When you touch a class or method in your baltik (or in TRUST itself), you usually don't need to run the full non-regression suite — only the test cases that exercise the code path you changed. The trust -check flag has a coverage variant that selects test cases by their C++ symbol coverage, letting you re-run exactly the relevant ones in seconds rather than hours.

The -check coverage forms

In addition to running named test cases, trust -check accepts a function, class or class::method name. It looks up which test cases exercise that symbol (via the coverage database built by the nightly CI) and runs only those:

trust -check FUNCTION_NAME
trust -check CLASS_NAME
trust -check CLASS_NAME::METHOD_NAME

-ctest accepts the same forms and routes through the CTest harness when you prefer that runner:

trust -ctest CLASS_NAME::METHOD_NAME

Worked example: the RRK2 rational Runge-Kutta scheme

Say you've modified the RRK2 class (rational Runge-Kutta order 2) and you want to re-run only the test cases that touch it. Find the class in the Doxygen documentation first — that confirms the public methods and gives you their exact spellings. Then:

trust -check RRK2::RRK2 # only the cases hitting the constructor

trust -check RRK2 # every case hitting any RRK2 method

The first invocation is the laser-focused one: it re-runs only the test cases whose execution actually entered RRK2::RRK2. The second is broader and useful as a smoke test of the class as a whole before opening a pull request.

When to use coverage-driven selection

Before a commit, to confirm your fix doesn't regress any of the test cases that exercise the modified code.
During bisection when narrowing down which test exposes a regression.
In CI for fast feedback on pull-request branches before running the full nightly battery.

The coverage information is populated by the periodic CI runs that collect gcov data on the test suite; if a recently-added test case doesn't yet appear in the selection, the next coverage run will pick it up.