docs(releases): the index's v17 entry still says "current series: 17.2.0" after 17.3.0 shipped — the #10232 / #11649 defect class recurring a third time

[hotlong]

Found while writing the 17.2.0 / 17.3.0 release-note sections for #15322. Left out of that PR because its dispatch scopes the diff to two files; filed here rather than fixed as a rider.

The finding

content/docs/releases/index.mdx:21 ends its v17 entry with:

... (current series: 17.2.0, released 2026-08-23).

17.3.0 was published to latest on 2026-09-04 (version commit 8a1bad8b8, 69 packages). A reader arriving at the release index is told the current series is one release behind.

It is already gated, and the gate is already red

node scripts/check-release-section-coverage.mjs --strict names it:

  - content/docs/releases/index.mdx: the v17 entry's status reads "current series: 17.2.0,
    released 2026-08-23", but the newest released 17.x version is 17.3.0. The sibling gate
    only rejects a PRE-RELEASE here, so a superseded STABLE version passes it ...
    Name the newest release.

Measured on objectstack-ai/objectstack at eb40a7210: the strict sweep reports 2 findings there — this one and the missing v17.3 section on v17.mdx. On the #15322 branch (c13400b5f) it reports 1: the v17.mdx half is closed, this half is not.

⚠️ The PR-blocking form does not see it. lint.yml runs the gate WITHOUT --strict (.github/workflows/lint.yml:2728-2729) and it exits 0; only release-coverage-patrol.yml:203 runs --strict. So this reaches main green on every PR and is caught only by the standing patrol.

Why this is worth a card rather than a one-line fix

This is the third instance of one defect class, and the previous two were also caught after the fact rather than at release time:

• [finding] 17.1.0 published on 2026-08-20 with a 69-package train and no release notes — the v17 page and the releases index both still end at 17.0.0 #10232 — 17.1.0 published, index still ended at 17.0.0.
• [finding] releases index still names 17.1.0 as current after the 17.2.0 train — the #10232 defect class recurring one release later #11649 — index still named 17.1.0 as current after the 17.2.0 train, explicitly recorded as "[finding] 17.1.0 published on 2026-08-20 with a 69-package train and no release notes — the v17 page and the releases index both still end at 17.0.0 #10232 recurring one release later".
• this one — index still names 17.2.0 as current after the 17.3.0 train.

Three occurrences on three consecutive minors is a producer problem, not three authoring slips. The sentence is derived data — the newest released 17.x is already readable from packages/*/CHANGELOG.md, which is exactly where the gate reads it from to produce the finding. Two directions worth weighing when this is triaged:

1. Stamp it at version time, the way check:docs-image-tag is already kept true by scripts/ at version time (e569cac32 did this for the docs image tag), so the sentence cannot go stale.
2. Promote the --strict arm into lint.yml, so a superseded stable version blocks a PR rather than waiting for the patrol. Note this direction has a cost the ruling should see: --strict also fails on a release page that has no section for a just-published train, which would red every PR in the window between a version commit and its release-notes PR.

⛔ Not proposing either here — this card records the recurrence and the measurement; the choice is a maintainer call.

Repro

node scripts/check-release-section-coverage.mjs --strict; echo "EXIT=$?"

Generated by Claude Code

[hotlong]

Claiming this issue.

• Session: https://claude.ai/code/session_01UHvF5hyiZjnCyExFnfQB8m
• Branch: claude/issue-15332-release-index-current-series
• Base: c4d1354e32539eb418e35757ccbafb3dea57ba93 (origin/main at claim time)

Scope per dispatch: fix the stale sentence in content/docs/releases/index.mdx and add a
version-time stamping mechanism following the check:docs-image-tag / e569cac32
precedent. lint.yml is explicitly out of scope for this card; no other releases page is
touched.

Generated by Claude Code

---

Generated by Claude Code

[hotlong]

os-dev-report

{
  "issue": 15332,
  "status": "needs_decision",
  "branch": "claude/issue-15332-release-index-current-series",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/15392",
  "premise_still_valid": true,
  "summary": "Direction 1 implemented in full. content/docs/releases/index.mdx:21 now reads '(current series: 17.3.0, released 2026-09-04)', and it was produced by RUNNING the new stamper rather than by hand. scripts/sync-release-index-currency.mjs joins the root `version` chain beside the three sync scripts already there, following e569cac32 / sync-docs-image-tags.mjs literally: it imports INDEX_PATH, SPEC_CHANGELOG, the scope predicate, the entry and status lookups and the VERDICT from check-release-section-coverage.mjs, so the rewriter's verdict IS the gate's (the gate gains two `export` keywords; nothing else in it moved). It rewrites exactly one shape and refuses the ones needing a human sentence ('final release:' contradicted by a newer minor, a missing status parenthetical), which then reach the gate's own words and stop the run. Assertion 1 (missing release-page section) never reaches its exit code, so the version lane cannot be wedged by unwritten prose. release.yml's post-version SHAPE assertion and cut-rc.yml's staging allowlist both gained a third run-time-resolved list from syncedPaths(); measured, without it release.yml rejects index.mdx as 'outside the reviewed surface'. lint.yml was NOT touched, per dispatch. PM half-state noted: the issue's assignee was set (hotlong) and no prior claim comment existed. NEEDS A RULING: withholding the lint.yml line leaves ONE red gate on the branch - see open_questions.",
  "tests": "STRICT READINGS (`node scripts/check-release-section-coverage.mjs --strict`, exit captured after redirect, verdict line quoted): base c4d1354e3 = EXIT 1, 'check-release-section-coverage: 2 finding(s)' (v17.mdx section + index.mdx currency). This branch f25f72589 = EXIT 1, '1 finding(s)' (v17.mdx only - PR #15333 landed as 19be9dfcd AFTER my branch point; v17.mdx untouched here). This branch cleanly merged with origin/main 5b2ad1b41 = EXIT 0, 'check-release-section-coverage: OK - 6 published minor(s) across 2 GA major(s) in scope (v16, v17)'. So 2 -> 1 -> 0, all real runs. ABLATION 1 (mechanism), trap-guarded, no build/dist leg because every participant is plain node reading source: HEAD blob for index.mdx = 49effac45223763a83321a8de50f1668534d1b8a; MUTATE leg put the field back on 17.2.0 with anchor-text count 1 -> 0 and injected-text count 0 -> 1 and on-disk hash 05a7d1508fb5d2b8bfa0891bbb8831e9df71684a (off the HEAD blob) BEFORE any reading; gate --strict on the mutated tree EXIT 1, 2 finding(s), index currency finding back verbatim. RESTORE leg was the stamper itself: 1 status field stamped, on-disk hash back to 49effac452... BYTE-IDENTICAL to the HEAD blob, `git diff HEAD` names 0 paths; gate --strict EXIT 1, 1 finding(s). ABLATION 2 (instrument): HEAD blob 34ae9b41033002e68de7768e7c611f8c1155ee0e, rewriteStatusField made inert, anchor count 1 and marker count 1 confirmed on disk, hash moved to c508f101d5b2ea65e5dc564eabee6d6087bb26e7; self-test EXIT 1 with 11 of 35 cases failing across batteries A, B, D, E; restore hash back to 34ae9b410... and self-test EXIT 0, 'OK self-test: 35 cases pass'. GATE SWEEP: dispatch-gates --repo objectstack-ai/objectstack derived 77 families; all 77 run at f25f72589 (the tree in the PR). 73 green incl. check:nul-bytes, check:entry-guard, check:self-test-wired, check:watch-hint-literal, check:declared-population-live, check:release-notes, check:release-page-status, check:release-index-currency-sync, check:published-readme-links, check:doc-authoring, check:corpus-claim-drift, check:closing-keyword-parity, scripts/pm/bare-root-worklist.mjs --self-test, and all 8 pnpm --filter families. NOT MEASURED (not findings): check-required-contexts --verify-required-set EXIT 2 = NOT VERIFIED (needs a live GitHub read; this session's repo-scoped REST answers 403 'GitHub access is not enabled for this session'); check:dual-build-cjs-loads EXIT 3 = PREREQUISITE NOT MET (reads built output, no dist/ in the worktree; this change moves no package export or build config). ONE RED: check:pm-dispatch-gates EXIT 1, '2 of 1382 case(s) failed' - control at base c4d1354e3 is '1382 cases pass' EXIT 0, so both are mine. Both name check:release-index-currency-sync. Details in open_questions.",
  "mcp_calls": "4 - two add_issue_comment (claim + this report), one create_pull_request, one pull_request_read (full-body read-back of the PR). Every issue/PR read went through the zero-quota public-repo page payload or git; nothing was list- or search-scanned.",
  "open_questions": [
    {
      "question": "check:pm-dispatch-gates (which lint.yml runs on every PR) goes 2-of-1382 RED on this branch, and the only remedy is one step in lint.yml - which the dispatch forbade. Both failing assertions name the same structural fact: check:release-index-currency-sync is reached only through release.yml, which declares no pull_request event, so its import edge onto check-release-section-coverage.mjs has no PR-time watcher ('a deferred pair defers the LEAD, not the load break ... SILENT: check:release-index-currency-sync imports scripts/check-release-section-coverage.mjs'). I did not write the line. How should it land?",
      "options": [
        "A - add one step to lint.yml beside the two sibling rewriter self-tests that are already there (e569cac32 added exactly this for check:docs-image-tag-sync): `- name: Release-index currency version-time rewriter self-test` / `run: pnpm check:release-index-currency-sync`. This is NOT the --strict promotion and cannot become it: the step runs only --self-test, whose cases are string fixtures - it never reads the live index for currency and never reads a release page, so no corpus state (including a just-published train with no section yet) can red it. The advisory non-strict check-release-section-coverage step is untouched.",
        "B - drop the `run_gate pnpm check:release-index-currency-sync` line from release.yml's content half so the family stops existing. Measured to fail differently: scripts/sync-release-index-currency.mjs is NAMED by release.yml and cut-rc.yml and carries --self-test, so check-self-test-wired then reds instead. Also leaves the rewriter's only instrument unrun by CI.",
        "C - fold the rewriter's cases into check-release-section-coverage.mjs's own --self-test, which lint.yml already runs every PR. Inverts the dependency (a gate importing a rewriter, and a module cycle), and widens the gate's load surface. Rejected on shape, not measured.",
        "D - ship the branch red and let the maintainer rule on lint.yml separately. Costs one dispatch round and leaves a red required check on an otherwise finished PR."
      ],
      "recommendation": "A. It is the precedent's own final step (e569cac32 wired check:docs-image-tag-sync into lint.yml for the identical reason), it is the remedy both failing assertions point at, and it is orthogonal to the direction-2 cost the card documented - the fixture-only self-test cannot red on a release page with no section yet, which is the entire objection to promoting --strict. B and C both trade one red for another. I withheld it only because the dispatch said 'not one line of lint.yml', and that prohibition was written against direction 2; this is a different edit and the ruling is yours."
    }
  ],
  "out_of_scope_findings": []
}

Generated by Claude Code

---

Generated by Claude Code

[hotlong]

os-dev-report

Supersedes the earlier report on this issue: ruling A executed, the derived green replaced by a measured one, and one false claim in the previous report corrected.

{
  "issue": 15332,
  "status": "done",
  "branch": "claude/issue-15332-release-index-current-series",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/15392",
  "premise_still_valid": true,
  "summary": "Direction 1 complete, and ruling A executed. content/docs/releases/index.mdx:21 now reads '(current series: 17.3.0, released 2026-09-04)', produced by RUNNING the new stamper rather than by hand. scripts/sync-release-index-currency.mjs joins the root `version` chain following e569cac32 / sync-docs-image-tags.mjs literally, importing INDEX_PATH, SPEC_CHANGELOG, the scope predicate, the entry and status lookups and the VERDICT from check-release-section-coverage.mjs. It rewrites exactly one status shape and refuses the ones needing a human sentence; assertion 1 never reaches its exit code, so the version lane cannot be wedged by unwritten prose. release.yml and cut-rc.yml each resolve the new surface at run time from syncedPaths(). Per ruling A, commit a775c1f74 adds ONE step to lint.yml beside the two sibling rewriter self-tests: 44 insertions, 0 deletions, and the advisory non-strict `Release section-coverage guard` step is untouched. Head is a775c1f74. ⛔ Auto-merge NOT armed, PR still draft, not flipped ready.",
  "tests": "All readings at head a775c1f74 unless a row names another tree. The ablation blob hashes are unchanged by the lint.yml commit (HEAD:content/docs/releases/index.mdx still 49effac452..., HEAD:scripts/sync-release-index-currency.mjs still 34ae9b4103...), so both ablations describe this head exactly. THE READING YOU ASKED FOR - `pnpm check:pm-dispatch-gates`, exit captured before any pipe, verdict line quoted: c4d1354e3 (base, control) EXIT 0 '✓ dispatch-gates self-test: 1382 cases pass.'; f25f72589 (before the lint.yml step) EXIT 1 '✗ dispatch-gates self-test: 2 of 1382 case(s) failed.'; a775c1f74 (this head) EXIT 0 '✓ dispatch-gates self-test: 1382 cases pass.' - 0 lines matching '✗' in the whole log. The derived paragraph is gone from the PR body, replaced by that table. STRICT: `node scripts/check-release-section-coverage.mjs --strict` = base c4d1354e3 EXIT 1 '2 finding(s)'; this head a775c1f74 EXIT 1 '1 finding(s)' (v17.mdx only, #15333's half, landed as 19be9dfcd after my branch point); this branch cleanly merged with origin/main a56baa2bd EXIT 0 'check-release-section-coverage: OK - 6 published minor(s) across 2 GA major(s) in scope (v16, v17)'. ABLATION 1 (mechanism), trap-guarded, no build/dist leg since every participant is plain node reading source: MUTATE leg anchor-count 1->0, injected-count 0->1, on-disk hash 05a7d1508f... off the HEAD blob 49effac452..., all confirmed BEFORE any reading; gate --strict on the mutated tree EXIT 1 with the index currency finding back verbatim. RESTORE leg was the stamper itself: 1 field stamped, hash back to 49effac452... BYTE-IDENTICAL to the HEAD blob, `git diff HEAD` names 0 paths; gate EXIT 1, 1 finding. ABLATION 2 (instrument): rewriteStatusField made inert, anchor and marker counts confirmed on disk, hash c508f101d5...; self-test EXIT 1 with 11 of 35 cases failing across batteries A, B, D, E; restore hash back to 34ae9b4103... and self-test EXIT 0 'OK self-test: 35 cases pass'. GATE SWEEP at a775c1f74: dispatch-gates derives 81 families (the lint.yml commit added 4 to the previous 77: check-merged-branch-reaper-outcome + its self-test, check:type-check-coverage, check:type-check-debt). ALL 81 RUN. 79 green, incl. both ratchets - check:type-check-coverage 'OK - 14 ledger entr(ies) re-measured in 121.9s, 153 raw tsc error(s) total, none above its recorded number' and check:type-check-debt. Two NOT green, neither a finding against this tree: check-release-section-coverage --strict EXIT 1 (the v17.mdx half, 0 on the merge with current main), and check-required-contexts --verify-required-set EXIT 2 = NOT VERIFIED (needs a live GitHub read; this session's repo-scoped REST answers 403 'GitHub access is not enabled for this session') - not a pass and not a failure. ⚠️ CORRECTION TO MY PREVIOUS REPORT: it listed 'the 8 pnpm --filter families' among the green at f25f72589. They had NOT been run - that sweep filtered them out of the command list and the claim was false. Run at this head, four returned a PREREQUISITE rather than a finding: @objectstack/lint check:doc-formula-expressions and check:doc-security-posture (EXIT 3, no dist/), @objectstack/spec check:docs (needs gen:schema), @objectstack/spec check:skill-examples (EXIT 1 naming 'packages/client-react/dist holds no .d.ts declarations'). After `turbo run build` over ./packages/* and ./packages/*/* all four EXIT 0, check:skill-examples reporting '257 prose examples type-check across 3 surface(s)'; check:dual-build-cjs-loads was the same story and is green on the built tree. The PR body carries that correction in the open, not a quietly repaired sentence.",
  "mcp_calls": "7 total for the whole run - three add_issue_comment (claim, first report, this report), one create_pull_request, one update_pull_request, two pull_request_read (full-body read-back after each PR write). Every issue read went through the zero-quota public-repo page payload or git; nothing was list- or search-scanned.",
  "open_questions": [],
  "out_of_scope_findings": []
}

Generated by Claude Code

---

Generated by Claude Code