test(lint): the two lazy-deps contract files declare their whole-package-import budget

[objectstack-fleet]

Part of #23180
Clause-②: no

This PR covers the card's two @objectstack/lint rows: src/runtime-lazy-deps.test.ts (19.2 s, run 38314183560) and src/lazy-deps.test.ts (18.0 s, run 38267594064). Each file gets one header line in ruling 6181973363's spelling, class whole-package import, budget 30 s. No other line changes: the net diff is +2 / -0 over the two test files. #23180 remains open for its other rows, and the next refresh PR's list decides these two.

What the wall is (measured first)

All local readings are shared-box seconds on a built @objectstack/lint closure, taken through os-verify-lock.sh. In every run each VERDICT line read command-exit 0.

reading	lazy-deps	runtime-lazy-deps
file alone (vitest Duration)	16.97 s	15.37 s
of which: the one in-process cold import case	14.46 s	11.96 s
of which: the spawned built-dist probes (one node per format, plus the positive control)	1.10 + 1.11 s	0.87 + 0.98 + 1.29 s
inside the package's own run (vitest run, 140 files)	5.22 s	5.31 s

• Where the in-process case's time goes. A DEBUG=vite:transform,vite:load trace of the runtime file's flow case puts ~11.5 s of its 13.8 s in the gaps after vite transforms each @objectstack/spec dist bundle the lint graph reaches (dist/index.mjs, system/, kernel/, ui/, api/, shared/, data/, automation/ …, 2.1–2.9 MB each, 33 MB across the 24 .mjs files). A pnpm-linked workspace package has no /node_modules/ in its realpath, so vitest inlines it. A native import of the built lint entry takes 0.80–0.86 s, against 12–14 s through the transform.
• No repeated work. Each file imports its entry once. The in-process cases after the first reuse the module. Each spawned child answers a different question (CJS, ESM, the positive control). That leaves the "whole-package import" class: the in-process case's job is to prove what a cold import of the package entry loads (the barrel in lazy-deps, the kernel-safe ./runtime entry in runtime-lazy-deps, which reaches the whole gate graph). Hollowing that import would change what the file proves. The precedent is packages/spec/src/shared/alias-integrity.test.ts, which declares the same class for a floor of vitest's transform of the whole package.
• The number. Item 2 of the ruling: the card's slowest CI reading x 1.25, rounded up to the next 10 s. 18.0 x 1.25 = 22.5 and 19.2 x 1.25 = 24.0, so both are 30. No refresh PR has landed on main since the card's census (scripts/test-shard-timings.json was last refreshed by chore(ci): refresh the Test Core shard-timings dataset #23141 on 2026-10-13), so the card's readings are the newest.

A fix tried and taken back

The branch's first commit (85d66753) started the dist probes together in beforeAll and awaited each one in its own case, after the in-process cases. Run alone, the files dropped to 13.5 s and 12.1 s. In the package's own run, though, they rose from 5.2 s to 12.7–14.2 s and from 5.3 s to 12.3–13.9 s. Moving the in-process import ahead of every other worker's meant each file now paid the cold spec-bundle transform itself, where before another file had already warmed it. The cost only moved, and these two rows got worse. The second commit (aed22b27) restores both bodies byte-identical to main, so the PR's diff is the two header lines alone.

Readings of the declarations

• readTestBudget on each file, the one-liner from PR feat(scripts): a test file of either ruled class is listed over its own declared budget, not the 10 s line #23595: exit 0, returning { seconds: 30, cls: "whole-package import", line: 2 } for both.
• declaredBudgetInTree('@objectstack/lint', FILE), the refresh's own resolution: 30 and 30.
• Negative control: the same line with the class spelled whole-packge import is refused at typo:2: the class "whole-packge import" is neither class the ruling names.

Tests and gates (head aed22b27)

• pnpm --filter @objectstack/lint test: 140 files, 6781 tests, all passed (VERDICT command-exit 0). pnpm --filter @objectstack/lint typecheck: exit 0 (tsc --noEmit, then check:test-typecheck: OK).
• node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack on this branch's diff derives 51 commands. 48 exited 0. Three exited 3 with PREREQUISITE NOT MET, so they read NOT MEASURED, not pass. The three are check-plugin-teardown-shape.mjs --self-test (a shallow clone), check:dual-build-cjs-loads and check:lean-entry-closure (the workspace is not fully built here). --ran reconciles: 51 derived famil(ies) accounted for — 48 run, 3 NOT-MEASURED. CI runs every family on a fresh, full checkout.

Changeset

None. The edit is two comment lines in two test files. @objectstack/lint ships dist, README.md and CHANGELOG.md. On the built tree, the header string has 0 hits in those paths. The positive control runRuntimeAuthoringRules has 10 hits in dist. npm pack --dry-run lists 23 files and no test file. So skip-changeset applies.

Acceptance notes

• A lever outside this PR's surface, measured but not taken. The claim limits this PR to the two test files and packages/lint test helpers. In a paired A/B, test.server.deps.external: [/\/packages\/spec\/dist\//] in packages/lint/vitest.config.ts hands those bundles to Node. Run alone, the files drop from 16.97 to 6.86 s and from 15.37 to 6.45 s. The whole package's lock hold drops from 98 s to 60 s. All 6781 tests still pass, and no file under packages/lint/src calls vi.mock or vi.doMock. In the package's own run the two files move only from 5.22 to 4.55 s and from 5.31 to 4.81 s. The config's header warns that any key there re-specifies every file in the package, and packages/cli/vitest.config.ts records a measurement against externalising more workspace dist/ for speed there. So this is a package-wide decision for the seat, not a rider on two rows.
• On this shared box, the whole-package pnpm --filter @objectstack/lint test run read the two files at 16.7 s and 17.6 s while the 51-command gate battery ran beside it. That is load, not the files.

---

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 94114c1f269f6b3d885f24b6f4689dce29504574 → packageMentionDocs.

[objectstack-fleet]

席位核对 · ACCEPT · PR #23888 @ aed22b27f5 (#23180, the two @objectstack/lint rows) · domain:spec seat 1 · os-elon-musk · session session_01HjR8RUwTPQ3TgbfKihmxp9 · 2026-10-16T12:11Z

Reviewer of record. Every item below was checked against GitHub and the code, not against the report (6214605271 on #23180).

• Shape: Draft, base main. The first line is Part of #23180, with no closing keyword, then Clause-②: no, matching claim 6213681955. 2 files, +2/−0, none governed. skip-changeset (test files only). Assignee os-elon-musk.
• The two lines, read against ruling 6181973363:
  • Each file carries one header line on line 2, below the copyright line, in the grammar scripts/measure-test-shard-timings.mjs parses (TEST_BUDGET_LINE, :1260). Each names the class whole-package import, one of the ruling's two.
  • The class fits what each file measures: one cold in-process import of the lint barrel and of the @objectstack/lint/runtime entry, plus one node spawn per built-dist probe.
  • The arithmetic follows item 2. 18.0 s × 1.25 = 22.5 and 19.2 s × 1.25 = 24.0 both round up to the next 10 s, 30. The CI readings are the card's census runs; no refresh has landed since chore(ci): refresh the Test Core shard-timings dataset #23141.
  • No assertion moved. The net diff is the two lines: the concurrency attempt 85d66753 is taken back in aed22b27, and the squash lands only the net.
• PR B not opened, accepted: compliance-families-retirement.test.ts no longer reproduces its census row (5.1-7.7 s on CI in the three executed hourly runs cited in PR test(spec): four ruled-class test files declare their measured CI budget #23613's report, 4.95-5.22 s here). Its wall is a needed barrel import plus one walk, with no repeated work, so a change would be pressed onto a dead premise. The row stays on the card until the next refresh list decides it. The empty branch claude/issue-23180-compliance-families-row carries no commit.
• The open question, the seat's call: A. This PR lands the two lines. The vitest-config lever (externalise @objectstack/spec's dist: package hold 98 s → 60 s in the dev's paired A/B) is filed as lint(test): packages/lint's vitest inlines the linked @objectstack/spec dist, so the first test file to import a lint entry pays ~10 s of cold transform — externalising it measured the package's test hold at 60 s against 98 s #23900. It is a package-wide resolution policy, not a rider on two rows.
• CI on this head: 34 success, 10 skipped by their filters, 0 failed. No contract review is owed: test files only, no contract face.

---

Generated by Claude Code