Long posts

@schmidt_fu Brian Marick, 1997, "Classic Testing Mistakes", from "Testing Foundations: Consulting in Software Testing". Addresses more the notion of a test team and QA.

Useful ideas:

• If the UI is limited, add a test interface to allow introspecting better.
• When a regression happens, poke around and add more tests, not just   a regression test. Bugs tend to move in swarms.
• Take notes; the same problems tend to recur, and you can reuse the general   testing approach again and again.
• Automation vs. manual testing follows a break-even logic. If it takes you   longer to automate a test than it does to just run the test yourself   the limited number of times it will run, don't automate.   - Record and replay systems can reduce the automation cost, but are often     hampered in practice such that they aren't worth using at all.   - Scripted tests, using an API, are more easily maintained, though now you're     coding up a test API. And API design remains hard.   - Data-driven tests can be very robust to UI changes; they are also often     more comprehensible for domain experts.   * WARNING: Automated UI tests cannot be relied upon to test all routes to     a view; you'll have to rely on lower-level (easily broken) automated tests,     or manual testing.
• Code coverage: "Coverage has told you where your tests are weak, but it's up   to you to understand how." Branch coverage can be more useful than raw   statement coverage info.
• Start stress and load testing ASAP.