Marc Haber:
On Sat, Dec 13, 2025 at 06:38:15PM +0100, Niels Thykier wrote:To my knowledge, people use `systemd-tmpfiles` for creating/chown file system items not shipped in the `.deb`.And how about systemd-sysusers?
I am not sure what you mean with this question. The debhelper stack has support for `systemd-sysusers` via `dh_installsysusers`, and the snippets are aligned with `dh_installtmpfiles` to ensure they are run in the proper order.
In compat 14, `dh_installsysusers` is enabled by default. In compat 13, you can use `dh_installsysusers` via a Build-Depends on `dh-sequence-installsysusers`. Before that, you have to also activate `dh_installtmpfiles` and ensure they are run in the correct order (#885580). I believe (but did not verify that) the Build-Depends way gets the order right for this particular case. But compat 13 is available in all supported versions of Debian, so using compat 13 is a better choice at this junction in my view.
This limitation of not being able to "act between" parts of the debhelper provided maintscripts has existed since the invention of debhelper and the design of debhelper is not built around giving the maintainer flexibility. I do not see myself fixing that.I understand, but that kind of makes it significantly harder to use systemd-sysusers. As the adduser maintainers, I see that with a laughing and a crying eye, but things are going towards declarative.> As maintainer scripts try to be idempotent, people might get around to using #DEBHELPER# twice?But I trust your judgement, and since I seem to be the only person seeing a probable issue in that, I'm probably wrong.Greetings Marc
I have never heard or seen anyone use the #DEBHELPER# token twice.The only case I am aware of where someone needed to "split" the #DEBHELPER# token that is not supported currently involved having to reload dbus at a very specific point (#1070473) since the `dbus` trigger cannot handle that case. Perhaps that is the problem you are having.
In this particular case, please see https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1070119#55.
If that is not the problem you are observing, please describe the concrete problem/case you are observing, so I can engage in that rather than having to guess and ending up solving the problem/case you did not have. :)
Best regards, Niels
OpenPGP_signature.asc
Description: OpenPGP digital signature
