search

crate-ci / azure-pipelines

Easy continuous integration for Rust projects with Azure Pipelines
MIT License
88 stars 24 forks source link

readme

Build Status Codecov Documentation

Ah, so you want to set up continuous integration (CI) testing for your Rust project, and you decided you wanted to use Azure Pipelines for it? Well, you're in the right place!

Azure Pipelines, like many other CI services, basically requires you to fully spell out all the steps to your CI. This is very handy if you have a complex CI pipeline, but is pretty inconvenient if you just want something that just works. This project aims to bridge that gap. It also tries to guide you through how to even get Azure Pipelines set up in the first place, which can be a daunting thing to get right!

If you're curious what your CI will ultimately look like, go take a look at tracing-timing's CI for example. By default, it tests on all platforms, checks that your code compiles with and without any features it may have, and ensures that your code works with an older Rust version. You can also mix-and-match these checks if you wish.

The repository provides three main templates:

Below are instructions for how to set up your repository with testing from this repository, for setting up code coverage, and for configuring various parameters of the default CI template.

If you've done this before:

If you've done this before, and just want the standard YAML again for azure-pipelines.yml, here it is:

jobs:
 - template: default.yml@templates

resources:
  repositories:
    - repository: templates
      type: github
      name: crate-ci/azure-pipelines
      ref: refs/heads/v0.4
      endpoint: YOU_NEED_TO_SET_THIS

If you're getting something new set up:

Getting Azure Pipelines and its connection to GitHub set up correctly is not entirely straightforward. This document takes you through exactly the steps you need to do. Stray from these at your own risk.

Setting up Azure DevOps

Azure loves to try to get you to sign in to GitHub using OAuth, thereby giving them access to all your public and private repos. This makes us sad. Here's how you do it "the new way" instead.

First, make sure you have the Azure Pipelines GitHub Application installed:

Then, make sure you have an Azure Project for your GitHub organization:

Note that Azure associates only one of your projects with a given organization's GitHub Apps install, so you cannot have multiple Azure Projects that are linked to different GitHub projects under the same GitHub user/organization. This is stupid, but such is life.

This template uses Build stages, which is a preview feature of Azure Pipelines. You therefore need to enable support for it. To do so, click your profile icon in the top-right corner and click "Preview features". In the drop-down at the top in the panel that appears, choose "for this organization", then enable "Multi-stage pipelines".

Adding CI for a GitHub repository

At this point I'll assume you have an Azure Project correctly set up for your GitHub user or organization (as described above).

Before we continue, there's a fun little step you have to do first. Go to "Project settings" (bottom left), the "Service connections" (under "Pipelines"). There should be one thing listed there, and it's your authenticated connection to GitHub. Note down its name.

Now, create a file azure-pipelines.yml in the root of the repository you want CI for. If you want all the bells and whistles, write:

jobs:
 - template: default.yml@templates

resources:
  repositories:
    - repository: templates
      type: github
      name: crate-ci/azure-pipelines
      ref: refs/heads/v0.4
      endpoint: PLACEHOLDER

Where PLACEHOLDER is the service connection name we found above. The template also has a number of configuration options with opinionated defaults. If you have a particularly "weird" project, you can also mix-and-match individual CI components.

Once that's all committed and pushed, it's time to set up the Pipeline in Azure:

Hopefully Azure was now happy with your efforts. If it is, you'll be taken to your new shiny "Pipeline summary" page, and it will show you your build and tests progress! Congrats, you now have Azure Pipelines CI! If you instead get a big red box at the top of the "Run pipeline" box with an error, try to see if you can figure out which part of the magic incantation you missed. If it all looks right to you, file an issue!

If you want to add support for code coverage:

This pipeline is also set up to use tarpaulin and codecov.io for test coverage reporting. To enable this, here's what you have to do:

Note that this gives access to your Codecov API key to anyone with push access to the repository! Use it wisely. Forks of your repository do not have access to secrets by default.

Now just add this to your entries in azure-pipelines.yml that use the templates default.yml:

parameters:
 codecov_token: $(CODECOV_TOKEN_SECRET)

You may also want to give yourself a nice badge! Just go to the Settings page on codecov.io again and click "Badge" on the left. To add it your crates.io page, add this to Cargo.toml:

codecov = { repository = "GH_USER/GH_PROJECT", branch = "master", service = "github" }

Code coverage for PRs

If you are really sure you want to allow coverage to run for the arbitrary code people may submit in PRs to see your secrets, here's what you do:

If you instead want to simply skip coverage on pull requests, do not check the box next to "Make secrets available to builds of forks". You should not need to change anything else.

If you want to configure default.yml

The main template, default.yml comes configured with some opinionated defaults. It will check your crate against a minimum Rust version, check your project without features, and with all of them, and it will run both rustfmt, clippy, and beta/nightly check on your codebase. If, for whatever reason, you disagree with some of these choices, or have a project with particular needs, copy-paste default.yml into your azure-pipelines.yml, and replace

 - template: install-rust.yml

with

 - template: install-rust.yml@templates

Then you can tweak it to your heart's desire!

There are some smaller configuration parameters available for default.yml too. Most of these also apply to nightly-only.yml.

Testing on multiple platforms

jobs:
 - template: default.yml@templates
   parameters:
     cross: <bool> = true

By default, your pipeline will test on Linux, MacOS, and Windows. To only test on Linux, set cross to false.

Minimum Supported Rust Version (MSRV)

jobs:
 - template: default.yml@templates
   parameters:
     minrust: <false | rust version> = 1.32.0

By default, your pipeline will test against a minimum Rust version to ensure that users of your crates are not required to run on the latest stable version. The default minimum version may be bumped occasionally, but will always stay at least 4 releases behind the newest release (~6 months). If you wish to test a particular minimum version, say 1.34.0, you would give that version number as the minrust parameter to default.yml. If you wish to disable the MSRV check, set minrust to false.

Minimum supported nightly

If you are using the nightly-only.yml template, the equivalent of MSRV checking is to check that you support some particular "oldest" nightly. You can check this by specifying a date for the min parameter:

jobs:
 - template: nightly-only.yml@templates
   parameters:
     minrust: <false | YYYY-MM-DD> = false

Environment variables

jobs:
 - template: default.yml@templates
   parameters:
     env:
       <name>: <var>

If you tests require particular environment variables to be set, you can set these using the env parameter. The given environment variables will be passed in whenever your tests are run. You can set multiple environment variables, and you can use variables.

Additional setup steps

jobs:
 - template: default.yml
   parameters:
     setup:
       - <steps>

Occasionally your project requires additional setup steps for tests to be run. This may include installing packages, downloading dependencies, fetch files, or anything else you might think of. To add such extra steps, use the setup parameter and give it a list of tasks (you can see all of Azure's built-in tasks here).