- Run with one command and do not rely on error exit codes to stop the process when a release is not necessary
- Break `index.js` in smaller modules in order to improve testability and simplify the code
- Add several missing unit and integration tests to reach 100% coverage
- Integration tests now test end to end, including publishing to Github (with http://www.mock-server.com on Docker)
- Use `tj/commander.js` to print an help message, verify and parse CLI arguments
- Semantic-release can now be called via Javascript API: `require('semantic-release')(options)`
- Remove npmlog dependency and add more log messages
- Logger is now passed to plugins
- Add debug logs with `visionmedia/debug`. `debug` is enabled for both semantic-release and plugins with `--debug`
- Use `kevva/npm-conf` in place of the deprecated `npm/npmconf`
- Pass lastRelease, nextRelease and commits to generate-notes plugin
- In dry-run mode, print the release note instead of publishing it to Github as draft, and skip the CI verifications
- The dry-run mode does not require npm and Github TOKEN to be set anymore and can be run locally
BREAKING CHANGE: Semantic-Release must now be executed with `semantic-release` instead of `semantic-release pre && npm publish && semantic-release post`.
BREAKING CHANGE: The `semantic-release` command now returns with exit code 0 on expected exception (no release has to be done, running on a PR, gitHead not found, other CI job failed etc...). It only returns with 1 when there is an unexpected error (code error in a plugin, plugin not found, git command cannot be run etc..).
BREAKING CHANGE: Calling the `semantic-release` command with unexpected argument(s) now exit with 1 and print an help message.
BREAKING CHANGE: Semantic-Release does not rely on `npmlog` anymore and the log level cannot be configured. Debug logs can be activated with CLI option `--debug` or with environment variable `DEBUG=semantic-release:*`
BREAKING CHANGE: The CLI options `--debug` doesn't enable the dry-run mode anymore but activate the debugs. The dry run mode is now set with the CLI command `--dry-run` or `-d`.
- Use async/await instead of callbacks
- Use execa to run command line
- Use AVA for tests
- Add several assertions in the unit tests
- Add documentation (comments) in the tests
- Run tests with a real git repo instead of mocking child_process and add test helpers to create repos, commits and checkout
- Simplify test directory structure
- Simplify code readability (mostly with async/await)
- Use eslint for for linting, prettier for formatting
