developing-funboost-testing

Funboost's consume() runs an infinite loop, so tests need a termination mechanism. Either call os._exit(66) after a time.sleep() inside the script, or run it via subprocess.run with a timeout parameter that kills the process externally.

Create a standalone Python script that decorates a function with @boost and BoosterParams using your broker_kind, pushes several messages, starts consume(), then sleeps and exits. Verify success by reading the log output for expected print results and no exceptions.

Funboost loads funboost_config.py via importlib from sys.path at startup. Setting PYTHONPATH to the project root puts that config on the path so Redis/RabbitMQ connections resolve; otherwise the framework falls back to localhost defaults.

No, funboost uses standalone Python scripts rather than pytest. Tests live in directories like tests/ai_codes/ai_demos/ for AI-written demos and tests/ai_codes/regression_testing/ for regression checks, each with its own termination logic.

Exit code 66 only means the script survived the sleep period without crashing; it does not prove messages were consumed. You must read the console output or log files to confirm expected prints, no exception tracebacks, and matching push/consume counts.

Estimate framework startup at 5-10 seconds, then add message count divided by actual throughput (min of qps and concurrent_num divided by function duration), plus a 2-5 second buffer. Keep the total under 50 seconds.