The roles-exposed.html test skips a lot of elements just for ATK on the basis that "it's not focusable" or "not to be exposed" or similar reasons. This is problematic: Things that are deliberately not to be exposed as separate accessible objects also require testing. If such objects (e.g. i, b, cite, mark) suddenly start showing up (being "exposed"), that might indicate a regression. We cannot catch such regressions if we do not test for them. Besides, it's being tested for the Mac and the expected results in a number of instances are an empty role. Why not do the same?
<rdar://problem/19069923>
*** Bug 125491 has been marked as a duplicate of this bug. ***
*** Bug 125493 has been marked as a duplicate of this bug. ***
Created attachment 242136 [details] Patch
Comment on attachment 242136 [details] Patch Clearing flags on attachment: 242136 Committed r176515: <http://trac.webkit.org/changeset/176515>
All reviewed patches have been landed. Closing bug.