You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Freeze the orphaned dev-lead/v1-* channel tags with a tombstone release, so a wrong repin fails loudly at startup instead of running a September build #2301
Maintainer decision, 2026-10-11: the legacy dev-lead/v1-* channel tags are frozen, not deleted. This closes the open question in the notes of #2220 ("dev-lead repins its own caller stub to the legacy dev-lead/v1-stable channel to silence template-drift …").
Problem
Four channel tags of the previous major still exist in this repository and still resolve:
Tag
Commit
dev-lead/v1-next
40f6c232
dev-lead/v1-ring0
6d362318
dev-lead/v1-ring1
6d362318
dev-lead/v1-stable
4fd4eacd (2026-09-22)
Nothing promotes them any more, so a caller stub that is repinned to one of them runs a September build with no error. That happened twice on 2026-10-10 (see #2220): the stub change was green at startup and the PR lost every guard released since.
On 2026-10-11 no default branch pins a v1 channel: the ten repositories with a dev-lead caller stub that were checked all pin dev-lead/v139-<tier> (.github, .github-private, TalkTerm, bmad-bgreat-suite, broodly, broodminder-data, ContentTwin, empower-personal-dashboard, google-app-scripts, incubator, repo-template). So the tags have no legitimate consumer.
What "frozen" has to mean
A run that resolves dev-lead-reusable.yml@dev-lead/v1-<tier> must stop in its first job with a message that names the frozen channel and the channel to pin instead. The code at the current v1 commits cannot do that, so the freeze is a tombstone release: the four v1 channel tags move, once, to a commit whose dev-lead-reusable.yml does nothing except fail with that message. After that they never move again.
The immutable dev-lead/v1.x.y release tags stay exactly where they are. They are the audit and rollback record.
Acceptance criteria
A tombstone commit exists. A branch (for example dev-lead/v1-frozen), cut from the current dev-lead/v1-stable commit, whose .github/workflows/dev-lead-reusable.yml keeps the same workflow_call inputs and secrets (so a caller's uses: still validates) and whose only job fails with: the channel is frozen, pin dev-lead/v139-<tier>, and a link to this issue. It calls no model and needs no secret.
The four v1 channel tags point at the tombstone, moved by the release-manager path (scripts/cut-release.sh … --channel v1-<tier> with a dry run first), not by a local push. This step is the maintainer's to run or approve.
Nothing can move them again by routine.standards/canary-rings.json and the canary rollout treat dev-lead major 1 as frozen: autocut and promote skip it, and evaluate reports it as frozen instead of as a blocked candidate.
A test proves the loud failure. A check that resolves each dev-lead/v1-<tier> tag and asserts the workflow at that commit is the tombstone, so a later accidental move is caught.
Order
Criterion 1 first, then 2 after the maintainer's go, then 3 and 4. Before step 2, re-check that no default branch and no open PR pins a v1 channel; a PR that still carries the repin (as #1796 did until 2026-10-11) will start failing loudly, which is the intended effect.
Out of scope
Deleting any tag. Other agents' legacy channels (for example the bare pr-review/stable that six repositories pin): they need their own decision.
Thanks for the thorough write-up and for recording the maintainer decision so clearly. The tombstone-release approach and the acceptance criteria look well scoped. One clarifying question: should the tombstone's failure message name a specific current dev-lead/v139-<tier> channel for each v1-<tier> tag (for example v1-next pointing to v139-next), or is a generic "pin the matching dev-lead/v139-<tier>" enough? That decides whether the tombstone needs per-tier text or one shared message.
Activity
github-actions commentedon Oct 11, 2026
Thanks for the thorough write-up and for recording the maintainer decision so clearly. The tombstone-release approach and the acceptance criteria look well scoped. One clarifying question: should the tombstone's failure message name a specific current
dev-lead/v139-<tier>channel for eachv1-<tier>tag (for examplev1-nextpointing tov139-next), or is a generic "pin the matchingdev-lead/v139-<tier>" enough? That decides whether the tombstone needs per-tier text or one shared message.