* test(e2e): fix type errors in four Electron E2E specs
- category-management: the poll's message and 15 s timeout were passed to
toBe(), which takes one argument, so the poll ran with the default 10 s
timeout and message. Pass them to expect.poll().
- settings: evaluate<void, HTMLVideoElement> set the argument type, not the
element type; type the callback parameter instead.
- remote-control: keep the poll callback's writes visible to the return and
fail with a clear error instead of casting null.
- downloads: read the dataset key through its index signature.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): fix type errors in the benchmark harness
None of these changes behaviour for a capture the harness produces today:
- cancellation report: a null playlist id on the request record now means
unknown, like undefined, instead of matching only null DB records.
- main capture: type the worker transport map so its unavailable-reason
literals stay narrow.
- worker request performance: a parsed capture carries non-null work
timestamps; only the unavailable fallback reports null.
- Xtream iteration: hand the teardown closure the narrowed app and data
directory.
- Xtream exact schema: return early when the identity is not an exact
record instead of re-checking it three times.
- Xtream assembly: reject an unpainted UI explicitly; the null-to-0
comparison already rejected it.
- Xtream scenario driver: drop the always-null cancellation observer from
prepare; trigger() installs it.
- Electron process lifecycle: type the child process by the exit
subscription it uses, so a test double need not return a ChildProcess.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): fix type errors in the benchmark harness specs
The dynamically imported helpers were checked with
assert.equal(typeof fn, 'function'), which does not narrow, so later calls
used fn?.(). assert.ok(typeof fn === 'function') narrows them. The rest
types fixtures by the module under test, keeps closure-assigned callbacks
visible after their callbacks run, and narrows nullable fixture values.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): type-check the whole electron-backend-e2e project
Widen the spec config's include from the journey harness to src/**/*.ts
and the root Playwright configs, with no baseline or ignore list, and
update the validation map and the journey checklist.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(perf): build a complete renderer fixture in the request-performance spec
The fixture was cast through unknown, so the type check accepted an
incomplete RendererCaptureMetrics. Fill every required field with neutral
values; the summary reads only the probe arrays and heap fields.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Nothing type-checked apps/electron-backend-e2e: test-performance-harness
runs its specs through `tsx --test` and Playwright transpiles the journeys
without types, so the J3 fixture drifted from JourneyMainIpcCaptureState
(TS2739, fixed in #1797) while CI stayed green.
Add apps/electron-backend-e2e/tsconfig.spec.json. `typecheck:spec`
discovers every tsconfig.spec.json, so the `unit-and-typecheck` job picks
it up without a workflow change. It follows the spec-program convention
(module preserve, bundler resolution) and sets target es2022 and lib
es2022 + dom, because the project's commonjs module and the base es2015
target reject the harness's import.meta, BigInt literals and ES2022 names.
`types: ["node"]` keeps Jest and Mocha globals out of a node:test program.
The include starts with the journey harness (src/journeys/** and
src/performance/*journey*.ts), which type-checks cleanly; reverting the
#1797 fixture reproduces the TS2739. The older benchmark files and the
Electron E2E specs still carry type errors; the include widens as they
are fixed, with no baseline or ignore list.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
