[objectstack-fleet]

Requested by 田騏 · project thread

Fixes #31311

Clause-②: no — test-only; no accept set, schema or export changes
Metadata-protocol: none

维护者速读

• 改了什么:examples/app-todo/test/overdue-escalation-days-overdue.test.ts 改成与 test(example-todo): boot the loop-body kernel once per file under a measured budget #31312 相同的形状。整个文件只在 beforeAll 里启动一次 kernel(带 30 秒实测预算),每条用例开始前经 driver 清空 todo_task 行。三条用例的断言一字未改,只改这一个测试文件。
• 为什么改:原来两条异步用例各自启动一个完整的 kernel。在负载高的 Test Core 分片上,第一条用例在 5016 ms 处被 vitest 默认的 5 秒截断(job 128575054246),把 feat(objectql): restore a recycle-bin entry under its original ids and emit data.record.restored (#30237 PR 3) #31220 踢出了合并队列。同一分片上,下一条用例也跑了 3822 ms。
• 风险与回滚:只改测试文件,不发布(skip-changeset)。隔离已验证:4 种乱序和每条单跑都绿。去掉清空这一步后,第三条先跑的顺序会变红。回滚即 revert。
• 先看:文件里的 BOOT_BUDGET_MS 和它上方的实测注释。
• 两分钟验证:pnpm --filter @objectstack/example-todo exec vitest run test/overdue-escalation-days-overdue.test.ts,预期 3/3 通过。
• 需要你做的:无,席位落地。

What changes

• examples/app-todo/test/overdue-escalation-days-overdue.test.ts only, in the shape test(example-todo): boot the loop-body kernel once per file under a measured budget #31312 gave loop-body-iteration.test.ts.
  • bootTodoData() runs once per file in beforeAll, under an explicit 30 000 ms budget (BOOT_BUDGET_MS). The measurement is written beside it. Before this change it ran inside each of the two async cases, under vitest's 5 s per-test default.
  • beforeEach deletes every todo_task row through the driver, so no hook observes the reset. Each case starts from the table a fresh boot leaves.
  • afterAll disconnects the one driver. This replaces the per-case afterEach driver list.
  • The file registers no fixture (no flow and no object beyond the boot), so there is nothing to clean up in a finally.
  • Every assertion is unchanged.

Why

Queue job 128575054246, which ejected #31220:

case	time
the get_record step declares no projection (sync, no boot)	2 ms
every {{ currentTask.… }} hole … (boot + insert + find)	5016 ms, Test timed out in 5000ms at :129
0 when the task is not overdue … (boot + 2 inserts + find)	3822 ms

In the same job, loop-body-iteration.test.ts's ESCALATION (fixed by #31312) also timed out, at 5024 ms. The package Duration was 68.24 s, with import 149.87s. The slowest sibling case was 2099 ms.

Locally, on 4 cores, I timed the boot phases inside the cases on the old file:

• Cold boot: kernel.use 3 ms, bootstrap 20–23 ms, sqlite-wasm connect 40–139 ms, syncSchemas 45–52 ms, Default Organization 36–54 ms.
• Warm second boot: bootstrap 4–5 ms, connect 1 ms, syncSchemas 18–20 ms, Default Organization 20–25 ms.
• The case that paid the cold boot took 192–279 ms in total, so nearly all of its wall was the boot.

Tests

Case times in ms (sync, async 1, async 2), whole example suite, JSON reporter. "Load" means 8 CPU-bound processes running beside the suite.

	idle	load
before (cold)	2, 223, 68
before (warm)	2, 199, 76 / 2, 205, 81	2, 466, 212 / 2, 693, 236
after	2, 45, 10 / 2, 32, 8	2, 36, 11 / 2, 32, 10

• Isolation: I ran --sequence.shuffle with seeds 1, 7, 42 and 1234. Seed 42 runs the order "every hole", "get_record", "0 when"; seed 1234 runs "0 when", "every hole", "get_record". All 3 cases passed in every order. I also ran each case alone with -t: 1 passed and 2 skipped, every time.
• Reverse check: I removed the beforeEach reset. Seed 1234 then fails "every hole", because the rows "0 when" inserted are still in the table when it asserts toHaveLength(1). The other three orders still pass, because "every hole" runs before "0 when" in them. So the reset is what keeps the cases independent.
• The whole example suite, 5 runs in a row on c497cf52: pnpm --filter @objectstack/example-todo exec vitest run, each exit 0 with Test Files 9 passed (9), Tests 249 passed (249).

Acceptance notes

• The boot budget is per file (beforeAll, 30 s), not per case. The cases now run without a boot and take at most 49 ms locally, even under load, so they keep vitest's default timeout.
• Other sibling files in examples/app-todo/test still boot a kernel in each case: task-recurrence.test.ts (slowest case 2099 ms in job 128575054246, 2060 ms earlier), task-completion-trigger.test.ts (GATES 1447 ms, 1763 ms earlier) and derived-flag-removal.test.ts (1128 ms). None of them is at the 5 s line, so they are not changed here. If one is ever cut off, the same shape applies.

Verification

All on c497cf52 (this branch after #31312 merged, with origin/main merged in, no rebase):

• node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 37 commands. Every one ran with exit 0, after pnpm exec turbo run build --filter='./packages/**' --concurrency=2 so the gates that read built output measured. --ran: "37 derived famil(ies) accounted for — 37 run, 0 NOT-MEASURED".
• pnpm lint: exit 0.
• pnpm --filter @objectstack/example-todo typecheck: exit 0.
• Not run locally: test:repo (the vitest repo project), because this diff does not touch it. Test Core runs it on this PR.

Human Minimum

• 1. Schemas start as Zod; packages/spec holds no runtime logic. (Not touched.)
• 2. Every name I added uses the casing for its kind.
• 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. (Test-only in a private example package; nothing publishes: 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

🤖 Generated with Claude Code

[objectstack-fleet]

domain:server#2 · marchtian · session_017ihL2fBJ6xsLWJsnNJGMMx

Contract review — ACCEPT

Head-sha: c497cf52a300f991090ce5f6bb6207e53e1041c1

ACCEPT · in-seat (test-only, no spec / skills / ADR hunk) · 2026-11-08T01:59Z

• One file: examples/app-todo/test/overdue-escalation-days-overdue.test.ts. It takes test(example-todo): boot the loop-body kernel once per file under a measured budget #31312's shape: one kernel boot in beforeAll under a measured BOOT_BUDGET_MS of 30 s, with the basis in the comment (150–250 ms cold, 466–693 ms under 8 CPU-bound processes, the shard cut-off at 5016 ms with the next case at 3822 ms, job 128575054246). beforeEach deletes todo_task rows through the driver, and afterAll disconnects. The file registers no fixture flow, so no finally is needed.
• No assertion changed, nothing skipped, vitest.config.ts untouched. The two async cases now run in 32–45 ms and 8–10 ms idle, 32–36 ms and 10–11 ms under load.
• Isolation is shown. Four shuffled seeds and each case alone are green. With the reset removed, seed 1234 (the "0 when" case first) goes red, so the reset carries weight.
• Gates: 37 derived, all exit 0, --ran 0 NOT-MEASURED. Lint 0, typecheck 0. The example suite ran 249/249 five times in a row (report 6715901208). check-governed-merges: NOT governed.
• Landing: the seat adds skip-changeset (test-only, private package), readies the PR and enables auto-merge. Fixes #31311.