tsconfig.playwright.json

{
    // Program for the E2E / integration specs under `tests/`.
    //
    // ⚠️ IT WAS NEVER RUN BY ANYTHING, and that was not an oversight about
    // wiring — it could not run. `extends` brings the app program's settings,
    // and with no `references` the 2000+ `../../../server/…` imports in
    // `tests/**` were compiled AS SOURCE, so a single non-composite pass tried
    // to hold the whole server tree plus the app tree plus every spec. Measured:
    // 2 minutes to a heap OOM at the default ~4 GB limit. A program that dies is
    // indistinguishable from a program nobody added to CI, and a real TS1005
    // syntax error sat in `tests/e2e/core.integration.spec.ts` behind it.
    //
    // Now it REFERENCES the api project and consumes the `.d.ts` that project
    // already emits — the same change that fixed `tsconfig.tests-typecheck.json`
    // (238s → 7s) and the app pass before it. `app/**` and `workers/**` stay
    // source-compiled: only ~20 test files import from them, against ~2045 that
    // import from `server/`.
    //
    // ⚠️ `references` is NOT inherited through `extends`, and `tsc -p` does NOT
    // build referenced projects — it expects their output to exist and fails
    // TS6305 when it does not. `npm run type-check:e2e` therefore runs
    // `tsc -b tsconfig.api.json` first. Do not invoke this config directly.
    "extends": "./tsconfig.json",
    "references": [{ "path": "./tsconfig.api.json" }],
    "compilerOptions": {
        "types": ["@types/node"],
        "noEmit": true,
        "incremental": true,
        // Its own build-info file. Sharing one with the app program makes each
        // run look cold to the other, which is how a 20s check becomes a 6-minute
        // one with nothing in the diff to explain it.
        "tsBuildInfoFile": "./.tsbuildinfo.playwright"
        //
        // ⚠️ MEASURED FOOTPRINT, and why `type-check:e2e` caps the heap at 2048
        // rather than the 8192 its siblings use. With the reference in place this
        // program needs between 256 MB (OOMs) and 512 MB (passes, 6s cold); it
        // also passes on Node's DEFAULT heap, which is the exact condition that
        // used to kill it. An 8 GB cap would be 16x the need — and it would let a
        // regression HIDE: drop the `references` line or widen `include` back to
        // `tests/**` and the program silently eats gigabytes and slows to minutes
        // instead of failing. A tight cap makes that regression loud.
    },
    // ⚠️ SCOPED TO tests/e2e, not all of tests/**, and the difference is a
    // decision rather than a name. Widened to the whole test tree this program
    // reports HUNDREDS of pre-existing errors in `tests/unit/**` — missing
    // Cloudflare globals, a better-sqlite3 handle passed where a D1 one is
    // typed, drizzle insert shapes. None of that is checked by anything today
    // (vitest strips types with esbuild), so it is real debt, but it is a
    // cleanup and not this gate. Turning it all on at once would mean either a
    // red gate nobody can land against or a suppression list, and a suppressed
    // gate is the state this config was already in.
    "include": [
        "worker-configuration.d.ts",
        "tests/e2e/**/*",
        // The user-guide capture scripts. Named explicitly rather than by
        // widening to `tests/**` — that is the change the comment above forbids,
        // and this directory carries none of the debt that makes it expensive.
        // It is here at all because a file no program includes is a file that
        // type-checks vacuously forever while git and the runner both insist it
        // exists.
        "tests/docs-shots/**/*",
        "playwright*.config.ts"
    ],
    "exclude": ["node_modules", "dist", ".types"]
}