wallpaper: drop redundant FBOs
startup: create QtMultimedia players lazily on-demand instead of at
startup
launcher: fixed animated images in plugins that use them
dms: set MALLOC_CONF=thp:never,narenas:4,dirty_decay_ms:3000 for
quickshell when unset. jemalloc extets no longer round up to 2 MB
huge pages
Worth keeping an eye on any potential regressions people encounter
* feat(audio): add output port selection
* fix(audio): handle onShowPortSelector in AudioOutputDetail
* i18n: add translator context for all audio port selector strings
* fix(audio): set isLoading = false in the callback of refreshSinkPorts
* chore: remove unused property in AudioPortSelector
* fix(audio): parse availability correctly from `pactl list sinks`
* fix(audio): address output port review feedback
ensureEnabledOutput previously checked if ANY output was enabled,
including virtual fallback outputs. A profile with eDP-1 disabled and
FALLBACK enabled was treated as valid, causing the only real monitor
to be turned off during hotplug transitions.
Now ensureEnabledOutput only considers outputs with make+model (real
monitors) when checking for enabled outputs. When all real outputs are
disabled, it enables the first real one. When no real outputs exist in
the profile at all, applyConfigEntry writes an empty config and lets
the compositor handle fallback via its built-in default rules.
Fixes the case where unplugging an external monitor leaves the laptop
display disabled because a stale FALLBACK auto-profile was applied.
The idle inhibitor ("Keep Awake") lives only in memory: SessionService holds it
as a plain `property bool idleInhibited: false` and nothing ever writes it out.
Every restart of the shell silently clears it — the user's explicit choice is
dropped with no message and no trace, and the pill just disappears from the bar.
Restarting is not an edge case. `dms restart` is a first-class CLI command, and
the unit shipped in this repo (assets/systemd/dms.service) sets
`Restart=on-failure` with `TimeoutStopSec=10`, so a shell that is slow to stop
is killed and brought back automatically. Each of those paths turns Keep Awake
off without saying so.
Every other Control Center toggle already survives a restart. Measured on
upstream master, over the state read by Modules/ControlCenter/Components/
DragDropGrid.qml and its neighbours:
doNotDisturb in SessionSpec: yes
isLightMode in SessionSpec: yes
nightModeEnabled in SessionSpec: yes
wallpaperCyclingEnabled in SessionSpec: yes
idleInhibited in SessionSpec: NO
So this follows the doNotDisturb idiom exactly rather than inventing a
mechanism: one spec entry, one property, one setter that saves.
Restore happens through two entry points because the singletons are lazily
created and their order is not fixed: `onLoaded` covers "service instantiated
before the session file is read", `Component.onCompleted` covers the opposite.
enableIdleInhibit() is idempotent, so whichever runs second is a no-op.
The default stays false and SessionStore.toJson omits defaults, so an untouched
session.json gains no new key.
Related: #1483 asks for a configurable timer. The field added here is the same
shape doNotDisturb uses with doNotDisturbUntil, so a later `idleInhibitedUntil`
slots in beside it without another migration.
Co-authored-by: Mark2Mac <[email protected]>
* fix(dgop): merge child processes by default
* fix(dgop): show the process list on first open
parseData threw away the first processes payload and waited 1s for a
second, on the grounds that the first carries no CPU deltas. That holds
for the cpu module, whose prime stays, but the processes sample is
usable as it arrives, so the wait only left the list blank.
* fix(dgop): refetch when the module set changes mid-request
updateAllStats returned early while a request was in flight, and that
request was built from the modules as they stood when it went out.
Opening the process list during the bar's periodic poll got back a
response with no process data, then a full updateInterval of waiting.
Queue a refetch and reissue when the response lands.
* fix(dgop): clear pendingUpdate when no request is sent
* fix(dgop): prime the queued refetch so the cpu cursor spans an interval
* core: bump dgop version
---------
Co-authored-by: bbedward <[email protected]>
Peter Fern