BREAKING CHANGE:
ESLint expects configs or plugins which require other configs or plugins to have _peer dependencies_ of those things _and_ of ESLint itself. All deps have been changed to peer deps, and this module now requires the installation of the following _if_ using npm older than v7:
```
"@babel/core": "7.16.0",
"@babel/eslint-parser": "7.16.3",
"eslint": "7.32.0",
"eslint-plugin-import": "2.25.3",
"eslint-plugin-mocha": "9.0.0",
"eslint-plugin-promise": "5.1.1"
```
npm@7 will install these automatically if they do not exist.
- add `jsconfig.json` for the TS language server to understand the project
- update `.wallaby.js` to transpile fixtures if necessary; run on save to reduce lag
- update single-file Mocha debug launch config
- add `test` env in `babel.config.json` which is used by the former
- add some handy vscode tasks
- add a bunch of types
- remove unused `babel-eslint` (we use `@babel/eslint-parser`)
- bump `ecmaVersion` in `eslint-config-appium`
- remove cruft from babel configs (`shippedProposals: true` goes a long way)
- enable `sinon-chai` globally in tests
- add `rewiremock` for module-level mocking (sinon cannot do this itself, but they work in tandem)
This PR makes `chai` and `should()` _globals_ when using Mocha:
- `chai` is configured with `chai-as-promised`
- A `.mocharc.js` requires `./test/setup.js` _from the monorepo root_.
- Updated the ESLint config to recognize the globals.
- Updated Wallaby config
All instances of `import chai...` and the like have been removed from all test files.
Just a bunch of random (mostly path-related) fixes.
## Monorepo
- Make `npm run clean` work cross-platform
- Sort root `package.json`
- Update Babel configs to suppress warning about `@babel/plugin-proposal-private-property-in-object`
- Add `.nyc_output` and `*.lerna_backup` to `.gitignore`
- Replace references to `appium-support`
- Sync `package.json` fields
- Add a [Wallaby](https://wallabyjs.com) configuration file
## appium
- `lib/config`: `package.json` resolution
- split some tests out of `config-specs.js` and into `config-e2e-specs.js` since they are very clearly not unit tests
- Improve `PROJECT_ROOT` resolution, which is still a hack
- Copy fixtures & fix paths to fixtures in tests
## @appium/base-driver
- `lib/basedriver/driver`: `package.json` resolution
- Fix broken tests when run in isolation (`chai.should()`)
## @appium/doctor
- `lib/doctor`: `package.json` resolution
- `lib/utils`: `package.json` resolution & exposed `resetLog()` which undoes what `configureBinaryLog()` does (for testing purposes)
- `utils-specs.js`: clean up after `onLogMessage` test
- copy fixtures into `build` upon transpile
- Upgrade `@appium/gulp-plugins`
## @appium/gulp-plugins
- Ensure `postTranspile` is run when `transpile` is run before tests
- Fix `.babelrc` like `babel.config.json`
## @appium/support
- Fixture resolution
- Fix broken tests when run in isolation (`chai-as-promised`)
- Copy fixtures when transpiling
1. Uses `sync-monorepo-packages` to keep `README.md` and `packages/appium/README.md` in sync. Since `README` gets published to npm but _also_ shows in the GitHub repo, both places need a `README`. In lieu of a custom `README`, we can just sync it for now. This is called via a `preversion` script.
1. Various fields are synced in `package.json`. This does not include dependencies—but those _in common_ should probably be kept in sync as well.
1. Lerna uses conventional commits and Git tags to figure out what to do. It should automatically determine what needs releasing and what version it should be released as.
1. Tags have been pushed to the repo for Lerna to use. In the case where the changesets are functionally identical to the latest code in the old repositories, the version number has been retained, and the tag is of the format `[@scope/]<package-name>@<version>`. In the case where the old repos had _unreleased_ changesets, I’ve tagged with a new patch release and adopted the `-rc.0` suffix. So if we were at v5.5.0 in `appium-gulp-plugins`, `@appium/gulp-plugins` is tagged as `@appium/gulp-plugins@5.5.1-rc.0`. I have not modified the version in `package.json`.
1. As with adopting any new release/publishing strategy, we should be meticulous about what we’re doing until we’re comfortable with how it works. Lerna supports dry runs; we should use them.
1. Documentation on usage will arrive in next changes to this PR.
- add `npm start` script to run appium server
- add `npm run doctor` convenience script to run doctor
- renamed `clean` script to `reinstall`
- added new `clean` script to just clean
- updated the "appium from source" doc
this pkg is unique in that it does not use `gulp`, so for now we tell lerna to always ignore it when running `lerna exec` across packages. it has no tests, but it's marginally useful to lint the `index.js`, so we'll do that (see `lint` script in root `package.json`).
the canonical way to handle plugin deps in a sharable config is to use `peerDependencies`, so it now uses `peerDependencies` instead of `dependencies`. these peer deps are installed by the dependencies of `@appium-gulp-plugins`. see [docs](https://eslint.org/docs/developer-guide/shareable-configs#publishing-a-shareable-config)
our root `.eslintrc` depends upon `@appium/eslint-config-appium` (which is not an unusual situation). to pull this off in a monorepo, a cheat is to `file:` schema in the dependency specifier of the root `package.json`.
also tweaked `lerna.json` to _always_ run `bootstrap` with the `--no-ci` flag. lerna will attempt to use `npm ci` if `CI` env var is present, but for that to work, we must have `package-lock.json` under version control.
Jonathan Lipps