Release 2026.8.31-3 preserves upstream desktop settings and explicit launch environment without invoking the CLI sandbox setup. Keep incomplete runtimes from blocking the packaged fallback, and use the namespace sandbox consistently in both locations.
Co-Authored-By: Codex XHigh <noreply@openai.com>
Use the same direct executable path for menu launches and URLs, require working user namespaces, and retain only the first-update relaunch gate backport. This avoids the upstream CLI fallback that makes a helper in the user runtime setuid-root.
Co-Authored-By: GPT-6 Codex (xhigh) <noreply@openai.com>
Track main while pinning the initial release commit. Ship the matching upstream installer and Linux namespace sandbox backport, and launch the native user build prepared by Omarchy.
Co-Authored-By: GPT-6 Codex (xhigh) <noreply@openai.com>
Revert the native installer and updater packaging introduced by #325. Keep the prebuilt desktop and existing launcher, updating only the release tag, commit and archive checksum from the previous recipe.
0.11.2 is 14 hours old, inside the package's 24h min_release_age, so the
upstream sync would not pick it up until tomorrow; ship it now by hand.
Tarball commit b000b79 verified against the v0.11.2 tag, checksum from
an independent download. The new Cargo dependencies (ashpd, zbus,
gdk4-wayland) are vendored crates needing no new system libraries.
0.11.2 also ships FileManager1/portal integration files. Stage the
FileManager1 service under /usr/share/strata the way upstream's own
PKGBUILDs do — the app copies it per-user on opt-in. The portal files
stay unpackaged, matching upstream: enablement is per-user and
consent-gated in the app.
The package carries a deliberate commit pin (59d557a) so every
architecture packages the same tree. The AUR sync kept trying to revert
that to the unpinned branch-tip recipe, recording a version downgrade in
the process. Mark it local so exactly one source owns the recipe; it is
Omarchy-maintained until it moves to an upstream release feed.
* Bump libfprint-git for Synaptics 06cb:010b support
* Describe libfprint-git as the driver the fingerprint setup installs
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
---------
Co-authored-by: powderluv <powderluv@powderluv.org>
Co-authored-by: David Heinemeier Hansson <david@hey.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Keep build-time registration private, verify the published launcher, and then let the native command register the final desktop entry. Keep that launcher first on the desktop environment PATH for subsequent registrations.
Co-Authored-By: GPT-6 Codex (xhigh) <noreply@openai.com>
Stage the initial clone before publishing its ownership marker and build the desktop through the same native CLI used by updates. Keep a concurrent checkout intact and reset pkgrel for the version change.
Co-Authored-By: Claude Opus 5 (default) <noreply@anthropic.com>
Install Hermes into its writable native layout instead of shipping a frozen /opt desktop. Preserve the native update path, migrate tagged bootstraps, and keep existing CLI launchers until the desktop is ready.
Co-Authored-By: GPT-6 Codex (xhigh) <noreply@openai.com>
makepkg always exports CARCH, so PKGBUILDs may branch on it at file
scope. Every place the tooling sourced a PKGBUILD did so without CARCH,
taking the wrong branch or aborting partway, and check-versions turned
the resulting empty pkgver/pkgrel into the version '-', which never
matches a published version and queues an endless rebuild.
package_pkgbuild_var reads one variable the way makepkg would see it,
for the architecture being checked, and reports whether the source
succeeded. check-versions, sync-rebuilds and omarchy-pkgs use it.
check-versions now warns and skips a package whose pkgver or pkgrel is
empty instead of comparing a partial version. A self-test covers a
PKGBUILD that branches on CARCH before assigning its version.
The file-scope case on CARCH ended in 'return 1', which aborted any
'source PKGBUILD' that did not export CARCH before pkgver and pkgrel
were assigned. check-versions reads PKGBUILDs that way, saw an empty
version, queued 8.12.34-35 on every tick, and promotion then refused to
overwrite the already-published artifact. Every channel has been in
backoff since 06:41 UTC today because of it.
Arch Linux ARM's makepkg.conf defaults PKGEXT to .pkg.tar.xz. build/sign.sh
only signs *.pkg.tar.zst, and push-build, sync-rebuilds and the notifier
parse the same suffix, so an aarch64 package built under the stacked
pipeline came out as .xz and would have been skipped at signing. Seen on a
plain x86_64 runner building aarch64 under QEMU (fork run 33642125077).
Same fix as the PKGEXT line in #240.
One list, PUBLISHED_ARCHES in helpers/paths.sh (default x86_64,
overridable with OMARCHY_ARCHES), now drives everything the repository
host schedules. check-versions compares PKGBUILDs against each
architecture's channel databases and writes one queue per channel and
architecture; auto-release works through the queues one architecture at
a time, each with its own backoff, so a failing build on one never
blocks the other; advance-channel --arch all re-runs an advance for every
published architecture and omarchy-release uses it for start and ship,
building the pinned pair once per architecture in its rc trigger; the
train observes channels through the reference (first) architecture
instead of a hard-coded x86_64. Queue and backoff files written under
the old per-channel names are treated as x86_64 until consumed.
Two things made an aarch64 builder image impossible to create: the
keyring bootstrap fetched omarchy-keyring from the target architecture's
own channel tree, which does not exist before that architecture has
published anything, and the QEMU probe only knew the x86_64-host,
aarch64-target case. The keyring (arch=any) now always comes from the
x86_64 tree, and the probe compares host and target architectures and
runs a container for the target platform.
clean-repo grouped versions with a regex that only knew any, x86_64 and
i686, so aarch64 packages would never have been pruned.
Package OpenClaw 2026.9.1 as a local PKGBUILD based on the AUR one,
tracking the npm registry's latest dist-tag through the repository's
declarative npm upstream provider: upstream's release cadence outruns
the AUR maintainer, and the dist-tag is the stable channel where a plain
version-max would ship next cycle's betas. A 24h min_release_age
quarantines fresh releases, which matters more than usual here because
the npm tarball is not vendored: package() resolves ~330 transitive
dependencies from the live registry without integrity pins. The pinned
sha256 was verified against the registry by hand. The initial pin was
taken inside its quarantine window through the documented
BYPASS_MIN_RELEASE_AGE maintainer path, deliberately, and lands through
this reviewed change as that path intends.
The AUR post_upgrade restart attempt is replaced with printed guidance:
it targeted a nonexistent openclaw.service, and the real
openclaw-gateway.service is a systemd user unit a root pacman hook
cannot reach (voxtype-bin sets the precedent).
The builder ships npm 12, which refuses install-time lifecycle scripts unless
the package is allow-listed, and for a local tarball the allow-list key is the
tarball's own file: spec rather than the package name. Without it openclaw's
postinstall never runs, the .openclaw-lifecycle-pending marker ships in the
package, and every invocation dies trying to finish the lifecycle inside the
root-owned /usr/lib/node_modules/openclaw. package() now passes
--allow-scripts and fails the build if the marker survives.
That postinstall also runs upstream's legacy-state migration against whatever
home it sees, so the npm call gets a scratch HOME under $srcdir with the
OPENCLAW_* location overrides unset: a maintainer's own ~/.openclaw is not
the build's to prune.
