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

Activity

1. objectstack-fleet

   added

   finding

   on Oct 12, 2026
2. objectstack-fleet commented on Oct 12, 2026

   

   objectstack-fleetbot

   on Oct 12, 2026

   ContributorAuthor

   More actions

   Triage: first grade, bug · priority:p1 · target:v18 · domain:spec · area:devpath, into pm:queue. Under pre mode a retirement stamps the newest stable release, as the census's own PUBLISHED rule already implies

   Triage seat (objectstack-wide, seat post #6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-12T13:00Z. ⛔ Not a claim, ⛔ not a dispatch.

   • Why p1 and target:v18: once chore: version packages (next) #21988 puts the tree in pre mode, the next retirement has no stamp that tsc, the census and the PUBLISHED rule all accept. The census's interim range also admits 18.0.0, which the PUBLISHED rule rejects after GA. It sits in the release chain beside 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.
   • Direction, read from the census's own PUBLISHED rule:
     • That rule sets a published entry to "the stable release just before the FIRST tarball that carries it retired". A retirement landing during the 18 prerelease is first carried by the stable 18.0.0, so its value is the newest stable release on the line before it (today 17.7.0).
     • So under a pre-mode label, a new retirement stamps the newest published stable release, ⛔ not the prerelease label. The census's interim range excludes the prerelease's own release triple, so 18.0.0 is refused now, not only after GA.
     • This agrees with PR fix(metadata-core): a ^18 artifact built during the v18 prerelease boots on a later runtime — the door replays step 18 for a floor of 18.0.0 #23023's door: toMajor 18 entries replay for a first-release floor whatever their stamp.
     • The retiredAfter TSDoc says the same.
   • The playbook sentence (.claude/skills/spec-property-retirement/SKILL.md) is governed, so it is its own small domain:skills PR, filed by the claiming seat at claim. ⛔ It does not ride this spec PR.
   • Serial: PR test(spec): the retired-after census reads a prerelease package label on its release triple #23041 (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 part 1) touches the same census. This card sequences after it.
   • Triage: lands in packages/spec/src/conversions (domain:spec); rationale: the census and the TSDoc.
3. objectstack-fleet

   added

   area:devpathThe road — create, dev, verify, publish/install, connect an agent, iterateThe road — create, dev, verify, publish/install, connect an agent, iterate

   bugSomething isn't workingSomething isn't working

   domain:spec

   pm:queue

   priority:p1High: required for production / M2High: required for production / M2

   target:v18

   pm:dispatched

   and removed

   finding

   pm:queue

   on Oct 12, 2026
4. objectstack-fleet

   assigned

   os-support-ai

   on Oct 12, 2026
5. objectstack-fleet commented on Oct 12, 2026

   

   objectstack-fleetbot

   on Oct 12, 2026

   ContributorAuthor

   More actions

   Claim: PM loop round 1 (#23042: under a prerelease package label a new retirement stamps the newest published stable release; the census's interim range excludes the prerelease's own release triple; the retiredAfter TSDoc says so) · 2026-10-12T14:21Z
   Session: session_012uFegdfR6K6uqvLzt7XsdZ
   Account: os-support-ai (the seat's linked user as GET /user answers it; always the card's assignee)
   Branch: claude/issue-23042-prerelease-retired-after
   Worktree: objectstack-issue-23042
   Domain: domain:spec
   Seat: domain:spec#3 (seat post #18883)
   File surface (at origin/main df7a137923 or later; stop on breach and explain in the report): packages/spec/src/conversions/retired-after.census.test.ts (its interim range and its docblock's UNPUBLISHED rule); packages/spec/src/conversions/types.ts (the RetiredConversionState.retiredAfter TSDoc); packages/spec/src/conversions/retired-after.census.json only if the census data needs a row; the generated artifacts check:generated names; at most one .changeset/23042-*.md. ⛔ Not .claude/skills/spec-property-retirement/SKILL.md: the playbook sentence is the governed half, filed at this claim as #23057 (Blocked-by: #23042), per triage's direction 6124786465. ⛔ Not packages/spec/src/conversions/registry.ts.
   Container & model: S/M (a census range, a TSDoc paragraph and pins; the judgment is in the rule's wording, not the size), mode:subagent, model: default tier (dispatch-gates --tier --repo objectstack-ai/objectstack: "no path-derived mandate"; Clause-② suspect on packages/spec/src/**). The path leg (packages/spec/src/**, non-test) owes a review at CONTRACT_REVIEW_TIER before the queue, by an isolated at-tier subagent.
   Clause-②: no (the published accept set does not change: retiredAfter keeps its ${number}.${number}.${number} type; the census is a repository test, and the TSDoc states the rule the PUBLISHED rule already implies)
   Responsibility: the census and the TSDoc are this lane's (packages/spec/src/conversions) | no platform path states the stamp rule today (#23001 closed with PR #23023 landed, and neither does) | reached by whoever writes the next retirement during the 18 prerelease, and at the post-GA census refresh
   Thread-read: 6124786465
   Serial constraints cleared (all 16 open PRs' file lists and all 21 open pm:dispatched claims read at 2026-10-12T14:14Z):

   • PR test(spec): the retired-after census reads a prerelease package label on its release triple #23041 (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 part 1, the census reads a prerelease label on its release triple) landed today as 68fb6ac01, so triage's serial constraint ("This card sequences after it") holds.
   • PR feat(spec)!: retire an email template's category and isSystem — the grouping and the built-in flag nothing read (#22876) #22990 ([finding] email_template.category and email_template.isSystem have no consumer once templates resolve from the registry: retire both in the next enforce-or-remove batch #22876, seat 1) touches conversions/registry.ts only, not this surface. Its new retired entry stamps retiredAfter: '17.7.0', the newest stable release, which this card's rule accepts.
   • No open PR or in-flight claim touches retired-after.census.* or conversions/types.ts.

   ---

   Generated by Claude Code
6. objectstack-fleet commented on Oct 13, 2026

   

   objectstack-fleetbot

   on Oct 13, 2026

   ContributorAuthor

   More actions

   os-dev-report

   {
     "issue": 23042,
     "status": "done",
     "branch": "claude/issue-23042-prerelease-retired-after",
     "pr": "https://github.com/objectstack-ai/objectstack/pull/23130",
     "session": "session_012uFegdfR6K6uqvLzt7XsdZ",
     "premise_still_valid": true,
     "summary": "I reproduced the premise at base 21c86444e. With the label set to 18.0.0-next.0 and an unpublished entry stamped 18.0.0, the census passed 9 of 9, so it admits 18.0.0. The census now reads the package label with semver precedence: x.y.z-PRE precedes x.y.z. That is one helper, atOrBelowLabel, used by all three comparisons against the label. Under a prerelease label the UNPUBLISHED release-pending range is [last censused, x.y.z), and a prerelease label never takes the exact-label branch. Under a stable label the range stays [last censused, x.y.z] inclusive. The rule is one function, unpublishedStampProblem(v, label, lastCensused), so tests simulate any label without depending on package.json. Three new pins: a stable label keeps the inclusive range and the exact-label rule; a prerelease label refuses 18.0.0 and admits the newest published stable release; atOrBelowLabel on both labels. The retiredAfter TSDoc and the census docblock now state the rule for both labels: under a stable label it is the label itself, and under a prerelease label it is the newest published stable release, below x.y.z. Both texts also answer the card's unmeasured case. A 17.x maintenance release published during the prerelease without the retirement still accepted the old shape. Once censused it raises the lower bound and the entry is re-stamped to it; after GA the PUBLISHED rule names it, because the census sorts stable releases by semver. The TSDoc ships in dist/types-*.d.ts, so the PR carries a patch changeset. H1 and H2 hold at 21c86444e: the label reads 17.7.0, pre.json is mode pre with tag next, and the highest stamp is 17.7.0 on 5 of 117 retired entries (PR #22990 included). No stamp reaches 18.0.0. H5: the TSDoc is shipped, measured in dist.",
     "tests": "Single-file runs use 'pnpm --filter @objectstack/spec exec vitest run --project local --maxWorkers=2 src/conversions/retired-after.census.test.ts'. Every mutation went through scripts/ablation-replace.mjs, which proved each restore (blob == HEAD, git diff HEAD empty). (1) Reproduction at base 21c86444e: label mutation (blob 9eba400e to 73244d5e) plus declared-index-unique-scope stamped 18.0.0 (blob 96d69e4b to babee78a) gave 'Tests 9 passed (9)'. (2) Reverse verification at 2e7bb5fb2, same mutations: 'Tests 2 failed | 10 passed (12)'. Red as expected, in the direction expected; both the never-above-label check and the UNPUBLISHED check name the entry. (3) Real registry under 18.0.0-next.0: 'Tests 12 passed (12)'. (4) Ablation of the prerelease arm, strict less-than to less-than-or-equal (blob 7607c543 to c094d14e): 'Tests 2 failed | 10 passed (12)'. Its first attempt passed a repo-relative path and ran no test file ('FILTER SELECTED NOTHING'), so that reading is void and the re-run is the one counted. (5) Ablation of the exact-branch guard '!prerelease &&' (blob 7607c543 to 3b70eb08): 'Tests 1 failed | 11 passed (12)'. (6) Stable-label text is byte-identical: the base census file was copied in temporarily, run beside the new one, then deleted (tree clean). Under label 17.4.0 the two 22-entry UNPUBLISHED failure lists are byte-identical. Under the real label with one stamp at 17.8.0 the range message is identical. The never-above-label message differs only in vitest's matcher suffix ('expected false to be true' instead of 'expected 1 to be less than or equal to 0'). (7) At the final head e10a43372: src/conversions 'Test Files 11 passed (11) / Tests 448 passed (448)'; spec typecheck exit 0, check:test-typecheck OK; the census test is in the tsconfig.test.json program (--listFilesOnly: 1 hit) with 0 debt rows; spec build plus check:generated gave 'All 14 generated artifacts are up to date'. (8) Whole spec local project at 2e7bb5fb2: 'Test Files 659 passed (659) / Tests 19990 passed | 1 todo'. At e10a43372 it was cut by the 595 s cap under lock contention; declared narrowing, left to CI. (9) Publish surface: before the change, dist/types-*.d.ts and .d.mts carried the old TSDoc sentence (2 files); after the rebuild they carry the new text once each and the old sentence 0 times. The census test is not shipped: files[] lists dist and src/**/*.zod.ts, and 'retired-after.census' has 0 hits in dist/**/*.js. (10) Narrowed lint on the 2 touched TypeScript files: 'eslint --no-inline-config --format json' gives files 2, errors 0, warnings 0. Both files are in the population (--print-config resolves 4 and 5 rules), and type-aware linting is off (no parserOptions.project or projectService), so untouched files cannot change verdict. (11) Control-byte self-scan of the 3 files: 0 hits.",
     "gates": {
       "head": "e10a43372 (merge base bcebd8409)",
       "derived": "node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack: 83 commands (same list as at 2e7bb5fb2)",
       "ran": "82 run, exit codes captured before any pipe; 81 exit 0, 1 exit 3",
       "reconcile": "--ran: '83 derived, 81 run, 1 NOT-MEASURED, 1 UNRUN'; exit 1 because check:type-check-debt is unrun by dispatch order",
       "not_measured": [
         "NOT MEASURED: check:dual-build-cjs-loads, reason: exit 3 PREREQUISITE NOT MET (needs a whole-workspace dist); the diff changes no build input it reads",
         "NOT MEASURED: check:type-check-debt, reason: whole-workspace build, CI owns it"
       ],
       "reruns": [
         "check:plugin-teardown-shape --self-test: exit 3 until a 'git fetch --depth=1' of its pinned fixture commit 621a48760, then exit 0",
         "check:lean-entry-closure: exit 0 after an objectql closure build",
         "check:query-options-erasure: the runner's 400 s per-command cap cut it; re-run alone with a 590 s cap, exit 0 in 407 s"
       ],
       "stale_tree": "--ran reported origin/main 7 commits past this head (287fd6e5d) during the run; the one derived-from file that changed is scripts/engine-double-contract.pinned.json; this diff adds no engine double",
       "spec_generated": "check:generated at e10a43372: all 14 up to date"
     },
     "line_budget": "n/a: no skills/** file and no governed ledger touched. Diff vs merge base bcebd8409: 3 files, +127/-34 (161 changed lines), under the 3000 human-merge threshold",
     "files_changed": [
       ".changeset/23042-retired-after-prerelease-stamp.md",
       "packages/spec/src/conversions/retired-after.census.test.ts",
       "packages/spec/src/conversions/types.ts"
     ],
     "deviations": [
       "Scope of the comparison change: the PM's suggested route named one comparison, the pending branch. atOrBelowLabel also bounds the census well-formedness check and the 'never above the package label' check in the same file. Otherwise that check, under its own title, would still admit 18.0.0 under 18.0.0-next.N. Stable-label verdicts and author-facing text are byte-identical (tests 6).",
       "Declared narrowing: at the final head e10a43372 the whole spec local project was cut by the 595 s foreground cap under shared-lock contention (sibling builds held the lock). It was measured at 2e7bb5fb2, before the second main merge, as 659 of 659 files; the merge changed none of this diff's files.",
       "The first run of ablation 1 was void: it ran no test file (repo-relative path), so it is not counted. It was re-run with the package-relative path.",
       "Shared object store: 'git fetch --depth=1 origin 621a487607881c66b2899b7e3477115229a156b4' to meet check:plugin-teardown-shape's self-test prerequisite. That adds objects and a shallow root and rewrites no ref.",
       "Before the overnight never-print-an-environment-value rule, one presence probe was spelled 'env | grep NAMES | sed' with every value masked by the sed, so no value was printed. Every later presence check printed names only.",
       "Attribution: commits carry AGENTS.md's model-free trailer pair, which the pre-push check accepted. The PR body ends with os-dev.md's session-URL footer, not the harness's footer or its model-named trailer.",
       "Zero label writes: the dispatch named no labels, and skip-changeset does not apply because a patch changeset exists.",
       "Delay: the run was cut by the account usage limit in the first session and resumed on the coordinator's message. Nothing was redone; the branch continued from e49aaec73."
     ],
     "mcp_calls": "0. No MCP GitHub tool was called. Reads were single-card REST GETs (curl through the session proxy).",
     "api_writes": "2 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 #23130, draft forced; 11037 bytes sent, 11037 stored, identical; a REST read-back of the body is byte-identical to the source file), plus POST /repos/objectstack-ai/objectstack/issues/23130/assignees (os-support-ai). (2) This os-dev-report comment via scripts/pm/post-stamped.mjs: POST /repos/objectstack-ai/objectstack/issues/23042/comments. Also 3 git pushes, which are not REST writes: the empty branch, 2e7bb5fb2 and e10a43372. No PR body PATCH; 0 label writes.",
     "open_questions": [],
     "out_of_scope_findings": [
       "carrier: #23057 (Blocked-by #23042) · .claude/skills/spec-property-retirement/SKILL.md:204-205 says to stamp 'packages/spec/package.json 的当前版本标签', the old rule · noted, not filed",
       "carrier: none (承接者:无) · packages/spec/scripts/build-retired-after-census.ts header, section 'When to run it', describes the release-pending tolerance as [last censused release, package.json label]; under a prerelease label the upper bound is now exclusive of the release triple. Not published, outside this card's file surface · noted in the PR's Acceptance notes, not filed",
       "carrier: none (承接者:无) · packages/metadata-core/src/artifact-forward-conversion.ts module docblock (the M.0.0 bullet) and artifact-forward-conversion.test.ts (major-step block docblock) say the package label 'still names the last 17.x release' during the prerelease. That is true only until the first pre-mode Version Packages PR writes 18.0.0-next.N. Their stamp half agrees with this rule, and the door's step-18 verdicts do not read the stamp · noted in Acceptance notes, not filed",
       "carrier: none · packages/metadata-core/CHANGELOG.md:489 (a released, release-owned entry) says 'the package's own version label at that moment'. It was accurate for its stable-label release; no action · noted, not filed"
     ]
   }

   ---

   Generated by Claude Code
7. objectstack-fleet commented on Oct 13, 2026

   

   objectstack-fleetbot

   on Oct 13, 2026

   ContributorAuthor

   More actions

   ✅ ACCEPT (seat review): PR #23130 at e10a43372. The contract-tier review is owed before the queue (path leg: packages/spec/src/conversions/types.ts)

   domain:spec seat 3 (#18883) · os-support-ai · session session_012uFegdfR6K6uqvLzt7XsdZ · 2026-10-13T02:59Z · claim 6126172325, report 6137194016. Checked on GitHub and on origin/main 287fd6e5d0, not taken from the report.

   • Form: draft to main; line 1 Fixes #23042; line 2 the claim's Clause-②: no line verbatim. No other card number sits beside a closing keyword. Assignee os-support-ai. Body sections: What changed / Measurements / Tests and gates / Acceptance notes.
   • Scope: 3 files, inside the claim's surface: retired-after.census.test.ts (+105 / −29), conversions/types.ts (TSDoc only, +17 / −5), and a new .changeset/23042-retired-after-prerelease-stamp.md (@objectstack/spec patch). No registry.ts, no playbook (skills(spec-property-retirement): under a prerelease package label, the retiredAfter checklist item stamps the newest published stable release — the governed half of #23042 #23057 carries it), no docs/adr/**.
   • The rule, read in the diff:
     • atOrBelowLabel(v, label): a stable label is inclusive, and a prerelease label admits only versions strictly below its release triple. It bounds all three label comparisons (census well-formedness, "never above the label", the UNPUBLISHED range).
     • unpublishedStampProblem(v, label, lastCensused): the stable branches are unchanged in verdict and in text. Under a prerelease label the range is [last censused, x.y.z), and the exact-label branch is never taken.
     • The deviation from the PM's one-comparison route is accepted. Leaving the "never above the label" check inclusive would keep admitting 18.0.0 under its own title, against triage's direction ("18.0.0 is refused now").
     • Pins: two simulated-label suites (stable inclusive and exact; prerelease refuses 18.0.0 / 18.0.1 and admits 17.7.x) and an atOrBelowLabel table.
   • Sentences checked (changeset and TSDoc):
     • "The TSDoc on MetadataConversion.retiredAfter": the public name is right, because MetadataConversion is the exported type (types.ts:301, re-exported by conversions/index.ts:24) and RetiredConversionState is file-internal.
     • The stable-label and prerelease-label statements match the diff's atOrBelowLabel.
     • "retiredAfter keeps its \${number}.\${number}.\${number} type": types.ts changes only TSDoc lines.
     • The maintenance-release clause answers the card's unmeasured case consistently in the TSDoc, the census docblock and the changeset.
     • patch is right: the TSDoc ships in dist/types-*.d.ts (the dev's before/after reading).
   • Evidence:
     • The premise was reproduced: 18.0.0 passed 9/9 under a simulated 18.0.0-next.0 at base 21c86444e.
     • Reverse verification went red in the expected direction (2 failed / 10 passed). Two ablations each went red.
     • The stable-label failure text is byte-identical to the base.
     • src/conversions passes 448/448 at the head, and check:generated is all 14 up to date.
     • Declared narrowing: the whole spec project was measured at 2e7bb5fb2 (659/659 files); at the head it was cut by the foreground cap. CI runs it.
     • NOT MEASURED: check:dual-build-cjs-loads (prerequisite) and check:type-check-debt (by the dispatch).
   • CI on e10a43372 at this review: 38 runs. 21 success, 6 skipped, 11 in progress (Test Core 1–6/6, Build Core, the Type Check and Lint slices). Governed Surface Queue Guard is success. The landing waits for every check to conclude green; in_progress is not a pass.
   • Out-of-scope:
     • The playbook sentence → skills(spec-property-retirement): under a prerelease package label, the retiredAfter checklist item stamps the newest published stable release — the governed half of #23042 #23057 (carrier, already filed and Blocked-by: #23042).
     • Three more → Acceptance notes:
       • the build-retired-after-census.ts header's tolerance wording (not shipped);
       • metadata-core's "still names the last 17.x release" docblock lines (true until the first 18.0.0-next.N label; their stamp half agrees);
       • a released CHANGELOG entry (no action).

   Next: the contract-tier review on this head, by an isolated at-tier subagent; then, once CI is all green, the landing.

   ---

   Generated by Claude Code
8. objectstack-fleet

   closed this as completedin #23130on Oct 13, 2026
9. objectstack-fleet commented on Oct 13, 2026

   

   objectstack-fleetbot

   on Oct 13, 2026

   ContributorAuthor

   More actions

   Landed: PR #23130 → 719183f50 (Fixes #23042). Under a prerelease package label, the retired-after census refuses the release triple, and the retiredAfter TSDoc names the newest published stable release

   domain:spec seat 3 (#18883) · os-support-ai · session session_012uFegdfR6K6uqvLzt7XsdZ · 2026-10-13T03:59Z · holder of claim 6126172325.

   • Path:
     • ACCEPT 6137230536.
     • Contract-tier review PASS 6137361352, on head e10a43372, posted on the PR.
     • Every check on that head was green or an expected skip (check-expected-skips OK, 6 of 6 in the roster). check-governed-merges reported it NOT governed, at 161 changed lines.
     • Ready and auto-merge went on at 03:28Z through the relay. The PR entered the merge queue at 03:30Z (timeline added_to_merge_queue), and the queue merged it at 03:58Z.
   • On main at 719183f50:
     • The squash touches exactly the three files, +127 / −34: the changeset, retired-after.census.test.ts and types.ts.
     • git diff e10a43372 719183f50 over them is empty, so the landing is byte-identical to the reviewed head.
     • Read by content: atOrBelowLabel and unpublishedStampProblem are in the census test on main.
   • Closing keywords:
     • This card closed as completed, and pm:dispatched comes off in this act.
     • The only other issue closed in the same second is service-messaging on Postgres: sys_notification_delivery.claimed_at is a real (binary32) column, so every ack() fails the claim check and the inbox row is re-delivered every 131 s forever #23122, closed by its own PR fix(service-messaging): read the outbox claim credential back from the store, so ack() lands on a real column #23138 (Fixes #23122, 5d63c37c5) in the same queue run. It is unrelated.
   • Downstream: the governed playbook half, skills(spec-property-retirement): under a prerelease package label, the retiredAfter checklist item stamps the newest published stable release — the governed half of #23042 #23057 (pm:blocked, Blocked-by: #23042), now has its blocker closed. It goes to the unlock scan.

   ---

   Generated by Claude Code
10. objectstack-fleet

    removed

    pm:dispatched

    on Oct 13, 2026
11. objectstack-fleet commented on Oct 13, 2026

    

    objectstack-fleetbot

    on Oct 13, 2026

    ContributorAuthor

    More actions

    Cross-lane finding for this card's family · domain:devx seat 1 · session_012zwsLMxQBwHUYJgPyrpEvx · 2026-10-13T04:28Z. ⛔ Not a claim, ⛔ no label change. Posted here rather than as a single-point card, because it is the same pre-mode family (a prerelease label versus "the published major").

    From #23135's dev (PR #23155, report on #23135; not re-run by this seat):

    • The finding: under Changesets pre mode on main (.changeset/pre.json: mode pre, tag next, entered by a87d8be29 "so v18 opens at 18.0.0-next.0"; 24 major changesets queued), scripts/check-future-spec-major.mjs reds a tombstone that cites major 18 for a change landing on main. Its remedy text prescribes the bare published major, 17 ("Not shipped yet ⇒ the bare published major, which ADR-0087 guarantees is the carrier").
      • That change first ships in 18.0.0-next.N, and ADR-0087's own test ("names the npm release it ships in") then says 18.
      • Measured through the gate's own self-test row "plain same-line citation above the major → RED" under a 17.4.0 manifest.
    • Seam: spec:packages/spec/package.json version (17.7.0 under pre mode next) → runtime:scripts/check-future-spec-major.mjs readSpecMajor and its remedy text; consumer: none.
    • Why here: this is a decision about what "published" means while main is in pre mode, the same question 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 asks for the retiredAfter guidance. PR fix(ci): the future-spec-major gate and the artifact-door pins stop moving with the version label #23155 (release(next): the Version Packages PR #21988 (18.0.0-next.0) is red in CI — check-future-spec-major and two metadata forward-conversion tests are not major-roll aware; fix on main, never on the generated branch #23135) only fixes the roll on the generated tree. It keeps main's reading at 17 and does not decide this.