Updating IHaskell to a Newer GHC

Posted on 2 May 2021
Tags: , ,

As the current maintainer of IHaskell, I see myself as having one primary responsibility: keeping it up-to-date with newer GHC releases. The chain of events that led to me becoming a maintainer started with the then-latest version of IHaskell not having support for GHC 8.0, and I still remember how frustrated I felt when dealing with this limitation.

Since then I’ve had the opportunity to add GHC 8.2, 8.4, 8.6, 8.8, 8.10, and now 9.0 support, but because I only have to do this every 6 months or so (at the earliest) I promptly forget the details of this work afterwards and have to spelunk through old, often heavily amended, commits to rediscover what past me (who is notoriously bad at documentation) did.

At the time of writing, GHC 9.2 is expected to be released soon and I don’t want to forget everything I’ve just (re)learned when that happens. Additionally, it is conceivable that at some point someone other than me would like to take a crack at updating IHaskell to the newest version of GHC. This blog post details the steps I took to make these tasks easier in the future.

Building IHaskell’s dependencies

I should start by saying that my current approach relies heavily on Nix and the infrastructure available in nixpkgs. If you don’t want to use Nix for whatever reason the general ideas might still translate to whatever method you use instead but the details will almost certainly vary widely.

The objective of this first step is to get us to the point where all of IHaskell’s dependencies are building, so that we can then focus on ghc-parser, ipython-kernel, and ihaskell exclusively.

I start with a version of Nixpkgs that has the necessary GHC version and package overrides to minimise work. As of this writing, the Nixpkgs maintainers base the Haskell package set they use on Stackage Nightlies with overrides added from head.hackage. Updates seem to go into the haskell-updates branch first and are then periodically merged into master. I started with this commit but it had issues with building alex that I sent pull requests for to Nixpkgs and to the project. In the meantime it’s very easy to make any required changes to a fork or a local copy of Nixpkgs. I start by copying release.nix from the IHaskell project root and changing the reference to Nixpkgs:

Changing Nixpkgs
let
  nixpkgs-src = builtins.fetchTarball {
    url = "https://github.com/NixOS/nixpkgs/tarball/8795d39ce70f04e3fd609422d522e5b2594f3a70";
    sha256 = "01w7q0nqydippj0ygbg77byb770snhc5rnqzc6isws58642l8z4s";
  };
in
{ compiler ? "ghc901"
, jupyterlabAppDir ? null
, nixpkgs ? import nixpkgs-src {}
, packages ? (_: [])
, pythonPackages ? (_: [])
, rtsopts ? "-M3g -N2"
, systemPackages ? (_: [])
}:

Then it’s possible to build this and see how many packages need changes:

$ nix-build release-9.0.nix --keep-going 2>&1 | wc -l

Fixing the affected packages might involve patching, jailbreaking it so that its dependency bounds are relaxed, using a newer version that is not included in the package set by default, or any number of other changes. Here’s what I ended up with this time:

Package set overrides
      cryptohash-md5    = nixpkgs.haskell.lib.doJailbreak super.cryptohash-md5;
      cryptohash-sha1   = nixpkgs.haskell.lib.doJailbreak super.cryptohash-sha1;
      basement          = super.basement_0_0_12;
      foundation        = super.foundation_0_0_26_1;
      memory            = nixpkgs.haskell.lib.appendPatch super.memory (nixpkgs.fetchpatch {
        url = "https://gitlab.haskell.org/ghc/head.hackage/-/raw/c89c1e27af8f180b3be476e102147557f922b224/patches/memory-0.15.0.patch";
        sha256 = "0mkjbrzi05h1xds8rf5wfky176hrl03q0d7ipklp9x4ls3yyqj5x";
      });
      cryptonite        = nixpkgs.haskell.lib.appendPatch super.cryptonite (nixpkgs.fetchpatch {
        url = "https://gitlab.haskell.org/ghc/head.hackage/-/raw/6a65307bbdc73c5eb4165a67ee97c7b9faa818e1/patches/cryptonite-0.28.patch";
        sha256 = "1wq9hw16qj2yqy7lyqbi7106lhk199hvnkj5xr7h0ip854gjsr5j";
      });
      profunctors       = self.callCabal2nix "profunctors" profunctors-src {}; # `profunctors-src` is defined above
      mono-traversable  = nixpkgs.haskell.lib.dontCheck super.mono-traversable;

After every few changes, I like to rerun nix-build and watch the number go down. It’s also possible to build an individual package, e.g. to build foundation (and any dependencies) one would run

$ nix-build release-9.0.nix -A passthru.haskellPackages.foundation

This is what the final release-9.0.nix looked like.

Eventually only ghc-parser, maybe ipython-kernel, and ihaskell should fail to build.

Updating ghc-parser

ghc-parser has the fewest dependencies of the three packages we are changing so it makes sense to start there. I’m relatively low-tech as far as development workflow goes and I prefer ghcid and a text editor, mostly because I haven’t yet figured out how to get anything more advanced to work. To get ghcid running, assuming you have it installed globally like I do, you relax the version bounds in ghc-parser.cabal and run

$ nix-shell release-9.0.nix -A passthru.haskellPackages.ghc-parser.env
$ cd ghc-parser
$ runhaskell Setup.hs configure
$ ghcid -c runhaskell Setup.hs repl lib:ghc-parser

Most of the compilation errors are related to this GHC module restructuring that started in GHC 8.10 and continued in GHC 9.0. If I had kept better notes from last time I would have looked at ghc-api-compat which offers a compatibility shim and whose .cabal file makes translating between old and new module names very easy. Instead I ended up looking at the GHC 9.0 API Haddocks and the GHC 8.10 API Haddocks. As an aside, I am irritated that the most recently released GHC API documentation isn’t available on Hackage. I also like to have a local checkout of the GHC source so that I can look at the code across different commits if required.

These are the changes I needed to make to ghc-parser.

Updating ipython-kernel

ipython-kernel doesn’t depend on the GHC API directly, so changes to it are usually related to breaking API changes in other dependencies. In this case, no changes were required!

Updating ihaskell

This is usually the most involved package to update, as its operation is intimately tied with the details of the GHC API. Most of the changes required were for three reasons:

  1. The aforementioned module hierarchy change

  2. Changing the terminology from “packages” to “units” as described in this commit

  3. Removing specialised gcatch, gtry, etc. functions in favour of the more general versions in exceptions, as detailed in this section of the release notes

As before, it’s possible to get ghcid running with

$ nix-shell release-9.0.nix -A passthru.haskellPackages.ghc-parser.env
$ runhaskell Setup.hs configure --enable-tests
$ ghcid -c runhaskell Setup.hs repl lib:ihaskell

After getting everything compiling, I like to build the ihaskell package by running

$ nix-build release-9.0.nix -A passthru.haskellPackages.ihaskell

because this sets up the test environment correctly (i.e. putting the built ihaskell executable in the $PATH) before running tests, although course you could do this manually. This usually catches any issues that have slipped through and small formatting changes in GHC output across versions.

Here are the changes I made to ihaskell.

Acceptance testing

Since IHaskell bridges the Jupyter and GHC ecosystems, we have an acceptance test that essentially runs an IHaskell notebook through nbconvert and ensures that the output is identical to the input. Because GHC output (amongst other things) differs across GHC versions, this acceptance test was frequently broken and/or a bad indicator of whether any changes were correct. Recently James Brock simplified and greatly improved the acceptance test to be more reliable. Unfortunately the latest releases of Jupyter now include additional metadata with each response including the time of reply, which cannot be expected to be the same across runs. In the past it’s been possible to filter the offending fields out using grep -e but a more sophisticated approach was required this time so I took the opportunity to learn a little more about jq and used that instead. This new approach should also be more flexible and better at accommodating future output changes.

Here are the changes I made to the acceptance tests.

Using the updated IHaskell

We’re done! I like to quickly try out a new notebook, as a quick test that everything works as expected (and also for the novelty of being the first person to try IHaskell on the newest GHC). To do this, I run

$ nix-build release-9.0.nix
$ result/bin/jupyter-notebook