Chrome Remote Desktop Testing Best Practices

When creating or updating tests in //remoting, follow these rules to ensure stability and avoid flakiness:

• Do Not Mask Failures: Never modify a test simply to hide or bypass a legitimate underlying Chromium bug.
• Print Debugging: If you are stuck or do not understand why a test is failing, do not guess. Add LOG(ERROR) << "State: " << variable; statements, rebuild, and re-run the test to empirically observe the state.
• Unit Tests & TDD: New functionality should always have unit tests, and Test-Driven Development (TDD) should be used when possible. However, CRD is very complex and often calls directly into OS APIs, making unit testing difficult or sometimes impossible. Apply judgment.
• Asserts: Avoid YODA style test asserts.
• TaskEnvironment: Prefer base::test::TaskEnvironment over raw threads.
• Avoid RunLoops: Prefer base::test::TestFuture and base::test::RunOnceCallback (from base/test/gmock_callback_support.h) over manual RunLoops.
• RunUntilIdle is Banned: Do NOT use RunUntilIdle(). It is non-deterministic.
• Observing Completion:
  • Explicitly quit run loops using run_loop.Quit() or run_loop.QuitClosure() if completion is observable via a callback.
  • Otherwise, wait for a condition via base::test::RunUntil().
• Flakes: If a test succeeds after a retry, investigate and fix the root cause. remoting_unittests should run in <10s; longer runs likely indicate a hang.

Technical Reference

Gmock Callback Support (base/test/gmock_callback_support.h)

Provides RunOnceCallback<N>(args...) for mocking callbacks that take base::OnceCallback. Use it in WillOnce() to provide results asynchronously.

Test Future (base/test/test_future.h)

A modern replacement for RunLoop. Use future.GetCallback() to pass to async methods and future.Get() or future.Take() to wait for and retrieve results.

Task Environment (base/test/task_environment.h)

Enables the use of base::ThreadPool and RunLoop in tests. Prefer SingleThreadTaskEnvironment when multi-threading is not required.