Meeting Summary

Gtk+ 3 a11y branch is merged with master

Benjamin and Matthias have merged the Gtk+ 3 'a11y' branch into master.
We all need to test.
It's a lot more stable than one might expect given the enormous nature of the change. However, there are still bugs - some of which Joanie has found.
Because Benjamin (Company) hangs out in #a11y and because he prefers to be "poked," Joanie created a list on the A11y Team wiki. Team members should add stuff they find there.
Joanie needs to update Orca's gtk-demo regression tests for gtk3-demo anyway. The act of doing so should uncover more bugs (i.e.
- as she writes the tests).

Li created atk-3 branch.
Proposed use cases:
1. Add there things that break current API.
2. Add things that can be added on atk2 but which would be good to test first.
After a little discussion during the meeting (see full log), use case 2 shouldn't be used for all things, but just for things that people are really not sure about.

Recent discussions on #a11y seem to clarify how a button should be implemented. Piñeiro will try to apply it to StButton, probably this weekend.
Javi reported that Accerciser's pygi work is working fine. It will be good if people could do some testing. His branch is available at github.
Regarding 645665, Joseph reported that Bastien reviewed his initial patch for adding brightness and contrast preferences to gsettings-desktop-schemas. Joseph responded to the review, and uploaded a new patch with the requested modifications.
Mike made some AT-SPI API changes for compatibility with Javascript.

Regarding performance, Piñeiro (who was not present at the previous meeting, when the topic came up) suggested should ask Fer about it, as he made some measures in the past. See this performance-list message for more information.
In related news, a new intern will start at Igalia on Monday. The main tasks will center around helping Piñeiro with some cally (and probably gnome-shell) work, but also other a11y things. He is a good candidate to check that before-after performance measurement, for instance.
Mike looked briefly at the tree view performance test in gtk and figured out that it's still slow to populate a tree view. This seems to be due to the fact that it sends a row-added signal when a row is added, and needs to calculate the row number each time to put in the signal.
Fer reported that he got Firefox multiprocess a11y working on Linux. He anticipates that the plugin atksocket support will likely be included in Firefox 7. The multiprocess branch integration is much further away from being finished and will probably be incorporated in Firefox 8 or 9.
In response to Joseph's request, Fer defined "multiprocess a11y" as "proper logical single tree made of accessible elements coming from different processes."

Team:
- Check how Mike's change improves desktop performance (i.e. with a11y on and no AT listening).
Brian:
- Send a mail summarizing his proposal regarding a testing-centered design.
- Announce his Mago tests on the gnome-accessibility-devel list, and also create a README and INSTALL.
Javi:
- Spin a new (and possibly the last accessibility-specific) spin of the testing distro once he has the new tarball from Joanie and any other required changes requested by our team.
Joanie:
- Investigate what other resources we have/can draw upon to achieve the desktop performance testing work.
Joseph:
- Explore the JavaScript Atspi import further to see how much is available.
Mike:
- File a bug regarding GAIL treeview children / children-changed signals, attach a patch, and call for testers.
Piñeiro:
- Review the documentation page.