Use of loginctl terminate-session to logout is apparently nonstandard

[LRitzdorf]

When attempting to log out of i3 via loginctl terminate-sesison (as rofi-power-menu currently does), i3 exits with code 1. This is reported as an error by sddm, and also leaves behind an X session with a background image and responsive cursor but no interactable elements, which must be killed via Ctrl+Alt+Bksp.

I originally opened this as an issue on i3, but after further testing, found that Openbox and Awesome behave in the same way when killed by loginctl. Thus, it would appear that terminating a session in this way is not a standard method of logging out, and should probably not be the default configuration for rofi-power-menu. Unfortunately, I'm not sure what an appropriate replacement would be, though I'm also not very experienced in the area of desktop interface standards to begin with.

[LBlend]

I've stumbled upon a similar issue on Hyprland. It would be nice if we could specify what action would be executed for each item.

[LRitzdorf]

If it helps, I've been able to work around this by having rofi-power-menu call a shell script of my own:

#!/bin/sh

if [ -z "$DISPLAY" ]; then
    exit 1
fi
case "$XDG_CURRENT_DESKTOP" in
    "i3")
        i3-msg exit ;;
    "Hyprland")
        hyprctl dispatch exit ;;
    *)
        loginctl terminate-session "${XDG_SESSION_ID:-}" ;;
esac

Adding more desktop environments here should be pretty self-explanatory :)

[LBlend]

If it helps, I've been able to work around this by having rofi-power-menu call a shell script of my own:

#!/bin/sh

if [ -z "$DISPLAY" ]; then
    exit 1
fi
case "$XDG_CURRENT_DESKTOP" in
    "i3")
        i3-msg exit ;;
    "Hyprland")
        hyprctl dispatch exit ;;
    *)
        loginctl terminate-session "${XDG_SESSION_ID:-}" ;;
esac

Adding more desktop environments here should be pretty self-explanatory :)

Ah great! This is exactly what I'm looking for!

This might be a dumb question but how did you get rofi-power-menu run your script? Did you just modify the source code on your own?

[LRitzdorf]

I did, yes. The actual executable (i.e. which rofi-power-menu) is a shell script, so editing is pretty straightforward.

Also, since I'm on Arch, pacman will at least warn me if it's about to overwrite my changes, and save a backup of them. Your mileage with apt or whatever else may vary.

[LBlend]

I see! I'm on Arch as well so I'll just have to keep an eye out for that I guess!
Thank you!

[LRitzdorf]

At the risk of going too far off topic, that's just pacman's pacnew/pacsave system. The pacdiff tool (part of the pacman-contrib package) is really nice for automatically finding and resolving them, as well!

[kennykingdev]

I've stumbled upon a similar issue on Hyprland. It would be nice if we could specify what action would be executed for each item.

This is my thought as well, running sway in my case. The option to set everything with a config file would be great!

[macedigital]

I stumbled on this issue as well (broken SDDM greeter after logging out) and I believe there is a better way by changing the command to loginctl kill-session "${XDG_SESSION_ID-}"¹.

The solution should be agnostic about greeter or window-manager, neither requiring adding more cases e.g. to the wrapper script from #22 (comment), (e.g. sway sessions would be stopped with swaymsg exit), nor passing in configuration options.

I've prepared a pull-request #36 and would welcome feedback if the change fixes the issue for others as well.

Footnotes

1. https://github.com/jluttine/rofi-power-menu/blob/395c1e07360b2dbd13c0a658665ab0a581024ec3/rofi-power-menu#L43C27-L43C36 ↩