(01:18:06 PM) eseidel: smfr: understandable. looking at your diff now
(01:18:36 PM) eseidel: smfr: I wonder if gtk ax code is failing to update layout itself.
(01:18:42 PM) smfr: could be
(01:18:56 PM) eseidel: and was dependant on some previous layout behavior
(01:18:59 PM) smfr: some of the changes seem like preogressions
(01:19:04 PM) smfr: image.description should be AXDescription: Image title attr. Was AXDescription: .
(01:19:40 PM) eseidel: smfr: that looks like its getting the empty string for the description
...I confirmed it was this revision...
(01:54:52 PM) smfr: mrobinson: ok. so someone needs to updateStyle before accessing the Ax info
Committed r65720: <http://trac.webkit.org/changeset/65720>
Wow. So this is really old. And whatever the original regression is, that is no longer a problem. Yay! That said.... A lot of other things have changed since these were skipped, including:
1. Output of role names has changed
2. Output of attributes has changed (and spits out a lot more)
3. Layout tables are no longer exposed as tables
4. Anonymous blocks are sometimes included in the accessibility tree
5. We have test runner API to get accessible objects by element id
The object-attributes.html test can be unskipped by updating it to take into account the above.
The title-and-alt.html test can be partially fixed in a similar fashion. However, there will still be some failures which fall under the umbrella of bug 157187.
My proposal is that we update both tests, but leave the second one as a failure to be dealt with in bug 157187. Patch shall be forthcoming.
Created attachment 277727 [details]
Comment on attachment 277727 [details]
Clearing flags on attachment: 277727
Committed r200260: <http://trac.webkit.org/changeset/200260>
All reviewed patches have been landed. Closing bug.