Jmax/lg 11294 add aria tags to cancel idv verify screen

[jmax-gsa]

🎫 Ticket

LG-11294

🛠 Summary of changes

• Added aria-label attributes to the Rails-generated (via the button_to helper) <form> tags for the buttons on the 'cancel IdV' screen (the buttons are 'No, keep going', 'Start over', and either 'Go to account page' or 'Exit <>'
• Added view spec support for testing button_to generated forms for aria attributes
• Added feature spec support for testing aria labeling on form landmarks

📜 Testing Plan

The review app for testing is here

• Create a user and enter IdV
• Select 'Cancel' from any of the IdV screens
• You are now on the 'Cancel IdV' screen. Use the browser's View Source to inspect the HTML, and verify that each of the buttons has the corret 'aria-label' attribute on its enclosing form.
• Exit IdV, log out, and then start IdV from the Sinatra sample SP app.
• Examine the HTML for the 'Cancel IdV' screen again, and verify that the aria-label tags are still present.

👀 Screenshots

Start over button

Keep going button

Account page button

[aduth]

I think it makes sense to have a unique label for each form and it's probably okay that those are the same as the button text, but I'm not sure I follow that the changes to how we compute accessible names are necessary. For a regression spec, I might imagine we could check for this using a new matcher that checks for unique names of landmarks if multiple of the same type exist in the page (form in this case). Something like expect(page).to have_unique_landmark_labels.

[aduth]

Couple comments, but LGTM otherwise 👍

[aduth]

I wonder if we could add this to the existing expect_page_to_have_no_accessibility_violations common helper? Maybe that'd start flagging issues with unrelated screens.

Not sure if it covers all the variations, but we have a spec which calls that helper for the IdV cancel screen:

identity-idp/spec/features/accessibility/idv_pages_spec.rb

Lines 16 to 23 in 2b7c12d

	scenario 'cancel idv' do
	sign_in_and_2fa_user
	visit idv_cancel_path
	expect(current_path).to eq idv_cancel_path
	expect_page_to_have_no_accessibility_violations(page)
	end

[aduth]

Do we need the first part of this condition? I think if we were to adapt this for other screens, we're more concerned with testing for unique labels if a page has more than one form, so if the page has only one form and it has a nil label, that seems fine?

[jmax-gsa]

Mmmm. Maybe (labels.length == 1 || labels.none?(&:nil)) ?

[aduth]

Yeah, though that doesn't check that the labels are unique across multiple forms. Originally I was thinking it'd be enough to keep the rest of the current logic (labels.uniq.length == labels.length) but that wouldn't flag if there was multiple forms and only one or some of them were labeled.

Maybe this?

labels.one? || labels.compact.uniq.length == labels.length