Electron E2E

This suite covers the renderer-to-main process boundary in the packaged Electron app: the contextBridge API surface, IPC handler wiring, window lifecycle, and contextIsolation enforcement. Regressions are security-relevant. The bridge is the only sanctioned path from renderer JavaScript to Node APIs, and a leaked Node global widens the blast radius of any future renderer XSS.

Trading flows are not re-tested here. The web Playwright E2E suite covers them once against the same Chromium engine that Electron ships. This suite focuses on behaviour that is only true under Electron.

Location: frontend/tests-electron/ Config: frontend/playwright.electron.config.ts Run: cd frontend && npm run test:electron Prerequisite: npm run electron:build (produces dist/ and dist-electron/)

What it verifies

1. Window opens with the correct title.
2. Startup overlay is displayed on launch (the loading screen shown while the dashboard mounts).
3. Dashboard mounts after startup completes.
4. contextBridge API: window.electronAPI is exposed to the renderer with exactly the methods the main process intends to share.
5. contextIsolation is enforced: no Node.js globals (require, process, __dirname) leak into the renderer. A renderer-side XSS therefore cannot require("fs") and read the user’s disk.
6. IPC handlers: minimize, maximize, and close calls reach the main process and act on the BrowserWindow.
7. Pop-out windows open for localhost URLs (used by the detached order-ticket window).

How it launches

Playwright’s _electron fixture (from playwright, not @playwright/test) launches the compiled dist-electron/main.js:

import { _electron as electron } from "playwright";

const app = await electron.launch({
  args: ["--no-sandbox", "--disable-gpu", "dist-electron/main.js"],
  env: { NODE_ENV: "test" },
});

const window = await app.firstWindow();
await expect(window).toHaveTitle(/VETA/);

NODE_ENV=test disables the dev-tools toggle and the auto-update check. ELECTRON_RUN_AS_NODE is explicitly unset; if it leaks through from the parent process, Electron starts as a Node shell rather than a desktop app and tests fail in confusing ways.

Test execution flow

graph TD
    A["npm run test:electron"] --> B["Playwright config loaded<br/>(playwright.electron.config.ts)"]
    B --> C["test.beforeAll:<br/>ElectronMockServer starts<br/>on a random port"]
    C --> D["_electron.launch():<br/>spawn dist-electron/main.js"]
    D --> E["Electron main process starts<br/>NODE_ENV=test"]
    E --> F["main process creates BrowserWindow<br/>+ wires IPC handlers"]
    F --> G["preload.js exposes electronAPI<br/>via contextBridge"]
    G --> H["app.firstWindow()<br/>returns Page handle"]
    H --> I["test body asserts:<br/>title, contextBridge,<br/>contextIsolation, IPC, pop-out"]
    I --> J{Pass?}
    J -->|Yes| K["app.close()<br/>+ mock server shut down"]
    J -->|No| L["trace + screenshot,<br/>retry once in CI"]

Workers are set to 1 because only one packaged app instance can run at a time without port collisions. app.firstWindow() waits for the renderer to mount before the test body runs, so per-test setup is implicit.

Mocking the backend

tests-electron/helpers/ElectronMockServer.ts boots a small HTTP server inside the test that the Electron app talks to instead of the real backend. This keeps the suite independent of whether the production server is up. The mock server is narrower than GatewayMock from the Playwright suite; Electron tests verify desktop-specific plumbing rather than UI flows.

Config

• Workers: 1. Electron tests run serially because only one packaged app instance can run at a time without port collisions.
• Test timeout: 60s per test (cold-start can be slow).
• Global timeout: 10 minutes overall (xvfb startup plus cold start dominates CI time).
• Retries: 1 in CI, 0 locally.
• Trace: on-first-retry. Screenshot: only-on-failure.

What is not here

Trading flows are not duplicated from the web Playwright suite. Electron is the same Chromium renderer with extra plumbing; doubling up trade-flow tests would double maintenance for no extra signal.

Features that depend on electronAPI (pop-out windows, native menus, native notifications) are tested here because they do not exist in the browser.

Adding a test

1. Create tests-electron/my-feature.spec.ts.
2. Use the standard @playwright/test test(). The role fixtures from the web suite do not apply; Electron has its own setup.
3. Launch the app in a test.beforeAll and close it in test.afterAll. Per-test cold start would balloon the suite duration.
4. If you need backend behaviour the mock server does not expose yet, extend ElectronMockServer. Do not reach across to GatewayMock; the two suites are intentionally separate.