onEvent() runs on the Hyprland IPC listener thread and mutated the label's
style context (add/remove class) directly, racing the GTK main thread's
drawing and corrupting the heap (double free / corrupted double-linked list).
Follow the Submap pattern: onEvent only stores the layout under the mutex and
emits the dispatcher; update() swaps the CSS class on the main thread, tracking
the previously applied class.
Fixes#4665
isLuaProtocol() probed the protocol by sending a real
"dispatch workspace __waybar_probe__". On Hyprland < 0.54 "workspace"
is a valid dispatcher, so the probe actually switched the user to a junk
workspace named __waybar_probe__ on the first workspace click/scroll.
Detect the protocol with the read-only "version" IPC query instead:
parse the numeric "version" field (falling back to the always-present
"tag" field) and treat Hyprland >= 0.54 as Lua. This has no side
effects. On any parse/query failure we log and fall back to the legacy
protocol, preserving prior behavior for older versions.
The debounce timer added for flicker prevention was armed from the IPC
listener thread via Glib::signal_timeout().connect(), while its timeout
lambda and the m_updatePending flag ran on the GTK main thread — an
unsynchronized cross-thread data race on GLib timer/main-loop state.
Additionally ~Workspaces() never disconnected the timer, so a pending
timeout could fire on a freed 'this' (use-after-free).
Restore the pre-refactor threading model: onEvent now only mutates state
under m_mutex on the IPC thread and calls dp.emit() (Glib::Dispatcher is
thread-safe). The debounce timer is owned entirely by the main-thread
update() path, which arms/re-arms it on each dispatch and coalesces
bursts into a single refresh. ~Workspaces() disconnects the timer
(guarded) so none outlives the object. Debounce behavior is preserved.
The format-<short_description>[-<variant>] override branches only passed a
positional arg, so a format using {short}/{long}/{variant} threw 'argument
not found', which disabled the whole module. Now supply the same named args
as the fallback/tooltip branches. Fixes#5120.
Post-0.15.0 review of the 0.15.0..HEAD range surfaced regressions and
bugs. This restores backward compatibility for existing configs/CSS,
fixes confirmed defects, and repairs the scdoc man-page build break on
master. Pango-markup tooltips are intentional and were kept.
Backward-compat restorations:
- AModule: honor legacy numeric Gdk::CursorType cursor values (int overload)
- memory: correct GiB divisor (was ~2.3% low); round bare {} placeholders
- wireplumber: scale max-volume into the linear domain so the cap works again
- idle_inhibitor: gate right/middle-click deactivate & scroll on dynamic-timeouts;
accept both dynamic-timeout(s); widen timeout to double (no fractional truncation)
- custom: keep #custom-<name>.<class> CSS selectors working (classes on box_)
- image: don't wordexp-split a single path; fall back to the literal path
- niri/window: restore hide-when-empty (new show-empty opt-in); escape tooltip
- wlr/taskbar: plain-text tooltip when markup is disabled
Bug fixes:
- tray: fix use-after-free in onAdd; guard the watcher retry timeout
- hyprland: clamp max-windows iterator (OOB); drop duplicate language tooltip block
- niri/window: supply {col}/{max_col} args in the empty branch (fmt::format_error)
- mpris: escape {dynamic}/{player} tooltip; fix dangling player; albumArtist source
- mango: fix use-after-free race (dispatch under callback_mutex_)
- mpd: contain throwing checkErrors in noexcept idle paths (no std::terminate/UAF)
- keyboard_state: always render every lock label, with guarded defaults
- bluetooth: bound GATT ReadValue timeout, opt-in + services-resolved gating,
preserve authoritative Battery1 percentage
- wireplumber: fix WpDevice reference leak / NULL handling
- battery, clock, dwl, wayfire, graph, custom_graph, transform, river: assorted
crash/logic fixes
Man page / build:
- niri-workspaces: fix scdoc "indented by an amount greater than 1"
(workspace-taskbar sub-options were mis-indented; breaks man-page build)
- document new show-empty (niri/window); correct network {txBitrate}/{rxBitrate}
Not compiled locally (no gtkmm on this host); C++ build relies on CI.
Man pages validated with scdoc 1.11.4.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Migrate idle_inhibitor, sway/language, sway/scratchpad,
hyprland/language and hyprland/window to the shared label/tooltip
setters. idle_inhibitor uses updateLabelAndTooltipForState; the others
keep their custom label/tooltip resolution (trim, precomputed tooltip,
rewriteString, per-lang/variant formats) and switch to the dedup-aware
setLabelMarkup/setTooltipMarkup.
Replace the getSocket1JsonReply("reload") call in onSpecialWorkspaceActivated
with the module's normal local refresh path: the state update to
m_activeSpecialWorkspaceName is picked up by the update() render (triggered by
the dp.emit() already issued at the end of onEvent), which re-evaluates active
state, icons and CSS classes. This avoids reloading the entire Hyprland config
on every special workspace activation while keeping the special-active icon and
CSS improvements.
The hover-check Glib::signal_timeout re-arms every 50ms and is not owned by
the Workspace. Since Workspaces are destroyed at runtime, a workspace removed
while its check is armed would let the timeout fire on freed memory (use-after-
free, also touching the destroyed m_button). Add a destructor that calls
stopHoverCheck() to disconnect the source.
The active-layout event payload is `KEYBOARDNAME,LAYOUT`, and kbName was
taken as everything before the first comma. Keyboard names reported by
Hyprland can themselves contain commas (e.g. some MSI keyboards), so a
`keyboard-name` configured with a comma never matched the truncated value
and the module stopped updating.
Match the configured keyboard-name as a full prefix (followed by the ','
separator) instead. Layout parsing already handles commas via the
last-comma-before-parenthesis heuristic, so only the keyboard-name filter
needed fixing.
Addresses the keyboard-name case originally reported by @tirsek in #3406.
Alex