w3c / accname

Accessible Name and Description Computation
https://w3c.github.io/accname/
61 stars 23 forks source link

Discuss how to break up the computedlabel tests in wpt/accname (to verify AccName computation) #190

Closed cookiecrook closed 1 year ago

cookiecrook commented 1 year ago

Discuss how to break up the automated computedlabel tests in wpt/accname to verify AccName computation related to testdriver implementation.

See also the resource utilities in wpt/wai-aria/scripts/aria-utils.js after WPT PR #38925 lands.

cookiecrook commented 1 year ago

@jaragunde I see you added a lot of label tests to Chromium recently (video). I wanted to make sure you're aware that many of these can be run in WPT now. Please join the discussion with ARIA WG AccName meetings and/or watch the https://github.com/web-platform-tests/interop-2023-accessibility-testing/issues repo.

spectranaut commented 1 year ago

some links:

chromium accname tests: https://source.chromium.org/chromium/chromium/src/+/main:content/test/data/accessibility/accname/

wpt accname tests: https://github.com/web-platform-tests/wpt/tree/master/accname/manual

cookiecrook commented 1 year ago

@spectranaut I don't believe we should worry about the WPT manual accname tests for this effort. Let's focus on the new automation capability.

spectranaut commented 1 year ago

heyy sorry I'm a bit confused, I really don't understand what you are asking.

You want to write tests for accname using the webdriver computedlabel in WPT, correct? And this issue is to discuss a plan for how to do it?

If so, we could start by transforming existing tests in chrome or the existing "manual" tests in WPT, but maybe you mean we should write them from scratch?

cookiecrook commented 1 year ago

@spectranaut wrote:

transforming existing tests in chrome

This makes sense to me and is part of why I copied in @jaragunde.

or the existing "manual" tests in WPT

I think we should review those as a last step to check for edge cases we don't catch in the first round of automated testing.

But in general, I think we could aim for:

Then again, if there aren't any strong opinions, I'll just start adding tests in the format how they make the most sense to me. If so, we can close this issue.

cookiecrook commented 1 year ago

Here's where I'm headed, based on the permalink IDs in the algorithm. Each heading below will be an individual test file. The No-Ops will not be tested.

comp_init (No-Op, No-Test) comp_computation (No-Op, No-Test)

comp_hidden_not_referenced.html

comp_labelledby.html

comp_embedded_control.html

comp_label.html

comp_host_language_label.html

comp_name_from_content.html

comp_text_node.html

comp_recursive_name_from_content.html

comp_tooltip.html

comp_append (No-Op) comp_complete (No-Op)

jaragunde commented 1 year ago

In case it's useful later, I created a list of the accname tests in Chromium and when they were added, so we can differentiate the ones that Joanmarie imported from the work group and the ones that were added later by Chromium devs. It also lists when the expectations were last modified, to know if the test needed changing from the original and why.

Info here: https://docs.google.com/spreadsheets/d/1cXaNgsraAQwVAfYPMuaeayoyAG-5pnztrMjewrK0mpY/edit?usp=sharing