Adrian Colyer is talking about writing robust pointcut expressions.
This seems to be something that SEEMS simple at first, but actually requires some thought.
One of the problems is that you can easily get away with an expression that “works” in the sense that for your little test, at this moment in time, it does what you were thinking.
This ability can lead to hacking :)
It is like mucking with Regex when you first get started. You often have a couple of cases, and you muck around with the regex line noise until it works for those 2 cases. However, in the future, it breaks on another 5 cases.
Does TDD and process help? Yes and No. It can help, but it all comes down to how lazy you are being, and how well you are thinking. If you have good tests to begin with it can definitely help. “Oops, I changed this pointcut and now this test isn’t working”. Future-proofing isn’t easy, but it shows that it IS OK to think a little about your application, instead of whipping out a solution that passes a test in front of your eyes.