py.test ignores warnings.simplefilter

[woodscn]

I have a parametrized test that should emit a warning under certain conditions. After setting simplefilter('always'), py.test still only sees the first instance of the warning. Executing the same test in the interactive interpreter (as part of a for loop) correctly warns the user each time.

In the following simple example, for whatever reason, py.test doesn't capture any of the warnings at all.

from warnings import warn, simplefilter
import pytest

simplefilter('always', MyWarning)

class MyWarning(UserWarning):
    pass

def warn_the_user(ind):
    warn("Iteration # " + str(ind), MyWarning)
    return ind

@pytest.mark.parametrize('ind', range(10))
def test_warn_the_user(ind):
    warn_the_user(ind)

$ py.test pytests_warning.py
============================= test session starts ==============================
platform darwin -- Python 3.4.3 -- py-1.4.30 -- pytest-2.7.2
rootdir: ~, inifile:
collected 10 items

pytests_warning.py ..........

========================== 10 passed in 0.01 seconds ===========================

[RonnyPfannschmidt]

you are not at all using the pytest integration

did ypu perhaps mean to do a:

from warnings import warn, simplefilter
import pytest

class MyWarning(UserWarning):
    pass

def warn_the_user(ind):
    warn("Iteration # " + str(ind), MyWarning)
    return ind

@pytest.mark.parametrize('ind', range(10))
def test_warn_the_user(ind):
   with pytest.warns(MyWarning):
        warn_the_user(ind)

note that we discovered a bug in 2.8.0 wrt warnings, so you might need to try the 2.7 series until 2.8.1 is released
however since you use a custom warnings class it might be alright,
however you might need to still add in the simplefilter line

see #1041 for details

[woodscn]

I'm more concerned that pytest seems to be overriding my simplefilter. Does pytest ignore the usual Python warnings.warn function by default?
I'm using 2.7.2, and pytest.warns does not seem to be present.

[nicoddemus]

xref: #2452

[Zac-HD]

Since pytest 3.8, we use the standard Python warnings mechanism. #3954 and #2586 document this problem with the current APIs, so I'm closing this issue as a duplicate (even though it was first!).