got ghc-mod working with nix

[chessai]

I added a nix expression that allows building ghc-mod with nix. It does not affect the code, however, a few of your dependencies have what I believe are unnecessarily low upper bounds. They are:

hlint, monad-journal, optparse-applicative, extra, and haskell-src-exts.

The reason I say that these bounds are unnecessarily low is that ghc-mod can build with their newest versions.

[DanielG]

Maybe I'm a bit confused about how things are supposed to work with Nix but why are you PRing this here instead of over at nixpkgs?

As for the over-restrictive bounds I'd be happy to merge a PR that just bumps the related dependencies if you tested that everything builds and works with them.

[chessai]

The version of ghc-mod on hackage is currently not buildable (at least with 8.2.2), and nixpkgs only allows for performing overrides on the version in hackage, which is not possible in this case. With this nix pression, users can install ghc-mod with "nix-env -f default.nix -i 'ghc-mod-pinned' "

…

On Feb 25, 2018 3:50 PM, "Daniel Gröber" ***@***.***> wrote: Maybe I'm a bit confused about how things are supposed to work with Nix but why are you PRing this here instead of over at nixpkgs? As for the over-restrictive bounds I'd be happy to merge a PR that just bumps the related dependencies if you tested that everything builds and works with them. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#928 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ARyL6996hdsSvmrVc3mUdXCpY9MEN1quks5tYcd-gaJpZM4SQLJo> .

[peti]

I don't understand the purpose of this PR. ghc-mod.nix is an auto-generated file that downloads the ghc-mod source code from github even though it's supposed to be living inside of that source code. This feels counter-intuitive, at best. Also, Nixpkgs has build instructions for ghc-mod already, so why bother duplicating them here? Maybe I am missing something?

[chessai]

@peti, it's possible I've made a mistake, but when installing ghc-mod with 'nix-env -iA nixos.haskellPackages.ghc-mod', ghc-mod failed to build.

Looking at that nixpkgs issue which referenced this PR, I can see that the problem is fixed.

When I made this PR, I didn't go to nixpkgs because I thought nixpkgs was pulling ghc-mod from Hackage. Now I know that assumption is incorrect.

Also, this is probably the wrong place to make the PR anyway. I apologise.

[peti]

When I made this PR, I didn't go to nixpkgs because I thought nixpkgs was pulling ghc-mod from Hackage.

Nixpkgs does pull ghc-mod from Hackage. If ghc-mod would release an update to Hackage that fixes the overspecified constraint on hlint, then that would have automatically fixed the build failure we experienced in the release-17.09 branch. Alas, no such update seems to be forthcoming, so we worked around the issue by patching the build inside of Nixpkgs to just ignore the upstream version constraints, and now it succeeds again.

[DanielG]

@peti sorry for the late reply, but wouldn't a hackage revision suffice for a bounds issue? Does your tooling not pick those up or something? I'd be happy to bump them bounds if anyone actually notified me of any problems!