docs/tests/README.md at 03bbbcf44a2f2172a1349e52f97aa6ea82de1571

mirror of synced 2026-01-10 00:03:04 -05:00

Files

Mike Surowiec 37f73f0bb3 Heroku dev deps (#19431 )

* fix: req.csrfToken doesn't always exist (e.g. 500 page)

* feat: update dockerfile and add nextjs to build

* fix: run linter

* move @babel deps -> dev deps

* move webpack looking things from deps -> dev deps

* move pa11y-ci to optional dep

* explicitly include optional deps for pa11y

* allow heroku dev deps to be installed

* fix: update postcss module

* fix: update dockerfile build

* tmp: disable renderReact

* see if another deploy is slower/faster

* move a few more packages to devDeps

* upgrade to package-lock v2

* use dayjs instead of date-fns

* move cross-env to devDeps

* remove unused 'del' package

* commit husky precommit hooks

* add hrtime to clone-for-build.js

* Revert "add hrtime to clone-for-build.js"

This reverts commit 70ee647bacce833f4ed2f621f62c63c1d85e5413.

* update babel/eslint

* fix: remove unused plugin

* try a .slugignore

* fix: heroku-postbuild to use npm run build

* fix: i cannot spell dereferenced

* add .next/cache to heroku cacheDirectories

* test cached build

* remove aws-sdk, see what breaks

* move jest-puppeteer to optional deps

* fix: update browser-test.yml to use newer node version

* move jimp to optional dependencies

* move puppeteer to optional dependencies

* fix: ci optional include

* fix: bad copy pasta

* remove previous react experiment

* update tests/README.md with note about optional deps

* bump node test version back to 14

* convert package-lock back to v1

* fix: use node 15.x to leverage npm optional deps

* fix: optional dep install

* test: see what happens with heroku/nodejs-typescript buildpack

* back to heroku/nodejs buildpack

* move jest to optional

* revert jest move

* remove .slugignore

* cleanup dockerfile, move xlsx-population to optional, add comment about optional deps

* Update Dockerfile

Co-authored-by: James M. Greene <JamesMGreene@github.com>

Co-authored-by: James M. Greene <JamesMGreene@github.com>

2021-05-24 15:40:50 -07:00

2.7 KiB

Raw Blame History

Tests

It's not strictly necessary to run tests locally while developing: You can always open a pull request and rely on the CI service to run tests for you, but sometimes it's helpful to run tests locally before pushing your changes to GitHub.

Test are written using jest, a framework maintained by Facebook and used by many teams at GitHub. Jest is convenient in that it provides everything: a test runner, an assertion library, code coverage analysis, custom reporters for different types of test output, etc.

Install optional dependencies

We typically rely on CI to run our tests, so we consider some large test-only dependencies optional (for example, puppeteer). In order to run the tests locally you'll need to make sure optional dependencies are installed by running:

npm ci --include=optional

Running all the tests

Once you've followed the development instructions above, you can run the entire test suite locally:

script/test # or `npm test`

Watching all the tests

You can also run a script that will continually watch for changes and re-run the tests any time a change is made. This command will notify you when tests change to and from a passing or failing state, and will also print out a test coverage report, so you can see what files are in need of tests.

npm run test-watch

Testing individual files

If you're making changes to a specific file and don't want to run the entire test suite, you can pass an argument to the jest testing tool:

jest __tests__/page.js

The argument doesn't have to be a fully qualified file path. It can also be a portion of a filename:

jest page # runs tests on __tests__/page.js and __tests__/pages.js

Linting

To validate all your JavaScript code (and auto-format some easily reparable mistakes), run the linter:

npm run lint

Broken link test

This test checks all internal links and image references in the English site. To run it locally (takes about 60 seconds):

npx jest links-and-images

It checks images, anchors, and links for every version of every page.

It reports five types of problems:

Broken image references
- Example: /assets/images/foo.png where foo.png doesn't exist.
Broken same-page anchors
- Example: #foo where the page does not have a heading Foo.
Broken links due to page not found
- Example: /github/using-git/foo where there is no foo.md file at that path.
Broken links due to versioning
- Example: an unversioned link to a Dotcom-only article in a page that has Enterprise versions.
Broken anchors on links
- Example: /some/valid/link#bar where the linked page can be found but it does not have a heading Bar.

2.7 KiB Raw Blame History