Technical Guide: Anti-Flakiness
This document outlines specific technical solutions that uphold the BDR value of “Infinite Reproducibility”.
1. The Problem: Intermittent Failures
A test that fails “sometimes” without a code change destroys trust in the report. If we cannot trust the report, BDR loses its meaning.
2. Smart Waits Strategy
We do not use static delays (e.g., Wait.5s()). Instead, Layer 0 (Technical) must implement:
- Auto-waiting: Waiting for an element to be ready for action.
- State-driven checks: Waiting for a system state change, not just element appearance.
3. Infrastructure Health-checks
Before blaming the test code, we must ensure the “patient is alive”.
- Global Setup: Before running the test suite, we verify the availability of critical services (DB, API, Auth).
- Fail Fast: If infrastructure is down, the run must fail immediately. This protects the “reputation” of tests from failures caused by environment instability.
4. Data Isolation
Every test in BDR must be independent.
- Fresh Users: Create unique data for each run.
- Cleanup: Automatic data cleanup after completion (via hooks or Teardown mechanisms).
5. Rich Debugging in Reports
To eliminate the cause of flakiness, the report must provide:
- Trace Viewer: Step-by-step execution playback.
- Video/Screenshots: Visual confirmation of DOM state at the moment of failure.
- Network Logs: Statuses of all API requests.
This guide is part of the implementation of Layer 0 and Layer 1 strategies.