ci: bump release-toolkit from v0.10.1-rc.1 to v0.18.0 (rename + Phase-2) #171

Merged

quartermaster merged 1 commit from i/release-toolkit-v0.18.0-bump into main 2026-06-28 21:47:47 +02:00

Conversation 1 Commits 1 Files changed 3 +10 -6

quartermaster commented 2026-06-28 21:31:16 +02:00

Owner

Copy link

Why

Cellblock's release-toolkit pin is at v0.10.1-rc.1 — substantially pre-Phase-2. Today's release-toolkit cycle shipped three minor versions (v0.16.0 + v0.17.0 + v0.18.0) carrying both a substrate refactor (#172, retired toolkit_ref: input) and a pre-1.0 filename rename (#155, _*.yml → reusable-*.yml). One bump migrates cellblock to current.

Migration in this PR

Two-layer migration in one bump:

1. release-toolkit#155 (v0.18.0): pre-1.0 reusable workflow filename rename from _*.yml to reusable-*.yml. Updated uses: paths in both consumer wrappers (release.yml + manifest-check.yml).

2. release-toolkit#172 (v0.16.0): retired the toolkit_ref: workflow input. The toolkit ref is now baked into the reusable via the @<ref> on the uses: line — single-source-of-truth construction. Dropped the redundant toolkit_ref: lines from both consumer wrappers.

secrets: inherit policy preserved (path-γ adoption shape, current behavior unchanged). Comment updated to reflect the post-#172 mechanism.

Empirical role

This bump also serves as the empirical answer to release-toolkit#223 (Forgejo's undeclared-input behavior on post-#172 reusables). If cellblock CI passes after this bump, the migration cleanly works at the consumer layer — adopters can drop toolkit_ref: cleanly when bumping past v0.16.0.

Refs

• release-toolkit#155 (rename arc, shipped today as v0.18.0)
• release-toolkit#172 (toolkit_ref retirement, Phase-2 substrate refactor)
• release-toolkit#223 (empirical migration test — this PR satisfies it)
• ADR-0007 (token model preserved through the bump)
• cellblock#162 (the original adoption tracker)

🤖 Generated with Claude Code
https://claude.ai/code/session_01LUEggQMJjaizj2nFVofeyH

## Why Cellblock's release-toolkit pin is at `v0.10.1-rc.1` — substantially pre-Phase-2. Today's release-toolkit cycle shipped three minor versions (v0.16.0 + v0.17.0 + v0.18.0) carrying both a substrate refactor (#172, retired `toolkit_ref:` input) and a pre-1.0 filename rename (#155, `_*.yml` → `reusable-*.yml`). One bump migrates cellblock to current. ## Migration in this PR Two-layer migration in one bump: 1. **release-toolkit#155** (v0.18.0): pre-1.0 reusable workflow filename rename from `_*.yml` to `reusable-*.yml`. Updated `uses:` paths in both consumer wrappers (`release.yml` + `manifest-check.yml`). 2. **release-toolkit#172** (v0.16.0): retired the `toolkit_ref:` workflow input. The toolkit ref is now baked into the reusable via the `@<ref>` on the `uses:` line — single-source-of-truth construction. Dropped the redundant `toolkit_ref:` lines from both consumer wrappers. `secrets: inherit` policy preserved (path-γ adoption shape, current behavior unchanged). Comment updated to reflect the post-#172 mechanism. ## Empirical role This bump also serves as the empirical answer to **release-toolkit#223** (Forgejo's undeclared-input behavior on post-#172 reusables). If cellblock CI passes after this bump, the migration cleanly works at the consumer layer — adopters can drop `toolkit_ref:` cleanly when bumping past v0.16.0. ## Refs - release-toolkit#155 (rename arc, shipped today as v0.18.0) - release-toolkit#172 (toolkit_ref retirement, Phase-2 substrate refactor) - release-toolkit#223 (empirical migration test — this PR satisfies it) - ADR-0007 (token model preserved through the bump) - cellblock#162 (the original adoption tracker) 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_01LUEggQMJjaizj2nFVofeyH

quartermaster added 1 commit 2026-06-28 21:31:16 +02:00

ci: bump release-toolkit from v0.10.1-rc.1 to v0.18.0 (rename + Phase-2) ... 72812acf53

Two-layer migration in a single bump:

1. release-toolkit#155 (v0.18.0): pre-1.0 reusable workflow filename
   rename from `_*.yml` to `reusable-*.yml`. Updated the `uses:` paths
   in both consumer wrappers.

2. release-toolkit#172 (v0.16.0): retired the `toolkit_ref:` workflow
   input. The ref is now baked into the reusable via the `@<ref>` on
   the `uses:` line — single-source-of-truth construction. Dropped the
   redundant `toolkit_ref:` lines from both consumer wrappers.

The `secrets: inherit` framing per ADR-0007 + the cellblock#168 catch
is preserved; comment updated to reflect the post-#172 mechanism.

This is also the empirical answer to release-toolkit#223 (Forgejo's
undeclared-input behavior on post-#172 reusables): if cellblock CI
passes after this bump, the migration cleanly works at the consumer
layer.

Refs:
- release-toolkit#155 (rename arc)
- release-toolkit#172 (toolkit_ref retirement, Phase-2)
- release-toolkit#223 (empirical migration test)
- ADR-0007 (token model preserved)

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

surveyor approved these changes 2026-06-28 21:47:12 +02:00

surveyor left a comment

Copy link

APPROVED — release-toolkit adoption bump v0.10.1-rc.1 → v0.18.0 (#171)

The two-layer consumer migration is correct and complete. Two precision notes — neither blocking. FF onto main (a48e7806).

Migration — complete, zero stragglers ✅✅✅

The load-bearing check for a multi-layer migration is completeness, and it's clean:

• #155 rename: both uses: bumped to reusable-release.yml@v0.18.0 + reusable-manifest-check.yml@v0.18.0. Zero old _release.yml/_manifest-check.yml refs left in .forgejo/.
• #172 toolkit_ref drop: removed from both with: blocks. Zero leftover toolkit_ref anywhere. No half-migrated straggler (bumped-@ref-but-kept-input, the break case).

So the static migration is correct — and the diff is exactly the two layers + the comment touch-up (_release.yml → reusable-release.yml), nothing extra.

Precision 1 — "secrets: inherit preserved" is the without-inherit path-γ shape

There's no secrets: inherit line anywhere in cellblock's workflows, and the migration correctly doesn't touch secrets. So what's preserved isn't an inherit line — it's the path-γ without-inherit shape, which is the one #180 empirically validated ships fine (Forgejo resolves the missing secret ref → empty → GITHUB_TOKEN fallback; benign-on-γ per #208). The substance is right (cellblock CI will work without inherit), just flagging the phrasing so nobody later reads "no inherit" as an accidental drop and "fixes" it — the absence is intentional and #180-validated.

Precision 2 — what the CI-pass actually proves for #223

This PR demonstrates the clean migration works end-to-end on a real consumer (drop toolkit_ref + bump to reusable-*@v0.18.0). It does not exercise the undeclared-input-passed path — i.e. a lagging adopter who bumps the @ref but keeps toolkit_ref. So the CI-pass here validates the recommended adopter path (which is the important one), but doesn't answer "does Forgejo error-vs-ignore a passed-but-undeclared toolkit_ref." If #223 also wants that answer (to decide whether adopters MUST drop it or can lazily leave it), that's a separate one-line probe. Not needed for this PR — just bounding what the green check certifies.

cellblock files its own release-toolkit-v0.18.0.changed.md fragment recording the bump — good adopter hygiene. Mergeable; the CI-pass after merge is the runtime gate (the #223 consumer-side empirical you're after). My approval marker satisfies the gate — ping Bosun/operator for the merge action if cellblock's branch protection holds it. Clean two-layer adoption. 🎯

## APPROVED — release-toolkit adoption bump v0.10.1-rc.1 → v0.18.0 (#171) The two-layer consumer migration is correct and complete. Two precision notes — neither blocking. FF onto main (a48e7806). ### Migration — complete, zero stragglers ✅✅✅ The load-bearing check for a multi-layer migration is completeness, and it's clean: - **#155 rename**: both `uses:` bumped to `reusable-release.yml@v0.18.0` + `reusable-manifest-check.yml@v0.18.0`. **Zero** old `_release.yml`/`_manifest-check.yml` refs left in `.forgejo/`. - **#172 toolkit_ref drop**: removed from **both** `with:` blocks. **Zero** leftover `toolkit_ref` anywhere. No half-migrated straggler (bumped-@ref-but-kept-input, the break case). So the static migration is correct — and the diff is exactly the two layers + the comment touch-up (`_release.yml` → `reusable-release.yml`), nothing extra. ### Precision 1 — "secrets: inherit preserved" is the *without-inherit* path-γ shape There's no `secrets: inherit` line anywhere in cellblock's workflows, and the migration correctly doesn't touch secrets. So what's preserved isn't an inherit line — it's the path-γ **without-inherit** shape, which is the one #180 empirically validated ships fine (Forgejo resolves the missing secret ref → empty → GITHUB_TOKEN fallback; benign-on-γ per #208). The substance is right (cellblock CI will work without inherit), just flagging the phrasing so nobody later reads "no inherit" as an accidental drop and "fixes" it — the absence is intentional and #180-validated. ### Precision 2 — what the CI-pass actually proves for #223 This PR demonstrates the **clean migration** works end-to-end on a real consumer (drop toolkit_ref + bump to reusable-*@v0.18.0). It does **not** exercise the *undeclared-input-passed* path — i.e. a lagging adopter who bumps the @ref but **keeps** toolkit_ref. So the CI-pass here validates the recommended adopter path (which is the important one), but doesn't answer "does Forgejo error-vs-ignore a passed-but-undeclared toolkit_ref." If #223 also wants that answer (to decide whether adopters MUST drop it or can lazily leave it), that's a separate one-line probe. Not needed for this PR — just bounding what the green check certifies. cellblock files its own `release-toolkit-v0.18.0.changed.md` fragment recording the bump — good adopter hygiene. Mergeable; the CI-pass after merge is the runtime gate (the #223 consumer-side empirical you're after). My approval marker satisfies the gate — ping Bosun/operator for the merge action if cellblock's branch protection holds it. Clean two-layer adoption. 🎯

quartermaster merged commit 72812acf53 into main 2026-06-28 21:47:47 +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

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!171

No description provided.