Bug 1592330 - [Tracking BZ] Drop unnecessary dependencies on 'initscripts' package
Summary: [Tracking BZ] Drop unnecessary dependencies on 'initscripts' package
Keywords:
Status: ASSIGNED
Alias: None
Product: Fedora
Classification: Fedora
Component: initscripts
Version: 29
Hardware: All
OS: Linux
medium
medium
Target Milestone: ---
Assignee: Lukáš Nykrýn
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
Depends On: 1592335 1592338 1592342 1592343 1592345 1592359 1592366 1592368 1592370 1592379 1592381 1592385 1592386 1592389 1592391 1592392 1592395 1592398 1592399 1592405 1592406 1592443 1592325 1592340 1592344 1592348 1592349 1592350 1592351 1592352 1592354 1592355 1592356 1592357 1592360 1592362 1592363 1592364 1592372 1592377 1592378 1592380 1592382 1592383 1592384 1592388 1592390 1592393 1592394 1592396 1592400 1592401 1592402 1592403 1592404 1592410 1592444 1592449
Blocks:
TreeView+ depends on / blocked
 
Reported: 2018-06-18 12:49 UTC by David Kaspar [Dee'Kej]
Modified: 2018-11-01 01:31 UTC (History)
5 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
: 1610274 (view as bug list)
Environment:
Last Closed:


Attachments (Terms of Use)

Description David Kaspar [Dee'Kej] 2018-06-18 12:49:01 UTC
Description of problem:
There are still quite a number of packages out there which depend on 'initscripts' package for no longer valid reason. For future work we need to decrease the number of dependant packages as much as possible.

This is a tracking BZ for all the packages depending on initscripts for some reason.

Version-Release number of selected component (if applicable):
https://koji.fedoraproject.org/koji/buildinfo?buildID=1092002

Additional info:
Please, do not post any comments directly in this bug. Create new bug if necessary.

Comment 1 Zbigniew Jędrzejewski-Szmek 2018-06-18 13:02:08 UTC
I strongly suggest putting the name of the package as the first word in your bugreports. Bugzilla doesn't always make it easy to see for which package the bug is in listings.

Comment 2 David Kaspar [Dee'Kej] 2018-06-18 13:41:06 UTC
(In reply to Zbigniew Jędrzejewski-Szmek from comment #1)
> I strongly suggest putting the name of the package as the first word in your
> bugreports. Bugzilla doesn't always make it easy to see for which package
> the bug is in listings.

Uh, the pain of doing this manually... o.O I really need to write myself some script for this... :D

Comment 3 Pavel Raiskup 2018-06-18 14:50:47 UTC
Is this effort backed up by Fedora guidelines change?

Comment 4 David Kaspar [Dee'Kej] 2018-06-18 15:13:26 UTC
(In reply to Pavel Raiskup from comment #3)
> Is this effort backed up by Fedora guidelines change?

Why would it be? In other words, why should packages have unnecessary dependencies? :)

Comment 5 David Kaspar [Dee'Kej] 2018-06-18 15:15:24 UTC
(In reply to David Kaspar [Dee'Kej] from comment #4)
> (In reply to Pavel Raiskup from comment #3)
> > Is this effort backed up by Fedora guidelines change?

In short, no it isn't backed up by Fedora guidelines change. At least I don't know about any guideline change related to this.

The only thing I can point you right now is this:
https://fedoraproject.org/wiki/Packaging:Guidelines#Initscripts

Comment 6 Pavel Raiskup 2018-06-18 15:19:57 UTC
Aha, I probably misinterpreted the reasons for this tracker.

Do I understand it correctly that it is not an attempt to switch
spec files (%post/%postun scriptlets, etc.) from using

  service <SERVICE> condrestart

into

  systemctl condrestart <SERVICE>

That said, is it still valid to depend on initscripts for /sbin/service
to work?

Comment 7 Lukáš Nykrýn 2018-06-18 15:30:59 UTC
Well speaking about that. If you need service, you are probably shipping initscripts and If I am not mistaken that is prohibited by our policy unless you have an exception.

Comment 8 Pavel Raiskup 2018-06-18 15:57:13 UTC
(In reply to Lukáš Nykrýn from comment #7)
> Well speaking about that. If you need service, you are probably shipping
> initscripts and If I am not mistaken that is prohibited by our policy unless
> you have an exception.

The assumption isn't correct.  Consider I'm maintaining spec file
which is supposed to work both on el6 and Fedora Rawhide, and I don't want
to have %if 0%{?fedora} >= X || 0%{?rhel} >= Y in scriptlets.

Comment 9 David Kaspar [Dee'Kej] 2018-06-19 11:52:08 UTC
(In reply to Pavel Raiskup from comment #8)
> The assumption isn't correct.  Consider I'm maintaining spec file
> which is supposed to work both on el6 and Fedora Rawhide, and I don't want
> to have %if 0%{?fedora} >= X || 0%{?rhel} >= Y in scriptlets.

I understand your reason. On the other hand, having the %if conditionals is common practice in specfiles, and installing initscripts just to install/enable some service is quite inconsiderate to end-users (e.g. in mimimal OS images or containers).

We would like to see that no package in the future would depend on initscripts, so the package could be uninstalled from the system if user wants, thus initscripts becoming useful for only few (mostly) 3rd party/proprietary software (which still relies on initscripts).

On the other hand, if we see many people / packages depending on the /usr/sbin/service so much (I understand the convenience of using it), it might be worth trying to completely separate the 'service' into a new package. I'll have to think about it more.

For now, it would be good to at least get some statistics, on how many packages still need the initscripts and for what reasons. :)

Comment 10 Pavel Raiskup 2018-06-19 20:35:52 UTC
(In reply to David Kaspar [Dee'Kej] from comment #9)
> On the other hand, if we see many people / packages depending on the
> /usr/sbin/service so much (I understand the convenience of using it), it
> might be worth trying to completely separate the 'service' into a new
> package.

Or provide /sbin/service in systemd package?  AFAIK that the service command
will be needed indefinitely to comply with LSB.

Comment 11 David Kaspar [Dee'Kej] 2018-06-20 12:48:15 UTC
(In reply to Pavel Raiskup from comment #10)
> Or provide /sbin/service in systemd package?  AFAIK that the service command
> will be needed indefinitely to comply with LSB.

This was actually my other idea, and it kind of makes sense that systemd would ship it, but... I really don't expect systemd to accept it.

Maybe it could be a separate package, and systemd might have weak dependency on it, like "suggests"? :)

Comment 12 Lukáš Nykrýn 2018-06-20 14:49:59 UTC
I don't like that, usage of service command should be tight to initscripts. The use-case mentioned here is a misuse in my eyes and brings an unnecessary dependency. And that is something we should always try to fix. Generally speaking, you should use the condition. If you have a specific use-case, then we should discuss it, but it is not for this Bugzilla.

Comment 13 Jan Kurik 2018-08-14 11:12:02 UTC
This bug appears to have been reported against 'rawhide' during the Fedora 29 development cycle.
Changing version to '29'.


Note You need to log in before you can comment on or make changes to this bug.