The fixes for bug 25534 implemented the bulk of AtkTable. Support for atk_table_get_column_header was removed due to being broken: * when a cell was returned, it did claim to be in the proper row, but * it didn't have a child of ROLE_TEXT with the cell contents as expected, and * often a cell wasn't returned What's puzzling to me is that the nearly identical implementation for atk_table_get_row_header works like a charm. Unless I'm just missing something (a perfectly reasonably possibility), this might be due to "non-traditional" table hierarchy exposed to Atk apps by WebKitGtk (for which I've opened bug 30895).
Created attachment 58654 [details] test case
Created attachment 58655 [details] Screenshot with Accerciser showing expected results (hoping a picture is worth a thousand words. :-) )
Created attachment 58657 [details] possible fix? Not flagging for review for a couple of reasons: 1. No test. 2. Mario had indicated an interest in working on some of these bugs blocking Orca support (yay! thanks!!). One less thing to worry about. :-) So really, I'm attaching this in the hopes that he'll take this bug, refine the patch as needed, and propose a proper fix with a test. :-)
(In reply to comment #3) > Created an attachment (id=58657) [details] > possible fix? > > Not flagging for review for a couple of reasons: > > 1. No test. > > 2. Mario had indicated an interest in working on some of these bugs blocking Orca support (yay! thanks!!). One less thing to worry about. :-) So really, I'm attaching this in the hopes that he'll take this bug, refine the patch as needed, and propose a proper fix with a test. :-) That's the plan but first I need to finish some pending things on other bugs, such as bug 36128, which is btw blocker for this one (so it makes sense to finish work first there :-))
Created attachment 59528 [details] Full patch proposal (with unit test) Hi, after spending most of the time struggling with another bug found while working on this one [*], now that one is fixed I was finally able to finish Joanmarie's work by adding a new unit test in testatk, as well as fixing some small issues in get_row_header() to keep it coherent with get_column_header(), and make it work well. Hence, attaching now the patch that fixes this stuff. Hope you like it. [*] https://bugs.freedesktop.org/show_bug.cgi?id=28659
Comment on attachment 59528 [details] Full patch proposal (with unit test) Looks good.
Created attachment 59535 [details] Some cleanup ON TOP of the full patch proposal Get rid of g_timeout_add() by using g_idle_add() instead. This patch must be applied after the previos one "Full patch proposal (with unit test)".
Comment on attachment 59535 [details] Some cleanup ON TOP of the full patch proposal I'm Krusty the Klown and I approve this message.
Comment on attachment 59528 [details] Full patch proposal (with unit test) Clearing flags on attachment: 59528 Committed r61884: <http://trac.webkit.org/changeset/61884>
Comment on attachment 59535 [details] Some cleanup ON TOP of the full patch proposal Clearing flags on attachment: 59535 Committed r61893: <http://trac.webkit.org/changeset/61893>
All reviewed patches have been landed. Closing bug.