Testing

Testing and debugging PyLith can be challenging due its many dependencies and complex interaction with PETSc. Our strategy is to test at a variety of levels to isolate bugs close to their origin while also building a comprehensive suite of tests. We use C++ unit tests for verifying small pieces of code, such as individual C++ class methods, work as intended. We use the Method of Manufactured Solutions for verifying implementation of governing equations and their solution using PETSc. We use full-scale tests for verifying integration of complete simulations in both serial and parallel. When an example, user simulation, or one of the tests indicates a bug exists, we leverage the test suite to find the bugs. In general, the most efficient strategy for debugging is to first try to expose the bug in a serial unit test, followed by an MMS Test, a serial full-scale test, and finally a parallel full-scale test. This may require creating new tests if the bug is not exposed by current tests. The PyLith developers make extensive use of debuggers, such as gdb and lldb, and memory management analysis tools, such as valgrind, to detect and squash bugs. These are discussed in Debugging tools.

• C++ unit tests
  • CppUnit macros
• Method of Manufactured Solutions
  • Debugging residual errors
  • Debugging Jacobian errors
  • Example
• Running C++ unit tests and MMS tests
  • Running C++ unit tests
• Running MMS tests
  • Using the debugger
  • Using valgrind
• Python unit tests
• Full-scale tests
  • Command line arguments
  • Using the debugger
  • Using valgrind
  • Example
• Debugging tools
  • VS Code integration
    • C++ unit test
    • MMS test
    • PyLith simulation
  • Debugger quick reference
  • Valgrind quick reference
• Viewing fields
  • Viewing differences
• Running tests in CI Docker containers