Description
Originally reported by: BitBucket: thiefmaster, GitHub: thiefmaster
Imagine this code:
def blackbox(callback, arg):
# this is assumed to be application code
try:
callback(arg)
except Exception:
pass
def test_callback():
def cb(x):
if not x:
import pytest
pytest.fail('foo')
blackbox(cb, True)
blackbox(cb, False)
The testcase will pass since except Exception will catch the Failed raised by pytest.fail. However, I don't think there is any valid use-case for application code to prevent a testcase failure from propagating to the test runner.
A more practical example where this is a major issue is monkeypatching a library function to fail when called. For example, one might want to monkeypatch smtplib.SMTP.connect to fail when code tries to send an email unexpectedly - but chances are good that the application is wrapping that code in a try...except Exception block to prevent an email sending failure from breaking things (e.g. because the email is not very important). So now the pytest.fail called in the monkeypatched method will never propagate since it's caught by application code.