Running Pants 2.16.0 on Ubuntu 22.04. I have a test that sets an environment variable using

os.environ["VAR_NAME"] = "some value"

. This isn't reflected when the test is ran, even when

VAR_NAME

is included without a value in

extra_env_vars

in the BUILD file for the tests. Is this intended behavior?

Morten, where is the

os.environ["VAR_NAME"] = "some value"

occurring? Could you provide some more details? If you are running a python process and it sets the os.environ and then reads it later, I can't imagine that pants can get in the way of that.

def test_example():
    os.environ["VAR_NAME"] = "<https://test.com>"
    settings = Settings()
    assert settings.enable_sink is True

Here is the test. The build file looks like this:

python_tests(
    name="tests",
    extra_env_vars=[
        "VAR_NAME",
    ]
)

Isn't that setting so you can transmit VAR_NAME from the caller into the test. Like

def test_example():
   assert os.environ['VAR_NAME'] == 'good'

and you can run

$ VAR_NAME=good pants test path/to/the/test.py

Yes, the call to

Settings()

which internally calls

pydantic_settings.BaseSettings

reads the environment variables 🙂

Yes. It is possible that on import pydantic_settings has already read the os.environ and doesn't observe you updating it.

Jeez, it was 100% my own fault. I was running

settings = Settings()

in the file I was importing from. This means that when I ran my test this would be run twice, the first time before my monkeypatch had a chance to take effect, and then again in my test, now with patched values.

Man, np, this happens all the time. Def done it myself