Capybara not finding signout link by id - ruby-on-rails

Question

For a rails app I'm currently working on I recently changed the design somewhat so that the signout link no longer has the anchor text "Sign out" but instead a glyphicon from twitter bootstrap. the html for link now looks like this:
<a class="btn btn-primary btn-sm" data-method="delete" href="/users/sign_out" rel="nofollow">
<span class="glyphicon glyphicon-log-out"></span>
</a>
meanwhile my capybara test looks like this:
context "when not logged in" do
it 'cannot create wikis' do
#free_user = create(:user)
login_as(#free_user, :scope => :user)
click_link "Sign out"
visit root_path
expect(page).to_not have_link('Create wiki')
end
end
Now that the text "Sign out" is no longer there, I need a new way to identify the link. Checking the documentation for capybara (or rather this handy cheatsheet), it looks like I can supply either the text of the link or its id. So I tried giving it an id:
<a class="btn btn-primary btn-sm" data-method="delete" href="/users/sign_out" id="signout" rel="nofollow">
<span class="glyphicon glyphicon-log-out"></span>
</a>
So now it's got an id of "signout" however when I make this change to the test, it still won't pass.
1) Standard (free) User when not logged in cannot create wikis
Failure/Error: click_link "signout"
Capybara::ElementNotFound:
Unable to find link "signout"
# ./spec/features/standard_user_role_spec.rb:107:in `block (3 levels) in <top (required)>'
I tried making sure that I was still a logged in user in the test by creating and logging in the user as seen above and by adding a check that the html on the page contains Hello, since it says "Hello" and the name of the user when the user is logged in:
expect(page).to have_content('Hello')
This gave me another error that I don't understand:
1) Standard (free) User when not logged in cannot create wikis
Failure/Error: expect(page).to have_content('Hello')
Capybara::ElementNotFound:
Unable to find xpath "/html"
# ./spec/features/standard_user_role_spec.rb:107:in `block (3 levels) in <top (required)>'
So what could be going on here?
Full spec available here

Answer 1

You can re-add the text but hide it from all but screen readers:
<a class="btn btn-primary btn-sm" data-method="delete" href="/users/sign_out" id="signout" rel="nofollow">
<span class="sr-only">Sign out</span>
<span class="glyphicon glyphicon-log-out"></span>
</a>
This also improves accessibility somewhat.
context "when not logged in" do
it 'cannot create wikis' do
#free_user = create(:user)
login_as(#free_user, :scope => :user)
visit root_path
click_link "Sign out", visible: false
expect(page).to_not have_link('Create wiki')
end
end
Note that we explicitly tell Capybara to look for hidden text with the visible option.

Answer 2

For the applications, which use JS, that means the request is being sent using for example json format, and is begin triggered by click on the link. (suppose the HTML markup is the same). It is useful the following call:
find('Sign Out', visible: false).trigger('click')
or by id (if you'll add it):
find(:css, "#sign-out", visible: false).trigger('click')