Fix follow path checking at depths greater than 2

[RealityAnomaly]

Motivation

Currently, you can't have follows paths with depths greater than 2 - you get an error along the lines of error: input 'B/D' follows a non-existent input 'B/C/D'.

Context

We need to recurse into the input tree to handle follows paths that trarverse multiple inputs that may or may not be follow paths themselves. I've added tests for the problem.

Checklist for maintainers

Maintainers: tick if completed or explain if not relevant

• agreed on idea
• agreed on implementation strategy
• tests, as appropriate
  • functional tests - tests/**.sh
  • unit tests - src/*/tests
  • integration tests - tests/nixos/*
• documentation in the manual
• documentation in the internal API docs
• code and comments are self-explanatory
• commit message explains why the change was made
• new feature or incompatible change: updated release notes

Priorities

Add 👍 to pull requests you find important.

[iFreilicht]

I cherry-picked the test onto current master and it fails indeed with error: input 'B/D' follows a non-existent input 'B/C/D'. I then tested the full branch locally and it does indeed fix the test.

Code change itself is trivial, this can be merged.

If you want, you can also add a bullet to rl-next.md so the fix shows up in the next release notes, but it's not a requirement :)

[RealityAnomaly]

If you want, you can also add a bullet to rl-next.md so the fix shows up in the next release notes, but it's not a requirement :)

Sure - I've added a small note.

[iFreilicht]

Looks good! Just to set your expectations, I don't have commit access, so this PR might stay open until one of the maintainers gets around to merging it, despite my approval.

[roberth]

Tricky one. I think the diagram I wrote in the suggestion could be helpful. Does it match your understanding of the problem?

@tomberek I think you mentioned a problem like this some time ago.

[tomberek]

@tomberek I think you mentioned a problem like this some time ago.

The example I've been using: #5790 (comment)

[aakropotkin]

Thanks for making this PR @VertexA115 <3

[roberth]

It would perhaps be nice to (also?) have a test that's based on the actual contents of the overridden flakes; maybe something from #5790?
Nonetheless the new test seems sufficient as a regression test.

[iFreilicht]

The test on macos is failing only due to a flaky test. Could someone with CI access retrigger it?

[tomberek]

@tomberek I think you mentioned a problem like this some time ago.

The example I've been using: #5790 (comment)

Note. This PR is solving a different problem than what I mentioned above.