Improve output configuration via wlr-output-configuration

[vaxerski]

Improves the handling of wlr-output-configuration (e.g. wdisplays, kanshi)

Notable changes:

• Output configuration will be valid for as long as the client has the global bound (so likely for as long as it's running)
• Wlr output configuration will no longer be a rule, it will be treated as a hard override of whatever the user sets in the cfg (for as long as point 1)
• Older clients have precedence (but who runs 2 monitor managers?)

Ref #7059

[leiserfg]

I just built it and now kanshi thinks the internal monitor was disabled, but it is still enabled (fully working not as before the patch when it was disabled but the last frame was still shown). Will try wdisplay.

[leiserfg]

Wdisplay does not work either, I disable the Enabled checkbox and as soon as I click Apply it gets enabled again.

[vaxerski]

log pls

[vaxerski]

oh wait sec

[vaxerski]

now

[leiserfg]

Do you want a log with kanshi, with wdisplay, or both?

[vaxerski]

whichever doesnt work, and explain what doesnt work

[leiserfg]

After e it worked with kanshi, let me play with it a bit more.

[vaxerski]

the power of e

[leiserfg]

It disabled the internal monitor correctly, but didn't enable it when I unplugged the external one
hyprland.log

My Kanshi rules are:

profile undocked {
  output "eDP-1" enable scale 1.175000
}

profile docked-left {
  output "DP-3"
  output "eDP-1" disable
}

profile docked-right {
  output "DP-2"
  output "eDP-1" disable
}

the logs of kanshi are:

Aug 29 16:32:54 rahmen systemd[1417]: Started Dynamic output configuration.
Aug 29 16:32:55 rahmen kanshi[952746]: applying profile 'docked-left'
Aug 29 16:32:55 rahmen kanshi[952746]: applying profile output 'DP-3' on connected head 'DP-3'
Aug 29 16:32:55 rahmen kanshi[952746]: applying profile output 'eDP-1' on connected head 'eDP-1'
Aug 29 16:32:55 rahmen kanshi[952746]: configuration for profile 'docked-left' applied
Aug 29 16:33:17 rahmen kanshi[952746]: no profile matched

it seems to me like the internal monitor never gets re-exposed to kanshi.

[leiserfg]

I did sleep 5 && hyprctl monitors and disconnected the external monitor and got unknown request.

[vaxerski]

eee

[vaxerski]

can you post WAYLAND_DEBUG=1 of kanshi when it fails to re-enable edp1

[leiserfg]

I tested with Shikane now (same rules) cause in nixos is easier for me to get the stderror of that one.
shikane.log

[vaxerski]

I don't see anything wrong? What didn't work?

[leiserfg]

The laptop's display stays off.

[vaxerski]

ye from that log I can't see what's wrong? shikane just... doesnt ask us to enable it.

Can you grab a log from something like kde or sway where it works?

[leiserfg]

I will test the PR further tonight.

[leiserfg]

I'm testing again. I started hyprland with this as part of my config:

 exec-once = env WAYLAND_DEBUG=1 shikane 2> /tmp/shikane.log

and I'm doing tail -f /tmp/shikane.log, and I don't see logs appearing when I plug and unplug the external monitor. So it is kinda worse than it is in master. Right now my external monitor is showing a fix frame but does not show as enabled in hyprctl monitors.

shikane logs:
shikane.log

[vaxerski]

hm, thats odd. I'll test on my laptop tomorrow... Does socket2 log anything?

socat -v unix-connect:$XDG_RUNTIME_DIR/hypr/$HYPRLAND_INSTANCE_SIGNATURE/.socket2.sock -

[leiserfg]

@vaxerski checking in the git history I found this PR that was what fixed kanshi when a similar problem was happening in the days of wlroots

https://github.com/hyprwm/Hyprland/pull/3718/files#diff-dbcd1ce1e0b916536f17c999473ad9e22c014646ddb2df15c45d726884a3481cR1293-R1300

[vaxerski]

@leiserfg I've rebased this on top of main. I remember some other fixes on main, is this better than the current head?

[leiserfg]

I was chatting with the author of shikane and he told me the issue is:

When a display gets unplugged the compositor should send Finished events for the output itself and all its modes to shikane.
In your log shikane only receives events that say that one display is disabled (but not unplugged) and one is enabled.
The Finished events are missing.
And once you plug it back in shikane gets all the initial information again.
For all of its runtime shikane knew that 2 displays existed.
Which should not have been the case because you unplugged your monitor.
There is nothing shikane can do about this.
It sounds interesting as to why the Finished events are not sent to shikane.

[leiserfg]

I think this PR adds more complexity to the whole monitor management and it will be better to keep it closer to what it is in master and try to introduce the proper events, what do you think?

[vaxerski]

No, this system is generally more robust IMO.

Let me try to fix that, thank the shikane dev for me.

[leiserfg]

If you wanna try yourself you can use shikane with
env SHIKANE_LOG=trace shikane 2>&1 | grep Event::

.config/shikane/config.toml

[[profile]]
name = "laptop builtin"
   [[profile.output]]
   match = "eDP-1"
   enable = true
   scale = 2.0

[[profile]]
name = "nobuiltin + hp24"
   [[profile.output]]
   match = "eDP-1"
   enable = false

   [[profile.output]]
   search = "/.*"
   enable = true

[vaxerski]

yeah I managed, something is seriously wrong around monitor SPs

[vaxerski]

fixed, try now.

[leiserfg]

Now it works (I even see the Finished events).

[vaxerski]

great, merging then