Dear adduser maintainers, recently a growing number of packages are using debhelper's installsysusers addon to create system users. installsysusers uses systemd-sysusers to create system users, so adduser it's completely sidestepped. At the same time several DDs decided that dh_sysuser (I'm the maintainer), that uses adduser, has to go away before the next stable release[1]. Now my problem is that the way installsysusers adds a dependency on systemd doesn't work well for alternatives inits[2], so I'm not comfortable to drop dh_sysuser and I'm looking for a solution. If adduser could be used as alternative to systemd by dh_installsysusers that would solve my issue. Even if you don't care about my use case, the linux world is standardizing more and more on systemd features so maybe this could be a useful addition to adduser? According to my understanding, in order to be used by dh_installsysusers adduser has to learn to A) create system user on a target chroot B) use a sysusers.d[3] file as input I'm available to produce patches for the two above, it could be a new option in adduser or a new "addsysuserd" program, I don't have an opinion on that. If patches are accepted I also will be available to fix bugs that could arise from those new features. this is a relevant effort for me, so I would like to know in advance if a sane patch for this will be accepted or not. Thanks for considering, Lorenzo [1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=981937 [2] does not work for hurd too https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1076045 [3] https://www.freedesktop.org/software/systemd/man/latest/sysusers.d.html
systemd | systemd-standalone-sysusers | systemd-sysusers | opensysusers As the package name suggests, systemd-standalone-sysusers does not depend on systemd. [2,3] [1] https://salsa.debian.org/debian/debhelper/-/blob/c3540cdad2785c79264a9cfd227aa4956460ed69/dh_installsysusers#L104 [2] https://packages.debian.org/sid/systemd-standalone-sysusers [3] https://salsa.debian.org/systemd-team/systemd/-/blob/bb6db3edfe40fe1a98cdcc6d2d07a7dac38aefc5/debian/control#L447-461 Isn't that what opensysusers already does? Regards,
Hi, In principle yes, but it is not in testing due to issues like [1]. It already had other security issues in the past[2] and upstream seems dead (last commit two years ago). Probably someone would need to write a proper replacement that doesn't have such issues, but there currently seems nobody who is both interested and capable to implement it (as for similar other cases). Ansgar [1]: https://bugs.debian.org/1055517 [2]: https://bugs.debian.org/992058
*shrug* As using adduser is just a recommendation and not a requirement, and adduser sees itself as a service provider not as a mandatory thing, and going a declarative approach certainly has its advantages, I don't care much about packages not using adduser. As I can see, dh_sysuser is just used by a single package and I find the idea of incorporating the functionality into that single package compelling. As the adduser maintainer, I embrace ways to create system users in a different way. That reduces my workload. While I strive to deliver a good service to the packages using adduser to create their system users I acknowledge that the number of such packages is going to get smaller due to packages migrating to the systemd way of doing things. I am okay with that. I think it would be a better idea to either keep this in a separate package. I am not sure whether it is within Luca's powers to kind of order you to have your package removed from Debian. I don't think that adduser should be part of a political fight. I am way too tired to be part of this. So, sorry, I am not convinced that adduser should adopt a totally different kind of interface at this time. Greetings Marc
Control: tags -1 wontfix thanks The discussion has slowed down and died without new arguments being on the table. I am therefore, for the time being, tagging this bug as wontfix without this being a final judgement. Greetings Marc
Control: tags -1 wontfix thanks The discussion has slowed down and died without new arguments being on the table. I am therefore, for the time being, tagging this bug as wontfix without this being a final judgement. Greetings Marc