[objectstack-fleet]

Requested by Elon · project thread

domain:server#1 · os-elon-musk · session_016WV4cyFzEX3mhkfxdV1M8E

Fixes #30873

Clause-②: no
Metadata-protocol: none

Claim: 6681653863 (re-posted with this branch as 6681707609) · Branch: claude/project-thread-7l5fxn

维护者速读

• 改了什么：只改一个测试文件 packages/rest/src/meta-write-door-capability-enumeration.test.ts，在文件顶部加一行 import '@objectstack/objectql'; 和说明注释。没有改任何断言，没有改任何源码。
• 为什么改：「币种代码回填」接口的对照用例是这个文件里唯一会越过权限闸门、走到接口里那句 import('@objectstack/objectql') 的用例。这个包第一次加载要冷编译整个查询引擎，空闲时 3.2 秒，全算在该用例 5 秒的超时窗口里；机器一忙就超时，Test Core (6/8) 变红，把无关的 PR 踢出合并队列。把第一次加载挪到模块顶部，它就在收集阶段完成，收集阶段不计超时。
• 风险与回滚：极小。纯测试代码，不发布任何文件；该用例仍然在两个挂载点、两种身份下各打一次，仍然要求 501 NOT_IMPLEMENTED。回滚即撤 PR。
• 先看：packages/rest/src/meta-write-door-capability-enumeration.test.ts 顶部新增的那一行 import 和注释；对照用例本身（文件末尾 the control: its own capability）一字未动。
• 两分钟验证：pnpm --filter @objectstack/rest exec vitest run --project local src/meta-write-door-capability-enumeration.test.ts，预期 66/66 通过，该用例约 5 ms（之前约 3.3 秒）。
• 需要你做的：无，席位落地。

What changes

packages/rest/src/meta-write-door-capability-enumeration.test.ts only, +10 / -0: a module-top import '@objectstack/objectql'; under a comment carrying the measured reason. No assertion, no case and no source file changes.

The #30675 control case ("its own capability (and isSystem) passes the gate, on both bases") is the one case in this file that gets past a gate into the record-currency backfill door, which loads backfillRecordCurrencyCodes through a guarded import('@objectstack/objectql'). In this package that specifier resolves through the workspace link to packages/objectql/dist/, so the first load is a cold vite transform of the engine's module graph. It was paid inside the case's 5000 ms testTimeout. Loaded at module top, it is paid during collection, which no vitest timeout clocks (the convention in AGENTS.md § Build & Test and in scripts/check-test-source-alias.mjs's clocked-window section; analytics-dataset-body-door.test.ts in this package is the same shape). The door's own import() then resolves to the module already loaded.

The case is unchanged: it still knocks on both bases (/api/v1 and the scoped /api/v1/environments/:environmentId) as both a setup.write holder and isSystem, and still expects 501 NOT_IMPLEMENTED with the nested NOT_IMPLEMENTED code, which is the answer that tells the gate's verdict apart from its refusal. @objectstack/objectql is already in @objectstack/rest's KNOWN_UNALIASED_TEST_IMPORTS set (other tests here import it statically), so no ledger moves.

Why

Card #30873: Test Core (6/8) on PR #30870 (a translations-only service-settings PR) timed out on this case twice, job 127778628692 on 9c22c667 and job 127785162372 on 9e0f2cd6, Test timed out in 5000ms, while every other case in the file passed.

Where the time goes, measured on origin/main 98f640d5 in this container (4 vCPU) with a throwaway probe test (not committed) that times each step inside a vitest worker of this package:

Step	Idle, two runs
first import('@objectstack/objectql')	3201 / 3257 ms
second import('@objectstack/objectql')	0 / 0 ms
first new RestServer(...) + registerRoutes() (scoping auto)	11 / 10 ms
each later boot	0-1 ms

So of the case's ~3.3 s idle, about 3.2 s is the cold load; the four boots cost about 13 ms together and the handler past the gate is negligible. The case pays the import on its first knock only, which is why the other three knocks in the same case are cheap.

The other cases in the same describe do not share the cost: the anonymous case and the refusal case are answered 401 / 403 before the door resolves an engine or reaches the import (2-6 ms each, before and after). The route's own test file, meta-record-currency-backfill-route.test.ts, already imports ObjectQL from @objectstack/objectql at module top.

No timeout was raised, and none is needed (below).

Tests

Before is origin/main 98f640d5, after is this branch. Same container, 4 CPUs. "Loaded" pins vitest to CPUs 0-1 beside four node -e 'for(;;){}' busy loops pinned to the same two CPUs (PIDs recorded, killed by a trap). Command: pnpm exec vitest run --project local src/meta-write-door-capability-enumeration.test.ts in packages/rest.

	The control case	Sum of the 66 cases	Result
Before, idle (3 runs)	3314 / 3308 / 3986 ms	3402 / 3401 / 4099 ms	66/66 pass
Before, loaded (3 runs)	5016 / 5023 / 5026 ms	5282 / 5224 / 5250 ms	1 failed, 65 passed, Test timed out in 5000ms, 3 of 3 runs
After, idle (3 runs)	5 / 5 / 5 ms	94 / 89 / 113 ms	66/66 pass
After, loaded (3 runs)	13 / 12 / 15 ms	239 / 236 / 221 ms	66/66 pass

Every other case is under 20 ms idle before and after (the slowest is the anti-drift enumeration case, 14-19 ms idle, 29-34 ms loaded after).

Where the load went: vitest's own Duration line, one run each

The load is moved out of the clocked window, not removed: the file's total wall is about the same, and the seconds move from tests to import (collection).

• Before, idle: Duration 8.86s (transform 6.43s, setup 0ms, import 5.24s, tests 3.47s), 66 passed.
• After, idle: Duration 8.50s (transform 6.23s, setup 0ms, import 8.26s, tests 91ms), 66 passed.
• Before, loaded: Duration 16.05s (transform 12.65s, setup 0ms, import 10.16s, tests 5.31s), 1 failed, 65 passed: FAIL |local| … > the control: its own capability (and isSystem) passes the gate, on both bases, Error: Test timed out in 5000ms.
• After, loaded: Duration 25.33s (transform 18.05s, setup 0ms, import 24.49s, tests 343ms), 66 passed. The loaded collection is slow and noisy on two contended cores; no vitest timeout covers it.

The before-idle Duration run was taken by writing the 98f640d5 blob of the test file over the committed fix, running it, and restoring with git checkout HEAD --; the restored file's hash equals the HEAD blob and git diff HEAD is empty.

No ablation: the diff changes no behaviour a test pins (it moves where a module is first loaded; every assertion is byte-identical to main).

Acceptance notes

• A test-only change: @objectstack/rest's files is dist, README.md and CHANGELOG.md, and after pnpm --filter @objectstack/rest build the new comment has 0 hits under packages/rest/dist (positive control: _migrate-record-currency-codes, 2 hits). Nothing published moves. No changeset, labelled skip-changeset.
• pnpm check:test-source-alias did not flag the old shape and could not: its clocked-window rule reads load sites in TEST files (plus one hop through a same-package helper that imports a non-literal specifier), while this load sat in the door's source, reached from a test's clocked window through a route handler. Reported to the seat as a finding; not changed here (outside this card's file surface).

Verification

All on the final head 73062e5028 (one commit on origin/main 98f640d5, which had not moved when this PR opened; the delta against main is the one file, +10 / -0), after a whole-tree turbo run build --filter=!@objectstack/docs (118/118 tasks, lock VERDICT command-exit 0).

• pnpm --filter @objectstack/rest test (the local project) under the verify lock: Test Files 393 passed (393), Tests 7456 passed, 332 skipped (7788), VERDICT command-exit 0. The repo project (test:repo) is left to CI: this file is not in vitest.repo-tests.json.
• pnpm --filter @objectstack/rest typecheck (tsc --noEmit plus check:test-typecheck, which compiles the test layer under tsconfig.test.json): VERDICT command-exit 0, 0 errors.
• node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 49 families on 73062e5028. Each ran sequentially with its exit code captured before any pipe: 48 exit 0 on the first pass and pnpm check:dual-build-cjs-loads exit 3 (PREREQUISITE NOT MET, no dist for 46 packages); after the whole-tree build it ran again, exit 0. pnpm lint: exit 0.
• dispatch-gates --ran: "49 derived famil(ies) accounted for — 49 run, 0 NOT-MEASURED", with pnpm lint listed as outside the derivation.
• Every family CI runs on this PR is read from the PR's check runs, not re-run here.

Human Minimum

• 1. Schemas start as Zod; packages/spec holds no runtime logic. (No packages/spec change.)
• 2. Every name I added uses the casing for its kind. (No name added.)
• 3. No formatter ran; pnpm lint passes.
• 4. One issue, one branch off an up-to-date main, one pull request.
• 5. A change that publishes carries a changeset, and both declaration lines are filled in here and in the changeset. (Nothing publishes: test-only, skip-changeset.)
• 6. packages/spec touched: its generated artifacts are regenerated and committed. (Not touched.)
• 7. No hand edits to content/docs/references/, packages/*/CHANGELOG.md or content/docs/releases/.
• 8. Title, body and comments are in English.
• The Before Submitting list in CONTRIBUTING.md is done.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

Nothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs.

What this run could not see

• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 0 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 98f640d581339769d54cf27c8f7b02cbc46849f4 → packageMentionDocs.

[objectstack-fleet]

domain:server#1 · os-elon-musk · session_016WV4cytLmJxfAmgPVjUbbW

席位核对 · ACCEPT · 2026-11-06T10:42Z · #30873

Head-sha: 73062e5

• Shape: 1 file, +10/−0, one commit on main 98f640d5. Inside claim 6681707609 (branch claude/project-thread-7l5fxn); test-only; skip-changeset is right for a private test file.
• Read in the diff: a module-top import '@objectstack/objectql' in packages/rest/src/meta-write-door-capability-enumeration.test.ts, with the measured reason in a comment. @objectstack/objectql is already a dependency of @objectstack/rest (package.json :46), so nothing new is pulled in.
• Cause, as measured by the dev on main: the control case's ~3.3 s was the first import('@objectstack/objectql') through the backfill door's guarded dynamic import (3.2 s cold transform), paid inside the 5000 ms window; the stack boots cost about 13 ms. The red reproduced 3/3 on two loaded cores. After: the case runs in 5 ms idle and 12–15 ms loaded, 66/66 every run.
• Card's ⛔ lines: no skip, no retry, no quarantine, no timeout raise; no assertion touched, so the control still proves its own capability and isSystem pass the gate on both bases. The other cases in the describe do not share the cost.
• Gates: every check-run on this head has concluded: 48 distinct checks, the latest run of each 34 success, 14 skipped, 0 failure (Test Core 6/8 green). Locally (dev report 6682594602): 49/49 derived families exit 0 with --ran 0 not measured; pnpm lint 0; rest suite 393/393 files; rest typecheck 0.
• Contract: Clause-②: no, Metadata-protocol: none, confirmed from the diff.
• Out-of-scope finding (dev, class a): check:test-source-alias cannot see a literal dynamic import inside the source under test reached from an it() body, and the rest door sources carry many guarded await import calls of @objectstack/spec subpaths in the same latent shape. Filed as check:test-source-alias misses a literal dynamic import in the source under test, so a cold dist load inside a test window passes the gate #30909.
• Landing: ready, then auto-merge, through the relay. This unblocks PR i18n(service-settings): ship the built-in settings pages in zh-Hant, ar, fr and pt-BR #30870.