In the tests we could call create_app() zero or more times, which meant
that log messages produced in the test process could be duplicated on
stdout many times depending on which set of tests had been run.
What we really want is:
logging configured and going to stdout in the WSGI app (as it is now);
logging configured with no handler in the test suite, so that pytest
can capture the log messages.
In the tests we could call create_app() zero or more times, which meant
that log messages produced in the test process could be duplicated on
stdout many times depending on which set of tests had been run.
What we really want is:
can capture the log messages.