On Fri, Aug 4, 2023 at 11:57 AM tito via Dng <dng@???> wrote:
>
> On Fri, 4 Aug 2023 08:43:24 +0000
> Manfred Wassmann via Dng <dng@???> wrote:
>
> > From my experience, I gave it a try in Debian Bullseye and i ended up with
>
> > a running but removed pid 1 systemd which wasn't able to do anything and
> > after trying all other options the system had to be shut down by unplugging
> > the power supply. So they may not have a compatibility layer in Debian,
> > which seems likely to me, or there is no (functional) stability guarantee
> > with systemd.
>
> Hi,
> in bookworm you cannot remove systemd even if you try very hard
> with dpkg --force-all etc. unless you delete it by hand.

I checked the Bookworm package and it has exactly the same pre-remove
script as the Bullseye package and for all who are not familiar enough
with the Debian package management system: If a prerm script of a
package returns an error, the removal of the package is aborted and
there is no --force-XXX option which would override that behaviour.
Now the systemd prerm script checks for the existence of a directory
which systemd creates when it starts up and errors out (exit 1) in
that case. As noted in the file this code exactly is there to prevent
the systemd package to be removed if systemd is the active init.

> You can install another init with systemd running and reboot to it.
> OTOH you can remove sysvinit with dpkg more easily and end
> with a removed pid 1. Tested.

Then you should probably go to Ian Jackson's Blog at
diziet.dreamwidth.org, which initially spurred this thread[1], and
notify him about that. He surely will be delighted if a real insider
points out his whole story of needing to hack the install process to
be able to replace systemd in Debian and how to do it was nothing but
an excuse to post some Anti-Devuan-FUD.

[1] https://lists.dyne.org/lurker/message/20230721.011329.adaba746.en.html