xwayland: listen to request_activate event

[yrhki]

When REAPER submenu is closed XCB_CLIENT_MESSAGE with type
NET_ACTIVE_WINDOW is sent to set focus to parent menu.

Closes: #6324

[emersion]

I don't understand why re-focusing is necessary here.

[emersion]

Giving focus to an unmapped surface doesn't make sense, the surface isn't shown on screen.

[yrhki]

@emersion I looked deeper into this and this seems to be proper solution.

[Xyene]

This looks like a proper fix, thanks for looking into it!

Out of curiosity, does this remove the need to do

sway/sway/desktop/xwayland.c

Lines 105 to 120 in 8daa632

	if (seat->wlr_seat->keyboard_state.focused_surface == xsurface->surface) {
	// This simply returns focus to the parent surface if there's one available.
	// This seems to handle JetBrains issues.
	if (xsurface->parent && xsurface->parent->surface
	&& wlr_xwayland_or_surface_wants_focus(xsurface->parent)) {
	seat_set_focus_surface(seat, xsurface->parent->surface, false);
	return;
	}
	// Restore focus
	struct sway_node *previous = seat_get_focus_inactive(seat, &root->node);
	if (previous) {
	// Hack to get seat to re-focus the return value of get_focus
	seat_set_focus(seat, NULL);
	seat_set_focus(seat, previous);
	}

? (In case it does, not asking you to make the change in this PR, just wondering if it does.)

[emersion]

We should make sure this doesn't allow an X11 app to steal focus from a Wayland app. IIRC we already have some kind of client matching elsewhere, maybe we can do something similar?

[yrhki]

@Xyene Removing

sway/sway/desktop/xwayland.c

Lines 108 to 112 in 8daa632

	if (xsurface->parent && xsurface->parent->surface
	&& wlr_xwayland_or_surface_wants_focus(xsurface->parent)) {
	seat_set_focus_surface(seat, xsurface->parent->surface, false);
	return;
	}

breaks menus in IntelliJ IDEA.

Also found that in IDEA set_hints event is called on sub-menu before unmap happens, but sway_xwayland_unmanaged dosen't have it. It would be helpful if wlr_signal_emit_safe would log when nothing is called.

[Xyene]

Thanks for testing.

Yeah, I think that doesn't sound unreasonable for a debug-level message, it seems likely this issue would have been found quicker if we had it.

[yrhki]

I think this will do it. Doesn't limiting to the same process apply here as well similar to xwm_handle_focus_in?

[emersion]

Doesn't limiting to the same process apply here as well similar to xwm_handle_focus_in?

Yup, that would be nice.

[emersion]

LGTM, thanks!

[salkin-mada]

Fantastic! Thanks so much @yrhki