I think that it is much clearer if the client test fails when no server is running rather than having a test which simply states pass whereas the client is talking to a wall, and then the user adds the --test_output=all and discovers that the test is green whereas the log contains an horrible error message.
Plus, it is an opportunity to talking about test caching.
I think that it is much clearer if the client test fails when no server is running rather than having a test which simply states pass whereas the client is talking to a wall, and then the user adds the
--test_output=all
and discovers that the test is green whereas the log contains an horrible error message. Plus, it is an opportunity to talking about test caching.