IMHO, there're two points here;
1. Unit testing is not TDD
2. TDD has nothing to do with software testing
TDD is good for developing rapid changing codebases with short development cycles with known requirements and a harsh time pressure, but it's a front loaded way as well.
By which I mean, you have to invest some hours beforehand to make sure you're not shipping crap because you didn't have time to see what would break or to make sure you didn't skipped the controller your GUI developer needed.
TDD tries to ensure one thing only: "Awareness". You accept the initial costs of TDD if awareness is a big issue for you. Otherwise, use something else which serves you to solve it.
If you're aware of what you will be implementing and if there's a mechanism you can check your code against, it'd be efficient, right? TDD uses unit tests for documenting that "Awareness", aiming to utilize the benefits of unit testing as well.
You're right, that it is not feasible trying to test every aspect of your solution via unit tests and that's why there are integration tests, systems tests and acceptance tests. So if you're planning to find defects via the unit tests you're writing for your TDD cycle, think again. Unit tests are good for checking "completeness" and a great tool helping with regressions. Nothing more and nothing less.
As a good engineering practice, we adopt the method of working, based on how we planned to work. If we find TDD fruitful to implement, we avoid putting any logic to controllers. We implement our business logic behind the controller, which also cleans it off of implementation specific crap. Besides us using a completely detached GUI layer helps us write controller unit tests running via HTTP.
We simplified our working environment and make it suitable for TDD, by ways we found to provide more capabilities to us. If we're not to use TDD, we employ other logics and structures, which would fit well with the method we'd use instead.
Long story short, TDD gives what TDD is intended for and as long as you cooperate. If you expect more than it can provide, you'd be disappointed. Regardless of your development method, you have to make sure your development/architecture models and your tool set complies with your method of choice. The rest relies in the question "What do you need your development model to solve for you?"
TDD is good for developing rapid changing codebases with short development cycles with known requirements and a harsh time pressure, but it's a front loaded way as well.
By which I mean, you have to invest some hours beforehand to make sure you're not shipping crap because you didn't have time to see what would break or to make sure you didn't skipped the controller your GUI developer needed.
TDD tries to ensure one thing only: "Awareness". You accept the initial costs of TDD if awareness is a big issue for you. Otherwise, use something else which serves you to solve it.
If you're aware of what you will be implementing and if there's a mechanism you can check your code against, it'd be efficient, right? TDD uses unit tests for documenting that "Awareness", aiming to utilize the benefits of unit testing as well.
You're right, that it is not feasible trying to test every aspect of your solution via unit tests and that's why there are integration tests, systems tests and acceptance tests. So if you're planning to find defects via the unit tests you're writing for your TDD cycle, think again. Unit tests are good for checking "completeness" and a great tool helping with regressions. Nothing more and nothing less.
As a good engineering practice, we adopt the method of working, based on how we planned to work. If we find TDD fruitful to implement, we avoid putting any logic to controllers. We implement our business logic behind the controller, which also cleans it off of implementation specific crap. Besides us using a completely detached GUI layer helps us write controller unit tests running via HTTP.
We simplified our working environment and make it suitable for TDD, by ways we found to provide more capabilities to us. If we're not to use TDD, we employ other logics and structures, which would fit well with the method we'd use instead.
Long story short, TDD gives what TDD is intended for and as long as you cooperate. If you expect more than it can provide, you'd be disappointed. Regardless of your development method, you have to make sure your development/architecture models and your tool set complies with your method of choice. The rest relies in the question "What do you need your development model to solve for you?"
All the best