The UTF presents you with 4 different variants how it can be used.

Unlike the Buridan's donkey though, you shouldn't have problems deciding which one to use, since there are clear reasons why would you prefer each one.

In most cases to compile a test module based on the UTF all you need to include is just the single header boost/test/unit_test.hpp. This header includes internally most of the other headers that contains the UTF definitions. Some advanced features, like the floating point comparison or the logged expectations testing, are defined in independent headers and need to be included explicitly.