https://pantsbuild.org/ logo
Powered by
#general
  • r

    rough-minister-58256

    09/11/2017, 11:08 PM
    but I don’t believe there’s a way currently in pants to say “this target should execute against both python2 and python3”
  • r

    rough-minister-58256

    09/11/2017, 11:08 PM
    it’s all based on compatibility tags at the target level, and each target ends up with a singular compatible interpreter
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:08 PM
    hmmmm.
  • r

    rough-minister-58256

    09/11/2017, 11:08 PM
    it’d be great to add that as feature tho, for tox parity
  • r

    rough-minister-58256

    09/11/2017, 11:08 PM
    makes perfect sense to me
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:08 PM
    so could one use compatibility to set up two individual test targets, then?
  • r

    rough-minister-58256

    09/11/2017, 11:09 PM
    hrm ya that should work actually
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:09 PM
    I've gotten a bit munged up trying to do so for libraries
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:09 PM
    hmmm. Okay. Let me try that approach
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:11 PM
    I've set 
    compatibility=['CPython>=2.7', 'CPython>=3.6']
    in a test target
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:11 PM
    I get no errors after doing so, but it has no effect on the interpreter that's actually used to execute the test target
  • r

    rough-minister-58256

    09/11/2017, 11:20 PM
    it’s executing with python3.6?
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:20 PM
    no
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:21 PM
    I'm getting the Python 2.7 interpreter on my machine consistently
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:21 PM
    The only way I've been able to get it to use a Python 3 interpreter is by using the PEX_PYTHON env variable
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:22 PM
    but if I try to run a test target with that set to a Python3 interpreter, I get an error about incompatible versions of pytest and pytest-cov
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:23 PM
    I'm trying to strip this down to the minimal possible setup to verify that I can accomplish the tasks I need to do
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:24 PM
    I've got a setup with a single python_library, and a single python_tests target that depends on that library
  • r

    rough-minister-58256

    09/11/2017, 11:24 PM
    looks like constraints are possibly ORed together - what if you just do 
    compatibility=['CPython>=3.6']
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:24 PM
    I'm checking the interpreter used in running the tests by having a test that asserts that 
    sys.version_info[0] == 3
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:24 PM
    Let me try that
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:26 PM
    Okay, that didn't work, but it did give me a different exception:
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:26 PM
    Copy code
    FAILURE: Unable to detect a suitable interpreter for compatibilities: CPython>=3.6 (Conflicting targets: packages/ackermann/tests/python:tests_py3, packages/ackermann/src/python:ackermann)
  • r

    rough-minister-58256

    09/11/2017, 11:26 PM
    that means it can’t find a python 3.6 interpreter on your $PATH
  • r

    rough-minister-58256

    09/11/2017, 11:26 PM
    (good sign)
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:27 PM
    the BUILD files for both the test and the library targets both specify 
    compatibility=['CPython>=3.6'],
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:27 PM
    when it tries to do so, what does it look for?
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:27 PM
    I have both 
    python
    and 
    python3
    available
  • b

    bulky-motorcycle-41302

    09/11/2017, 11:27 PM
    do I need to set up an alias of 
    python3.6 -> python3
    ?
  • r

    rough-minister-58256

    09/11/2017, 11:28 PM
    no, at a high level I believe it looks for all instances of 
    python
    on all path components of $PATH and then executes them to get their sys.version_info
1...181118121813...2420Latest