1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
|
//! Pointer hit-testing against a window's titlebar/border/content.
//! Split out of the original single `manager.rs` - see `super` (`mod.rs`) for
//! `WindowManager`'s field definitions; everything here is plain `impl WindowManager`
//! methods, unchanged from before the split.
use super::*;
impl WindowManager {
// ---- Hit testing ------------------------------------------------------
/// Topmost window whose frame contains `(x, y)`, along with what part of
/// its titlebar/border was hit (button, drag area, resize edge).
///
/// Restricted to the current workspace, same as `visible_windows`/
/// `visible_windows_front_to_back` - a window on another workspace is
/// never minimized (that's a separate flag from "not currently shown"),
/// so without this a click landing on its old on-screen geometry hit
/// *that* window instead of whatever the user could actually see.
/// Reported live: clicking a window while a differently-workspaced one
/// happened to occupy the same screen coordinates sent the click to the
/// invisible one.
pub fn hit_test(&self, x: i32, y: i32) -> Option<(WindowId, TitlebarHit)> {
self.hit_test_with(x, y, |_, geometry| geometry)
}
/// Same as [`Self::hit_test`], but lets the caller substitute a
/// different rect than `w.geometry` for whichever window is being
/// tested - `geometry_for(id, w.geometry)` is called once per window in
/// the same topmost-first order, and its return value is what actually
/// gets tested instead of `w.geometry` directly.
///
/// This exists for exactly one reason: a backend that animates window
/// geometry (currently only the Wayland one, via `window_anims` in
/// `CompState`) draws the border/titlebar at the *interpolated* rect
/// every frame (`WindowAnim::current_rect`), but `w.geometry` here is
/// always the animation's *target* - core has no concept of animation
/// at all, deliberately (`Window.geometry` is meant to be the single
/// source of truth every other subsystem reads). Calling plain
/// `hit_test` during an active animation (toggling maximize/fullscreen,
/// a Snap-Layouts zone, or a new window's open-slide - see
/// `WindowManager::toggle_maximize`/`apply_snap_zone`/
/// `toggle_fullscreen` for where `anim_from` gets set) meant the
/// decoration/resize-margin hit-test used the window's *final* position
/// while the border was still visibly animating toward it - reported
/// live as "the border isn't always truly on the edge of the window",
/// i.e. hovering what you can see as the edge doesn't match what's
/// actually clickable there for as long as `animation_duration_ms`
/// (200ms by default) hasn't elapsed since the last toggle/snap/open.
/// Content clicks never had this problem - `space.map_element` already
/// maps the client's surface at the same interpolated rect the border
/// draws at (`state/geometry.rs::sync_geometry`), so `space.element_
/// under` and the border were already agreeing with each other; only
/// this compositor's own decoration hit-test was reading a different
/// number than what it was drawing on screen.
pub fn hit_test_with(&self, x: i32, y: i32, geometry_for: impl Fn(WindowId, Rect) -> Rect) -> Option<(WindowId, TitlebarHit)> {
for w in self.order.iter().rev().filter_map(|id| self.windows.get(id)) {
if w.minimized || w.workspace != self.current_workspace {
continue;
}
let margin = w.resize_margin.unwrap_or(self.resize_margin);
let geometry = geometry_for(w.id, w.geometry);
if let Some(hit) = ResizeEdge::hit_test(geometry, x, y, w.decorated, w.border_width, margin, self.theme.buttons_left, self.theme.button_order, w.is_dialog) {
return Some((w.id, hit));
}
// Not a titlebar/border/resize-margin hit on `w` - but if the
// point still falls inside `w`'s own plain content rect, `w`'s
// real, opaque content is what's actually drawn there (this is
// topmost-first order, so nothing checked so far is above it),
// and continuing the loop into a *lower* window's own border/
// resize zone at this same point would return a hit for
// something the user cannot see or reach - `w`'s content is
// in the way regardless of whether `w` itself claimed the
// point as one of its own edges. Reported live as being able
// to grab a resize edge, or trigger a titlebar-adjacent action,
// on a window fully covered by another one on top of it.
// Content clicks (the content-hit branch in `input.rs`) are
// unaffected - they already resolve via `Space::element_under`,
// smithay's own real Z-order, which never had this gap.
if geometry.contains_point(x, y) {
return None;
}
}
None
}
/// Topmost non-minimised window on the current workspace containing a
/// point, ignoring decorations. Used for modifier+drag, where the grab
/// applies anywhere in the window rather than only on the titlebar
/// (`hit_test`). See `hit_test`'s doc comment for why the workspace
/// check is load-bearing, not redundant with `minimized`.
pub fn window_at(&self, x: i32, y: i32) -> Option<WindowId> {
self.order
.iter()
.rev()
.filter_map(|id| self.windows.get(id))
.find(|w| !w.minimized && w.workspace == self.current_workspace && w.geometry.contains_point(x, y))
.map(|w| w.id)
}
/// The corner of `id` nearest a point, for modifier+right-drag resize:
/// grabbing the closest corner is what makes the gesture feel like it
/// pulls the edge you aimed at (matching Hyprland's `resizewindow`).
pub fn nearest_corner(&self, id: WindowId, x: i32, y: i32) -> ResizeEdge {
let Some(w) = self.windows.get(&id) else { return ResizeEdge::BottomRight };
let (cx, cy) = w.geometry.center();
match (x < cx, y < cy) {
(true, true) => ResizeEdge::TopLeft,
(false, true) => ResizeEdge::TopRight,
(true, false) => ResizeEdge::BottomLeft,
(false, false) => ResizeEdge::BottomRight,
}
}
}
|