[objectstack-fleet]

Requested by Tesla · project thread

Fixes #25831
Part of #25677 (objectui merge aftercare). Gates the joint 18.2.0 Version PR, #25753.

Clause-②: yes (narrowing)
Metadata-protocol: narrowed

What changed

One new file, .changeset/25831-object-ui-18-1-to-merge.md, executing maintainer ruling item 28 = B (card comment 6404422838) as amended by the Claim amendment 6413971007:

• Frontmatter: the 38 public @object-ui/* packages of the fixed group, each minor. The private object-ui (VS Code extension) moves with the group through PRIVATE_FIXED_MEMBERS; the ignored example-*, site, runner and test-support are not named.
• Body: a **BREAKING** lead naming the range (objectui a3c0da9, the 18.1.0 release commit, to 9e1fe03, the freeze), one paragraph for each of the five ! commits (PRs fix(core,rest,services): fail loud when a permission-store read fails #13475, fix(pm-dispatch): gate the worktree-rescue step with the three adjudicated fences #13473, os-dev.md's Part-of bullet lacks the bookkeeping-sentence / quote-backtick shapes that PR #13505 sharpened in review-checklist.md — deferred by the frame-region fence #13506, [finding] check-dev-prereqs 的 declaredEntry() 遇到嵌套 exports 条件会静默落到 pkg.main —— 今天靠 27/27 巧合无害 #13507, platform-admin re-anchor (L5 removal half): delete the legacy grant-row read — one minor after L4's landing #13515) saying what was retired and who it reaches, each pointing at its original ui/.changeset/*.md for the migration table, then a summary of the other 30 PRs with their objectui numbers.
• The 65 originals under ui/.changeset/ are untouched (no root reader; kept as history).

Why one ADR-0087 marker, not five

check-adr-0087-registration accepts exactly one disposition marker per changeset file. Measured on main bbf70614: a probe changeset carrying five markers exits 1 with "5 adr-0087: markers -- exactly one disposition is expected" (self-test battery R9). Every mechanical category (registered, runtime-export-retired, type-surface-only, runtime-interface-only) compares the merge base with HEAD, and the five retirements are already on main, so the only admissible category for this diff is not-required (no-migration-prescription), which refuses an inline FROM/TO table. Hence the paragraphs point at the original changesets instead of restating their tables. The PM accepted this (option A) in Claim amendment 6413971007; per-break markers would be option B (six files) and can be reopened on the card.

The marker, in an HTML comment in the changeset body, reads in words: adr-0087: not-required (no-migration-prescription), reason: the five retirements are @object-ui/* surfaces (a renderer node type and its schema, a widget source member, unread exports, three providers duplicating @object-ui/react / @object-ui/app-shell names, nine @object-ui/types schema keys), none is an @objectstack/spec key, object definition or stored shape, so the packages/spec registries have no entry to write; objectui had no conversion registry; each old spelling is refused by name, and each change's migration ships in the original changeset this entry points to.

Verification (head 0329dd0514)

• node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands: 20 derived commands, all exit 0; --ran reconciliation: 20 derived, 20 run, 0 NOT-MEASURED. Includes check-adr-0087-registration --base origin/main (1 declared-breaking changeset, disposition accepted), check-empty-changeset, check-changeset-no-major, their self-tests, check:changeset-gate-self-tests, check:nul-bytes, check:published-files.
• node scripts/check-changeset-no-major.mjs --base origin/main --event E with this body's two declaration lines: exit 0 (LEVEL AXIS green). The same with Clause-②: no / Metadata-protocol: none: exit 1 (a minor beside none is refused), which is why the Claim's lines were amended.
• node scripts/check-changeset-fixed.mjs: exit 0, 111 public packages + 1 declared private member in sync.
• pnpm changeset status --verbose --since origin/main: 112 fixed-group members to 18.2.0, the 38 @object-ui/* among them, each citing this file.
• changesets v3.0.3 has no version --dry-run, so pnpm changeset version was run on a scratch state of this worktree and reset: all 38 public @object-ui/* manifests read 18.2.0 and ui/packages/core/CHANGELOG.md gains the 18.2.0 entry with the marker hidden as a comment (337 files touched, discarded with git reset --hard). That run was at 213e4d2b; the head differs only by moving the two declaration lines to the end of the body, so the CHANGELOG headline is the BREAKING lead rather than the Clause-②: line.
• Push-environment battery (GITHUB_ACTIONS=true GITHUB_EVENT_NAME=push node scripts/pm/check-dispatch-gates.mjs): result posted on the card with the report.

Acceptance notes

• Arithmetic: root .changeset/*.md count goes from 96 to 97 (+1, this file; README excluded); ui/.changeset/ keeps 65 changesets + README.
• Each of the 38 CHANGELOGs receives the same entry, because the group is fixed and the ruling names all 38.

---

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 40317091962 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）：

• Build Core — 失败步骤: Build packages (excluding docs)（日志不可读，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）：

• ⚠️ 本次没有可用的聚合签名（日志里没有能解析出测试文件名的 FAIL 行）—— 这不是「没有同签名的其他 PR」，是这一轮没测到。跨 PR 聚合本次不可用，请手工比对其他 PR 的同类评论。
• ⚠️ 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 22 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]

Queue build 40317091962 (Build Core) is red on a group PR ahead of this one, not on this diff (one .changeset/*.md file, no source).

• Failure: @object-ui/app-shell#build, src/preview/DraftChangesPanel.tsx(699,9) and (774,36): TS2451 "Cannot redeclare block-scoped variable 'isNew'".
• Cause: at the queue commit 8322e395e4, fix(app-shell): no drafts-ledger read for a member, no published read for a new draft #26056 (dcb798f548, ahead in the group) adds const isNew = entry.kind === 'new'; at line 699 in the same scope where fix(app-shell): pending changes folds platform-written system fields and derived keys into one collapsed line #25984 (already on main, 7bbc0b9081) destructures isNew from detail at line 774. The two PRs are each green alone and conflict together. main (be7897799b) does not carry fix(app-shell): no drafts-ledger read for a member, no published read for a new draft #26056, and its copy of the file has no redeclaration.
• No fix exists to port. fix(app-shell): no drafts-ledger read for a member, no published read for a new draft #26056's author needs to rename one of the two bindings. This PR does nothing until that happens, and it should be re-queued once fix(app-shell): no drafts-ledger read for a member, no published read for a new draft #26056 is out of the group or fixed.

---

Generated by Claude Code