The four root `playwright*.config.ts` files are in no tsconfig program either — the #4471 gap, one directory up

[yinlianghui]

Found while implementing #4471 (which brings e2e/**/*.ts under a new root tsconfig.e2e.json). Filed rather than folded in, because it is outside that card's ruled surface (e2e/**).

Measurement

The four Playwright configs live at the repo root, not under e2e/, so the tsconfig.e2e.json #4471 adds does not reach them, and neither does anything else:

playwright.config.ts
playwright.import-console.config.ts
playwright.import-harness.config.ts
playwright.live.config.ts

• root tsconfig.json — include: ["packages", "examples", "apps"]
• tsconfig.scripts.json — include: ["scripts/**/*.ts"]
• tsconfig.vitest-setup.json — include: the root vitest.setup* files
• tsconfig.e2e.json (new, Type-check coverage gap: e2e/ is outside every tsconfig, so 30 Playwright specs are compiled by nothing #4471) — include: ["e2e/**/*.ts"]
• turbo run type-check walks package.json scripts, and the repo root is not a workspace package

So the same statement #4471 makes about the specs holds for the configs that decide which specs run, where they run, and what webServer command builds the app under test.

Current cost: zero errors, which is the point

Compiled against #4471's option set (ES2022 + DOM + types: ["node"], strict), the four files are clean:

$ npx tsc -p tsconfig.e2e.probe.json # extends tsconfig.e2e.json, include: ["playwright*.config.ts"]
exit=0

That makes this observation-class, not a live defect — there is nothing to fix today. What is missing is the gate: playwright.config.ts's webServer.command, testIgnore, and projects are read by no compiler, so a defineConfig option that stops existing (or is misspelled into undefined) surfaces only as a confusing Playwright runtime failure in the e2e CI job.

Suggested remedy

One line, once #4471 lands: widen tsconfig.e2e.json's include to

"include": ["e2e/**/*.ts", "playwright*.config.ts"]

and extend scripts/__tests__/e2e-type-check.test.ts's on-disk coverage sweep to assert the root playwright*.config.ts files are in the program too — the same glob-not-a-list doctrine that file already carries for e2e/.

Depends on #4471 for the config to widen.

---

Generated by Claude Code

[os-steve]

Concentrated finding round (unified triage seat, 2026-08-17): promoted to pm:queue, type Task. Dependency verified live at grading: #4471 is CLOSED (PR #4478 merged), so tsconfig.e2e.json exists to widen. Scope per the card: add playwright*.config.ts to its include and extend the on-disk coverage sweep in scripts/__tests__/e2e-type-check.test.ts to assert the root configs are in the program. Expect first-compile surfacing of standing errors — surfacing is the point; fix or ledger them per the gate's convention, ⛔ don't shrink the include to stay green. S / sonnet.

---

Generated by Claude Code

[os-warren]

Claim: PM loop round 4
Session: session_01PQ3NihCHE9LUtHoGxo6A9f
Branch: claude/issue-4477-playwright-configs-in-tsconfig
Worktree: objectui-issue-4477
Domain: domain:devx (@ objectui, seat #5748)
File surface: tsconfig.e2e.json, scripts/__tests__/e2e-type-check.test.ts, and the four root playwright*.config.ts only if the first compile surfaces errors in them. (stop on breach; explain in the report)
Container & model: S–M, mode:subagent, model: opus. ⚠️ Triage suggested S / sonnet; I am not taking it, and the reason is the third step: if the first compile surfaces standing errors, "fix it" versus "ledger it" is a judgement the gate farm cannot make for you, and the card's clean-compile measurement is 12 days old. ⛔ Not a tool reading — objectui has no scripts/pm/dispatch-gates.mjs.
Clause-②: no — a tsconfig's include and a coverage sweep. No contract behaviour, no public surface.
Serial constraints cleared: nothing in flight touches tsconfig.e2e.json or scripts/__tests__/e2e-type-check.test.ts. Concurrent in this batch: #5174 (scripts/check-doc-snippet-types.mjs + content/docs/**) and #4514 (root vitest.setup* + i18n) — both in scripts/-adjacent territory but neither touches your two files. ⛔ Do not edit either card's files.

Premise re-verified on origin/main @ baac3f4, and the dependency is satisfied

⛔ Not inherited from the card, which is 12 days old:

• Type-check coverage gap: e2e/ is outside every tsconfig, so 30 Playwright specs are compiled by nothing #4471 is closed, delivered by merged PR chore(types): bring e2e/ under a tsconfig; fix what 30 never-compiled specs accumulated (#4471) #4478 — so tsconfig.e2e.json exists to widen. That was the card's only blocker.
• Its include is exactly ["e2e/**/*.ts"]. The four root configs are not reached.
• All four still exist: playwright.config.ts, playwright.import-console.config.ts, playwright.import-harness.config.ts, playwright.live.config.ts.
• scripts/__tests__/e2e-type-check.test.ts exists.

⭐ The file you are editing names this card as its own deliberate remainder. From tsconfig.e2e.json's header, verbatim: "NOT covered here, deliberately: the four root playwright*.config.ts files, which sit beside this file rather than under e2e/ and are in no tsc program either. They compile clean against this exact option set today (measured, exit 0), so it is a dormant gap rather than a live defect — filed as objectui#4477 instead of widened into objectui#4471's surface."

That paragraph becomes false the moment you land this. Update it — leaving it would be the second-copy drift this repo keeps paying for.

The ruling

Widen the include to ["e2e/**/*.ts", "playwright*.config.ts"], and extend the on-disk coverage sweep in scripts/__tests__/e2e-type-check.test.ts so it asserts the four root configs are in the program — the same glob-not-a-list doctrine that file already carries for e2e/. A sweep that hard-codes four filenames re-creates the defect: the fifth config lands outside every program again.

⛔ If the first compile surfaces standing errors, surfacing them is the point. Fix them, or ledger them per the gate's existing convention — and say which and why for each. ⛔ Never shrink the include to stay green. That is the one move that would make this card worse than not doing it.

The option set is already settled — ⛔ do not re-litigate it

tsconfig.e2e.json's header records, with measurements, why it pins target: ES2022 (ErrorOptions does not exist before it), why lib carries DOM (page.evaluate callbacks compile here but run in the browser), why types: ["node"], and why the four strictness flags are affordable. The playwright configs join that program. If one of them genuinely cannot compile under that set, ⛔ do not loosen the shared options — report it.

Two details from that header worth carrying: comments in these tsconfigs are // and never /* */, because a glob containing **/ closes a block comment early and a tsconfig that fails to parse falls back to compiling the whole repository rather than failing loudly. And the project is deliberately standalone (imports no workspace package, needs no build) — the sweep pins that premise, so if widening drags in a @object-ui/* import the wiring has to be reconsidered, ⛔ not silently allowed to start depending on dist.

PM mechanism assumptions — test, do not trust

1. The card measured the four configs compiling clean against this exact option set (exit=0) — 12 days ago. Re-measure before planning; if errors surface now, that is the card getting bigger, not you going off-scope.
2. I expect pnpm type-check:e2e and the turbo root task //#type-check:e2e to pick the widened include up with no workflow change, since the header says type-check dependsOn it. Verify that rather than assuming CI covers it — a gate that does not run in CI is not a gate.

Premise-first

The body is a lead, not a specification. premise_still_valid: false with no PR is a legitimate delivery — though the four readings above say the premise holds. Read the card and its one comment on GitHub yourself and check the body for sanitizer truncation.

Deliver

Draft PR against main, first body line Fixes #4477. Push the branch early. Run vitest from the repo root. Report each gate by name with its own conclusion — ⛔ never an aggregate or the required set. node scripts/check-changeset-presence.mjs and follow its verdict; ⛔ never major.

Return the JSON report and post it here as a comment whose first line is the literal <!-- os-dev-report --> marker. ⚠️ Known platform behaviour: the body sanitizer eats that HTML comment — if a read-back shows it gone, re-post with the marker as literal text. Machine-checkable fields only.

---

Generated by Claude Code