Front-end infrastructure and code to complement mozilla/addons-server.
This code and its associated production website are included in Mozilla’s web and services bug bounty program. If you find a security vulnerability, please submit it via the process outlined in the program and FAQ pages. Further technical details about this application are available from the Bug Bounty Onramp page.
Please submit all security-related bugs through Bugzilla using the web security bug form.
Never submit security-related bugs through a Github Issue or by email.
- You need Node 8.x which is the current LTS (long term support) release.
- Install yarn to manage dependencies and run scripts.
The easiest way to manage multiple node versions in development is to use nvm.
- type
yarn
to install all dependencies - type
yarn amo:stage
to start a local server that connects to a hosted staging server
Here are some commands you can run:
Command | Description |
---|---|
yarn amo | Start the dev server/proxy (for amo) using data from Docker |
yarn amo:dev | Start the dev server/proxy (for amo) using data from the dev server (https://addons-dev.allizom.org/) |
yarn amo:dev-https | Same as amo:dev but with HTTPS, available at: https://example.com:3000/. Read about setting up this environment |
yarn amo:no-proxy | Start the dev server without a proxy (for amo) using data from Docker |
yarn amo:stage | Start the dev server/proxy (for amo) using data from the staging server (https://addons.allizom.org/) |
yarn build | Build an app specified with the NODE_APP_INSTANCE environment variable. |
yarn build-all | Build all the applications. |
yarn build-ci | Run the build-all and bundlesize npm scripts. |
yarn bundlesize | Run bundlesize to check the generated AMO bundle sizes. Building AMO is required first. |
yarn disco | Start the dev server (for Discovery Pane) using data from the dev server (https://addons-dev.allizom.org/) |
yarn disco:https | Same as disco but with HTTPS, available at: https://example.com:3000/. Read about setting up this environment |
yarn flow | Run Flow. By default this checks for errors and exits |
yarn flow:check | Explicitly check for Flow errors and exit |
yarn flow:dev | Continuously check for Flow errors |
yarn eslint | Lint the JS |
yarn snyk | Run snyk (without a command) |
yarn snyk-ci | Run snyk test and monitor |
yarn snyk-wizard | Run snyk wizard to fix an issue reported by snyk |
yarn start-func-test-server | Start a Docker container for functional tests |
yarn stylelint | Lint the SCSS |
yarn storybook | Run storybook |
yarn lint | Run all the JS + SCSS linters |
yarn prettier | Run Prettier to automatically format the entire codebase |
yarn prettier-dev | Run Pretty-Quick to automatically compare and format modified source files against the master branch |
yarn prettier-ci | Run Prettier and fail if some code has been changed without being formatted |
yarn version-check | Check you have the required dependencies |
yarn test | Run all tests (Enters jest in --watch mode) |
yarn test-coverage | Run all tests and generate code coverage report (Enters jest in --watch mode) |
yarn test-coverage-once | Run all tests, generate code coverage report, then exit |
yarn test-once | Run all tests, run all JS + SCSS linters, then exit |
yarn test-ci | Run all continuous integration checks. This is only meant to run on TravisCI. |
You can enter the interactive jest mode by typing yarn test
. This is the easiest way to develop new features.
Here are a few tips:
- When you start
yarn test
, you can switch to your code editor and begin adding test files or changing existing code. As you save each file, jest will only run tests related to the code you change. - If you had typed
a
when you first started then jest will continue to run the full suite even when you change specific files. Typeo
to switch back to the mode of only running tests related to the files you are changing. - Sometimes running tests related to your file changes is slow. In these cases, you can type
p
ort
to filter tests by name while you working fixing a specific test suite. More info. - If you see something like
Error watching file for changes: EMFILE
on Mac OS thenbrew install watchman
might fix it. See jestjs/jest#1767
By default, yarn test
will only run a subset of tests that relate to the code you are working on.
To explicitly run a subset of tests, you can type t
or p
which are explained in the jest watch usage.
Alternatively, you can start the test runner with a specific file or regular expression, like:
yarn test tests/unit/amo/components/TestAddon.js
If you want to run all tests and exit, type:
yarn test-once
There is limited support for using Flow to check for problems in the source code.
To check for Flow issues during development while you edit files, run:
yarn flow:dev
If you are new to working with Flow, here are some tips:
- Check out the getting started guide.
- Read through the web-ext guide for hints on how to solve common Flow errors.
To add flow coverage to a source file, put a /* @flow */
comment at the top. The more source files you can opt into Flow, the better.
Here is our Flow manifesto:
- We use Flow to declare the intention of our code and help others refactor it with confidence. Flow also makes it easier to catch mistakes before spending hours in a debugger trying to find out what happened.
- Avoid magic Flow declarations for any internal code. Just declare a type alias next to the code where it's used and export/import it like any other object.
- Never import a real JS object just to reference its type. Make a type alias and import that instead.
- Never add more type annotations than you need. Flow is really good at inferring types from standard JS code; it will tell you when you need to add explicit annotations.
- When a function like
getAllAddons
takes object arguments, call its type objectGetAllAddonsParams
. Example:
type GetAllAddonsParams = {|
categoryId: number,
|};
function getAllAddons({ categoryId }: GetAllAddonsParams = {}) {
...
}
- Use Exact object types via the pipe syntax (
{| key: ... |}
) when possible. Sometimes the spread operator triggers an error like 'Inexact type is incompatible with exact type' but that's a bug. You can use theExact<T>
workaround fromsrc/core/types/util
if you have to. This is meant as a working replacement for $Exact. - Add a type hint for components wrapped in HOCs (higher order components) so that Flow can validate calls to the component. We need to add a hint because we don't yet have decent type coverage for all the HOCs we rely on. Here is an example:
// Imagine this is something like components/ConfirmButton/index.js
import { compose } from 'redux';
import * as React from 'react';
// This expresses externally used props, i.e. to validate how the app would use <ConfirmButton />
type Props = {|
prompt?: string | null,
|};
// This expresses internally used props, such as i18n which is injected by translate()
type InternalProps = {|
...Props,
i18n: I18nType,
|};
export class ConfirmButtonBase extends React.Component<InternalProps> {
render() {
const prompt = this.props.prompt || this.props.i18n.gettext('Confirm');
return <button>{prompt}</button>;
}
}
// This provides a type hint for the final component with its external props.
// The i18n prop is not in external props because it is injected by translate() for internal use only.
const ConfirmButton: React.ComponentType<Props> = compose(translate())(
ConfirmButtonBase,
);
export default ConfirmButton;
- Try to avoid loose types like
Object
orany
but feel free to use them if you are spending too much time declaring types that depend on other types that depend on other types, and so on. - You can add a
$FLOW_FIXME
comment to skip a Flow check if you run into a bug or if you hit something that's making you bang your head on the keyboard. If it's something you think is unfixable then use$FLOW_IGNORE
instead. Please explain your rationale in the comment and link to a GitHub issue if possible. - If you're stumped on why some Flow annotations aren't working, try using the
yarn flow type-at-pos ...
command to trace which types are being applied to the code. Seeyarn flow -- --help type-at-pos
for details.
We use Prettier to automatically format our JavaScript code and stop all the on-going debates over styles. As a developer, you have to run it (with yarn prettier-dev
) before submitting a Pull Request.
We use snyk to continuously monitor our application's dependencies.
As a member of the add-ons-team
, you can fix an issue reported by running:
yarn snyk-wizard
The wizard allows you to decide whether you want to upgrade dependencies or ignore the issue for 30 days. See the existing reasons to ignore an issue in the .snyk
file. Snyk is a bit intrusive and changes many things (like re-adding snyk test
to the npm test
script): double check your changes before submitting a Pull Request. You have successfully fixed an issue when yarn snyk-ci
does not complain. Make sure you open a Pull Request with a branch pushed to this repository and not from your fork, because the snyk-ci
job (Travis CI) does not run on forks.
Note: You should authenticate yourself once by running yarn snyk auth
(no dash). It will open a link in your favorite browser and authenticate you locally.
To see a report of code coverage, type:
yarn test-coverage-once
This will print a table of files showing the percentage of code coverage. The uncovered lines will be shown in the right column but you can open the full report in a browser:
open coverage/lcov-report/index.html
A proxy server is provided for running the AMO app with the API on the same host as the frontend. This provides a setup that is closer to production than running the frontend on its own. The default configuration for this is to use a local addons-server for the API which can be setup according to the addons-server docs. Docker is the preferred method of running addons-server.
Authentication will work when initiated from addons-frontend and will persist to addons-server but it will not work when logging in from an addons-server page. See mozilla/addons-server#4684 for more information on fixing this.
If you would like to use https://addons-dev.allizom.org
for data you should use the yarn amo:dev
command. See the table of commands up above for similar hosted options.
If you need to override any settings while running yarn amo
, yarn amo:dev
, or yarn amo:stage
, first create a local config file named exactly like this:
touch config/local-development-amo.js
Make any config changes. For example:
module.exports = {
trackingEnabled: true,
};
Restart the server to see it take affect.
Consult the config file loading order docs to learn more about how configuration is applied.
When running yarn disco
, your local server will be configured for a hosted development API. If you want to run your own addons-server API or make any other local changes, you'll need to create a custom config file named exactly like this:
touch config/local-development-disco.js
Here's what local-development-disco.js
would look like when overriding the apiHost
parameter so that it points to your docker container:
module.exports = {
apiHost: 'http://olympia.test',
};
Restart the server to see it take affect.
If you want to access your local server on an Android device you will need to change a few settings. Let's say your local machine is accessible on your network at the IP address 10.0.0.1
. You could start your server like this:
API_HOST=http://10.0.0.1:3000 \
SERVER_HOST=10.0.0.1 \
WEBPACK_SERVER_HOST=10.0.0.1 \
yarn amo:dev
On your Android device, you could then access the development site at http://10.0.0.1:3000
.
NOTE: At this time, it is not possible to sign in with this configuration because the Firefox Accounts client redirects to localhost:3000
. You may be able to try a different approach by editing /etc/hosts
on your device so that localhost
points to your development machine but this has not been fully tested.
When developing locally with a webpack server, the randomly generated asset URL will fail our Content Security Policy (CSP) and clutter your console with errors. You can turn off all CSP errors by settings CSP to false
in any local config file, such as local-development-amo.js
. Example:
module.exports = {
CSP: false,
};
The documentation you are reading right now lives inside the source repository as Github flavored Markdown. When you make changes to these files you can create a pull request to preview them or, better yet, you can use grip to preview the changes locally. After installing grip
, run it from the source directory like this:
grip .
Open its localhost
URL and you will see the rendered README.md
file. As you make edits, it will update automatically.
If you run yarn storybook
, you can see storybook locally here: http://localhost:9001/.
You can find the source files under the ./stories directory.
Check it out on Github Pages: https://mozilla.github.io/addons-frontend/
More info coming soon :)
The following are scripts that are used in deployment - you generally won't need unless you're testing something related to deployment or builds.
The env vars are:
NODE_APP_INSTANCE
this is the name of the app e.g. 'disco' NODE_ENV
this is the node environment. e.g. production, dev, stage, development.
Script | Description |
---|---|
yarn start | Starts the express server (requires env vars) |
yarn build | Builds the libs (all apps) (requires env vars) |
Example: Building and running a production instance of the AMO app:
NODE_APP_INSTANCE=amo NODE_ENV=production yarn build
NODE_APP_INSTANCE=amo NODE_ENV=production yarn start
Note: To run the app locally in production mode you'll need to create a config file for local production builds. It must be saved as config/local-production-amo.js
and should look like:
import { apiStageHost, amoStageCDN } from './lib/shared';
module.exports = {
// Statics will be served by node.
staticHost: '',
// FIXME: sign-in isn't working.
// fxaConfig: 'local',
// The node server host and port.
serverHost: '127.0.0.1',
serverPort: 3000,
apiHost: apiStageHost,
amoCDN: amoStageCDN,
CSP: {
directives: {
connectSrc: [apiStageHost],
scriptSrc: ["'self'", 'https://www.google-analytics.com'],
styleSrc: ["'self'"],
imgSrc: [
"'self'",
'data:',
amoStageCDN,
'https://www.google-analytics.com',
],
mediaSrc: ["'self'"],
fontSrc: ["'self'", 'data:', amoStageCDN],
},
},
// This is needed to serve assets locally.
enableNodeStatics: true,
trackingEnabled: false,
// Do not send client side errors to Sentry.
publicSentryDsn: null,
};
After this, re-build and restart using yarn build
and yarn start
as documented above. If you have used localhost
before with a different configuration, be sure to clear your cookies.
NOTE: At this time, it's not possible to sign in using this approach.
When implementing user interfaces you will need to refer to the Sketch mocks that are located in the assets directory. You will need a license to run Sketch and you also need to install some fonts (which are free). Install Fira Sans, Open Sans, Fira Mono and Chivo.
On MacOS, you can use the Homebrew tap Caskroom-fonts:
brew tap homebrew/cask-fonts
brew cask install font-fira-sans font-open-sans font-chivo font-fira-mono
You can check to see what commit of addons-frontend
is deployed by making a request like this:
curl https://addons-dev.allizom.org/__frontend_version__
{
"build" : "https://circleci.com/gh/mozilla/addons-server/6550",
"commit" : "87f49a40ee7a5e87d9b9efde8e91b9019e8b13d1",
"source" : "https://github.com/mozilla/addons-server",
"version" : ""
}
This will return a 415 response if a version.json
file doesn't exist in the root directory. This file is typically generated by the deploy process.
For consistency with monitoring scripts, the same data can be retrieved at this URL:
curl https://addons-dev.allizom.org/__version__
This project will hold distinct front-ends e.g:
- Discovery Pane
- AMO or
addons.mozilla.org
We've made a conscious decision to avoid "premature modularization" and keep this all in one repository. This will help us build out the necessary tooling to support a universal front-end infrastructure without having to worry about cutting packages and bumping versions the entire time.
At a later date if we need to move things out into their own project we still can.
- Based on Redux + React
- Code written in ES2015+
- Universal rendering via node
- Unit tests with high coverage (aiming for 100%)