fix textDocument/selectionRange's method

[kokobd]

fix #387

[jneira]

Good catch, it should be catched by some type of testing, but it is a generic issue we should tackle somewhere else

[kokobd]

Nix build passes but direct build fails. It looks like an orphan instance in JsonSpec.hs conflicts with latest version of aeson.

lsp/lsp/test/JsonSpec.hs

Line 162 in 08e8a72

instance Arbitrary J.Value where

Should it be fixed here or in another PR?

[kokobd]

According to aeson's changlog, it provides Arbitrary instance for Value starting from version 2.0.3.0

[jneira]

wow, @michaelpj got to support aeson >= 2.0 without any cpp, not sure if we could avoid it fixing this

[jneira]

Add a instance is a breaking change and it should be a major version bump 🙄
Well i would update the aeson bounds to exclude 2.0.3.0 to make ci work and check the pr changes and left the fix for another pr (if @michaelpj agree)

[kokobd]

@jneira I've added version bound by another commit in this PR. CI has passed now.

[michaelpj]

Ugh, thanks for fixing this.

Re: the CPP, I think there's a more elegant solution. We only need the instance in the tests for the lsp package. So we can put a aeson >= 2.0.3.0 constraint on the test component of lsp, and unconditionally delete the instance. That will force a new aeson... but only for people who build the test component. Users who only depend on the library won't get that constraint and will continue being able to use any vesrion.

[michaelpj]

(Incidentally, I don't love the fact that merge-me auto-merges even if someone pushes another commit!)

[michaelpj]

#389

[Bodigrim]

Add a instance is a breaking change and it should be a major version bump

This is a non-orphan instance, which according to PVP is non-breaking.