[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Avoiding SystemD isn't hard

On 10/22/2014 07:41 AM, Martin Steigerwald wrote:

Am Mittwoch, 22. Oktober 2014, 07:04:12 schrieb Jape Person:

On 10/21/2014 09:22 PM, Miles Fidelman wrote:

Don Armstrong wrote:

On Tue, 21 Oct 2014, Miles Fidelman wrote:

which is immediately followed by completely inaccurate information,
including:

------
"With jessie, it will become /easier/ to choose the init system, because
*neither init system is essential now*. Instead, there is an essential
meta-package "init", which requires you to install one of systemd-sysv |
sysvinit-core | upstart. In other words, you have more choice than ever
before."


This statement is actually correct; until this change it was not
possible to completely replace sysvinit with another init system without
running dpkg-divert commands to divert the init away.


ok.


When, in fact, there IS no way to prevent systemd from being installed
during installation (except for upgrades).


That preseeding doesn't do this is a bug, it's filed (#668001), and the
patch for it was just written on Friday the 17th. Because Debian is
going to freeze in less than three weeks, the maintainers are wary of
applying this patch this close to release without extensive testing.

Furthermore, the effect of this patch is trivially obtained by using a
late_command to remove systemd-sysv and install sysvinit-core.


except for the various reported issues with all the things aptitude
wants to remove when you try this.... (I really hate unwinding highly
entangled dependencies).


If you actually want to see this patch applied to the version of the
Debian installer that Jessie will release with, you should coordinate
with the nice people in #debian-boot to see what type of testing they
would want to see before they are willing to vet the patch.


Good point - will see how receptive they are.


Which leaves the only option being to install, with systemd, then follow
the instructions in
https://randomstring.org/blog/blog/2014/10/14/removing-systemd-from-a-d
ebian-jessie-system/>>>
|# apt-get install sysv-rc sysvinit-core sysvinit-utils

# apt-get purge systemd libpam-systemd systemd-sysv


This is the wrong command to run. You want:

aptitude install sysvinit-core systemd-sysv-;

Removing libpam-systemd and systemd something depends on them isn't
useful; they don't determine what the init system is, after all.


seems like we could use some definitive instructions on how to actually
do this; we now have:

https://lists.debian.org/debian-user/2014/09/msg00969.html
and
https://randomstring.org/blog/blog/2014/10/14/removing-systemd-from-a-debi
an-jessie-system/

(where I copied the above instructions from)
followed by:
https://lists.debian.org/debian-user/2014/09/msg00998.html
and now your comments above
- plus the various bug reports about how to avoid mass uninstallations
by aptitude (e.g.,
https://lists.debian.org/debian-devel/2014/10/msg00409.html)

A distinct lack of clarity, given the TC resolution about expecting
Jessie to support multiple init systems, but not actually having a
clearly defined way to do anything except hack until one gets things to
work.

Miles Fidelman


I haven't paid a lot of attention to threads concerning systemd because
of the (unfortunate, though occasionally entertaining) hyperbole and
innuendo employed by so many of the involved parties. I mostly despaired
of finding anything substantive in the various threads.

But I'm glad you (and Don Armstrong) have been among the more reasonable
participants in the argument.

I was wondering if either of you -- or anyone else -- has tried the
init-select package to see if it is actually able to avoid adverse
effects upon the installed software while allowing the user (presumably
someone with root privs) to select among init systems. That would be
nice, wouldn't it?

I intend to look at it this weekend, if I can find some time for a
little project. I'm guessing that it's pretty basic, and maybe not meant
for use on systems with DEs, DMs, WMs, and lots of GUI user applications
installed on them. It it works for those, then the author truly deserves
congratulations.

I hope everyone finds a way to get a system configuration s/he can live
with happily!


I did, as did a co-worker of mine.

It appears to be basically working, but not always does what I would expect
from it. So I reported some bug reports about it:

init-select init-select adds sysvinit while systemd already there
https://bugs.debian.org/762558

init-select: assumes systemd is default while its not
https://bugs.debian.org/762578

http://bugs.debian.org/init-select shows one other bug regarding
functionality:

init-select: Offers not installed inits, confusing error messages, unhelpfull
description
https://bugs.debian.org/759031

Thats not much so far, I wonder whether many people test this.

A co-worker of mine tested with openrc, but there was a problem on boot with
openrc. It hung on some service and it wasn´t even possible to Ctrl-C it like
often possible with sysvinit. I don´t remember the details tough. But he
switched back. While openrc may be reliable and stable with Gentoo, I think
the debian variant of it can actually use more testing.

Ciao,
That's interesting. Thanks for the heads-up. I'm going to read the bug reports and mull them over before I decide whether or not to actually test the package. 

Regards,
Jape
Reply to: