probe: undeclared-input-passed behavior post-#172 (lazy-leave adopter guidance) #250

New issue

Closed

opened 2026-06-28 21:48:21 +02:00 by quartermaster · 0 comments

quartermaster commented

2026-06-28 21:48:21 +02:00

Owner

The question

After #172 retired the toolkit_ref: input (v0.16.0), the recommended adopter path is drop + bump (drop the toolkit_ref: line, bump @<ref> on the reusable). The clean-migration path is empirically validated (cellblock#171, 2026-06-28).

What's NOT empirically validated: the lazy-leave path — adopter bumps @<ref> but FORGETS to drop the toolkit_ref: input. Does Forgejo:

(a) Error fail-loud at workflow-parse time? ("undeclared input")
(b) Silently ignore the input + run normally?
(c) Something else (warn-and-continue, deprecation note, etc.)?

Per Surveyor d7d9 (cellblock#171 review): this is a separate one-line probe from the clean migration test.

Why it matters

Adopter-documentation guidance changes based on the answer:

(a) errors loudly → "you MUST drop toolkit_ref: when bumping past v0.16.0" — a clear documentation imperative + adopters get a fast-fail at CI
(b) silently ignores → "you SHOULD drop toolkit_ref: for hygiene, but stale configs still work" — softer documentation + adopters can lazily migrate
(c) other → calibrate guidance to the actual behavior

This is a 5-minute probe with documentation value, not a substrate bug — just unknown post-#172 behavior.

Probe shape

Create a deliberately-malformed consumer wrapper:

jobs:
  release:
    uses: frankenbit/release-toolkit/.forgejo/workflows/reusable-release.yml@v0.18.0
    with:
      toolkit_ref: v0.18.0  # ← retired, deliberately left in
      runs_on: go

Run it (e.g., on a throwaway test branch) + observe the workflow-run behavior. Record the outcome here + update integration.md adopter guidance accordingly.

Could be done as a no-op test PR on a sandbox repo, or even on cellblock as a temporary test commit that gets reverted.

Refs

release-toolkit#223 (parent — clean-migration empirically validated 2026-06-28)
release-toolkit#172 (toolkit_ref retirement, v0.16.0)
cellblock#171 (the clean-migration anchor)
Surveyor d7d9 (scope precision call)

priority/low · size/S

## The question After #172 retired the `toolkit_ref:` input (v0.16.0), the recommended adopter path is **drop + bump** (drop the `toolkit_ref:` line, bump `@<ref>` on the reusable). The clean-migration path is empirically validated (cellblock#171, 2026-06-28). What's NOT empirically validated: the **lazy-leave path** — adopter bumps `@<ref>` but FORGETS to drop the `toolkit_ref:` input. Does Forgejo: - (a) Error fail-loud at workflow-parse time? ("undeclared input") - (b) Silently ignore the input + run normally? - (c) Something else (warn-and-continue, deprecation note, etc.)? Per Surveyor d7d9 (cellblock#171 review): this is a separate one-line probe from the clean migration test. ## Why it matters Adopter-documentation guidance changes based on the answer: - **(a) errors loudly** → "you MUST drop `toolkit_ref:` when bumping past v0.16.0" — a clear documentation imperative + adopters get a fast-fail at CI - **(b) silently ignores** → "you SHOULD drop `toolkit_ref:` for hygiene, but stale configs still work" — softer documentation + adopters can lazily migrate - **(c) other** → calibrate guidance to the actual behavior This is a 5-minute probe with documentation value, not a substrate bug — just unknown post-#172 behavior. ## Probe shape Create a deliberately-malformed consumer wrapper: ```yaml jobs: release: uses: frankenbit/release-toolkit/.forgejo/workflows/reusable-release.yml@v0.18.0 with: toolkit_ref: v0.18.0 # ← retired, deliberately left in runs_on: go ``` Run it (e.g., on a throwaway test branch) + observe the workflow-run behavior. Record the outcome here + update integration.md adopter guidance accordingly. Could be done as a no-op test PR on a sandbox repo, or even on cellblock as a temporary test commit that gets reverted. ## Refs - release-toolkit#223 (parent — clean-migration empirically validated 2026-06-28) - release-toolkit#172 (toolkit_ref retirement, v0.16.0) - cellblock#171 (the clean-migration anchor) - Surveyor d7d9 (scope precision call) priority/low · size/S