b4de1c0d34b8f96af84a8fbfc64e0800aeee8712
66
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 David Heinemeier HanssonandClaude Opus 5.5
|
0b9dace018 |
Unify input methods with installer keyboard selection (#14433)
* Unify input methods with the installer keyboard selection * Preserve inherited input keys and later keyboard changes * Exercise input engines across desktop toolkits * Show input modes in the keyboard indicator * Keep Pinyin setup prompts visible * Remember input selections before an app gains focus * Remove the Japanese input welcome notification * Show explicit choices on notification cards * Suppress the Cloudpinyin first-use prompt * Show the input switching shortcut in keyboard hints * Verify input setup and composition stay quiet * Use Super+I for input switching independently of Fcitx keys * Restore linked input settings and attribute test notifications to Fcitx * Correct the input setup shortcut hint * Keep input tooltips to the name and switching shortcut * Place the input language selector with the right bar widgets * Show input switching only for multiple layouts or methods * Unify input method and keyboard layout switching * Select keyboard layouts and input methods in the Typing menu * Apply typing selections immediately from the menu * Use installer keyboard choices and keep typing selections accurate * Queue typing changes and wait for saved selections * Keep queued typing selections when reopening the menu * Cycle keyboard layouts alongside composition engines * Avoid shadowing Python typing and retain pending mixed input selections * Keep input clicks working through controller and compositor failures * Move the keyboard layout widget to the right side of existing bars Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Map keyboard choices to the layouts the installer sets Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Keep the switching order when changing input methods Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Set input method fonts only after the selection saves Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Restore confirmed picker checks when an immediate change fails Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Route Super+I through the shell's input reader so unfocused choices stick Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Restore the input group when the font preference cannot be written Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop the azerty conversion now that the installer no longer offers it Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Apply the owner's keyboard layout once before persisting the input choice A quattro merge kept a second apply_keyboard after the retry loop, which re-ran loadkeys and firstboot, ignored their failure, and overwrote the XKB layout setup supplies for keymaps kbd-model-map lacks. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Give the Lao keyboard choice a Lao desktop layout The row used la-latin1, kbd's Latin American Spanish keymap, so Lao installs typed Spanish. There is no Lao console keymap; like Korean, the console stays US and the desktop gets la, which Omarchy keeps behind US as a non-Latin layout. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Load engines installed after Fcitx started when they are selected The Typing menu refused an engine whose package an update had installed, since Fcitx discovers engines only at startup. Restart input once and wait for the engine; name the package when it really is not installed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Cycle input methods back with Super+Shift+I Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Drop notification action buttons now that the Pinyin prompt is suppressed Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Cycle input only through the shell's reader Without the shell there is no bar or menu either, so the one-shot cycling path only duplicated the reader's logic. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Add engines from the command line through the Typing menu's path omarchy setup input <engine> had its own restart, append, select and live-rollback code in configure.py, parallel to the menu's. It now installs the packages and appends through typing_setup, leaving configure.py with install-time seeding and update defaults. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Keep Ctrl+Space out of Fcitx's toggles instead of probing for it The terminal prefix bind read the Fcitx config and profile on every press to avoid fighting a Ctrl+Space toggle. configure --defaults now removes that key from every trigger list, so the bind only checks for a terminal. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Run the input-method scripts with the system Python A bare python resolves through PATH, so a mise, pyenv or venv interpreter could run them without the system's PyGObject. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> |
||
|
+12
|
e1b0e5e9bb |
Converge omarchy-mac and omarchy-mx-mac into upstream Omarchy (#14431)
* Add omarchy-hw-platform to tell Apple Silicon and Qualcomm machines apart
aarch64 covers Apple Silicon Macs, Snapdragon laptops, Raspberry Pis and
VMs, so the CPU architecture can't pick platform setup. omarchy-hw-platform
prints apple-silicon, qualcomm, generic-aarch64 or generic from the vendor
prefixes in the device tree's root compatible (/proc/device-tree, then
/sys/firmware/devicetree/base) and the CPU, and fails when they contradict
each other. Its qualcomm answer uses the same evidence as the dragon
branch's omarchy-hw-qualcomm-soc, and the fixtures check the two agree.
omarchy-hw-apple-silicon is the Apple predicate on top of it.
An image built away from its hardware names its target in a root-owned
manifest, /var/lib/omarchy/image/target (format=1, platform=). While the
root is being built rather than booted, the detector answers from it and
never reads the build host's device tree; a booted system always answers
from its hardware. As root it restarts in an empty environment, so no
variable picks the platform for privileged setup.
Co-authored-by: Birk Skyum <74932975+birkskyum@users.noreply.github.com>
Co-authored-by: Jimmy Van Veen <8206925+JimmayVV@users.noreply.github.com>
* Reject platform gates keyed only on the CPU architecture
An architecture check is right for an ABI, a binary or a repository's
$arch, not for picking a platform. The lint lists each shipped check after
review and fails on a new one, so platform setup goes through
omarchy-hw-platform.
* Compose default packages from base, architecture and platform lists
omarchy-reinstall-pkgs installed the base list on every machine. A
platform's default set is now the base list, then install/omarchy-aarch64.packages
on aarch64, then the platform's own list when it has one
(install/omarchy-qualcomm.packages adds linux-firmware-qcom).
omarchy-pkg-defaults prints the set for the detected platform, or for the
one named, and reinstall installs it. x86_64 still gets exactly the base
list, and an image build gets its target's set.
* Early-load thunderbolt only where the kernel builds the module
linux-aarch64 has no thunderbolt module, and mkinitcpio fails the whole
image over a MODULES entry it cannot find. The optional form, thunderbolt?,
is not enough: a systemd initramfs copies MODULES as written into
modules-load.d, where "thunderbolt?" names no module, and a kernel without
the module would still list it there. thunderbolt_module.conf now adds plain
thunderbolt only when modinfo finds it as a module for the kernel being
built.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
* Compose mkinitcpio HOOKS from an early baseline drop-in
omarchy_hooks.conf assigned HOOKS outright, and mkinitcpio sources drop-ins in
name order, so it ran after any numbered drop-in (90-*.conf) and threw its
hooks away. HOOKS now comes from 00-omarchy-hooks.conf, which sorts first, and
later drop-ins add to it. The NVIDIA kms filter and the vconsole handling stay
in omarchy_hooks.conf. With the drop-ins Omarchy writes, machines keep the
same HOOKS, MODULES and FILES; a user drop-in that sets HOOKS and sorts before
omarchy_hooks.conf now takes effect instead of being overwritten.
Migration 1786605598 now sources the baseline first, since without it the
migration would read empty HOOKS and also rebuild hybrid systems. It skips a
machine that lacks the baseline.
* Resolve the LUKS parent from raw lsblk output
lsblk -s prints the ancestor tree with box-drawing prefixes ("└─/dev/nvme0n1p6"), so the crypto_LUKS parent came back as a path that does not exist and first-boot setup stopped with "Could not locate the LUKS device to re-key". -r prints raw names. Factory reset uses the same lookup.
* Check the factory-reset passphrase against keyslots only
cryptsetup open tries enrolled tokens before the key file, so a TPM2 or
FIDO2 token could authorize the re-key with a wrong passphrase.
* Test LUKS parent lookup during owner setup and factory reset
Exercise raw ancestry output without invoking provisioning or reset. Cover btrfs subvolume roots, absent encryption, failed probes, and explicit cryptdevice specifications.
* Cover LUKS parent lookup through an LVM volume group
* Journal the first-boot LUKS re-key so an interrupted attempt resumes
A first boot killed after the staged install key's slot was retired left the
key file behind, and every retry then stopped at "the staged LUKS key no
longer unlocks". A lost key file with the boot-time auto-unlock still
configured let setup finish with the disk unlocking unattended.
The re-key now runs from install/provisioning/luks-rekey.sh and records its
phase and slot numbers (never key material) in
/var/lib/omarchy/provisioning/luks-rekey.state. It reuses the owner's slot
instead of adding a duplicate, retires LUKS1 slots too, and destroys the
staged key only after the volume proves it opens nothing. Setup refuses to
finish while any part of the provisioning unlock remains. The journal ends at
phase=done and leaves together with pending, so a retry after the staged key
is retired must use the password the disk already holds.
Key checks exclude LUKS2 tokens: cryptsetup open tries an enrolled token
(say a TPM2 slot a previous owner left behind) before the key and reports its
slot whatever key it was given. Slot listings read only the keyslot section,
since a luks2-keyring token looks like a keyslot in luksDump.
* Change the root volume's LUKS key before the login and root passwords
Installation gives the login user, root and the disk holding / one password,
but omarchy-drive-password changed only the LUKS key, so the three drifted
apart. For the disk holding / it now checks the current password opens the
disk, changes the LUKS key, confirms the new key opens the disk and the old
one no longer does, and only then sets the login and root passwords. Any other
encrypted drive changes its LUKS key alone. Every drive now asks for the
current password with gum instead of cryptsetup's prompt.
A journal in ~/.local/state/omarchy/drive-password.state (disk UUID, key slots
and phase, never a password, written with the re-key journal helpers) lets the
next run finish an interrupted change. The rerun accepts only the old key
(removing the slot the interrupted change added) or the new one (retiring the
old slot), so another key such as a recovery passphrase never becomes the login
password. luksChangeKey is judged by whether the new key opens the disk, not
by its exit status, and key checks exclude LUKS2 tokens, which cryptsetup
open would otherwise accept in place of any password.
Runs are serialised, cryptsetup calls hold a disk lock, a / whose LUKS
partition cannot be resolved fails closed, and passwords are never traced. A
new password equal to the current one is refused, and --pbkdf argon2id is
passed only to LUKS2, so LUKS1 drives can change their password again.
* Add omarchy-lifecycle-dispatch for platform boot operations
Provisioning, factory reset and update call a fixed set of operations
through it. A platform whose boot chain the generic Limine path can't drive
registers a directory of root-owned entrypoints in the dispatcher; every
other platform gets a no-op and keeps its generic path. Apple Silicon is
registered to omarchy-mac-boot's /usr/lib/omarchy/mac-boot, with its
required operations failing clearly when the package or an entrypoint is
missing. docs/lifecycle-dispatch.md is the contract, including how a
Qualcomm boot package would plug in.
* Let the platform's boot package own first-boot setup's boot chain and key slots
Owner provisioning asks provision-prepare before the owner form, hands the
staged unlock to provision-commit and provision-verify where the platform
implements both, and records the kept LUKS slot through luks-slots before
the staged key is destroyed; the disk password change records the owner's
new slot the same way. On x86, Snapdragon and generic aarch64 every call is
a no-op and first boot is unchanged.
* Reset through the platform's boot package where it owns the boot chain
Where a platform implements the reset operations, reset-prepare and
reset-verify stage and prove the factory root's boot files, the throwaway
slot is added only after that, and reset-commit gets the key on standard
input once the factory root is active. A failure before the switch rolls the
boot state back and revokes the slot. Elsewhere the Limine UKI path is
unchanged.
* Check the platform's boot chain after an update
omarchy-update-boot runs update-verify after AUR packages, the last
sudo-capable step, so it also covers the initramfs rebuilds they trigger.
It follows third-party build code, so like AUR it authenticates through the
no-update wrapper. A failed verification finishes the update's remaining
steps, then exits non-zero without offering the reboot. Where the platform
does not implement it nothing runs and nothing asks for root.
* Hand a machine to its platform's migration through lifecycle dispatch
migrations/1790347292.sh resolves the new optional migrate operation and,
where the platform's boot package implements it, runs it as root. The
entrypoint moves the machine onto the platform's official package set,
idempotently and resumably; its failure leaves the migration pending. Once
it succeeds, a root-owned marker spares other accounts the migration and its
root prompt. Everywhere else, Macs without the boot package included, the
migration is a no-op, and an undetermined platform keeps it pending.
* Refuse packages tagged for another platform in a pacman hook
A platform package names its platform with groups=('omarchy-platform-<name>'),
<name> being what omarchy-hw-platform prints. omarchy-settings ships a
pre-transaction hook, the guard it runs and a copy of the detector, so the
guard is resident before the runtime and any platform package. The guard
reads tags from the synced databases and refuses a transaction that installs
a package tagged for another platform; a transaction without a tagged
package passes without asking which machine this is. Image builds answer
from the image's root-owned target manifest, as the detector does.
The guard ships on aarch64 only: no platform package is built for x86_64,
so x86 transactions run no hook. Hardware setup starts by proving the guard
is resident and checking the packages an installer placed without it. On
generic aarch64, where no platform packages exist, a missing guard is
logged, not fatal, and on x86 it is expected.
docs/platform-guard.md is the contract.
Co-authored-by: Birk Skyum <74932975+birkskyum@users.noreply.github.com>
Co-authored-by: Jimmy Van Veen <8206925+JimmayVV@users.noreply.github.com>
* Enable Voxtype's Vulkan backend only when it is installed
omarchy-hw-vulkan reports a Vulkan runtime, not whether this voxtype build
has a Vulkan backend. The aarch64 voxtype-bin package ships only the CPU and
ONNX binaries, so on any aarch64 machine with a Vulkan driver setup ran
`voxtype setup gpu --enable` and printed "Vulkan backend not installed" in
the middle of an otherwise working install. Require the backend binary too.
x86_64 voxtype-bin always ships /usr/lib/voxtype/voxtype-vulkan, so x86
installs still switch to the GPU backend as before.
Ported from omacom/omarchy-mac#483.
Co-authored-by: Oliver Lukschander <33756270+oliverlukschander@users.noreply.github.com>
* Link the system Widevine CDM into Brave
The widevine package registers its CDM for Chromium and Firefox only.
Brave looks for the CDM in its own install directory, finds nothing, and
DRM sites (Netflix, Apple TV+, Spotify) fail with a blank player and no
useful error. This hits machines where Widevine comes from that package,
such as aarch64 ones, since no Widevine component is published for them.
Link /opt/WidevineCdm/chromium into /opt/brave-bin and /opt/brave-origin-bin
through as_root when Brave or Brave Origin is installed, and add a migration
that repairs existing installs. Both do nothing when the widevine package is
absent or the browser already has a CDM. A standard x86 install has no
widevine package, so nothing changes there.
Ported from omacom/omarchy-mac#474, extended to Brave Origin.
Co-authored-by: Caleb Marshall <68087341+cajomar@users.noreply.github.com>
* Report failed setup steps on the console
With file logging, run_logged sends a failing leaf's output and its
"Failed:" marker to the install log only, so the caller sees setup stop
with an exit status and nothing else. A missing command (exit 127) was the
case that made this hard to diagnose.
Print the failed script, its exit status and the log path on stderr. Stdout
logging mode already shows the failure marker and is unchanged, and
successful steps stay quiet.
Ported from omacom/omarchy-mac#389 (the logging part).
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
* Let a migration defer itself without stopping the update
A migration that exits 75 changed nothing and cannot finish yet: omarchy-migrate
leaves it pending with its login notification, runs the migrations after it and
exits 0, so omarchy update carries on. Any other failure still stops the queue.
The platform migration passes a deferred refusal of its entrypoint through, and
defers itself where the dispatcher can't resolve its entrypoint (the platform
can't be told, or the entrypoint isn't trusted): nothing has run then, so the
update goes on to its boot verification, which fails there, as documented.
* Keep Apple Silicon's systemd initramfs on the HOOKS baseline
00-omarchy-hooks.conf set the busybox line on every machine. On a Mac,
omarchy-mac-boot's drop-ins then add the Asahi firmware hooks, but its
encryption drop-in reads busybox encrypt as a Mac unlocked through
cryptdevice= and leaves the line alone. The image came out as
"... block encrypt asahi omarchy-vendorfw filesystems ...", with no
systemd, sd-encrypt or omarchy-mac-encrypt, so an encrypted Mac could not
unlock its disk. Reproduced against an M1 Pro's installed drop-ins.
The baseline now asks omarchy-hw-platform. Apple Silicon starts from the
systemd line its boot package builds on. A Mac whose own mkinitcpio.conf
carries the asahi hook, or busybox encrypt, keeps that line, as a legacy
install needs. x86, Snapdragon and generic aarch64 keep today's busybox
line, as does a machine with no detector. The detector is the runtime's,
or else the copy omarchy-settings ships for its platform guard, so a
settings package newer than the runtime still places a Mac. A detector
that cannot place the machine stops the build.
The HOOKS test now composes the baseline on each platform and on legacy
Mac lines. A new test composes it with omarchy-mac-boot 20260926-1.45's
drop-ins, kept as fixtures, and checks the exact HOOKS the M1 Pro builds.
The NVIDIA test pins its platform.
* Hide optional installs that have no package on this machine
Not every optional install exists on every architecture: some vendors ship no
aarch64 build, and some repositories don't carry a package at all. Each
Install row now carries a `when:` guard naming every package its install
command adds, checked with omarchy-pkg-available, so a row with nothing to
install disappears instead of failing halfway through a pacman transaction.
The rows that build from the AUR (Chrome, Edge, Brave, Brave Origin, Zen,
NordVPN) are guarded on the CPU architectures their vendors publish for.
omarchy-pkg-available asks the sync databases: one `pacman -Slq` per batch,
and `pacman -Sp` for a name that list doesn't hold, so a provided name
(libappindicator-gtk3) and a version constraint resolve the way `pacman -S`
will. The menu's guard prelude sources the same function, so the whole batch
still costs one fork. omarchy-reinstall-pkgs reports and skips a default
that no configured repository offers instead of failing the transaction.
On x86_64 every guarded row resolves against the stable and edge
repositories, so the menu is unchanged there.
Tests derive each row's packages from its install command on x86_64 and
aarch64 and fail when the guard names different ones.
Co-authored-by: Ryan Hughes <1630358+ryanrhughes@users.noreply.github.com>
* Run 1Password with software GL where there is no render GPU
Electron draws a blank window when the machine has no DRM render node: a VM,
a headless or passthrough GPU, or a GPU whose driver isn't up yet (Apple M3
and later on Linux today). omarchy-hw-render-gpu tells whether a renderD node
exists, and omarchy-cmd-electron-gl-args prints --ozone-platform=wayland
--disable-gpu only when it doesn't.
omarchy-cmd-electron-gl-wrap installs a marked PATH wrapper in /usr/local/bin
that execs the real binary with those flags, so it follows the machine rather
than the moment it was installed. It refuses to replace a launcher it didn't
write, or to wrap a binary over itself, and stages the write so a failure
leaves the old launcher in place. omarchy-cmd-desktop-exec-repair points the
user's desktop entry at the wrapper by replacing only the executable token,
keeping arguments, other actions and customizations, with a backup.
The 1Password installer uses them when the machine has no render node. With
one, which is every x86 machine with a working GPU, nothing changes.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
* Record with wf-recorder when gpu-screen-recorder can't start
gpu-screen-recorder knows NVIDIA, AMD, Intel and Broadcom GPUs and aborts
during GL init on anything else, such as Apple Silicon's, before a recording
starts. Screen recording now falls back to wf-recorder, when it is installed,
if gpu-screen-recorder is missing or exits before the file appears. The
picker, file naming, notifications and post-processing are shared. wf-recorder
records one audio device, so desktop plus microphone go through a temporary
null sink that stop unloads, and its track is pinned to 48 kHz. Portal
capture and --resolution stay gpu-screen-recorder only.
The finalize pass keeps 48 kHz too: loudnorm declares 192 kHz output, so
without -ar the AAC encoder wrote 96 kHz from a 48 kHz recording, which many
editors decode as silence, and which on Apple Silicon makes the speaker
safety daemon lock the amps. #11092 makes the same one-line change; whichever
lands second drops it.
Stop and status matched `pgrep -f ^gpu-screen-recorder`, which misses
wf-recorder and matches any command line that starts with that path. The
script now records the recorder's pid, and the new
omarchy-capture-screenrecording-process checks or signals it by its
executable, through a pidfd opened before the check so a reused pid can't be
signalled. When that pid is no longer a recorder (a recording started before
this change, or one that died), every recorder the user runs counts, as it
does for the bar indicator and the Stop row, which use the same helper.
Where gpu-screen-recorder works, which is every x86 GPU it supports today, it
records as before.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
Co-authored-by: Issac Trotts <15530+ijt@users.noreply.github.com>
* Hold logind's lid switch while a display it doesn't count is in use
systemd-logind treats a closed lid as docked only when an external display
is on a connector type it lists (HDMI, DisplayPort, DVI, VGA...). Displays on
"USB" connectors, which Apple Silicon uses for USB-C displays, aren't on that
list, and a missing comma fuses DVI-A and Composite into one entry. With such
a display in use, closing the lid suspended a laptop that was still being
used on it.
omarchy-system-lid-inhibit holds a handle-lid-switch inhibitor while one of
those connectors is connected and enabled in Hyprland, on a laptop where
HandleLidSwitchDocked is ignore, which is exactly the case logind would have
ignored had it counted the display. It releases the inhibitor when the display
goes. A compositor or logind that doesn't answer changes nothing, since
releasing then would suspend a closed lid. The inhibitor runs as a transient
user unit bound to the monitor watcher's unit, so a watcher that dies can't
leave the lid inhibited until logout.
The monitor watcher runs it alongside the clamshell sync. On a machine whose
external displays are all on connectors logind counts, it does nothing.
* Keep the focused workspace in front when a monitor goes away
When a monitor is unplugged, or the lid turns off the laptop panel, Hyprland
first warps focus to whichever monitor it lists first and only then moves the
removed monitor's workspaces over. The workspace you were on lands hidden
behind that monitor's own (Hyprland 0.56.2, CMonitor::onDisconnect).
monitor.focused fires before Hyprland records the new focus, so the monitor
still reported as active is the one being left: its workspace is remembered
until the event loop next turns, and a monitor.removed in that window focuses
it again once the removal has finished. An open special workspace already
comes along, and any focus or workspace change made before the restore runs
wins.
* Drive the keyboard backlight from the ambient light sensor
Laptops with an ambient light sensor and a keyboard backlight LED now light
the keys in the dark and dim them as the room brightens, inverse to display
auto-brightness. omarchy-brightness-keyboard-auto polls the sensor every five
seconds (preferring an IIO device that names itself ALS, and converting a raw
channel with its offset and scale) and maps lux onto the LED, with a deadband
so it doesn't chase noise. It pauses while the screen is locked or the lid is
closed.
A level set with the keyboard-brightness keys wins until the light changes
enough that it no longer fits. omarchy-brightness-keyboard records each level
it sets in the session's runtime directory (a private state directory without
one), so the loop can tell keys turned off on purpose from keys left off by
the lock screen or a restored zero, which it lights again.
The user unit runs it in the graphical session. Its ExecCondition asks
--available, so it is a no-op on machines without both a sensor and a
keyboard LED. First run enables it, and a migration enables it for existing
installs, writing the enable symlink itself when there is no live user
manager.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
Co-authored-by: Gabe Ohlson <3355909+6abe@users.noreply.github.com>
* Defer an image build's hardware setup to the machine's first boot
An image is built away from the machine it will run on, so its hardware
leaves would probe the builder. While the image manifest the detector
already reads (/var/lib/omarchy/image/target) exists, omarchy-apply-hardware
queues every leaf of install/hardware/all.sh instead of running it and arms
omarchy-provision-hardware.service. On the first boot, before owner setup
and the login screen, omarchy-provision-hardware retires the manifest and
runs the queue on the real hardware: a failed step stays queued for the
next boot (exit 75), and the initramfs is rebuilt once after the last step
when a step changed its inputs or asked for it. Bluetooth starts its
service on that boot, whose target is already up. Root ignores any
environment override of these paths. Without a manifest, as on every
ISO install, hardware setup runs exactly as before. Nothing here fetches
package databases: an image carries its sync databases and every package
its steps install.
* Give an image-built root a UTF-8 locale when LANG is still C
A root built from a distribution tarball instead of the ISO (Arch Linux
ARM's ships LANG=C) never went through the ISO's locale step, so it ran
non-UTF-8. The config phase now selects en_US.UTF-8, generating it if
needed, but only when LANG is unset, C or POSIX: any named locale is
somebody's choice and stays. The ISO writes the locale before this runs,
so an ISO install is unchanged.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
* Let a platform's packages run their own boot, system and user setup
Three dispatch operations give a platform one place for the setup it
needs beyond Omarchy's, instead of inline platform branches in the setup
leaves. The last hardware leaf runs setup-boot, then setup-system, as
root, so an image build queues them with the others and its first boot
runs them, with image-first-boot (offline-safe, asking for the one boot
rebuild instead of building). setup-boot belongs to the boot package and
is required on Apple Silicon, where hardware setup has no other way to
run the Mac's boot setup; a failed setup-boot stops the leaf before
setup-system. setup-user runs as the user being set up, refuses root,
and gets only the user's home and session variables: from the last user
leaf at finalization (skipped inside an image build, where the detector
would read the build host) and again from first run once the session is
up. On Apple Silicon setup-boot resolves in omarchy-mac-boot's directory
and the other two, both optional, in omarchy-mac's, /usr/lib/omarchy/mac;
x86, generic aarch64 and Qualcomm register none, so all three are no-ops
there.
A failed finalization now keeps first run pending, like any failed
first-run step, instead of letting the other steps mark it complete.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
* Find batteries, lids and panels that aren't named like a PC's
Laptop detection assumed PC names: a battery called BAT*, a lid in /proc/acpi, a
touchpad or trackpad in the device name, a GPU backlight. Apple Silicon has
none of them (macsmc-battery, the SMC lid switch, apple-mtp-multi-touch,
apple-panel-bl), and some Framework and Chromebook firmware calls its battery
CMB0, so those machines lost the battery panel, laptop-only services, lid
handling, the touchpad toggle and display brightness.
- Battery presence goes by type instead of name, and skips a battery scoped
Device (a wireless mouse or keyboard), so a desktop with one isn't a laptop.
- Battery status takes the first UPower battery that powers the machine, reads
thresholds and cycles from it (BAT* stays the fallback), shows the magnitude
of a signed power_now, and rounds the percentage half-up for display like the
bar, while the charge-hold check keeps the unrounded value.
- A laptop is also one with any input device that reports a lid switch
(SW_LID), whatever its name: Apple's SMC input device has the same name on a
Mac mini, where it reports none. Whether the lid is closed comes from ACPI
where there is an ACPI lid, as before; otherwise logind is asked, with a
short timeout, only when a device reports a lid switch.
- A device named touchpad or trackpad still wins; only when there is none does
the touchpad match take a mouse udev marks as a touchpad (Apple's
multi-touch trackpad), never a touchscreen.
- The platform package can describe its displays in
/usr/share/omarchy-platform/displays.conf: backlights that never drive the
panel (a Touch Bar's), the panel's own where there is no GPU backlight, and
a display driver without DDC channels, where a probe walks unrelated I2C
buses for about 3 s: there only a connector with a ddc node is probed, and
a monitor without one reports no brightness instead of dimming the built-in
panel. Omarchy ships no such file and names no device in it; without it the
backlight choice and DDC probing are unchanged.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
Co-authored-by: Ryan Hughes <1630358+ryanrhughes@users.noreply.github.com>
Co-authored-by: Sparsh Kumar <22640187+skkumarsparsh@users.noreply.github.com>
Co-authored-by: dl-alexandre <166029845+dl-alexandre@users.noreply.github.com>
* Keep the user's XDG config and state directories in platform user setup
setup-user runs with an allowlisted environment, which dropped
XDG_CONFIG_HOME and XDG_STATE_HOME. A platform's user setup reads the user's
configuration and keeps its setup markers there, so for a user who sets
either it looked in, and wrote to, the default directories instead. Both now
pass through with the rest of the session; XDG_DATA_HOME and everything else
still don't.
* Let a platform package add Hyprland defaults of its own
A platform's runtime package may need Hyprland defaults of its own (a Mac's
capture keys and three-finger workspace swipe, say), and Omarchy's config
should carry no branch for them. Three directories under the platform root,
/usr/share/omarchy-platform, now take them. The one installed platform package
owns that root, Omarchy reads it by literal path and ships nothing there, so
without a platform package nothing changes. Each file is loaded by its path,
so it can't shadow a module, and one that fails is logged and shown as a
notification while the rest of the config loads.
- hypr/defaults/*.lua loads before Omarchy's defaults. A
chord bound there replaces Omarchy's default for it, and a decorator it adds
runs before every later bind, so it can bind what has to run first. It sees
the command a bind runs, including the command a shell shortcut such as
{ menu = "root" } stands for when it binds the shell's global shortcut.
- hypr/settings/*.lua loads after Omarchy's defaults and
before the theme, for settings that replace Omarchy's.
- hypr/gestures/*.lua loads after the user's files, only for what must
see them: the gesture registry, which now records on every platform, lets a
default gesture step aside for the user's own.
The user's files load after the first two and can unbind or rebind any of it.
* Let a platform package name keys in the keybindings menu
The keybindings menu shows a key by its keysym, which on some keyboards is not
what the key says: on a MacBook, XF86MonBrightnessUp and Down are the F2 and
F1 keys, so the menu listed rows like "SHIFT + XF86MonBrightnessUp". A
platform package can now list names for such keys in /usr/share/omarchy-
platform/key-names, one "<keysym> <name>" per line, and the menu shows those
instead. The file is part of the menu cache's key, so installing or changing
it refreshes the menu. Without it the menu is as before.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
Co-authored-by: kx0101 <90087463+kx0101@users.noreply.github.com>
* Keep the top bar out of a panel's camera cutout
A laptop panel with a camera notch at its top (Apple Silicon MacBooks, for
one) covered the middle of the top bar, and a bar shorter than the notch left
a sliver of every window beside the camera. Omarchy knows no panel's cutout
itself: the platform's own package describes them in
/usr/share/omarchy-platform/display-cutouts.json, by
connector, physical mode and the rows the cutout covers. On a described panel
a top bar is never shorter than the cutout, its center section sits beside
the right one, and hiding it parks all of it; [bar] notch-height sets that
floor by hand, on that panel only. Without a description, which is every
machine without such a package, the bar is as before.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
* Keep PC and Intel Mac quirks off Apple Silicon
A few hardware steps and commands written for PCs and Intel Macs do the wrong
thing on Apple Silicon:
- The Intel Mac Wi-Fi quirk turns off the Broadcom firmware supplicant for
every Mac whose DMI vendor is Apple and whose chip is on its list. Apple
Silicon reports Apple Inc. through U-Boot's DMI tables, and the M1 Pro and
Max's BCM4387 is on the list, but there the firmware supplicant is the one
that works: turning it off makes firmware commands time out and breaks
scanning. The install step and its migration now skip Apple Silicon.
- fnmode=2 for Apple-like keyboards is left to the Mac's own platform package
on Apple Silicon, where the built-in keyboard's top row depends on it.
- The SPI keyboard check reads the DMI product name without failing the
hardware pass where there is none.
- The Windows VM, an x86_64 guest under KVM, runs on x86_64 only, and direct
boot (U-Boot finds the bootloader and makes its boot entries itself)
refuses Apple Silicon; the menu hides both there. Hibernation setup skips
it; its kernel has no hibernation today, and resume there is not wired.
Everywhere else each step runs as before.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
* Let a platform's runtime package adjust the apps it needs to
Some apps need a platform's own touch once they are installed. On Apple
Silicon, Chromium-family browsers must decode video in software (the hardware
decoder renders green frames in them), and Steam only runs through a
FEX launcher, which then depends on Steam and must go before Steam can be
removed. Until now the Mac runtime did this with inline branches in the
installers.
Two optional dispatch operations take their place. post-install <app> runs
as the user once an installer has installed the app and finished its own
setup of it: every browser install (after its flags file, before it says the
browser is installed) and Steam's (after the package, before the 32-bit
drivers and the first launch). pre-remove <app> runs before Steam's removal
drops its packages. Where a platform registers them they refuse root (a
platform with nothing registered is a no-op, so sudo installs still work
there) and get the same home and session variables as setup-user, and a
failure fails the install or stops the removal. On Apple Silicon they
resolve in omarchy-mac's directory; every other platform registers none, so
installs and removal run as before there.
Steam's 32-bit driver step also stops failing the install on a machine with no
Intel, AMD or NVIDIA GPU (Apple Silicon, but also Snapdragon or a VM): it has
nothing to add there and now says so.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
* Ask the platform's boot package before an update takes over its files
When an Omarchy package's upgrade stops only on files no package owns, the
update moves them aside and retries. On a Mac some of those files belong to its
boot chain or were written by its image (initramfs and pacman configuration,
say), and moving them aside can leave it unable to boot or update. The Mac
runtime refused those paths with an inline list.
update-takeover <path>... asks the platform's boot package instead, before
anything moves: read-only, exit 0 lets the takeover go ahead, anything else
keeps every file and the update fails as it did before the retry existed. The
update resolves it as the user and runs it with sudo only where it resolves,
inside the authorization the move needs anyway. On Apple Silicon it is
required: a Mac without its boot package, or with one too old to answer, keeps
its files. Every other platform registers none, so the takeover goes ahead as
before, asking no one.
* Install the Mac's package set on Apple Silicon
omarchy-pkg-defaults already composes install/omarchy-<platform>.packages
after the base and aarch64 lists, but Apple Silicon had no list, so a Mac got
the generic aarch64 set: no omarchy-mac or omarchy-mac-boot, whose entrypoints
the dispatcher requires there.
install/omarchy-apple-silicon.packages names the Mac's runtime and boot
packages, plus the defaults an owner may remove: the startup-volume tool,
hardware video decode, Widevine, and wf-recorder, since gpu-screen-recorder
can't capture on Apple Silicon and screen recording falls back to it. What a
Mac can't work without (its speaker stack, PipeWire's PulseAudio server and
realtime scheduling, the GPU's Vulkan driver) comes as omarchy-mac's
dependencies, so a new Mac driver needs no Omarchy change. Other platforms
compose exactly what they did.
Co-authored-by: Scott Jones <444693+scottjones@users.noreply.github.com>
* Give existing aarch64 machines the zram swap Omarchy configures
omarchy-settings configures swap on zram and tunes memory reclaim for it.
x86_64 gets zram-generator from its installer and new aarch64 installs from
their default packages, but an aarch64 machine installed before
omarchy-aarch64.packages named it has no generator, so the configuration asks
for a device nothing creates.
Migration 1790328426 installs zram-generator where the platform's default
packages name it and it is missing, and starts the zram swap wherever it is
down, unless no device is configured (then the generator makes no swap unit)
or the administrator masked it. A swap that won't start, or whose unit can't
be read, only warns: the generator's unit starts with every boot, so a reboot
retries it, and failing the migration would stop every update. A failed
install waits (75) for the next migrate, as it does where the platform can't
be told, changing nothing. It keeps the name it has in omarchy-mac, so Macs
that already ran it skip it. It also retires an /etc copy of the drop-in
identical to the packaged one, which earlier Mac setups made and which would
hide the packaged file's updates; an edited copy stays. On x86_64 the default
packages don't name the generator and there is no copy, so it does nothing.
* Keep an ARM machine's repositories when it changes channel
omarchy-refresh-pacman and install finalization copied x86_64 templates
everywhere, so on an aarch64 machine a channel change or a finished install
wrote Arch Linux's x86 [core]/[extra]/[multilib] and dropped Arch Linux ARM
and the platform's own repositories.
Channel changes now tell the platform apart. x86_64 is unchanged. On aarch64
the machine's configuration is kept, with its repositories, mirrors and
signature policies, and only the [omarchy] server moves to the new channel;
a configuration whose Omarchy server can't be identified is refused
unchanged, and one with no Omarchy repository gets it at the end (not on
Apple Silicon, whose own repository must not come first). Only pacman.conf is
backed up and rewritten, straight from this process through sudo, never
through a file another process could change first.
Before that, a channel must be qualified for the platform: a signed upgrade
and reboot tested there, not only a published repository. x86_64 keeps
stable, rc and edge; Snapdragon and other aarch64 machines take edge, as
dragon does; Apple Silicon takes none until the Mac packages qualify on edge.
The refresh then checks, as the user, that the channel's omarchy.db
publishes the runtime pair, and on Apple Silicon omarchy-mac and
omarchy-mac-boot. omarchy-channel-set refuses an unqualified channel before
anything else, the dev confirmation included.
Install finalization writes a whole template, as before on x86_64. Other
aarch64 platforms get Arch Linux ARM (dragon's pacman-aarch64.conf and
mirrorlist-aarch64, byte for byte) with its keyring, plus Omarchy on a
qualified channel. Apple Silicon gets omarchy-mac's template on a qualified
channel and keeps its image's configuration otherwise. In an image build,
finalization leaves the keyring to the machine's first boot, so machines
flashed from one image don't share a pacman master key. On an unqualified
channel, a machine that gets no Omarchy repository is told why.
Co-authored-by: Birk Skyum <74932975+birkskyum@users.noreply.github.com>
Co-authored-by: Jim Martin <1334144+jdvmi00@users.noreply.github.com>
* Keep Arch Linux ARM's signing keys current on ARM machines
omarchy-update-keyring reinstalls archlinux-keyring before every system
upgrade, since keys change without a version bump. An aarch64 machine's Arch
Linux ARM repositories are signed with keys from archlinuxarm-keyring, which
nothing refreshed, so a rotated ALARM key failed the upgrade that followed.
On an aarch64 platform where archlinuxarm-keyring is installed, it is
reinstalled in the same transaction. On x86_64 the keyring step runs exactly
as before, even where that package happens to be installed.
* Mark the runtime's platform files as deciding at runtime
The omarchy-settings recipe leaves the mkinitcpio, Limine, memory and USB
drop-ins out of its aarch64 package, because they used to assume x86_64. That
leaves an aarch64 install without the HOOKS baseline its initramfs needs, and
without the zram and reclaim settings every other Omarchy machine gets.
Each of those files now decides at runtime whether it applies: the HOOKS
baseline asks the platform, thunderbolt is added only where the kernel has the
module, the Limine files are read only where Limine is installed, the oomd
drop-ins are inert without systemd-oomd, aarch64 installs the zram generator
the zram drop-in needs, and the USB autosuspend option is ignored by a kernel
that builds usbcore in. default/settings-runtime-profile says so, file by
file; a recipe that finds it ships the same files on every architecture
(omacom/omarchy-pkgs#638), and older sources keep today's aarch64 package. A
test checks that every file it names exists and that each runtime decision is
in place.
* Set the wireless regulatory domain from the owner's timezone at first-boot setup
A deferred install (an Apple Silicon image, an OEM ISO install) runs its
hardware steps before the owner picks a timezone, on the UTC placeholder,
which names no country, so install/hardware/set-wireless-regdom.sh leaves the
regulatory domain unset and the radios stay on the world domain. Owner setup
now runs that step again once the owner's timezone is set, and applies it to
the radios through wireless-regdb's helper, since no reboot follows setup. An
up-front install, which picks the timezone before its hardware steps, is
unchanged.
* Ask again for a keyboard layout first-boot setup could not apply
First-boot setup kept going when the chosen layout did not take: loadkeys
errors were ignored, a keymap localectl did not list kept the default, and a
failed or partial write of /etc/vconsole.conf was only logged. The owner then
typed the disk password in one layout while the boot files asked for it in
another, and could only unlock the disk by typing US-style.
The keyboard step now fails closed. On a console loadkeys must succeed, the
keymap must be in localectl's list exactly (a list localectl cannot produce
counts as a failure too, logged apart), and after systemd-firstboot or the
localectl fallback, /etc/vconsole.conf is read back in a clean shell and must
name the chosen keymap with an XKB layout. Any failure is logged, the owner is
told the layout could not be set, and the layout is asked for again, so the
password form is never reached with a layout that did not take.
* Refresh a platform's own signing keyrings before an update
omarchy-update-keyring reinstalls archlinux-keyring (and Arch Linux ARM's on
aarch64) before the system upgrade, so a rotated key never fails that
upgrade's signature checks. A platform's own repository has a keyring of its
own that nothing refreshed: on Apple Silicon, asahi-alarm-keyring for
[asahi-alarm].
A platform package now lists those keyrings in
/usr/share/omarchy-platform/keyrings, one package name per line, and
omarchy-update-keyring reinstalls each one that is installed in the same
transaction. It adds no repository the machine didn't already trust. A line
that isn't one keyring package name is skipped with a warning, so a mistake in
the list can't stop every update. Without the file, as on x86 today, the step
runs as before.
* Drive an untyped default input from the audio panel
Quickshell types only exact media classes, so a virtual source (an
Audio/Source/Virtual: EasyEffects' source, or the microphone mapping a
platform package adds) has no PwNode.audio. With one as the default input the
audio panel showed the input at 0% and its slider changed nothing, the peak
meter stayed flat, the microphone widget read it as muted, and the panel could
not list it unless it was already the default.
Such an input now goes through wpctl for its volume and mute (UntypedInput in
shell/Commons, one reader and one pactl subscribe for the whole shell), the
panel meters it with the new omarchy-audio-source-level, and the input list
takes an untyped node PulseAudio lists as a source.
A platform whose audio runs through its own processing can also name nodes
that are neither devices nor apps in /usr/share/omarchy-platform/audio.json,
which Omarchy ships none of: the panel and the microphone widget leave them
out, and a "replaced" node only while its replacement exists. Without the
file nothing is hidden.
On every machine: the input list leaves out a source whose ports are all
unavailable (an empty headset jack) unless it is the default, as the output
list does for sinks; choosing an input no longer moves a filter's own capture
(a source output with node.link-group), which would feed the filter from the
new input or from itself; the shell's own level meters no longer light the
microphone widget; and omarchy-audio-sink-availability reads pactl in the C
locale for sinks too, since "not available" is translated.
* Give first-boot setup an XKB layout for keymaps systemd cannot map
systemd-firstboot writes an XKB layout only for keymaps listed in
kbd-model-map. Nine the setup form offers are not (Colemak, Azerbaijani,
Bulgarian, Czech, Swiss German, Kyrgyz, Norwegian, Polish, Ukrainian), so it
wrote KEYMAP= alone, the readback refused it, and owner setup could never get
past the keyboard step for them.
Setup now adds the XKB settings itself when firstboot left none, matching the
console map so the shared password types the same on the desktop: kbd's cz is
QWERTY, bg-cp1251 phonetic and azerty French AZERTY, and non-Latin layouts keep
a US layout to switch to, as kbd-model-map's own rows do. A test checks every
layout the form offers ends with an XKB layout.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Codex XHigh <noreply@openai.com>
* Leave keys a firmware key turned off dark
The ambient keyboard backlight treated any 0 the keyboard-brightness keys had
not recorded as an accident and relit it within five seconds. Firmware Fn keys
(Fn+Space on ThinkPads and Framework laptops, Fn+F10 on Dells) write the LED
directly, so on those laptops turning the keys off never stuck.
The lock screen's off now records a blank, and only that blank or a 1%
leftover is taken back; any other off is a choice that pauses auto like the
keys do. A wake clears the blank only once the restore succeeded, the command
runs one at a time so a blank racing a wake can't leave the LED and record
disagreeing, and the loop drops a blank once it has taken the keys back.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Codex XHigh <noreply@openai.com>
* Drop the optional-install availability guards
Each menu open checked every Install row against the sync databases, which
costs most of a second of pacman work on every machine to hide rows only an
ARM machine is missing. The fix for those is publishing the ARM packages, not
hiding the rows, so this removes omarchy-pkg-available, the Install rows'
availability and architecture guards, and reinstall's skipping of defaults no
repository offers.
The Windows VM row keeps its x86_64 guard, since no package fixes a guest that
needs an x86_64 host, and reinstall keeps its aarch64 channel handling.
This reverts commit aea001cbde4d0852b311ef42b5e8c380c719f833.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Codex XHigh <noreply@openai.com>
* Put the current root back when the factory root cannot take its place
Factory reset moved the running root aside, marked the reset committed, then
renamed the staged factory root to @. If that second rename failed, cleanup
saw a committed reset: it skipped the boot package's rollback and kept the
throwaway LUKS slot, and since the swap had not happened it deleted the staged
root as well, leaving nothing at @.
The reset now commits only once the factory root is at @. A failed activation
renames the old root back first, then cleanup rolls the platform's boot state
back and revokes the throwaway slot. If that rename fails too, both roots are
kept, cleanup still rolls back and revokes, and the error names the rename
that recovers the machine. The generic Limine path has already rebuilt its
boot files from the factory root by then, as it had before this change.
* Leave other recorders alone when the saved recorder pid is stale
A saved pid that no longer named a recorder made status and stop select every
gpu-screen-recorder and wf-recorder the user ran, so stopping could interrupt,
and after five seconds kill, a recording started somewhere else.
A stale pid now means no recording of ours is running: the toggle starts a new
one, and --stop-recording finds nothing to stop. Stop acts on the recorder the
status check selected rather than selecting again. Without a saved pid, as for
a recording started before the pid was saved, every recorder the user runs
still counts.
* Name the audio a wf-recorder recording could not capture
With desktop and microphone audio both requested, wf-recorder records a null
sink fed by two loopbacks, and a loopback that failed to load was ignored: the
recording went ahead without that source, or silent when both failed. A
missing null sink or default device dropped audio just as quietly.
A source that cannot be captured is now left out with a notification naming
it, the loopbacks that did load are unloaded, and the recording carries the
source that remains, or no audio track at all.
* Show the Edge, Spotify and Dropbox installers on x86_64 only
These vendors currently publish Linux builds for x86_64 alone (the AUR's
microsoft-edge-stable-bin and spotify are arch=('x86_64'), and Dropbox lists
no Linux ARM client), so no native aarch64 build is available and choosing
one on an ARM machine only fails. Like the Windows VM row they keep a uname
check, which the menu answers once per open without asking pacman. The other
Install rows are unchanged.
* Revoke the generic reset's throwaway slot when the reset fails
The generic Limine path added its throwaway LUKS slot without recording it,
so a reset that failed before activation never revoked it. Once the factory
root's UKI was on the ESP, a failed activation that put the old root back left
a machine that booted the factory kernel and unlocked unattended with a key
readable on the ESP, while the screen said the current root was back.
The generic slot is now recorded like the platform's and revoked by cleanup
whenever the reset fails before the factory root is active. A failed
activation on that path says the boot files were already rebuilt from the
factory system and to run limine-update before rebooting.
* Clear a stale recorder pid when a recording starts
The recorder's pid is saved only once its file exists. Until then a pid left
by a recording that ended without a stop still counted as stale, so pressing
the toggle again while the new recorder started found nothing of ours and
could start a second recording. Starting a recording now clears that pid,
and during startup the toggle falls back to the user's recorders as before.
* Word the dropped-audio notice for every reason a source is left out
The notice said the dropped source "could not be captured", which was wrong
when the null sink itself failed to load: the microphone was never tried, and
desktop audio recorded alone as it always has. It now says the source could
not be added to the recording, true for a missing device, a failed loopback
and a failed null sink alike.
* Show the Heroic and Minecraft installers on x86_64 only
Heroic publishes Linux builds for x86_64 alone (its arm64 builds are for
macOS and Windows), and Mojang's Linux launcher is a single x86_64 tarball.
They join Edge, Spotify and Dropbox under the same uname check, so every row
whose vendor ships no native Linux aarch64 build is treated alike.
* Ask the recorder script whether its stop has a recording to end
The bar indicator and the menu's Stop Screenrecording row asked the process
helper whether any recorder the user runs was alive, while the toggle's stop
acts only on the recording it saved. With a stale saved pid alongside an
unrelated recorder, both offered a stop that did nothing.
omarchy-capture-screenrecording --status answers with the selection the stop
uses, and both now ask it. Status needs no recordings directory and never
notifies, so the recordings-directory check moved below argument parsing and
skips it.
* Copy every platform's pacman templates into place, as x86_64 does
On aarch64 a channel change edited the machine's pacman.conf in place: an awk
pass found the [omarchy] server and rewrote its channel, appended a section
where there was none, checked the channel's omarchy.db first, and gated it all
on a per-platform table of qualified channels. Install finalization appended
Omarchy to a template, or on a Mac copied one bundled inside omarchy-mac. What
a machine ended up with was never in the repository.
Each platform now has a pacman.conf and mirrorlist per channel, copied whole:
x86_64's stay in default/pacman, Snapdragon and other aarch64 machines use
default/pacman/aarch64, and Apple Silicon default/pacman/apple-silicon, with
Omarchy and Asahi ALARM ahead of Arch Linux ARM. Omarchy publishes aarch64
packages on edge alone so far, so every ARM channel points at edge for now and
channel changes work there; moving a channel is a template edit. A channel
without both files is refused before anything changes, including a dev
checkout's own templates. Reinstall goes back to resetting to stable, and the
update confirmation loses the missing-repository notice it no longer needs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Codex XHigh <noreply@openai.com>
* Revert "Copy every platform's pacman templates into place, as x86_64 does"
This reverts commit 9a07e138cc6c3a19b58d2f3d7eb536dd3042ccc5.
* Copy every platform's pacman templates into place, as x86_64 does
On aarch64 a channel change edited the machine's pacman.conf in place: an awk
pass found the [omarchy] server and rewrote its channel, appended a section
where there was none, checked the channel's omarchy.db first, and gated it all
on a per-platform table of qualified channels. Install finalization appended
Omarchy to a template, or on a Mac copied one bundled inside omarchy-mac. What
a machine ended up with was never in the repository.
Each platform now has a pacman.conf and mirrorlist per channel, copied whole:
x86_64's stay in default/pacman, Snapdragon and other aarch64 machines use
default/pacman/aarch64, and Apple Silicon default/pacman/apple-silicon, with
Omarchy and Asahi ALARM ahead of Arch Linux ARM. Omarchy publishes aarch64
packages on edge alone so far, so every ARM channel points at edge for now and
channel changes work there; moving a channel is a template edit. A channel
without both files is refused before anything changes, including a dev
checkout's own templates. Reinstall goes back to resetting to stable, and the
update confirmation loses the missing-repository notice it no longer needs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Codex XHigh <noreply@openai.com>
* Drop the migrations no installed machine needs
Fresh installs mark every migration done, and every Apple Silicon machine is a
fresh install, so the Mac handoff to omarchy-mac-boot (and the migrate
operation it called), zram for aarch64 installs that predate the package list,
the Brave Widevine link and the Broadcom migration's Apple Silicon guard never
run anywhere. The exit-75 deferral in omarchy-migrate existed only for two of
them. Keyboard backlight auto stays on new installs through first run rather
than switching on for existing ones, and migration 1786605598 goes back to
quattro: the mkinitcpio drop-ins arrive by package.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Drop the generic changes Mac support doesn't need
The pacman platform guard (and the detector copy omarchy-settings shipped for
it, which the HOOKS baseline fell back to), the lint against uname gates, first
run staying pending after a failed finalization, run_logged echoing failures to
stderr, half-up battery rounding, hiding empty-jack inputs and leaving a
filter's capture alone when choosing an input all rode along with the Mac work
without serving it. Each goes back to quattro's behaviour and can come back on
its own merits.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Drop the platform Hyprland slots
A platform package could load its own Hyprland binds, settings and gestures:
binds that replaced Omarchy's chords and decorated every later bind, settings
ahead of the theme, and gestures after the user's files, which needed every
hl.gesture call wrapped so a default could step aside. On a Mac those carried
macOS habits (a three-finger workspace swipe, tap-to-click off, menus opening
on the built-in screen, other screenshot and keyboard-backlight chords), not
anything the hardware needs, and Omarchy should behave the same on every
machine. The one bind a Mac does need, its lid switch under the SMC's name,
joins Omarchy's own lid binds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Keep aarch64 machines off the release channels until a release supports them
Every aarch64 channel's template pointed at the edge repository, but a switch
to stable or rc installs the release-named omarchy and omarchy-settings. On
edge/aarch64 those are the 4.0.4 release line, which has no omarchy-hw-platform,
lifecycle dispatch or omarchy-pkg-defaults, so `omarchy-channel-set stable` on
a Mac replaced the quattro runtime it runs with one that cannot run it, and
succeeded.
aarch64 platforms now have edge templates alone, so a switch to stable or rc is
refused before anything changes; adding those templates once a release
supports aarch64 opens them. Refresh with no channel named, as reinstall runs
it, and install finalization for a channel the platform has no templates for
take the platform's default channel: stable on x86_64, edge on aarch64.
* Leave a filter's own capture alone when choosing an input again
d0b94d98c returned omarchy-audio-input-set-default to moving every source
output, as not serving Mac support. It does: asahi-audio's microphone DSP on
Apple Silicon records the raw array through a filter-chain capture stream
(audio_effect.jNNN-mic, node.link-group "filter-chain-..."). On an M1 Pro,
choosing the Mac's microphone in the audio panel moved that stream onto the
DSP's own output, and the microphone recorded digital silence (-91 dB, from
-55 dB of room noise) until PipeWire restarted. With the skip it keeps
recording (-54 dB).
This restores the skip of source outputs carrying node.link-group, and its
test, as they were before that commit.
* Say aarch64 offers edge alone where the templates are described
The aarch64 edge templates, the template helper and the update-process docs
still said every platform has a template per channel and every ARM channel
installs from edge, which the previous commit made untrue.
* Point the Mac packages at omacom/omarchy-mac-pkgs
omarchy-mac and omarchy-mac-boot now live in omacom/omarchy-mac-pkgs, with
each package at the top level. The dispatcher, its documentation and the
Apple Silicon package list name that repository, and the drop-in test's
fixtures, byte for byte the same as omarchy-mac-boot/files/etc/mkinitcpio.conf.d
at 4399105, cite it there.
* Stop naming a platform migration among the dispatcher's callers
The migrate operation went with the Mac migrations, and no caller hands a
machine to a platform migration any more, so the header lists only the
flows that still dispatch. The optional-operations comment is rewrapped.
* Keep the Intel Mac Broadcom migration off Apple Silicon again
A fresh install marks every migration done only for the user it creates, so
an account added later runs them all on its first update. On an Apple
Silicon Mac this one matches: an M1 Pro reports sys_vendor "Apple Inc." and
a BCM4387 (14e4:4433), so it would append feature_disable=0x82000, which
stops that chip scanning once brcmfmac reloads. Restore the guard and its
test case.
* Say when Steam's post-install hook runs now that the drivers go first
Steam's installer now adds the 32-bit drivers before Steam itself, so the hook runs after both and before the first launch.
* Release fallback audio after the recorder exits
When wf-recorder ends outside the normal Stop path, status rejects the saved PID before unloading the combined-audio loopbacks. Release those owned modules when the saved recorder is gone, so the next status or stop call ends microphone capture without signaling another recorder. Leave a mix without a saved PID alone: its recorder can still be starting.
Co-Authored-By: GPT-6 in Codex <noreply@openai.com>
Co-Authored-By: Codex Medium <noreply@openai.com>
* Serialize recording state changes and stale audio cleanup
Status can finish checking an exited recorder after another command has replaced its audio modules, and stop finalization can outlive the recorder while a new recording starts. Share a lock across start, stop and cleanup so each command finishes updating its resources before another changes them. Status uses a read-only check when the lock is busy, keeping the bar responsive during the picker. Close the lock descriptor in persistent children so recording and preview cleanup cannot retain it.
Co-Authored-By: GPT-6 in Codex <noreply@openai.com>
Co-Authored-By: Greptile <165735046+greptile-apps[bot]@users.noreply.github.com>
* Keep automatic and manual keyboard brightness updates together
A brightness key can update the LED and its chosen-level record after the automatic loop checks for a manual change but before it applies a new target. The automatic write then overwrites the user's choice and deletes the record. Take the manual command's lock for the whole tick, skipping a busy tick and releasing the descriptor on every return. Create the private fallback directory before opening the lock outside a session runtime directory.
The normal lock screen waits five seconds before blanking, so an ordinary tick should already see the session locked. The regression exercises a brightness key arriving within a tick instead of relying on an implausibly delayed lock-state check.
Co-Authored-By: GPT-6 in Codex <noreply@openai.com>
Co-Authored-By: Claude Opus 5.5 High via Pi <noreply@anthropic.com>
* Read the split hook baseline in the NVIDIA migration
The NVIDIA firmware migration decides whether to rebuild by sourcing the hardware and filter drop-ins. With HOOKS now defined in the earlier baseline, that partial read treats every hybrid machine as if kms had been removed. Source the new baseline first when it exists, keeping the legacy two-file layout working and skipping rebuilds when the effective hooks still include kms.
Co-Authored-By: GPT-6 in Codex <noreply@openai.com>
Co-Authored-By: Claude Opus 5.5 High via Pi <noreply@anthropic.com>
* Use a hardware helper for x86-only menu entries
Keep the x86_64 availability checks for Windows and vendor-specific installers in a hardware predicate, so menu conditions use the same command style as other hardware checks.
Co-Authored-By: GPT-6 in Codex <noreply@openai.com>
* Keep restored keyboard levels from pausing automatic brightness
Idle wake, quick unlock and hibernate can restore a saved level older than the ambient loop current level without the lock blanking first. Treat that restoration as a one-tick event, preserving existing manual pauses and the normal blank-and-restore path. Consume unmatched and unchanged restore records too, so a later firmware key cannot look like the earlier restore.
Co-Authored-By: GPT-6 in Codex <noreply@openai.com>
Co-Authored-By: Claude Opus 5.5 High via Pi <noreply@anthropic.com>
* Preserve restored keyboard levels across a later lock blank
The ambient loop skips a locked session, so a restored level may still be unconsumed when the next blank saves it. Carry that origin through blanking and restore failures so fingerprint unlock cannot make the old saved level look like a manual choice. A brightness key replaces the marker with its chosen level and retains the ordinary blank behavior.
Co-Authored-By: GPT-6 in Codex <noreply@openai.com>
Co-Authored-By: Claude Opus 5.5 High via Pi <noreply@anthropic.com>
* Preserve firmware brightness choices before a lock blank
Firmware keys change the LED without updating the record. Carry a pending restoration through the next blank only if the observed LED still matches that restoration, or is still zero from its previous blank. A firmware change otherwise follows the ordinary blank-and-restore path so automatic lighting pauses on the chosen level.
Co-Authored-By: GPT-6 in Codex <noreply@openai.com>
Co-Authored-By: Claude Opus 5.5 High via Pi <noreply@anthropic.com>
* Change only the disk password in omarchy-drive-password
Changing the system disk's password also set the login and root passwords
to match. That changes account-password policy on every platform, x86
included, and overwrites passwords someone chose to keep different, so it
moves to its own change with an explicit confirm.
The disk change itself stays as it was: the new key must open the disk and
the old one must not before the change counts, the journal finishes an
interrupted change, and where the platform's boot checks prove the key
slots, the owner's new slot is recorded before the run finishes. The
journal's confirmed phase is now "record" and no longer needs a login user.
* Cancel a pending recording start instead of queueing behind it
Starting a recording holds the recorder lock through the picker and until the
recorder creates its file. A second toggle or a stop then waited on that lock,
with no end while a portal pick or a stuck recorder never finished, and once
the start gave up the queued toggle began another recording.
While a start is pending, a toggle or stop that cannot take the lock within a
second now leaves a cancel marker and returns. The start checks it after the
picker, after the webcam overlay and while waiting on its recorder, then stops
that recorder and releases its audio. A recorder other than the portal's that
has not created its file after 15 seconds is stopped and the next backend
tried.
* Move automatic keyboard lighting to its own change
Ambient-light keyboard backlight is a new feature on every laptop with the
sensor, not something Apple hardware needs to run Omarchy, so it leaves this
change for a separate review. The keyboard-brightness command returns to
upstream's version, and first run no longer enables the user service.
* Move workspace refocus on monitor removal to its own change
Keeping the viewed workspace in front after a monitor is unplugged changes
desktop behaviour on every machine, not just Apple hardware, so it leaves
this change for a separate review.
* Move prebuilt-image hardware deferral to its own change
Building an image away from its machine needs a manifest the detector reads,
a queue of hardware steps for the first boot, a first-boot service and a
per-machine pacman keyring. An install on the machine itself needs none of
it, so it leaves this change for a separate review with an installer and
package boundary of its own. Hardware and user setup now run where they are
installed, and the platform setup operations take no image argument.
* Move the Qualcomm platform to its own change
Snapdragon support is separable from Apple compatibility, so the detector no
longer names a qualcomm platform: a "qcom," device tree is generic aarch64,
which Snapdragon already behaves as everywhere else here, and a tree naming
both Apple and Qualcomm is still refused. linux-firmware-qcom's platform list
leaves with it.
* Keep a cancel that arrives while a recording start finishes
A stop could land after the start's last cancel check, while it saved its
recording as started, and be dropped with the marker. The start now removes
its starting marker before a final check, a request reads that marker only
after leaving its own, and one that finds the start gone waits for the lock
and acts unless the start already used its request. A recorder ended by a
cancel or a readiness timeout gets two seconds after TERM before KILL, so a
recorder ignoring TERM cannot hold the start and its lock.
* Defer an image build's hardware setup to the machine's first boot
Brings back prebuilt-image support split out of #14431 so it gets its own review. An image built away from its machine names its target in a root-owned manifest the detector reads; hardware setup is queued for the machine's first boot, run there by omarchy-provision-hardware, which also makes the machine's own pacman keyring; and the platform setup operations learn they run on an image's first boot. Apple Silicon images built by the omarchy-mac installer need this; an install on the machine itself does not.
* Rebuild the initramfs when deferred setup changes module options
mkinitcpio's modconf hook copies /etc/modprobe.d into the initramfs, but the
deferred first-boot setup only compared mkinitcpio's configuration and the
kernel command line when deciding whether to rebuild. A step that only wrote
module options, as nvidia.sh does, left the old ones in the boot image.
* Put the Omarchy repository first on aarch64, as on x86_64
quattro now searches [omarchy] before the Arch repositories (
|
||
|
David Heinemeier Hansson
|
5657a91f8d |
Drop Lazydocker from new installs (#14230)
* Drop Lazydocker from new installs * Preserve Lazydocker during Quattro upgrades * Remove Lazydocker from the manual |
||
|
|
393a43d469 |
Pin root= before the packages that can drop it (#6951)
* Pin root= before the packages that can drop it limine-entry-tool falls back to /proc/cmdline for root= only while nothing appends to KERNEL_CMDLINE. Installing omarchy-settings lands a drop-in that appends with +=, switching that fallback off, and a kernel bump in the same transaction then bakes a UKI with no root= at all. preserve_kernel_cmdline_root repaired that afterwards, so a completed upgrade booted — but the machine was unbootable for the seconds in between, and an upgrade interrupted there left it in an emergency shell. Pinning cannot wait until after the packages land, and the old guard could not be moved earlier as it was: it asked the tool for its effective cmdline, which still resolves root= through the fallback right up until the drop-in arrives, so it reported healthy on exactly the machines about to break. Ask the config layers whether root= is stated explicitly instead, for the default profile alone, and pin before the package transaction. Verification stays after it, since that is what rebuilds the UKIs. The pin no longer gates on limine-mkinitcpio being present, since it now runs before the transaction that can install it, and a machine still missing it is exactly one that needs pinning first. Verified in a VM upgrading a 3.8.0 install, sampling the UKI every 2s across the upgrade: a legacy install without the pin lost root= for ~10s, and booting that state landed in "Failed to mount '' on real root". With this change the same upgrade never loses it, and an install that already pins root= is untouched. Closes #6894 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Merge the limine config layers when checking for a root= pin, and re-pin after the transaction The pin check returned on any line that mentioned root= under KERNEL_CMDLINE[default]. limine-entry-tool merges its layers in order, /usr/share drop-ins, /etc/limine-entry-tool.conf, /etc drop-ins, then /etc/default/limine, where = replaces and += appends, so a later layer's = could remove a pin the check still counted, and a quote before root= let systemd.setenv="root=..." read as one. Both skip the pin, which is the direction that bricks the next boot. The check now merges the layers the same way, strips only the outer double quotes the tool strips, drops quoted values the kernel does not split, and looks for root= as a parameter of its own. It was checked against limine-entry-tool --get-cmdline on a worker for every fixture in the test. Installing limine-mkinitcpio-hook runs limine-install, so a machine with no /boot/limine.conf when the pin runs ahead of the transaction can have one after it, and verification only warned about the UKI it found without root=. Verification now runs the pin for any machine the first call skipped. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Codex Medium <noreply@openai.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com> Co-authored-by: omarchybot <omarchybot@users.noreply.github.com> Co-authored-by: Codex Medium <noreply@openai.com> |
||
|
David Heinemeier HanssonandClaude Fable 5.1
|
d78ea9e158 |
Point GitHub URLs at omacom/omarchy instead of basecamp/omarchy
The repo moved to the omacom org. GitHub redirects the old URLs, but `omarchy channel set dev` was still cloning from basecamp/omarchy, which left every dev checkout with a stale origin remote that confuses gh (pr create fails with "No commits between omacom:quattro and basecamp:<branch>"). Update the clone URL, the quattro upgrade tarball, the update-confirm release link, the systemd Documentation link, and the manual. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LSFKDatumRZHB8zqk5CP5C |
||
|
Ryan Hughes
|
f5cd244ed0 | Retire the stock user icon font missed by Quattro upgrades | ||
|
Ryan Hughes
|
3f91f06a62 |
Merge pull request #10225 from mdisec/security/harden-quattro-fprintd-path
Harden lock authentication command lookup |
||
|
|
41b6cc6965 |
Add native video wallpaper support (#6792)
* Add native video wallpaper support * Pause video wallpapers while a fullscreen app is focused * Sample one frame when a video background sets the bar text colour A video wallpaper made the transparent bar's colour sampling decode the entire file. ImageMagick's video delegate runs ffmpeg with no frame limit, so a twenty-second 1080p background took 11.3s of CPU where one frame takes 0.14s, and it did that on every theme change. The result was unusable anyway: a multi-frame input emits one value per frame, which the single-value match then rejected, so transparent bars silently fell back to the plain text colour on every video wallpaper. Selecting frame zero fixes the cost and the colour together, and fixes animated GIFs, which had the same bug. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Load wallpaper video lazily, and without an audio output Three costs the still-image path should never have paid. BackgroundMedia imported QtMultimedia at file scope and was instantiated on every output, so the module and its audio dependency closure mapped into every shell process whether or not a video was ever shown — measured at +2.72 MiB RSS. Moving the element into its own file behind a Loader that takes a URL defers the whole import: an inactive loader maps none of it, an active one maps all 25 libraries. An inline Component cannot defer that, because the type has to resolve when the file compiles. Qt's Video convenience type always builds an AudioOutput, and `muted` only aliases that sink's volume, so every monitor decoded an audio stream it would never play and opened an audio client for it. A bare MediaPlayer with no audio output spawns no QFFmpeg::AudioR, QAudioContext or PWDevMon thread, and plays files with no audio track just the same. The shared image also turned mipmapping on, which the desktop background never had. A full mip chain is about a third more texture memory — 10.6 MiB extra at 4K, per output — for a wallpaper drawn at its own size. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Stop wallpaper playback while the session is locked or screensaved Playback stopped only for a focused fullscreen window. Locking the session did not stop it, and the lock screen starts a player of its own, so an N-monitor desktop reached 2N decode pipelines the moment it locked — and stayed there, because a display blanked for idle stops being presented but does not stop Qt's FFmpeg engine, which drives its own clock. A laptop locked with the lid shut decoded video until the battery ran out. The lock and idle services already know both states, so the background service takes the shell reference the loader offers it and reads them. Looking a service up by id needs the registry to be reactive, or a background that loads before the lock service would bind to null and stay there. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Fan out video thumbnails narrower than single-threaded image jobs The generator fans out one job per core, which was bounded because VIPS_CONCURRENCY=1 made each of them single-threaded. ffmpegthumbnailer leaves FFmpeg's automatic decoder threading on, so a folder of uncached videos put a codec thread pool on every core at once. Queueing video work separately keeps the still-image path at full width and gives the video path a quarter of it. * Recognize a named video file as a theme preview The backgrounds fallback beside it already picks videos, so a theme shipping preview.mp4 was the one case that still went unseen. * Document video backgrounds in the manual The manual described backgrounds as images only. Worth saying plainly that a video wallpaper costs far more power than a still one and that each monitor decodes its own copy, since neither is visible from the picker. * Stop the lock screen's own playback once the displays go dark Pausing the desktop wallpaper on lock only moved the cost. The lock screen builds a player per monitor of its own, so locking an N-monitor session went from N decoders to N rather than to none — and the lock service blanks the displays five seconds later without touching them, which is where a lock spends nearly all of its time. A laptop locked and shut still decoded video into a dark panel. The service already owns both transitions, so it records whether the displays are dark and the lock view stops playback while they are. The manual said playback stops while the screen is locked, which was the same overstatement; it now says once a locked screen has gone dark. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Keep videos out of the lazy thumbnail path A lazy row stands in with the media file itself until its thumbnail exists, and the picker draws that with an Image — which shows a picture and shows nothing for a video, with no reload once the real thumbnail lands. So the first open after discovering an uncached video showed a blank tile. The same branch also spawns one generator per file immediately, before either queue is reached, and the theme switcher always asks for lazy thumbnails. That put the narrower video fan out on the one path that never used it: forty uncached previews meant forty ffmpegthumbnailer processes. Sending videos to the queue instead fixes the blank tile and puts them back under the cap. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Rebuild the theme preview cache after teaching it about video Preview discovery changed what it recognizes, but its cache keys on theme directory mtimes alone. A theme that already shipped a video preview would keep whatever the old rules cached until something happened to touch the directory. Bumping the version rebuilds it once. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Drop an activeAudioTrack setting that never took effect Qt's FFmpeg backend ignores setActiveTrack while no source is open, and the literal binding is not reapplied once the media loads and the tracks become known, so the line did nothing. What actually keeps the audio decoder and its client from ever being built is the absent audio output, which a file carrying an audio track confirms on its own: no QFFmpeg::AudioR, QAudioContext or PWDevMon thread appears without it. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Give up the blank state when a display comes back The lock screen stops its wallpaper while the displays are dark, but it was tracking the blanking it asked for rather than the panels themselves. Opening a docked lid turns the internal panel back on without going through runWake, and so does a resume, which left a visible lock wallpaper frozen on one frame until the next keypress. A frozen wallpaper someone is looking at is worse than the decoding it saves, so a screen change gives the state up. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Time bound the video thumbnail generator Routing videos through the queue means they are generated before the picker opens rather than behind it, which turned an unreadable or stalled file into a picker that never opens. ffmpegthumbnailer had no bound of its own and the drain waits for every job. A generator that gives up is already handled: the run reports failure, the partial file is removed, and the row drops out of the list. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Pause only the output a fullscreen window covers The fullscreen test was global, so a game on one monitor stopped the wallpaper on every other one — including the ones still in plain view. That is the failure the lock work was careful to avoid, and it made the manual's claim that playback stops when nothing can see it untrue for the commonest multi-monitor case. A lock or a screensaver does cover every output, so those stay a single decision; fullscreen is now matched against the focused monitor, the way the bar already routes by output. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Kill a video thumbnail generator that ignores the timeout Plain timeout sends TERM and then waits for a process that may never take it, which leaves the bound it was added for unenforced on exactly the stuck files it was meant to catch. Co-Authored-By: Codex XHigh <noreply@anthropic.com> * Pause video wallpapers in battery power-saver * Fix paused video wallpaper source priming * Skip snapshots for video background transitions (cherry picked from commit 6f759538bfa76c2da03634e98ebfc2ebf63ec68e) * Generate thumbnails for direct-scan videos (cherry picked from commit 10fcca018a865dca311fb6863e8c8b0057291223) * Remember a video the thumbnail converter rejected A permanently unreadable video cost ten seconds of generator time on every picker open before its row dropped, because nothing recorded the failure. Both the menu image generator and the direct picker scan now leave a marker beside the missing thumbnail, keyed like the thumbnail on the file's size and mtime, so a repaired file starts clean. A timeout is left to retry, as it may only have been a busy machine. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * Follow the panels' real DPMS state under a locked video wallpaper The lock screen stopped video playback when it asked for the displays to blank, and resumed on input, but never checked what the panels did. A blank that failed left a lit panel on one frozen frame, and a resume that turned the same outputs back on played nothing until the next keypress. Quickshell exposes no DPMS signal, so while a video is the locked wallpaper the lock polls hyprctl and decides per surface from the answer. A wake or blank request drops the last answer so its optimistic state applies until the next poll confirms it. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * Pause a video wallpaper for the fullscreen window that covers it The fullscreen check read the globally active window and the focused monitor, so it only knew about the window that had focus. A fullscreen window left on one monitor while focus moved to another resumed the wallpaper decoding behind it, and with fullscreen windows on two outputs only the focused one paused. Each output's visible workspace reports whether a fullscreen window covers it, and Quickshell flips that on the compositor's fullscreen event, so each panel now decides from its own monitor's active workspace instead. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * Reopen a video wallpaper a theme switch replaced behind its path Two themes that both ship backgrounds/wallpaper.mp4 leave the current background at the same path after a switch, so the displayed path never changed and the running player kept decoding the old file from its open descriptor. Stills go through the snapshot transition and survive this; a video switch is instant and did not. A forced switch onto the path already on show now bumps a reload counter, and BackgroundMedia rebuilds the video player for it. A cache-busting query is not an option there, since FFmpeg reads it as part of the filename. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * Keep picker rows uncached while a rejected video is left out Skipping a video with a failure marker let the picker cache its rows without it, and cached rows are trusted on the directory's mtime alone. A file repaired in place never touches that, so the marker's fresh key was never consulted and the video stayed missing. The generator now hands the marker back to the row loop, which drops the row and leaves the rows uncached, so each open re-stats the file and a repaired one is converted again. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * Hand each background loader only its own kind of file BackgroundMedia fed one URL to both the still loader and the video player. On a switch from image to video the Image was handed the video's URL in the moment before its loader unloaded, so Qt tried to decode the mp4 as a picture and logged an unsupported format on every such switch; the reverse handed the player a still to demux. The still URL is now empty whenever the path is a video and the video URL empty whenever it is a still, so a switch changes only the loader that stays. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * Stop a video wallpaper before tearing its player down Switching from a video to a still destroys the BackgroundVideo item while its player is mid-read, which FFmpeg reports as a failed open in the shell journal on every such switch. Stopping the player on destruction lets the demuxer wind down first, and the switch is quiet. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * Play a video wallpaper's sound track from the first monitor Video wallpapers were always silent: the player was built without an audio output, since a muted output still decodes the track and opens an audio client on every monitor. A video with music should be able to play it. The player now builds its AudioOutput only once the media reports a sound track, so a silent file still opens no audio client, and only the first screen's panel opts in, so a multi-monitor desktop does not layer copies of the track. The output is muted while a paused player primes its first frame, and the lock screen stays silent. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * Keep a departing video player off the still's file The switch away from a video still logged a cancelled open, and stopping the player on destruction only hid it: stopping reports the media as loaded, which the loaded handler answered by playing again. The real cause was one evaluation pass. Both URLs derived from the `video` flag, which is itself bound to the path, and QML updates the two in no fixed order, so the video URL could evaluate against the stale flag and hand the player the still for a moment. Its destructor then cancelled that open. Each URL now tests the path directly, the source binding only applies while the path is a video and restores nothing when it stops, and the destruction stop goes away with the hazard it introduced. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * Pin the audio wiring in the test and name the output in the manual The audio assertion passed with the BackgroundMedia forwarding binding removed, which would have left every wallpaper silent, and did not pin the silent default or the first-screen selection. It covers all three now. The manual said the sound track plays "from your first monitor", which reads as routing to that monitor's audio device. It is the first monitor's wallpaper that plays, through the default output. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> --------- Co-authored-by: Omabot <omabot@omarchy.org> Co-authored-by: Codex XHigh <noreply@anthropic.com> Co-authored-by: z8 <yam@kernelius.com> Co-authored-by: David Heinemeier Hansson <david@hey.com> |
||
|
Mehmet Ince
|
70047956fa |
Harden lock authentication command lookup
Keep the target user's local bin on user-scoped upgrade paths while giving privileged lock and firewall helpers only root-owned search directories. Pin the lock helper's root PATH and fprintd-list executable, with regression coverage for each defense independently. Reported-by: Rooke Poole <rookpool97@gmail.com> |
||
|
Ryan Hughes
|
58c399de30 |
Revert "Drop unrelated privileged heredoc scanner"
This reverts commit
|
||
|
Ryan Hughes
|
15f26cbe1b | Remove redundant migration command preflight | ||
|
Ryan Hughes
|
4c23077f80 | Drop unrelated privileged heredoc scanner | ||
|
Ryan Hughes
|
40d0c9bbdf | Require Quattro migrations to complete | ||
|
acrogenesis
|
d593847728 |
Fail the suite on privileged writes through unquoted heredocs
An installer that writes a root-owned file through a heredoc with an unquoted delimiter (<<EOF rather than <<'EOF') has the installing user's shell expand the body first, so a user-controlled value is baked in as a literal. Send that into /etc and root later reads or executes a path the unprivileged user picked: a udev rule carrying $HOME/.local/share/omarchy/bin/... resolves through a symlink that user owns, so replacing the symlink gets their code run as root. Add the static check. A heredoc is flagged when its delimiter is unquoted, its body contains an install-time expansion (escaped \$VAR does not count, since that is left for a root daemon to expand at runtime), and its output reaches /etc, /usr, /opt, /srv, /boot or /var/lib via sudo tee, sudo dd, a redirect, or an install/cp/mv of the generated scratch file. Destinations written as variables are resolved from the file's own assignments. Sites that genuinely need install-time expansion declare it inline: # omarchy:heredoc-expands paths=none -- $servers is a validated IP list paths= is machine-checked against the expansions the scanner finds to be path-shaped, so this cannot become a rubber stamp: adding a $HOME/... to an already-annotated heredoc makes the declaration false and trips the check again. Path expansions anchored under a root-owned prefix, as in "/etc/systemd/system/$unit", are correctly not path-shaped. Annotate the sites the scan reports, each of which expands a scalar: DNS addresses in omarchy-dns, a literal PAM line in omarchy-setup-security-fingerprint, kernel cmdline parameters and usernames in omarchy-upgrade-to-quattro. omarchy-provision-owner expanded a unit name that was already a constant, so its delimiter is now quoted and the name hardcoded; the generated unit file is byte-identical. omarchy-windows-vm declares paths=storage,shared, the only site that interpolates a user-chosen path. Fixtures prove non-vacuity in both directions: the write routes other than a pipe into sudo tee, the shapes that must stay quiet, udev rules and a shutdown unit taken verbatim from this repository's history, and the rubber-stamp case where a paths=none annotation on a baked $HOME path still fails. |
||
|
Ryan Hughes
|
c720f0b981 |
Merge pull request #8951 from omacom/cups-browsed-temporarily-removed
Temporarily remove automatic printer discovery |
||
|
David Heinemeier Hansson
|
bf20c94ea0 |
Merge quattro into the Chromium first-run EULA branch
Quattro stopped making the Chromium managed-policy directory world-writable while this branch was open, and the block it deleted from the theme install leaf sat directly above the comment this branch rewrites, so the two edits landed in one hunk. The resolution keeps the hardening — the policy directory is set up through install/config/browser-policy.sh now — along with the first-run seed and the comment that names both things the seed does. |
||
|
David Heinemeier HanssonandCodex XHigh
|
24c18df5b7 |
Temporarily remove automatic printer discovery
cups-browsed is the daemon that watches the network and creates print queues by itself. Hardening it took a root daemon with a predictable cache down to a confined service account, but a daemon that turns anything advertising itself on the network into a print queue is a lot of exposure for a convenience, so it comes out of the default install while that is reworked. Only the discovery half: CUPS itself stays and printing keeps working, with each printer added by hand in Print Settings. The migration disables the unit before removing the package because that is the only order that works: pacman deletes the unit file but not the enable symlink, and once the unit is gone systemd can no longer resolve it by name to clean that up. It then removes the queues discovery generated. cups-browsed keeps those when it stops, since KeepGeneratedQueuesOnShutdown defaults to Yes, and they route through its own implicitclass backend, which goes with the package, so they cannot print again. Idle ones go. A queue with jobs on it is left alone and named: implicitclass only needs cups-browsed to choose a destination, so a job already past that point finishes on its own, and deleting the queue would abort it. One printer's job does not hold up the removal. A printer added by hand has an ipp:// or usb:// device and is left where it is. A queue whose jobs cannot be asked about is left alone rather than assumed idle, including one named so that lpstat would misread it -- "all" is its word for every destination, and a leading dash or a comma reads as another option or a list. Where CUPS does not answer at all, or a queue will not delete, discovery is still stopped but the package stays and no marker is written. omarchy-migrate records a migration for the user as soon as it exits zero, so that is where the machine stays until someone removes the package by hand, and the message says so rather than implying a retry. The queue list is read under LC_ALL=C because lpstat translates "device for", and captured rather than piped, so a cupsd it cannot reach is reported instead of reading like a machine with nothing to clean up. It removes with plain pacman -R rather than omarchy-pkg-drop, which passes -n and would discard /etc/cups/cups-browsed.conf instead of keeping it as a .pacsave. A removal meant to be temporary should not delete the machine's copy of its own configuration. Without -s either, so it only ever removes the package it names: sweeping newly unneeded dependencies is nothing today, but it is not a promise a rolling dependency graph can keep. Queue names come off the network, since cups-browsed names its queues after what the printer advertised. CUPS allows every printable character but space, tab, / and #, and lpstat and lpadmin take a destination as an option value, so a name with a leading dash or a comma is reported rather than passed to them and guessed at. Migration state is per user, so a machine-wide marker records the one removal. Without it, an account whose first migration run came after someone deliberately reinstalled discovery would quietly take it back out again. The install-time override for cups-browsed.conf now waits for cups-browsed rather than for CUPS. Guarding it on a file CUPS still ships would write a configuration file for a package nothing installed, and pacman would later land the package's own copy beside it as a .pacnew. The hardened configuration stays in the tree. omarchy-settings still ships the cups-browsed.conf override, the sysusers account and the service drop-in, so they are what discovery returns onto. Co-Authored-By: Codex XHigh <noreply@openai.com> |
||
|
acrogenesis
|
bafc9a1000 |
Write browser theme colour through a passwordless helper
Managed policy dirs are enterprise trust roots, so they stay 0755 root:root. The menu path takes root for that one write through a sudoers glob of six hex digits, the same shape as omarchy-dns, and falls back to pkexec where the grant is not installed. Drop omarchy-browser-policy; a group member could plant any JSON, not just a colour. |
||
|
acrogenesis
|
95b791af16 |
Stop world-writable browser policy directories
Chromium managed policy is mandatory for every profile. World-writable dirs let any local uid plant policy, including force-installed extensions. Write goes through the omarchy-browser-policy group at 2775 so theme colour still works without other-write. |
||
|
Helmut Januschka
|
b91180a808 | Skip Chromium EULA on first run | ||
|
|
b5ded31e2f |
Don't put the user in the docker group; make it opt-in (#8056)
* Don't put the user in the docker group; make it opt-in The docker group is root-equivalent: anything in it can `docker run -v /:/host` and rewrite the host as root with no password. On a single-user box that's not an escalation (the owner is already a wheel/sudo user), but it hands any code running as the user — a rogue plugin, a poisoned dependency — a silent, headless, passwordless path to root that sudo's password prompt would otherwise gate. Stop granting the docker group by default. The daemon still runs (docker.socket); the Docker TUI and the Windows VM reach it through a polkit prompt, and the plain `docker` CLI runs under sudo. Sudoless Docker is a warned opt-in via Setup > Security (omarchy-setup-security-sudoless-docker). No automatic path may re-grant it: install and first-boot provisioning never record or apply the group (provisioning also filters a docker line left in an older factory snapshot), and the Quattro upgrade no longer adds it. The Windows VM keeps needing the root daemon for a privileged container (KVM, NET_ADMIN), so it is reworked to run without the group and without becoming a new way in: - The compose lives in a root-owned dir and is only written by an elevated, input-validated writer. A root-invoked bring-up must never consume a file a user-process could rewrite to bind-mount / into the guest — the old ~/.config/windows compose was exactly that. Volume paths are rebuilt from $HOME on migration rather than trusted from the (user-writable) legacy file, path validation rejects traversal, and the privileged sub-action is checked against an allowlist before dispatch (a slash in it would otherwise run as a path). - pkexec elevates a verified root-owned command path, not a PATH-resolved one, so an authorized prompt can't be redirected to an attacker's binary. - The guest password is kept in a private 0600 per-user file for RDP instead of a world-readable compose, and a declined authorization is reported as such, never as a completed stop. Existing installs auto-migrate the VM (no redownload) and refresh the stale Docker launcher entry. 🤖 Generated by Opus 4.8 in Claude Code. Reviewed by Codex XHigh. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Co-Authored-By: Codex XHigh <codex@openai.com> Claude-Session: https://claude.ai/code/session_01Gb7x6poap4hGCndPx5qt5T * Migrate existing installs off the docker group The default flip only reaches new installs; existing users keep their docker group membership and stay exposed. Extend the migration that already refreshes the Docker launcher to also remove the current user from the group when present, reusing omarchy-remove-security-sudoless-docker so there is one source of truth for the change and its notice. It takes effect at next login (the current session keeps working), and passwordless docker can be turned back on from Setup > Security > Sudoless Docker. Migrations run with sudo available — during `omarchy update`, or in the terminal the pending-migrations notification opens — so the privileged removal does not prompt at an unattended login. The no-op path (already out of the group) needs no privilege. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Gb7x6poap4hGCndPx5qt5T * Refuse symlinked VM mount sources; correct the docker CLI docs Review follow-ups. valid_path keeps a traversal string (/./, //, ..) out of the compose, but it is a string check: a symlink planted at ~/.windows or ~/Windows redirects the privileged bind mount exactly as traversal would, because docker follows it. So verify the mount sources as root immediately before bringing the VM up — refuse a source that is a symlink or resolves through one — which is where the string check cannot help. A missing source stays fine (docker creates a plain dir). Also correct the development-tools manual: the CLI is not transparently elevated (there is no docker wrapper and `d` is still plain docker), so say plainly that docker on the command line takes `sudo` until sudoless Docker is enabled. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Gb7x6poap4hGCndPx5qt5T --------- Co-authored-by: David Heinemeier Hansson <david@hey.com> Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com> Co-authored-by: Codex XHigh <codex@openai.com> |
||
|
David Heinemeier HanssonandClaude Fable 5
|
de854d3f0c |
Free the Copy URL shortcut from ghost extension registrations (#6821)
* Rebind ghost Copy URL shortcut registrations to the pinned id Chromium never hands a suggested shortcut to one extension while another — even a long-gone one — still holds the registration. Profiles that first loaded Copy URL before its id was pinned registered Alt+Shift+L under an id derived from the extension's load path at the time, so the pinned extension never receives the shortcut and the keypress does nothing (#6816). The quattro upgrade tried to repair this against one hardcoded path-derived id, which only ever matched a single home directory. The historical ids are unknowable in general — they hash long-gone absolute paths through whatever symlinks existed then — but the registration itself names the command, so a migration now rebinds any copy-url command that points away from the pinned id, unless that id belongs to an extension that is actually installed or the pinned extension already holds a binding of its own. Browsers rewrite Preferences on exit, which reverts any repair made while one runs, so the migration asks for this user's browser windows to be closed first — failing and staying pending when there is no terminal to ask in or the prompt is declined. The backup a repair leaves behind marks it as attempted but unverified: until a browser-free run confirms the registration stayed repaired, the migration keeps itself pending rather than trusting a disk state an open browser may still overwrite. The upgrade-time repair is dropped: the upgrade already runs migrations, so the migration is the single implementation. Fixes #6816 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Pin the WhatsApp Slim extension id Keyless unpacked extensions get path-derived ids, which go stale if the load path or packaging ever changes — the same class of bug that broke the Copy URL shortcut for pre-package installs. Pin the id with a manifest key like the other bundled extensions. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Opus 5
|
b99fd91cf1 |
Simplify the Quattro upgrade, and stop it leaving an error bar behind (#6716)
* Accept either name for the lock authentication command
omarchy-setup-lock was renamed to omarchy-apply-lock in
|
||
|
David Heinemeier HanssonandClaude Opus 5
|
536fcd5c6c |
Move install-time plumbing out of the setup namespace
setup is where a user goes to configure something: direct boot, security keys, hibernation. These three are not that. omarchy-apply-system is the ISO's entry point in the target chroot, omarchy-apply-hardware is what it calls for device quirks, and omarchy-apply-lock is called by install/config/lockscreen-pam.sh. apply is the verb they already used to describe themselves, and it carries the contract: declared state under install/ converged onto the machine, idempotent, safe to repeat. The group gets no GROUP_DESCRIPTIONS entry on purpose. That table drives the top-level group list on its own, so an entry would put apply back in front of users even with every command in it hidden, the way provision already stays out. A test covers it. The ISO installs the runtime from the mirror it ships with, so it moves to the new names in lockstep and no compatibility route is needed. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Opus 5
|
318bf2c43a |
Keep the Quattro upgrade from aborting silently into an unsafe state (#6617)
The script is fetched from the branch but calls into the installed /usr/share/omarchy tree, which can lag it. A packaged build without bin/omarchy-done aborted apply_user_transition under set -e two thirds of the way through: NetworkManager was already enabled, iwd was not yet disabled, and nothing was printed, so the run read as finished. The completion markers are now written directly instead of through omarchy-done, and the two remaining unguarded packaged commands warn rather than abort. Retiring iwd moves up next to the NetworkManager enable it depends on, so no failure in between can leave both enabled. An aborted run now says so instead of returning to the prompt on a green progress line. Fixes #6575 Co-authored-by: Claude Opus 5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Opus 5
|
f76d058a6d |
Force a database refresh before installing keyrings during the Quattro upgrade (#6615)
The upgrade repoints the mirrorlist and the [omarchy] server, then ran pacman -Sy. A plain -Sy keeps the legacy database whenever the new server's copy isn't newer, so the checksums stay stale and every re-download of a rebuilt package aborts as corrupted. Fixes #6576 Co-authored-by: Claude Opus 5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Fable 5
|
b85ae70ebd |
Stop pipefail from turning grep -q SIGPIPE exits into false negatives (#6614)
* Stop pipefail from turning grep -q SIGPIPE exits into false negatives grep -q exits at the first match, and when the producer is still writing it dies with SIGPIPE. Under pipefail that 141 becomes the pipeline's status, so hardware checks like lspci | grep -q read as "not found" on exactly the machines they target. The T2 defaults migration hit this and silently skipped real T2 Macs (#6608). Redirect grep to /dev/null instead of -q wherever a pipeline feeds grep in a pipefail context, so grep reads all input and the producer never gets killed. The install-time T2 checks aren't run under pipefail today but are switched too, since they're the same detection line the issue calls out. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Re-run the T2 defaults migration its broken hardware check skipped The SIGPIPE bug marked 1785944594 as applied without doing anything on affected T2 Macs. The original migration is idempotent, so a fresh migration can just source it now that the guard is fixed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Address Copilot review: fix OCR grep pipeline and prove the T2 repair screen_contains piped tesseract into grep -Fqi under the acceptance suite's pipefail, the same SIGPIPE false negative the rest of the branch fixes. The T2 test's lspci stub now keeps writing past the pipe buffer after the match so every scenario exercises the SIGPIPE case, and a new case runs the rerun migration against fixtures a bitten install would have. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
|
3bfea9b840 |
Fail loudly when a pre-update snapshot isn't actually created (#6580)
* Fail the snapshot when Snapper is installed but has no configs omarchy-snapshot create loops over the configs snapper reports. With none, the loop body never runs, so it prints "Create system snapshot" and exits 0 without capturing anything. Every update then reports a snapshot it never took, and the absence only surfaces when a rollback is needed and the snapshot list turns out to be empty. * Say so when the update proceeds without a snapshot The update ignores exit 127 so a system without snapper updates quietly. Any other snapshot failure was being swallowed by the same expression, which let the update continue with no indication that it was now unprotected. Keep continuing, but say it out loud. * Point the snapshot repair hint at how the installer runs it Also hold the green header until a snapshot will actually be attempted, so the no-config failure doesn't open with a success banner. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Continue the quattro upgrade when the pre-upgrade snapshot fails The upgrade runs under set -e, so the new non-zero exit from an unconfigured Snapper would have aborted a re-run at the snapshot step instead of proceeding like omarchy-update does. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: David Heinemeier Hansson <david@hey.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
|
9d5c6e25e7 |
Keep root= in the kernel cmdline when upgrading to Quattro (#6579)
* Keep root= in the kernel cmdline when upgrading to Quattro The packaged drop-in /etc/limine-entry-tool.d/omarchy-defaults.conf sets KERNEL_CMDLINE[default] with the += operator. limine-entry-tool.conf documents what that costs: "+= appends parameters to an existing cmdline ... Ignores /etc/kernel/cmdline and /proc/cmdline". As soon as the drop-in lands, the tool stops auto-detecting the cmdline. Fresh installs are unaffected, because the ISO writes /etc/default/limine from default/limine/default.conf with @@CMDLINE@@ substituted. The upgrade path never created that file. A pre-quattro install that relied on the auto-detected root= therefore ends up with a cmdline that has no root= at all, in both limine.conf entries and in the cmdline embedded in the UKIs. The next boot fails with "ERROR: Failed to mount '' on real root" and drops to an emergency shell, where the error gives no hint that the cmdline is the cause. Capture the boot-critical parameters from /proc/cmdline before the reboot, while the still-correct cmdline of the running kernel is readable, and write them to /etc/default/limine, which is loaded last so += keeps the drop-in parameters instead of replacing them. Copy root=, rootflags, rootfstype, resume, resume_offset, the cryptdevice and rd.luks keys and rw/ro verbatim rather than reconstructing them, so LUKS and hibernation setups survive too. Re-running the upgrade on an already-broken system has no root= left to copy, so fall back to deriving it from the mounted root, including the subvolume on btrfs. Any config layer that already pins root= is treated as authoritative and left untouched. * Anchor the cmdline guard and harden the repair path The early-return guard searched for the bare string root=, which matches the commented example limine-entry-tool.conf ships at line 53: #KERNEL_CMDLINE[default]+=rw root=UUID=... That file is present on every stock machine, so the guard always fired and the function never wrote anything. Match assignments instead, and check /etc/kernel/cmdline separately since it holds bare parameters rather than shell assignments. Three fixes on the repair path: Assigning boot_params discarded every parameter the collection loop had just captured, so cryptdevice, cryptkey, resume and ro were dropped, and rw was forced over a captured ro. Prepend the derived root= instead, and only add rw when the booted cmdline stated no mount mode. On an encrypted root, findmnt reports the unlocked mapper device, whose UUID says nothing about which container to unlock. Emitting it produced a cmdline that still could not boot while satisfying the final check, so the user rebooted into the same emergency shell believing it was repaired. Warn and write nothing in that case. The allowlist gained rd.luks.key, rd.luks.crypttab, rd.md.uuid, rd.dm.uuid, rootwait, rootdelay and dm-mod.create. Verification now also reads the .cmdline section of each UKI. With omarchy-uki.conf among the drop-ins that embedded copy is what actually boots, so a green limine.conf alone did not prove the machine would come up. * Filter guard paths and narrow the dm-crypt and UKI checks The guard passed /etc/default/limine to grep unconditionally, and that file is absent on exactly the machines this targets. A missing operand makes grep exit 2 without -q, so a drop-in pinning a real root= went undetected and the function appended a second one, overriding the explicit setup it promises to leave alone. Rather than relying on -q returning 0 despite the error, which is a GNU grep special case and not true of every implementation, filter the paths first and only grep the ones that exist. The exit status is then unambiguous. The dm-crypt check gated on the /dev/mapper/* prefix, which also matches plain LVM, dm-raid and multipath. Those roots need no unlock parameters and were repairable before, so the prefix test denied them a working root=UUID= and told them they were encrypted. Gate on the device mapper target type instead. root_filesystem_encrypted() is not reused here on purpose: it treats every /dev/mapper/* path and any non-empty /etc/crypttab as an encrypted root, which suits its own call site but would reintroduce the same false positive. UKI verification now runs through as_root, since a restrictive ESP fmask would otherwise make find return nothing and the check pass in silence, and is scoped to the omarchy_linux*.efi images limine-entry-tool generates so a shared ESP or a stub without a .cmdline section cannot raise a false "do not reboot" warning. The allowlist gained rd.lvm.lv and rd.lvm.vg. * Strip the subvolume before resolving the root device type findmnt appends the subvolume for btrfs mounts, so the source read back for an encrypted btrfs root is /dev/mapper/cryptroot[/@]. lsblk cannot resolve that path, the device type came back empty, and the crypt gate never fired. The function then wrote root=UUID= with no unlock parameters, limine.conf ended up carrying a root= so the final check stayed quiet, and the machine still booted to an emergency shell. That is the layout Omarchy installs when encryption is picked, so the gate missed exactly the roots it exists for. The previous /dev/mapper/* prefix test matched the bracketed form by accident. Moving to the device mapper target type is still the right call, it just needs the unbracketed source, which findmnt --nofsroot provides. Also give root_type an explicit empty default. It is assigned inside a branch and read outside it, and set -u treats a declared-but-unassigned local as unbound, so a findmnt that cannot answer would abort the upgrade with the quattro packages already installed and everything from configure_snapper_policy onward skipped. * Look for the crypt layer across the whole device stack lsblk -no TYPE reports only the target's own type. On the standard full-disk encryption layout, LUKS container -> LVM PV -> root LV, that type is lvm and the crypt layer sits in the parents, so the gate never fired: the function wrote root=UUID= with no unlock parameters, the final check found a root= and stayed quiet, and the machine still booted to an emergency shell. Walk the parents with lsblk -s and look for a crypt layer anywhere in the chain. That keeps LVM, dm-raid and multipath roots on the repair path, since they carry no crypt layer and root=UUID= is enough once mkinitcpio assembles them. root_type is replaced by root_stacks_crypt, which says what is actually being tested and drops the LVM-versus-crypt caveat the old target-type check needed. Also drop a vacuous test assertion: piping a bracketed literal through grep -qv '\[' selects nothing, so the branch was unreachable and the case passed whatever the script did. The --nofsroot assertion above it is what holds that fix. * Keep the crypt gate off a pipeline exit status Capture the device stack and match it from a here-string rather than piping lsblk into grep -q. Under pipefail a short-circuiting grep can leave the producer with SIGPIPE and turn the pipeline into 141, which reads as "no crypt layer" and disarms the gate silently. lsblk writes its whole table in one go, so this is out of reach in practice, but nothing about the gate should depend on how much output a helper happens to buffer. Also correct a stale test comment that described the target-type check the previous revision used, four lines above the comment explaining why that check was insufficient. * Harden the kernel cmdline preservation against false root= pins The /etc/kernel/cmdline early return trusted a file limine-entry-tool ignores once a += drop-in sets KERNEL_CMDLINE[default], leaving exactly the targeted machines unbootable. The pin guard now reads only the *.conf layers the tool loads, only the default key, and tokenizes the assignment value so quoted decoys and volatile-root= cannot pin. /proc/cmdline is tokenized quote-aware so dm-mod.create="..." survives verbatim, the root= verification is token-anchored, and an unverified cmdline now blocks the reboot instead of only warning. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Ask limine-entry-tool for the effective cmdline instead of parsing its configs --get-cmdline default answers whether root= survives the tool's own config merge, replacing the glob, grep and quote-aware tokenizer walk over the config layers, and the quote-aware /proc/cmdline parsing reverts to plain word splitting. The verification and the reboot gate stay: they are what catches anything the simpler paths miss. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: David Heinemeier Hansson <david@hey.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Fable 5
|
6ee243cc37 |
Split the end-user omarchy skill into topic guides (#6602)
* Split the end-user omarchy skill into topic guides Move default/omarchy-skill to default/agents/skills/omarchy and break the monolithic SKILL.md into on-demand topic files for Hyprland config, shell plugins, theming, and hooks. Update the skill symlink wiring, relink existing installs through a migration, and correct claims that had drifted from the implementation: plugin hot-reload, terminal reload, menu customization, refresh scopes, theme overlays, background locations, hook timing, and the packaged (not git-managed) system directory. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Add capture and contributing guides to the omarchy skill Cover screenshots, screen recording, OCR text capture, and LocalSend or Taildrop sharing, plus how to route bug reports, suggestions, and support questions upstream with diagnostics and captures of the problem attached. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Scope Hyprland reload guidance to the Lua config files hyprsunset.conf and xdph.conf are read by separate processes, so hyprctl neither applies nor validates them. Document restarting hyprsunset after editing its config, including in the night light example. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
David Heinemeier Hansson
|
2cff67fa45 |
Fix sleep lock session environment race (#6493)
Found by @tobi. |
||
|
David Heinemeier HanssonandClaude Opus 5
|
2e04f093c1 |
Ship zram tuning as a vendor drop-in
pacman checks file conflicts during transaction prepare, so the --overwrite entry added alongside /etc/systemd/zram-generator.conf never got to run: the copy of omarchy-update-system-pkgs driving the upgrade is the one already on disk. Every machine carrying archinstall's copy aborted the transaction. zram-generator.conf(5) reserves /etc for the local admin and has vendors ship snippets under /usr/lib/systemd/zram-generator.conf.d/, where drop-ins outrank the main config file. Nothing collides, and the tuning wins whether or not archinstall's file is still around. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LjeyQZsNBxqyYy9z8KaKm7 |
||
|
David Heinemeier HanssonandClaude Opus 5
|
5884397e46 |
Drop the bar.json nobody reads
It was the user override file for the first Quickshell bar, back when the bar layered it over bar-defaults.json. Unifying everything into shell.json left an empty file that installs and the quattro upgrade kept faithfully copying around. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Opus 5
|
4dcbed7728 |
Tune swap on zram instead of leaving it at kernel defaults
Omarchy installed zram-generator but never shipped a config for it, so the only zram tuning any machine had was whatever the ISO wrote once, and the VM sysctls were untouched kernel defaults aimed at spinning disks. The generator's own default caps the device at 4G, which on a large machine sends reclaim to the hibernation swapfile far earlier than it needs to go. Ship min(ram / 2, 8192) instead, matching what Fedora settles on, and state the priority explicitly since it has to sit above the pri=0 that omarchy-hibernation-setup gives the disk swapfile. The four sysctls follow from swap being RAM rather than a disk. swappiness above 100 says evicting an anonymous page beats dropping a page-cache page that would have to be read back, which is true once swap is compressed memory. page-cluster drops to one page per fault because the default of 8 buys readahead for a seek zram doesn't have and pays a decompression for each page. Zeroing watermark_boost_factor stops fragmentation from producing reclaim bursts while memory is still free, and raising watermark_scale_factor gives kswapd room to reclaim in the background instead of letting allocations stall in direct reclaim. The last two are what actually addresses the stutter people notice and misread as swap being used too eagerly: proactive swapping to zram is the cure, and synchronous direct reclaim is the disease. Every ISO-installed machine already has an unowned zram-generator.conf, which pacman would refuse to overwrite, so both upgrade paths need it listed. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
Kyunghyun Park
|
e5924ef7d1 | Fix XDG browser defaults when BROWSER is set | ||
|
David Heinemeier HanssonandClaude Fable 5
|
ff1d1c325e |
Suppress Hyprland reload errors for the whole quattro upgrade
Suppression was enabled only after cleanup_legacy_user_paths had already moved ~/.local/share/omarchy aside, so the live session could auto-reload during the swap window and pop "source= globbing error: found no match" for every legacy hyprland.conf source line. Enable suppression as the first step of the upgrade sequence instead; the EXIT trap still restores it if the upgrade dies early. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Fable 5
|
f96827c909 |
Simplify the quattro upgrade script
- Extract run_as_user_omarchy and run_as_user_wayland_session helpers to replace ten hand-rolled run_as_user env blocks - Declare the retired user-unit list once instead of three times - Use git ls-files --others to find user-added theme backgrounds instead of diffing against the tracked list by hand - Stop rewriting the SDDM 10-theme/10-wayland confs the omarchy package already installs, and drop the duplicate [Theme] section from 99-omarchy-login.conf - Drop the legacy-path rewrite in the theme hyprland.conf shim; the user transition already rewrote those references - Remove dead code: the unreachable channel re-validation, the declare -F guard in cleanup_on_exit, and the unset block at the end of the user-transition script Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Fable 5
|
88d60e1468 |
Point gum prompts at the terminal in the quattro upgrade
The user-transition script is fed to bash on stdin, so the browser-close gum confirm read the script itself: it auto-declined and swallowed the remaining script text, leaving bash to fail on a mid-construct syntax error. The outer confirm and reboot prompts had the same exposure when the script runs via curl | bash. Read all three from /dev/tty. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Fable 5
|
4f1b1fceb8 |
Pass the --yes flag into the quattro user transition
The Copy URL shortcut repair checked the outer script's yes variable, which never reached the user-transition heredoc's bash process, so the upgrade died with "yes: unbound variable" under set -u whenever a browser was running with the old extension id in its Preferences. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
David Heinemeier HanssonandClaude Fable 5
|
17c8af3f14 |
Simplify shell restart around quickshell instance management
omarchy-restart-shell now kills by config dir via quickshell kill and relaunches with --no-duplicate, polling kill until every instance is gone since kill returns without waiting and -n silently exits if one remains. The locked-session refusal reads Hyprland monitor state alone, which covers a hung shell holding the lock; the redundant shell IPC probe, availability guards, and omarchy-shell path indirection are gone. Shell IPC calls now time out (2s default, OMARCHY_SHELL_IPC_TIMEOUT to override) so probing an unresponsive shell fails fast instead of hanging the caller. omarchy-hyprland-launch checks dispatch output for "ok" rather than the exit code, which is 0 even on Lua errors, so a failed dispatch falls through to the bash -lc fallback instead of silently launching nothing. The Quattro upgrade cutover delegates to omarchy-restart-shell instead of carrying its own copy of the lock check, pkill, and readiness loop. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
David Heinemeier Hansson
|
73efa7d926 | Remove stale nofile config during Quattro upgrade | ||
|
David Heinemeier Hansson
|
1620b3467b | Refresh applications during Quattro upgrade | ||
|
|
1024bf5e7a |
Backfill pre-quattro hardware support and tmux fixes for existing installs
Existing quattro installs get sof-firmware, the Mesa Vulkan drivers,
the tmux terminal-features and pane-binding fixes, and the XPS 13 text
scaling via migration, while Omarchy 4 upgraders get the same hardware
support and scaling through omarchy-upgrade-to-quattro.
The Vulkan driver backfill adapts
|
||
|
David Heinemeier Hansson
|
b39313abcf | Simpler finish | ||
|
David Heinemeier Hansson
|
d0cc4b9674 | More explicit clear | ||
|
David Heinemeier Hansson
|
79e408b323 | Clear before the logo | ||
|
David Heinemeier Hansson
|
4db7a03235 | Let's try a logo | ||
|
David Heinemeier Hansson
|
77353e9670 | Slim down celebration text | ||
|
David Heinemeier Hansson
|
0b7c440009 | Use gum to prompt for consent | ||
|
David Heinemeier Hansson
|
28d853402e |
Make the Copy URL extension reliable across install and upgrade
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.
|