On 2026-09-03 registry.npmjs.org returned 503s and then timeouts from its
bulk advisory endpoint for over an hour. `npm audit` exits non-zero both
for a real advisory and for an endpoint it cannot reach, so the Frontend
job failed four times running and, because it is a required check, no
pull request could land at all - including Go-only ones that touch no
JavaScript. Every one of those failures was the outage. In two of the
runs the other audit call in the same job succeeded and reported zero
vulnerabilities.
The audits now run through scripts/npm-audit-retry.sh, which separates
the two cases and does nothing else. A conclusive result is acted on
immediately, and the gate stays exactly as strict as before: any
vulnerability at any severity still fails, and no severity threshold is
introduced. Only an unreachable endpoint is retried, with backoff.
When retries are exhausted the outcome depends on whether the answer is
actually unknown. A change that touches frontend-modern/package.json,
frontend-modern/package-lock.json, or the runner itself fails, because
the dependency graph moved and no result means no answer. A change that
touches none of them warns and continues, because the graph is then
identical to the base commit that already produced a passing answer.
Advisories published later against unchanged dependencies are what
Dependabot security updates are for, not a per-pull-request audit.
Deliberately not done: relaxing the severity threshold. That was my first
instinct, but the evidence does not support it. None of the four failures
was an advisory, the lockfile reports zero vulnerabilities at every
severity, and a threshold would have weakened the gate without fixing
anything. The contract's rule that audit suppression is not a valid
closure stands.
The pinning test now requires the runner's invocation and the dependency
detection wiring, and additionally asserts the runner carries no
--audit-level flag, so the strictness cannot be quietly traded away
later.
Go 1.26.8 supersedes the prior patch release, so every release builder and local toolchain guard must move together to prevent candidate artifacts from retaining an older compiler and runtime.
Contract-Neutral: toolchain-only patch update; no product or runtime contract changed
Change-source: pulse-maintainer
An unchanged control revision cannot repair release-convergence debt after its
bounded retry budget is spent. Continuing to raise from the scheduled
reconciler only creates a recurring controller failure while the original
failed run already preserves the actionable debt signal.
Treat that exhausted current revision as a stable no-op, but let a newer
control revision dispatch again so repaired controls are not stranded behind
stale attempts.
Contract-Neutral: release-control automation only; no runtime or public API contract changed
Change-source: pulse-maintainer
Move x/crypto to v0.56.0 after the upstream SSH connection deadlock advisories, and keep the local dependency-floor check from accepting a downgrade.
Contract-Neutral: dependency-only security floor update; no runtime contract changed
Change-source: pulse-maintainer
Retry only explicit registry failures with bounded one-minute attempts while preserving immediate advisory failures. Defer the aggregate audit verdict so frontend tests and builds still report during npm service incidents.
Credential-containment suppression could otherwise hide failed, cancelled, or incomplete public release jobs and strand recoverable convergence debt without an unattended retry. Require complete job evidence and limit suppression to the paid-runtime failure plus its aggregate verdict.
Keep convergence recoverable after short-lived Actions artifacts expire by preserving the exact digest-bound package already verified from GHCR.
Change-source: pulse-maintainer
A committed release with an unchanged operator-owned containment failure cannot converge through unattended retries. Classify that evidence without weakening the block, and rearm only when the relevant private inputs or public controls change.
Change-source: pulse-maintainer
GitHub reruns preserve the failed workflow SHA, so a repaired convergence control cannot resolve an already committed release. A missed workflow_run event can also leave mutable aliases stranded without another attempt.
Change-source: pulse-maintainer
Integrate the reviewed delivery-trust candidate while preserving commits af0e8f8d39 and 6575ebd928 unchanged in history.
Change-source: pulse-maintainer
Reject job-level permission expansion and unsafe checkout selection in the closed-PR capacity workflow, and report cancellation requests separately from runs that completed during the API race.
Change-source: pulse-maintainer
Cancel queued and running validation workflows for a closed pull request head so obsolete matrices cannot hold the hosted-runner limit and delay required checks. Keep the privileged close hook bound to reviewed default-branch code and cover reopen, branch-reuse, identity, and API-race boundaries.
Change-source: pulse-maintainer
test_staged_root_sourced_doc_with_stale_mirror_fails errored in the
"Script smoke tests & backend build" job on PR #1857 (run 33609703555)
with OSError [Errno 39] Directory not empty: '.git' raised from
TemporaryDirectory cleanup. The test body passed; a background git
process spawned by init/add/commit (auto-gc, fsmonitor, or maintenance)
was still writing under .git when shutil.rmtree ran. The suite passes on
main most of the time and locally, so this is a race, not a logic bug.
Disable gc.auto, core.fsmonitor and maintenance.auto for the throwaway
repos, and point GIT_CONFIG_GLOBAL/GIT_CONFIG_SYSTEM at os.devnull so a
runner's host config cannot re-enable them. Construct both temp
directories with ignore_cleanup_errors=True as a belt-and-braces fallback
so a straggler can never fail a test whose assertions already passed.
Three defects that the v6.4.3-rc.1 cut exposed, each with its contract note
and proof:
1. scripts/release-preflight-worker.sh published the smoke stack on host port
7655 and probed it directly. pulse-dev also hosts the maintainer's dogfood
Pulse instance on 127.0.0.1:7655 and a second instance on 17655, so the
release smoke failed to start after every other stage had passed on the
qualified head. The worker now honours PULSE_RELEASE_PREFLIGHT_E2E_PORT,
otherwise takes the first candidate host port pair it verifies free and
fails fast when none is, probes health and update status on that port, and
hands Playwright the same base URL.
2. .github/workflows/publish-helm-chart.yml resolved the chart version before
checking out the repository, but de41ea1883 made that step call
scripts/write_github_output.py, so every chart publish since failed with
"can't open file". Run 33580123246 hit it after all other staging passed;
the checkout now precedes the resolver.
3. docs/release-control/control_plane.json maps the 6.4.3 version prefix to
release/v6.4. Run 33579042375, dispatched from main, failed inside the
compiler dispatch because main advanced one minute after the pipeline
pinned its source SHA; with pull requests landing every few minutes a
candidate cannot hold its SHA between prepare and compile. release/v6.4
was created from main at the exact-SHA-qualified commit 56e51e622e and
carries these same fixes; v6.4.3-rc.1 published from it.
The deployment-installability contract records the worker port and toolchain
rules and the Helm checkout order; release_preflight_test.py and
build_release_assets_test.go pin them.
Rootful evidence must reject non-running systemd states and prove no unrelated container runtime survives. Require exact manager state, mask distro containerd, recheck readiness after scenarios, and remove the complete Podman socket boundary.
The disposable host reached multi-user.target in a degraded state because distro Podman housekeeping used overlay storage inside the outer container. Mask unused units, force VFS for explicit daemons, and fail closed on systemd unit failures so retained rootful evidence remains trustworthy.
The pre-batch upstream rootful source-closure manifest was created before the local Unix recovery test. Include that compiled installer input so merged qualification evidence remains bound to the complete harness.
Change-source: pulse-maintainer
(cherry picked from commit f5ad4e343e)
A workflow dispatch by branch can resolve after that branch moves, allowing an unreviewed tip to enter the release pipeline. Require every publishing dispatch to name its expected source SHA and make the workflow reject a different source or workflow commit before checkout.
Change-source: pulse-maintainer
(cherry picked from commit a461fc9c0a)
The Unix agent-ID recovery test imported syscall.Mkfifo from a generic test file. Go must compile that file before its runtime skip can run, so Windows CI could no longer build the installer test package. Keep the security regression on supported Unix targets while restoring the Windows delivery signal.
Change-source: pulse-maintainer
(cherry picked from commit c0ca94ee9f)
(cherry picked from commit d2cc6b4884)
Stable promotions built whatever the dispatch branch was at that second.
The resolver checked that HEAD descends from the promoted release
candidate but never that its content matches, so v6.4.0 shipped 64
changed files, including product code, that v6.4.0-rc.12 had not
soaked. Every v6 version was mapped to main, which now moves every few
minutes under the autonomous maintainer, so each fix to a candidate
brought everything landed since and stable was never an exact soaked
commit. Five of six stable minor releases shipped under version-bound
owner exceptions that waived the soak.
From v6.5.0 the release train applies (RELEASE_PROMOTION_POLICY.md,
"Release Train"): a two-week train sized to measured velocity, a
release/v6.N branch per train declared in the control plane so the
workflow refuses a dispatch from anywhere else, a stable promotion that
may differ from its candidate only in release metadata unless
hotfix_exception names active customer harm, and a seven day soak for
minor releases. The 6.4.x line stays on main so the v6.4.3-rc.1
candidate already prepared there is unaffected. The gap is registered
as coverage gap release-train-exact-candidate-promotion.
agent_state_dir_lifecycle_test.go calls syscall.Mkfifo, which does not
exist on Windows, so scripts/installtests has failed to compile in the
Windows leg of Unified Agent Native Verification since 53267e149d and the
install.ps1 contract tests there have not run. Every test in the file
drives install.sh through bash and systemd, so tag the file unix-only,
matching the other lifecycle lab files. GOOS=windows go vet now passes.
go test places t.TempDir under GOTMPDIR, and the preflight worker nested that
under its per-run directory, so unix socket fixtures in cmd/pulse-agent-helper,
internal/agenthelper, and internal/dockeragent exceeded the 108-byte sun_path
limit and failed with "bind: invalid argument". GitHub runners leave GOTMPDIR
unset, and the longest rootless Podman fixture sits exactly at the limit under
/tmp, so the worker now leaves GOTMPDIR unset by default and only honours an
explicit PULSE_RELEASE_PREFLIGHT_GO_TMP_DIR override.
Two cmd/pulse-agent helper tests ran without a state directory, so the agent
consulted the platform default /var/lib/pulse-agent. On a host where that
path exists as a private directory the pending-update handoff lookup fails
with permission denied. Give those tests an isolated state directory.
install.sh refuses connection state, token, and lifecycle files whose
parent directory is group- or world-writable. The fixtures build those
parents with t.TempDir, which inherits the process umask, so on a worker
with umask 002 the state directory came out 775 and seven TestInstallSH
cases failed on the pulse-dev release preflight while passing on GitHub
runners and macOS. Set umask 022 for the package on unix so the tests
prove the installer contract rather than the host's umask.
Recognize bare sequence entries and every valid block scalar header across executable action inputs, run scripts, and step environment boundaries.
Change-source: pulse-maintainer
The exact-SHA preflight streams the worker script over a non-login ssh
shell, so /etc/profile.d/mise.sh is never sourced and a stale system Node
binary on the worker shadows the infra-managed mise toolchain. pulse-dev
carried a manually installed Node 20 symlink in /usr/local/bin, so the
worker failed its Node 24 check while a login shell resolved 24.19.0.
Activate mise shims when mise is installed so the worker uses the same
toolchain the bootstrap provisions, regardless of how it is invoked.
Open the v6.4.3 candidate line from main. The v6.4.2 tag was staged on
2026-08-31 but never activated: its release run was cancelled after the
private Pro build failed the compiler memory gate, so the latest published
stable is still v6.4.1. This candidate carries the complete v6.4.2 change
set plus the corrections landed since that tag, including the stale PBS
Backup Running state (#1815), the Windows Unified Agent auto-update 404
(#1820), and shared-token same-hostname agent identity collapse (#1753).
Packet: VERSION, compose and install-docker defaults, Helm chart metadata,
release notes with a declined visual plan, changelog, pointer docs and
the shipped docs mirror, and the deployment-installability cutoff note.
Rollback target is v6.4.1 and the mobile decision is no-mobile-impact.
Tests: the packet tests now describe the 6.4.3 train, v6.4.2 is recorded
as an unpublished stable so it is never derived as the previous stable or
rollback target, and the Python v6.4.2 notes expectation matches the
phrase the notes actually use.
pulse-triage[bot]