ublue-update F41 notifiy fix

B

bsherman•10/26/24, 8:33 PM

@Gerblesh @fiftydinar i'm threading this

B

bsherman•10/26/24, 8:34 PM

https://github.com/ublue-os/ublue-update/issues/132

GitHub

`loginctl` commands need to be changed since Fedora 41 (systemd v25...

--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...

G

Gerblesh•10/26/24, 8:34 PM

Yeah I'm just wondering if we could just use the python

pwd

pwd

pwd

pwd module to avoid using loginctl subprocess

B

bsherman•10/26/24, 8:34 PM

As an aside, we can probably just get a list of users with a UID above 1000 and a homedir, and sidestep needing to call systemd as a subprocess

@Gerblesh said the above

B

bsherman•10/26/24, 8:35 PM

@Gerblesh I don't know what you meant by

Would that work with machinectl?

M

M2•10/26/24, 8:35 PM

machinectl shell UID@ /path/to/process should work

MM2 machinectl shell UID@ /path/to/process should work

G

Gerblesh•10/26/24, 8:35 PM

Even if the user is not logged in?

M

M2•10/26/24, 8:35 PM

i mean it works for machinectl shell 1000@ /usr/bin/notify-send

MM2 i mean it works for machinectl shell 1000@ /usr/bin/notify-send

G

Gerblesh•10/26/24, 8:35 PM

B

bsherman•10/26/24, 8:36 PM

i'm really confused

B

bsherman•10/26/24, 8:36 PM

we're using loginctl to get a list of active sessions

B

bsherman•10/26/24, 8:36 PM

not to send a notification

B

bsherman•10/26/24, 8:36 PM

or are those comments 2 distinct ideas

Bbsherman we're using loginctl to get a list of active sessions

G

Gerblesh•10/26/24, 8:37 PM

We 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

B

bsherman•10/26/24, 8:39 PM

honestly, i'd prefer to just see a fix and not a refactor

B

bsherman•10/26/24, 8:40 PM

bazzite is trying to launch F41 on Tuesday

G

Gerblesh•10/26/24, 8:40 PM

https://github.com/systemd/python-systemd

GitHub

GitHub - systemd/python-systemd: Python wrappers for systemd functi...

Python wrappers for systemd functionality. Contribute to systemd/python-systemd development by creating an account on GitHub.

B

bsherman•10/26/24, 8:41 PM

yeah, i've already been down this path, it mgith work but it's not current

B

bsherman•10/26/24, 8:41 PM

hopefully it's version agnosttic

B

bsherman•10/26/24, 8:42 PM

but they haven't updated sinc v235 and we are hitting issues with v256

G

Gerblesh•10/26/24, 8:43 PM

Yeah I'll just add a version check to the PR for F41

B

bsherman•10/26/24, 8:44 PM

i'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.

GGerblesh Yeah I'll just add a version check to the PR for F41

B

bsherman•10/26/24, 8:45 PM

this is all i was going to do...

this thing runs infrequently enough, it could probably even be in the

get_active_sessions

get_active_sessions

get_active_sessions

get_active_sessions function itself

Bbsherman this is all i was going to do... this thing runs infrequently enough, it could ...

G

Gerblesh•10/26/24, 8:47 PM

Yeah, thinking about it now, I don't think it's practical. CoreOS does maintain some good systemd bindings for golang though

G

Gerblesh•10/26/24, 8:48 PM

Nvm they're also a bit old

M

M2•10/26/24, 10:45 PM

Shelling out is fine.

G

Gerblesh•10/26/24, 11:37 PM

@bsherman I have systemd 256 but the

-j

-j

-j

-j flag does not work?

G

Gerblesh•10/27/24, 12:17 AM

gdbus call --system \
    --dest org.freedesktop.login1 \
    --object-path /org/freedesktop/login1 \
    --method org.freedesktop.login1.Manager.ListUsers

gdbus call --system \
    --dest org.freedesktop.login1 \
    --object-path /org/freedesktop/login1 \
    --method org.freedesktop.login1.Manager.ListUsers

gdbus call --system \
    --dest org.freedesktop.login1 \
    --object-path /org/freedesktop/login1 \
    --method org.freedesktop.login1.Manager.ListUsers

gdbus call --system \
    --dest org.freedesktop.login1 \
    --object-path /org/freedesktop/login1 \
    --method org.freedesktop.login1.Manager.ListUsers

G

Gerblesh•10/27/24, 12:17 AM

works

G

Gerblesh•10/27/24, 12:17 AM

the dbus API shouldn't change anytime soon

G

Gerblesh•10/27/24, 12:17 AM

at least I'd hope

G

Gerblesh•10/27/24, 1:53 AM

is 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)?

GGerblesh the dbus API shouldn't change anytime soon

B

bsherman•10/27/24, 2:10 AM

the dbus way looks better anyway

GGerblesh is there a way to make it so that an RPM spec builds things differently based on...

B

bsherman•10/27/24, 2:11 AM

yes... but does dbus approach not work on both 40 and 41?

G

Gerblesh•10/27/24, 2:11 AM

it works on both yes

G

Gerblesh•10/27/24, 2:11 AM

I tried dbus and now I'm trying the rpmSPEC

G

Gerblesh•10/27/24, 2:11 AM

if dbus is preferred then I'll do that instead

G

Gerblesh•10/27/24, 2:11 AM

I just figured we didn't want a rewrite

G

Gerblesh•10/27/24, 2:12 AM

loginctl is slightly more readable because the JSON fields are not just an array, but the dbus API does work

B

bsherman•10/27/24, 2:12 AM

if the change is to use dbus in

get_active_sessions

get_active_sessions

get_active_sessions

get_active_sessions instead of executing a command a few times, that seems smart

oh, but you still have to exec it?

Bbsherman if the change is to use dbus in `get_active_sessions` instead of executing a co...

G

Gerblesh•10/27/24, 2:13 AM

well you could exec it or use the dbus library

B

bsherman•10/27/24, 2:13 AM

$ gdbus call --system \
    --dest org.freedesktop.login1 \
    --object-path /org/freedesktop/login1 \
    --method org.freedesktop.login1.Manager.ListUsers
([(uint32 1000, 'benjamin', objectpath '/org/freedesktop/login1/user/_1000')],)

$ gdbus call --system \
    --dest org.freedesktop.login1 \
    --object-path /org/freedesktop/login1 \
    --method org.freedesktop.login1.Manager.ListUsers
([(uint32 1000, 'benjamin', objectpath '/org/freedesktop/login1/user/_1000')],)

$ gdbus call --system \
    --dest org.freedesktop.login1 \
    --object-path /org/freedesktop/login1 \
    --method org.freedesktop.login1.Manager.ListUsers
([(uint32 1000, 'benjamin', objectpath '/org/freedesktop/login1/user/_1000')],)

$ gdbus call --system \
    --dest org.freedesktop.login1 \
    --object-path /org/freedesktop/login1 \
    --method org.freedesktop.login1.Manager.ListUsers
([(uint32 1000, 'benjamin', objectpath '/org/freedesktop/login1/user/_1000')],)

I know that's not JSON, but it is certainly parseable

G

Gerblesh•10/27/24, 2:14 AM

oh it is JSON

B

bsherman•10/27/24, 2:14 AM

oh! HAH!

B

bsherman•10/27/24, 2:14 AM

yeah, the type through me off

G

Gerblesh•10/27/24, 2:14 AM

I used busctl but it's largely the same

G

Gerblesh•10/27/24, 2:14 AM

is the dbus library better though? Since we are going a seperate route anyways

B

bsherman•10/27/24, 2:15 AM

;-D

B

bsherman•10/27/24, 2:15 AM

considering it'd be good to have this working now

what's easiest way for you?

G

Gerblesh•10/27/24, 2:16 AM

tbh it would probably be the dbus library, would remove a lot of the execs

ublue-update F41 notifiy fix

Similar Threads

Similar Threads

Similar Threads