debian-mirror-gitlab/spec/support/capybara.rb
2023-07-09 08:55:56 +05:30

234 lines
8.9 KiB
Ruby

# frozen_string_literal: true
# rubocop:disable Style/GlobalVars
require 'capybara/rails'
require 'capybara/rspec'
require 'capybara-screenshot/rspec'
require 'selenium-webdriver'
# Give CI some extra time
timeout = ENV['CI'] || ENV['CI_SERVER'] ? 45 : 10
# Support running Capybara on a specific port to allow saving commonly used pages
Capybara.server_port = ENV['CAPYBARA_PORT'] if ENV['CAPYBARA_PORT']
# Define an error class for JS console messages
JSConsoleError = Class.new(StandardError)
# Filter out innocuous JS console messages
JS_CONSOLE_FILTER = Regexp.union(
[
'"[HMR] Waiting for update signal from WDS..."',
'"[WDS] Hot Module Replacement enabled."',
'"[WDS] Live Reloading enabled."',
'Download the Vue Devtools extension',
'Download the Apollo DevTools',
"Unrecognized feature: 'interest-cohort'",
'Does this page need fixes or improvements?',
# Needed after https://gitlab.com/gitlab-org/gitlab/-/merge_requests/60933
# which opts out gitlab from FloC by default
# see https://web.dev/floc/ for more info on FloC
"Origin trial controlled feature not enabled: 'interest-cohort'",
# ERR_CONNECTION error could happen due to automated test session disabling browser network request
'net::ERR_CONNECTION'
]
)
CAPYBARA_WINDOW_SIZE = [1366, 768].freeze
SCREENSHOT_FILENAME_LENGTH = ENV['CI'] || ENV['CI_SERVER'] ? 150 : 99
@blackhole_tcp_server = nil
# Run Workhorse on the given host and port, proxying to Puma on a UNIX socket,
# for a closer-to-production experience
Capybara.register_server :puma_via_workhorse do |app, port, host, **options|
file = Tempfile.new
socket_path = file.path
file.close! # We just want the filename
TestEnv.with_workhorse(host, port, socket_path) do
# In cases of multiple installations of chromedriver, prioritize the version installed by SeleniumManager
# selenium-manager doesn't work with Linux arm64 yet:
# https://github.com/SeleniumHQ/selenium/issues/11357
if RUBY_PLATFORM =~ /x86_64-linux|darwin/
chrome_options = Selenium::WebDriver::Chrome::Options.chrome
chromedriver_path = File.dirname(Selenium::WebDriver::SeleniumManager.driver_path(chrome_options))
ENV['PATH'] = "#{chromedriver_path}:#{ENV['PATH']}" # rubocop:disable RSpec/EnvAssignment
end
Capybara.servers[:puma].call(app, nil, socket_path, **options)
end
end
Capybara.register_driver :chrome do |app|
options = Selenium::WebDriver::Chrome::Options.new
# Force the browser's scale factor to prevent inconsistencies on high-res devices
options.add_argument('--force-device-scale-factor=1')
options.add_argument("window-size=#{CAPYBARA_WINDOW_SIZE.join(',')}")
# Chrome won't work properly in a Docker container in sandbox mode
options.add_argument("no-sandbox")
# Run headless by default unless WEBDRIVER_HEADLESS specified
options.add_argument("headless") unless ENV['WEBDRIVER_HEADLESS'] =~ /^(false|no|0)$/i || ENV['CHROME_HEADLESS'] =~ /^(false|no|0)$/i
# Disable /dev/shm use in CI. See https://gitlab.com/gitlab-org/gitlab/issues/4252
options.add_argument("disable-dev-shm-usage") if ENV['CI'] || ENV['CI_SERVER']
# Explicitly set user-data-dir to prevent crashes. See https://gitlab.com/gitlab-org/gitlab-foss/issues/58882#note_179811508
options.add_argument("user-data-dir=/tmp/chrome") if ENV['CI'] || ENV['CI_SERVER']
# Set chrome default download path
if ENV['DEFAULT_CHROME_DOWNLOAD_PATH']
options.add_preference("download.default_directory", ENV['DEFAULT_CHROME_DOWNLOAD_PATH'])
options.add_preference("download.prompt_for_download", false)
end
# Set up a proxy server to block all external traffic.
@blackhole_tcp_server = TCPServer.new(0)
Thread.new do
loop do
Thread.start(@blackhole_tcp_server.accept, &:close)
end
end
options.add_argument("--proxy-server=http://127.0.0.1:#{@blackhole_tcp_server.addr[1]}")
options.add_argument("--proxy-bypass-list=127.0.0.1,localhost,#{Gitlab.config.gitlab.host}")
Capybara::Selenium::Driver.new(
app,
browser: :chrome,
options: options
)
end
Capybara.register_driver :firefox do |app|
options = Selenium::WebDriver::Firefox::Options.new(log_level: :trace)
options.add_argument("--window-size=#{CAPYBARA_WINDOW_SIZE.join(',')}")
# Run headless by default unless WEBDRIVER_HEADLESS specified
options.add_argument("--headless") unless ENV['WEBDRIVER_HEADLESS'] =~ /^(false|no|0)$/i
Capybara::Selenium::Driver.new(
app,
browser: :firefox,
options: options
)
end
Capybara.server = :puma_via_workhorse
Capybara.javascript_driver = ENV.fetch('WEBDRIVER', :chrome).to_sym
Capybara.default_max_wait_time = timeout
Capybara.ignore_hidden_elements = true
Capybara.default_normalize_ws = true
Capybara.enable_aria_label = true
Capybara::Screenshot.append_timestamp = false
Capybara::Screenshot.register_filename_prefix_formatter(:rspec) do |example|
example.full_description.downcase.parameterize(separator: "_")[0..SCREENSHOT_FILENAME_LENGTH]
end
# Keep only the screenshots generated from the last failing test suite
Capybara::Screenshot.prune_strategy = :keep_last_run
# From https://github.com/mattheworiordan/capybara-screenshot/issues/84#issuecomment-41219326
Capybara::Screenshot.register_driver(:chrome) do |driver, path|
driver.browser.save_screenshot(path)
end
RSpec.configure do |config|
config.include CapybaraHelpers, type: :feature
config.before(:context, :js) do
# This prevents Selenium from creating thousands of connections while waiting for
# an element to appear
webmock_enable_with_http_connect_on_start!
next if $capybara_server_already_started
TestEnv.eager_load_driver_server
$capybara_server_already_started = true
end
config.after(:context, :js) do
# WebMock doesn't stub connections, so we need to restore the original behavior
# to prevent many specs from failing:
# https://github.com/bblimke/webmock/blob/master/README.md#connecting-on-nethttpstart
webmock_enable!
end
config.before(:example, :js) do
session = Capybara.current_session
allow(Gitlab::Application.routes).to receive(:default_url_options).and_return(
host: session.server.host,
port: session.server.port,
protocol: 'http')
# CSRF protection is disabled by default. We only enable this for JS specs because some forms
# require Javascript to set the CSRF token.
allow_any_instance_of(ActionController::Base).to receive(:protect_against_forgery?).and_return(true)
# reset window size between tests
unless session.current_window.size == CAPYBARA_WINDOW_SIZE
begin
session.current_window.resize_to(*CAPYBARA_WINDOW_SIZE)
rescue StandardError # ?
end
end
end
# The :capybara_ignore_server_errors metadata means unhandled exceptions raised
# by the application under test will not necessarily fail the server. This is
# useful when testing conditions that are expected to raise a 500 error in
# production; it should not be used on the happy path.
config.around(:each, :capybara_ignore_server_errors) do |example|
Capybara.raise_server_errors = false
example.run
ensure
Capybara.raise_server_errors = true
end
config.append_after do |example|
if example.metadata[:screenshot]
screenshot = example.metadata[:screenshot][:image] || example.metadata[:screenshot][:html]
screenshot&.delete_prefix!(ENV.fetch('CI_PROJECT_DIR', ''))
example.metadata[:stdout] = %{[[ATTACHMENT|#{screenshot}]]}
end
end
config.after(:example, :js) do |example|
# when a test fails, display any messages in the browser's console
# but fail don't add the message if the failure is a pending test that got
# fixed. If we raised the `JSException` the fixed test would be marked as
# failed again.
if example.exception && !example.exception.is_a?(RSpec::Core::Pending::PendingExampleFixedError)
console = page.driver.browser.logs.get(:browser)&.reject { |log| log.message =~ JS_CONSOLE_FILTER }
if console.present?
message = "Unexpected browser console output:\n" + console.map(&:message).join("\n")
raise JSConsoleError, message
end
end
# prevent localStorage from introducing side effects based on test order
unless ['', 'about:blank', 'data:,'].include? Capybara.current_session.driver.browser.current_url
execute_script("localStorage.clear();")
end
# capybara/rspec already calls Capybara.reset_sessions! in an `after` hook,
# but `block_and_wait_for_requests_complete` is called before it so by
# calling it explicitly here, we prevent any new requests from being fired
# See https://github.com/teamcapybara/capybara/blob/ffb41cfad620de1961bb49b1562a9fa9b28c0903/lib/capybara/rspec.rb#L20-L25
# We don't reset the session when the example failed, because we need capybara-screenshot to have access to it.
Capybara.reset_sessions! unless example.exception
block_and_wait_for_requests_complete
end
end