crawler skip links nested within a shadow tree

[yinonov]

Describe the bug

running the action on a website with links that are nested within shadow tree just skips visiting those links and only visits links nested within the light tree

CodePen repro example

it's not a repro but the actual site (preview channel) where this occurs and shadow links are skipped
https://vivid-j0vxkmsqmhn0.deno.dev/

Expected behavior

Shadowed links should be visited and checked

Screenshots

this gets visited

this doesn't

both are equivalent according to the accessibility tree view (devtools) and share the same computed properties

Are you willing to submit a PR?

Of course, with a little guidance

Did you search for similar existing issues?

yes, none to be found

Additional context

this bug might no be related to the GH action but to the underlying accessibility insights service.
forward to the relevant repo if need

[sfoslund]

Hi @yinonov, thanks for filling an issue! I believe this is not a bug and is in fact by design. Please check out our docs here, which discuss why Accessibility insights does not support elements that are contained within a closed shadow DOM: https://accessibilityinsights.io/docs/web/reference/faq/#are-automated-checks-compatible-with-shadow-dom

[ghost]

The team requires additional author feedback; please review their replies and update this issue accordingly. Thank you for contributing to Accessibility Insights!

[yinonov]

Hi @yinonov, thanks for filling an issue! I believe this is not a bug and is in fact by design. Please check out our docs here, which discuss why Accessibility insights does not support elements that are contained within a closed shadow DOM: https://accessibilityinsights.io/docs/web/reference/faq/#are-automated-checks-compatible-with-shadow-dom

Thanks @sfoslund, this section u shared refers to a closed mode of shadowRoot, which isn't the case here. Mode is open.

[ghost]

This issue has been marked as ready for team triage; we will triage it in our weekly review and update the issue. Thank you for contributing to Accessibility Insights!

[ghost]

This issue requires additional investigation by the Accessibility Insights team. When the issue is ready to be triaged again, we will update the issue with the investigation result and add "status: ready for triage". Thank you for contributing to Accessibility Insights!

[RobGallo]

Investigating with engineers

[yinonov]

any clue why this happens? does it happen in axe itself?

[RobGallo]

@yinonov

You are referring to which links are crawled, not which links are scanned, correct?

Not crawling open shadow dom links is a current limitation of accessibility-insights-service).

The most realistic path to picking up links in open shadow DOMs would be switching the service from puppeteer to playwright (which would pierce open shadow DOMs by default with the sort of query apify uses to scrape for links),

Piercing closed shadow DOMs is a lot harder implementation-wise – that’s functionality that neither Puppeteer nor Playwright facilitate.

This does make sense for us to address long-term. I have filed an issue in the service repo to track the request: #2170.

I will close this issue. But the community should feel free to leave comments in the service tree repo to help us prioritize this change.

[yinonov]

Thanks for the detailed explanation.

IN the meantime, I s there a workaround where I could specify a list of links for the action to visit?

[brocktaylor7]

Hello @yinonov,

Sorry for the delayed response. There is a mechanism for specifying a list of links for the action to visit.

You can use the input-urls input in your YAML configuration file to specify the links. The expected value of the input is a string of URLs each separated by a space.

An example of the step with the URLs provided might look something like this:

- name: Scan for accessibility issues
  uses: microsoft/accessibility-insights-action@v3
  with:
      url: https://your-website-url
      input-urls: https://your-first-link https://your-second-link https://your-third-link

This will put the four listed URLs in the queue to be scanned first, then any crawled URL will be added to the end of the queue and will be scanned after those first four.

Hopefully that solves your issue for now. Let us know if you have any further questions!

[yinonov]

thanks @brocktaylor7 that's exactly what I did