Omarchy enables omarchy-usb-authorization.service and
omarchy-thunderbolt-authorization.service from first-run, and the
Thunderbolt one from its setup command and a migration. The settings
packages left both only in /usr/share/omarchy/default/systemd/user, where
systemd does not look.
On edge the Thunderbolt migration (1790344599) therefore fails with "Unit
omarchy-thunderbolt-authorization.service does not exist" and every later
migration stays pending. First-run fails as well and enables none of its
units.
Install both in /usr/lib/systemd/user, wherever the source has them: the
4.0.4 release does not, so its package is unchanged.
Buzz (https://github.com/block/buzz, Apache-2.0) is a Tauri desktop app
on a Nostr relay. Upstream ships Linux builds for x86_64 only, so this
builds the desktop-v0.5.28 tag from source for x86_64 and aarch64,
following upstream's release-linux job: the six sidecar binaries from
the root workspace, then the app with --features mesh-llm. The updater
stays compiled out (its key and endpoint are unset), so pacman owns
updates. On the fast ring; the watch follows desktop-v* releases.
On aarch64, -march gains +crypto so sha2-asm can assemble its SHA
instructions; sha2 still detects them at runtime.
Built natively for aarch64 with bin/build and launched with isolated
HOME/XDG dirs: the window opens and renders the bundled frontend.
MediaPipe 1.1.0 links its SQLite memory and vector stores from prebuilt
archives built with Google's internal toolchain. They have no archive symbol
table, and once indexed they still need Google's libc++ (std::__u) and
internal symbols (__alloc_token_*__Znwm, sqlite_helpers, sqlite3_vector_init),
so the link of libmediapipe.so fails. Drop the semantic retriever's C API from
the library; mediapipe and its other tasks import and work as before.
sync-upstream.yml, sync-rebuilds.yml and track-branches.yml each opened one
batch PR, so one package with a failing build held back every other update
in it (#788, #837). All three now group their changes with
package-pr-groups.sh and hand them to package-prs.yml, which opens a PR per
package and arms auto-merge. Packages pinned from the same upstream branch
still share one. Green packages land on their own and only the failures stay
open; a red PR conflicting with master, and each lane's old batch PR, get
closed.
sync-upstream.yml loses its human-merged lane: it now pushes with the PAT
like the other two, so its builds start without an approval hold and its PRs
auto-merge when green. min_release_age still quarantines fresh releases. Its
approve job, approve-sync-push.cjs, and auto-merge-pr.cjs's exception for
auto/sync-upstream go with it.
The branch tracker stalled omarchy-dev and omarchy-settings-dev twice in a
week:
- All ~50 auto_merge packages rode one PR, so omasnap-git's failing smoke
suite kept the quattro pins unmerged from Oct 4 to Oct 8 (#788). Each
package now gets its own PR on auto/track-branches-<package>; packages
pinned from the same branch share one, since they move in lockstep. A run
closes tracker PRs that master has moved past, and the old single PR.
- A branch tip younger than min_release_age held the whole branch, so a
run that landed shortly after a quattro commit pinned nothing. The watch
now falls back along first-parent history to the newest commit that has
cleared the window, measured from clone time so siblings agree.
- The every-2-hours schedule fired every 6 to 9 hours in practice; run it
hourly.
The self-tests, the build-isolation fixture and the three tracker
workflows pulled archlinux:base-devel from Docker Hub without logging
in. On 2026-10-09 that failed two jobs on one pull request: one with
'toomanyrequests', one with a 504 from Docker Hub's token endpoint.
Arch publishes the same image as ghcr.io/archlinux/archlinux, which
GitHub's runners pull without that limit.
The tracker opens its PR and then labels it, so a new tracker PR starts
two build runs at once. The second cancelled the first, whose always-run
`result` job then failed and left a red required check on the PR (#901)
until the surviving run finished. The tracker also force-pushes every two
hours, which would cancel long builds the same way the sync branches did.
Treat auto/track-branches like auto/sync-*: the newer run waits and
reuses the finished build's artifacts.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Auto-merge upstream updates for trusted packages
The unattended lane already runs every upstream provider, not only branch
watches, so trusted release feeds can ship without waiting on the reviewed
sync PR. Opt in the browsers, large vendors' apps, Omacom's own projects,
and the vendor binaries that hold fresh releases for 24h: 39 packages,
including rawmakase-bin and gliff.
Drivers, firmware, the bootloader, core desktop components, beta and
nightly channels, and community projects stay on the reviewed lane.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Auto-merge grok-bot, perplexity, disktree-bin, cliamp, openclaw, voxtype-bin and ttfx
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
power-profiles-daemon only drives intel_pstate, amd_pstate and ACPI
platform_profile, so ARM laptops on cppc_cpufreq (NVIDIA N1X, Snapdragon X)
get placeholder profiles that change nothing and no performance profile.
This packages upstream 0.30 with a CPPC driver patch: balanced (EPP 64) and
power-saver (EPP 192) use CPPC autonomous frequency selection, performance
runs the performance governor. The driver only claims policies with the
schedutil and performance governors and writable autonomous selection knobs,
restores probe-time settings if the first activation fails, and reapplies the
profile on resume. The package provides and conflicts with
power-profiles-daemon, and fixes the powerprofilesctl shebang at package time
so Omarchy no longer needs to sed it.
Tested on an NVIDIA N1X laptop: median keystroke-to-frame latency in Chromium
went from ~130ms (schedutil) to ~56ms on balanced at the same idle power, and
~32ms on performance, which Omarchy selects on AC. All 137 upstream tests
pass, including five new CPPC tests.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Add ghost: an owner-local AI persona with an Omarchy-native HUD
Ghost (https://github.com/ferdousbhai/ghost, Apache-2.0) is a local AI
persona for Omarchy: a daemon on 127.0.0.1, a Quickshell HUD summoned
with Super+Ctrl+G, a `ghost` CLI, an opt-in Chromium relay, and a
computer-use helper. The package builds from the v0.1.0 GitHub release's
sanitized source archive plus its prebuilt Bun runtime archive; the
`.omarchy/package.json` upstream block follows later releases the same
way. check() runs the project's own test suite against the payload.
Built locally with `makepkg --cleanbuild` from these exact inputs on
Omarchy 4.0.0.alpha: ghost-0.1.0-1-x86_64.pkg.tar.zst.
* ghost 0.1.1: arch=('any'), one runtime archive for every architecture
The payload is Bun-target JavaScript with no native modules; verified
on aarch64 under emulation (ferdousbhai/ghost#52). The runtime archive
is now named linux-any and listed with the source tarball under
sources.any.
* ghost: 0.1.2
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017TQbyws7icrhj7H7V7yi6v
* ghost: 0.2.0
* ghost: 0.2.1
* ghost: 0.2.2
* ghost: 0.2.3
* ghost: 0.3.0
* ghost: 0.4.0 with independently installable runtime
* ghost: 0.4.1
* ghost: 0.4.2
* ghost: 0.4.3
* ghost: 0.5.0
* ghost: 0.5.1
* ghost: 0.5.2
* ghost: 0.6.0
* ghost: 0.6.0-2, declare ripgrep for package checks
check() in the 0.6.0 source runs packaging tests that need rg, and the
clean build container does not have it, so every build since 0.5.0 stopped
with "ripgrep is required by package checks but is not installed".
Same version, so pkgrel goes to 2 (ghost e97eb09 carries the same
checkdepend for the next release's render).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Recreate pacman's local signing key before upgrading packages so keyring install scripts can sign trusted keys. Keep the key out of the builder image and exercise keyring population with a keyless test image.
* Add linux-aarch64-pkgbase-shim
Arch Linux ARM's kernel packages install the kernel as /boot/Image and ship
no usr/lib/modules/<ver>/{pkgbase,vmlinuz}, which is how mkinitcpio and
limine-entry-tool find kernels. One pacman hook, ordered 85- so it runs
before 90-mkinitcpio-install, writes both files into every package-owned
modules directory that lacks pkgbase; the usual hook then builds the UKI and
writes the Limine entry. It also covers kernels already present when the
shim itself is installed, cleans up the leftover directories of removed
kernels, and never touches a directory that has pkgbase, so it retires by
removal once archlinuxarm/PKGBUILDs#2215 (or its successor) lands.
Added to the aarch64 workflow's ISO package set.
* linux-aarch64-pkgbase-shim: leave the rebuild to limine-update when the 90 hook is masked
omarchy-iso's installer points 90-mkinitcpio-install.hook at /dev/null in
the target during the chroot phase and runs limine-update itself afterwards.
When the shim is installed in that phase it writes pkgbase and vmlinuz and
would otherwise start a UKI build that finalize_limine_boot throws away.
* linux-aarch64-pkgbase-shim: pkgrel 2
The aarch64 repo build only rebuilds packages whose version moved, so the
previous script fix never reached the repository.
* Trim aarch64 kernel shim comments
* Refresh shim kernel images safely on reinstall
* Run package regression suites in existing CI
* Cover shim ownership and failed-copy handling
* Handle shared kernel-directory ownership in the ARM boot shim
* Limit the CI change to the package test scripts
* Warn when a kernel directory's pkgbase names another package
---------
Co-authored-by: Jimmy Van Veen <jimmy.van.veen@gmail.com>
Co-authored-by: Jim Martin <jimmartin@gmail.com>
* Add qcom-firmware-extract
Copy the vendor-signed Qualcomm firmware named by a Snapdragon laptop device tree from its Windows driver store into /usr/lib/firmware/updates.
* Select matching Qualcomm firmware variants
* Trim firmware extractor comments
* Keep firmware selection and initramfs updates safe on partial installs
* Run package regression suites in existing CI
* Match kernel firmware compression and search order
* Limit the CI change to the package test scripts
* Match firmware names exactly when updating the manifest
* Count compressed firmware only when the kernel can load it
* Scope the README's firmware claim to this package
* Find DSP device-tree images under their Windows names
The Surface Pro 11 device tree asks for adsp_dtb.mbn and cdsp_dtb.mbn,
but Windows ships them as adsp_dtbs.elf and cdsp_dtbs.elf, so
extraction left both DSP device-tree images missing. Look for the
Windows name when the device-tree name is not found. Exact names still
take precedence, and the existing companion matching still selects
between variants.
Validated with test.sh and on a Surface Pro 11 against its own Windows
driver store: 5 of 5 firmware files staged, up from 3, and the ADSP and
CDSP boot with the device-tree images taken from the .elf files.
* Let later firmware sources fill in what a stage lacks
Any stage manifest, even an empty one, made --stage exit as already
staged and made --install use only the stage, so Windows was never read
again. A --stage rerun now keeps the staged files and looks for the
rest, and --install scans Windows only for what the stage lacks. -d
still replaces both. A rerun without root, which cannot read Windows,
keeps the stage and succeeds as before. Each run reads the stage
manifest once.
Mount points honour QCOM_FW_ROOT so the tests can stub the partition
scan. Both root checks now take the user ID from QCOM_FW_TEST_EUID when
the tests set it, and the install-mode check no longer skips when
QCOM_FW_ROOT is set.
* Let board packages list firmware that must not come from Windows
Some device trees name firmware that should not be copied from Windows:
files a board package produces or holds back until they are tested, and
files the machine's boot firmware loads before Linux, which Windows
drivers do not carry.
On the Surface Laptop 8 these are its USB4 router image and SOCCP pair.
A board package can now list such names in
/usr/share/qcom-firmware-extract/provided.d/<compatible>.list. On a
machine with that compatible they are never searched for, installed or
reported by --list-missing, and the log names the list for each one. An
unreadable list or an entry containing a space is skipped with a
warning.
A copy an earlier --install made of a newly listed name stays in place:
the same bytes would not show that the file is still this tool's copy,
since a board's own tool may keep the same file there. The next
--install stops recording it and prints where it is, for the board
package or the user to remove.
Since --list-missing omits listed names, a board must not list firmware
that a hardware check relies on unless its package installs it before
the check runs.
* Find staged firmware by file name when its device-tree path changed
Since the stage became the first source, --install takes a staged file
only by its exact device-tree path. After a full-disk install the stage
is the only copy, so a device tree that files an image under another
path could no longer be served from it. After Windows, fall back to a
staged file with the same name that the stage manifest lists, but not
when Windows refused differing variants of it as ambiguous.
* Test that listed firmware never sends the extractor to the disks
With every name installed or listed, --list-missing prints nothing,
--stage and --install neither list nor mount partitions, and listed
names are not installed from Windows or the stage. A control run
without the list does scan. Listing the names again leaves the copies
it installed in place but stops recording them, and another missing
file still sends the extractor to Windows for that file alone.
* Read only internal disks from an installed system
--install mounts the Windows partitions whenever the stage lacks a
file, which after a full-disk install can be every later run, and it
mounted every NTFS partition, USB disks included. It now reads only
partitions on disks that lsblk reports as neither removable, hotplug
nor USB. --stage in the live session still reads all of them, internal
disks first, and -d takes a driver store from anywhere.
---------
Co-authored-by: Jimmy Van Veen <jimmy.van.veen@gmail.com>
Co-authored-by: Brandon Williams <turbine.bmw@gmail.com>
Co-authored-by: Jim Martin <jimmartin@gmail.com>
The controller created every builder without a DigitalOcean account key, so
DigitalOcean set a root password, emailed it to the token's owner, and turned
password logins on: one email and one password-reachable box per builder,
146 of them on Oct 8.
DO_SSH_KEYS lists account key IDs attached at creation (Ryan, DHH and Emir
in the template). It needs ssh_key:read on the controller's token; without it
every create is refused with 403, so it stays empty until the token has it.
Builders also turn password SSH off in cloud-init, which holds even while no
account key is attached, and Emir joins the operators whose GitHub keys reach
root.
RAWmakase is a Lightroom-compatible RAW photo editor. Upstream publishes
self-contained Linux tarballs for x86_64 and aarch64 (LibRaw, lcms2,
libjpeg and ONNX Runtime bundled under /usr/lib/rawmakase) with a
SHA256SUMS manifest, so the package re-wraps that usr/ tree and
sync-upstream follows releases declaratively with a 24h quarantine.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Ryan Hughes