Hopefully an easy one. Trying to

.pants test

a directory full of, well, tests. The

BUILD

has

sources=["***/**.py"

and includes all the test files. Except that there are a couple of files in there without actual tests; they're imported by the test files. Running the test target works, but those files print

=== no tests ran in 0.01s ===

and are listed by pants as "failed", ie

❯ ./pants test --force some/file/without/tests.py
/home/vputz/.cache/pants/setup/bootstrap-Linux-x86_64/2.4.0.dev2+gitd46197d0_py38
15:29:46.22 [INFO] initializing scheduler...
15:29:46.55 [INFO] scheduler initialized.

𐄂 some/file/without/tests.py:../../tests failed.

within an appropriate venv,

❯ pytest some/file/without/tests.py
======================================================== test session starts =========================================================
platform linux -- Python 3.9.0+, pytest-6.2.2, py-1.10.0, pluggy-0.13.1
rootdir: /root/dir/wotever
plugins: some-plugins
collected 0 items                                                                                                                    

======================================================= no tests ran in 0.01s ========================================================

How can I avoid a "failed" with a "no tests ran" result--or skip the test running on files without tests (while still including them for dependencies)? If I

.pants test the_actual_test_target

it runs all the tests but lists the files without tests as "failed" (and thus the test target fails). If I

pytest

the directory representing the test target it doesn't even pick out the files without tests.

The 

BUILD

 has 

sources=["***/**.py"

  and includes all the test files.  Except that there are a couple of files in there without actual tests; they’re imported by the test files.

do you have separate targets for the tests and the libraries? i’m guessing not if you’re using that glob…

By "libraries" if you mean the files with "test helper" functions, then no (the tests are dependent on a library from a different BUILD). I'll take a look at

./pants tailor

. I will say that I briefly tried putting the two "tools" files in the

dependencies

field and then changing the

sources

to only have the test sources, but then it complained that there wasn't a BUILD file down in there. Got to run for today likely, but I'll try

tailor

and see what it does.

@echoing-farmer-15630: ok, sounds good. yea, would recommend committing everything… and then effectively deleting all BUILD files before running

tailor

, and seeing what it does.

Actually, the rubber-ducking technique worked; creating a

python_library

target with

name='test_tools'

and then including

:test_tools

as a dependency in the

python_tests

target seems to have done the trick for now. Not quite how I anticipated that working to be honest, and it could still bite you I think if you marked all tests in a file as skipped, but I don't have time to test for now. I'll still check out

tailor

for fun tomorrow.

yea, would recommend it. when proparly laid out, you should need very little boilerplate in each BUILD file.

FYI this question is https://www.pantsbuild.org/docs/python-test-goal#depending-on-test-utilities in the docs fwict

The

:test_tools

dependency should be inferrable, assuming you import from the test tools in the tests, so you could try leaving off that explict dependency (but keeping the

test_tools

target itself around).