A directory's stat size passes the MaxEntrySize check, so CopyFile tried to
read it and every folder restore failed with "is a directory". Skip the read
for directories; the uri-list and plain path offers already cover them.
Co-authored-by: Claude Opus 5.5 <[email protected]>
* lock: take the lock back when the shell restarts into a locked session
Check the state once when the module completes, on locked or lockedHint,
and handle it the way onSessionLocked does, including the wait while the
session is inactive. After a logind Unlock that the shell ignored for a
lock it started, only lockedHint still says locked; and the module loads
asynchronously, so the state can arrive before its Connections exist.
* loginctl: read LockedHint from logind on unlock
logind's Unlock signal does not clear LockedHint, only the locker does.
Forcing it false told a shell restarted in that window that the session
was no longer locked.
* added launcher dmenu mode
* fixed prechecker issues
* added unit tests for dms dmenu command
* added workaround for matugen printing error message to stdout
* correct help message: matching is case-*in*sensitive
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
* no longer save dmenu selections to history; now properly close dmenu sockets
* increased parity with rofi -dmenu flags; added -i as an ignored flag to match dmenu
* go mod tidied
---------
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
EntryToFile and the preview helpers stripped file:// without
percent-decoding, so history entries whose paths contain non-ASCII
characters stat'ed a literal %E5... path and copyEntry failed with
"file not found". CopyFile additionally hard-failed any file larger
than MaxEntrySize (5MB default) right after stat, so even decoded
entries for bigger files could never be restored to the clipboard.
Decode the path in EntryToFile, uriListPreview and
tryReadImageFromURI. In CopyFile only read file content when it fits
the size cap; oversized files still get offered as uri-list /
gnome-copied-files / plain path, and the stored entry keeps the file
URI instead of an empty body.
Keep encode and decode symmetric: every URI this service stores or
offers now goes through encodeFileURI (percent-encoded) and every
read path decodes exactly once through pathFromFileURI, so a literal
%20 in a file name survives the round trip. Regression tests cover
space / literal %20 / # / CJK / oversized names across EntryToFile,
uriListPreview, the image probe and handleCopyEntry.
* feat(screenshot): add Ctrl+C shortcut to copy selection without saving
Add Ctrl+C shortcut in interactive region selection mode to immediately
copy the current selection (or hovered snap target) to clipboard without
saving to a file.
- Works during mid-drag, on hovered targets, or on existing selections
- Bypasses file saving and forces clipboard output
- Emits aborted status in JSON mode so external callers (such as Quick
Capture) exit cleanly without launching an annotation editor
- Exits cleanly without error if triggered with no selection
- Updates HUD items with Ctrl+C copy shortcut
- Hides corner resize handles and maintains crosshair cursor while dragging with Ctrl held
* fix(screenshot): bypass stdout and scroll phase when copyOnly is active
- Disable stdout output when copyOnly is active to ensure the image
is copied to the clipboard instead of being written to stdout.
- Write 'Copied to clipboard' to stderr if --stdout was originally passed.
- Bypass entering scrollPhase in finishSelection() if copyOnly is set,
immediately capturing and copying the selected region.
* feat(screenshot): snap to active popouts and modals in region mode
Allow region screenshot selector to detect and snap to visible shell
popouts and modals on hover and click or keyboard confirmation.
- PopoutManager.qml: add getActiveSurfaces() collecting visible
DankPopout and DankModal global bounding boxes
- DMSShellIPC.qml: add screenshot.getSurfaces IPC endpoint
- core: fetch snap targets from shell IPC during region screenshot init
- selector: highlight snap targets on hover with surface name and dimensions
- selector: click or Enter/Space on hovered target locks selection to surface bounds
- selector: allow hover snap over pre-selection (SelectionState.fromPreSelect)
- selector: reset selection on degenerate (1x1) click release for quick re-selection
- unit tests for display name formatting, hover detection, and render bounds
* fix(screenshot): correct snap off-by-one and drag-away cancellation
- region_input: fix drag-away snap cancellation by setting hasSelection=true
and clearing preSelect so freeform dragging renders and is not discarded
- region_input: add snapToTarget helper subtracting 1 device pixel (devPxX/Y)
from currentX/currentY to prevent inclusive coordinate off-by-one (420x580 -> 420x580)
- region_render: compute hover bounds with w = round(width * scaleX) and
x2 = x1 + w - 1 to eliminate +1 device pixel in render bounds
- region_snap_test: assert exact 420x580 bounds for both hover and snapped states
* style(screenshot): format region_snap_test.go with golangci-lint-fmt
A compositor resolves a bind's keysym on the first level of the active
layout. Layouts that put punctuation higher up therefore cannot reach it
at all: de:neo has bracketleft on layer 4 and plain de on AltGr, so
Mod+bracketleft is a bind that can never fire. The editor recorded and
stored it anyway and gave no sign that anything was wrong (#3585).
`dms keybinds keymap` now reports what the seat's xkb keymap puts on the
first level of every physical key, and the editor says so instead of
letting a dead bind through.
The keymap a compositor hands out writes its keysyms as hex, so only the
range the keymap package can name back is usable for the comparison.
That is the ASCII range, which is where the problem lives, and the
command reports the vocabulary alongside the keys so the caller can tell
"the layout cannot reach this" from "we could not name it" and stay
quiet in the second case.
* lock: lock before sleep when the command skips logind
Custom hibernate never emits PrepareForSleep, so the session stayed unlocked. Wait for the compositor lock on every sleep path, and release the delay inhibitor if that lock finished first.
* lock: release the sleep inhibitor only after lockerReady
* lock: read the sleep inhibitor under its mutex in the race test
scrollbar: interface style toggle and no scrollbars in the dash
dash: never allow scrollbars
cc: widget sheet cards, header close button and content-sized dialogs
* 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.
Co-Authored-By: Claude Opus 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Q5LQHivuucTLXwLkvZMfcn
* 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.
Co-Authored-By: Claude Opus 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Q5LQHivuucTLXwLkvZMfcn
---------
Co-authored-by: Claude Opus 5 <[email protected]>
Autostart apps (e.g. KeePassXC) that call RegisterStatusNotifierItem
before the watcher owns its name fail silently and never re-register.
Until now the recovery scan only ran on suspend/resume or watcher
owner change, so first-boot orphans were never rescued.
Schedule a one-time delayed scan when the watcher first appears.
recoverTrayItems() already deduplicates via registeredConnIDs, so this
cannot double-register apps that registered normally.
* matugen: support multiple injected palettes with settings ui
Run user-configured external commands on each theme build and merge their
JSON output into matugen's import data under a namespace, so templates can
reference e.g. {{mypalette.color0}} alongside dank16. Multiple palettes are
supported, each under its own namespace. Command output may be nested JSON
(pywal/wallust shape). A settings page manages the list; a legacy
single-object config file is still read.
* settings: map PaletteInjectTab in search index
* matugen: bound palette-inject command with timeout and WaitDelay
* fix palette column spacing (review suggested
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
* remove wallpaper kind guard (review suggestion)
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
* fixup! remove wallpaper kind guard (review suggestion)
---------
Co-authored-by: bbedward <[email protected]>
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
* feat(keybinds): redesign cheatsheet with category navigation, M3 styling, and click-to-execute
- Add category sidebar with Material Symbols icons and badge counts
- Move search field to top of sidebar with Ctrl+F shortcut
- Implement 2-column masonry layout for keybind items
- Deduplicate duplicate actions and stack multi-key combos vertically
- Collapse consecutive workspace keybindings (1-9) into range entries
- Add media & hardware icons for volume, mic, brightness, and PrtSc keycaps
- Preserve Niri window filter properties and distinguish same-app Alt-Tab binds
- Enable click-to-execute for keybind actions across supported compositors
* fix(keybinds): preserve niri action properties and disable execution on range rows
* fix(keybinds): convert niri action properties to CLI flags and disable IPC execution for window switcher binds
* fix(keybinds): pass skip-confirmation as boolean switch in niri executeAction
* fix(keybinds): handle niri spawn-sh actions in executeAction
dnf lists the packages an upgrade obsoletes indented beneath it, under a
separate heading. Those rows are removals carrying their installed
version, so they were parsed as upgrades from a version to itself and
counted as pending. Skip indented rows. The heading is translated, so
matching it would only work in English.
dnf prints one row per (package, repo), so a package offered at the same
version by two enabled repos is listed twice and inflates the pending
count. Key the parse on (name.arch, version) and keep the first row.
bbedward