Why your Cucumber feature loses cookies when run under Selenium

When your Cucumber feature seems to forget cookies / sessions when you run it with Selenium check if the test travels in time like here:

Given the date is 2017-10-20
When I sign in
Then I should see "Welcome!"

What happens here is that the Rails application serving pages runs in 2017, but the process running your browser still lives today. This huge gap in time will expire most cookies immediately.

If all you need is to freeze the time to a date, a workaround is to travel to the future instead.

Capybara can match elements outside of <body>

Capybara will match elements outside of a page's <body> tag.

For example, the step definitions below match <link> nodes in a page's <head>:

Then /^my browser should auto-discover the "([^"]*)" feed$/ do |slug|
  page.should have_css(
    'head link' +
    '[rel="alternate"]' +
    "[href='http://www.example.com/#{slug}/feed.rss']" +
    '[title="RSS feed (all cards)"]' +
    '[type="application/rss+xml"]',
    visible: false
  )
end

Then /^my browser should not auto-discover any RSS fe...

Mailcatcher: An alternative to inaction_mailer

Looks simpler than inaction_mailer:

gem install mailcatcher
mailcatcher

Setup Rails to send mails to 127.0.0.1:1025. Usually you want the following config in config/environments/development.rb and maybe in test.rb or cucumber.rb.

config.action_mailer.delivery_method = :smtp
config.action_mailer.smtp_settings = {
  :address => 'localhost',
  :port => 1025
}

Now you can see sent mails in your browser when opening http://127.0.0.1:1080

Note: In order to s...

Single step and slow motion for cucumber scenarios using @javascript selenium

Single step and slow motion for Cucumber scenarios can come in handy, especially in @javascript scenarios.

# features/support/examiners.rb
AfterStep('@slow_motion') do
  sleep 2
end

AfterStep('@single_step') do
  print "Single Stepping. Hit enter to continue"
  STDIN.getc
end

If you're using spreewald, these tags are available as @slow-motion and @single-step (with dashes instead of underscores).

Note: You can also [prevent the selenium webbrowser wind...

Test that a form field has an error with Cucumber and Capybara

You can use the step definition below to say this:

Then the "Last name" field should have an error

Capybara

Then /^the "([^\"]*)" field should( not)? have an error$/ do |field, negate|
  expectation = negate ? :should_not : :should
  page.send(expectation, have_css('.field_with_errors', :text => field))
end

Prevent the selenium webbrowser window from closing after a failed @javascript step

When cucumber encounters a failing step in a @javascript feature, the selenium browser window instantly closes. Sometimes you do not want that, because you need to see what is going on. You can click through the data your feature created, when you add the following file to your features/support directory:

#features/support/examiners.rb

After('@leave_the_window_open') do |scenario|
  if scenario.respond_to?(:status) && scenario.status == :failed
    print "Step Failed. Press return to close browser"
    STDIN.getc
  ...

How Non-negotiable Features Kill Software Products

features are pre-sold without any option to negotiate what’s important and what may be left out, you inevitably end up with too much complexity. Such pre-sold features not only tie your hands, but the client is also not able to change what he needs over time.

Selenium WebDriver 2.5.0, 2.6.0 fails when selecting options from select boxes

We are consistently having trouble with selenium-webdriver > 2.5.0 where whenever we try to select an option from a <select> Capybara complains:

No such option 'Foo' in this select box. Available options: 'Foo', 'Bar', 'Baz' (Capybara::OptionNotFound)

This seems to happen with both old and new versions of Firefox. Our workaround so far is to freeze the gem at version 0.2.2.

How to install a frozen version of Firefox for your Selenium tests

Whenever Firefox updates, all your Cucumber features that use Selenium break. This is annoying.

In order to remedy this, version 0.5.0 of our geordi gem comes with a script that helps you create an unchanging version of Firefox for your Selenium tests. In particular, this new copy of Firefox will have the following properties:

  • It won't update itself with a newer version
  • It can co-exist with your regular Firefox installation (which you can update at will)
  • It will use a profile separate from the one...

Capistrano: Different usernames for each server

If you have different users for different servers, don't use set :user. Encode the username into the server definition instead:

server "username@servername.tld", :app, :web, :cron, :db, :primary => true

Turn off SSL for scenarios with Selenium

Selenium does not speak SSL because it uses WEBrick that doesn't. When you use Selenium for Cucumber scenarios that visit pages with SSL, they will fail.

To turn off SSL only for scenarios that are executed on WEBrick, put this method into your application controller.

def ensure_proper_protocol
  request.headers['SERVER_SOFTWARE'].andand.include?('WEBrick') || super
end

Select an unknown option with Capybara

When you don't know which options are available, but need to have an option selected, use this step.

When /^I select the second option from "([^"]*)"$/ do |id|
  second_option = find(:xpath, "//*[@id='#{id}']/option[2]").text
  select(second_option, :from => id)
end

How to accept or deny JavaScript confirmation dialogs in Capybara/Selenium

These methods are available to you:

page.driver.browser.switch_to.alert.accept
page.driver.browser.switch_to.alert.dismiss
page.driver.browser.switch_to.alert.text # the confirmation text

Spreewald gives you steps like these:

When I confirm the browser dialog
When I cancel the browser dialog

Note that recent versions of Selenium will automatically dismiss confirmation dialogs. [Here is how to fix that](https://makandracards.com/makandra/617366-configure-selenium-webdriv...

ActiveRecord 3+ auto-converts times to UTC by default. Hilarity ensues.

Remember how Rails 2 came with an awesome feature that broke all code using Time.now or Time.parse?

This behavior is now the default for Rails 3. Disable it by adding the following to your config/application.rb:

config.active_record.default_timezone = :local
config.active_record.time_zone_aware_attributes = false    

Let a Rails 3 application make a request to itself

Ever wondered how Rails talks to itself in a Cucumber feature? In Rails 3 you can do it like this:

def rack_env(path)
  { "rack.input" => {},
    "PATH_INFO"=>"#{path}",
    "REQUEST_METHOD"=>"GET" }
end

request = rack_env('/users/new')
response = Rails.application.call(request)
status, headers, body = response

puts status # e.g. 200
puts headers.inspect # hash of headers
puts body.body # html of response body

Instead of Rails.application you can also call any Rack application.

Windows 7: Open terminal from Explorer

Let's say you have an Explorer window showing a directory and want a cmd terminal to be opened there.\
While you needed a PowerToy on Windows XP, this is a build-in feature for Windows 7.

Simply press and hold the Shift key and right-click a folder icon or into empty space of an Explorer window.\
Then, choose "Open Command Window Here" from the context menu.

Let Webrat make a POST request

Just add the parameter :post to the visit method:

visit publish_entry_path, :post

Webrat doesn't follow redirect because it considers the url external

Rails doesn't know which host it is running on. For generating links, it strips the hostname off the request URL, which can lead to errors when you have absolute URLs in your Cucumber tests.

If you really need to use absolute URLs somewhere, say in an email you send, either throw away the host when parsing it (e.g. body.scan(/http:\/\/[^\/]+\/([^\s"<]+)/)) or tell Webrat you're back on your site.

Downgrade Firefox 6 to Firefox 5 on Ubuntu

Note that if you plan to downgrade Firefox because your Selenium tests broke after a Firefox upgrade, there is a better way that doesn't involve downgrading. Mozilla has stated that they will no longer provide security patches for any but the most recent versions of Firefox. So running an old Firefox should not be a long-term solution for anything.

If you still want to downgrade your Firefox for other reasons, here is how I downgra...

Prevent your Firefox from auto-updating

Note that if you plan to freeze your Firefox versions because your Selenium tests break whenever Firefox updates, there is a better way that lets you keep an up-to-date Firefox. Mozilla has stated that they will no longer provide security patches for any but the most recent versions of Firefox. So running an old Firefox should not be a long-term solution for anything.

If you still wish to disable the auto-update in Firefox, a poste...

Best GitHub feature

When browsing a repository, pressing "t" allows you to quickly search for file names. Very awesome!

Go here to try it out.

Detect the current Rails environment from JavaScript or CSS

Detecting if a Javascript is running under Selenium WebDriver is super-painful. It's much easier to detect the current Rails environment instead.

You might be better of checking against the name of the current Rails environment. To do this, store the environment name in a data-environment of your <html>. E.g., in your application layout:

<html data-environment=<%= Rails.env %>>

Now you can say in a pi...

How to stub class constants in RSpec

Hint: There's another card with this helper for Cucumber features.


Sometimes you feel like you need to stub some CONSTANT you have defined in an other class. Since actually constants are called constants because they're constant, there's no way to easily stub a constant.

Here are three solutions for you.

Easiest solution

Rethink! Do you really need CONSTANT = %w[foo bar] to be constant? In many cases, setting it as a...

Capybara - The missing API

The Capybara API is somewhat hard for parse for a list of methods you can call on a Capybara node. Below you can find such a list. It's all copied from the Capybara docs, so all credit goes to the Capybara committers.

When you talk to Capybara from a Cucumber step definition, you always have page as the document root node, or whatever you scoped to by saying within(selector) { ... }. You can select child notes by calling page.find(selector) or page.all(selector). You can call the same ...