Hey again. What are the limitations of the test results caching? I'm thinking about 3rd party dependencies or internal dependencies in another repository f.e.? From a quick trial and error pants impressed me but still there must be limits, right?

What limits do you imagine? A test run is - in Pants parlance - a Process execution. The cache key is effectively a hash of

{"env": [...], "args": [...], "file_inputs": [...]}

where the file inputs are not the files themselves, but their fingerprints (hash + size).

I don't know - I guess the

"env"

part seems to be the most opaque to me ... everything available locally feels rather easy to grasp - if there is a change, it'll be noticed.

The env are the env vars.

oh ok ... then

"venv"

😄

At a very high level, the exact versions/source-code of external dependencies (including git dependencies) influence the cache key: if they change, the cache key will change and thus tests will be rerun.

Ok that sounds reasonable - somehow challenging to implement I could imagine - but ok, I’m convinced I guess 😛