Azure CI

This page documents how to use Stack on Azure CI.

Quick Start

Note that you have to create azure pipelines for your project and then you need to put the relevant configuration files:

  • For simple Azure configuration, copy-paste the azure-simple file into azure-pipelines.yml.
  • For complex Azure configuration, you need to take the below linked four files and put all of them into the .azure directory except the azure-pipelines.yml file which should be put in the root of the repository.

For a more detailed explanation, you can read further.

Simple and Complex configuration

We provide two fully baked configuration ready to be used on your projects:

  • The simple Azure configuration is intended for applications that do not require multiple GHC support or cross-platform support. It builds and tests your project with just the settings present in your stack.yaml file.
  • The complex Azure configuration is intended for projects that need to support multiple GHC versions and multiple OSes, such as open source libraries to be released to Hackage. It tests against cabal-install, as well as Stack on Linux and macOS. The configuration is significantly more involved to allow for all of this branching behavior. In the provided template, we use the Linux build configuration to test against various combinations of cabal-install and stack. The macOS and Windows configuration is used only for testing Stack build configuration. These are the files for the complex configuration:
  • azure-pipelines.yml : This is the starter file used by the Azure CI.
  • azure-linux-template.yml : Template for Azure Linux build
  • azure-osx-template.yml : Template for Azure macOS build
  • azure-windows-template.yml : Template for Azure Windows build

NOTE: It is likely going to be necessary to modify this configuration to match the needs of your project, such as tweaking the build matrix to alter which GHC versions you test against, or to specify GHC-version-specific stack.yaml files if necessary. Don't be surprised if it doesn't work the first time around. See the multiple GHC section below for more information.

Creating Azure Pipelines

Each of these configurations is ready to be used immediately. But before we go into where to put them, we have to create pipeline for your project in Azure CI platform:

  • Go to You have to initially sign-in to your microsoft account there.
  • Once you have logged in to your Microsoft account, you have to sign in to Azure devops from there.
  • You will be greeted with a dashboard where you can create your projects.
  • Click the "Create Project" button and fill the relevant information in the dialog and then click the "Create" button.
  • This will lead you to the project dashboard page where you can create pipelines.
  • Click on "Pipelines" in the left menu. This will load the pipelines page on the right.
  • Click on the button "New Pipeline" and you have to follow through the wizard there. You need to choose your github repository (or Azure repos) and follow the wizard. Note that in the Configure step you have to select the "Starter Pipeline". This will open up an editor window. You can leave the existing yaml configuration there as it is and click the "Save and run" button. That will popup a dialog. Select the relevant option and click "Save and run" button. (Note that this step would have created azure-pipelines.yml in your repository, you have replace that with the appropriate configuration file.)

The rest of this document explains the details of common Azure configurations for those of you who want to tweak the above configuration files or write your own.

Note: both Azure and Stack infrastructures are actively developed. We try to document best practices at the moment.


Note that you need at least one agent to build your code. You can specify which virtual image you want to choose using this configuration:

  vmImage: ubuntu-16.04

The other popular options are macOS-10.13, vs2017-win2016 for Mac and Windows respectively. You can find the complete list here.

Note that as of now, Azure CI doesn't offer any caching support. You can use something like cache-s3 for explicit caching, although it may cost you. For a project with an example usage of cache-s3, you can see the Azure configuration of stack repository.

Installing Stack

Currently there is only one reasonable way to install Stack: fetch precompiled binary from the Github.

- script: |
    mkdir -p ~/.local/bin
    curl -L | tar xz --wildcards --strip-components=1 -C ~/.local/bin '*/stack'
  displayName: Install Stack

Installing GHC

There are two ways to install GHC:

  • Let Stack download GHC
  • Install GHC using apt package manger. This method is only applicable for Debian based images.

See the simple azure script for an example of the first option (letting Stack download GHC). Here, we will show the second option:

sudo add-apt-repository -y ppa:hvr/ghc
sudo apt-get update
sudo apt-get install cabal-install-$CABALVER ghc-$GHCVER

For the above commands to work, you need to set the CABALVER and GHCVER environment variable properly.

Multiple GHC - parametrised builds

For different GHC versions, you probably want to use different stack.yaml files. If you don't want to put a specific stack.yaml for a particular resolver and still want to test it, you have specify your resolver argument in ARGS environment variable (you will see an example below). For cabal based builds, you have to specify both GHCVER and CABALVER environment variables.

      BUILD: stack
      STACK_YAML: stack.yaml
      BUILD: stack
      STACK_YAML: stack-lts-13.yaml
      BUILD: cabal
      GHCVER: 8.4.4
      CABALVER: 2.4
      BUILD: cabal
      GHCVER: 8.6.5
      CABALVER: 2.4
      BUILD: stack
      ARGS: "--resolver nightly"
      BUILD: style
      BUILD: pedantic
      STACK_YAML: stack.yaml

Running tests

After the environment setup, actual test running is simple:

  - stack test

Other details

Some Stack commands will run for long time. To avoid timeouts, use the timeoutInMinutes for jobs.