I'm getting an error without much of a stack trace while trying to build a very basic docker image and I can't figure out why for the life of me. Below is the error with the generic stack trace, plus some of the trace level logging. I'm pretty sure the error is occurring somewhere in the

create_docker_build_context

rule, and the last thing I'm seeing is it spitting out the list of packages it built for the image, but it's swallowing the actual error it seems.

12:11:56.53 [TRACE] Completed: pants.backend.docker.util_rules.docker_build_context.create_docker_build_context
12:11:56.53 [TRACE] Completed: pants.backend.docker.goals.package_image.build_docker_image
12:11:56.53 [TRACE] Completed: pants.core.goals.package.environment_aware_package
12:11:56.53 [DEBUG] Completed: `package` goal
12:11:56.53 [TRACE] Completed: select
12:11:56.53 [DEBUG] computed 1 nodes in 5.899770 seconds. there are 7134 total nodes.
12:11:56.53 [ERROR] 1 Exception encountered:

Engine traceback:
  in select
    ..
  in pants.core.goals.package.package_asset
    `package` goal

Traceback (most recent call last):
  File "/Users/nick.dellosa/Library/Caches/nce/695a0515ecce6747a3e42742caf246ea42091f11bfbdd8d95ea0903e90eb510a/bindings/venvs/2.19.0rc5/lib/python3.9/site-packages/pants/core/goals/package.py", line 165, in package_asset
    packages = await MultiGet(
  File "/Users/nick.dellosa/Library/Caches/nce/695a0515ecce6747a3e42742caf246ea42091f11bfbdd8d95ea0903e90eb510a/bindings/venvs/2.19.0rc5/lib/python3.9/site-packages/pants/engine/internals/selectors.py", line 374, in MultiGet
    return await _MultiGet(tuple(__arg0))
  File "/Users/nick.dellosa/Library/Caches/nce/695a0515ecce6747a3e42742caf246ea42091f11bfbdd8d95ea0903e90eb510a/bindings/venvs/2.19.0rc5/lib/python3.9/site-packages/pants/engine/internals/selectors.py", line 172, in __await__
    result = yield self.gets
TypeError: 'NoneType' object is not iterable

Does this work in 2.18? Or does it rely on some 2.19 functionality?

Ooh, I was able to get the actual stack trace with 2.18!

~Great! That makes debugging easier, and means this is not a regression (in the docker backend).

It's my custom tagging plugin that is breaking it for this particular target.

Ack, thanks. I'm working on 2.19 release so would be a shame with a last minute blocker. On topic, does

pkill pantsd

between runs affect what gets outputted? Also maybe

--no-local-cache

might affect it. I assume you're already using

--print-stacktrace

in both versions.

yeah let me jump back to 2.19 and try some stuff

That doesn't sound implausible. This seems like a good thing to file a GitHub issue for, and we can get some more eyes on it there. There's been some work to change how rules work that might affect it.

Confirmed it only effects union rules or anything called from a union rule.

yea, unions reenter the rule graph anew, so I'm not surprised if an error doesn't include the outter scope in the traceback.