Skip to content

✨ release: make v6 an official release line (v6.x.y tags) - #17074

Merged
clubanderson merged 5 commits into
v5from
ci/v6-release-line
Nov 7, 2026
Merged

clubanderson merged 5 commits into
v5from
ci/v6-release-line

Conversation

@clubanderson

Copy link
Copy Markdown
Member

Makes v6 an official release line so it gets semver tags (v6.x.y), GitHub Releases, SBOM/NOTICE artefacts and immutable image retags the way v5 does. The owner wants to announce "latest is available in v6" with a real version number.

What changed

File Why
.github/release-lines.yml release_lines: [v4, v5, v6]. v6 stops being an extra on the PR-gating workflows and docker.yml LONG_LIVED; every pinned lane that does not cover v6 now writes down a -v6 exclusion with its reason (fossa, gh-aw-compile, hive-action-smoke-tests, crux-sync, forgejo-smoke, stream-image, v4-freeze-gate, v5-topup, v6-topup).
dco-post-merge.yml, dco-push-delta.yml, docs-link-check.yml, podman-{arm64,rootful,rootless}-lane.yml, podman-contract.yml, quadlet-gate.yml, suid-contract.yml, scorecard.yml v6 added to the branch filters (and the dco-post-merge matrix) — the core release-line gates the guard exists to carry forward.
tagged-release.yml Ported the #9154/#9307 generalisation that only ever landed on v6 (RELEASE_LINE: ${{ github.ref_name }}, concurrency: tagged-release-${{ github.ref_name }}, --is-release-line refusal, trusted-event filter on workflow_run), so the v5 and v6 copies are the same file. Fixed the one v5 literal that port missed (-f branch=v5 in the required-checks GraphQL query). New relay job: GitHub delivers workflow_run/schedule only to the default branch's copy, so when the v5 copy sees docker.yml complete a push/dispatch build on another declared release line, it dispatches tagged-release.yml --ref <line>; that line's own copy then takes the existing backstop path (resolve tip → require published images → compile fragments → derive → release). Event-driven only; the hourly schedule stays a default-branch backstop.
promote-stable.yml Same #9154 port (derive the branch from github.ref_name, refuse non-release-lines). Behaviour on v5 is unchanged: schedule/workflow_run run on the default branch, so stable keeps following v5.
src/scripts/check-release-lines.sh, test-release-lines-guard.sh --is-release-line <branch> query mode + its self-test (from v6, #9307). TRUSTED_SCRIPTS_TREE re-pinned.
src/docs/releases.md, release-line-guard.md, release-channels.md Document the per-line policy, the relay, the channel mapping and the extended cut checklist.
changelog.d/added-v6-release-line.md Fragment.

First v6 version

derive-release-version.sh with RELEASE_LINE=v6 only looks at v6.* tags; none exist, so the first release is forced to v6.0.0 (bump=first), never a bump of v5's number. Verified locally against the v6 tip after compiling its changelog.d/: Derived release version: v6.0.0 (bump=first). The first v6 release consumes the ~2,130 fragments v6 has accumulated through top-ups (its ## Unreleased has never been emptied) — that is a large but correct first changelog; no release: none marker is involved (the #17015 rule only forbids none in fragments and none carry it). One v6-only fragment (changelog.d/fixed-14225.md, missing its - bullet) makes compile-changelog.sh fail on v6 today; the top-up PR that carries this change to v6 fixes it.

Channel mapping (unchanged by this PR — documented)

v5 v6
stable ✅ via promote-stable.yml (default branch only) ✗
candidate, :latest ✅ docker.yml row v5 true candidate ✗
edge, v6-latest ✗ ✅ docker.yml row v6 false edge (already on v6)
vN.x.y immutable tags + GitHub Release + SBOMs ✅ v5.x.y ✅ v6.x.y (new)

Do not dispatch promote-stable.yml from v6: candidate is a single v5-owned tag, and stable stays on v5.

Release-line-guard checklist (src/docs/release-line-guard.md)

  • 1. release_lines in .github/release-lines.yml → [v4, v5, v6]
  • 2. Every pinned workflow either names v6 in its trigger or carries a written -v6 exclusion (guard: PASS locally, self-test: PASS)
  • 3. Every env_lists list covers v6 (docker.yml LONG_LIVED already had it; now as a release line, not an extra)
  • derive-release-version.sh: first tag on the line is vN.0.0 → v6.0.0
  • tagged-release.yml serves github.ref_name, gated by --is-release-line; relay for non-default lines
  • promote-stable.yml gated the same way; stable remains default-branch (v5) only
  • docker.yml channel rows on v6 already say v6 false edge (hive, hive-contributor, hive-hub) — no change
  • v6-topup.yml / v5-topup.yml exclusions updated so the guard agrees on every line
  • Branch protection: v6 is currently unprotected, so the release commit lands through the direct-push path ([ci-maintainer] v5 releases stalled 15h since v5.136.0: hourly tagged-release backstop fires ~every 6–7h (repo-wide schedule starvation) and no run meets sha==tip under 12–23 merges/h (needs human: workflow change) #10863); the PR fallback path is still there if protection is added
  • Docs that enumerate lines updated
  • v6's own copies (release-lines.yml, workflows, scripts) — follow-up top-up PR v5→v6 (also fixes changelog.d/fixed-14225.md)

Validation

src/scripts/check-release-lines.sh PASS, src/scripts/test-release-lines-guard.sh PASS, check-changelog-fragment.sh OK, actionlint clean on the changed workflows (pre-existing shellcheck infos only).

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
@kubestellar-prow kubestellar-prow Bot added the dco-signoff: yes Indicates the PR's author has signed the DCO. label Nov 7, 2026
@kubestellar-prow

Copy link
Copy Markdown
Contributor

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please assign danathar for approval. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Details Needs approval from an approver in each of these files:

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

@kubestellar-prow kubestellar-prow Bot added the size/XL Denotes a PR that changes 500-999 lines, ignoring generated files. label Nov 7, 2026
@hivecommons-hive

Copy link
Copy Markdown
Contributor
f38575c

ℹ️ Sentinel notice — informational; author is trusted (owner), merge is not blocked

Hive flagged this PR (author @clubanderson, head f38575ca321a) because it matches behaviors that can override security controls, escalate privileges or damage the codebase. This is a heuristic, not an accusation — a maintainer should confirm the change is intended before it merges.

  • sensitive_path — changes 13 sensitive path(s) (Touches a sensitive path (OWNERS, workflows, policies, hive config, security docs, …))
    • .github/release-lines.yml
    • .github/workflows/dco-post-merge.yml
    • .github/workflows/dco-push-delta.yml
    • .github/workflows/docs-link-check.yml
    • .github/workflows/podman-arm64-lane.yml
    • .github/workflows/podman-contract.yml
    • .github/workflows/podman-rootful-lane.yml
    • .github/workflows/podman-rootless-lane.yml
    • .github/workflows/promote-stable.yml
    • .github/workflows/quadlet-gate.yml
    • .github/workflows/scorecard.yml
    • .github/workflows/suid-contract.yml
    • .github/workflows/tagged-release.yml
  • permission_escalation — widens workflow permissions or uses pull_request_target (Widens GitHub Actions permissions or switches to pull_request_target)
    • .github/workflows/tagged-release.yml
  • ci_gate_weakening — CI configuration is deleted, softened or has gate checks removed (Disables or softens CI gates (continue-on-error, || true, removed checks))
    • .github/workflows/tagged-release.yml
    • src/scripts/check-release-lines.sh

Hive did not add the sentinel-alert label because this author is trusted. The finding remains recorded for audit and the dashboard Security tab; set sentinel.trusted_authors_block: true to require the blocking label for trusted authors too.

@github-actions

github-actions Bot commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

Open PR overlap check

Base branch: v5

Conflicts are checked with every PR merged onto the current tip of v5, so a PR that only needs a rebase is not reported as conflicting with its siblings.

No conflicting open PRs or same-file overlaps were found.

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
@hivecommons-hive

hivecommons-hive Bot commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

Hive status — updated in place; edits do not notify. Full history is in the Activity Monitor.

@clubanderson this PR is held at stage/071-needs-author and the hive will not merge it yet.

Why:

  • the review swarm returned requires_human: it has a question for you, not a fix for the hive
  • the reviewer asked you: hanthor should confirm that after this lands a workflow_dispatch of promote-stable.yml from v6 is allowed to move the stable channel to a v6 candidate (promote-stable.yml drops the RELEASE_BRANCH: v5 literal for ${{ github.ref_name }}), and sign off on the workflow-permission surface as a second approver.

Next: the hive does not push to your branch. Push a new head (or reply here); CI and the review swarm re-run automatically and the PR returns to the normal path.

Review findings from reviewer (folded here because this head already carries its hive review):

This is a release-policy and workflow-permissions change (12 files under .github/workflows/, a new actions: write job that dispatches workflows cross-branch, and a TRUSTED_SCRIPTS_TREE re-pin), so it needs a CODEOWNER other than the author to sign off (.github/CODEOWNERS: .github/workflows/** clubanderson hanthor Danathar). I'm routing rather than judging it; here is what I verified so the decision is quick.

Verified correct

  • TRUSTED_SCRIPTS_TREE (.github/scripts/integration-governance.cjs:12) = c06705a4… which is exactly git rev-parse HEAD:src/scripts at cddc939; v5's current tree is 6cf759e2…, so the re-pin is required and accurate.
  • src/scripts/check-release-lines.sh --is-release-line (:53-60) behaves as documented: v6 → exit 0, main → exit 1. The full guard passes with 0 FAILs at this head, and test-release-lines-guard.sh reports RESULT: PASS.
  • The relay job (.github/workflows/tagged-release.yml, new relay:) only fires on workflow_run of a successful push/workflow_dispatch build whose head_branch != github.ref_name, rejects anything not matching ^v[0-9]+$, then requires release-line membership from its own checkout before gh workflow run tagged-release.yml --ref "$LINE". GITHUB_TOKEN-created workflow_dispatch events do start runs, so the mechanism works. Both v4 and v6 have tagged-release.yml and docker.yml, so the dispatch target exists on every declared line.
  • v6's own tagged-release.yml already carries the RELEASE_LINE: ${{ github.ref_name }} generalisation (v6 :176, :270), so the relayed run executes the parameterised copy, not a v5-hardcoded one.

Sequencing gap (medium) — needs the v6 top-up to land with this
v6's .github/release-lines.yml:36 currently reads release_lines: [v4, v5]. The relayed run on v6 executes v6's copy, whose guard (v6 tagged-release.yml:270) runs check-release-lines.sh --is-release-line v6 against v6's checkout and exits 1 with ::error. So from the moment this merges until the top-up adds v6 to v6's manifest, every v6 push produces a red Tagged Release run on v6 and mints nothing. The body mentions the top-up (for the fixed-14225.md fragment) but not this dependency — worth landing them together or in quick succession. Not a defect in this diff; a merge-order constraint.

Not verified (out of my budget, flagging for the approver): the -f branch=v5 GraphQL literal fix in tagged-release.yml, and the eleven -v6 exclusions in release-lines.yml each having the intended reason.

Review findings from reviewer-2 (folded here because this head already carries its hive review):

Conclusion: I read the workflow and script changes against the v5 tree and found no defect; this needs a second approver because it changes release/publishing behaviour and 13 workflow files, which is a trust surface I do not sign off on alone. The one genuine policy choice is above.

What I verified

  • .github/scripts/integration-governance.cjs:12 — TRUSTED_SCRIPTS_TREE is re-pinned to c06705a4339105f9cdc51220053c319a2b90dce1, which is exactly git rev-parse pr-head:src/scripts. The merge-base (dc4f20c) and current v5 tip share the same src/scripts tree (6cf759e…, no src/scripts commits in between), so the post-merge tree will match the pin. If anything under src/scripts/ lands on v5 before this merges, the pin must be regenerated.
  • New relay job (tagged-release.yml, relay: block): job-scoped permissions: contents: read / actions: write, persist-credentials: false, manifest sparse-checked-out from the running ref (the default branch), LINE rejected unless it matches ^v[0-9]+$ and check-release-lines.sh --is-release-line before reaching --ref. head_branch on push/workflow_dispatch runs is repository-owned. Looks correct to me.
  • decide now filters workflow_run to event == push || workflow_dispatch and head_branch == github.ref_name; the previously hard-coded -f branch=v5 / --ref v5 sites all read $RELEASE_LINE. The dispatched run on v6 takes the event_name != 'workflow_run' backstop path with RELEASE_LINE=v6.
  • promote-stable.yml: concurrency group and RELEASE_BRANCH become ${{ github.ref_name }} and a first step refuses any branch not under release_lines. This is sound as a guard, but it is also the behaviour change the question above is about — before this PR, no dispatch from any branch could promote anything but v5.

Note for the author (no action required if intended): the body says the v5 and v6 copies of tagged-release.yml become "the same file"; that holds only after the v5→v6 forward-merge carries this commit, since the PR targets v5 only.

— hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
The parameterised tagged-release.yml renames push_v5 to push_line; the
contract test must follow. Re-pins TRUSTED_SCRIPTS_TREE.

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

@hivecommons-hive hivecommons-hive Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Decision needed: @clubanderson should confirm that the new tagged-release relay is meant to start dispatching v4's own Tagged Release on every successful v4 docker.yml build (v4's copy has a hardcoded workflow_dispatch backstop that will cut a v4.x.y tag whenever Unreleased is non-empty), and rebase — the PR is currently DIRTY against v5.

Routing to a human rather than judging: this edits 12 workflows plus the release/promotion path, so it is a trust-surface change by policy regardless of author. Facts I verified, then the one question.

Verified (read the workflow hunks and the v4 tree, not just the body):

  • The v5→${{ github.ref_name }} generalisation in tagged-release.yml and promote-stable.yml is consistently applied (concurrency group, RELEASE_BRANCH/RELEASE_LINE, every branches/v5, --ref v5, --base v5, gh pr list --base, ancestor check, --target). I found no remaining literal v5 in a code path.
  • Both workflows now fail closed with check-release-lines.sh --is-release-line "$RELEASE_LINE" before doing anything, and the guard is checked out from the workflow's own ref with persist-credentials: false and a sparse checkout of only the manifest and script.
  • The new relay job's token scope is contents: read + actions: write, which is the minimum gh workflow run needs. workflow_dispatch runs started with GITHUB_TOKEN do trigger, so the relay mechanism works.
  • TRUSTED_SCRIPTS_TREE in integration-governance.cjs is re-pinned; a maintainer should confirm the new tree hash is the one computed from this branch's src/scripts.

The one thing a person has to decide — a behaviour change on the live v4 line:

relay fires for any workflow_run.head_branch that is a declared release line and != github.ref_name. With release_lines: [v4, v5, v6], a successful docker.yml push build on v4 now dispatches tagged-release.yml --ref v4. v4's copy of that workflow (not touched by this PR) accepts workflow_dispatch at .github/workflows/tagged-release.yml:142 and its backstop then resolves the v4 tip (:197) and considers it for release (:231). Today v4's workflow_run trigger is dead (that event only fires on the default branch's copy, and v5's copy filtered head_branch == 'v5'), so v4 releases only happen via schedule/manual dispatch. After this PR every v4 build becomes a release opportunity. The PR body describes the relay as being "for non-default lines" with v6 in mind and does not mention v4 — if the v4 effect is intended, say so in the body; if not, the relay if (or the script check) should exclude lines whose own copy is not the generalised one.

Also: mergeable: DIRTY — this needs a rebase onto current v5 before it can land.

I have not approved or changed anything.

— hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88

Re-parameterises the new tagged-release.yml state machine on RELEASE_LINE
(github.ref_name): concurrency group, decide/backstop/precheck/push_line/
gh-release steps, --is-release-line guard, relay job for non-default lines.
test-release-push-retry.sh follows (push_line, RELEASE_LINE env, #9154
decide assertion). TRUSTED_SCRIPTS_TREE re-pinned.

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

@hivecommons-hive hivecommons-hive Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

HUMAN DECISION NEEDED — @clubanderson sign off that tagged-release.yml on v6 becomes a publishing workflow (tags, GitHub Releases, GHCR retags) and that the new relay job (.github/workflows/tagged-release.yml:217-258, actions: write) may auto-dispatch Tagged Release on every declared line, including v4.

Security delta re-review at cddc939 (previous head 30c64c7). Eight of the nine new commits are upstream v5; the only PR-specific change is the merge resolving tagged-release.yml/test-release-push-retry.sh against #17083.

Verified in the delta

  • relay filter unchanged: workflow_run + conclusion == 'success' + event in (push, workflow_dispatch) + head_branch != ref_name (tagged-release.yml:218-222), then ^v[0-9]+$ and check-release-lines.sh --is-release-line from the default-branch checkout (:248-256) before --ref "$LINE" (:258).
  • decide still requires event in (push, workflow_dispatch) before trusting head_branch (:270-273); RELEASE_LINE is guarded by --is-release-line (:357) before it reaches jq/regex/--ref/--base (:310,391,491).
  • TRUSTED_SCRIPTS_TREE = 'c06705a4339105f9cdc51220053c319a2b90dce1' (integration-governance.cjs:12) equals rev-parse cddc939:src/scripts.
  • No v5 literal remains in any non-comment code path of tagged-release.yml.

Previous findings

  • [security] workflow + release-publishing change needs maintainer sign-off — still present (it is the nature of the PR, not a defect; decision above).
  • [intent-alignment] relay also dispatches Tagged Release for v4 — still present: release_lines: [v4, v5, v6] (.github/release-lines.yml:42), so a v4 push build now auto-dispatches v4's own copy (which is self-consistent: concurrency: tagged-release-v4, RELEASE_LINE: v4 at v4:tagged-release.yml:145,258), turning v4 from manual-dispatch-only into auto-releasing. Either exclude v4 in the relay or state this is intended.

New (low) — the merge dropped the comment that explains why the event filter makes head_branch trustworthy (fork PR head_branch can be named v5). The code is intact at :270-273; only the rationale guarding against a future "simplification" is gone. Suggested restore at tagged-release.yml:262-266:

    # workflow_run: only a successful docker.yml build of a real push to, or
    # a workflow_dispatch on, the release line. The event filter is what makes
    # `head_branch` trustworthy: docker.yml also runs on `pull_request` with
    # no branch filter, and a fork PR's `head_branch` is whatever the fork
    # named its branch — `v5` included — so comparing the name alone would
    # admit a fork's commit. Push and workflow_dispatch runs are always on one
    # of this repository's own branches. Both count (#10863): post-release
    # dispatches are harmless because Unreleased is already empty, and the
    # backstop below may need to prime docker.yml manually when a busy line's
    # current tip has not yet published images.

— hive: agent=ci-maintainer backend=copilot model=claude-fable-5.1 copilot=1.0.88

@hivecommons-hive

Copy link
Copy Markdown
Contributor

Needs a decision, not a merge:

  • the reviewer swarm returned requires_human @cddc939
  • the review found: the reviewer marked this change high-risk
  • reviewer question: @hanthor should confirm that after this lands a workflow_dispatch of promote-stable.yml from v6 is allowed to move the stable channel to a v6 candidate (promote-stable.yml drops the RELEASE_BRANCH: v5 literal for ${{ github.ref_name }}), and sign off on the workflow-permission surface as a second approver.
  • the PR edits CI workflow / action files: awaiting owner approval — a hive owner / trusted merger's GitHub Approve review on the current head (or a dashboard /hive decision approve) releases the fence and the hive merges it; approval on an earlier head does not count unless the hive's own merge of the base branch is all that moved the head, so re-approve after a push

Head: verdict requires_human @cddc939, 81/81 checks green, DCO yes, soak 10m0s elapsed.

How to release it:

  • A hive owner / trusted merger submits a GitHub Approve review on the current head (or approves from the Hive dashboard, which posts /hive decision approve on their behalf); that explicit approval releases the workflow-path fence and the lane merges on its next pass once CI, soak and the other gates are clear. Approval on an earlier head carries over only the hive's own merge of the base branch (which leaves the diff unchanged): re-approve after any other push.
  • After reading the verdict, a trusted merger can submit a GitHub Approve review for the current head and the lane clears the hold on its next pass; /hive approve replies on a PR are acknowledged but not consumed. An ordinary contributor cannot release their own PR.

Clear the cause and the hive re-evaluates on its next pass; the stage moves to stage/090-ready-for-human-merge, stage/095-pre-merge-soak or the merge queue on its own. Approval does not bypass CI, hold labels or other merge gates.

@clubanderson
clubanderson merged commit bde4411 into v5 Nov 7, 2026
96 of 97 checks passed
@kubestellar-prow
kubestellar-prow Bot deleted the ci/v6-release-line branch November 7, 2026 21:59
@github-actions

github-actions Bot commented Nov 7, 2026

Copy link
Copy Markdown
Contributor

Thank you for your contribution! Your PR has been merged.

We'd love to hear how your experience was: share feedback

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

dco-signoff: yes Indicates the PR's author has signed the DCO. size/XL Denotes a PR that changes 500-999 lines, ignoring generated files.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant