The RETURN-trap smoke fixture has a simulated manager and script binary, not a real managed Linux process. Provide its exact-service/new-version comparator and require exactly one success-path invocation, with none on prerelease refusal. Keep actual process-identity refusals in the independent source/native controls; do not relax production verification.
Change-source: pulse-maintainer
Contract-Neutral: Legacy shell smoke fixture correction only; production updater and its running-executable, signing, restoration and deadline contracts are unchanged.
Read bounded, complete unit state before mutation and confirm inactive state before rollback replacement. Keep the verified backup and stop automatic recovery when service state cannot be reconciled; preserve prior-active restoration, masks, signatures and configuration.
Exercise failed/partial reads, deadlines, no-op stops and restoration through confined production-shell fixtures and the existing signed transaction controls. Fix fixture path and manager observations so registered Go controls execute the actual source.
Refs: #2785
Change-source: pulse-maintainer
The complete current-source smoke suite exposed two stale fixtures: release objects omitted the required draft flag, and the proxy safety check assumed an emoji presentation selector absent from the shipped heading. Supply complete positive metadata while retaining draft/chart/preview and missing-maturity refusals; locate exactly one named Markdown section and discriminate missing or duplicate boundaries. Preserve all runtime, authentication, update and workflow gates.
Contract-Neutral: Test fixtures and section extraction only; product runtime, subsystem contracts, authentication boundaries and release-metadata admission are unchanged.
Change-source: pulse-maintainer
The complete signed transaction, race controls, installer suite and vet passed, but the older shell smoke fake still refused stop and consequently could not reach its restore/start assertion. Allow only that successful stop boundary and retain the activation assertions. Production updater and its explicit stop-failure controls are unchanged.
Contract-Neutral: Existing failed-installer fixture correction only; production snapshot, rollback and activation contracts are unchanged.
Change-source: pulse-maintainer
Reject directory destinations before the same-directory file rename, use portable mode inspection in the connected transaction tests, and teach the existing service fakes that stopping and failed activation are separate operations. The recorded first candidate proof remains bound to its own source.
Contract-Neutral: Preserve the reviewed backup and rollback semantics; portable utility invocation and faithful test boundary only.
Change-source: pulse-maintainer
A RETURN trap is not scoped to the function that installs it. perform_update
left its trap referencing function-local installer_tmp/signature_tmp (and
service_name/service_was_active) installed, so it re-ran when a later function
returned. With those locals gone and 'set -u' active, the updater aborted with
'installer_tmp: unbound variable' immediately after logging 'Auto-update check
completed' — pulse-update.service entered failed state even though the update
had succeeded (reported #2128, affects v6.3.2/v6.4.1).
Disarm the trap with 'trap - RETURN' on its first invocation, and use
${installer_tmp:-}/${signature_tmp:-} so a stray invocation cannot abort the
script. Add a regression test asserting no RETURN trap remains after both the
successful and the prerelease-refusal paths.
Change-source: pulse-maintainer
Two bugs from #1630 that combined to take installs down silently:
1. perform_update()'s install-failed rollback branch restored the backup
but never restarted pulse.service. Since the generated
pulse-update.service gates on ExecCondition=systemctl is-active,
every later timer run was then skipped and the install stayed down
until manual intervention. Restart is now guaranteed by a
service_was_active-guarded restart in that branch plus an
ensure_service_restarted RETURN trap so no exit path can miss it
(re-fix of #1323, originally c0b3a0e66, lost in 778a2577b and only
partially restored in 672e81985).
2. install.sh aborted under errexit when writing the /bin/update helper
on a read-only filesystem - after the new binary was installed and
the service stopped, landing in bug 1's no-restart branch. The stock
pulse-update.service uses ProtectSystem=strict, so /bin and
/usr/local/bin are read-only on stock unattended updates; transient
read-only remounts hit the same path. The helper write, PATH
appends, and the /usr/local/bin/pulse symlink are now idempotent and
non-fatal with a warning (install_binary_symlink).
Contract: deployment-installability now pins fail-closed service
availability for unattended updates and non-fatal writes outside the
hardened unit's writable set, with proofs in pulse_auto_update_test.go
and root_install_sh_test.go plus shell regression coverage in
scripts/tests/test-pulse-auto-update.sh (installer-exits-nonzero path)
and scripts/tests/test-install-update-resilience.sh (read-only helper
and symlink paths, verified under set -e, root-safe via ENOTDIR).
Fixes#1630
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Production telemetry (7d, 2026-07-17) shows a persistent tail of live
installs on 6.0.0-rc.x and 6.0.3/6.0.4 while 6.0.5 has been stable since
2026-07-09. Root cause for the reachable stranding class: release
selection trusted GitHub's created_at ordering, and this repo interleaves
v5-line maintenance releases with v6 releases (v5.1.36 was created the
day before v6.0.5). Whenever a v5.1.x release is the most recently
created stable:
- stable channel (getLatestReleaseForChannel) returned the first
non-prerelease in list order -> "no update" for every v6 install
until the next v6 release ships
- the RSS rate-limit fallback returned the first feed entry matching the
channel, same failure, both channels
- the unattended updater trusted /releases/latest, which GitHub defines
as most-recently-created stable -> timer no-ops in the same window
All three now select the highest eligible version: the API path tracks
newest stable + newest prerelease in one pass (prerelease = GitHub flag
OR tag shape), the feed fallback picks the max matching the channel, and
pulse-auto-update.sh scans /releases pairing each tag with its own
draft/prerelease flags before the fail-closed shape filter, keeping
/releases/latest as fallback. helm-chart-* tags fail semver parsing and
are dropped everywhere.
Verified not broken (pinned by new tests): rc-channel installs are
offered the newer stable (6.0.0-rc.6 -> 6.0.5) and move onto the next rc
line when one opens (-> 6.1.0-rc.2); the auto-update prerelease filter
(fail-closed is_prerelease_tag, metadata-flag refusal) still refuses
prerelease targets on stable. Go tests fail on the pre-fix code
(returned 5.1.37 / "no update").
Contract: deployment-installability now pins version-max release
selection with proofs in manager_stranded_upgrade_test.go and
pulse_auto_update_test.go.
Back-port the pulse-auto-update.sh half of v5 fix 0f2982ce3 to v6. After a
verified update, if Pulse was running beforehand, wait_for_service_active
polls up to 20s, tries one explicit start, and if it still does not come
back up restores the backup binary/VERSION, restarts, and returns failure
instead of leaving Pulse stopped (common on unprivileged LXC where the
installer's restart silently fails). Adds a BASH_SOURCE guard so the
script can be sourced, plus scripts/tests/test-pulse-auto-update.sh.
Note: the install.sh start_pulse STOPPED_PULSE_SERVICE guard (interactive
installer path) still needs the stop-tracking infrastructure wired into
v6's update flow; tracked as a follow-up.
pulse-triage[bot]