Meeting Summary

• Meeting started by Piñeiro at 14:36:50 UTC.
• Meeting ended at 15:30:33 UTC.
• Participants (Lines Said):
  • Piñeiro - API (109)
  • Joanie - joanie (65)
  • Benjamin - Company (42)
  • Joseph - clown (20)
  • Patrick - prlw1 (8)
  • Mike - mgorse (3)

• Log

Testing

Better Way to Keep on Top of Changes?

• The other day Matthias happened to mention in #a11y that the gnome-contacts contact list had been redone with new custom widgets and that we should test it.
• Joanie thinks we need a way better system than relying upon the kindness of Matthias to give us such heads-up pings.
• Piñeiro suggested as a possible long-term solution: a11y test suit, executed on build.gnome.org, mail sent on any error.
• Joanie asked if it would be possible for someone (other than she and Piñeiro) to do manual testing on the core gnome apps
• Piñeiro said that we will be required to do manual testing for a while
• Patrick commented that Brian Nitz might have something solving those issues. Since Brian was not present, we need to get details from him.

Gtk A11y Dump Tool

• See Benjamin's commit for an overview

• Benjamin created a tool which takes a ui file created in glade, grabs its accessible and dumps everything from it into a text file. (example: glade file, text output)

• Benjamin stated that the idea for this kind of test is to do 3 things:
  1. regression-test: if something got messed up while refactoring
  2. crash-test: if we can make a weird combination of widgets crash, just make a glade file that exposes it, and put it there.
  3. specifying the intended behavior for GTK: we want to create a ui file and the text file manually and then make sure that the gtk implementation actually conforms to it.

• Benjamin stated the third point was the most important and provided the following example: "If we say we want row items in treeviews we'll create a glade file with a treeview and a text file that says 'here are row objects, they have this role, implement these states and interfaces etc' and then we'll go and change the GailTreeView implementation."

• Piñeiro pointed out that Orca already has a set of regression tests and asked if these tests would replace what Orca has. Joanie responded that the Orca regression tests are designed to be sure that changes to Orca don't break access to, say, Gtk. Thus the Orca tests and Benjamin's tests are independent of one another.

• Piñeiro mentioned that it would be good to have similar tests for clutter/st, WebKitGtk, Gecko, etc.

• Benjamin needs an idea of how you would like the output of gail to look for certain widgets/ui files.
• Joanie suggested that this might have some implication for our ATK prioritizations, e.g. today there is no such thing as ATK_ROLE_TABLE_ROW, but there will be.
• We can approach this in a couple of ways:
  1. Benjamin and others working on moving Gail into Gtk+ will provide glade files and the a11y team should say if the output is correct or not
  2. The a11y team can create glade files and say "it should look like this."
• However, Benjamin concluded by saying, "Most of all: please write tests for us so we can fix things." So the team should pitch in in where we can. Tests can be contributed by pinging Benjamin on IRC with a paste.

Testing Distro Update?

• Joseph supplied a patch to Orca to fix the DBus communication with gs-mag. Joanie reviewed it and made some suggestions. Joseph is modifying the patch in light of those suggestions and hopes to have a new patch by end of today or Friday.
• Joanie will roll the new tarball for the testing distro.
• Javi was not present to comment on when he could do a (the final?) spin of the testing distro.

GNOME 3.2 Progress

• Piñeiro is still trying to get some time to work on current St bugs.
• Mike pushed a patch to make atk-bridge not listen for events when no ATs are running.
• The team needs to check how Mike's change improves desktop performance (i.e. with a11y on and no AT listening).

Action Items

New

• Joanie will summarize the 'Gtk A11y dump tool' topic for the minutes.
• The team needs to check how Mike's change improves desktop performance (i.e. with a11y on and no AT listening).

Ongoing / Rolled-Over

• 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.
• Joseph:
  • Investigate if it is feasible to call libatspi using javascript (ie: on gnome-shell)
• Piñeiro:
  • Review the documentation page.