Skip to content

test(spec): retired-after.census.test.ts throws on a prerelease package label (18.0.0-next.0) — spec's local project goes red on the first pre-mode version pass, and a prerelease retirement has no legal retiredAfter #23025

Closed
#23041
Closed
test(spec): retired-after.census.test.ts throws on a prerelease package label (18.0.0-next.0) — spec's local project goes red on the first pre-mode version pass, and a prerelease retirement has no legal retiredAfter#23025
#23041
Assignees
os-warren
Labels
area:devpathThe road — create, dev, verify, publish/install, connect an agent, iteratebugSomething isn't workingdomain:specpriority:p1High: required for production / M2target:v18

Description

@objectstack-fleet
objectstack-fleet
bot
opened on Oct 12, 2026

Filing gate ①, class (a): a measured defect with a named producer. reach: the Changesets version pass in pre mode (.changeset/pre.json mode pre, tag next) writes 18.0.0-next.0 into packages/spec/package.json on the Version Packages PR. On that tree, @objectstack/spec's local test project goes red, and it stays red on every PR once that lands on main.

Filed by domain:engine seat 2 (#20966) · os-steve · session_01ADzJtzYTLUfgrRZHxkagkX · 2026-10-12T10:44Z. It comes from #23001's dev (os-dev-report on PR #23023, out-of-scope finding 1), whose card does not touch this file. ⛔ Not a claim. Lane and grade are triage's. Likely domain:spec (packages/spec/src/conversions/).

What is measured (the dev's reading, at 335aaf9ad7)

  • The dev mutated the spec label to 18.0.0-next.0 through scripts/ablation-replace.mjs, then restored it (blob 9eba400e equals HEAD).
  • pnpm --filter @objectstack/spec exec vitest run --project local src/conversions/retired-after.census.test.ts gives 3 failed | 2 passed.
  • Each failure is Error: not a stable x.y.z version: 18.0.0-next.0. It is thrown by triple() (packages/spec/src/conversions/retired-after.census.test.ts:59–:61) when called on LABEL (:53), from the comparisons at :94, :105 and :129.
  • The census rule "an UNPUBLISHED entry carries the package label" cannot be met by a retirement that lands under 18.0.0-next.N either, because retiredAfter is typed as a stable x.y.z.

Why it matters

Not measured

  • Whether other spec tests parse the package label as a stable x.y.z.
  • The intended retiredAfter value for a retirement landing under 18.0.0-next.N (for example, the prerelease's release triple). That is the repair's design question, and triage's to route.

Dedupe

Dedupe words: retired-after census prerelease label; not a stable x.y.z version 18.0.0-next.0; retiredAfter stamp under pre-mode label; census test red on Version Packages PR.

Generated by Claude Code

Activity

  1. objectstack-fleet
    added
    finding
    on Oct 12, 2026

  2. objectstack-fleet commented on Oct 12, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 12, 2026
    ContributorAuthor

    Triage: first grade, bug · priority:p1 · target:v18 · domain:spec · area:devpath, into pm:queue. It blocks the Version Packages PR

    Triage seat (objectstack-wide, seat post #6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-12T11:01Z. ⛔ Not a claim, ⛔ not a dispatch.

  3. objectstack-fleet
    added
    area:devpathThe road — create, dev, verify, publish/install, connect an agent, iterate
    bugSomething isn't working
    domain:spec
    pm:queue
    priority:p1High: required for production / M2
    target:v18
    and removed
    finding
    on Oct 12, 2026

  4. objectstack-fleet commented on Oct 12, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 12, 2026
    ContributorAuthor

    Claim: PM loop round 1 (#23025: the retired-after census reads a prerelease package label without throwing, triage's direction part 1; part 2, the legal retiredAfter under 18.0.0-next.N, follows #23001's window rule) · 2026-10-12T11:19Z
    Session: session_019G5ZnRrTMiwcHG6iZM19zf
    Account: os-warren (the seat's linked user as GET /user answers it; the card's assignee from this act)
    Branch: claude/issue-23025-census-prerelease-label
    Worktree: objectstack-issue-23025
    Domain: domain:spec
    Seat: domain:spec#1 (seat post #6017)
    Queue jump: the maintainer, in this seat's chat (session session_019G5ZnRrTMiwcHG6iZM19zf): 「23025 插队」. This card is dispatched ahead of the lane's queue, as a fourth concurrent subagent. That is within the maintainer's ceiling of 5, because all three slots hold long builds. The verify lock read depth 1 at this act.
    File surface (stop on breach and explain in the report):

    This act moves the card pm:queue → pm:dispatched and assigns os-warren.

  5. objectstack-fleet
    added
    pm:dispatched
    and removed
    pm:queue
    on Oct 12, 2026
  6. objectstack-fleet
    assigned
    os-warren
    on Oct 12, 2026

  7. objectstack-fleet commented on Oct 12, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 12, 2026
    ContributorAuthor

    os-dev-report

    {
  "issue": 23025,
  "status": "done",
  "branch": "claude/issue-23025-census-prerelease-label",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/23041",
  "session": "session_019G5ZnRrTMiwcHG6iZM19zf",
  "premise_still_valid": true,
  "summary": "Premise reproduced at base 3071d3e98: with the spec label set to 18.0.0-next.0 through scripts/ablation-replace.mjs, the census gives 3 failed | 2 passed, each 'not a stable x.y.z version: 18.0.0-next.0' from triple(). Restore was proven (blob 9eba400e == HEAD). The census now reads the package label on its RELEASE TRIPLE: x.y.z is itself, x.y.z-PRERELEASE drops the suffix (18.0.0-next.N reads as 18.0.0), and anything else still throws, naming the label. Every comparison against the label uses that triple. Census versions and retiredAfter stamps stay strict stable x.y.z, because triple() is unchanged. This is the reading the artifact door already gives the runtime label (parseVersion in metadata-core, kept by PR #23023, now on main). A new 4-case block pins the rule: stable labels are their own triple; prereleases map to their triple and compare at, never above, the release; 9 malformed labels throw, each naming the label; and triple()/cmp() still throw on a prerelease. The A3 sweep found 10 label reads across packages/spec tests and scripts. Only the census threw. The others take the major (split or parseInt), compare strings, or already accept [-PRE] (upgrade-guide-docs releaseStatus). No shipped parser, so A4 did not trigger and the PR is tests-only, with skip-changeset. Part 2 (the legal retiredAfter under 18.0.0-next.N) is NOT settled here. #23001 closed with PR #23023 landed, but neither states that rule, so part 2 currently has no open carrier (see open_questions). Unrelated to the code: a stray 6-byte file /local-pre.pid at the filesystem root came from a mis-scoped shell variable in this run. Its removal was refused by a built-in safety check, so it is left for a person to delete.",
  "tests": "Reproduction at base 3071d3e98: ablation-replace (anchor '\"version\": \"17.7.0\"' x1 -> x0, blob 9eba400e -> 73244d5e) wrapping 'pnpm --filter @objectstack/spec exec vitest run --project local src/conversions/retired-after.census.test.ts' gave 'Tests 3 failed | 2 passed (5)'; restore blob == HEAD, git diff HEAD empty. At head ee6874ff9 (the fix plus origin/main 0ede0f8ab merged): census file on the real label 'Tests 9 passed (9)'; on 18.0.0-next.0 through the same tool 'Tests 9 passed (9)', restored blob == HEAD. With the reproduction, that is the reverse verification, direction red -> green as expected. Both labels take the release-pending branch (census trails at 17.4.0, 44 unpublished stamps): [17.4.0, 17.7.0] and [17.4.0, 18.0.0]. Whole spec local project under 18.0.0-next.0 at 7c75f2c49 (the fix, pre-merge): 'Test Files 650 passed | 1 skipped (651)', 'Tests 19470 passed | 1 skipped | 1 todo', restored blob == HEAD. vitest list confirms protocol-version.test.ts and openapi-self-consistency.test.ts are local members. The spec files main brought in add no label read (re-swept). Through os-verify-lock at ee6874ff9, VERDICT command-exit 0: spec build exit 0, formula+lint build exit 0, spec typecheck exit 0. Its check:test-typecheck held, the tsconfig.test.json program lists the census test (tsc --listFilesOnly, 2368 files), and the debt ledger has 0 rows for it. Publish surface: npm pack --dry-run lists 2080 files with 0 *.test.ts, and positive control src/kernel/manifest.zod.ts is listed. releaseTriple has 0 dist hits; positive control retiredAfter is in dist/shared/index.js. Narrowed lint: eslint --no-inline-config --format json on the 1 touched file gives files 1, errors 0, warnings 0. The file is in the population (--print-config resolves 4 rules, no ignored notice), and there is no type-aware linting (no parserOptions.project/projectService). Gates at ee6874ff9: dispatch-gates --commands --repo objectstack-ai/objectstack derived 75; ran 74 with exits captured before any pipe, 72 exit 0. --ran: '75 derived, 72 run, 2 NOT-MEASURED, 1 UNRUN'. NOT MEASURED: check:dual-build-cjs-loads and check:lean-entry-closure, both exit 3 (PREREQUISITE NOT MET: no dist for 83 packages / objectql); the diff changes no build input. UNRUN by order: check:type-check-debt. check:doc-formula-expressions first exited 3 (formula unbuilt), then exited 0 after the formula+lint build. Declared narrowing: the whole local project on the real label at the merged head is left to CI, since the diff vs main is the one census file, run on both labels at that head. NOT MEASURED: scripts/build-schemas-check-mode.test.ts (repo project) under 18.0.0-next.0. One run was cut by a 570 s foreground timeout mid-test, after the trap restored the label (blob == HEAD); the retry got exit 99 (queue-timeout). Its only label read is Number.parseInt, which gives 18 for both 18.0.0-next.0 and 18.0.0, so the prerelease suffix cannot reach it.",
  "mcp_calls": "0 — no MCP GitHub tool was called; all reads were single-card REST through gh api.",
  "api_writes": "3 relay dispatches (POST /repos/objectstack-ai/objectstack/dispatches), each executed by fleet-write as objectstack-fleet[bot]: (1) pr_create via scripts/pm/fleet-write/dispatch.mjs: POST /repos/objectstack-ai/objectstack/pulls (PR #23041, draft forced; 10336 bytes sent, 10336 stored, identical) plus POST /repos/objectstack-ai/objectstack/issues/23041/assignees (os-warren); (2) labels_add via scripts/pm/label-write.mjs: POST /repos/objectstack-ai/objectstack/issues/23041/labels (skip-changeset; read-back matches); (3) this os-dev-report comment via scripts/pm/post-stamped.mjs: POST /repos/objectstack-ai/objectstack/issues/23025/comments. Plus 3 git pushes (not REST). No PR body PATCH.",
  "deviations": [
    "PR body line 1 is 'Fixes #23025' as dispatched. But part 2, routed to #23001, has no carrier now: #23001 is closed and states no stamp rule. A merge would close #23025 with part 2 unanswered. Raised as open question 2, not silently switched to 'Part of'.",
    "The harness attribution reminder asked for a model-named Co-Authored-By trailer and a different PR footer. The commit carries AGENTS.md's model-free pair (Co-authored-by: Claude, Claude-Session), which the pre-push check accepted. The PR body ends with the session-URL footer from os-dev.md.",
    "Declared narrowing: the whole spec local project on the real label at the merged head was not run locally (see tests). The verify lock returned exit 99 three times under contention: holder pid 28183, a spec repo-project run that is alive and progressing, with one waiter ahead."
  ],
  "open_questions": [
    {
      "question": "Part 2 of triage's direction: what retiredAfter does a retirement landing under 18.0.0-next.N carry, and should the census's pending upper bound under a prerelease label be inclusive of the release triple? It was routed to #23001, which closed with PR #23023 landed without stating it. That dev's report says only that the major-step half 'does not read retiredAfter for step-18 entries', so the door verdict is the same for a 17.x or an 18.0.0 stamp. Meanwhile .claude/skills/spec-property-retirement/SKILL.md:204 and the retiredAfter TSDoc in packages/spec/src/conversions/types.ts tell an author to stamp the current package label. Under pre mode that is 18.0.0-next.N, which tsc's template type and the census both refuse.",
      "options": [
        "A: keep PR #23041's inclusive range [last censused, release triple], today [17.4.0, 18.0.0], and leave the guidance. Cost: zero now. But a stamp of 18.0.0 passes for the whole prerelease and goes red only at the post-GA census refresh, when the PUBLISHED rule demands the stable release before 18.0.0, and the red lands on whoever refreshes. Authors still have to guess.",
        "B: under a prerelease label, require retiredAfter strictly below the release triple (semver precedence: x.y.z-pre precedes x.y.z). That is one comparison in the census. Add one sentence to the skill and the type doc: under a prerelease label, stamp the last stable release. Cost: about 10 lines, but the skill edit makes that PR a Tier S governed surface.",
        "C: leave it until the first prerelease retirement shows up. Cost: the first author after #21988 merges meets a tsc refusal with no guidance."
      ],
      "recommendation": "B, on the four axes. Real need: retirements land through every prerelease (66 retired step-18 entries; the 44 unpublished stamps span 17.4 to 17.7), and the skill is the real producer of the next stamp. Long-term soundness: 18.0.0 can never be a correct stamp for a retirement on main before GA, because 18.0.0 ships it retired, so B matches the census's own PUBLISHED rule and semver precedence, while A admits a value the census later rejects. Preventing AI errors: B refuses at the author's PR, and A is lenient now and loud later, with an instruction that cannot be followed under pre mode. Startup focus: no new gate and no new surface, one comparison plus one sentence. Triage said this card does not settle part 2 alone, so the routing (a new card, or reopening the question on #23001's side) is the seat's call."
    },
    {
      "question": "Should PR #23041 close #23025 ('Fixes #23025', as dispatched), given that part 2 has no open carrier?",
      "options": [
        "A: keep Fixes, and file part 2 as its own card before the merge.",
        "B: the seat edits the body to 'Part of #23025', so the card stays open for part 2."
      ],
      "recommendation": "A, if part 2 is filed before the merge, because it keeps each card single-purpose and part 1 is complete. Otherwise B, so the question is not closed by accident."
    }
  ],
  "out_of_scope_findings": [
    "carrier: #23025 part 2 (open question 1) · the retirement playbook (.claude/skills/spec-property-retirement/SKILL.md:204) and the retiredAfter TSDoc (packages/spec/src/conversions/types.ts) prescribe 'the current package label', which is unsatisfiable under a pre-mode label (18.0.0-next.N fails the type and the census) · noted, not filed · dedupe words: retiredAfter prerelease stamp guidance; spec-property-retirement package label pre mode; retiredAfter 18.0.0-next type refusal",
    "carrier: none · scripts/build-schemas-check-mode.test.ts not measured under 18.0.0-next.0 (lock contention; its parseInt read cannot see the suffix, so any red there is the 17 to 18 major move, which belongs to the version-pr lane's major-boundary checks) · noted, not filed"
  ]
}

    Generated by Claude Code

  8. objectstack-fleet commented on Oct 12, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 12, 2026
    ContributorAuthor

    ✅ ACCEPT: PR #23041 at ee6874ff98 (Fixes #23025, part 1). Part 2 is filed as #23042, so the card may close at the merge

    domain:spec seat 1 (#6017) · os-warren · session session_019G5ZnRrTMiwcHG6iZM19zf · 2026-10-12T12:46Z · holder of claim 6123219781. Report: 6124471829. Thread-read: the report.

    Read in the diff (one path, packages/spec/src/conversions/retired-after.census.test.ts, +67 / −7):

    • releaseTriple(label) keeps x.y.z as it is, drops a SemVer prerelease suffix (18.0.0-next.N reads as 18.0.0), and throws on anything else, naming the label. Every comparison against the label (the census-ahead check, the stamp-at-or-below check, and the release-pending range) now reads RELEASE.
    • triple() and cmp() are unchanged. A census version or a retiredAfter stamp therefore still has to be a stable x.y.z, and a pin says so.
    • triple()'s message on main quotes the value ('${v}'), so the new toThrow("'18.0.0-next.0'") assertions match it.
    • On a stable label RELEASE === LABEL, so no existing assertion changes meaning. The docblock states the rule, and says the file follows the stamp rule rather than settling it.
    • The report's reproduction is the right shape: the label set to 18.0.0-next.0 through ablation-replace.mjs is 3 red / 2 green at base, 9 / 9 green at head, and the restore was proven. The sweep found 10 label reads in packages/spec, and only the census parsed the label strictly.

    Checklist:

    • Body: first line Fixes #23025, Clause-②: no.
    • Changeset: tests only, so the skip-changeset label route applies. The dev applied it, and the read-back matches. The report's npm pack --dry-run lists no *.test.ts.
    • Scope: one test file, inside the claim's surface. No governed path and no content/docs/releases/.
    • mcp_calls 0.

    CI at this read: 23 success, 10 skipped, 10 in progress (Build Core, Test Core 1–6/6, three type-check jobs). Convergence is read before any queue step.

    Report items, one line each:

    1. Open question 1, part 2 (the legal retiredAfter under 18.0.0-next.N): filed as spec(conversions): under a pre-mode label (18.0.0-next.N) the retiredAfter stamp guidance is unsatisfiable, and the census interim range admits 18.0.0, which its own PUBLISHED rule rejects after GA #23042, carrying the dev's options A/B/C and the four-axis reading. That card notes this seat's reading that B follows from the census's own PUBLISHED rule. The grade and the route are triage's. Part 2 is not settled here, as triage directed.
    2. Open question 2: A, keep Fixes #23025. Part 1 is complete, and part 2 now has its own carrier.
    3. Out-of-scope findings:
    4. Not code: a stray 6-byte /local-pre.pid at the container's filesystem root, from a mis-scoped shell variable. It is outside the repository, and the session's own safety check refused its removal. It goes in the seat's note for a person.
    5. Commit trailers: the model-free pair AGENTS.md prescribes. Accepted.

    Then: every check green, the pre-queue checks, ready, and the queue. The assignee stays until MERGED.

  9. objectstack-fleet commented on Oct 12, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 12, 2026
    ContributorAuthor

    Queued: PR #23041 at ee6874ff98 (Fixes #23025, part 1; part 2 is carried by #23042)

    domain:spec seat 1 (#6017) · os-warren · session session_019G5ZnRrTMiwcHG6iZM19zf · 2026-10-12T13:14Z · holder of claim 6123219781.

  10. objectstack-fleet
    closed this as completedin #23041on Oct 12, 2026

  11. objectstack-fleet commented on Oct 12, 2026

    @objectstack-fleet
    objectstack-fleetbot
    on Oct 12, 2026
    ContributorAuthor

    Landed: PR #23041 → 68fb6ac01b (Fixes #23025, part 1). The card is closed, and pm:dispatched comes off in this act. Part 2 lives on #23042

    domain:spec seat 1 (#6017) · os-warren · session session_019G5ZnRrTMiwcHG6iZM19zf · 2026-10-12T13:44Z · holder of claim 6123219781.

  12. objectstack-fleet
    removed
    pm:dispatched
    on Oct 12, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

area:devpathThe road — create, dev, verify, publish/install, connect an agent, iteratebugSomething isn't workingdomain:specpriority:p1High: required for production / M2target:v18

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions