Front-end testing

Bedrock runs a suite of front-end Jasmine behavioral/unit tests, which use Jasmine Browser Runner as a test runner. We also have a suite of functional tests using Selenium and pytest. This allows us to emulate users interacting with a real browser. All these test suites live in the tests directory. To run the tests locally, you must also first download geckodriver and chromedriver and make it available in your system path. You can alternatively specify the path to geckodriver and chromedriver using the command line (see the pytest-selenium documentation for more information).

The tests directory comprises of:

  • /functional contains pytest tests.

  • /pages contains Python page objects.

  • /unit contains the Jasmine tests and Jasmine Browser Runner config file.

Installation

First follow the installation instructions for bedrock, which will install the dependencies required to run the various front-end test suites.

To download geckodriver and chromedriver and have it ready to run in your system, there are a couple of ways:

  • Download geckdriver and add it to your system path:

    cd /path/to/your/downloaded/files/
    mv geckodriver /usr/local/bin/
    
  • If you’re on MacOS, download geckodriver directly using Homebrew, which automatically places it in your system path:

    brew install geckodriver
    
  • Download chromedriver and add it to your system path:

    cd /path/to/your/downloaded/files/
    mv chromedriver /usr/local/bin/
    
  • If you’re on MacOS, download chromedriver directly using Homebrew/Cask, which automatically places it in your system path:

    brew tap homebrew/cask
    
    brew cask install chromedriver
    

Running Jasmine tests using Jasmine Browser Runner

To perform a single run of the Jasmine test suite using Firefox and Chrome, first make sure you have both browsers installed locally, and then activate your bedrock virtual env.

$ pyenv activate bedrock

You can then run the tests with the following command:

$ npm run test

This will run all our front-end linters and formatting checks before running the Jasmine test suite. If you only want to run the tests themselves, you can run:

$ npm run test

See the Jasmine documentation for tips on how to write JS behavioral or unit tests. We also use Sinon for creating test spies, stubs and mocks.

Running functional tests

Note

Before running the functional tests, please make sure to follow the bedrock installation docs, including the database sync that is needed to pull in external data such as event/blog feeds etc. These are required for some of the tests to pass.

To run the full functional test suite against your local bedrock instance in Mozorg mode:

$ py.test --base-url http://localhost:8000 --driver Firefox --html tests/functional/results.html tests/functional/

This will run all test suites found in the tests/functional directory and assumes you have bedrock running at localhost on port 8000. Results will be reported in tests/functional/results.html.

To run the full functional test suite against your local bedrock instance in Pocket mode, things are slightly different, because of the way things are set up in order to allow CI to test both Mozorg Mode and Pocket Mode at the same time. You need to define a temporary environment variable (needed by the pocket_base_url fixture) and scope pytest to only run Pocket tests:

$ BASE_POCKET_URL=http://localhost:8000 py.test -m pocket_mode --driver Firefox --html tests/functional/results.html tests/functional/

This will run all test suites found in the tests/functional directory that have the pytest “mark” of pocket_mode and assumes you have bedrock running in Pocket mode at localhost on port 8000. Results will be reported in tests/functional/results.html.

Note

If you omit the --base-url command line option in Mozorg mode (ie, not in Pocket mode) then a local instance of bedrock will be started, however the tests are not currently able to run against bedrock in this way.

By default, tests will run one at a time. This is the safest way to ensure predictable results, due to bug 1230105. If you want to run tests in parallel (this should be safe when running against a deployed instance), you can add -n auto to the command line. Replace auto with an integer if you want to set the maximum number of concurrent processes.

Note

There are some functional tests that do not require a browser. These can take a long time to run, especially if they’re not running in parallel. To skip these tests, add -m 'not headless' to your command line.

To run a single test file you must tell py.test to execute a specific file e.g. tests/functional/test_newsletter.py:

$ py.test --base-url http://localhost:8000 --driver Firefox --html tests/functional/results.html tests/functional/firefox/new/test_download.py

To run a single test you can filter using the -k argument supplied with a keyword e.g. -k test_download_button_displayed:

$ py.test --base-url http://localhost:8000 --driver Firefox --html tests/functional/results.html tests/functional/firefox/new/test_download.py -k test_download_button_displayed

You can also easily run the tests against any bedrock environment by specifying the --base-url argument. For example, to run all functional tests against dev:

$ py.test --base-url https://www-dev.allizom.org --driver Firefox --html tests/functional/results.html tests/functional/

Note

For the above commands to work, Firefox needs to be installed in a predictable location for your operating system. For details on how to specify the location of Firefox, or running the tests against alternative browsers, refer to the pytest-selenium documentation.

For more information on command line options, see the pytest documentation.

Running tests in Sauce Labs

You can also run tests in Sauce Labs directly from the command line. This can be useful if you want to run tests against Internet Explorer when you’re on Mac OSX, for instance.

  1. Sign up for an account at https://saucelabs.com/opensauce/.

  2. Log in and obtain your Remote Access Key from user settings.

  3. Run a test specifying SauceLabs as your driver, and pass your credentials.

For example, to run the home page tests using Internet Explorer via Sauce Labs:

$ SAUCELABS_USERNAME=thedude SAUCELABS_API_KEY=123456789 SAUCELABS_W3C=true SELENIUM_EXCLUDE_DEBUG=logs py.test --base-url https://www-dev.allizom.org --driver SauceLabs --capability browserName 'internet explorer' --capability platformName 'Windows 10' --html tests/functional/results.html tests/functional/test_home.py

Writing Selenium tests

Tests usually consist of interactions and assertions. Selenium provides an API for opening pages, locating elements, interacting with elements, and obtaining state of pages and elements. To improve readability and maintainability of the tests, we use the Page Object model, which means each page we test has an object that represents the actions and states that are needed for testing.

Well written page objects should allow your test to contain simple interactions and assertions as shown in the following example:

def test_sign_up_for_newsletter(base_url, selenium):
    page = NewsletterPage(base_url, selenium).open()
    page.type_email('noreply@mozilla.com')
    page.accept_privacy_policy()
    page.click_sign_me_up()
    assert page.sign_up_successful

It’s important to keep assertions in your tests and not your page objects, and to limit the amount of logic in your page objects. This will ensure your tests all start with a known state, and any deviations from this expected state will be highlighted as potential regressions. Ideally, when tests break due to a change in bedrock, only the page objects will need updating. This can often be due to an element needing to be located in a different way.

Please take some time to read over the Selenium documentation for details on the Python client API.

Destructive tests

By default all tests are assumed to be destructive, which means they will be skipped if they’re run against a sensitive environment. This prevents accidentally running tests that create, modify, or delete data on the application under test. If your test is nondestructive you will need to apply the nondestructive marker to it. A simple example is shown below, however you can also read the pytest markers documentation for more options.

import pytest

@pytest.mark.nondestructive
def test_newsletter_default_values(base_url, selenium):
    page = NewsletterPage(base_url, selenium).open()
    assert '' == page.email
    assert 'United States' == page.country
    assert 'English' == page.language
    assert page.html_format_selected
    assert not page.text_format_selected
    assert not page.privacy_policy_accepted

Smoke tests

Smoke tests are considered to be our most critical tests that must pass in a wide range of web browsers, including Internet Explorer 11. The number of smoke tests we run should be enough to cover our most critical pages where legacy browser support is important.

import pytest

@pytest.mark.smoke
@pytest.mark.nondestructive
def test_download_button_displayed(base_url, selenium):
    page = DownloadPage(selenium, base_url, params='').open()
    assert page.is_download_button_displayed

You can run smoke tests only by adding -m smoke when running the test suite on the command line.

Waits and Expected Conditions

Often an interaction with a page will cause a visible response. While Selenium does its best to wait for any page loads to be complete, it’s never going to be as good as you at knowing when to allow the test to continue. For this reason, you will need to write explicit waits in your page objects. These repeatedly execute code (a condition) until the condition returns true. The following example is probably the most commonly used, and will wait until an element is considered displayed:

from selenium.webdriver.support import expected_conditions as expected
from selenium.webdriver.support.ui import WebDriverWait as Wait

Wait(selenium, timeout=10).until(
    expected.visibility_of_element_located(By.ID, 'my_element'))

For convenience, the Selenium project offers some basic expected conditions, which can be used for the most common cases.

Debugging Selenium

Debug information is collected on failure and added to the HTML report referenced by the --html argument. You can enable debug information for all tests by setting the SELENIUM_CAPTURE_DEBUG environment variable to always.

Guidelines for writing functional tests

  • Try and keep tests organized and cleanly separated. Each page should have its own page object and test file, and each test should be responsible for a specific purpose, or component of a page.

  • Avoid using sleeps - always use waits as mentioned above.

  • Don’t make tests overly specific. If a test keeps failing because of generic changes to a page such as an image filename or href being updated, then the test is probably too specific.

  • Avoid string checking as tests may break if strings are updated, or could change depending on the page locale.

  • When writing tests, try and run them against a staging or demo environment in addition to local testing. It’s also worth running tests a few times to identify any intermittent failures that may need additional waits.

See also the Web QA style guide for Python based testing.

Testing Basket email forms

When writing functional tests for front-end email newsletter forms that submit to Basket, we have some special case email addresses that can be used just for testing:

  1. Any newsletter subscription request using the email address “success@example.com” will always return success from the basket client.

  2. Any newsletter subscription request using the email address “failure@example.com” will always raise an exception from the basket client.

Using the above email addresses enables newsletter form testing without actually hitting the Basket instance, which reduces automated newsletter spam and improves test reliability due to any potential network flakiness.

Headless tests

There are targeted headless tests for the download pages. These tests and are run as part of the pipeline to ensure that download links constructed via product details are well formed and return valid 200 responses.