Sometimes your test runner might depend on some external files. So, you might need to specify working directory for running the test runner and/or add extra command line arguments to specify where it should find external files or even extra arguments for other stuff.

So, I suggest:

1. Make it possible to specify working directory for running test runner binary
2. Add support for specifying any kind of extra command line arguments to test runnner. For example, when using Boost.Test, you can pass extra user defined arguments after adding a double dash argument: './test_runner – --extra-arg --test-dir SOURCE_DIR'