THIS PROJECT IS NOW DEPRECATED IN FAVOR OF ghcup-hs

--

ghcup makes it easy to install specific versions of ghc on GNU/Linux as well as macOS (aka Darwin), and can also bootstrap a fresh Haskell developer environment from scratch. It follows the unix UNIX philosophy of do one thing and do it well.

Similar in scope to rustup, pyenv and jenv.

Ubuntu users may prefer hvr's ppa.

This project was started when CM was switching from stack to cabal nix-style builds.

Installation

Choose one of the following installation methods.

Simple bootstrap of ghcup, GHC and cabal-install

# complete bootstrap
curl https://gitlab.haskell.org/haskell/ghcup/raw/master/bootstrap-haskell -sSf | sh

# prepare your environment
. "$HOME/.ghcup/env"
echo '. $HOME/.ghcup/env' >> "$HOME/.bashrc" # or similar

# now create a project, such as:
mkdir myproject && cd myproject
cabal init -n --is-executable
cabal v2-run

Manual install

Just place the ghcup shell script into your PATH anywhere.

E.g.:

( mkdir -p ~/.ghcup/bin && curl https://gitlab.haskell.org/haskell/ghcup/raw/master/ghcup > ~/.ghcup/bin/ghcup && chmod +x ~/.ghcup/bin/ghcup) && echo "Success"

Then adjust your PATH in ~/.bashrc (or similar, depending on your shell) like so, for example:

export PATH="$HOME/.cabal/bin:$HOME/.ghcup/bin:$PATH"

Security aware users may want to use the files from the release page and verify the gpg signatures.

Usage

See ghcup --help.

Common use cases are:

# install the last known "best" GHC version
ghcup install
# install a specific GHC version
ghcup install 8.2.2
# set the currently "active" GHC version
ghcup set 8.4.4
# install cabal-install
ghcup install-cabal
# update cabal-install
cabal new-install cabal-install

Generally this is meant to be used with cabal-install, which handles your haskell packages and can demand that a specific version of ghc is available, which ghcup can do.

Manpages

For man pages to work you need man-db as your man provider, then issue man ghc. Manpages only work for the currently set ghc. MANPATH may be required to be unset.

Design goals

simplicity
non-interactive
portable
do one thing and do it well (UNIX philosophy)

Non-goals

invoking sudo, apt-get or any package manager
handling system packages
handling cabal projects
being a stack alternative

How

Installs a specified GHC version into ~/.ghcup/ghc/<ver>, and places ghc-<ver> symlinks in ~/.ghcup/bin/.

Optionally, an unversioned ghc link can point to a default version of your choice.

This uses precompiled GHC binaries that have been compiled on fedora/debian by upstream GHC.

Alternatively, you can also tell it to compile from source (note that this might fail due to missing requirements).

In addition this script can also install cabal-install.

Known users

vabal

Known problems

Limited distributions supported

Currently only GNU/Linux distributions compatible with the upstream GHC binaries are supported.

Precompiled binaries

Since this uses precompiled binaries you may run into several problems.

Missing libtinfo (ncurses)

You may run into problems with ncurses and missing libtinfo, in case your distribution doesn't use the legacy way of building ncurses and has no compatibility symlinks in place.

Ask your distributor on how to solve this or try to compile from source via ghcup compile <version>.

Libnuma required

This was a bug in the build system of some GHC versions that lead to unconditionally enabled libnuma support. To mitigate this you might have to install the libnuma package of your distribution. See here for a discussion.

Compilation

Although this script can compile GHC for you, it's just a very thin wrapper around the build system. It makes no effort in trying to figure out whether you have the correct toolchain and the correct dependencies. Refer to the official docs on how to prepare your environment for building GHC.

FAQ

Why reimplement stack?

ghcup is not a reimplementation of stack. The only common part is automatic installation of GHC, but even that differs in scope and design.

Why not contribute to stack and create a library for the common part?

While this might be an interesting idea, ghcup is about simplicity.

Why write a >1000k LOC bash script?

ghcup is POSIX sh.

Why write a >1000k LOC POSIX sh script?

Mainly because the implementation is fairly straight-forward and the script is highly portable. No need to bootstrap anything or set up yet another CI to build ghcup binaries for all possible arches and distros just to perform a very simple task: identify distro and platform and download a GHC bindist.

Why not support windows?

Consider using Chocolatey or ghcups.

License

haskell/ghcup

ErrorLooks like something went wrong!

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 4

Packages 0

Uh oh!

Contributors 23

Languages

Error
Looks like something went wrong!

Packages