I could use some ideas on how to deal with this symlink: https://github.com/st2sandbox/st2/blob/pants/st2tests/st2tests/fixtures/packs/core which links to: https://github.com/st2sandbox/st2/tree/pants/contrib/core I'm going to talk about the two distribution paths, to help with understanding why other people added the symlink, but I'm not asking about distribution/archives/etc.

contrib/core

primarily has files+code (in a StackStorm pack) that gets installed with StackStorm itself. But, it is also distributed as part of the

st2tests

module which provides general purpose test fixtures for use in the StackStorm ecosystem. There are a couple of issues with the symlink. The first of which was disambiguating a dependency which was simple enough: https://github.com/st2sandbox/st2/blob/pants/contrib/core/tests/BUILD#L3 The next couple of issues are intertwined: • For all the other fixtures, I used a python import so that pants could infer the dependency from test to fixture. But that get's a little odd with the symlinked code living in two places. •

contrib/core

is a source root (identified by a

[source].marker_filenames

file), but

st2tests/fixtures/packs/core

should not be a source root. Please help me think through what to do about this. I'm not even sure what questions to ask. Any thoughts?

Hmmm

I had a regex, but a core contributor for ST2 didn’t like that. Then I used

pack.yaml

as a marker file, which conflicted with how I had to load the fixtures. Now I’ve got a more limited set of files with an empty

PACK_ROOT

file. There are 7 (maybe 9) of those, so I guess I’ll just have to list them.

Not sure this will help though - there is always a source root. Pants takes the innermost dir that looks like one based on the source_roots config. So if all else fails it'll treat the repo root like a source root for

st2tests/fixtures/packs/core

Hmm. OK I think that resolves it. Thanks. Here's my big long explanation of why for future people that want to simplify: https://github.com/st2sandbox/st2/blob/pants/pants.toml#L40-L71

Thanks for the explanation!