fix: corrects color-contrast tests with overlapping elements

[downzer0]

I believe this corrects the color-contrast check that tests for overlapping and obscuring elements. The previous code was checking for elements only at a specific set of coordinates, not within the bounding box. So, for example, if there was nothing at 0,0 then it would pass, even if there was overlap elsewhere.

Supporting documentation: https://stackoverflow.com/a/12067046

One nit change: I updated elm to node to make it more clear what is coming in (the node being checked). Happy to revert this. No additional changes were done.

[marcysutton]

The tests still pass, so the original fix for implicit labels seems to be preserved without needing getClientRects. Nice job 👍

[marcysutton]

FYI: I used getClientRects because it would target the diff in coverage for an inline label wrapping an input, as opposed to the outer bounding box of the label. I guess with this change, the fgNode is the input instead of the label in that case?

[downzer0]

The fgNode in this case would be the foreground "whatever". The function receives two element parameters - one that's the "background" and one that's potentially overlapping (generally a child element). The only way you'd have overlapping would be with absolute positioning, but you'd still need a parent/child combo, right? Are there any scenarios that might not be covered here?

[marcysutton]

When the input was under audit and its parent was an implicit, inline label, the label would be skipped and not considered a background unless it was added to the stack manually. (also the label itself as a node under audit, which contains the actual text content). I couldn't guarantee which element was in the front of the element stack since the node context would be either the child or the parent. If you look at the examples in this MDN doc, you can see getClientRects can be used to check inline content overlap. https://developer.mozilla.org/en-US/docs/Web/API/Element/getClientRects

But if this works, awesome. I want to run it through the aXe extension to be sure.

[WilcoFiers]

It looks like this broke a few tests in Chrome (v59). We used to only get one error there (which I'm fixing in #373), but now I'm getting 5 failures. Please keep in mind that grunt test-webdriver is currently broken, so you'll need to go over the results yourself to check. Firefox also has errors it didn't used to have.

[marcysutton]

I do see the test failures now–there are quite a few when you run grunt test. If you rebase from develop, you can fix some of the flaky iframe tests. But there are some legit failures in both the unit and integration tests.