The sof-audio-pci-intel-* driver family needs sof-firmware to boot the
DSP, but mainline linux only optdeps it, so affected installs came up
with a Dummy Output sink only. Replace the Panther Lake check with
omarchy-hw-intel-sof, which detects any Intel PCI audio controller.
Ports ffa81021 from master, which adapts basecamp/omarchy#6200.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
pacman -Q resolves provider names, so dropping a virtual package could
remove whichever package provides it. Read pacman -Qq once and match
requested names exactly, ignoring providers and duplicate arguments.
Ports the provider-ignoring semantics of 9791b227 from master with a
single pacman query instead of one per package.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
libretro-bsnes2014 and libretro-mame2016 no longer exist in the repos,
which made the whole retroarch install fail. Ports fe30a07d from master.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Position the bar by dragging (or click-and-holding) empty bar space
toward a screen edge, with a ghost slab previewing the target edge.
With drag for position and double-click for transparency, the bar
config panel, its inline gear button, and the omarchy-launch-bar-settings
CLI are no longer needed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The clamshell watcher polls omarchy-hyprland-monitor-clamshell every two
seconds and calls enable_internal unconditionally which re-enabled DPMS
on the internal panel each time. When the lock screen blanks the displays,
the next poll of omarchy-hyprland-monitor-clamshell re-lights the laptop
screen, leaving it showing the lock screen indefinitely while external
monitors stayed off.
Only issue the DPMS wake when the clamshell flag was actually consumed,
i.e. on the lid-open transition the wake was meant for.
If the lid-open even fails to turn on the screen for whatever reason,
then pressing a key or moving the mouse will re-light the screen. This
is configured in `input.lua` via `key_press_enables_dpms = true` and
`mouse_move_enables_dpms = true`
The Copy URL extension was intermittently failing to copy anything, and
its notification looked oversized with no icon. Several distinct problems:
- The MV3 service worker was pinned to stale cached code. Chromium caches
the worker script for a --load-extension command-line extension and does
NOT re-register it when the file changes in place — not on restart, not on
a manifest version bump, not even after a clean shutdown. So updated code
never took effect: the old worker kept calling chrome.scripting.executeScript,
which throws once the scripting permission is gone, and the copy silently
failed. Renaming the worker to a versioned filename (background-2.js) is a
new script URL, which forces a fresh registration for everyone — new
installs and existing installs alike. Rename it again on any future worker
change (see the note at the top of background-2.js).
- Clipboard writes now go through an offscreen document (MV3's sanctioned
path) using a textarea + execCommand('copy'), replacing executeScript +
navigator.clipboard. This works on chrome:// pages and needs no scripting
permission. execCommand is used deliberately: navigator.clipboard.writeText
rejects in an unfocused offscreen document.
- Add a toolbar action (chrome.action.onClicked) so the extension is
clickable, not keyboard-only — it was greyed out in the extensions menu
with nothing to invoke.
- Slim the "URL copied" notification: lead the summary with a glyph so the
omarchy notification shell collapses it to a single-line toast with an
icon, instead of an oversized card with a blank icon slot.
- Guard the Quattro shortcut-repair migration: it edits Chromium Preferences
to move the Alt+Shift+L binding to the new extension id, but a running
browser rewrites Preferences on exit and reverts the edit. Prompt (via gum)
to close the browser first, only when a browser is running and there is
actually a stale binding to repair.
The omarchy-bar command had grown three overlapping ways to inspect the
bar (show/layout/list/options, plus selected/active/available/widgets
aliases) and mixed layout mutation in with bar-level settings. Untangle
it into two focused commands:
- omarchy-bar keeps only the bar-level settings that write shell.json:
use, reset, position, transparent, and settings. reset now delegates
to `use omarchy.bar` rather than duplicating the del(.bar.id) write.
- omarchy-bar-plugin owns all layout mutation: add, move, remove, set,
and replace, with the placement flags and jq resolve/anchor helpers.
`omarchy bar plugin ...` routes here via the dispatcher.
Drop the inspection commands entirely: the layout is visible on the bar,
the config is shell.json, and widget/option ids come from
`omarchy plugin list`. Nothing consumed the show output programmatically
except tests. This also removes omarchy-bar-position, whose jq write was
a duplicate of `omarchy bar position`.
Strip environment-invariant guards (require_command, require_omarchy_path)
that defended against jq or OMARCHY_PATH being absent — neither happens on
a real system. Keep the user-input validation (--section/--index) and the
atomic shell.json write.
Update callers (service install/remove, refresh-shell, plugin-clone,
plugin enable), keybindings, the menu, tests, and docs to the new split.
The weather widgets asked wttr.in to geolocate by IP, which lands on
whatever city the ISP registers its address block under -- Downey when
you're in Malibu. Now the location can be set explicitly:
* Click the location in the weather panel to search via open-meteo
geocoding, pick a suggestion (disambiguated by region), and it sticks.
The X next to the field returns to IP auto-detect, as does committing
an empty search.
* State lives in ~/.local/state/omarchy/settings/weather.json as
{name, latitude, longitude}, owned by the new omarchy-weather-location
command (bare call prints the current location, --set/--clear write).
The panel watches the file, so hand edits apply live.
* Stored coordinates make wttr.in and open-meteo exact. Open-meteo also
now supplies current conditions alongside the daily forecast it
already served, which fills the panel within a second of a location
change instead of waiting out wttr.in's multi-second responses; wttr
results still win once they land. Failed wttr fetches retry instead
of sitting stale until the next refresh cycle.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
canonical_widget_id was a pure pass-through, making its reserved-id
check a tautology; require_omarchy_path and choose_value were defined
but never called.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Apps like nvim clear the per-pane cursor colour with OSC 112, which made
tmux reset the outer terminal's cursor. Terminals that can't reload their
config (foot) then restore the cursor colour from whatever theme was
active at launch, leaving a mix of old and new cursor colours across
panes. Set the global cursor-colour option as a fallback so cleared panes
land on the current theme instead.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The trailing slash from the themes/*/ glob put a double slash in the
path comparison, so stock backgrounds never matched git ls-files output
and were all moved into ~/.config/omarchy/backgrounds.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
With dotglob set, * already matches dotfiles, so the extra dot patterns
made the second ln -s fail and silently skip the legacy hypr shim.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Legacy repos cached omarchy packages under the same name and version as
quattro rebuilds with different bytes, so pacman aborted --noconfirm
transactions with checksum errors.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Weather wraps its panel in a BarWidget that only exposed togglePanel(),
so it couldn't ride the new shell-routed hotkey path and was left on the
stale-prone per-plugin target. Expose open/close/opened on the wrapper
(open maps to the panel's hotkey path so the center hover reveal stays
suppressed) and switch omarchy-notification-weather to the new syntax.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A previous submap attempt was brittle because the script owned the
keyboard state: if it died between enter and exit, the submap stuck.
This version lets the compositor own both edges. The Lua config binds
plain Return when a layer surface with slurp's "selection" namespace
opens and unbinds it when the last one closes, so the bind exists
exactly while a pick is on screen. slurp dying for any reason (cancel,
crash, pkill, SIGKILL of the picker) fires layer.closed and the bind
evaporates — there is no code path where it leaks.
Return invokes omarchy-capture-region --take-fullscreen, which flags a
marker in XDG_RUNTIME_DIR and dismisses slurp; the picker treats an
empty slurp result with a fresh marker as the focused monitor, reusing
the existing fullscreen path. Since monitor focus follows the cursor,
that is the monitor under the slurp crosshair. The recording flow gets
monitor:NAME from the existing match, so Return during a recording
pick starts a native full-monitor capture.
The signal mode no-ops unless slurp is actually running, and both the
signal protocol and the layer-scoped bind lifecycle are verified live:
bind present only while slurp is open, real picks resolved by the
signal on both the screenshot and recording paths, no leftover binds.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Both capture scripts carried the same pipeline: workspace rectangles
from hyprctl, a hyprpicker freeze, slurp, and the bare-click snap loop.
The copies had already diverged once (rotated-display handling), so the
pipeline now lives in one hidden helper, omarchy-capture-region-pick,
which prints the picked geometry in slurp format.
The two callers differ only in what they layer on top. The screenshot
needs the freeze to outlive the pick so grim captures frozen content:
--keep-freeze leaves hyprpicker running and prints its PID first, with
the caller owning the kill. The recording prefers native full-monitor
capture: --match-monitor prints monitor:NAME when the geometry exactly
matches a display.
One deliberate fix along the way: the screenshot snap regex rejected
negative coordinates, so bare clicks on monitors positioned left of or
above the origin never snapped. The helper uses the recording variant,
which handles them.
Verified with an 11-case shim harness (snap precedence, rotated-monitor
match, freeze ownership, cancellation, negative coordinates) and the
live end-to-end screenshot sanity test.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Five scripts each hand-rolled colors.toml parsing plus their own
alias/fallback cascade, and the tables had drifted: theme-osc grew a
divergent color8/color9 chain, tmux and gnome carried private mode
detection, and the preview tool duplicated the lot. The differences
never mattered for shipped themes but were a standing bug factory for
sparse user themes.
omarchy-theme-color is now the single resolver. Its cascade is a
verbatim port of the canonical theme-set-templates logic (semantic
keys, legacy colorN aliases, derived shades, selection/cursor
derivation, mode precedence), exposed as --all, --raw, or single-key
lookup with fallback.
Generated output is byte-identical across all 21 shipped themes on
every surface: 17 template files, OSC byte stream, tmux and gsettings
command logs, and previews. Sparse legacy themes now resolve the same
palette everywhere instead of five slightly different ones.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
cmd_replace fed $PLACEMENT_FROM_SECTION/$PLACEMENT_FROM_INDEX to jq
without ever calling parse_placement, so they were always empty
strings that implied flag support which does not exist. Pass literal
empties with a comment stating replace resolves by widget id only.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The screenshot picker swaps monitor dimensions for transform 1/3
displays; the recording copy had drifted and did not, so portrait
monitors got landscape rectangles in the picker. The full-monitor
match had the same blind spot, silently downgrading rotated-monitor
selections to region capture instead of native -w capture.
Both now share the screenshot script's format_geo jq helper, which
also lets the monitor match compare one geometry string instead of
four fields.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Four scripts matched internal panels with contains("eDP") while
omarchy-monitor-state used test("^(eDP|LVDS|DSI)-"), so laptops with
LVDS or DSI panels were internal to monitor-state but invisible to the
clamshell, mirror, and external-active logic. All five sites now share
the inclusive anchored pattern.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
FILES="$*" space-joined multiple CLI path arguments into one bogus
argument before handing them to LocalSend. Build the file list as an
array from the start, which also collapses the single/multi send
branch into one path.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The keymap cache parsed xkbcli compile-keymap into a bash associative
array, serialized that array back into a string, and re-parsed it
inside a second awk's BEGIN block. One awk now reads the compiled
keymap directly, seeded with the fallback table so xkbcli failure
still degrades gracefully. (The xkbcli child must read /dev/null or
it consumes the binding records arriving on stdin.)
Also gone: the never-called lookup_keycode_cached, the DEBUG timing
scaffold, the redundant explicit noops in the Lua mock that the
metatable already provides, the walker-era XML entity escaping that
nothing renders as markup anymore, and the monitor-height computation
in favor of a fixed height like every other menu.
The cache key drops the double sha256 and now includes the active
keymap, so switching XKB layout invalidates cached symbol names
instead of serving stale ones.
Binding records are byte-identical before and after: 215 of 215.
686 lines to 573.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Each color lookup spawned a 60-line awk program with a character-by-
character quote-state machine for comment stripping, about 22 spawns
per run counting the dotted-key retries. One pass now tracks the
current [section], normalizes dotted keys to the same path, and emits
key/hex pairs into a single associative array the shell reads once.
The strict value pattern (a lone, optionally quoted and 0x/#-prefixed
six-digit hex, followed by nothing or a comment) makes quote-aware
comment stripping unnecessary. Section form still wins over dotted
form and first valid occurrence still wins, matching the original
search order. Output is byte-identical across fixtures covering both
key forms, comments, prefixes, and fallback paths, plus real themes.
197 lines to 151.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The braille and block renderers were two near-identical awk programs
sharing a verbatim PBM tokenizer, header check, and blank-line trim,
differing only in the per-cell rendering. One awk with a mode variable
now branches where they actually differ.
The option surface served two in-repo callers, both passing constants:
the duplicate positional loop behind --, the undocumented --block and
--braille aliases, and the --flag=value forms all go. The two magick
probes for the alpha range collapse into a single invocation.
Output is byte-identical across images and flag sets for everything
that remains, including both real callers' exact invocations.
353 lines to 243.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
