On 5/26/21 5:23 PM, Mark Hindley wrote:
> On Wed, May 26, 2021 at 04:23:56PM +0900, Simon Walter wrote:
>> On 5/26/21 12:37 AM, Mark Hindley wrote:
>>> This is actually a Debian bug and should be fixed there.
>>>
>>> If cockpit requires systemd, is should declare that dependency.
>>
>> AFAICT, it requires a systemd socket. It doesn't require any systemd
>> packages (debs). Would that still be considered a dependency WRT to
>> packages?
>
> I am no systemd expert ;) and have little first hand knowledge of it. But my
> limited understanding is that systemd socket activation is the systemd
> absorption/reimplementation of inetd(8) and requires systemd itself to be
> running.
>

Cockpit uses a systemd socket get started on demand. The socket is not
needed for the cockpit-ws cockpit-bridge etc to run. However, I am told
that the version used in Beowulf is old and a newer version will fail to
work w/o systemd. So, I guess lets see how far it makes it on Devuan. It
may become totally useless, and then I'll file a bug upstream. You can
see my bug report here:

https://www.mail-archive.com/debian-bugs-dist@lists.debian.org/msg1806544.html

In an email not on bug tracker for some reason:

"More recent cockpit versions have a much more complex unit with
cockpit.{socket,service} controlling the cockpit-tls component, and that in
turn launches per-client-certificate cockpit-ws instances (as a separate
user)
through systemd socket activation. So this init script does not apply
at all
to current versions, and there is no reasonable way how to write one.

So in summary, there is no way of running cockpit in a non-systemd/Linux
environment that I'd be willing to support. For these I'd rather recommend
looking at webmin, ebox, or similar project."

Best regards,

Simon