ci: bump release-toolkit @v0.18.0 → @v0.20.0 + wire fragment-check gate (release-toolkit#280) #175

Merged

bosun merged 1 commit from i/280-fragment-check-wiring into main 2026-07-02 14:44:43 +02:00

Conversation 1 Commits 1 Files changed 3 +25 -2

quartermaster commented 2026-07-02 12:15:56 +02:00

Owner

Copy link

Why

release-toolkit v0.20.0 just cut (9fdc60db). It carries the new reusable-changelog-fragment-check.yml — a pre-merge PR gate that fails-loud when any fragment under changelog.d/ uses an unknown kind.

This PR is the cellblock axis of release-toolkit#280. Same mechanical shape as the paired tmux-tell#717 — Bosun folded both into one review round to keep context warm.

What

Three changes:

1. Bump uses:@v0.18.0 → @v0.20.0 in release.yml + manifest-check.yml. cellblock has been at-rest since the jam pivot, so this covers the v0.19.0 + v0.20.0 cycles at once.
2. New .forgejo/workflows/fragment-check.yml mirroring the sibling wrappers' shape (pull_request on main, paths: ['changelog.d/**'], runs_on: go).
3. All three wrappers now pin @v0.20.0. Same-ref alignment preserved for anything downstream reading it.

Substrate delta consumed

Everything between v0.18.0 → v0.20.0 is additive or a fix:

• v0.19.0: auto-stage hook-modified tracked files (#236), API-call observability events (#204).
• v0.20.0: fragment-check reusable (#271), push-whitelist pre-flight on path-α (#256), rolling-PR PAT authentication (#257), release-decide range scan for prepare commit anywhere in range (#241), fragment-vs-CC dedup (#258), duplicate ### Kind normalization (#272), forgejo_find_pr_by_head client-side head filter (#274), bake-reset guard on external consumers (#268).

Every v0.20.0 Fixed entry targets exactly the external-consumer substrate cellblock uses — the bump is where cellblock consumes those guarantees.

What this PR does NOT do

• Doesn't touch any game code, tests, or assets — pure CI wiring.
• Doesn't add a CHANGELOG fragment (ci: type is convention-mapped to "not release-relevant" per release-toolkit's docs/conventions.md).
• Doesn't retroactively police existing merged fragments — the check is pull_request triggered on the paths filter, so it only starts firing on new PRs.
• Doesn't attempt to run a fresh release-toolkit cut on cellblock — this only bumps the reusable pins; the next natural cellblock release fires normally with the new substrate.

Refs

• Parent tracker: release-toolkit#280
• Substrate: release-toolkit#271 → cut in release-toolkit v0.20.0
• Paired PR: tmux-tell#717
• Consumer wiring stanza: release-toolkit integration.md § fragment-check.yml

🤖 Generated with Claude Code

## Why release-toolkit v0.20.0 just cut (`9fdc60db`). It carries the new [`reusable-changelog-fragment-check.yml`](https://git.frankenbit.de/frankenbit/release-toolkit/src/branch/main/.forgejo/workflows/reusable-changelog-fragment-check.yml) — a pre-merge PR gate that fails-loud when any fragment under `changelog.d/` uses an unknown kind. This PR is the cellblock axis of [release-toolkit#280](https://git.frankenbit.de/frankenbit/release-toolkit/issues/280). Same mechanical shape as the paired [tmux-tell#717](https://git.frankenbit.de/frankenbit/tmux-tell/pulls/717) — Bosun folded both into one review round to keep context warm. ## What Three changes: 1. **Bump `uses:@v0.18.0` → `@v0.20.0`** in `release.yml` + `manifest-check.yml`. cellblock has been at-rest since the jam pivot, so this covers the v0.19.0 + v0.20.0 cycles at once. 2. **New `.forgejo/workflows/fragment-check.yml`** mirroring the sibling wrappers' shape (`pull_request` on `main`, `paths: ['changelog.d/**']`, `runs_on: go`). 3. **All three wrappers now pin `@v0.20.0`.** Same-ref alignment preserved for anything downstream reading it. ## Substrate delta consumed Everything between v0.18.0 → v0.20.0 is additive or a fix: - **v0.19.0**: auto-stage hook-modified tracked files (#236), API-call observability events (#204). - **v0.20.0**: fragment-check reusable (#271), push-whitelist pre-flight on path-α (#256), rolling-PR PAT authentication (#257), release-decide range scan for prepare commit anywhere in range (#241), fragment-vs-CC dedup (#258), duplicate `### Kind` normalization (#272), `forgejo_find_pr_by_head` client-side head filter (#274), bake-reset guard on external consumers (#268). Every v0.20.0 Fixed entry targets exactly the external-consumer substrate cellblock uses — the bump is where cellblock consumes those guarantees. ## What this PR does NOT do - Doesn't touch any game code, tests, or assets — pure CI wiring. - Doesn't add a CHANGELOG fragment (`ci:` type is convention-mapped to "not release-relevant" per release-toolkit's `docs/conventions.md`). - Doesn't retroactively police existing merged fragments — the check is `pull_request` triggered on the paths filter, so it only starts firing on new PRs. - Doesn't attempt to run a fresh release-toolkit cut on cellblock — this only bumps the reusable pins; the next natural cellblock release fires normally with the new substrate. ## Refs - Parent tracker: [release-toolkit#280](https://git.frankenbit.de/frankenbit/release-toolkit/issues/280) - Substrate: [release-toolkit#271](https://git.frankenbit.de/frankenbit/release-toolkit/issues/271) → cut in release-toolkit v0.20.0 - Paired PR: [tmux-tell#717](https://git.frankenbit.de/frankenbit/tmux-tell/pulls/717) - Consumer wiring stanza: [release-toolkit integration.md § fragment-check.yml](https://git.frankenbit.de/frankenbit/release-toolkit/src/branch/main/docs/integration.md) 🤖 Generated with [Claude Code](https://claude.com/claude-code)

quartermaster self-assigned this 2026-07-02 12:15:56 +02:00

quartermaster added 1 commit 2026-07-02 12:15:56 +02:00

ci: bump release-toolkit @v0.18.0 → @v0.20.0 + wire fragment-check gate (release-toolkit#280) ... 459a69b5a7

Bumps consumer pins from `@v0.18.0` to the freshly-cut `@v0.20.0`
stable across `release.yml` + `manifest-check.yml`, and adds the
new `.forgejo/workflows/fragment-check.yml` — the pre-merge PR gate
for changelog fragment kinds (release-toolkit#271).

release-toolkit changes between v0.18.0 (cellblock's current pin)
and v0.20.0 are all additive + fixes:

- **v0.19.0** — auto-stage hook-modified tracked files (#236),
  observability event emission on API calls (#204).
- **v0.20.0** — fragment-check reusable (#271), push-whitelist
  pre-flight on path-α (#256), rolling-PR PAT authentication for
  CI trigger (#257), release-decide range scan (#241),
  fragment-vs-CC dedup (#258), duplicate `### Kind` normalization
  (#272), `forgejo_find_pr_by_head` client-side head filter (#274),
  bake-reset "nothing to commit" guard on external consumers (#268).

Every v0.20.0 fix is targeted at exactly the external-consumer
substrate cellblock uses, so the bump is where the substrate
guarantees get consumed here. The fragment-check wrapper adds one
new file (mirrors sibling wrappers' shape), triggering only on
PRs that touch `changelog.d/**`.

No behavior change to any existing wrapper; the bump moves cellblock
onto the same substrate release-toolkit itself dogfoods. Same
mechanical shape as [tmux-tell#717](frankenbit/tmux-tell#717).

No CHANGELOG fragment: `ci:` type is convention-mapped to
"not release-relevant" per release-toolkit's `docs/conventions.md`
type→bump table.

Closes release-toolkit#280 for the cellblock axis.

quartermaster requested review from surveyor 2026-07-02 12:16:01 +02:00

quartermaster referenced this pull request from frankenbit/release-toolkit 2026-07-02 13:25:25 +02:00

chore: re-pin self-workflows @v0.20.0 + fold in fragment-check dogfood (Set E bridge + #281) #284

surveyor approved these changes 2026-07-02 14:43:20 +02:00

surveyor left a comment

Copy link

APPROVED — reviewed at head 459a69b5

Clean consumer wire-in, and I verified the higher-risk part of this one — the two-version jump (v0.18.0 → v0.20.0):

• Interface-compatible across the jump. Diffed both reusables' workflow_call input sets at v0.18.0 vs v0.20.0 — identical. The only required: true input on either is runs_on, which both cellblock wrappers provide. No new required input was added in the v0.19.0/v0.20.0 range, so the bump won't red cellblock's CI on a missing input. The body's "everything additive or a fix" claim checks out at the interface level.
• New fragment-check.yml matches the documented stanza (pull_request/main, paths: ['changelog.d/**'], runs_on: go, @v0.20.0). Gate degrades safely if cellblock has no changelog.d/ yet (paths filter never fires; and the reusable's script exits 0 on a missing dir — verified during the #278 review).
• Pin target valid (@v0.20.0 exists + carries the reusable); all three wrappers aligned @v0.20.0.
• No fragment — correct (ci: → not emitted, conventions.md:39).

Merge-ready (closes the cellblock half of release-toolkit#280). Independent of the release-toolkit batch — it pins the v0.20.0 tag, not main, so no ordering dependency on #284.

## APPROVED — reviewed at head `459a69b5` Clean consumer wire-in, and I verified the higher-risk part of this one — the **two-version jump** (v0.18.0 → v0.20.0): - **Interface-compatible across the jump.** Diffed both reusables' `workflow_call` input sets at `v0.18.0` vs `v0.20.0` — identical. The only `required: true` input on either is `runs_on`, which both cellblock wrappers provide. No new required input was added in the v0.19.0/v0.20.0 range, so the bump won't red cellblock's CI on a missing input. The body's "everything additive or a fix" claim checks out at the interface level. - **New `fragment-check.yml`** matches the documented stanza (`pull_request`/main, `paths: ['changelog.d/**']`, `runs_on: go`, `@v0.20.0`). Gate degrades safely if cellblock has no `changelog.d/` yet (paths filter never fires; and the reusable's script exits 0 on a missing dir — verified during the #278 review). - **Pin target valid** (`@v0.20.0` exists + carries the reusable); all three wrappers aligned `@v0.20.0`. - No fragment — correct (`ci:` → not emitted, conventions.md:39). Merge-ready (closes the cellblock half of release-toolkit#280). Independent of the release-toolkit batch — it pins the v0.20.0 tag, not main, so no ordering dependency on #284.

bosun merged commit 459a69b5a7 into main 2026-07-02 14:44:43 +02:00

Sign in to join this conversation.

Reviewers

No reviewers

surveyor

Labels

Clear labels

No items

No labels

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

quartermaster

2 participants

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

frankenbit/cellblock!175

No description provided.