[objectstack-fleet]

Requested by Elon · project thread

Part of #25920 — slice 3 of the unwired-tooling cleanup (slices 1 and 2: #26004, #26024). The card stays open.

Summary

Deletions only: 3 files, −1,147 lines, on main bbf7061.

• ui/scripts/page-key-read-census.mjs + its test + the census:page-key-reads alias → delete. It was the instrument a renderer-vs-PageSchema page-key gate would have read; that card (objectui#9438) was closed as not planned on 2026-09-22. No gate or workflow reads it; nothing but its own test and the alias names it.

Scope changes found while preparing this slice

Two scripts the #26004 table listed as delete candidates are not dead and move to the maintainer list instead:

• tsconfig-test-parity-census.mjs — five test comments in ui/packages/plugin-designer and one in ui/packages/fields/tsconfig.test.json deliberately tell readers to read the lib level off pnpm census:tsconfig-test-parity instead of restating it (objectui#9513). A live hand-run instrument.
• dollar-dialect-alias-census.mjs — deleting its test turns two cells of ui/scripts/__tests__/markdown-test-inputs.test.ts red ("EVERY tracked markdown document is an input", objectui#9096): its trackedFiles() walk is the read that makes that class total, so removing it would change which PRs the CI markdown exclusions let skip the test shards. That is a CI behaviour change, not cleanup.
• Likewise check-component-surface-parity.mjs: ui/packages/types/src/navigation.ts and ui/packages/plugin-map/src/index.tsx point readers at pnpm check:component-surface-parity.

Remaining delete work: changeset-polarity-census (its card objectui#9727 closed after the rebuild), split in two PRs to stay under the 3,000-line threshold. Then the mechanical tsconfig dedupe.

Verification (head 714e17e)

• Exact-name git grep across the whole tree for the script, its test and the alias: no hits outside the deleted files.
• ui suites that read the scripts inventory and manifest (markdown-test-inputs, package-scripts-vitest-projects, check-unreferenced-sources, ci-cd-pipeline-doc, vitest-invocation-guard): 5 files, 192 tests passed. check:unreferenced-sources and check:test-path-roots exit 0.
• dispatch-gates --commands families: per-family exit codes in the report on the card.
• No changeset: private ui tooling and the private ui root manifest only.

---

Generated by Claude Code

🤖 Generated with Claude Code

https://claude.ai/code/session_012eXtKYmXEnCc1GExP4go12

[objectstack-fleet]

席位核对 — ACCEPT at 714e17ee

Seat review of #26057, slice 3 of #25920 (devx seat #6023).

• Deletions only: 3 files, −1,147. Not governed. Under the human-merge threshold.
• page-key-read-census is a report-only census (its header says "Exit: 0 whatever it finds"). Its card, objectui#9438, was closed not_planned on 09-22. On the PR head, git grep for the script name and for census:page-key-reads finds nothing left anywhere in the tree.
• The scope changes are recorded in the body and accepted. tsconfig-test-parity and component-surface-parity are live hand-run instruments that source comments point readers at. dollar-dialect-alias-census carries the objectui#9096 markdown-totality read. All three move to the maintainer list rather than being deleted.
• skip-changeset: private ui tooling only.
• CI at 714e17ee: 47 success, 13 skipped, 0 failed.
• Contract review not triggered.

Landing: pr_ready + automerge_enable now, under the maintainer's standing grant. #25920 stays open for slices 4 and 5 and the tsconfig dedupe.

Card: #25920

---

Generated by Claude Code

[github-actions]

⛔ merge queue 构建失败 — 先分诊，再决定要不要重排

队列构建 40316529606 红了。队列跑的是全量套件（PR 侧 CI 只跑 affected 子集），
所以失败的测试可能在本 PR 没碰过的包里 —— 那不是重排能修的。每次盲目重排都会让排在后面的所有 PR 重建一轮。

分类：failure —— 按下面的日志分诊。

失败的 job（日志抽取，best effort）：

• Test UI (11/16) — 失败步骤: Run the ui tests (shard 11/16)

  × after a successful save: the shell leaves at once 1066ms
  FAIL   @object-ui/console  src/pages/settings/__tests__/SettingsView.shellLeave-13231.test.tsx > CONTROLS: the page registers only while it holds edits (objectui#13231) > after a successful save: the 
    ↳ 失败原因: (这条 FAIL 之后 12 行内没有可识别的原因行 —— 点进 job 看)
  

↳ 失败原因 是判读的关键：超时（Test timed out in … / Hook timed out in …）多半是负载/时序，不是本 PR 的回归；
断言（AssertionError: …）才指向真实的行为改变。两者的 FAIL 行长得一模一样，只有这一行能区分。

⚠️ 断言这一侧有一类例外，判据是断言在测什么，不是它是不是 AssertionError。 断言的对象是产品行为（一个值、一个形状、一次拒收）⇒ 照上面读：真实的行为改变，去查，⛔ 不要重排掉；
断言的对象是这次实验自身的有效性前提（跑完的耗时、负载下的先后、任何只在时间预算内才成立的条件）⇒ 它跟超时是同一类，同样对负载敏感，重排一次是合法的判别手段。
识别是机械的：断言的消息或它比较的值本身点名了一段时长、一个时间戳、一个耗时计数。实测过的一对 —— AssertionError: SecurityPlugin.init() ran: expected false to be true 测的是产品行为（真回归）；
AssertionError: this run took over a second, so second-precision stamps could have differed too: expected 1006 to be less than 1000 测的是实验前提：它守护的那条不变式当时是绿的，同一个 head 原样重排一次即成功。
穿着 AssertionError 外衣的时间测量，仍然是时间测量。（⛔ 这只改「怎么读一次红」，不改「哪些测试可以重排」——后者由别处管。）

跨 PR 相同签名（24h，按失败测试文件聚合）：

• src/pages/settings/__tests__/SettingsView.shellLeave-13231.test.tsx — 24h 内已弹出 2 个互相独立的 PR（fix(app-shell): the notify recipients editor writes and keeps { emailField } #25996、chore(ui-tooling): drop the page-key read census its card closed as not planned (slice 3 of #25920) #26057）。汇总 issue：Queue-flake anchor: src/pages/settings/__tests__/SettingsView.shellLeave-13231.test.tsx #26096（新建）
• ⚠️ 24h 评论账本没读完（超过 5 页仍未读到窗口尽头），所以上面的「不同 PR 数」是下界，不是全量。

历史信号：

• 本 PR 过去 24h 无队列失败记录（首次）。
• 过去 24h 队列共有 25 个失败构建（不含本次）。

分诊清单：

1. 失败测试在本 PR 改动的包里 → 真回归，修 PR。
2. 失败测试与本 PR 无关 → 看上面的「跨 PR 相同签名」；已有汇总 issue ⇒ flaky/环境问题实锤，去那张 issue 上谈，修好前重排只会再烧一轮全队列。
3. 两者都不是 → 可能与同组 PR 语义冲突；等前面的 PR 落地或失败出队后再重排一次即可，不要连续重排。

---

Generated by Claude Code · merge-queue-triage workflow (#4859)

[objectstack-fleet]

Seat re-queue (devx seat #6023): queue build 40316529606 ejected this PR on Test UI 11/16, where @object-ui/console SettingsView.shellLeave-13231.test.tsx:195 findByText('Profile page') timed out (1066 ms against the 1000 ms default) on a fully loaded shard. This diff deletes only ui/scripts/page-key-read-census.mjs, its test and its alias, and touches no console code. #25996 was ejected with the same signature, and it is filed as flake anchor #26096. The head 714e17ee is unchanged and green, so per the triage note's one-requeue allowance for a timeout-shaped red, the seat re-arms auto-merge once.

---

Generated by Claude Code