, like all other
packages. Nothing special.

> I think a change such as this should be worked on early in a release
> cycle which is why I'm filing this bug report now. If work to implement
> this happens very soon I will offer to lend a hand as a reviewer and
> could potentially also handle the util-linux side of things (as moving
> the files will likely need a coordinated upload of both src:sysvinit and
> src:util-linux).

Yes. Help is definitely needed. I guess, order of actions is following:

 * you upload util-linux=<no-hwclock> version and breaks with current
   version of initscripts
 * I include files into initscripts and upload initscripts=<hwclock> and
   conflict with util-linux << <no-hwclock>.

This is how things usually done on non-conffiles. Are there additional
complications with conffiles?
Note, that I send and fetch email in batch, once in a few days.
Please, mention in body of your reply when you add or remove recepients.

More information about the Debian-init-diversity mailing list