Arch Linux

Not sure how you propose to do that. It don't build without it and it links to the library, making it a hard dep.

You complain that you have no use for NetworkManager on a desktop system, what use do you have for powerdevil?

Edit: nevermind, I see it's a dep of plasma-desktop.

NM-qt is a hard dependency for powerdevil, made powerdevil optional for plasma-desktop instead.

I have a better solution to propose regarding this issue: make networkmanager an optdep for networkmanager-qt (or powerdevil and nothing for networkmanager-qt, probably makes more sense). This last one does indeed not depends on networkmanager, just checked that right now.

powerdevil works well without NetworkManager installed, but propose additional wlan/wwan/bt powermanagement handling if NetworkManager is installed (or in use, not sure about what is needed exactly).

Personally, I’m not using NetworkManager and don’t want to, and wlan/… PM is handled by TLP here.

I don't like this solution, it feels like an ugly hack. NM-qt is a wrapper for the NM dbus API, so it certainly depends on NM even if it doesn't link to the library, and is useless without it.

Does this actually cause any issue besides taking a few MB of disk space? It shouldn't have any effect if you don't enable the service.

Well, I found myself without networking ability after update+reboot (netctl wasn’t working without any explicit error message), because NetworkManager was auto-started by something (the corresponding systemd unit not being enabled) and had to figure that out to finally mask NM service as a solution (don’t know what started it) and reboot to get netctl working again. Since I don’t know what caused this and that it can happens in probably lot of unexpected situations, I would recommend avoiding force-pushing NM.

So, while I also expected no effect without enabling the service, it happened that the service was started anyway, masking it being the only solution at this point (or figuring out what started it, but I don’t know how to do that and whether this is “fixable”).

I agree that NM-qt does nothing on it owns without NM, but AFAIK it’s only used as a dependency by other things (like powerdevil) that use it to access NM API, and never as standalone. So, I don’t really think this is a real issue to move the dependency to the programs that actually use NM (through NM-qt).

Found myself in trouble after update as well.

Please see how Fedora solved the issue: https://bugzilla.redhat.com/show_bug.cgi?id=1350973

My vote goes for making networkmanager optdep for networkmanager-qt

I just tested disabling NM. After restarting Plasma I have no network at all, even though plasma-nm is enabled on the system tray. So it's definitely not Plasma which is starting NM automatically.

Can you check the journal to see what is starting NM on your system? It definitely shouldn't be happening.

Just remembered this for instance: https://github.com/linrunner/TLP/issues/193.

Also, as far as I’m concerned, I can live with that situation: I’ve just -Rddn’ed NetworkManager like I did for kwallet (because I don’t want to use this one and Chromium ask me to create a new Wallet each time a password field is present on a page), or even with other things like ttf-oxygen that I don’t want on my system. So, one more error line after each pacman transaction is something I can live with.

@notfood Fedora enables NM by default on install, so we can't compare to them. Besides, pacman doesn't have anything equivalent to Fedora's "recommends" (dependencies that can be removed)

On my system, journal log doesn’t says what started it (just says that it started short after systemd first messages in log, before sddm), but since I have tlp, it should be that. However, you don’t know how many things like that may exist.

systemctl list-dependencies --reverse

So, tlp indeed.

Actually both NetworkManagerQt and BluezQt were made optional by this commit https://quickgit.kde.org/?p=powerdevil.git&a=commit&h=3257be562b2c48eb11c9e32f7e4fd82ea7fb090b but it was reverted later due to 5.7's dependency freeze https://quickgit.kde.org/?p=powerdevil.git&a=commit&h=f2018a8d3c5f2824278bb167cb48d938ca2670c4

Closing this as I can't think of a reasonable way to fix this without tampering with dependencies. If something is starting NM on your system against your will, you should mask the service.
If someone needs powerdevil and absolutely don't want NM installed on their system, they can either:
- Use a dummy empty package that provides NM (like the akonadi one in AUR)
- Use a powerdevil-light package that includes the aforementioned commit and disables NM-qt and bluez-qt support.

IMO the ideal solution would be for upstream to turn NM and bluez support into a plugin so we can make the dependencies optional (like they are for plasma-workspace). Feel free to suggest this upstream.

If someone can think of a better solution, ask for reopen.

As of Plasma 5.8.0 this has been fixed. The change-log for Powerdevil says:

"Make NetworkManagerQt and BluezQt optional."

https://www.kde.org/announcements/plasma-5.7.5-5.8.0-changelog.php#powerdevil

Please consider making both networkmanager-qt and bluez-qt optional dependencies.

Thank you for your help.

They can't be optional deps. They're build time optional, so they're either deps or they're not. This is at the maintainer's discretion.

Yes, but at least we can now provide powerdevil-light in AUR easily. Expect it soon. ;)

Here you go: https://aur.archlinux.org/packages/powerdevil-light/.

Please note you need to remove networkmanager-qt and/or bluez-qt (any is enough) first and compile this one after, else it will detect them at build time and compile with them as deps. Or alternatively, build in a clean chroot.

@ArchangeGebriel you can pass -DKF5BluezQt_FOUND=OFF to cmake to pretend that bluez-qt isn't available even if it is.