Tests

The tests use Playwright. There are API tests, which call the HTTP API, and UI tests, which drive a browser. They run in Chromium.

Run the full suite

The full suite runs in Docker. It builds the app, starts it with a fresh database, runs the tests and removes everything afterwards.

Install the browser

yarn test:install

Start Docker

The test scripts need a running Docker daemon.

Run the tests

yarn test

This runs scripts/test-e2e-sharded.sh with 4 shards. Each shard gets its own app container, starting at port 3123, and all shards share one test database server. The logs go to logs/test-results/.

Open the report

yarn test:report

More shards finish faster but use more memory and CPU:

yarn test:sharded:3
yarn test:sharded:5
yarn test:sharded       # 10 shards
bash scripts/test-e2e-sharded.sh 7

To run everything against one app container instead, use yarn test:single. To pick tests in Playwright's UI mode, use yarn test:ui. Both use docker/docker-compose.local.yml.

Run tests against your own stack

While you work on a test, it is faster to keep the app running and run only that test.

Set the test values in `.env`

The tests need the test helpers turned on and the same tokens the test scripts use:

.env

ENABLE_TEST_ENDPOINTS=true
SERVER_TOKEN=server123
API_TOKENS=ci-admin:ci-admin-token-0123456789abcdef
API_TOKENS_READONLY=ci-readonly:ci-readonly-token-0123456789abcdef

Only use these values on a test install. ENABLE_TEST_ENDPOINTS turns on /api/test/*, which can sign anyone in as admin.

Start the app

yarn docker:local:up

The tests expect the app on http://localhost:3069. Set PLAYWRIGHT_BASE_URL to use another address.

Run the tests you want

SKIP_WEBSERVER=1 yarn test:api
SKIP_WEBSERVER=1 yarn test:manual tests/api/veto.spec.ts
SKIP_WEBSERVER=1 yarn test:manual --grep "@veto"

SKIP_WEBSERVER=1 stops Playwright from starting its own server.

Other scripts: yarn test:ui:manual and yarn test:api:manual open Playwright's UI mode, yarn test:veto and yarn test:cs-major run tagged groups.

Write a test

The test layout:

tests/api/: API tests.
tests/ui/: UI tests.
tests/helpers/: shared helpers for sign-in, teams, servers, veto and the database.
tests/setup.spec.ts: global setup, runs first.

Things to know:

Sign in on both sides. Playwright's page and request have separate cookies, and most helpers use request. Call setupTestContext(page, request) in beforeEach, which signs in on both.
Act as a player for the veto. The veto checks the caller's Steam ID, and an admin is not on either team. The veto helpers use admin impersonation (impersonatePlayer, stopImpersonating) for you. The veto board is on the player page, /player/:steamId.
Use fake servers. A server with host 0.0.0.0 answers RCON with canned replies and is always online. createTestServer does this. Any other host makes the tournament start check try a real RCON call, which fails.
Do not skip. Do not write if (!visible) test.skip(). Assert instead, so a missing element fails the run.
Tag tests with @api, @ui, @crud, @veto, @cs-major or @auth, so they can be filtered with --grep.
Keep files small. Split a test file when it grows past 300 to 400 lines.

tests/README.md in the repository has examples for each helper.

In CI

On each pull request, GitHub Actions runs lint, the typecheck ratchet, the i18n key check, the API docs check, and the E2E suite in 4 parallel jobs. Pull requests that only change docs skip the E2E jobs.

On this page