Maybe bug in tests re Local dists Can I get another set of e Pants #development

Join Slack

Maybe bug in tests re: Local dists. Can I get anot...

# development

bitter-ability-32190

03/13/2023, 5:37 PM

Maybe bug in tests re: Local dists. Can I get another set of 👀 in 🧵 ?

bitter-ability-32190

03/13/2023, 5:38 PM

src/python/pants/backend/python/util_rules/local_dists_test.py

we make a

LocalDistsPexRequest

with no ICs associated, but the test assumes py3. Seems like we'd want to get the ICs from the sources here?

bitter-ability-32190

03/13/2023, 5:38 PM

Yeah OK I think it's a bug. Every "real world" caller provides ICs

bitter-ability-32190

03/13/2023, 5:49 PM

This commit fixes it: https://github.com/pantsbuild/pants/pull/18433/commits/bca084e0445e25addd23e25278a24f447a00c152

witty-crayon-22786

03/13/2023, 6:15 PM

probably a bug. but it seems like beyond ICs, any caller setting

internal_only

(aka: please build this for exactly one interpreter, because this is not being published outside of Pants) could provide an actual interpreter

witty-crayon-22786

03/13/2023, 6:17 PM

Copy code

def __init__(
  self,
  addresses: Iterable[Address],
  *,
  interpreters: InterpreterConstraints | PythonBinary,
  sources: PythonSourceFiles = PythonSourceFiles.empty(),
) -> None:

witty-crayon-22786

03/13/2023, 6:18 PM

when tests are running, iirc, they’ve already chosen an interpreter. so passing in

internal_only=True, interpreter_constraints=…

would be asking to re-resolve, and then relying on the same one that was pre-resolved to be re-resolved.

witty-crayon-22786

03/13/2023, 6:20 PM

yea, it looks like the

internal_only

flag already does some of this… it’s not clear though that when you pass it, you don’t need to pass ICs

Open in Slack

Previous Next