layer-shell surfaces aren't capable of constraining the pointer.

[dmorilha]

Hello there, let me start by saying this is a very exciting time to be getting aware of sway! I went through the code and is very well written! nice!

I am an experienced Linux user and I am finally trying to move away from X11 on all my devices after realizing how better things are with wayland/sway.

While trying to mush wlroots examples together, I realized the layer-shell example doesn't go well with the pointer-constraint protocol. What I would like to do is to be able to lock the cursor if it goes beyond the viewport, and subsequently release it.

I tried various approaches on how to circumvent the problem but I was finally able to make the both nonstandard protocols work together when I removed this if statement on sway/input/cursor.c : https://github.com/swaywm/sway/blob/master/sway/input/cursor.c#L1294

I looks like overlays never have the "proper focus"... Is this a bug, or intentionally designed? Would I be able to constrain the cursor into an overlay in a different way?

[emersion]

We could extend the check in handle_pointer_constraint to lookup seat->focused_layer in addition to seat_get_focus.

[emersion]

Or introduce a new seat_get_focus_surface that returns a wlr_surface to simplify our logic.

[Jaakkonen]

I came across this behavior when trying to implement a virtual KVM using wlr-virtual-pointer-unstable-v1, virtual-keyboard-unstable, wlr-layer-shell-unstable-v1 and pointer-constraints-unstable-v1. Essentially making 1px wide windows on screen borders, detecting pointer enter and then locking pointer and emitting all captured pointer movement and keyboard movement to a connected device until a 1px window gets in focus on that side. However I could not get the pointer locked on the 1px layer surface I had possibly due to this behavior. I'll try make a PR adding support for this case so that this topic can be discussed about with concrete runnable code and possibly even merged :)

[feschber]

I came across this behavior when trying to implement a virtual KVM using wlr-virtual-pointer-unstable-v1, virtual-keyboard-unstable, wlr-layer-shell-unstable-v1 and pointer-constraints-unstable-v1. Essentially making 1px wide windows on screen borders, detecting pointer enter and then locking pointer and emitting all captured pointer movement and keyboard movement to a connected device until a 1px window gets in focus on that side. However I could not get the pointer locked on the 1px layer surface I had possibly due to this behavior. I'll try make a PR adding support for this case so that this topic can be discussed about with concrete runnable code and possibly even merged :)

I tried to do the same thing and here we are ... Any updates on this?