Why does adding “sleep 1” in an after hook cause this Rspec/Capybara test to pass?

问题

I'm using rails 4.0.5, rspec 2.14.1, capybara 2.2.1, capybara-webkit 1.1.0 and database_cleaner 1.2.0. I'm seeing some weird behavior with the following feature test (which simulates a user viewing a comment on a post, hovering over an icon to make a menu appear, and clicking a menu item to delete the comment):

let(:user){create(:user)}
let(:post){create(:post, author: user)}
let!(:comment){create(:comment, post: post, author: user)}

...

it "can delete a comment" do
  assert(page.has_css? "#comment-#{comment.id}")
  find("#comment-#{comment.id}-controls").trigger(:mouseover)
  find("#comment-#{comment.id} .comment-delete a").click
  assert(page.has_no_css? "#comment-#{comment.id}")
end

This test fails about 80% of the time, always due to some record being retrieved from the database as nil-- I get NoMethodError: undefined method X for nil:NilClass, for various values of X. Sometimes the nil is the comment that's being deleted, sometimes it's the post that the comment's attached to, sometimes it's the author of the comment/post.

If I add sleep 1 to the end of the test, it passes:

it "can delete its own comment" do
  assert(page.has_css? "#comment-#{comment.id}")
  find("#comment-#{comment.id}-controls").trigger(:mouseover)
  find("#comment-#{comment.id} .comment-delete a").click
  assert(page.has_no_css? "#comment-#{comment.id}")
  sleep 1
end

It also passes if I put sleep 1 in an after block.

Any idea why I get these NoMethodErrors, and/or why the test passes if I make it sleep for a second after all the work is done?

回答1:

I suspect that it's possible in your application for the comment to disappear from the page (which is the last thing you're asserting) before it's deleted from the database. That means that the test can clean up before the deletion happens. If this is the case you can fix it by waiting for the actual deletion to happen at the end of the test. I have this method around (a reimplementation of a method that was removed from Capybara 2 but is still sometimes necessary)

def wait_until(delay = 1)
  seconds_waited = 0
  while ! yield && seconds_waited < Capybara.default_wait_time
    sleep delay
    seconds_waited += 1
  end
  raise "Waited for #{Capybara.default_wait_time} seconds but condition did not become true" unless yield
end

so I can do

wait_until { Comment.count == 0 }

in tests.

Another approach is to add Rack middleware that blocks requests made after the test ends. This approach is described in detail here: http://www.salsify.com/blog/tearing-capybara-ajax-tests I saw it do a very good job of addressing data leakage in a good-sized suite of RSpec feature specs.

回答2:

https://github.com/jnicklas/capybara#asynchronous-javascript-ajax-and-friends

When working with asynchronous JavaScript, you might come across situations where you are attempting to interact with an element which is not yet present on the page. Capybara automatically deals with this by waiting for elements to appear on the page.

来源：https://stackoverflow.com/questions/23554882/why-does-adding-sleep-1-in-an-after-hook-cause-this-rspec-capybara-test-to-pas

标签

ruby-on-rails

ruby

rspec

capybara

capybara-webkit