* lock: keep the fingerprint reader usable across suspend
The reader stopped responding after the lid was closed and opened, after a long
idle period, and after suspend.
The shell held a PAM session for as long as the screen was locked, so a suspend
always landed on a running verify. fprintd then suspends the device under the
verify, pam_fprintd's ReleaseDevice fails with "device is still busy", and the
claim is never released, so every later Claim fails until fprintd restarts. The
reader is now released when the monitors blank and armed again when they come
back, so a suspend has nothing left to wedge.
The PAM profile drops to max-tries=1 and sets an explicit timeout. pam_fprintd
only counts a wrong finger, so one finished session is now one wrong scan and
maxFprintTries means the number of scans it names rather than five times that.
A timeout expiry arrives as PamResult.Error, exactly like a device fault, so a
session that ran at least as long as the configured timeout is treated as a
planned renewal instead of spending the error budget and painting an error.
A suspend that still lands while the screen is on wedges as before. Recovery now
waits out fprintd's idle timeout instead of climbing to it, which brings the
reader back about 37 seconds after resume instead of about 98.
SessionService gains lidOpened from UPower, because logind can ignore the lid on
AC and resume alone is not enough. The lock PAM state machine is fixed alongside:
PamResult.Failed was dropped, a pending retry outlived the contexts that owned
it, and an availability check reset the wrong-scan counter.
* lock: time the fingerprint session from the verify, not the claim
pam_fprintd's timeout runs from the verify, but the session age was measured
from start(), which folds the claim into it. A slow claim plus a verify that
failed before its own timeout could add up past the timeout and be taken for a
planned renewal, skipping the backoff. A session that never reached a verify had
the same problem.
The age now starts at the message pam_fprintd sends when the verify begins, so a
session that never verified can never qualify as an expiry, and a slow claim no
longer ages a fault. The message is translated and quickshell clears it before
completed, so only its timing is used, never its text.
That message arrives after pam_fprintd has already armed its deadline, because
VerifyStart is asynchronous, so the age read here runs short of the real timeout
and a slack makes it up. The bias is deliberate. A fault misread as a timeout
costs one extra retry, while a timeout misread as a fault parks the reader behind
a backoff that grows to a minute, which on an idle lock screen would leave it
unusable most of the time.
Also corrects the comment that claimed a timeout is reported without a message.
---------
(cherry picked from commit d7074a53eb)
* fix: keep tray order stable when tooltipTitle changes dynamically
Revert the earlier simplistic key approach: keep getTrayItemKey()
unchanged (id::tooltipTitle) so duplicate-id disambiguation (#1542)
is preserved and persisted id::tooltipTitle entries keep matching.
Instead make persisted-key lookups tolerant of dynamic tooltipTitle:
resolveOrderIndex() in sortByPreferredOrder() does exact match first,
falling back to prefix matching an id::* entry only when the id has a
single instance. findSavedTrayKey()/isTrayIdHidden() replace direct
SessionData.isHiddenTrayId() calls and showTrayId() now receives the
actual saved key.
Prefix fallback is gated by trayItemCountForId() to single-instance
ids only, so multiple items sharing an id always match exactly and
hiding/ordering one never affects the others.
* fix: bail on ambiguous saved tray keys in fallback
* tray: use occurrence index as stable key, add one-time session migration
* tray: pin stable key per item instance so it never flips when siblings come and go
* tray: assign unique per-id sequential keys to avoid collision after sibling exit
* tray: reuse free key slots so a restarted singleton keeps its plain id
(cherry picked from commit 60342b7416)
With two or more screens connected and none of them in the lock screen
display list (a laptop docked at a different desk), every surface was
inactive and no password field rendered, so the only way out was a
TTY. getFilteredScreens only falls back to all screens with
showOnLastDisplay and exactly one screen, which the lock screen picker
never sets. Treat an empty result as all screens, the same as the
default "all" preference.
(cherry picked from commit 6a320399c9)
Opening the launcher by hovering the frame edge requested the Overlay
layer via CompositorService.framePeerSurfacesUseOverlayForScreen(),
which resolves to whether the frame is configured for the screen and
is true in normal use. DankLauncherV2ModalConnected's
frameOwnsConnectedChrome requires the Top layer, so the overlay
request silently dropped connected-chrome anchoring: the launcher
opened dead-centered with a different animation instead of emerging
from the configured frame side.
The keybind/bar-button path already opens on the Top layer by passing
no overlay argument. The edge-hover trigger only runs in connected
frame mode to begin with (gated by usesConnectedFrameChromeForScreen
in this file's own eligible check), so there's no case where it needs
standalone/overlay behavior. Match the keybind path.
(cherry picked from commit 579d6f7b5c)
* tailscale: handle profile switches and bus updates reactively
* tailscale: address review feedback on test coverage and dead code
(cherry picked from commit 218f7f752f)
Hyprland's workspace refactor (hyprwm/Hyprland#16140) broke
Hyprland.focusedWorkspace.monitor as a way to find the focused screen
(quickshell-mirror/quickshell#1119). Quickshell already exposes
Hyprland.focusedMonitor directly, so read the monitor from there
(cherry picked from commit 05c807b7cd)
* fix(battery): report when no charge threshold exists instead of faking success
* fix(battery): hide the charge limit controls when there is no battery
Follow-up to the review on #3385: the new explanation showed on
desktops and VMs with no battery at all, where it reads as a fault
report about hardware the machine was never going to have.
Both the explanation and the apply button now require
BatteryService.batteryAvailable. Hiding the button there is the same
point the pull request makes elsewhere: offering a control that cannot
work is what produced the false success in the first place. A machine
without a battery has no charge threshold to write to, so the limit was
never applicable.
BatteryService.batteryAvailable is already the gate used twice in this
file for the same reason.
(cherry picked from commit 6b50696b8a)
* add support for shelly
* fixed terminal failure
* updated to have it not allow adding ignore aur
* updated error strings
* Fixed label
port 1.6
(cherry picked from commit 8fe918ff24)
* updated dank to support aqueous
* further updates
* aqueous support work
* updated to support workspace options better in aqueous
* added blur to dms shell for aqueous
* added keybinding update
* remove unneeded blur
* clean up per claud review
* swapped to ipc
* aqueous: move IPC integration to QML, delete test fixtures (since we
dont have a QML test structure even setup), delete in-repo docs
* fix staticcheck
* aqueous: consolidate more aqueous specific stuff
* add aqueous logo
---------
Co-authored-by: Zoey <[email protected]>
(cherry picked from commit b9365610c8)
Root cause: we were creating a `Date` that did not exist in my timezone
(Santiago, Chile) because of DST. The JS engine decided to pick the
previous day, so I ended up with two times 5 September.
Fix: introduce a `DateOnly` object that internally uses UTC (to prevent
DST issues). Other parts of the system still use `Date`, which
probably should be refactored to use `DateOnly` as well (maybe in an
upcoming PR?). In the meantime, we provide a function to export a
`DateOnly` as a `Date`, using noon as the time instead of midnight
(because there are no DST changes at noon AFAIK).
We tested this manually (see the attached screenshots), but didn't find
a suitable place for automated tests.
port 1.6
(cherry picked from commit e4793327b8)
* fix(dash): dim today button when on current month and hide tooltip on click
* fix(dash): use Theme.shortDuration for calendar today button opacity animation
* refactor(dash): use standard DankActionButton tooltipText for calendar today button
* chore(dash): remove unused todayButton id
* fix(dash): allow returning to today when another day in current month is selected
(cherry picked from commit 07d8e28668)
* fix(bar): keep widget content on whole pixels
Bar widgets centre their content inside a box of
widgetThickness - horizontalPadding * 2, and Qt rounds the offset of a
centred anchor to whole pixels. When the content and that box have
different parity the exact offset falls on a half pixel, Qt rounds it
away from zero and the content renders half a pixel above the centre.
Widgets drawn from rectangles, such as the workspace switcher, compute
their own snapped height and stay exactly centred, which is why they
look right next to everything else in #3074.
With the stock configuration this is the normal case: the bar is 40 px,
the box is 14 px and barIconSize returns round((40 / 48) * 18) = 15.
Two sizes now round to whole pairs of pixels. The icon size, so that it
matches the parity of the box, and widgetThickness, because the box
inherits its parity from there (the padding is doubled, so it never
changes it). Going through every innerPadding from 0 to 16, nine of the
seventeen configurations currently put the icon on a half pixel and none
of them do afterwards. The 48 px default bar is untouched, the sizes
that change do so by one pixel.
Measured in a nested Hyprland at scale 1, ink centroid of the widget
content relative to the centre of its pill, before and after:
launcher -0.50 +0.00
weather -0.81 -0.52
clipboard -1.49 -0.93
notifications +0.03 +0.46
battery -0.63 -0.49
bluetooth -0.53 -0.03
Every icon moves down by half a pixel. What remains matches the ink
asymmetry of the individual glyphs, measured separately at an exact
position, and is not a placement error.
Two things this does not cover. On a fractional or HiDPI scale
horizontalPadding can land on a half pixel, which puts the box back
between the grid lines, and a text only widget such as the clock has a
line box of 15 px at the default font size, which cannot sit exactly on
the centre of a 28 px pill no matter how it is rounded.
Closes#3074
* fix(bar): give bar text a line box that can be centred
A widget without an icon is centred as one line box, and the same parity
rule applies to it as to the icon in the commit before this one. At
fontSizeSmall the line box of the bar font is 15 px tall, so on a 28 px
pill it cannot sit on the centre no matter which way the offset is
rounded, and the clock reads about half a pixel high next to everything
else. One step up the box is 16 px and it lands on the grid.
Only the branch the standard bar uses changes. The 0.9 branch below a
scale of 0.75 already comes out at 14 px, and the fontSizeMedium branch
above 1.25 is left alone because there the next even box is two steps
away, which would be a bigger change than this is worth.
Same session and same measurement as the previous commit, ink centroid
of the widget content relative to the centre of its pill:
clock and date -0.58 +0.05
battery -0.63 -0.05
weather -0.82 -0.28
The bar font goes from 12 px to 13 px, which is the visible part of this
and the reason it is a separate commit. It can be dropped without
touching the icon fix.
* bar: snap widget thickness to even device pixels and pick text sizes by line box parity
* bar: snap bar thickness to even so the pill centres on whole pixels
* bar: share the bar thickness formula so dock, osd and popouts stay in step
* dock: read the bar clearance at the compositor screen scale
---------
Co-authored-by: bbedward <[email protected]>
* feat(theme): add a selectable source color mode for wallpaper palettes
DMS always seeds the palette from matugen's most dominant candidate
(--source-color-index 0, hardcoded at both call sites). When a large
muted region outvotes a smaller vivid one, a visibly warm wallpaper
produces a cold palette: one image here goes to #8ad0ee where the warm
subject is at hue 46.
Adds a Source Color setting with seven values. Dominant is the default
and is byte-identical to today. Colorful is computed in Go before
matugen runs: quantize the image (Wu for the starting clusters, then a
weighted k-means in Lab), score the result the way caelestia-cli does,
and hand matugen the resulting hex. Its scoring walks a chroma/tone bar
down from 20 until something clears it, which biases the pick toward a
colorful, well-lit color rather than whichever muted color covers the
most pixels. The remaining five map to matugen's --prefer.
- core/internal/matugen/sourcecolor.go: extraction, scoring, and the
one helper both matugen call sites now share
- buildOnce resolves the seed once, before the StockColors branch, so
the dry run and the real run cannot disagree about it
- extraction failure logs a warning and falls through to matugen's own
extraction; it never fails a theme build
- --prefer arrived in matugen 4.1, so it is gated separately from the
existing isV4 check; on 4.0.x those five modes degrade to the
dominant color instead of aborting on an unknown argument
- --source-mode on generate and queue, plus sourceMode on the socket
handler
- Source Color dropdown in Theme & Colors, defaulting to Dominant
Adds github.com/Nadim147c/material/v3 (Apache-2.0, no dependencies) for
HCT conversions, Wu quantization and the dislike analyzer. Its
QuantizeCelebi is not used: it returns after one unconverged iteration
and its Lab.DistanceSquared is a dot product rather than a distance, so
it collapses most of an image into a single cluster.
* fix(theme): address review feedback on source color mode
Two points from review on #3020.
The quantize comment pointed at a FOLLOWUPS file that does not exist in
this repository. Replaced the reference with the measurement it was
standing in for, so the claim is self-contained.
setMatugenSourceMode called Theme.generateSystemThemesFromCurrentTheme()
directly on top of the regenSystemThemes onChange hook that Spec.set
already dispatches. Both feed a 100ms debounce so it was one generation
either way, but the direct call was redundant and is gone.
* fix(theme): keep {{image}} resolvable in colorful mode
Colorful resolves the seed itself and calls matugen with a hex source, which
leaves matugen's {{image}} unset: the builtin pywalfox template and any user
template using the keyword render the literal string "Null".
Pass the wallpaper path alongside dank16 in --import-json-string, as suggested
in review, so those templates keep working unedited. The path is made absolute
to match what matugen renders for {{image}} in the other modes.
* fix(theme): cut translator terms on the source color dropdown
Per-mode descriptions were seven of the fifteen new terms this branch adds,
and five of them differed by a single adjective. Drop them: the dropdown row
already describes what the setting does, and the docs cover each mode.
Vladimir Kostelei