test(spec): split build-schemas-check-mode.test.ts into six files so the repo project runs them in parallel by objectstack-fleet[bot] · Pull Request #25428 · objectstack-ai/objectstack

objectstack-fleet · 2026-10-23T13:17:22Z

Requested by Tesla · project thread

Clause-②: no
Metadata-protocol: none

The card's re-scope (option A, triage comment 6358468116): split packages/spec/scripts/build-schemas-check-mode.test.ts into several files so the repo vitest project runs them in parallel. Test-only; no turbo.json, partition, ci.yml or shard-count change.

What changes

One 5135-line file becomes six test files and one shared module. vitest runs a file's tests one at a time, so this file (841.97 s, 92 tests, the slowest file of PR run 39893808670) was spec's test:repo floor on its own. The repo project's pool is vitest's default (cores - 1, so 3 workers on a 4-vCPU runner; VITEST_MAX_WORKERS from scripts/vitest-worker-cap.mjs only ever lowers it), so separate files run on separate workers.
build-schemas-check-mode.fixture.ts (not a .test.ts, so neither project collects it) carries the family header, the sandbox builders, the [finding] build-schemas-check-mode.test.ts loses its own fixture commit under the merge-queue full-suite run — 10 tests fail on one phantom sha (first occurrence) #9068 git discipline, run(), and every constant more than one block reads. The old root beforeAll/afterAll become useSharedSandbox(rootScriptsDir), which each test file calls once, so every file builds its own shared sandbox (an fs.mkdtemp, so two files never share one). Module state stays per file because vitest isolates each test file's module graph.

The 15 describe blocks move verbatim, grouped by subject and balanced on a measured local run of the original file (seconds are that run's per-block sums):

File	Blocks	Tests	Local share (s)
`.manifest.test.ts`	`check:authorable-surface` 在 --check 模式下仍会写 `json-schema.manifest.json` —— 一个「检查」在改工作区 #4711, [finding] Five filter operators ($gt/$gte/$lt/$lte/$between) reach NO published reference page — `build-schemas.ts` skips their whole schema over an unrepresentable `z.date()`, and the skip is silent #16431, `gen:schema` rmSync 整个 `json-schema/` 会顺手抹掉 `gen:openapi` 的产物,rest 的 openapi 路由测试随后 503 假红——`check:generated` 原地跑 build-schemas 也触发 #5371, [finding] the published bundle `json-schema/objectstack.json` declares `x-schema-count: 1596` but carries 1585 definitions — and a docs page publishes that field as "the total number of definitions" #12588	18	149.2
`.deleted-baseline.test.ts`	authorable-surface 的 tombstone 门禁可被手编基线绕过 —— 删掉基线行就删掉了证据(#4638 / #4643 已两次这样过绿) #4650	15	137.5
`.anchor.test.ts`	spec 的 #4650 删除闸门在「按 SHA 钉住的消费者构建」里无法锚定 origin/main，硬失败 —— cloud 的镜像构建与 pin bump 全线卡死 #5235, `check:authorable-surface` 在 `--check` 模式下也会重写 `authorable-surface.base.json` —— 一次纯核验会改工作区,且任何无关 PR 都能因此静默推进删除门的锚点 #5358	15	125.6
`.git-topology.test.ts`	os-regen 驱动指示的 `gen:schema` 在 merge 未 commit 时运行,会把 authorable-surface 锚点倒退回旧 merge-base —— 生成器写入、门全绿、静默撤销 main 的锚点推进 #5370, gen:schema 在锚点「领先于」解析基线时打印「trails the baseline … by 0 key(s)」——方向说反,信息为零 #5847, `resolveSurfaceBase()` 的 tip fallback 是静默的正确性降级 —— 拿不到 merge base 时它照样把「删除」判成违规并 `exit 1` #6452 (the three own-sandbox blocks)	17	157.9
`.retired-keys.test.ts`	build-schemas.ts 检查 (b) 用叶名匹配 conversion surface —— 无关簇的 `.type` 就能让一个 tombstone 冒充「已登记迁移」 #4659, [finding] a nested RETIRED_KEYS_BY_MAJOR row is unverifiable — check (b2) passes a fabricated one, so a typo'd nested retirement registers silently and forever #17969	13	130.9
`.deletion-proofs.test.ts`	json-schema.manifest.json 的「deliberate removal」删行仍是纪律而非门禁 —— #4650 的同类洞,上移一层(整 schema 级) #4725, build-schemas.ts 检查 (c) 的「墓碑已满 2 个 major」证明仍用叶名匹配 —— 无关簇的登记可以替一次退休提前起算 #5898, [finding] proof 4 的 `delivers()` message-includes 腿今天只由一个零守着 —— 而通往夹具的三条路各被本仓自己写着的一条规矩堵住 #18865	14	127.1

Block-local constants travel with their block; constants two files read live in the fixture.

Each test file names its out-of-package input. check:cross-package-test-inputs reads a test file's own source, not the modules it imports, so five of the six files at first stopped showing that they run against the repo-root scripts/ and the gate asked to move them to local. Each file now resolves ../../../scripts itself and hands it to useSharedSandbox(), which is the path the fixture symlinks into every sandbox. They stay in the repo project.
vitest.repo-tests.json lists the six files in place of the one.
Budget lines. Each file carries its own // test-budget: line. No CI reading of a split file exists yet, so each figure is apportioned: the pre-split file's slowest CI reading (983.3 s, run 38798542648, the reading the old 1230 line rested on) times the file's share of the local run, times 1.25, rounded up to 10 s (190 to 240 s). Each line says so and asks to be re-measured once CI has read the file on its own.
Prose references to the old file name in packages/spec/scripts/** (build-schemas.ts and eight sibling tests) point at the split file or the fixture they mean.
No changeset: scripts/ ships in no package's files[] (spec's files is dist, json-schema, liveness, prompts, llms.txt, README.md, src/**/*.zod.ts, CHANGELOG.md).

Verification (head `78a4221bd`)

No case lost. vitest list --project repo on the old file and on the six new files, file prefix stripped, sorted: 92 and 92 names, diff empty. vitest list --project local matches none of the files.
Timing, before and after (local, same 4-core box, same lock, labelled local): the original file alone took 839.06 s (92/92 passed). The six split files run together as the repo project: 370.44 s wall, 92/92 passed, slowest file git-topology 205.4 s. The before figure from CI is on the card: 841.97 s for the file, spec two-task total 1212.41 s (PR run 39893808670) and 1494.69 s in merge-group run 39894014103. The after figure from CI is this PR's own Test Core timing table, read by the seat at review.
pnpm --filter @objectstack/spec build, typecheck (includes check:scripts-typecheck): exit 0. pnpm --filter @objectstack/spec test: 711/711 files, 21634 passed, 1 todo.
node scripts/pm/dispatch-gates.mjs --commands derived 66 families on the final head 78a4221bd; all 66 ran and exit 0. --ran reconciliation: 66 run, 0 NOT-MEASURED, 0 UNRUN. On the first sweep (head 9f2ed009f, 63 families) check:cross-package-test-inputs answered 1 (fixed above), check-plugin-teardown-shape --self-test answered 3 (its pinned control commit was missing from the shallow clone; fetched), and check:dual-build-cjs-loads / check:lean-entry-closure answered 3 (no built dist); all exit 0 on the final sweep.

Acceptance notes

The card's done-when number ("spec's two-task total in a merge-group Test Core timing table drops") can only be read after this lands in a merge group; the before figure above is the one to compare against.
scripts/git-env.d.mts:9 (repo root, outside this card's file surface) still names the old file in prose, and scripts/report-test-timings.mjs:935/943 uses the old name as a fixture log line (correct as data). Noted, not changed.
The cross-package scan does not follow a test file's imports into same-package helper modules. Here each file now names its input itself; other tests whose out-of-package reads live only in an imported helper would be unseen by the same scan. Noted, not filed.

CI after-numbers (run 39963105499, head `dce4c6406`, `pull_request`, Test Core aggregate job 120460015448)

Slowest file in spec: 841.97 s before (the one file, PR run 39893808670). After, the six files read 457.69 s (git-topology), 420.84 s (deleted-baseline), 386.75 s (deletion-proofs), 147.50 s (retired-keys), 140.73 s (manifest) and 104.13 s (anchor). The slowest is now 457.69 s.
Spec's two-task total (the card's done-when quantity): 1113.11 s in this PR run, against 1212.41 s in PR run 39893808670 and 1494.69 s in merge-group run 39894014103. The merge-group comparison can be read only after this lands.
Load cost: on a shared runner, the files run side by side and each takes longer than its serial share. The six together are 1657.6 s of file time, against 842 s for the one file. The wall drops because they overlap.
Shard tests steps in this run: 10.6 to 20.4 min (slowest shard 4/6). That is a reading of this run, not a comparison of like with like.
Budget lines were re-set from this reading in ab63b4e7e: each is the CI reading × 1.25, rounded up (580, 530, 490, 190, 180 and 140 s). They replace the first apportioned estimates (190 to 240 s), which assumed each file keeps its serial share and came in under the three heaviest readings.

🤖 Generated with Claude Code

https://claude.ai/code/session_01P8iUGfP2wffkVBD5uSfDgq

…the repo project runs them in parallel The one 5135-line file was the slowest file of the Test Core run (841.97 s, 92 tests) and vitest runs a file's tests serially, so spec's test:repo floor was this file alone. The shared sandbox, git discipline, run() spawner and cross-file constants move to build-schemas-check-mode.fixture.ts (not a test file); each split file registers its own shared sandbox via useSharedSandbox(). The 15 describe blocks move verbatim, grouped by subject and balanced on a measured local run; vitest.repo-tests.json lists the six files, and each carries its own test-budget line. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01P8iUGfP2wffkVBD5uSfDgq

…ode files Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01P8iUGfP2wffkVBD5uSfDgq

check:cross-package-test-inputs reads a test file's own source, not the fixture module it imports, so five of the six split files stopped showing the out-of-package input they run against. Each file now names the repo-root scripts/ directory and hands it to useSharedSandbox(), which is what the fixture symlinks into every sandbox. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01P8iUGfP2wffkVBD5uSfDgq

github-actions · 2026-10-23T13:18:26Z

📓 Docs Drift Check

⚠️ 1 changed file(s) yielded no anchor (packages/spec/vitest.repo-tests.json), so the pages documenting them are NOT COVERED by this run — this is not a clean bill of health for those files. Nothing else in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 1 changed package(s)).

What this run could not see

1 changed file(s) yielded no anchor (packages/spec/vitest.repo-tests.json) — pages documenting those are invisible to this run
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 — 144 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 b9c7a8f541a0af1ec82cabad8c9aaab65de895c1 → packageMentionDocs.

objectstack-fleet · 2026-10-23T13:22:00Z

CI · Type Check · source gates failed before any check ran · domain:spec seat 1 · 2026-10-23T13:21Z

The job (run 39959968741, job 120435206191, head 78a4221bd) died in checkout: three RPC failed; HTTP 408 / expected 'packfile' fetch errors, then git exit 128. No type check or gate body executed, so this is a runner/network failure, not this PR's. The seat re-runs the failed jobs once when the run completes (scripts/pm/rerun-failed.mjs); a second failure is treated as real.

Generated by Claude Code

Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01P8iUGfP2wffkVBD5uSfDgq

…irst CI reading The apportioned figures (190-240 s) assumed each file keeps its serial share of the pre-split wall. CI run 39963105499 read the files at 104-458 s: running side by side on a shared runner, the three heaviest each take longer than their serial share. Each budget is now that reading x 1.25, rounded up. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01P8iUGfP2wffkVBD5uSfDgq

objectstack-fleet · 2026-10-23T14:59:03Z

维护者速读

改了什么: spec 包里最慢的那个测试文件 build-schemas-check-mode.test.ts(5135 行、92 个用例)拆成 6 个测试文件，再加一个不算测试的共享夹具文件。用例本身一字未改，拆分前后 92 个用例名逐个比对一致。
为什么改: 这一个文件串行跑 842 秒，是 spec 测试的硬底线(ci(test-shards): split spec's build-schemas-check-mode.test.ts (842 s, 92 tests, one file) into several files, so spec's test:repo floor stops setting the slowest Test Core shard — re-scoped from file-sharding spec's test task #25372)。拆开后 CI 的多个 worker 能并行跑。这个 PR 自己的 CI 读数:spec 两项测试总耗时 1212 s → 1113 s，最慢的单个文件 842 s → 458 s。
风险与代价(含回滚): 只动测试和测试清单，不碰任何发布的代码，不需要 changeset。行数大(+5542 / −5150)是因为整块代码搬家。并行后每个文件在共享 runner 上比单跑慢，所以 6 个测试预算行按这次 CI 读数 ×1.25 重设。回滚就是 revert 这一个提交。
席位意见: 建议合并。CI 全绿(49 项，38 成功 + 11 跳过)，用例一个没少，范围符合分诊席的改写意见 A(不改 turbo / ci.yml / 分片数)。
你要做的: 行数超过 3000 行人工合并线，请 os-zhuang 或 hotlong 中一位在 PR 上点 Approve，之后席位自动入队合并。

席位核对 · ACCEPT(待人批) · `domain:spec` seat 1 (seat post #6017) · `os-tesla` · session `session_013odhA26dvzy7pQmevdTTvy` · 2026-10-23T14:58Z

Head ab63b4e7e8 · card #25372 · claim 6364060214 · re-scope 6358468116 (option A).

Shape: draft, base main, body line Fixes #25372, skip-changeset (scripts/ ship in no package's files[]). 18 files, +5542 / −5150. check-governed-merges: no governed path, but 10692 lines > 3000 ⇒ human merge or an authorized approval from GOVERNED_APPROVERS.
Premise measured: 92 → 92 cases (the seat's own count over the six files; the dev's vitest list diff is empty). vitest.repo-tests.json lists all six; the old name is gone from it.
Gates pinned to the head: 49 check-runs on ab63b4e7e8, 38 success + 11 skipped, 0 red. The earlier Type Check · source gates red on 78a4221bd was a checkout failure (comment 6365877899), cleared by a real merge of main (dce4c6406).
Accepted deviations: each split file names ../../../scripts itself (check:cross-package-test-inputs does not follow imports); budget lines re-set from the CI reading × 1.25 in ab63b4e7e.
Out of scope: the cross-package scan not following same-package helper imports is noted in the PR body, not filed.
Landing: stays draft, needs-user-decision on the PR, review requested from os-zhuang and hotlong. On an authorized APPROVED the seat runs pr_ready + automerge_enable.

Generated by Claude Code

claude added 3 commits October 23, 2026 12:04

docs(spec): point prose references at the split build-schemas-check-m…

9f2ed00

…ode files Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01P8iUGfP2wffkVBD5uSfDgq

objectstack-fleet Bot assigned os-tesla Oct 23, 2026

objectstack-fleet Bot added the skip-changeset PR has no user-facing published change; bypasses the changeset gate label Oct 23, 2026

github-actions Bot added size/xl tests tooling Fix lands in tooling, not a product package — enters pm:queue only with Unblocks: or a named surface labels Oct 23, 2026

claude added 2 commits October 23, 2026 13:37

Merge origin/main into claude/project-thread-2vwjru

dce4c64

Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01P8iUGfP2wffkVBD5uSfDgq

objectstack-fleet Bot requested review from hotlong and os-zhuang October 23, 2026 14:59

objectstack-fleet Bot added the needs-user-decision Blocked on a maintainer decision — do not dispatch; decision cards only, never on a PR label Oct 23, 2026

os-zhuang approved these changes Oct 23, 2026

View reviewed changes

os-zhuang marked this pull request as ready for review October 23, 2026 17:48

os-zhuang enabled auto-merge October 23, 2026 17:48

os-zhuang added this pull request to the merge queue Oct 23, 2026

Merged via the queue into main with commit 87af852 Oct 23, 2026
55 checks passed

os-zhuang deleted the claude/project-thread-2vwjru branch October 23, 2026 18:30

Navigation Menu

Repository navigation

test(spec): split build-schemas-check-mode.test.ts into six files so the repo project runs them in parallel - #25428

test(spec): split build-schemas-check-mode.test.ts into six files so the repo project runs them in parallel#25428
os-zhuang merged 5 commits into
mainfrom
claude/project-thread-2vwjru

objectstack-fleet Bot commented Oct 23, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 23, 2026 •

edited

Loading

Uh oh!

objectstack-fleet Bot commented Oct 23, 2026

Uh oh!

objectstack-fleet Bot commented Oct 23, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

objectstack-fleet Bot commented Oct 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What changes

Verification (head 78a4221bd)

Acceptance notes

CI after-numbers (run 39963105499, head dce4c6406, pull_request, Test Core aggregate job 120460015448)

Uh oh!

github-actions Bot commented Oct 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

📓 Docs Drift Check

Uh oh!

objectstack-fleet Bot commented Oct 23, 2026

Uh oh!

objectstack-fleet Bot commented Oct 23, 2026

维护者速读

席位核对 · ACCEPT(待人批) · domain:spec seat 1 (seat post #6017) · os-tesla · session session_013odhA26dvzy7pQmevdTTvy · 2026-10-23T14:58Z

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

objectstack-fleet Bot commented Oct 23, 2026 •

edited

Loading

Verification (head `78a4221bd`)

CI after-numbers (run 39963105499, head `dce4c6406`, `pull_request`, Test Core aggregate job 120460015448)

github-actions Bot commented Oct 23, 2026 •

edited

Loading

席位核对 · ACCEPT(待人批) · `domain:spec` seat 1 (seat post #6017) · `os-tesla` · session `session_013odhA26dvzy7pQmevdTTvy` · 2026-10-23T14:58Z