Here's the failure on CI: https://travis-ci.org/pointfreeco/swift-nonempty/builds/440846584?utm_source=github_status&utm_medium=notification

Interestingly enough, even if I try to #if os(Linux) the issue, an AST check blows things up:

pointfreeco/swift-nonempty@cfa55ac

I can't even retain conditional conformance on non-Linux platforms 🙁

@swift-ci create

I'll try to reproduce this pretty soon, but meanwhile is it possible to check with a master snapshot? We changed a lot of things about how generic parameters are implemented that couldn't be cherry-picked to 4.2, so it's possible this is "already" fixed. (Even though it won't be out for a while, sorry.)

Okay, a master snapshot does this:

/home/jrose/swift-nonempty/Sources/NonEmpty/NonEmpty+Dictionary.swift:109:12: error: generic parameter 'Key' could not be inferred
    return Dictionary(uniqueKeysWithValues: Array(lhs))
           ^
Swift.Dictionary:1:15: note: 'Key' declared as parameter to type 'Dictionary'
public struct Dictionary<Key, Value> where Key : Hashable {
              ^
/home/jrose/swift-nonempty/Sources/NonEmpty/NonEmpty+Dictionary.swift:109:12: note: explicitly specify the generic arguments to fix this issue
    return Dictionary(uniqueKeysWithValues: Array(lhs))
           ^
                     <<#Key: Hashable#>, Any>

…which is SR-9068. So I'm not sure if this is fixed yet or not.

Okay, replacing that body with return true made everything compile. So I'll try to get this down to a reproducible test case and add it to the suite, but it looks like it's fixed on master.

I never managed to add the test case and I can't even get this to fail anymore (Swift 4.2 on Ubuntu 18.04). At this point I think it's not worth spending time on anymore.