Skip to content

E2E-Tests.md

Latest commit

 

History

History
135 lines (103 loc) · 4.35 KB

E2E-Tests.md

File metadata and controls

135 lines (103 loc) · 4.35 KB
Raw

Local & CI E2E tests

Our dev setup and the nature of the SDK required test runs as real on-device integration tests.

To make these tests CI-friendly, our test framework streams structured test events from the app to a Node “collector” which produces JUnit XML + raw artifacts.

Packages

The packages are part of our monorepo. They are, however, ready for an individual publishing if/when we decide to reuse the framework in other SDKs.

  • packages/mobile-testbed/
    • Test framework (TestSuite/TestRunner + TestEvent/TestMonitor).
  • packages/mobile-test-reporter/
    • In app reporter(s) that implement TestMonitor and send events to the collector over HTTP.
  • packages/mobile-test-runner/
    • Node collector CLI that aggregates events and writes JUnit + artifacts, both locally and on CI.

How it works

  1. CI starts the collector (mobile-test-runner collect ...).
  2. The mobile app runs test suites with TestRunner and emits TestEvents to a TestMonitorGroup.
  3. HttpTestReporter buffers events and POSTs them to the collector.
  4. When tests finish, the app sends an explicit run completion request.
  5. The collector writes:
    • artifacts/e2e/junit.xml
    • artifacts/e2e/events.jsonl
    • artifacts/e2e/summary.json
    • artifacts/e2e/runs/<runId>/* (run metadata + raw events)
    • Logs from the Android and iOS devices

Running locally

  1. Build the infra packages (so dist/ exists):
yarn e2e:infra:build

or

yarn workspace mobile-testbed build
yarn workspace mobile-test-reporter build
yarn workspace mobile-test-runner build
  1. Add collector URL to testapp/.env:
  • TEST_COLLECTOR_URL=http://127.0.0.1:8137 (the URL must include the scheme)
  1. Start the collector (in a separate terminal):
mkdir -p artifacts/e2e
node packages/mobile-test-runner/dist/cli.js collect --host 127.0.0.1 --port 8137 --out artifacts/e2e --expected-runs 1 --timeout 30m
  1. Run the app on device/simulator (tests auto-start in TestExecutor):
yarn runReactAndroid
# or
yarn runReactIos

OR

yarn freshCordovaAndroid
# or
yarn freshCordovaIos

When the run completes, check:

  • artifacts/e2e/junit.xml
  • artifacts/e2e/events.jsonl

If you want to run both platforms and have the collector exit only after both complete:

mkdir -p artifacts/e2e
node packages/mobile-test-runner/dist/cli.js collect --host 127.0.0.1 --port 8137 --out artifacts/e2e --expected-runs 2 --timeout 45m

Then run both:

yarn runReactAndroid
yarn runReactIos

Alternatively, you can run it as "automation" with the following commands:

yarn e2e:local:rn
# or
yarn e2e:local:cordova
# or
yarn e2e:local:full

Watchmode

The collector supports an "indefinite" watch mode when running test locally. It ignores the --expected-runs and --timeout params. You can run it with:

node packages/mobile-test-runner/dist/cli.js collect --host 127.0.0.1 --port 8137 --out artifacts/e2e --watch

CI workflow

Workflow file: .github/workflows/mobile-e2e.yml

Mobile CI is the single PR integration workflow. It builds and runs the test applications, so the former standalone Cordova and React Native integration build workflows are no longer needed.

PRs and pushes to develop or release/* run the four E2E checks independently:

  • e2e-android-rn
  • e2e-android-cordova
  • e2e-ios-rn
  • e2e-ios-cordova

When a new commit is pushed to a same-repository PR, GitHub creates a visible workflow run for the pull_request: synchronize event. Its first attempt is cancelled by the first step of each E2E job before checkout, dependency installation, emulator, simulator, or build work starts. The cancellation log explains that this saves CI time and that the job can be rerun. This avoids spending CI time on a commit that may immediately be superseded. The cancelled run remains available in the Actions and PR checks UI, and each E2E job can be rerun independently.

To run the checks for that exact commit, open the cancelled workflow run and choose Re-run all jobs, or rerun an individual job. Reruns use the original PR commit and bypass the first-attempt cancellation gate. Initial PR runs (opened, reopened, or marked ready for review), pushes, and manual workflow_dispatch runs execute normally. Fork pull requests are skipped to avoid running untrusted code with the workflow's build and test resources.