In pants <2.20, we were able to have a scenario like

python_sources(
   name="my_parametrized_source",
   overrides={
      "some_source.py": {
         "dependencies": parametrize(variant1=[":some_thing"], variant2=[":another_thing"])
      }
   }
)

pex_binary(
   name="my_binary",
   entry_point="some_source.py",
   dependencies=["//some_source.py:my_parametrized_source@variant1"]
)

and things just worked. In >=2.20, you get

UnownedDependencyError

on inferring the entry point to the

pex_binary

and then it lists out all the parametrized options.

ah, bummer about the breakage, and forcing you to make changes without warnings. That's not so good.

Hmmm, curious which commit in 2.20 caused this

Haven't tried out the above since

executable

worked fine. I don't think the target address for

entry_point

works that way, though.

@happy-kitchen-89482, I think this is https://github.com/pantsbuild/pants/pull/20390

The docs say it will interpret the stuff after

:

as a function definition in the module

Ah, right. That's unfortunate. --- I've filed https://github.com/pantsbuild/pants/issues/20806 because it seems like this warning is both redundant and unsilenceable

Just tried out 2.19.3rc1 and it appears to be fixed there

I think the 2.19 series doesn't have this new warning. I'd guess it's still visible in the latest 2.20.x series, since we haven't had a chance to land https://github.com/pantsbuild/pants/pull/20853

I was able to run a

pants dependents

command as my signal. Previous versions were failing

Ah, hm. I think your original message indicates that only >=2.20 was failing? Or were there (other?) problems in 2.19 too?

Oh, yeah you're right here. 2.19.2 was just as happy

I don't think someone has fixed this yet, sorry, so no release scheduled. That PR is still open and references https://pantsbuild.slack.com/archives/C0D7TNJHL/p1714260455658849 as suggesting it needs to be done a different way

I think 2.23.0 looks to have fixed this