Join GitHub today
Currently on Travis, the WML unit tests can take a significant amount of time to complete since they each individual unit test starts up its own instance of Wesnoth to run that single test. Given that there are a decent number of tests, it can end up taking 10+ minutes for them to complete, which has caused travis timeouts occasionally.
It is done this way, since currently the WML tests script uses Wesnoth's return code to determine if the test returned the expected value.
The change would then be to instead allow specifying multiple tests to run, and then outputting the results of each test after the final test runs.
I think probably the only required C++ change needed would be to add support for running multiple tests in a row - maybe just a huge list at the command line or read from a config file - and then the existing unit tests could be updated to instead log their output to stdout, use PersistenceWML, etc.
The results could also be directly output into a csv file, I suppose.