When running tests on a Windows platform, the CppUnit
test runner instantiates all of the test fixtures
simultaneously, before the selection of which tests
are to be run. At a minimum, I would expect that only
the tests selected for a test run would have their
fixtures instantiated. I would consider it good
behavior to have each test's fixture set up
immediately before it is run, and destroyed
immediately afterwards. While this is not a problem
for many tests, I have been testing internally
threaded classes, and the current behavior is a
problem for obvious reasons.
Log in to post a comment.