However if you start doing arbitrary rewrites of code aimed at breaking the test suite you're no longer doing TDD :-)

This isn't really surprising since you should not be writing code with TDD that isn't being exercised by a failing test.

No one is talking about arbitrary rewrites. Refactorings can easily result in green-bar code that wasn't immediately motivated by failing tests and that contains untested branches.

This doesn't mean that I think more traditional white box testing, branch coverage, statement coverage, etc. are useless - far from it. They're excellent tools and you can find many bugs with them.

Seems we aren't really in much disagreement after all. My original point was only that too often I've seen TDD adopted "at the expense" of more traditional testing methods rather than as a complementary "design" process.

Refactorings can easily result in green-bar code that wasn't immediately motivated by failing tests and that contains untested branches.

Fair point. Misunderstood what you were getting at.

That said, it's still been my experience that doing TDD and merciless refactoring produces good branch coverage more easily that post-code test writing. My hunch would be that refactorings that introduce untested branches are more than outweighed by those that remove branches or cause duplicate coverage by different test. Just a hunch tho'.

Seems we aren't really in much disagreement after all.

No, I don't think so ;-)

My original point was only that too often I've seen TDD adopted "at the expense" of more traditional testing methods rather than as a complementary "design" process.

Bad things can and do happen if people jump into a new development style and abandon an old one. The problem I see most with people adopting TDD is, well, not doing TDD :-). It's so very easy for a newbie to write code that "obviously" works, or fail to write tests for something that is hard to test rather than re-writing the code to make it easier to test.

Traditional testing methods like looking at code and branch coverage can actually be great tools for supporting TDD. Its been my experience that if you've got low code coverage then you almost certainly are not doing TDD properly, so it's great feedback into the process of adopting and maintaining good TDD practices.

Of course with infinite time and money you do every kind of testing practice that you think will benefit you. Unfortunately that's not a common scenario. My personal experience is that if you have a team doing TDD well then it's better to spend your testing resources in the acceptance/customer tests and in exploratory testing rather than more formal whitebox techniques. As ever YMMV ;-)