Skip to content

Latest commit

 

History

History
1465 lines (1084 loc) · 59.9 KB

README.md

File metadata and controls

1465 lines (1084 loc) · 59.9 KB
Raw


Craft: Universal Release Tool (And More)

Travis GitHub release npm version license

craft is a command line tool that helps to automate and pipeline package releases. It suggests, and then enforces a specific workflow for managing release branches, changelogs, artifact publishing, etc.

Table of Contents

Installation

Binary

craft is distributed as a minified single JS binary.

npm (not recommended)

Recommendation is to used this file directly but one can also install craft as an NPM package and can be installed via yarn or npm:

yarn global add @sentry/craft
npm install -g @sentry/craft

Usage

$ craft -h
craft <command>

Commands:
  craft prepare NEW-VERSION  🚢 Prepare a new release branch
                          [aliases: p, prerelease, prepublish, prepare, release]
  craft publish NEW-VERSION  🛫 Publish artifacts         [aliases: pp, publish]
  craft targets              List defined targets as JSON array
  craft config               Print the parsed, processed, and validated Craft
                             config for the current project in pretty-JSON.
  craft artifacts <command>  📦 Manage artifacts          [aliases: a, artifact]

Options:
  --no-input     Suppresses all user prompts                    [default: false]
  --dry-run      Dry run mode: do not perform any real actions
  --log-level    Logging level
          [choices: "Fatal", "Error", "Warn", "Log", "Info", "Success", "Debug",
                                 "Trace", "Silent", "Verbose"] [default: "Info"]
  -v, --version  Show version number                                   [boolean]
  -h, --help     Show help                                             [boolean]

Version naming conventions

Craft currently supports semantic versioning (semver)-like versions for the NEW-VERSION argument passed to its prepare and publish commands. This means, releases made with craft need to follow a general pattern as follows:

<major>.<minor>.<patch>(-<prerelease>)?(-<build>)?
  • The <major>, <minor> and <patch> numbers are required
  • The <prerelease> and <build> identifiers are optional

Preview releases (<prerelease>)

Preview or pre-release identifiers must include one of the following identifiers

preview|pre|rc|dev|alpha|beta|unstable|a|b

and may additionally include incremental pre-release version numbers. Adding identifiers other than the ones listed above result in Craft either rejecting the release (if not parse-able) or the release being treated by individual targets as a stable release.

Examples:

1.0.0-preview
1.0.0-alpha.0
1.0.0-beta.1
1.0.0-rc.20
1.0.0-a

// invalid or incorrectly treated
1.0.0-foo
1.0.0-canary.0

Special Case: Python Post Releases

Python has the concept of post releases, which craft handles implicitly. A post release is indicated by a -\d+ suffix to the semver version, for example: 1.0.0-1. Given that we only consider certain identifiers as pre-releases, post releases are considered stable releases.

Build identifiers (<build>)

Craft supports adding a build identifier to your version, for example if you release the same package version for different platforms or architectures. You can also combine build and pre-release identifiers but in this case, the pre-release identifier has to come first.

Examples:

// valid
1.0.0+x86_64
1.0.0-rc.1+x86_64

// invalid or incorrectly treated
1.0.0+rc.1+x86_64
1.0.0+x86_64-beta.0

Caveats

  • When interacting with remote GitHub repositories, craft uses the remote origin by default. If you have a different setup, set the CRAFT_REMOTE environment variable or the --remote option to the git remote you are using.

Global Configuration

Global configuration for craft can be done either by using environment variables or by adding values to a configuration file (see below).

All command line flags can be set through environment variables by prefixing them with CRAFT_ and converting them to UPPERCASE_UNDERSCORED versions:

CRAFT_LOG_LEVEL=Debug
CRAFT_DRY_RUN=1
CRAFT_NO_INPUT=0

Since Craft heavily relies on GitHub, it needs the GITHUB_TOKEN environment variable to be set to a proper GitHub Personal Access Token for almost anything. The token only needs repo scope (repo:status and public_repo subscopes, to be precise).

Additional environment variables may be required when publishing to specific targets (e.g. TWINE_USERNAME and TWINE_PASSWORD for PyPI target).

Environment Files

craft will read configuration variables (keys, tokens, etc.) from the following locations:

  • $HOME/.craft.env
  • $PROJECT_DIR/.craft.env
  • the shell's environment

where $HOME is the current user's home directory, and $PROJECT_DIR is the directory where .craft.yml is located.

These locations will be checked in the order specified above, with values found in one location overwriting anything found in previous locations. In other words, environment variables will take precedence over either configuration file, and the project-specific file will take precedence over the file in $HOME.

The env files must be written in shell (sh/bash) format. Leading export is allowed.

Example:

# ~/.craft.env
GITHUB_TOKEN=token123
export NUGET_API_TOKEN=abcdefgh

Workflow

craft prepare: Preparing a New Release

This command will create a new release branch, check the changelog entries, run a version-bumping script, and push this branch to GitHub. We expect that CI triggered by pushing this branch will result in release artifacts being built and uploaded to the artifact provider you wish to use during the subsequent publish step.

craft prepare NEW-VERSION

🚢 Prepare a new release branch

Positionals:
  NEW-VERSION  The new version you want to release           [string] [required]

Options:
  --no-input       Suppresses all user prompts                  [default: false]
  --dry-run        Dry run mode: do not perform any real actions
  --log-level      Logging level
          [choices: "Fatal", "Error", "Warn", "Log", "Info", "Success", "Debug",
                                 "Trace", "Silent", "Verbose"] [default: "Info"]
  --rev, -r        Source revision (git SHA or tag) to prepare from (if not
                   branch head)                                         [string]
  --no-push        Do not push the release branch     [boolean] [default: false]
  --no-git-checks  Ignore local git changes and unsynchronized remotes
                                                      [boolean] [default: false]
  --no-changelog   Do not check for changelog entries [boolean] [default: false]
  --publish        Run "publish" right after "release"[boolean] [default: false]
  --remote         The git remote to use when pushing
                                                    [string] [default: "origin"]
  -v, --version    Show version number                                 [boolean]
  -h, --help       Show help                                           [boolean]

craft publish: Publishing the Release

The command will find a release branch for the provided version. The normal flow is for this release branch to be created automatically by craft prepare, but that's not strictly necessary. Then, it subscribes to the latest status checks on that branch. Once the checks pass, it downloads the release artifacts from the artifact provider configured in .craft.yml and uploads them to the targets named on the command line (and pre-configured in .craft.yml).

craft publish NEW-VERSION

🛫 Publish artifacts

Positionals:
  NEW-VERSION  Version to publish                            [string] [required]

Options:
  --no-input         Suppresses all user prompts                [default: false]
  --dry-run          Dry run mode: do not perform any real actions
  --log-level        Logging level
          [choices: "Fatal", "Error", "Warn", "Log", "Info", "Success", "Debug",
                                 "Trace", "Silent", "Verbose"] [default: "Info"]
  --target, -t       Publish to this target
    [string] [choices: "npm", "gcs", "registry", "docker", "github", "gh-pages",
                                                 "all", "none"] [default: "all"]
  --rev, -r          Source revision (git SHA or tag) to publish (if not release
                     branch head)                                       [string]
  --no-merge         Do not merge the release branch after publishing
                                                      [boolean] [default: false]
  --keep-branch      Do not remove release branch after merging it
                                                      [boolean] [default: false]
  --keep-downloads   Keep all downloaded files        [boolean] [default: false]
  --no-status-check  Do not check for build status    [boolean] [default: false]
  -v, --version      Show version number                               [boolean]
  -h, --help         Show help                                         [boolean]

Example

Let's imagine we want to release a new version of our package, and the version in question is 1.2.3.

We run prepare command first:

$ craft prepare 1.2.3

After some basic sanity checks this command creates a new release branch release/1.2.3, runs the version-bumping script (scripts/bump-version.sh), commits the changes made by the script, and then pushes the new branch to GitHub. At this point CI systems kick in, and the results of those builds, as well as built artifacts (binaries, NPM archives, Python wheels) are gradually uploaded to GitHub.

To publish the built artifacts we run publish:

$ craft publish 1.2.3

This command will find our release branch (release/1.2.3), check the build status of the respective git revision in GitHub, and then publish available artifacts to configured targets (for example, to GitHub and NPM in the case of Craft).

Configuration File: .craft.yml

Project configuration for craft is stored in .craft.yml configuration file, located in the project root.

GitHub project

Craft tries to determine the GitHub repo information from the local git repo and its remotes configuration. However, since publish command does not require a local git checkout, you may want to hard-code this information into the configuration itself:

github:
  owner: getsentry
  repo: sentry-javascript

Pre-release Command

This command will run on your newly created release branch as part of prepare command. By default, it is set to bash scripts/bump-version.sh. Please refer to the Pre-release version bumping script conventions section for more details.

preReleaseCommand: bash scripts/bump-version.sh

Post-release Command

This command will run after a successful publish. By default, it is set to bash scripts/post-release.sh. It will not error if the default script is missing though, as this may not be needed by all projects. Please refer to the Post-release script conventions section for more details.

postReleaseCommand: bash scripts/post-release.sh

Release Branch Name

This overrides the prefix for the release branch name. The full branch name used for a release is {releaseBranchPrefix}/{version}. The prefix defaults to "release".

releaseBranchPrefix: publish

Changelog Policies

craft can help you to maintain change logs for your projects. At the moment, craft supports two approaches: simple, and auto to changelog management.

In simple mode, craft prepare will remind you to add a changelog entry to the changelog file (CHANGELOG.md by default).

In auto mode, craft prepare will use the following logic:

  1. If there's already an entry for the given version, use that
  2. Else if there is an entry named Unreleased, rename that to the given version
  3. Else, create a new section for the version and populate it with the changes since the last version. It uses GitHub Milestones to provide a concise and rich changelog. If the PRs are associated with a milestone, the milestone title and description are used as the changelog entry alongside a brief list of associated PRs. Any individual commits and PRs are listed under the "Various improvements & fixes" section at the bottom. Check out Craft's own releases as example.

Configuration

Option Description
changelog optional. Path to the changelog file. Defaults to CHANGELOG.md
changelogPolicy optional. Changelog management mode (none, simple, or auto). Defaults to none.

Example (simple):

changelog: CHANGES
changelogPolicy: simple

Valid changelog example:

## 1.3.5

* Removed something

## 1.3.4

* Added something

Example (auto):

changelog: CHANGES
changelogPolicy: auto

Changelog with staged changes example:

## Unreleased

* Removed something

## 1.3.4

* Added something

Additionally, .craft.yml is used for listing targets where you want to publish your new release.

Minimal Version

It is possible to specify minimal craft version that is required to work with your configuration.

Example:

minVersion: '0.5.0'

Required Files

You can provide a list of patterns for files that have to be available before proceeding with publishing. In other words, for every pattern in the given list there has to be a file present that matches that pattern. This might be helpful to ensure that we're not trying to do an incomplete release.

Example:

requireNames:
  - /^sentry-craft.*\.tgz$/
  - /^gh-pages.zip$/

Status Provider

You can configure which status providers craft will use to check for your build status. By default, it will use GitHub but you can add more providers if needed.

Configuration

Option Description
name Name of the status provider: only github (default) for now.
config In case of github: may include contexts key that contains a list of required contexts (checks)

Example:

statusProvider:
  name: github
  config:
    contexts:
      - Travis CI - Branch

Artifact Provider

You can configure which artifact providers craft will use to fetch artifacts from. By default, GitHub is used, but in case you don't need use any artifacts in your project, you can set it to none.

Configuration

Option Description
name Name of the artifact provider: github (default), gcs, or none

Example:

artifactProvider:
  name: none

Target Configurations

The configuration specifies which release targets to run for the repository. To run more targets, list the target identifiers under the targets key in .craft.yml.

Example:

targets:
  - name: npm
  - name: github
  - name: registry
    id: browser
    type: sdk
    onlyIfPresent: /^sentry-browser-.*\.tgz$/
    includeNames: /\.js$/
    checksums:
      - algorithm: sha384
        format: base64
    config:
      canonical: 'npm:@sentry/browser'
  - name: registry
    id: node
    type: sdk
    onlyIfPresent: /^sentry-node-.*\.tgz$/
    config:
      canonical: 'npm:@sentry/node'

Per-target options

The following options can be applied to every target individually:

Name Description
includeNames optional. Regular expression: only matched files will be processed by the target. There is one special case that includeNames supports.
excludeNames optional. Regular expression: the matched files will be skipped by the target. Matching is performed after testing for inclusion (via includeNames).
id optional. A unique id for the target type so one can refer to that target individually with the -t option with the publish command like -t registry[browser]. (see the example config above)

If neither option is included, all artifacts for the release will be processed by the target.

Example:

targets:
  - name: github
    includeNames: /^.*\.exe$/
    excludeNames: /^test.exe$/

GitHub (github)

Create a release on Github. If a Markdown changelog is present in the repository, this target tries to read the release name and description from the changelog. Otherwise, defaults to the tag name and tag's commit message.

If previewReleases is set to true (which is the default), the release created on GitHub will be marked as a pre-release version if the release name contains any one of pre-release identifiers.

Environment

Name Description
GITHUB_TOKEN Personal GitHub API token (see https://github.com/settings/tokens)

Configuration

Option Description
tagPrefix optional. Prefix for new git tags (e.g. "v"). Empty by default.
previewReleases optional. Automatically detect and create preview releases. true by default.
tagOnly optional. If set to true, only create a tag (without a GitHub release).false by default.

Example:

targets:
  - name: github
    tagPrefix: v
    previewReleases: false

NPM (npm)

Releases an NPM package to the public registry. This requires a package tarball generated by npm pack in the artifacts. The file will be uploaded to the registry with npm publish, or with yarn publish if npm is not found. This requires NPM to be authenticated with sufficient permissions to publish the package.

Environment

The npm utility must be installed on the system.

Name Description
NPM_TOKEN An automation token allowed to publish.
NPM_BIN optional. Path to the npm executable. Defaults to npm
YARN_BIN optional. Path to the yarn executable. Defaults to yarn
CRAFT_NPM_USE_OTP optional. If set to "1", you will be asked for an OTP (for 2FA)

Configuration

Option Description
access optional. Visibility for scoped packages: restricted (default) or public
checkPackageName optional. If defined, check this package on the registry to get the current latest version to compare for the latest tag. The package(s) to be published will only be tagged with latest if the new version is greater than the checked package's version

Example

targets:
  - name: npm
    access: public

Python Package Index (pypi)

Uploads source dists and wheels to the Python Package Index via twine. The source code bundles and/or wheels must be in the release assets.

Environment

The twine Python package must be installed on the system.

Name Description
TWINE_USERNAME User name for PyPI with access rights for the package
TWINE_PASSWORD Password for the PyPI user
TWINE_BIN optional. Path to twine. Defaults to twine

Configuration

none

Example

targets:
  - name: pypi

Sentry internal PyPI (sentry-pypi)

Creates a GitHub pull request to import the package into a repo set up like getsentry/pypi

Environment

Name Description
GITHUB_TOKEN Personal GitHub API token (see https://github.com/settings/tokens)

Configuration

Option Description
internalPypiRepo GitHub repo containing pypi metadata

Example

targets:
  - name: pypi
  - name: sentry-pypi
    internalPypiRepo: getsentry/pypi

Homebrew (brew)

Pushes a new or updated homebrew formula to a brew tap repository. The formula is committed directly to the master branch of the tap on GitHub, therefore the bot needs rights to commit to master on that repository. Therefore, formulas on homebrew/core are not supported, yet.

The tap is configured with the mandatory tap parameter in the same format as the brew utility. A tap <org>/<name> will expand to the GitHub repository github.com:<org>/homebrew-<name>.

The formula contents are given as configuration value and can be interpolated with Mustache template syntax ({{ variable }}). The interpolation context contains the following variables:

  • version: The new version
  • revision: The tag's commit SHA
  • checksums: A map containing sha256 checksums for every release asset. Use the full filename to access the sha, e.g. checksums.MyProgram-x86. If the filename contains dots (.), they are being replaced with __. If the filename contains the currently released version, it is replaced with __VERSION__. For example, sentry-wizard-v3.9.3.tgz checksums will be accessible by the key checksums.sentry-wizard-v__VERSION____tgz.

Environment

Name Description
GITHUB_TOKEN Personal GitHub API token (seeh ttps://github.com/settings/tokens)

Configuration

Option Description
tap The name of the homebrew tap used to access the GitHub repo
template The template for contents of the formula file (ruby code)
formula optional. Name of the formula. Defaults to the repository name
path optional. Path to store the formula in. Defaults to Formula

Example

targets:
  - name: brew
    tap: octocat/tools # Expands to github.com:octocat/homebrew-tools
    formula: myproject # Creates the file myproject.rb
    path: HomebrewFormula # Creates the file in HomebrewFormula/
    template: >
      class MyProject < Formula
        desc "This is a test for homebrew formulae"
        homepage "https://github.com/octocat/my-project"
        url "https://github.com/octocat/my-project/releases/download/{{version}}/binary-darwin"
        version "{{version}}"
        sha256 "{{checksums.binary-darwin}}"

        def install
          mv "binary-darwin", "myproject"
          bin.install "myproject"
        end
      end

NuGet (nuget)

Uploads packages to NuGet via .NET Core. By default, craft publishes all packages with .nupkg extension.

Environment

The dotnet tool must be available on the system.

Name Description
NUGET_API_TOKEN NuGet personal API token
NUGET_DOTNET_BIN optional. Path to .NET Core. Defaults to dotnet

Configuration

none

Example

targets:
  - name: nuget

Rust Crates (crates)

Publishes a single Rust package or entire workspace on the public crate registry (crates.io). If the workspace contains multiple crates, they are published in an order depending on their dependencies.

Environment

"cargo" must be installed and configured on the system.

Name Description
CRATES_IO_TOKEN The access token to the crates.io account
CARGO_BIN optional. Path to cargo. Defaults to cargo.

Configuration

Option Description
noDevDeps optional. Strips devDependencies from crates before publishing. This is useful if a workspace crate uses circular dependencies for docs. Requires cargo-hack installed. Defaults to false.

Example

targets:
  - name: crates
    noDevDeps: false

Google Cloud Storage (gcs)

Uploads artifacts to a bucket in Google Cloud Storage.

The bucket paths (paths) can be interpolated using Mustache syntax ({{ variable }}). The interpolation context contains the following variables:

  • version: The new project version
  • revision: The SHA revision of the new version

Environment

Google Cloud credentials can be provided using either of the following two environment variables.

Name Description
CRAFT_GCS_TARGET_CREDS_PATH Local filesystem path to Google Cloud credentials (service account file)
CRAFT_GCS_TARGET_CREDS_JSON Full service account file contents, as a JSON string

If defined, CRAFT_GCS_TARGET_CREDS_JSON will be preferred over CRAFT_GCS_TARGET_CREDS_PATH.

Note: CRAFT_GCS_TARGET_CREDS_JSON and CRAFT_GCS_TARGET_CREDS_PATH were formerly called CRAFT_GCS_CREDENTIALS_JSON and CRAFT_GCS_CREDENTIALS_PATH, respectively. While those names will continue to work for the foreseeable future, you'll receive a warning encouraging you to switch to the new names.

Configuration

Option Description
bucket The name of the GCS bucket where artifacts are uploaded.
paths A list of path objects that represent bucket paths.
paths.path Template-aware bucket path, which can contain {{ version }} and/or {{ revision }}.
paths.metadata optional Metadata for uploaded files. By default, it sets Cache-Control to "public, max-age=300".

Example

targets:
  - name: gcs
    bucket: bucket-name
    paths:
      - path: release/{{version}}/download
        metadata:
          cacheControl: `public, max-age=3600`
      - path: release/{{revision}}/platform/package

GitHub Pages (gh-pages)

Extracts an archive with static assets and pushes them to the specified git branch (gh-pages by default). Thus, it can be used to publish documentation or any other assets to GitHub Pages, so they will be later automatically rendered by GitHub.

By default, this target will look for an artifact named gh-pages.zip, extract it, and commit its contents to gh-pages branch.

WARNING! The destination branch will be completely overwritten by the contents of the archive.

Environment

none

Configuration

Option Description
branch optional The name of the branch to push the changes to. gh-pages by default.
githubOwner optional GitHub project owner, defaults to the value from the global configuration.
githubRepo optional GitHub project name, defaults to the value from the global configuration.

Example

targets:
  - name: gh-pages
    branch: gh-pages

Sentry Release Registry (registry)

The target will update the Sentry release registry repo(https://github.com/getsentry/sentry-release-registry/) with the latest version of the project craft is used with. The release registry repository will be checked out locally, and then the new version file will be created there, along with the necessary symbolic links.

Two package types are supported: "sdk" and "app". Type "sdk" means that the package is uploaded to one of the public registries (PyPI, NPM, Nuget, etc.), and that the corresponding package directory can be found inside "packages" directory of the release regsitry. Type "app" indicates that the package's version files are located in "apps" directory of the registry.

It is strongly discouraged to have multiple registry targets in a config as it supports grouping/batching multiple apps and SDKs in a single target.

Environment

none

Configuration

Option Description
apps List of app configs as a dict, keyed by their canonical names (example: app:craft)
sdks List of sdk configs as a dict, keyed by their canonical names (example: maven:io.sentry:sentry)
(sdks|apps).urlTemplate optional URL template that will be used to generate download links for "app" package type.
(sdks|apps).linkPrereleases optional Update package versions even if the release is a preview release, "false" by default.
(sdks|apps).checksums optional A list of checksums that will be computed for matched files (see includeNames). Every checksum entry is an object with two attributes: algorithm (one of sha256, sha384, and sha512) and format (base64 and hex).
(sdks|apps).onlyIfPresent optional A file pattern. The target will be executed only when the matched file is found.

Example

targets:
  - name: registry
    sdks:
      'npm:@sentry/browser':
    apps:
      'npm:@sentry/browser':
        urlTemplate: 'https://example.com/{{version}}/{{file}}'
        checksums:
          - algorithm: sha256
            format: hex

Cocoapods (cocoapods)

Pushes a new podspec to the central cocoapods repository. The Podspec is fetched from the Github repository with the revision that is being released. No release assets are required for this target.

Environment

The cocoapods gem must be installed on the system.

Name Description
COCOAPODS_TRUNK_TOKEN The access token to the cocoapods account
COCOAPODS_BIN optional. Path to pod executable.

Configuration

Option Description
specPath Path to the Podspec file in the repository

Example

targets:
  - name: cocoapods
    specPath: MyProject.podspec

Docker (docker)

Copies an existing source image tagged with the revision SHA to a new target tagged with the released version. No release assets are required for this target except for the source image at the provided source image location so it would be a good idea to add a status check that ensures the source image exists, otherwise craft publish will fail at the copy step, causing an interrupted publish. This is an issue for other, non-idempotent targets, not for the Docker target.

Environment

docker executable (or something equivalent) with BuildKit must be installed on the system.

Name Description
DOCKER_USERNAME The username for the Docker registry.
DOCKER_PASSWORD The personal access token for the account.
DOCKER_BIN optional. Path to docker executable.

Configuration

Option Description
source Path to the source Docker image to be pulled
sourceFormat Format for the source image name. Default: {{{source}}}:{{{revision}}}
target Path to the target Docker image to be pushed
targetFormat Format for the target image name. Default: {{{target}}}:{{{version}}}

Example

targets:
  - name: docker
    source: us.gcr.io/sentryio/craft
    target: getsentry/craft
# Optional but strongly recommended
statusProvider:
  name: github
  config:
    contexts:
      - Travis CI - Branch # or whatever builds and pushes your source image

Ruby Gems Index (gem)

Pushes a gem Ruby Gems. It also requires you to be logged in with gem login.

Environment

gem must be installed on the system.

Name Description
GEM_BIN optional. Path to "gem" executable. Defaults to gem

Configuration

none

Example

targets:
  - name: gem

AWS Lambda Layer (aws-lambda-layer)

The target will create a new public lambda layer in each available region with the extracted artifact from the artifact provider, and update the Sentry release registry with the new layer versions afterwards.

Environment

Name Description
AWS_ACCESS_KEY The access key of the AWS account to create and publish the layers.
AWS_SECRET_ACCESS_KEY The secret access key of the AWS account to create and publish the layers.

Configuration

Option Description
linkPrereleases optional Updates layer versions even if the release is a preview release, false by default.
includeNames optional Exists for all targets, see here. It must filter exactly one artifact.
layerName The name of the layer to be published.
compatibleRuntimes A list of compatible runtimes for the layer. Each compatible runtime consists on the name of the runtime and a list of compatible versions.
license The license of the layer.

Example

targets:
  - name: aws-lambda-layer
    includeNames: /^sentry-node-serverless-\d+(\.\d+)*\.zip$/
    layerName: SentryNodeServerlessSDK
    compatibleRuntimes:
      - name: node
        versions:
          - nodejs10.x
          - nodejs12.x
    license: MIT

Unity Package Manager (upm)

Pulls the package as a zipped artifact and pushes the unzipped content to the target repository, tagging it with the provided version.

WARNING! The destination repository will be completely overwritten.

Environment

none

Configuration

Option Description
releaseRepoOwner Name of the owner of the release target
releaseRepoName Name of the repo of the release target

Example

targets:
  - name: upm
    releaseRepoOwner: 'getsentry'
    releaseRepoName: 'unity'

Maven central (maven)

PGP signs and publishes packages to Maven Central.

Note: in order to see the output of the commands, set the logging level to trace.

Environment

Name Description
OSSRH_USERNAME Username of Sonatype repository.
OSSRH_PASSWORD Password of Sonatype repository.
GPG_PASSPHRASE Passphrase for your default GPG Private Key.
GPG_PRIVATE_KEY optional GPG Private Key generated via gpg --armor --export-secret-keys YOUR_ID. If not provided, default key from your machine will be used.

Configuration

Option Description
mavenCliPath Path to the Maven CLI. It must be executable by the calling process.
mavenSettingsPath Path to the Maven settings.xml file.
mavenRepoId ID of the Maven server in the settings.xml.
mavenRepoUrl URL of the Maven repository.
android Android configuration, see below.
kmp Kotlin Multiplatform configuration, see below.

The Kotlin Multiplatform configuration is optional and false by default. If your project isn't related to Android, you don't need this configuration and can set the option to false. If not, set the following nested elements:

  • distDirRegex: pattern of distribution directory names.
  • fileReplaceeRegex: pattern of substring of distribution module names to be replaced to get the Android distribution file.
  • fileReplacerStr: string to be replaced in the module names to get the Android distribution file.

Example (without Android config)

targets:
  - name: maven
    mavenCliPath: scripts/mvnw.cmd
    mavenSettingsPath: scripts/settings.xml
    mavenRepoId: ossrh
    mavenRepoUrl: https://oss.sonatype.org/service/local/staging/deploy/maven2/
    android: false

Example (with Android config)

targets:
  - name: maven
    mavenCliPath: scripts/mvnw.cmd
    mavenSettingsPath: scripts/settings.xml
    mavenRepoId: ossrh
    mavenRepoUrl: https://oss.sonatype.org/service/local/staging/deploy/maven2/
    android:
      distDirRegex: /^sentry-android-.*$/
      fileReplaceeRegex: /\d\.\d\.\d(-SNAPSHOT)?/
      fileReplacerStr: release.aar

Example (with Kotlin Multiplatform config)

targets:
  - name: maven
    mavenCliPath: scripts/mvnw.cmd
    mavenSettingsPath: scripts/settings.xml
    mavenRepoId: ossrh
    mavenRepoUrl: https://oss.sonatype.org/service/local/staging/deploy/maven2/
    android:
      distDirRegex: /^sentry-android-.*$/
      fileReplaceeRegex: /\d\.\d\.\d(-SNAPSHOT)?/
      fileReplacerStr: release.aar
    kmp:
      rootDistDirRegex: /sentry-kotlin-multiplatform-[0-9]+.*$/
      appleDistDirRegex: /sentry-kotlin-multiplatform-(macos|ios|tvos|watchos).*/

Symbol Collector (symbol-collector)

Using the symbol-collector client, uploads native symbols. The symbol-collector needs to be available in the path.

Configuration

Option Description
serverEndpoint optional The server endpoint. Defaults to https://symbol-collector.services.sentry.io.
batchType The batch type of the symbols to be uploaded. I.e: Android, macOS, iOS.
bundleIdPrefix The prefix of the bundle ID. The new version will be appended to the end of this prefix.

Example

targets:
  - name: symbol-collector
    includeNames: /libsentry(-android)?\.so/
    batchType: Android
    bundleIdPrefix: android-ndk-

pub.dev (pub-dev)

Pushes a new Dart or Flutter package to pub.dev.

Because there is no automated way to login and obtain required tokens, you need to perform a valid release beforehand, for every package that you configure. This will open up your browser and use Google's OAuth to log you in, and generate an appropriate file with stored credentials.

Based on your environment, you can find this file at either $HOME/.pub-cache/credentials.json or $HOME/Library/Application\ Support/dart/pub-credentials.json for OSX and $HOME/.config/dart/pub-credentials.json for Linux, depending on your setup.

For this target to work correctly, either dart must be installed on the system or a valid dartCliPath must be provided.

Environment

Name Description
PUBDEV_ACCESS_TOKEN Value of accessToken obtained from pub-credentials.json
PUBDEV_REFRESH_TOKEN Value of refreshToken obtained from pub-credentials.json

Configuration

Option Description
dartCliPath optional Path to the Dart CLI. It must be executable by the calling process. Defaults to dart.
packages optional List of directories to be released, relative to the root. Useful when a single repository contains multiple packages. When skipped, root directory is assumed as the only package.
skipValidation optional Publishes the package without going through validation steps, such as analyzer & dependency checks.
This is useful in particular situations when package maintainers know why the validation fails and wish to side step the issue. For example, there may be analyzer issues due to not following the current (latest) dart SDK recommendation because the package needs to maintain the package compatibility with an old SDK version.
This option should be used with caution and only after testing and verifying the reported issue shouldn't affect the package. It is advisable to do an alpha pre-release to further reduce the chance of a potential negative impact.

Example

targets:
  - name: pub-dev
    packages:
      uno:
      dos:
      tres:

Hex (hex)

Pushes a package to the Elixir / Erlang package manager Hex.

Environment

mix (bundled with the elixir language) must be installed on the system.

Name Description
HEX_API_KEY API Key obtained from hex.pm account
MIX_BIN optional. Path to "mix" executable. Defaults to mix

Configuration

none

Example

targets:
  - name: hex

Commit on Git Repository (commit-on-git-repository)

Takes a tarball and pushes the unpacked contents to a git repository.

Environment

Name Description
GITHUB_API_TOKEN GitHub PAT that will be used for authentication when a the repositoryUrl host is github.com.

Configuration

Option Description
archive Regular expression to match a .tgz file in the build artifacts. The content of the found file will be pushed to the git repository. Needs to match exactly one file.
repositoryUrl Url to the git remote git repository. Must use http or https protocol! (no git@...)
branch Which repository branch to push to.
stripComponents optional. How many leading path elements should be removed when unpacking the tarball. Default: 0 (see tar --strip-components option)
createTag optional. Whether to attach a tag to the created commit. The content of the tag is gonna be equal to the release version passed to craft ("NEW-VERSION"). Default: false

Example

targets:
  - name: commit-on-git-repository
    archive: /^sentry-deno-\d.*\.tgz$/
    repositoryUrl: https://github.com/getsentry/sentry-deno
    stripComponents: 1
    branch: main
    createTag: true

PowerShellGet (powershell)

Uploads a module to PowerShell Gallery or another repository supported by PowerShellGet's Publish-Module.

The action looks for an artifact named <module>.zip and extracts it to a temporary directory. The extracted directory is then published as a module.

Environment

The pwsh executable must be installed on the system.

Name Description Default
POWERSHELL_API_KEY required PowerShell Gallery API key
POWERSHELL_BIN optional Path to PowerShell binary pwsh

Configuration

Option Description Default
module required Module name.
repository optional Repository to publish the package to. PSGallery

Example

targets:
  - name: powershell
    module: Sentry

Integrating Your Project with craft

Here is how you can integrate your GitHub project with craft:

  1. Set up a workflow that builds your assets and runs your tests. Allow building release branches (their names follow release/{VERSION} by default, configurable through releaseBranchPrefix).

    on:
  push:
    branches:
      - 'release/**'

  2. Use the official actions/upload-artifact@v2 action to upload your assets. Here is an example config (step) of an archive job:

    - name: Archive Artifacts
  uses: actions/upload-artifact@v2
  with:
    name: ${{ github.sha }}
    path: |
      ${{ github.workspace }}/*.tgz
      ${{ github.workspace }}/packages/tracing/build/**
      ${{ github.workspace }}/packages/**/*.tgz

    A few important things to note:

    • The name of the artifacts is very important and needs to be name: ${{ github.sha }}. Craft uses this as a unique id to fetch the artifacts.
    • Keep in mind that this action maintains the folder structure and zips everything together. Craft will download the zip and recursively walk it to find all assets.

  3. Add .craft.yml configuration file to your project

    • List there all the targets you want to publish to
    • Configure additional options (changelog management policy, tag prefix, etc.)

  4. Add a pre-release script to your project.

  5. Get various configuration tokens

  6. Run craft prepare <version> --publish and profit!

Pre-release (Version-bumping) Script: Conventions

Among other actions, craft prepare runs an external, project-specific command or script that is responsible for version bumping. By default, this script should be located at: ./scripts/bump-version.sh. The command can be configured by specifying the preReleaseCommand configuration option in craft.yml.

The following requirements are on the script interface and functionality:

  • The script should accept at least two arguments. Craft will pass the old ("from") version and the new ("to") version as the last two arguments, respectively.
  • The script must replace all relevant occurrences of the old version string with the new one.
  • The script must not commit the changes made.
  • The script must not change the state of the git repository (e.g. changing branches)

Example

#!/bin/bash
### Example of a version-bumping script for an NPM project.
### Located at: ./scripts/bump-version.sh
set -eux
OLD_VERSION="${1}"
NEW_VERSION="${2}"

# Do not tag and commit changes made by "npm version"
export npm_config_git_tag_version=false
npm version "${NEW_VERSION}"

Post-release Script: Conventions

Among other actions, craft publish runs an external, project-specific command or script that can do things like bumping the development version. By default, this script should be located at: ./scripts/post-release.sh. Unlike the pre-release command, this script is not mandatory so if the file does not exist, craft will report this fact and then move along as usual. This command can be configured by specifying postReleaseCommand configuration option in craft.yml.

The following requirements are on the script interface and functionality:

  • The script should accept at least two arguments. Craft will pass the old ("from") version and the new ("to") version as the last two arguments, respectively.
  • The script is responsible for any and all git state management as craft will simply exit after running this script as the final step. This means the script is responsible for committing and pushing any changes that it may have made.

Example

#!/bin/bash
### Example of a dev-version-bumping script for a Python project
### Located at: ./scripts/post-release.sh
set -eux
OLD_VERSION="${1}"
NEW_VERSION="${2}"

# Ensure master branch
git checkout master
# Advance the CalVer release by one-month and add the `.dev0` suffix
./scripts/bump-version.sh '' $(date -d "$(echo $NEW_VERSION | sed -e 's/^\([0-9]\{2\}\)\.\([0-9]\{1,2\}\)\.[0-9]\+$/20\1-\2-1/') 1 month" +%y.%-m.0.dev0)
# Only commit if there are changes, make sure to `pull --rebase` before pushing to avoid conflicts
git diff --quiet || git commit -anm 'meta: Bump new development version' && git pull --rebase && git push

Development

Logging Level

Logging level for craft can be configured via setting the CRAFT_LOG_LEVEL environment variable or using the --log-level CLI flag.

Accepted values are: Fatal, Error, Warn, Log, Info, Success, Debug, Trace, Silent, Verbose

Dry-run Mode

Dry-run mode can be enabled via setting the CRAFT_DRY_RUN environment variable to any truthy value (any value other than undefined, null, "", 0, false, and no). One may also use the --dry-run CLI flag.

In dry-run mode no destructive actions will be performed (creating remote branches, pushing tags, committing files, etc.)

Sentry Support

Errors you encounter while using Craft can be sent to Sentry. To use this feature, add CRAFT_SENTRY_DSN variable to your environment (or "craft" configuration file) that contains a Sentry project's DSN.

For example:

export CRAFT_SENTRY_DSN='https://1234@sentry.io/2345'

Releasing

craft obviously uses craft for preparing and publishing new releases!

Did you mean recursion?