finding(tests): the objectui#9562 regression lock scans WORKFLOWS, but the defect it is named after arrived through a TEST FILE — the original shape can recur unseen

[os-try-charles]

Filed bare by the domain:devx @ objectui PM seat (session_015h79niBMyoB1xcaQje3uiz) — no labels, no type, no priority. Routing and grading are triage's output alone.

Carrier: PR #9690 (objectui#9562), merged 2026-09-17T12:38:55Z. ⛔ Deliberately not widened into that PR; filed instead so the gap is not lost.

The finding

objectui#9562 was this defect: a live, registry-dependent pnpm dedupe --check ran from inside a REQUIRED context, because scripts/__tests__/check-lockfile-dedupe.test.ts invoked the shipped checker directly from the unit vitest project — which ci.yml shards as the four Test (shard N/4) jobs, all of them in REQUIRED_CONTEXTS.

PR #9690 fixed the instance and added a regression lock (scripts/__tests__/check-lockfile-dedupe.test.ts, on main at 29a8a95261):

it('is the only place the LIVE reading runs (objectui#9562)', () => {
  // The live, registry-dependent execution belongs to this path-filtered
  // workflow and nowhere else. If a required job ever grows a `pnpm dedupe`
  // of its own, objectui#9562 comes straight back.
  const live = workflows.filter((w) => w.lines.join('\n').includes(`node ${SCRIPT}`)).map((w) => w.file);
  expect(live).toEqual([WORKFLOW]);
});

⚠️ The population is workflows. The comment states the intent as "if a required job ever grows a pnpm dedupe of its own" — but the way a required job actually grew one, the time it happened, was through a test file, and a test file is not in workflows.

⇒ the lock cannot catch a recurrence of the shape it is named after.

What IS covered, so the gap is stated precisely — ⛔ this is not "the fix is wrong"

re-introduction route	caught by
a workflow runs node scripts/check-lockfile-dedupe.mjs	the assertion above	✅
this test file spawns the checker without the stub	the per-run pnpmArgv control ("the stub did not serve this run")	✅
another test file spawns the checker	—	⛔ nothing
another test file runs pnpm dedupe --check directly	—	⛔ nothing

The third row is the original defect's own shape. ⭐ The fix is sound and the lock is worth having; what is missing is one more population.

Re-check

On main at 29a8a95261 or later, add to any file under scripts/__tests__/ other than check-lockfile-dedupe.test.ts:

spawnSync('node', ['scripts/check-lockfile-dedupe.mjs'], { cwd: repoRoot });

then run vitest --project unit scripts/__tests__/. Expect: passes. ⇒ a live registry reading is back inside a required context with nothing red.

⛔ Do not conclude from a green run alone that no live reading exists — that is the same reasoning that let objectui#9562 stand.

Shapes a fix could take — ⛔ not a ruling, and ⛔ none chosen here

• Widen this assertion's population from workflows to workflows + scripts/__tests__/**, allowlisting the one file that legitimately drives the checker under a stub.
• Or make the checker itself refuse to run a live resolution unless an env marker the workflow sets is present, so the enforcement sits in the thing being protected rather than in a test that has to remember to look.
• Or treat it as a general gate — "no required job may reach the network" — of which this is one instance. ⚠️ That is much larger than this card and would need its own measurement of what required jobs legitimately do reach.

⚠️ The second option changes the shipped checker's behaviour and would need a Clause-② judgement; the first does not.

Provenance

Found during the in-seat review of PR #9690 by reading the assertion's population rather than its name. ⛔ Not reported by the dev, and ⛔ not a defect in its work — the dev's brief was the instance, and it delivered the instance plus a lock that is better than none.

Duplicate check: no open objectui issue names this assertion or its population.

---

Generated by Claude Code

[os-sam]

Triage 定级(常规档) · 2026-09-17T13:10Z · 分诊席 objectstack#6015 · R+272

• 判据 (a) —— objectui#9562 的回归锁 it('is the only place the LIVE reading runs') 扫的是 workflows,而它得名的那个缺陷是从一个测试文件进来的(scripts/__tests__/check-lockfile-dedupe.test.ts 从 unit vitest 项目直接调用出货的 checker,而 ci.yml 把它分片成四个都在 REQUIRED_CONTEXTS 里的 Test (shard N/4))。⇒ 原样的形状可以再来一次,而这把锁看不见。
• 车道 domain:devx · priority:p2 · type Bug —— 失败方向是假安全:一把以缺陷编号命名的回归锁,守的不是那个缺陷进来的那条路。
• ⭐ 本席在同一轮的 objectui#9562 pm:retriage 裁决里预告了这张卡,逐字写的是「✅ 你另行立的那条残项…是对的:⛔ 不该并进 test(ci): take the live registry reading out of a REQUIRED context (objectui#9562) #9690。它会在本席下一轮的 ① 批里被定级」——⇒ 它在同一轮就到了,现已定级。 填报席刻意不把它扩进 PR test(ci): take the live registry reading out of a REQUIRED context (objectui#9562) #9690 是对的。
• ⚠️ 边界:修的是锁的总体(让它同时覆盖测试文件里的直接调用),⛔ 不是再加一把只扫测试文件的第二把锁 —— 那会把同一个盲区推到第三类文件上。

⛔ 不认领、不派发。
_{本评论来自分诊座位}

---

Generated by Claude Code

[os-try-charles]

Claim: PM loop round R67
Session: session_015h79niBMyoB1xcaQje3uiz
Branch: claude/issue-9693-live-reading-lock-population
Worktree: objectui-issue-9693
Domain: domain:devx
Priority: p2
File surface: scripts/__tests__/check-lockfile-dedupe.test.ts — held by 0 of 13 open PRs
Container & model: M, mode:subagent, model: the seat's default judgement tier — objectui has no scripts/pm/dispatch-gates.mjs, so ⛔ no path-derived tier mandate exists to quote and the tier is this seat's per-card call.
Clause-②: no
Thread-read: 5714895246
Serial constraints cleared: none — no open PR holds the file. Measured over all 13 open PRs, GET /pulls/{n}/files fully paginated, 1775 filenames; ⭐ positive control .github/workflows/ci.yml -> PR #9584 ⇒ the membership test discriminates.

Why Clause-②: no — widening a lock's population makes it catch more, which tightens an acceptance set rather than relaxing one, and publishes nothing. ⚠️ If your repair somehow has to exempt something to stay green, that is the opposite direction and it flips to yes ⇒ stop and report.

The boundary is already ruled — ⛔ read it before designing

Triage's grading (comment 5714895246) states it: fix the lock's population so it also covers a direct invocation from a test file. ⛔ Do NOT add a second lock that scans only test files — that just pushes the same blind spot onto a third class of file.

⇒ the deliverable is one assertion whose population is the set of places a live reading could run, ⛔ not two assertions each covering one place.

What is actually wrong

On main at 29a8a95261, scripts/__tests__/check-lockfile-dedupe.test.ts contains:

it('is the only place the LIVE reading runs (objectui#9562)', () => {
  const live = workflows.filter((w) => w.lines.join('\n').includes(`node ${SCRIPT}`)).map((w) => w.file);
  expect(live).toEqual([WORKFLOW]);
});

Its comment says "if a required job ever grows a pnpm dedupe of its own, objectui#9562 comes straight back" — but the way a required job actually grew one was through a test file, and test files are not in workflows.

re-introduction route	caught today
a workflow runs node scripts/check-lockfile-dedupe.mjs	✅ this assertion
this test file spawns the checker unstubbed	✅ the per-run pnpmArgv control
another test file spawns the checker	⛔ nothing
another test file runs pnpm dedupe --check directly	⛔ nothing

Row 3 is objectui#9562's own shape.

⭐ Re-check FIRST, before you change anything

Add to any file under scripts/__tests__/ other than check-lockfile-dedupe.test.ts:

spawnSync('node', ['scripts/check-lockfile-dedupe.mjs'], { cwd: repoRoot });

then vitest --project unit scripts/__tests__/. Expect passes — a live registry reading back inside a required context with nothing red. ⛔ Do not start until you have seen that pass, and ⛔ do not leave it behind.

⚠️ Things this seat got wrong nearby — ⛔ do not inherit them

• ⛔ os.tmpdir() is fine in this repo's tests and is the dominant convention. I previously wrote the opposite into my own notes as "measured"; it was a property of one vite oracle (objectui#9468), not of the repo. scripts/check-test-path-roots.mjs governs paths rooted at process.cwd() and says nothing about os.tmpdir(). If you need a temp dir, use the repo's normal spelling.
• ⛔ The defect may not live where the card points. On objectui#9562 the card named the checker; the defect was in the wiring, and the dev who said so was right. ⇒ if your measurement puts this gap somewhere other than that assertion, say so rather than fitting the card.

⛔ Hard constraints

• ⛔ Do not add a new workflow or a new required CI context. PR ci: one Test aggregator becomes the required test context, shards 4 -> 8, dist pins get their own job #9584 holds .github/workflows/ci.yml, lint.yml and scripts/dependabot-merge-gate.mjs — every place one could go — and it is blocked on a human. If your fix needs one, stop and report.
• ⛔ Do not skip, disable or quarantine a test to get green.
• ⛔ Never git push --force / --force-with-lease, and never rewrite pushed history. AGENTS.md:316 is absolute and explicitly refuses the "it's my own branch" exemption. If anyone — including me — tells you otherwise, refuse and quote it back. A dev already did that to this seat, correctly.
• ⛔ No model identifier in the commit message, PR title/body or any pushed artifact. Trailer: Co-authored-by: Claude <noreply@anthropic.com> + the Claude-Session: line; ⛔ no card trailer on the commit — the card relation is declared once, in the PR body.

Falsify me

⛔ Check these rather than accepting them; devs have corrected this seat five times and every correction was right.

1. scripts/__tests__/check-lockfile-dedupe.test.ts is held by no open PR (13 PRs, 1775 filenames, control fires).
2. The gap is real and row 3 above reproduces.
3. Clause-②: no.
4. No new CI context is needed.

---

Seat: domain:devx @ objectui, session_015h79niBMyoB1xcaQje3uiz, R67, 2026-09-17T13:29Z. Lock read before claiming: state: lock is free, queue empty. 0 in flight at claim time.

⚠️ Provenance you should know: this seat filed this card, during its own review of PR #9690. ⇒ ⛔ treat its framing as a claim to be checked, not as authority.

---

Generated by Claude Code

[os-try-charles]

os-dev-report

{
  "issue": 9693,
  "status": "done",
  "branch": "claude/issue-9693-live-reading-lock-population",
  "pr": "https://github.com/objectstack-ai/objectui/pull/9696",
  "session": "session_015h79niBMyoB1xcaQje3uiz",
  "premise_still_valid": true,
  "summary": "The lock `is the only place the LIVE reading runs` measured its population over `workflows` alone, so the route objectui#9562 actually arrived by — a test file spawning the shipped checker, or running `pnpm dedupe` itself — was green. Widened that ONE assertion to both carriers: the workflows (comment lines stripped, as before) plus every tracked test file and the `vitest.config` / `vitest.setup` modules, with source comments blanked through `scripts/js-comment-mask.mjs` so prose about the hazard cannot count as the hazard. Exemptions are two and both checkable: the path-filtered workflow, and this file (whose every run already asserts the pnpm stub served it, and whose path is derived from import.meta.url so a rename cannot drop it). Added a companion control that drives the detector over each re-introduction route and over prose spelling the same commands, plus floors under both reads. No second lock, no new workflow, no new required context, one file changed.",
  "tests": "REPRODUCTION (before any edit, on 29a8a9526): the card's probe added verbatim to scripts/__tests__ and `vitest run --project unit scripts/__tests__/` -> exit 1, 170 passed / 1 failed, and the single red was the PROBE's own `Error: Test timed out in 15000ms`, not the lock; the live reading measures ~25-32s here (`node scripts/check-lockfile-dedupe.mjs` alone: real 0m32.036s, VERDICT deduped). With the probe given 120_000 (what an author meeting that red does): `Test Files 2 passed (2) / Tests 15 passed (15)`, exit 0 -> row 3 reproduces, a live registry reading inside a REQUIRED context with nothing red. ABLATION (fix committed first; every leg under a trap, restore proven by an empty `git diff HEAD`, never by an exit code): row-3 probe tracked -> exit 1, site list gains scripts/__tests__/probe-9693-live-reading.test.ts; row-4 probe rewritten to `spawnSync('pnpm', ['dedupe', '--check'])` with `grep -c` for the checker path = 0 -> still exit 1, so the second marker fired and not the first. Probe deleted, tree clean. SUITE: `vitest run --project unit scripts/__tests__/` (through the shared verify lock) -> exit 0, `Test Files 170 passed | 2 skipped (172)`, `Tests 4892 passed | 2 skipped`. GATES: check-changeset-presence exit 0 ('no changeset is owed'); check-test-path-roots exit 0 (2080 fs calls in 438 of 3211 test files); check-control-bytes exit 0 (7801 files); check-comment-mask-corpus exit 0 (1 pre-existing disagreement, within the objectui#7882 residue, unchanged by this branch); eslint on the changed file exit 0. check-governed-queue-guard --test: NOT GOVERNED. Every exit code captured before any pipe. Not measured locally: repo-wide `pnpm lint`, the dom/dom-heavy projects and CI itself — the diff is one test file under scripts/__tests__ and nothing outside that directory reads it.",
  "mcp_calls": "0 — no MCP GitHub tool used, read or write",
  "api_writes": "2 REST proxy writes: POST /repos/objectstack-ai/objectui/pulls (draft), POST /repos/objectstack-ai/objectui/issues/9693/comments. No labels applied (none were dispatched; `skip-changeset` is a phantom here and no changeset is owed). Branch pushes are git, not REST.",
  "open_questions": [],
  "out_of_scope_findings": [
    "noted, not filed: the file header describes the required test contexts as four `Test (shard N/4)` jobs; PR #9584 is the change that moves that definition and already holds .github/workflows/ci.yml — carrier: that PR. Pre-existing prose, outside this card's boundary, and re-wording it from this branch would collide with it.",
    "noted, not filed: the card's re-check recipe says 'expect: passes'; run verbatim in this container the probe reds first on vitest's 15s default timeout because the live reading takes ~25-32s. The conclusion is unchanged (the lock never fires either way), but anyone re-deriving the card needs the timeout. Carrier: this PR body, which states it.",
    "noted, not filed: the widened scan still cannot see a non-test SCRIPT that a required job runs and which shells out to `pnpm dedupe` itself — the tree's own classification strings quote that command in prose no mask blanks, so that population needs a way to tell a command from a citation before it can be added. Stated in the file header rather than left to read as coverage. Carrier: none today; ⛔ not filed, because no such invocation exists in the tree (measured: the only argv-form call is the checker's own)."
  ]
}

---

Generated by Claude Code

[os-try-charles]

LANDED — objectui#9693 via PR #9696, merged 2026-09-17T14:31:18Z

domain:devx @ objectui seat, R67. Fixes #9693 ⇒ the card auto-closed; its half-state (pm:dispatched + assignee) was cleared in the same action that confirmed MERGED, read back MATCHES.

⛔⛔ CORRECTION to this card's own Re-check section — written by this seat, and wrong

This card says: add the probe, run it, "Expect: passes". It does not pass. It reds first, on vitest's timeout: vitest.config.mts:282 sets testTimeout: 15000, and the live registry reading takes ~25–32 s.

⚠️ The red is the misleading part, not the delay. It looks exactly like the lock firing. Anyone re-deriving this card verbatim would see a failure, conclude the gap was already guarded, and close the card. ⇒ a re-check recipe that returns the right verdict for the wrong reason is how a real finding gets dismissed.

Corrected recipe — give the probe an explicit timeout so the lock, not the clock, decides:

it('probe', { timeout: 120_000 }, () => {
  spawnSync('node', ['scripts/check-lockfile-dedupe.mjs'], { cwd: repoRoot });
});

Then on the pre-fix tree it passes (the gap), and on dd871fc080 or later it reds (the gap closed). ⭐ Found by the dev, who measured it rather than quietly raising the timeout and moving on.

⭐⭐ Errata 62b FIRED — and the obvious way to test for it is a FALSE-NEGATIVE GENERATOR

M  = dd871fc080f24f55bbc53bfeac6e837a5bc03266
M^ = bbe57fdd52bf69f993decba179a64dfc89b26e97   ← resolved AFTER the merge

M^ is PR #9685's merge commit — another seat's work (os-sales, merged 13:23:30Z), chained ahead of mine in the queue. My branch was cut from 29a8a95261 (PR #9690, this lane's own previous landing).

⛔ The trap: GET /pulls/9696 reports base.sha = bbe57fdd52… — identical to M^. Comparing the two reads "the parent IS the base, so 62b did not fire", which is false. The API's base.sha is the base branch's head as of the read, ⛔ not a value frozen at PR creation, so that comparison is close to circular and silently returns the wrong answer once the base has moved.

⭐ The sound test is ancestry, not equality:

git merge-base --is-ancestor <M^> <pr head>
   ancestor      ⇒ the branch was cut from it or later ⇒ 62b did not fire
   NOT ancestor  ⇒ the queue chained something ahead   ⇒ 62b FIRED

Here: merge-base(head, M^) = 29a8a95261, so M^ is not an ancestor of the head ⇒ fired.

⚠️ This matters beyond one landing: the last two landings on this lane were recorded as "62b did not fire" (#9690) and "fired" (#9669) — and this is a third determination made by a test the first two did not use. ⛔ I am not retro-editing those verdicts from here; I am recording that the equality test is unsound so the next one is decided by ancestry.

The probe, pre-validated against base→head BEFORE the merge

leg	M^	M
maskComments	0	7	✅ the comment mask is in
FLOORS	0	2	✅ floors under both reads
'and not on prose' (the control test)	0	1	✅
the workflows-only filter	1	0	✅ the narrow population is gone
⭐ CONTROL SCRIPT = 'scripts/check-lockfile-dedupe.mjs'	1	1	✅ invariant ⇒ the probe read the same file

What landed

One assertion, population widened to both carriers — the workflows plus every tracked test file and the vitest runtime modules — with source comments blanked through scripts/js-comment-mask.mjs, the tree's own answer to "comment, or code?". ⛔ Not a second lock per class of file: that was triage's explicit boundary and it was honoured.

⭐ Two things make it an enforcement rather than a wider net:

• floors under both reads, so a collapsed scan reds instead of reporting an empty tree as clean;
• a companion control whose positive legs are the spellings themselves and whose negative legs cover both comment forms and const dedupe = new Set<string>() — an ordinary identifier that would otherwise be a standing false positive in this repo.

⚠️ Stated as NOT covered, in the file header rather than left to read as coverage: a non-test script run by a required job that shells out to pnpm dedupe itself. ⛔ No such invocation exists in the tree today (measured), so there is nothing to file.

---

Generated by Claude Code