gitlab-org--gitlab-foss/qa
Stan Hu d697c1d836 Merge branch 'update-capybara' into 'master'
Update capybara and selenium-webdriver

See merge request gitlab-org/gitlab-ce!27201
2019-04-12 14:05:47 +00:00
..
bin
load Fix url property format and use conn. pool 2019-03-26 08:25:36 +05:00
qa Update project settings qa selectors 2019-04-11 18:30:23 +01:00
spec Enable RSpec/ExampleWording cop 2019-04-12 10:27:19 +00:00
.gitignore Add load testing script for artillery.io 2019-03-19 05:09:30 +00:00
.rspec
Dockerfile Allow unauthenticated downloads of Chrome 2019-04-12 00:46:17 -07:00
Gemfile Script to generate Perf Testdata 2019-03-05 08:39:04 +00:00
Gemfile.lock Also update nokogiri for QA 2019-04-12 21:03:08 +08:00
qa.rb Set feature flag via command line 2019-03-27 19:03:03 +00:00
Rakefile Fix url property format and use conn. pool 2019-03-26 08:25:36 +05:00
README.md Add 'bundle exec' to the 'bin/qa' calls in qa/README.md 2019-04-09 13:07:41 +02:00
STYLE_GUIDE.md Fix typo on tests style guide 2019-04-02 14:36:21 +02:00

GitLab QA - End-to-end tests for GitLab

This directory contains end-to-end tests for GitLab. It includes the test framework and the tests themselves.

The tests can be found in qa/specs/features (not to be confused with the unit tests for the test framework, which are in spec/).

It is part of the GitLab QA project.

What is it?

GitLab QA is an end-to-end tests suite for GitLab.

These are black-box and entirely click-driven end-to-end tests you can run against any existing instance.

How does it work?

  1. When we release a new version of GitLab, we build a Docker images for it.
  2. Along with GitLab Docker Images we also build and publish GitLab QA images.
  3. GitLab QA project uses these images to execute end-to-end tests.

Validating GitLab views / partials / selectors in merge requests

We recently added a new CI job that is going to be triggered for every push event in CE and EE projects. The job is called qa:selectors and it will verify coupling between page objects implemented as a part of GitLab QA and corresponding views / partials / selectors in CE / EE.

Whenever qa:selectors job fails in your merge request, you are supposed to fix page objects. You should also trigger end-to-end tests using package-and-qa manual action, to test if everything works fine.

How can I use it?

You can use GitLab QA to exercise tests on any live instance! For example, the following call would login to a local GDK instance and run all specs in qa/specs/features:

# Make sure to install the dependencies first with `bundle install`

bundle exec bin/qa Test::Instance::All http://localhost:3000

Note: If you want to run tests requiring SSH against GDK, you will need to modify your GDK setup.

Writing tests

  1. Using page objects
  2. Style guide

Running specific tests

You can also supply specific tests to run as another parameter. For example, to run the repository-related specs, you can execute:

bundle exec bin/qa Test::Instance::All http://localhost -- qa/specs/features/browser_ui/3_create/repository

Since the arguments would be passed to rspec, you could use all rspec options there. For example, passing --backtrace and also line number:

bundle exec bin/qa Test::Instance::All http://localhost -- qa/specs/features/browser_ui/3_create/merge_request/create_merge_request_spec.rb:6 --backtrace

Note that the separator -- is required; all subsequent options will be ignored by the QA framework and passed to rspec.

Overriding the authenticated user

Unless told otherwise, the QA tests will run as the default root user seeded by the GDK.

If you need to authenticate as a different user, you can provide the GITLAB_USERNAME and GITLAB_PASSWORD environment variables:

GITLAB_USERNAME=jsmith GITLAB_PASSWORD=password bundle exec bin/qa Test::Instance::All https://gitlab.example.com

If your user doesn't have permission to default sandbox group gitlab-qa-sandbox, you could also use another sandbox group by giving GITLAB_SANDBOX_NAME:

GITLAB_USERNAME=jsmith GITLAB_PASSWORD=password GITLAB_SANDBOX_NAME=jsmith-qa-sandbox bundle exec bin/qa Test::Instance::All https://gitlab.example.com

All supported environment variables are here.

Sending additional cookies

The environment variable QA_COOKIES can be set to send additional cookies on every request. This is necessary on gitlab.com to direct traffic to the canary fleet. To do this set QA_COOKIES="gitlab_canary=true".

To set multiple cookies, separate them with the ; character, for example: QA_COOKIES="cookie1=value;cookie2=value2"

Building a Docker image to test

Once you have made changes to the CE/EE repositories, you may want to build a Docker image to test locally instead of waiting for the gitlab-ce-qa or gitlab-ee-qa nightly builds. To do that, you can run from this directory:

docker build -t gitlab/gitlab-ce-qa:nightly .

Quarantined tests

Tests can be put in quarantine by assigning :quarantine metadata. This means they will be skipped unless run with --tag quarantine. This can be used for tests that are expected to fail while a fix is in progress (similar to how skip or pending can be used).

bundle exec bin/qa Test::Instance::All http://localhost -- --tag quarantine

If quarantine is used with other tags, tests will only be run if they have at least one of the tags other than quarantine. This is different from how RSpec tags usually work, where all tags are inclusive.

For example, suppose one test has :smoke and :quarantine metadata, and another test has :ldap and :quarantine metadata. If the tests are run with --tag smoke --tag quarantine, only the first test will run. The test with :ldap will not run even though it also has :quarantine.

Running tests with a feature flag enabled

Tests can be run with with a feature flag enabled by using the command-line option --enable-feature FEATURE_FLAG. For example, to enable the feature flag that enforces Gitaly request limits, you would use the command:

bundle exec bin/qa Test::Instance::All http://localhost --enable-feature gitaly_enforce_requests_limits

This will instruct the QA framework to enable the gitaly_enforce_requests_limits feature flag (via the API), run all the tests in the Test::Instance::All scenario, and then disable the feature flag again.

Note: the QA framework doesn't currently allow you to easily toggle a feature flag during a single test, as you can in unit tests, but that capability is planned.

Note also that the -- separator isn't used because --enable-feature is a QA framework option, not an rspec option.