I am currently using

__defaults__(
    {
        (
            python_sources,
            python_source,
            python_test,
            python_tests,
            python_test_utils,
        ): {"run_goal_use_sandbox": False}
    }
)

in a part of my project. Now I have a issue that if I run a python source it works perfectly fine. All files I require during runtime (relative to the source) are found. If I run

pants test ::

The

python_test

,

python_tests

, and

python test_utils

goals are still run in a sandbox. Hence, the tests fail because the paths can not be found. Does anyone have a solution for this issue?

Can you try it with putting it inside the pants.toml? If that works then the bug is in the

___defaults___

[python]
run_goal_use_sandbox = False

Tried it with

default_run_goal_use_sandbox = false

, same issue.

Did you mean you tried with what I suggested? Where did you find

efault_run_goal_use_sandbox = false

in docs?

[python]
default_run_goal_use_sandbox = false

This is the setting for the global python config: https://www.pantsbuild.org/docs/reference-python#basic-options

I was suggesting based on this https://www.pantsbuild.org/docs/reference-python_tests#coderun_goal_use_sandboxcode Although now I don't know why both of these options exist

They are the same. Yours is for a goal based configuration the other implies it being the default for all python goals.

What I am bit confused is it stresses

run

Run is generally used for running some application/binary

pants run <some.pex>

True, but if i can run a file without sandbox mode I should be able to also test it without sandbox mode.

Yeah so maybe there is some other way for it or maybe not. The option you are using is only for runnable targets

Yes, the docs state it exists: https://www.pantsbuild.org/docs/reference-python_tests

They most probably copy some internal doc template and don't distinguish between runnable target and non-runnable ones.

You mean the documentation being wrong, or the behaviour?

Doc being wrong. As you said the source code doesn't have the implementation

I am not sure how the test runner handles

python_sources

or if they are handled at all. But as a user I would expect that if I set

run_goal_use_sandbox = False

in a

python_source

that

python_test

would respect that configuration.

But I think you are confusing between what is

run

and what is

test

as a goal.

run

is only applicable for

python_sources

. The

python_test

doc shouldn't mention its usage. Whether such feature should exist or not, I would agree it should but it would be some other option like you said

use_sandbox

or something

Yes, I know, but

test

as a goal should test runnable goals. I will open an issue.

So this is actually not a bug, and is all as designed, although for the test goals it could maybe be better called out. Here's the gist: • The run goal is analogous to running a python file

python <file.py>

. You can run any python files including test ones, so therefore it inherits the field • The test goal specifically maps to running pytest on the file. And therefore the code isn't being run via the run goal and doesn't inherit this flag • The run goal runs the process, no matter what. Therefore it's ok to run in your source tree, since we aren't trying to cache any results. • The test goal is always run in the sandbox, because the results are cached. If we allowed running outside the sandbox, those cached values could be inconsistent with the state of your files, and that breaks a fundamental guarantee of Pants.

I get that it is designed that way. Now the question is how do I test my code. It runs perfectly in non sandbox mode, as it is supposed to. The test will always fail. I could try to include the required files at the required locations to make it testable. This seems quite hacky…

If your test is complaining about missing files, those files, and the dependency from your code to those files must be made known to Pants. That's how pants knows when the cache key for the test is invalidated (and what files belong in the sandbox)

Ok! So I will just add the files as resources for the tests.