| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Real web research (KDE's own source tree, current as of Plasma
6.6.5/2026): com.canonical.AppMenu.Registrar + dbusmenu is still the
current, unreplaced global-menu mechanism in KDE Plasma 6, and generic
Qt apps still export via the same QGenericUnixTheme path since Qt 5.7 --
exactly what srdwm's own appmenu_registrar.rs/appmenu.rs already
implement. No newer protocol to catch up to, no code gap found. srdwm's
own scope (discovery/registration) is correctly split from AGS's
(rendering) - see the FEATURE_GAP.md entry from the previous commit.
|
|
Two research-only entries, no code changes:
- docs/FEATURE_GAP.md gains a "vs. full desktop environments" section
(KDE/GNOME/XFCE/macOS/Windows), requested directly and distinct from
the file's existing tiling-WM (niri/sway/Hyprland) comparison. Verified
rather than assumed: real app-to-app clipboard already works
(delegate_data_device!), drag-and-drop between real windows already
works; genuine gaps are compositor-level blur-behind, the
already-tracked fractional-scale wl_pointer bug's real-world cost, and
no PipeWire screencasting - with an explicit line drawn between
srdwm's own scope and AGS's (notifications, applets, alt-tab UI,
screenshot tooling are shell concerns, not compositor gaps).
- docs/TODO.md: researched "different titlebars, non-traffic-light,
right side, especially firefox/chrome" and found the requested system
already exists and is already documented (button_style, button_side/
order, glyph-always, and a real, already-correct xdg-decoration
negotiation with Firefox's own specific behavior already documented).
One real, unverified gap found via actual web research into
Chromium's own Wayland decoration history: likely_draws_own_titlebar
only matches org.gnome.* today, and Chromium's xdg-decoration support
has a documented history of inconsistency vs Firefox/GTK. Deliberately
not blind-fixed - forcing decorated=false for Chrome would be worse
than doing nothing if it already negotiates correctly; needs a live
screenshot check with a real Chrome/Chromium install first.
|
|
Reported live: "looks weird and unpolished... need a lot more items".
Compared directly against the exact AGS reference this project's own
menu rebuild already targets rather than guessing:
- Highlighted rows used a flat, fully-saturated fill instead of the
reference's subtle 22%-accent-into-background wash. New decoration::
color::mix_rgb (channel-wise linear blend, generalizing brighten/
darken's fixed-target blends to an arbitrary second colour/ratio) lets
render_context_menu reproduce that same ratio.
- Every separator row was a label string of Unicode box-drawing
characters rendered as text glyphs, which render inconsistently at
small sizes - a label that's entirely U+2500 now draws a real 1px
hairline instead; a label that mixes it with real text ("--- Move to
Workspace ---", a deliberate section-header convention) still renders
as text, unchanged.
- "Select All" added to the bare-desktop menu, the one action every
mainstream desktop's own menu offers that this one lacked.
New tests needed real care: the panel's own rounded-corner distance
field softens alpha within its radius of any canvas edge, not just the
visible corners, so a naive full-row pixel scan against bg picked that
up as a false positive on the first attempt - fixed by scanning only
rows/columns confirmed (via a throwaway debug dump) to sit inside the
panel's genuinely flat interior.
Full workspace build/test/clippy clean, built and installed. Real
submenus and per-row icons remain real, separate scope - this
project's floating-menu UI has no nested-panel concept yet.
|
|
Reported live: "try move desktop items all at once somewhere else"
didn't work. Two compounding bugs, both real: CompState::
desktop_icon_drag only ever tracked one icon id, and the click handler
that starts a drag unconditionally collapsed any existing multi-
selection down to just the grabbed icon before the drag even began.
desktop_icon_drag is now Option<DesktopIconDrag> (crates/wayland/src/
desktop_icons.rs, new type): a grab offset, the grabbed icon's own live
position, and a members list - every currently-selected icon (the
grabbed one included), each a fixed offset from the grabbed icon's own
top-left at drag start, so the group moves as one rigid unit.
input/pointer.rs's click handler now only resets to single-selection
when the grabbed icon isn't already part of the current selection --
grabbing one inside an existing multi-selection keeps the whole group
selected and dragging, matching Windows/GNOME/macOS/KDE convention.
end_desktop_icon_drag snaps every dragged icon to its own nearest free
grid cell independently, tracking newly-claimed cells across the group
so two icons landing near each other never claim the same one.
Full workspace build/test/clippy clean, built and installed. Not unit-
testable (this module has no CompState test fixture for its own
selection/drag logic, an already-documented, accepted gap) - needs a
live drag to confirm.
|
|
Two independent pieces landed together this pass - both real, both
scoped, see docs/TODO.md for the full narrative on each:
Fake monitors: a genuinely independent, additional wl_output with no DRM
connector/CRTC behind it at all - distinct from srd.monitor.split
(divides one real output's own placement rectangle). Researched niri's
own Headless backend first (cloned at ~/reference-wms/niri): its render()
never actually composites anything, a no-render stub for that project's
test suite only. This one is real: it renders whatever is placed on it,
on demand, whenever a zwlr_screencopy_manager_v1 client asks for a frame.
New crates/wayland/src/udev/virtual_heads.rs (create/remove a real Output
+ global, render-on-demand for screencopy, integrated into platform.rs's
monitors() as a genuine srdwm_core::Monitor so core placement/workspace
code needs zero special-casing). New IPC/CLI: srd dispatch create
fake-monitor <name> <w>x<h> / remove fake-monitor <name>. Core-side
request queue in crates/core/src/manager/fake_monitor.rs.
Placement bug, root-caused and fixed: every new window opened alone
landed in the exact same spot, "not at all like Windows" (reported live).
SmartPlacement::place tried a grid cell first, and grid's own cell count
is existing.len() + 1 - with nothing else open (opening one app at a
time, the ordinary case), that's always 1, so a 1x1 grid returns the same
single cell forever regardless of session history. Cascade had the same
bug in a second form (its own step was existing.len() % max_steps, also
always 0 with nothing open). Fixed: WindowManager::next_cascade_step (a
Cell - add_window's own target_monitor stays borrowed across the call)
advances on every real placement and is never reset by a window closing;
place() now skips grid entirely when nothing else is open, going
straight to cascade, since grid's real job (dividing space among
concurrent windows) has nothing to divide when there's no concurrency.
Full workspace build/test/clippy clean (223 core / 141 wayland / 29
platform / 24 ctl / 28 config / 10 x11 tests, 0 failed, 0 clippy
warnings), built and installed.
|
|
docs/TODO.md is this shift's single consolidated pending-work list (see
its own header for why it exists alongside PANEL_SUPPORT_TODO.md/
SESSION_HANDOFF.md rather than replacing them) - every commit in this
batch has its own dated entry there with the full root-cause/
verification narrative. docs/DEFAULTS.md corrected against the real
config engine and extended for every new general.* key this shift added
(aspect_ratio rule action, phone_mode). docs/FEATURE_GAP.md is a new
survey against niri/Hyprland/sway, requested directly. docs/
IMPLEMENTATION_STATUS.md and docs/PRIOR_ART.md updated to match.
Cargo.lock reflects the new resvg/usvg/tiny-skia dependencies (real
icon-theme SVG rendering).
|
|
Standalone (not a workspace member, same reasoning as the existing
tools/toplevel-activate), built to live-verify the Wayland crate's new
pinned-virtual-pointer delivery (crates/wayland/src/virtual_pointer.rs)
against a real client rather than reading the source: prints its own
pid so an external script can pin it via srd dispatch pin input, then
drags from one point to another on a stdin signal.
Not yet run against a live nested instance - launching one hits this
project's own nightshift deny-list guard against a bare nested-compositor
invocation, parked rather than worked around; see docs/TODO.md.
|
|
XWayland stability, GPU rendering, and multi-cursor Phase 2
The bulk of a multi-session shift's real work landed in crates/wayland.
Full root-cause/verification narrative for every item below lives in
docs/TODO.md (each has its own dated entry); this is the summary:
Desktop shell:
- Real desktop icons v2 (state/desktop_icons.rs, desktop_icons.rs):
fixed origin-baked-before-the-bar-connects, fixed-icon sort order, and
a proper Rename/Delete-to-Trash menu (window_memory.rs backs the
rename-persistence side). Rubber-band marquee multi-select.
- icon_theme.rs: real freedesktop icon-theme lookup (inherits chain,
hicolor fallback) rendering actual theme SVGs via resvg/tiny-skia,
replacing the hand-drawn placeholder glyphs.
- Context/desktop menus (decoration.rs, desktop_menu.rs, state/menu.rs)
rebuilt to match the project's own AGS panel styling: rounded floating
panel, tinted-fill row highlight, real separators, a much fuller
titlebar window-menu action set.
Layer-shell / multi-monitor:
- Layer-shell hit-testing and render positioning (input/pointer.rs,
udev/render.rs's element placement) now correctly convert LayerMap's
logical geometry into physical pixels on a fractionally-scaled output
- root cause of a bottom-anchored dock being unclickable and
unpainted while a top-anchored bar on the same output worked.
udev/outputs.rs's relayout_outputs gained the same physical/logical
split for cross-output positioning, now backed by a real unit test
(next_logical_x) built from the original measured incident numbers.
- state/geometry.rs: a window's border/decoration no longer briefly
clips when moved between differently-scaled monitors mid-drag.
XWayland / stability:
- xwayland.rs, udev/session.rs, udev/platform.rs: fixed a 100%-
reproducible cold-start XKEYBOARD crash-loop (XWayland's own stdin
inherited a real, already-owned VT; env passthrough and idle-callback
spawn timing were both real, independent gaps) that had silently taken
down all X11-app support and the global-menu registrar every session.
- state/toplevel.rs, state/lifecycle.rs: XWayland dialog detection via
WM_TRANSIENT_FOR, not just a native xdg_toplevel parent.
Rendering:
- udev/render.rs, decoration.rs: real GPU (GBM+EGL+DrmCompositor)
window-content and cursor rendering on the udev backend, falling back
to the untouched Pixman path automatically on any init failure.
- decoration/tests.rs, state/mod.rs: rounded-corner/border fixes for
interactive resize lag and cross-monitor moves.
Multi-cursor Phase 2 (virtual_pointer.rs, new; state/mod.rs, udev/
platform.rs, winit/nested_platform.rs): pins a zwlr_virtual_pointer_
unstable_v1 object to a specific window, bypassing the shared seat/
focus/pointer_pos path entirely via hand-rolled wl_pointer.enter/motion/
button/frame/leave against every WlPointer the target client has bound
(PointerHandle::client_pointers). Lets an agent operate one window while
a human uses another, genuinely simultaneously, with zero client
cooperation and no second wl_seat (confirmed a dead end: real clients
only ever bind the first seat advertised).
Full workspace build/test/clippy clean.
|
|
The Lua and IPC/CLI surface for the three new core primitives
(crates/core: aspect_ratio rule action, general.phone_mode, virtual-
pointer window pinning):
- srd.rule(..., { aspect_ratio = "9:16" }): parses a "W:H" string into a
validated (u32, u32), rejecting a malformed value as a real Lua error
at config-load time rather than silently ignoring it.
- general.phone_mode config default, plus srd set phone_mode <bool> for
the live equivalent (same shape as animations/shadows/rounded_corners).
- pin_input IPC dispatch ({"cmd":"pin_input","pid":<pid>,"id":<window
id>}, id omitted to unpin) and its CLI surface, srd dispatch pin input
<pid> <window-id> / unpin input <pid>. Keyed by the owning client's
process id, not an opaque per-object id nothing outside the Wayland
backend could ever learn - a controlling tool already knows its own
pid for free.
See docs/TODO.md for the full design reasoning behind each of these.
|
|
Several independent, real pieces landed in crates/core this shift - see
docs/TODO.md for each one's full root-cause/verification narrative:
- "Primary" monitor is now picked by which head sits at physical (0, 0)
(the user's own configured anchor), not whichever connector DRM
happened to probe first - fixes desktop icons and new-window placement
landing on the wrong monitor depending on hotplug/probe order.
- A new window's target monitor now prioritizes the pointer's own current
monitor over the last-focused window's monitor, which goes stale the
moment the user's attention moves to empty desktop, a panel, or a dock.
- aspect_ratio window-rule action ("W:H") plus ResizeEdge::apply_aspect_
ratio: holds a floating window's aspect ratio through an interactive
resize. The real, scoped "phone monitor" primitive - matches any VM/
emulator/scrcpy window by app_id, nothing Android- or VM-specific here.
- general.phone_mode (WindowManager::phone_mode): a new window defaults
to maximized instead of floating/tiled small, unless a rule explicitly
floats it or sets maximized - the one placement default a phone-shaped
screen actually needs. Exposed read-only via IPC so a shell panel can
adapt its own chrome to the same signal.
- input_pin.rs: the core half of pinning a virtual pointer to a specific
window (Multi-cursor Phase 2) - a backend-agnostic request queue,
same cross-boundary shape output_position_requests/lock_requested
already use, since core has no real Wayland protocol object to reach
into itself.
Full workspace test suite covers all of the above (aspect-ratio resize
math for every edge case, phone-mode default-vs-rule-override behavior,
the pin-input request queue, the monitor-picking fixes).
|
|
Closes the one real gap an X11/Wayland feature-parity audit found this
session (desktop icons, window-position memory, and static exclusive-zone
reservation were already shared or Wayland-only by nature - see
docs/TODO.md's own audit entry for the full breakdown).
MenuAction/ContextMenu (row set, labels, row_at hit-testing) move from
crates/wayland/src/context_menu.rs into crates/core/src/context_menu.rs --
pure state and geometry with nothing Wayland-specific in it, so X11
needing the same rows is shared data, not duplicated logic. The Wayland
crate's own context_menu.rs is now a one-line re-export so every existing
crate::context_menu::... call site keeps working unchanged.
X11 has no compositor-level input dispatch to intercept every click the
way Wayland's input/pointer.rs does, so the X11 side
(crates/x11/src/platform/context_menu.rs, new) draws the menu into its own
small override-redirect popup window and grabs the pointer for the
duration so a click anywhere dismisses it, matching the Wayland backend's
own convention. events.rs's ButtonPress handler now reads the real button
number instead of hardcoding every press as a left click - a real latent
bug (right-clicking a titlebar button would have silently performed its
left-click action).
Live-verified end to end in an isolated Xvfb + srdwm --x11 instance: full
row set including the workspace picker, Minimize runs and closes the
menu, a second window's menu dismisses cleanly on outside click, normal
focus/click behaviour continues working afterward.
See docs/TODO.md for the full investigation and verification narrative.
|
|
The Rust rewrite (crates/) has fully superseded it - keeping both around
was actively misleading (main looked like it still shipped a C++ build),
and nothing here still depends on it.
|
|
Live-confirmed via a controlled test (move a window between differently-
scaled monitors, screenshot immediately after vs. a couple of minutes
later): the border briefly shows clipped/missing right after a
cross-monitor tiling move, then self-corrects on a later redraw.
Working hypothesis recorded (client configure/resize/commit round-trip
lagging the compositor's own already-updated model, likely wider on a
cross-scale move than a same-monitor tiling swap), but not confirmed --
reproduction via srd dispatch move window proved inconsistent (its
direction semantics swap within a monitor as often as they cross one),
and a live mouse-drag can't be synthesized here to test directly.
Not a corruption risk: an earlier resize-lag fix already
bounds every border/titlebar crop against the decoration buffer's real
last-built size, so the worst case is a stale/incomplete frame, never an
out-of-bounds read.
|
|
Requested directly: default icons "look very rudimentary... make slightly
blue as well, polished look".
Two changes to decoration::render_desktop_icon and its five draw_*_glyph
helpers:
- New fill_rounded_rect primitive: softly rounded corners (the same
clamp-then-distance smoothstep construction rounded_corners_pixman::
apply_corner_mask already established for content masking, not a new
technique) plus a vertical top-to-bottom gradient instead of one flat
fill - the same light-source cue buttons.rs's own glossy_shade uses
for the titlebar dots, as a plain linear gradient here. Applied to each
glyph's main body shape; small details (folder tab edge, computer
stand, trash ridges, home roof) stay flat/sharp.
- A dedicated ICON_COLOR constant (a clean mid-blue) instead of reading
theme.titlebar_fg_focused - that field is whatever the user's own
titlebar accent happens to be configured to, which could be any
colour; these glyphs want a consistent, recognisable blue palette of
their own, independent of theme.
Build/test/clippy already verified clean as part of the layer-shell scale
fix commit just before this one (same source tree, installed together).
|
|
Reported live, in stages: general input sluggishness, then specifically
dock/bar buttons not responding on the secondary monitor. Root-caused
jointly with a peer session (dotfiles-16), who independently instrumented
AGS itself (both bar and dock report correct visible/realized/revealed
state - the client is asking for the right thing) and srdwm's own
layer_hit_test log (the dock received zero hits across ~40 minutes while
the same output's wallpaper and bar took hundreds).
Confirmed against smithay 0.7.0's own source (desktop/wayland/layer.rs::
arrange): LayerMap::arrange() divides the output's physical mode by its
own scale before arranging layers, so LayerMap::layer_geometry() is
logical, not physical. Two call sites used it as physical, this
compositor's convention everywhere else:
- input/layers.rs::layer_surface_under_layers compared the physical
pointer position directly against logical layer geometry. On a
sub-1.0 scale output, logical space is larger than physical, so a
bottom-anchored dock's rect sat entirely past the pointer's reachable
range - permanently unclickable. A top-anchored bar only lost its own
right-hand end, which is what made this look like "the dock is
broken" rather than a scale bug affecting every layer surface there.
- elements.rs::output_layer_elements pushed the same logical position
straight into the physical framebuffer - for the dock, past the
bottom edge entirely, painting nothing.
Both fixed the same way udev/platform.rs::monitors() and udev/outputs.rs
already fix the identical unit mismatch for usable-area computation
(existing precedent, not a new technique): multiply by output.
current_scale().fractional_scale(), rounding to the nearest physical
pixel, before use.
Also removed a temporary per-pointer-motion-event diagnostic log in
layer_hit_test, still live from an earlier debugging session and
explicitly marked for removal but never removed - a real, measurable
cost on the hot input path, likely the direct cause of the separately
reported general slowness.
Full workspace test suite and clippy clean.
|
|
Live testing found v1 genuinely broken, not just rough:
1. Icons weren't rendering reliably at all - ensure_desktop_icons only
ever computed the grid's origin once, on whichever render pass
happened to be first. AGS's own top bar registers its exclusive zone
after that first pass, so origin got permanently baked in at the
pre-bar geometry. Confirmed live via a temporary diagnostic log.
Fixed by re-deriving origin from the primary monitor's current
geometry on every call instead of just the first.
2. Fixed icons (Home/Computer/Trash) always sorted before real files --
confirmed wrong via direct question. The whole list now sorts
alphabetically by label, case-insensitive, fixed icons included.
3. "Set as Wallpaper" was the wrong feature: removed entirely
(DesktopMenuAction::SetWallpaper, general.wallpaper_command,
is_image_path). The user wants that handled by their real file
manager once opened, not reimplemented here.
Also adds real menu functionality per "where are all the options":
Rename (inline text edit, new CompState::renaming_icon field and
keyboard redirect mirroring NativeLock::password's existing precedent),
Delete (moves to ~/.local/share/Trash per the freedesktop.org spec, new
trash.rs module, same-filesystem case, no confirmation - this is the
reversible move-to-trash, not a permanent delete), Empty Trash on the
Trash icon, and Open Terminal Here / Open in File Manager on the
bare-desktop menu (new general.terminal config key).
133 wayland-crate tests (up from 106), full workspace build and clippy
clean.
|
|
mid-gesture
Reported live: moving a window onto the other monitor "looks very messed
up". This machine's two real monitors have genuinely different scales
(eDP-1 at 1.0, HDMI-A-1 at ~0.843) - the exact condition needed to expose
this.
WindowManager::update_drag/update_resize only corrected w.monitor once,
at end_drag (update_resize never corrected it at all, not even at the
end) - but state/geometry.rs::sync_geometry reads that field on every
motion tick to pick which monitor's scale converts the client's physical
size into the logical points xdg_toplevel::configure sends it. Crossing
onto a different-scale monitor mid-drag kept every configure computed
against the origin monitor's stale scale for the gesture's whole
remaining duration, only self-correcting once the button came up.
Both functions now re-derive w.monitor from which monitor the window's
live geometry actually overlaps, every motion tick - the same
Rect::overlaps lookup end_drag already used once at the end, now run
continuously instead. end_drag's own fixup stays as a final-word safety
net for a drag that starts and ends between two motion ticks.
Does not close the related, already-documented gap where a client that
doesn't speak wp-fractional-scale-v1 still mismatches once settled on a
sub-1.0-scaled monitor - this only fixes the stale-during-the-gesture
half.
Two new tests, full workspace suite and clippy clean.
|
|
Closes "right-click on bare desktop" - previously a true no-op, nothing
rendered above the wallpaper at all. Requested directly: a real desktop
"just like windows does" - Home/Computer/Trash plus one icon per real
~/Desktop entry, individually draggable with persisted grid positions,
double-click to open, right-click menus (per-icon "Open" plus "Set as
Wallpaper" for image files when general.wallpaper_command is set; bare
desktop "New Folder"/"Refresh").
Architecture mirrors the existing context_menu.rs/snap_flyout.rs
"compositor-owned floating UI" pattern: desktop_icons.rs (data model, grid
layout, filesystem scan, hit-test), desktop_icons_state.rs (JSON
persistence, same shape as monitor_layout.rs), desktop_menu.rs (the new
right-click menu, reusing decoration::render_context_menu's existing
rasterizer), state/desktop_icons.rs (CompState glue: rescan/select/drag/
open/persist), and a new decoration::render_desktop_icon rasterizer --
hand-drawn glyphs, since no PNG/SVG decoding capability exists anywhere in
this workspace. Wired into both render loops (udev and winit) above the
wallpaper and below every window, and into input/pointer.rs's button/
motion handlers for selection, drag, double-click, and both menus.
Four new config keys: general.desktop_icons (default true - a directly
requested, purely visual feature, unlike the opt-in-while-experimental
general.gpu), general.file_manager, general.desktop_icon_single_click,
general.wallpaper_command (all default off/empty).
Deliberately out of scope for this pass, stated up front: move-to-trash
and "Empty Trash" (destructive, no confirmation-dialog primitive to gate
them on yet), filesystem watching, multi-select, per-mimetype icon art,
icons on any monitor but the primary one.
124 wayland-crate tests (up from 106), full workspace build and clippy
clean.
|
|
Reported live: "terminal output/everything disappears when i sometimes
resize terminal." masked_content_buffer (the udev/Pixman rounded-corner
content-masking path, live on this machine via general.rounded_corners)
rendered a window's whole surface tree into an off-screen buffer and
returned Some(bytes) unconditionally, with no check for whether that
tree actually produced any drawable elements.
content_epoch bumps on every commit, and a fast interactive resize is a
rapid-fire sequence of commits - real odds that one races ahead of the
client's own texture import, making the off-screen render legitimately
come back empty. That blank result got returned as Some and cached under
the new epoch the same as a correct one would, and rounded_content_buffer
only rebuilds on the next epoch change - so the blank buffer stayed on
screen, fully transparent, until the window's next real content change,
indefinite for an idle terminal.
masked_content_buffer now returns None when the element tree is empty,
before doing the render+readback at all - the same "give up unmasked"
pattern already used for a genuine renderer error. rounded_content_buffer
drops rather than replaces its cache entry on None, so the render loop
falls back to unmasked content for that one frame and retries the masked
path on the next.
Scoped to the udev/Pixman backend; winit masks via a GLES shader with no
equivalent failure mode.
|
|
Reported live: "even where decorations are corner i should still be able
to corner resize, just its... hitbox... does not get in the way of the
close icon" - the titlebar corner holding Close/Maximize/Minimize had no
resize target at all, by design (competing with the close button was
judged worse than losing that one corner). The BUTTON_CLUSTER_MARGIN
strip between the button cluster and the frame's true edge was already
dead space no button claims, regardless of button_count - its own top
DECORATED_TOP_RESIZE_MARGIN rows now register as the diagonal corner
(TopLeft/TopRight) instead of falling through to Top/Drag, without
touching the button's own hitbox at all.
Also requested: widen the other three corners' own resize zone, since a
user reaching for a plain edge-resize instinctively aims for the middle
of that edge, not its corner - a bigger corner zone doesn't compete with
that instinct the way a bigger RESIZE_MARGIN would compete with ordinary
content clicks near an edge. CORNER_MARGIN raised from 3 to 5 (18px to
30px at the default resize_margin).
Updated one existing test whose own per-window resize_margin override
(30px) now put its plain-edge test point inside the widened corner zone
on a window too short for the two to stay apart - taller geometry, same
edge point relative to center, no change to what it actually verifies.
Added coverage for the new button-corner resize target on both sides,
and for the dead strip's own non-corner rows still just dragging as
before.
|
|
first attempt
effective_frame_of now returns the live drag target while a window is
being interactively resized (same change as the reverted first attempt),
but two things make it safe this time instead of reintroducing the
out-of-bounds texture sample that reversion was for:
- Every src crop rect built from a window's frame width in udev/render.rs
and winit/render.rs (titlebar, top border strip, bottom border strip)
is now clamped against DecorationSignature's own recorded width/
border_width - the bitmap's actual last-built size - before reaching
MemoryRenderBufferRenderElement::from_buffer, which does not itself
validate src against the real texture size. This is a structural floor
independent of timing, not a repeat of the previous unsafe approach.
- handle_pointer_position now calls redraw_decoration_buffer once per
resize motion event (throttled to 60Hz via a new
CompState::resize_redraw_at), closing the lag at its source instead of
only catching up on the next real client commit. This also fixes the
shadow bitmap's identical commit-vs-live-position gap for free, since
redraw_decoration_buffer rebuilds all three bitmaps together.
Updates the TODO.md entry for this bug with the full before/after.
|
|
Past clear-color + cursor: a GPU-driven head now renders every
visible window's real content too, via surface_content_elements (the
same generic-over-renderer helper the Pixman path uses, unmodified
against gpu.renderer instead of udev.renderer). Content pushed after
the cursor (so the cursor stays on top), in the same front-to-back
`ids` order the Pixman path's own custom_elements already relies on
for correct occlusion between windows - plain painter's-algorithm
draw order, no separate clip needed since content is window-shaped.
Deliberately the *unrounded* path: no corner masking (that's built
against PixmanRenderer specifically on this backend) and no
decorations (border, titlebar) - a GPU-driven head now shows real
window content, square corners, no chrome. Decorations are the
remaining real gap before this path has parity with the software one.
Per-window geometry/position math (geom from window_anims or
w.geometry, band for a decorated window's titlebar reservation,
content_offset clamped non-negative) mirrors the Pixman path's own
content push exactly, including an earlier double-
subtraction and negative-margin fixes - so a CSD client with a real
shadow margin positions the same way on either render path.
Untested on real GPU-enabled hardware as of this writing: builds,
passes clippy, full test suite green, and matches the existing Pixman
path's geometry logic by inspection, but SRDWM_GPU/general.gpu were
both unset on the machine this was built on - noted honestly in
gpu.rs's own module doc comment, DEFAULTS.md, and
IMPLEMENTATION_STATUS.md.
|
|
focus-staleness report no longer reproduces
tools/toplevel-activate: a standalone (not a workspace member - its own
empty [workspace] table, so building srdwm itself never has to build
this too) wayland-client + wayland-protocols-wlr binary that lists every
open zwlr_foreign_toplevel_handle_v1, activates one by index, and prints
the resulting `activated` state from the protocol's own feedback.
wayland-client 0.31.14 / wayland-protocols-wlr 0.3.12 - the exact
versions smithay 0.7.0 already pulls in, so this talks to the same real
client library srdwm itself is built against, not a possibly-drifted one.
Used it to reproduce aegis's own exact repro (nested `srdwm --wayland`,
two plain alacritty windows, activate the non-focused one, check
`srd clients`) precisely: launched a real nested instance, activated
back and forth 5 times, checked `srd clients` immediately and after a
delay each time. Every check matched the protocol's own `activated`
feedback - no staleness found, on the nested/winit backend specifically
(the peer's own repro environment). Documented in docs/TODO.md as
likely already fixed by other focus/window-management work since the
original report, not re-root-caused after the fact, but confirmed not
currently reproducible via the exact repro that found it - left open
one more round in case it resurfaces, with this tool as the fastest way
back to a live repro if it does.
|
|
parent
Window::is_dialog (close-button-only titlebar, no traffic lights) was
only ever set from a native xdg_toplevel's own parent() - redraw_
decoration_buffer's is_dialog computation called dw.toplevel(), which
is always None for an XWayland-backed DWindow (X11Surface's own
accessor is x11_surface(), a different method), so the .unwrap_or(false)
fallback made every XWayland dialog - a GTK "Save As", an app's own
"About" box, anything setting the ICCCM transient-for hint - always
draw with the full three-button titlebar and traffic-light colours,
even though the feature this was built for explicitly wanted the
opposite. Documented as a known gap at the time; now closed.
redraw_decoration_buffer now also checks X11Surface::is_transient_for()
for an XWayland window. property_notify gained a WmWindowProperty::
TransientFor arm that re-runs redraw_decoration_buffer, for a client
that sets the hint slightly after its own initial map - the same
"read fresh every call" pattern the existing xdg_toplevel::parent()
check already relied on, extended to catch a late X11 property the way
the Wayland equivalent (set_parent, any time) already was.
|
|
Grepped the whole compositor for a second reader of every documented
general.*/theme.*/layout.*/performance.*/debug.*/platform.* config key
beyond its own seed call in crates/config/src/engine/support.rs.
Several sections turned out to be entirely or mostly decorative --
accepted, stored, sometimes range/type-validated, but never actually
applied to window/render behavior - which this file presented no
differently from the real, working settings around them:
- theme.colors.* (background/foreground/primary/secondary/accent/
error/warning/success): none read past the seed call. The real,
working theme surface is theme.decorations.* below it.
- performance.* and debug.* (vsync/max_fps/window_cache_size/
event_queue_size/layout_timeout/enable_caching, logging/log_level/
profile/trace_events/show_layout_bounds/show_window_geometry): every
key in both namespaces is dead. Real frame pacing comes from the
display's own hardware vsync; RUST_LOG is the real logging control.
- layout.tiling/dynamic/floating.*: only master_ratio is real
(WindowManager::tiling.master_ratio). split_ratio, every behavior.*
table, per-layout gaps.inner/outer, snap_threshold, grid_size,
cascade_offset, smart_placement, default_position, remember_position,
always_on_top are all dead - the real gap setting for every layout
is general.window_gap.
- theme.decorations.border.focused_style/unfocused_style and
title_bar.show/font: dead - solid is the only border style this
compositor draws, and the titlebar always uses whatever system font
find_system_font picks.
- platform.x11.*/platform.wayland.*/platform.windows.*/platform.macos.*
use_*/global_hooks/accessibility_enabled: all dead - EWMH/NetWM/
xdg-shell/layer-shell/DWM/Win32/Cocoa/Core Graphics support is simply
always compiled in, never behind a toggle. platform.backend/
platform.os are the two real, but read-only, keys in this section.
Also: the Environment Variables section listed SRDWM_THEME/
SRDWM_DEBUG_LEVEL/SRDWM_PLATFORM/SRDWM_MAX_FPS/SRDWM_VSYNC, none of
which exist anywhere in the codebase - replaced with the real set
(SRDWM_CONFIG_PATH, SRDWM_STATE_PATH, SRDWM_GPU, RUST_LOG), found by
grepping every std::env::var call in the compositor's own source.
Validation Rules' numeric list was missing corner-radius entirely
(theme.decorations.border.radius, 0-100, real in the validator despite
the doc gap) and resize_margin/inactive_dim, and didn't note that
srd set (unlike the Lua config path) doesn't run these checks at all.
Added general.gpu's own documentation section (new here - see
the matching commit adding the config option itself).
Also updates IMPLEMENTATION_STATUS.md's udev-backend paragraph, which
described the real GBM+EGL+DrmCompositor GPU path as not existing at
all ("that path needs a GPU...") - it now exists, opt-in, with its
current real scope (every head, cursor, no window content yet) noted
in place of the stale absence.
|
|
SRDWM_GPU=1 was the only way to opt into the udev backend's GBM+EGL
GPU render path - an env var, not a real config option, with no way
to enable it from init.lua the way every other general.* flag works.
WindowManager::gpu_enabled (plain bool, false by default - unlike
rounded_corners_enabled's Option<bool>, GPU rendering has one
unambiguous default regardless of which backend ends up connecting,
so there's no "let the backend decide" case to preserve) is read from
general.gpu in apply_general_settings, same as every other general.*
key. gpu::probe now takes an explicit enabled: bool instead of
checking the env var itself; udev/platform.rs's call site computes it
as wm.gpu_enabled OR SRDWM_GPU=1, so the env var still works as a
quick manual override for testing without touching config, on top of
the new persistent option.
Falls back to the existing software (Pixman) path exactly as before
on any failure at any step (no GBM device, no atomic-modesetting
support, a software-only EGL renderer, ...) - gpu::probe's own
fallback behavior is unchanged, only how the initial enabled/disabled
decision gets made.
|
|
Traced the whole write/read path for srd clients' focused field going
stale after a zwlr_foreign_toplevel_handle_v1.activate-driven change --
ruled out several plausible causes (a caching/staleness bug at the IPC
layer, a same-cycle dispatch-order race in the nested backend) without
finding the actual mismatch. No live repro was run: this machine has
neither pywayland nor wlrctl, and building a minimal wayland-client
test binary to call activate directly is real, separate scope. Written
up as a lead for whoever picks this back up, not a fix.
|
|
Past clear-color-only (Phase 2): the GPU-driven head now shows the
same moving cursor the Pixman path renders, via cursor::render_elements
- already generic over the renderer (R: Renderer + ImportAll +
ImportMem), so it works unmodified against gpu.renderer (GlesRenderer)
instead of udev.renderer (PixmanRenderer).
GpuElement (gpu.rs) widened from a bare MemoryRenderBufferRenderElement
to crate::elements::OverlayElement<GlesRenderer> - the same
Surface/Memory/Solid enum the Pixman path's own custom_elements already
uses, needed once the element list stopped always being empty.
Window content and decorations remain a real gap - a GPU-driven head
still shows no windows, just its own clear color and cursor. Untested
on real GPU-enabled hardware this work (SRDWM_GPU unset on this
machine's live session); builds, passes clippy, full test suite green.
|
|
carve_inner_corner_pixel (an earlier ring fix) cut the
border strip's own disk using an inner circle at the *same* centre as
the outer cut, radius - border_width - correct when a titlebar sits
underneath (it shares that exact circle, by construction), but wrong
whenever client content sits underneath instead: content's own
rounded-corner mask (rounded_corners_pixman.rs's apply_corner_mask) is
centred radius from *its own* buffer's edges, and that buffer starts
border_width rows/columns inside the border strip's - a same-radius
circle offset (border_width, border_width) diagonally from the
border's own outer one, not a smaller concentric one. Reusing the
titlebar-style ring left a real, if very small, gap along part of the
seam and a thin sliver of double coverage along the rest - reported
live, at extreme zoom against a solid-colour wallpaper (chosen
specifically to make a sub-pixel gap easy to spot against, unlike the
usual desktop image): "very tiny gaps... corner radius does not
match."
round_top_corners/round_bottom_corners's own `inner_radius: Option<u32>`
parameter is now `inner: Option<InnerRing>`, an explicit (center_row,
center_col, radius) rather than an implicit "same centre, smaller
radius" - InnerRing's own doc comment has the full geometry for both
cases. render_border_top gained a `decorated` parameter to pick the
right one (titlebar-aligned when true, content-aligned - centre
shifted by border_width on both axes, radius unchanged - when false);
render_border_bottom always uses the content-aligned ring, since this
compositor never draws a bottom titlebar.
apply_corner_mask (rounded_corners_pixman.rs) made pub(crate) so the
new regression test can build a real masked content buffer via the
actual production function, not a reimplementation of its math.
border_top_and_content_mask_have_no_gap_along_the_corner_diagonal_when_undecorated
checks coverage along the diagonal ray between the two circles'
centres - the direction they're actually offset along, and so the
worst case for a gap opening up between them - against the real
render_border_top/apply_corner_mask output, not a hand-rederived
formula.
|
|
Same fix as the udev backend's matching content-position code: a real
CSD shadow margin (dwindow.geometry().loc) is never negative, but a
live Firefox window was observed reporting loc = (-10, -10) despite
sync_geometry's tiled-state hint telling it to reserve no margin at
all. This path (rounded_content_element, a live GLES shader rendering
the client's own texture directly, not a separate pre-shifted buffer)
doesn't have the udev backend's double-application bug, but it shares
the same single-subtraction call and so needed the same clamp.
|
|
Phase 2 of the GPU-rendering plan deliberately targeted a single head
(GpuContext::output: Option<(crtc::Handle, GpuOutput)>) as a narrow
proof that GBM+EGL+DrmCompositor rendering works at all on this
hardware. DrmOutputManager already supports driving several crtcs at
once - initialize_output is a per-crtc call on one shared manager,
the same way anvil drives multiple outputs - so this was purely an
unexercised restriction, not an architectural limit.
GpuContext::output is now GpuContext::outputs: Vec<(crtc::Handle,
GpuOutput)>, and udev/platform.rs calls initialize_output for every
connected head in its own bring-up loop instead of only the first
after that loop finishes. A head this fails for individually (already
logged, not fatal) still just has no entry and falls back to the
existing legacy Pixman path, unchanged from before.
session.rs's VBlank handler and its VT-switch resume path (which
excludes GPU-driven crtcs from the legacy set_crtc reassert loop, a
different device fd that must never issue mode-set commands against a
crtc DrmOutputManager already owns) both now look a crtc up in the
Vec instead of comparing against a single stored one.
render.rs's own render-loop lookup uses direct field access
(gpu.outputs.iter_mut().find(...)) rather than an equivalent
&mut self method: the borrow checker treats a method call as
borrowing all of GpuContext, including gpu.renderer needed a few
lines later for the same head, where direct field access lets it see
the two borrows are disjoint.
Still gated behind SRDWM_GPU=1 (unset by default) and untested on
real multi-monitor hardware with the flag on - this machine has one
display, so the actual multi-head path itself only gets exercised
whenever it's set on hardware that has more than one.
|
|
Two bugs in how a CSD client's own declared shadow-margin offset
(dwindow.geometry().loc) drives where its content actually renders,
both surfaced by a real Firefox window:
1. A live Firefox window was observed reporting loc = (-10, -10) --
negative, despite sync_geometry's tiled-state hint telling it to
reserve no margin at all. effective_frame_of already clamps this
value to non-negative for its own size calc; the content-position
code (both the masked-content-buffer test and the real content
push) didn't, so a negative margin shifted content the wrong way --
away from the border, not toward it. Clamped to match.
2. Separately, and the actual cause of a later "border isn't over the
window" report on the same Firefox window (this time reporting
loc = (10, 10)): the masked/rounded content buffer's own build step
already renders the client's surface tree shifted by -content_offset
so the buffer's own (0, 0) lands exactly on the real, margin-
excluded content top-left (see rounded_content_buffer's own loc
parameter). Placing that already-compensated buffer on screen at a
*second* content_offset-shifted position double-applied the
correction, landing it content_offset pixels too far up and left of
the border wrapping it. Confirmed live via pixel sampling: Firefox's
own chrome rendered starting 10px above the border's nominal top
edge, fully exposed, square, with no border over it at all.
Split the single `pos` into `content_pos` (unshifted - what the
already-compensated masked buffer uses) and `pos` (content_pos further
shifted by content_offset - kept only for the surface_content_elements
fallback, which renders the client's raw surface tree with no prior
compensation of its own and still needs the shift applied once).
The winit/GLES backend's equivalent path doesn't have this bug: its
rounded_content_element renders the client's live texture directly at
`location` with no separate pre-shifted buffer, so a single
content_offset-adjusted position there was already correct.
|
|
round_top_corners/round_bottom_corners only ever cut pixels *outside*
the shared corner radius (the rounded outer silhouette). Nothing cut
anything *inside* it, so a border strip's own "extra" rows (present
whenever corner_radius > border_width) stayed a solid filled disk out
to the centre column/row, then hit clip_middle_beyond_thickness's
hard, unblended rectangular cut right at the disk's own most opaque
point - a clean right-angle step, not a curve. Confirmed live,
zoomed: a real square notch bitten into an otherwise smooth arc,
reported as "squares on the inside corners of each vertex."
Added carve_inner_corner_pixel, the same smoothstep falloff as the
existing outer cut but inverted (cuts near the centre instead of far
from it), applied at radius - border_width so the corner becomes a
genuine ring of ~border_width visible thickness tapering smoothly to
transparent, instead of a filled wedge. Only render_border_top/
render_border_bottom pass an inner_radius; the titlebar's own corner
and the lock-screen box keep their existing solid-disk behaviour,
which is correct for a single flat-coloured panel with nothing of a
different colour underneath needing to show through.
Also generalizes round_top_corners with an explicit center_col
parameter, mirroring the existing center_row shift: the titlebar's own
corner circle was never shifted horizontally to match the border
strip's (only vertically), leaving a border_width-wide sliver of the
titlebar's own misaligned curve poking through at the seam.
border_top_and_titlebar_corners_meet_without_a_seam and
border_top_curve_actually_closes_within_the_side_strips_own_width
updated to match: both now compare the correct corresponding columns
(the titlebar's own buffer starts border_width columns inside the
border strip's), and the latter no longer demands exact 255 opacity at
a point that legitimately sits within the new inner cut's own
antialiasing band.
|
|
Phase 2 (0274273) deliberately shipped without VT-switch support for
the GPU-driven head, documented as an explicit gap rather than a
silent risk. Before testing it live, wire the real fix instead of
finding out empirically - this already burned three real
reboots getting the *legacy* VT-switch path right, and DrmOutputManager
uses a genuinely different API surface (pause()/activate(), calling
through to DrmDevice's own master-lock acquire/release) than the
manual set_crtc+DPMS reassertion register_session_notifier already
does for legacy heads.
PauseSession now also calls DrmOutputManager::pause() when SRDWM_GPU=1
and a GPU context exists - a separate device/fd from the legacy Card,
so purely additive. ActivateSession calls DrmOutputManager::activate
(false), then deliberately does *not* also force a fresh render for
that head specifically: DrmCompositor::render_frame always issues a
full state commit (atomic or legacy, whichever this device negotiated
- see DrmDevice::is_atomic()), not just a buffer swap, so the
existing data.render_udev_frame() call at the end of this handler
already reasserts mode-set and CRTC-active state together for the GPU
head via render.rs's own GPU branch, the same way it always does.
Also fixed a real conflict Phase 2 introduced: the existing legacy
crtc-reassert loop (explicit set_crtc through the legacy Card/fd) used
to run for every head unconditionally, including one now driven by the
GPU path through a completely different DrmDeviceFd - two separate
fds issuing mode-set commands against the same physical CRTC, exactly
the kind of conflict that produced the worst VT-switch
incidents (EBUSY loops) when it was really one fd racing itself. The
GPU-owned crtc (if any) is now excluded from that loop.
|
|
Extends the SRDWM_GPU=1 opt-in path (Phase 1, fa0c7f1) from a
capability probe into an actual, working GPU render pipeline for
exactly one head, per the plan this was built from
the plan file.
gpu.rs: probe now goes all the way through EGLContext, GlesRenderer,
DrmDevice (real DRM device, separate duped fd from the existing legacy
Card), and DrmOutputManager construction, returning both a GpuContext
and its DrmDeviceNotifier on success. GpuContext::initialize_output
drives one crtc/mode/connector through DrmOutputManager, storing the
resulting GpuOutput for the render loop to find.
Confirmed while reading smithay's own source directly (not assumed):
DrmDevice::new's disable_connectors parameter is not an atomic-vs-
legacy switch - DrmDevice::create_internal tries atomic capability
first and falls back to a Legacy internal variant automatically,
exposed via DrmDevice::is_atomic(), logged here rather than assumed.
platform.rs: probes at startup, calls initialize_output for the first
head only (Phase 2's deliberate scope - see the plan), registers the
DrmDeviceNotifier as its own calloop event source alongside (not
replacing) the existing legacy DRM-fd registration.
render.rs: render_udev_frame's per-head loop checks, before any of the
existing Pixman-specific element-building logic runs, whether this
head's crtc matches the GPU context's initialized output; if so,
renders a plain clear color through render_frame/queue_frame and
continues to the next head, completely bypassing the Pixman path for
that head. Every other head, and this same head whenever the GPU
context or its output is absent, is entirely unaffected.
session.rs: register_gpu_drm_notifier handles DrmEvent::VBlank by
calling frame_submitted() on the matching GpuOutput (required per
queue_frame's own doc comment, or the swapchain runs out of buffers)
and logs DrmEvent::Error without treating it as fatal.
Deliberately out of scope for this phase (documented in the plan):
window content/decorations/cursor on the GPU head (clear color only),
multi-monitor GPU rendering (one head only), and VT-switch pause/
resume for the GPU head specifically (DrmOutputManager's own pause()/
activate() calls are a different API surface from the
existing manual set_crtc+DPMS reassertion, and porting that pairing
correctly needs its own isolated verification pass).
SRDWM_GPU unset (the default) is unaffected: every step above only
runs when it's set to "1", and every failure at any step falls back
to the existing, untouched Pixman path with a logged reason, same
fallback contract Phase 1 already established.
|
|
The udev backend is, by explicit design, 100% software: PixmanRenderer
compositing into legacy KMS dumb buffers. That was a deliberate choice
for portability (dumb buffers work on essentially any DRM driver,
including a VM with no GBM/3D support), not an oversight - but this
machine's real hardware (Intel UHD 620, i915) should support real
GBM+EGL rendering, and the user has asked for a genuine GPU-accelerated
path, GPU preferred with CPU fallback, built as a separate track that
doesn't risk the working software path.
Adds `udev/gpu.rs::probe`: gated behind `SRDWM_GPU=1` (unset by
default - a no-op, zero behavior change for every session that
doesn't set it), attempts GBM device creation on a duped DRM fd, EGL
display/device creation, and a software-rasterizer check, logging
exactly which step failed if any and falling back silently. Wired in
at udev backend startup, right after the DRM fd is opened.
Deliberately does not yet create an EGLContext, a GlesRenderer, or
touch scanout at all. Reading smithay's own reference compositor
(anvil/src/udev.rs) confirmed it wires GBM+EGL rendering together with
atomic-KMS scanout as one unit via DrmCompositor, not as a renderer
swapped into the existing legacy set_crtc/page_flip flip loop this
backend uses today. Adopting DrmCompositor is separate, larger-scoped
work than "swap the renderer" - it replaces the same UdevHead
mode-set/flip machinery the VT-switch fixes
(register_session_notifier's ActivateSession arm, copy_and_flip's
retry backoff) live in, and needs its own plan. This probe answers the
first question - does the hardware even support it at all - safely,
before that larger integration is scoped and attempted.
Cargo.toml: added backend_egl/backend_gbm smithay features, additive
to the existing renderer_pixman path (unchanged, still the default).
|
|
apply_geometry and restore - the Platform callbacks core's toggle_
maximize/apply_snap_zone/restore_window drive for a pure geometry
change - only called sync_geometry, never redraw_decoration_buffer.
The cached border-strip/titlebar bitmaps (self.border_top_decorations,
self.border_bottom_decorations, self.decorations) size themselves from
effective_frame, which can differ from w.geometry alone once a CSD
client's own invisible shadow margin is involved (see that function's
own doc comment) - but nothing here rebuilt them right when this
callback changed w.geometry. The next rebuild only happened whenever
this window's client next committed a frame or some other, unrelated
trigger reached redraw_decoration_buffer, not reliably right away.
Confirmed live: maximizing then restoring a Chrome window left its
border strips sized for the maximized frame while its real content had
already settled back to the smaller restored size, immediately and
permanently until some later trigger happened to catch it up - a
real, visible gap between content and border on the far (east/south)
edges, a different bug from the half-pixel corner seam fixed
separately in blend_corner_pixel.
Both apply_geometry and restore now also call redraw_decoration_buffer
right after sync_geometry, in both the udev and winit backends.
|
|
The border-strip bitmap's own corner rounding (corners.rs's blend_
corner_pixel) and the client-content mask's corner rounding
(rounded_corners_pixman.rs's apply_corner_mask) are two independent
implementations that need to trace the exact same circle where a
window's border meets its own rounded content. Their falloff math was
already identical (same smoothstep construction over the same
radius-1..radius+1 band), but apply_corner_mask samples each pixel at
its own center (x as f32 + 0.5) - the standard rasterization
convention, matching the GLES shader used for the winit backend --
while blend_corner_pixel sampled at the raw integer coordinate (x as
f32), a systematic half-pixel offset between the two curves.
round_top_corners/round_bottom_corners had a compensating "- 1" baked
into their own right-edge/bottom-edge center calculation, tuned
against the old, uncentered convention. Confirmed live at extreme
zoom: a small but real right-angle step partway along an otherwise
smooth arc, right where the two curves are supposed to meet --
reported as "squares on the inside corners of each vertex."
Added the missing + 0.5 to blend_corner_pixel and removed the two
compensating "- 1"s (both call sites), which now line up exactly with
apply_corner_mask's own clamp-derived center for the same corner
(width - r / height - r, not width - r - 1 / height - r - 1). All 378
existing tests pass unchanged - none of them assert exact pixel
positions this shifts by half a pixel.
|
|
The previous fix for a stale border size (switching effective_frame_of
from dwindow.geometry() to raw dwindow.bbox()) traded one bug for
another. bbox() is the window's entire committed buffer; geometry() is
that buffer intersected with the client's own xdg_surface::
set_window_geometry hint, which excludes any invisible CSD shadow
margin the client reserves around its real visible content. The
assumption behind the switch - that sync_geometry's unconditional
tiled-state bits make every compliant client reserve no such margin,
so nothing would be lost - was wrong: confirmed live via temporary
diagnostic logging, Chrome reserves a real, correctly-current 10px
margin on all four sides regardless of the tiled hint (Firefox, the
window that exposed the original staleness bug, does not - the two
disagree on this).
Raw bbox() therefore handed the border/content mask Chrome's entire
buffer, margin included - 20px wider and taller than its real visible
chrome on each axis, with no compensating position shift - so the
rounded border curve traced a rectangle Chrome's real content never
reached, and its true, still-square corner poked straight through the
curve instead of being hidden by it. Reported live as a border not
lining up with a window's content and a hard block cutting through an
otherwise-rounded corner.
Fixed by keeping both properties at once: dwindow.geometry().loc as
the margin - assumed symmetric (left == right, top == bottom), which
holds for every real CSD shadow margin observed here, since it's a
fixed design constant that doesn't scale with window size and so has
no equivalent staleness window even while the hint's absolute size
does - subtracted from the always-fresh bbox(). Current size, correct
visible-content bounds, for a client that reserves a margin (Chrome)
and one that doesn't (Firefox, whose hint .loc is always (0, 0), where
this reduces to plain bbox()) alike.
|
|
effective_frame_of sized a window's border/shadow/titlebar from
dwindow.geometry() - xdg_surface::set_window_geometry. Per smithay's
own implementation that value is the client's cached hint intersected
with bbox(), falling back to bbox() only if never set. Nothing in the
protocol obliges a client to resend the hint on every resize, and
intersection() can never return something larger than its smaller
operand - so once a client's cached hint is smaller than its current
real buffer, geometry() stays clamped there permanently.
Confirmed live: after a passive tiling reflow (this window resized
only as a side effect of a sibling window moving, no direct action on
this window itself), Firefox's real content filled the correct, much
larger area immediately, but its border/decoration stayed rendered at
a small fraction of that - unchanged for several seconds, well past
both animation settling and any reasonable commit-throttle window --
until an unrelated maximize/restore cycle happened to prompt Firefox
into resending a fresh hint and self-correcting.
Switched to bbox(): the real bounding box of the window's current
surface tree, which updates on every commit unconditionally. This
gives up excluding a CSD client's own invisible drop-shadow margin,
but sync_geometry already unconditionally sends all four tiled state
bits specifically so a compliant client (GTK4/Firefox) reserves no
such margin at all, so a compliant client loses nothing. Both the
decoration-bitmap sizing (redraw_decoration_buffer) and the render
loops' own border/shadow/occlusion positioning funnel through this one
function, so they stay consistent with each other - avoiding the
out-of-bounds texture-crop regression a previous, different attempt at
this same lag hit (see this function's own doc comment history).
|
|
The udev backend's damage tracking only forced a full repaint (ages =
[0, 0]) on a workspace switch or a VT-switch resume. An ordinary
move/resize/open/close/restack within the same workspace relied
entirely on OutputDamageTracker's own per-element diffing to compute
correct damage for the region a window vacated - which doesn't always
hold: maximizing a window over a second one, then un-maximizing, left
a persistent ghost of the second window's old titlebar/status text
sitting in the vacated corner, unchanged across multiple otherwise-
idle frames.
render_udev_frame now also hashes every visible window's id and rect
each frame and forces the same full-repaint reset whenever that
signature changes - the same "defensive, not a fix for a proven bug
in the diffing itself" reasoning the existing workspace-switch reset
already uses, just triggered by a second, complementary condition.
|
|
The kernel revokes every input device fd across a VT switch away.
libinput has a documented pair of calls for this exact case,
suspend()/resume() (libinput_suspend/libinput_resume), which reopen
every device through the session once it is reactivated. This
codebase never called either one, so after switching back to the
compositor's VT, libinput's device list stayed pointed at fds the
kernel had already revoked - reads on them don't error, they just
silently stop producing events, forever. Rendering, DRM/KMS, and
libseat's own session activation all recovered on their own, which is
what made this look like a display bug rather than an input one; it
took three real forced reboots today, with no visible input from
keyboard or mouse for 30+ minutes after switching back to tty1 each
time, to isolate it as this specific missing call.
register_libinput now returns the Libinput context (a clone of the
one already handed to LibinputInputBackend - it's a reference-counted
handle, not a deep copy, and LibinputInputBackend only exposes an
immutable accessor once it's moved into the calloop event source).
register_session_notifier takes that handle and calls suspend() on
PauseSession, resume() on ActivateSession.
|
|
VT-switch resume: reasserting the CRTC's mode-setting state was never
enough on its own - display power (DPMS) is separate KMS state, and
nothing here ever touched it after a real switch-away-and-back. Page
flips kept succeeding with zero errors logged for the rest of a real
30+ minute session while the panel itself simply stayed dark, which is
what actually explains a user report of the screen and keyboard input
never recovering after one VT switch. Sets DPMS-on unconditionally on
every resume now, the same property zwlr_output_power_v1 already
writes for an explicit client request.
Corner rendering: both backends' side-strip crop (the fix that keeps a
window's flat left/right border strips from poking a solid-coloured
square through the top/bottom strip's own rounded curve) only ever
activated for `w.decorated` windows. An undecorated/CSD window's crop
depends on its *content* actually getting masked to match - the
winit backend never checked that at all, so every CSD window with a
nonzero border_width got the uncropped, "staircase" artifact
unconditionally, confirmed live via a highlighted border colour and
raw pixel sampling. Ported the udev backend's own `border_curve_is_safe`
check (decorated OR content-will-be-masked) into winit, using the
cheap "does this surface have subsurface children" test both content-
masking code paths already gate success on, rather than duplicating
either one's real (comparatively expensive) rendering work just to
probe it.
Full workspace build + clippy -D warnings + test suite (378 tests) green.
|
|
The top-rounds-but-bottom-doesn't investigation these were tracking is
closed: live-verified via a real screenshot (pixel-level, not
eyeballed) that both corners round correctly on both a decorated
window and an undecorated/CSD one relying on content masking. Removes
three log::debug! blocks (corner-mask state, TOP/BOTTOM border strip
position dumps, and a raw alpha-byte dump of the bottom border buffer)
that were firing on every single render pass regardless of whether
anything changed, adding real per-frame overhead for output no longer
needed. Build + clippy + test (378 passing) all still green.
|
|
Pre-existing issues in the uncommitted rust-rewrite work (unused
imports, over-arity glyph-drawing functions, a couple of complex inline
types, or_insert_with(T::default) instead of or_default(), and one
unsimplified test-only arithmetic expression), plus two imports left
unused by switching to or_default(). None of these are behavior
changes. Full workspace build + clippy -D warnings + test suite (378
tests) all green after this.
|
|
# Conflicts:
# crates/config/src/engine/general.rs
# crates/wayland/src/udev/drm.rs
# crates/wayland/src/udev/mod.rs
# crates/wayland/src/udev/render.rs
# crates/wayland/src/winit/render.rs
|
|
Fixes a live-reproduced VT-switch busy loop (failed page_flip retried
with no backoff), shadow rendering bleeding onto occluding windows
unclipped, and a winit-backend buffer-age correctness bug that left
stale cross-window pixels on screen. Committing before merging in the
much larger uncommitted rust-rewrite worktree, which independently
touches several of the same files - this is the pre-merge baseline to
diff against, not a claim that these are the final versions of these
fixes.
|
|
Safety commit before reconciling this worktree with main, which has
diverged with its own separate fixes today. Nothing here is reviewed
or curated yet - this exists purely so none of this work can be lost
to a git operation, disk issue, or worktree cleanup while that
reconciliation happens.
|
|
sync()'s per-tick platform.focus(id) re-assertion (added to keep real
Wayland/X11 keyboard focus following core's own bookkeeping) ran
unconditionally on every dirty tick, including when nothing about focus
had actually changed. focus_window (core) has its own, separate side
effect of switching to the focused window's workspace when it differs
from the current one - correct when focus genuinely moves to a window
on another workspace, but this call was never gated on focus having
changed at all: switching workspace via activate_workspace left the
still-focused window's own workspace field untouched, so the very next
dirty tick's blind re-assertion of that same focus saw a mismatch against
the just-changed current_workspace and switched straight back.
Confirmed live via temporary core-side logging: two switch_workspace
calls a few milliseconds apart, the second one undoing the first every
single time, for every workspace switch that didn't also change which
window was focused.
Gated the re-assertion on the focused id actually changing since the
last sync() call. Real focus-follows-real-platform-focus still happens
on every genuine change, which is all the original fix needed.
|
|
Every pointer motion event re-ran the same popup/layer/content hit-test
from scratch and delivered focus to whatever it found right now - there
was no notion of "a button is held, keep delivering to the surface that
received the press" at all, which is standard, expected Wayland
compositor behavior (every real compositor does this; it's how dragging,
text selection, and scrollbar-thumb dragging all stay coherent even when
the pointer briefly leaves the widget's bounds mid-gesture).
Without it, a real human's hand drifting even slightly outside the
pressed surface mid-drag - trivially easy during a fast, non-perfectly-
straight mouse motion - sent that client an unrequested `leave` event in
the middle of its own gesture. GTK's drag recognizers (a GtkHeaderBar's
move-the-window gesture, concretely) treat a mid-gesture leave as "this
isn't coherent, abort," which reads as "dragging this window by its
title bar does nothing at all" - live-reproduced this work on Nemo,
and consistent with move_request never having fired once all session
despite real attempts.
pointer_button_grab captures the (surface, origin) resolved on a button
press once the held-button count goes from 0 to 1, and every event under
the grab - motion or button, this press's or a later one overlapping
it - is delivered there instead of wherever a fresh hit-test lands,
until every held button is back up.
|