495968fca5cea600f994b949e988493346d9ac8c
4
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 admin
|
53e7ccc46e |
Keep the GLES path when the scale arrives first, stop re-rendering what does not change, and let the compositor skip what the surface covers
Measured on a Librem 5 (Cortex-A53, GC7000Lite, 720×1440 at scale 2) with a frame trace on both sides of the socket: the shell's main surface repainted in 110 to 136 ms, quick settings drew at 20 fps with the GPU saturated, and every panel open uploaded its wallpaper and icons again. Most of it was the toolkit doing work the frame did not need. Backend choice. When the compositor announces the output scale before the first `configure`, `scale_factor_changed` allocated an SHM pool for the surface and `on_configure` took that pool as the sign that the software path had been chosen, locking the surface to tiny-skia for its whole life. Every ltk main surface on a scale-2 phone ran in software while its overlays, which inherit the scale from the main and never receive that event, ran on the GPU. The scale handler now only reallocates a pool that already exists; before the first configure neither path is chosen. Wallpapers. `WallpaperBundle::for_size` returned the crop at the source resolution, a 1103×2160 slice of a 4K image for a 720×1440 panel, 10 MB uploaded per texture and a hundred milliseconds each time. The crop is now downscaled to the surface (bilinear, once per size, cached), also for landscape surfaces smaller than the source. Viewport. `Element::Viewport` created a fresh sub-canvas every frame, on GLES a framebuffer, a 2048² glyph atlas and empty image and glyph caches, so a panel clipped to a height that animates re-uploaded its wallpaper and re-rasterised every glyph on every frame. Without a bottom fade the clip is now a scissor on the parent canvas (a mask on the software backend): no offscreen pass, no blit, and the child's textures stay in the parent's caches. `local_viewport` is honoured through a temporary override of the layout viewport. Viewports with a fade keep their sub-canvas, which, like the scroll sub-canvases, is now kept across frames and resized in place instead of reallocated when its size changes. Caches. The image texture cache moves from the canvas to the GL context: every canvas and sub-canvas shares one content-keyed LRU (32 MiB), so an overlay or subsurface that is destroyed and recreated finds its images resident. A GLES canvas dropped with an overlay is pooled in `AppData::canvas_pool` and handed to the next overlay, with its compiled programs and glyph atlas, which took 130 to 290 ms to build on that GPU. Subsurface slots keep their scroll and viewport sub-canvases between rasters. Surfaces. A `container.surface()` with shadows costs a Gaussian drop-shadow pass, a fill and one pass per inset shadow, all full-quad fragment shaders; a dozen glass cards were 40 ms of GPU per frame for chrome that never changes. `Canvas::fill_surface` on GLES now renders the chrome once per size, corners and style into a texture held by the context cache (`gles_render/surface_cache.rs`, 16 MiB LRU) and blits it, drawing directly only for surfaces with `PlusLighter` or `Overlay` insets, which depend on the backdrop. Opaque regions. ltk never called `wl_surface.set_opaque_region`, so the compositor blended every ltk surface and repainted everything underneath a full-screen homescreen. A surface whose `background_color` is opaque now declares itself opaque, and the new `App::overlay_opaque_region( id )` lets an overlay declare the physical rects it paints opaque, such as a panel column over an opaque wallpaper. `sync_opaque_region` sends the rects only when they change. Frame pacing. `eglSwapBuffers` blocked 13 to 37 ms per frame waiting for Mesa's own throttle callback; ltk paces itself on `wl_surface.frame`, so `make_current` sets the swap interval to 0. A subsurface move is now applied only when its parent has no frame callback outstanding, and the parent commit that applies it requests the next callback: a drag commits once per vblank with the latest finger position instead of once per input sample, without the app throttling samples itself. While the app reports `subsurface_motion_only`, the main's callback, and the callback of an overlay that hosts a subsurface, no longer force a full repaint; they mark the surface and the reconcile keeps the cadence with a bare commit when nothing moved. Swipe releases arm a settle the same way instead of dirtying every cache. Two hooks, `on_overlay_frame_committed` and `on_subsurface_committed`, let an app hand a panel over between its host and a sliding subsurface on real commits rather than loop ticks. Lines. A diagonal `draw_line` on GLES stamped one `fill_rect` per pixel of length, 190 draw calls for a spinner and over 2000 for a pattern-lock stroke; it is now a single rotated quad through the rect shader, with round caps and antialiased edges. A pixel test covers the scissor viewport, `for_size` gets a downscale test, and `docs/architecture.md`, `docs/widgets.md` and `docs/backends.md` describe the new hook, the viewport paths, the shared caches and the line rasteriser. |
||
|
admin
|
58b8dd7b90 |
Focus requests reach widgets clipped by a scroll viewport, and stop retrying after a few frames
A scroll viewport keeps in `widget_rects` only the widgets at least partly inside its visible area (`draw/layout.rs`, the translation of the sub-canvas rects to surface space). A focus request whose target sits below the fold therefore never finds it: `App::take_focus_request` was honoured only for widgets already laid out on screen. The Wi-Fi page of settings is one such case by construction: tapping a secured network inserts the password form under the row and asks for focus on the field, so for a row in the lower part of the list the field is born outside the viewport and the request misses. Worse, the miss branch parked the id in `focus_retry`, marked the view dirty and requested a redraw on every iteration, so the application rebuilt and repainted at the compositor's frame rate for as long as the widget stayed hidden. The layout pass now records the widgets it drops for carrying a `WidgetId`: `DrawCtx::scroll_clipped` holds `(id, scroll flat_idx, rect)` with the rect in the scroll's content coordinates, the same frame the navigable-item list uses, and `draw_frame` hands it to `FrameState`. When a focus request misses, `AppData::reveal_clipped_widget` looks the id up in that list on the main surface and the overlays, finds the viewport in `scroll_rects`, and moves its offset on the axes the scroll allows so the rect lands inside; the next frame lays the widget out and the retry focuses it. Nested scrolls resolve one viewport per frame. The rule that computes the new offset is `scroll::offset_into_view`, extracted from the keyboard navigation in `input/keyboard/nav.rs`, which now calls it instead of carrying its own copy; unit tests cover the three cases. `focus_retry` carries a budget of drawn frames (`FOCUS_RETRY_FRAMES`, four), decremented only on iterations that actually drew, so a request made while a frame callback is pending does not lose its chance. A fresh request from the application replaces a pending retry with a full budget. When the budget runs out the request is dropped, which ends the redraw loop for a widget that never appears. Documented in the `take_focus_request` doc comment, the cookbook and the changelog; Debian revision 0.3.0-3. |
||
|
admin
|
cfa0faff26 |
ltk: subsurface slides over overlays, axis-locked swipes, physical-space layout, touch reset on resume
Subsurfaces can now be parented to an overlay surface, not just the main surface. `SubsurfaceSpec` gains `parent: SubsurfaceParent { Main, Overlay(id) }` so a sliding panel can ride above app windows the way an overlay panel does, and an optional `gpu: bool` so content that uses the `surface-panel` backdrop-filter glass (a GLES-only pass) keeps it while sliding instead of dropping to the software rasteriser. `reconcile_subsurfaces` resolves each spec's parent independently — skipping an `Overlay` parent that is absent, unconfigured, or zero-sized — tracks the rastered size on the slot, and commits each touched parent once per frame. The ~14 GLES shader programs are compiled once into a shared `AppData::subsurface_gles_canvas` reused across every subsurface, so a lazily re-created sliding panel never recompiles them (hundreds of ms on a mobile GPU).
Vertical and horizontal swipes are now mutually exclusive: a gesture locks onto its dominant axis within the first 8 px of travel (new `SwipeAxis`) and ignores the perpendicular axis for the rest of the gesture, so a vertical swipe that drifts sideways no longer also drives the pager (and vice-versa). The upward swipe progress is no longer clamped at 1.0 — follow-the-finger panels can keep tracking the finger past the commit threshold — and a release below threshold delivers a final `progress = 0.0` cancellation pulse. A vertical swipe also no longer re-rasters the full-screen main surface on every motion event: only the overlays it drives are refreshed, while the horizontal pager (which does move the main surface) still redraws it. Re-rastering the main on every frame of a vertical drag was wasted work that stalled the loop and made the gesture feel laggy to start on a slow GPU.
Layout-affecting `Length` values (widths, paddings, gaps, widget sizes, including `Vw` / `Vh`) now resolve against the physical viewport via the new `Canvas::viewport_layout()` — the space the layout tree is actually computed in — so a `Vw(100)` fills the surface on a HiDPI (scale 2) output instead of covering half of it. Font sizes are unchanged: they still resolve against the logical viewport and are scaled at raster time. Switched column, row, spacer, wrap_grid, container, image, and vslider over to it; `img_widget` and `vslider` now document `Length::vw` / `vh` sizing and the showcase example demonstrates a viewport-relative image.
Touch gesture state is reset when the touch capability is added or removed — suspend / resume on devices that power the touchscreen down. A yanked capability never delivers the pending `up` / `cancel`, so the shared `reset_touch_state()` (also used by the `wl_touch.cancel` handler) drops the stranded `primary_touch_id` / slot state across the main surface and every overlay, keeping the first post-resume gesture clean. Also drops an accidental duplicate `on_scale_changed` call from the scale-change handler.
|
||
|
admin
|
042652ec73 |
windows: centralise input focus on a single FocusTarget, raise+focus on every window interaction
Focus was scattered across three stores — `Layouts::focus`, `Layers::focus` and `Windows::activated` — plus a hack that smuggled X11 surfaces through `Layers::focus`, with the keyboard target derived each frame from a `layers.focus.or(layout)` precedence. As a result "focused", "activated" and "has the keyboard" could drift apart: a window could be raised without taking the keyboard, the recurring "click a console window, then click it again before you can actually type" symptom.
New `windows/focus.rs` with `FocusTarget { Layout(Weak<Window>), Layer(WlSurface, app_id), X11(WlSurface) }`. `Windows` now holds one `focus_target: Option<FocusTarget>` as the single source of truth; `activated` survives only as a reconciliation cache for the xdg `Activated` state and is never set by hand. `set_focus` takes one `Option<FocusTarget>` and is the only entry point for changing focus: it brings the target to the front (`z_promote` + desktop stacking), refreshes the MRU and stores the target, and keyboard focus plus activation are applied from it on the next `focus()`. `focus()` resolves a baseline (the active layout's window, which carries `Activated`) and an overlay (a focused layer/X11 surface that takes the keyboard over the baseline without de-activating it), mirroring the old precedence but from one field, so keyboard focus follows focus by construction. `Layers::focus` is removed and X11 no longer rides inside it; the exclusive-keyboard layer path and `reap_layer` drive `focus_target` directly, and every click/touch/activation site collapses to `set_focus(Some(FocusTarget::…))`.
surface_at popup tie-break. When a layout hit is a popup or subsurface, its `wl_surface` is not tracked in `toplevel_z_order`, so the layout-vs-X11 z compare resolved to `None` and the X window underneath wrongly won — clicking a GTK menu (e.g. Firefox's) over an X11 window pressed the X window instead of the menu item. The tie-break now compares the owning toplevel's root z, taken from the hit's `FocusTarget::Layout` weak, so the popup's parent stacking decides.
Raise and focus on every window interaction, matching standard desktop behaviour. Client-driven (un)maximize, the SSD maximize/minimize/close buttons and the maximize keybind now bring the window to the front in both the render z-order and the hit-test stacking through a new `raise_toplevel` helper; previously only the stacking was updated, so an unmaximised window could stay visually behind another. Starting a move or resize — including a press on the resize edge — now routes through `set_focus`, so the window comes to the front and takes the keyboard before the drag instead of being resized or moved while it stays behind.
Input robustness around grabs and drags. A press under an active pointer grab (Wayland popup, drag-and-drop or explicit client grab) is routed to the grab instead of forge's own surface_at handling, so a click cannot leak to a window underneath a grabbing popup and a click outside the popup dismisses it cleanly. Only the press is blocked, never the release, so an in-flight move/resize drag or topbar swipe always reaches its end handler and can never stay glued to the cursor if a grab appears mid-gesture. `on_touch_up` now ends any active move/resize drag before its other early-return paths (armed SSD button, topbar swipe, app switcher) and drops a stale armed button, fixing a race where pressing a second, overlapping window left the drag running forever after release.
fix(input): keep the primary touch slot when migrating a drag across surfaces
A touch drag started on an overlay that hides mid-gesture (e.g. the app launcher when dropping an icon onto the dock or the homescreen) froze: it neither moved nor dropped. Destroying the origin surface migrated the drag state (long_press_fired / long_press_origin) and touch_focus to the main surface, but not its primary_touch_id; subsequent touch motion/up events then fell through the auxiliary path and never reached the gesture machine, leaving on_drag_move and on_drop uncalled.
reconcile_overlays and discard_overlay now adopt the destroyed overlay's primary_touch_id when a drag is in flight, so the rest of the touch sequence keeps driving the main surface. Touch-only; the mouse already migrated correctly via pointer_focus.
ltk: add a chassis module for full-screen ambient surfaces
New `src/chassis.rs`, re-exported from the crate root, gathers the scaffolding that every full-screen ambient surface — greeter, lock screen, kiosk — otherwise repeats by hand over the existing theme and `WallpaperBundle` primitives. `set_default_theme(mode)` finds, installs and activates the `default` theme document, returning the failure message instead of exiting so the caller decides how to abort. `theme_logo_rgba(size)` decodes the active theme's horizontal logo to RGBA, and `theme_icon_tinted(name, size, tint)` loads a symbolic theme icon and tints it. `branding_bundle_or_solid(name)` resolves a theme branding image (`"wallpaper"`, `"lockscreen"`, …) to a `WallpaperBundle`, falling back to a solid fill of the palette background when the theme ships none; `wallpaper_bundle_or_solid()` is the `"wallpaper"` convenience. `backdrop(content, &wallpaper, w, h)` stacks content over the wallpaper resolved for the surface size. No new capability — thin convenience over what the theme module and `WallpaperBundle` already expose — but it removes the per-application `load_theme_logo` / `build_wallpaper_bundle` / theme-bootstrap duplication.
ltk: animatable, input-transparent child surfaces via App::subsurfaces()
New `App::subsurfaces() -> Vec<SubsurfaceSpec<Msg>>` (default empty) describes input-transparent child surfaces composited over the main surface, with `SubsurfaceSpec { id, view, x, y, content_version }` and the stable `SubsurfaceId`. The motivating use is a slide/reveal that tracks a finger without the per-frame full-screen CPU re-raster a single-surface opacity or translate would cost: the content buffer is rasterised once and the compositor moves it.
`SubcompositorState` is bound in `event_loop/run.rs` from the compositor's `wl_compositor` (absent → `App::subsurfaces` silently degrades to none); `delegate_subcompositor!` is added on `AppData`, which gains a `subcompositor` binding and a `subsurfaces` map. New `event_loop/subsurface.rs` reconciles the live subsurfaces against the specs: each spec becomes a `wl_subsurface` sized to the main surface with an empty input region, so all pointer/touch falls through to the parent and the host keeps a single gesture/input model. The content — its own SHM pool and `Canvas`, drawn through the existing `DrawCtx` / `layout_and_draw` path — is rasterised only when the surface size or the spec's `content_version` changes; a position-only change emits `wl_subsurface.set_position` and commits the child surface, then a bare parent commit for placement. Committing the child is deliberate: desync subsurface state (the position) is applied on the child surface's own commit, not the parent's, so without it the move is queued but never lands. Positions are given in layout (physical) pixels and divided by the surface scale for the logical `set_position`.
The reconcile pass runs on every run-loop iteration rather than being gated behind a main-surface redraw, so a finger-driven move repositions at input-event rate, decoupled from the frame-callback cadence that paces full redraws — without this the subsurface only moved when the main surface happened to redraw, so a drag over a static background froze the panel in place.
|