BBGitHub
--output=json does not work anymore, just using -j is sufficient (however, this will break F40 & older, so some dynamic logic needs to be applied for this) See this issue for more details: syst...
GYeah I'm just wondering if we could just use the python
pwd module to avoid using loginctl subprocess
Mmachinectl shell UID@ /path/to/process should work
GEven if the user is not logged in?
Mi mean it works for machinectl shell 1000@ /usr/bin/notify-send
G
Bi'm really confused
Bwe're using loginctl to get a list of active sessions
Bnot to send a notification
Bor are those comments 2 distinct ideas
GWe use loginctl to figure out the active sessions yes, but I honestly think it would be fine to just run uBlue update for every user above UID 1000 with a homedir, regardless if they have an active session
Bhonestly, i'd prefer to just see a fix and not a refactor
Bbazzite is trying to launch F41 on Tuesday
GGitHub
Python wrappers for systemd functionality. Contribute to systemd/python-systemd development by creating an account on GitHub.
Byeah, i've already been down this path, it mgith work but it's not current
Bhopefully it's version agnosttic
Bbut they haven't updated sinc v235 and we are hitting issues with v256
GYeah I'll just add a version check to the PR for F41
Bi'm not opposed to bigger picture improvements i'm sure it could be cool and work better, things usually can...
but I appreciate you looking at the short term fix need.
but I appreciate you looking at the short term fix need.
Bthis is all i was going to do...
this thing runs infrequently enough, it could probably even be in the
this thing runs infrequently enough, it could probably even be in the
get_active_sessions function itselfGYeah, thinking about it now, I don't think it's practical. CoreOS does maintain some good systemd bindings for golang though
GNvm they're also a bit old
MShelling out is fine.
GGworks
Gthe dbus API shouldn't change anytime soon
Gat least I'd hope
Gis there a way to make it so that an RPM spec builds things differently based on fedora version (if fedora version is less than F41, build with older session.py)?
Bthe dbus way looks better anyway
Byes... but does dbus approach not work on both 40 and 41?
Git works on both yes
GI tried dbus and now I'm trying the rpmSPEC
Gif dbus is preferred then I'll do that instead
GI just figured we didn't want a rewrite
Gloginctl is slightly more readable because the JSON fields are not just an array, but the dbus API does work
Bif the change is to use dbus in
oh, but you still have to exec it?
get_active_sessions instead of executing a command a few times, that seems smartoh, but you still have to exec it?
Gwell you could exec it or use the dbus library
BI know that's not JSON, but it is certainly parseable
Goh it is JSON
Boh! HAH!
Byeah, the type through me off
GI used busctl but it's largely the same
Gis the dbus library better though? Since we are going a seperate route anyways
B;-D
Bconsidering it'd be good to have this working now 
what's easiest way for you?
what's easiest way for you?
Gtbh it would probably be the dbus library, would remove a lot of the execs
