It turns out that as a side effect of The Great Hierarchy Flattening, tables are now doing a very nice and accurate job of implementing AtkText. :-) The problem with this, however, is potentially two-fold: * ATs aren't expecting it to be implemented. This might explain some extreme chattiness I'm seeing in Orca when a table gets focus. (It might not. I still need to debug. Regardless....) * Given a really, really big table, I suspect that this accidental implementation of AtkText might result in a performance hit for users of ATs. We should probably stop tables from doing this, and let ATs get the text from the cells within the table, just like they do in other Gtk+ apps.
Created attachment 49706 [details] proposed fix One line fix that seems to solve the problem. So... How does one write a test to verify that AtkText is *not* implemented? In other words, doesn't testing that tables do not implement AtkText cause the test to fail because AtkText is no longer implemented for tables? :-/ Ideas welcome if this fix must have a test in order to be accepted. Thanks!
Created attachment 49740 [details] proposed fix with unit test I was chatting with Xan about this. He provided a way to test that we are no longer implementing AtkText for tables. (Thanks!) If such a test is worth doing, then I propose this patch; if it's not, I propose the first one. :-) For now, obsoleting the original.
Comment on attachment 49740 [details] proposed fix with unit test okay.
Comment on attachment 49740 [details] proposed fix with unit test (In reply to comment #3) > (From update of attachment 49740 [details]) > okay. Thanks! I assume I need to set the commit-queue? flag?
Comment on attachment 49740 [details] proposed fix with unit test Clearing flags on attachment: 49740 Committed r56036: <http://trac.webkit.org/changeset/56036>
All reviewed patches have been landed. Closing bug.