Meeting minutes
Review agenda and next meeting dates
Matt_King: Requests for changes to agenda?
Joe_Humbert: I have to leave early; can we discuss Testing action menu button with activeDescendant first?
Matt_King: Sure
Matt_King: Next community group meeting: Wednesday August 28
Matt_King: Next AT Driver Subgroup meeting: Monday September 9
Testing action menu button with activeDescendant
James_Scholes: We had reports that people were getting different results when they were using the VoiceOver "recorder" feature
James_Scholes: Specifically, it seems to occur when arrowing through the menu with "quick nav" off
James_Scholes: Upon further investigation, the presence of the recorder appears to be a red herring; VoiceOver's behavior is simply inconsistent
Matt_King: This sounds like a potential browser problem
Matt_King: So you're saying that when we run the setup script to open the menu, the response from VoiceOver can be different than if you open the menu just by pressing Ctrl+Option+Space
James_Scholes: That's right. It's very odd because they do the same thing
James_Scholes: The interesting thing about this case is that pressing the button manually also changes your focus. In theory, there should be no difference between pressing one button and pressing another
Matt_King: Doesn't the menu code set aria-activedescendant?
James_Scholes: Yes
James_Scholes: I also tried clicking the button, but that actually made the output even worse
Joe_Humbert: Do we know which results are correct?
James_Scholes: The people I polled on social media were hearing only the name of the item and not the role
James_Scholes: So we don't know which is correct, but we know which one we get most of the time
Matt_King: This sounds like a great conversation to have with James Craig, and it might be best had in person. That's good because TPAC is coming up
Joe_Humbert: I was also wondering whether this could be related to different versions
James_Scholes: We were testing on 14.6.1
Joe_Humbert: Ok, good. I was, too
Matt_King: So you talked to people outside of this project, James_Scholes?
James_Scholes: Yes, because I thought it had something to do with the AppleScript
James_Scholes: But the people I asked couldn't get it to trigger the behavior even when they used their own techniques
James_Scholes: When we enabled the recorder by pressing the global shortcut key, VoiceOver itself did something strange. Just starting recording caused VoiceOver to speak
James_Scholes: It's possible that there is some crossover between multiple bugs, but we were able to reproduce the reported behavior even without enabling the recorder
James_Scholes: The people I asked were not using precisely the same version of VoiceOver as IsaDC and Joe_Humbert and me
James_Scholes: If the reported output includes "menu item", you should be able to just delete "menu item" and then it should be the same
James_Scholes: I know one person was using 14.5, for instance
Joe_Humbert: Is there a way to have quicker feedback when we're getting these conflicts? Just so we don't have to wait for this meeting
Matt_King: There are W3C Slack channels, but I think the question is whether or not James_Scholes or IsaDC or Sam at PAC would be able to monitor
Matt_King: I personally would not be able to keep up
Hadi: What about using the mailing list?
James_Scholes: I am capable to join another Slack channel, but I also think the mailing list is an under-used resource
Joe_Humbert: I have posted things to the mailing list, but I haven't received responses when I did. I had to wait for the next meeting, anyway
Matt_King: If IsaDC can answer via e-mail... I know I can only check the list a few times a week. But the more people who can help out in terms of responding, the better
Matt_King: Anything else on this?
IsaDC: We need an NVDA tester
Joe_Humbert: Besides the conflicts, I'm done testing with VoiceOver, so I can volunteer for NVDA
IsaDC: We've recently changed two tests, so that explains why Joe_Humbert is reporting that two of the AT responses he recorded are now missing
[general discussion about the design of the test plan navigation buttons for Testers]
Testing of disclosure navigation menu
Hadi: I have a question about interpreting "list boundary is conveyed"
James_Scholes: I believe we talked about this last week and agreed to remove some of these assertions as unnecessary
Hadi: I am thinking specifically about test number 4
Hadi: We are leaving a list and expect to see a list boundary. But in a previous test, when we press "Shift+Tab", we leave the list, but there is no expectation about the list boundary
Matt_King: I think James_Scholes is correct, and I think the "list boundary" assertion should be removed from this test, as well
Matt_King: When you tab into the list, I think a lot of screen readers tell you that you are going into it. I think we agreed that since it's container information, it should be optional
Hadi: If we press Shift+B or Shift+F, are we expecting that we are leaving a list?
Matt_King: I don't think we're asserting anything like that
Hadi: List boundary is part of the test
Matt_King: Correct; we are going to remove those assertions
IsaDC: I think we need an issue for all of these changes because we've accrued a lot of ideas
Matt_King: I'm looking at test 10 and 11 right now...
James_Scholes: Is it fair to say we're removing duplicate assertions? If there are multiple assertions related to the list, they need to be de-duplicated
Matt_King: Yes, I think that's true in all of these cases
Matt_King: When you tab into the list from the button and the list is expanded, JAWS (and I'm pretty sure NVDA) says "list of four items" and then "overview"
Matt_King: "list of four items" is also what the text of a boundary is
Matt_King: If we use the assertion "conveys list boundary", it sort of captures all of that
Matt_King: If we were to use, instead of "list boundary", we could use two assertions: one for "list" role and one for the number of items
Matt_King: I think the "list boundary" assertion is more clear
James_Scholes: I agree
James_Scholes: Wanting to know that I've crossed out of the list is very different than wanting to know the size of the list
Matt_King: To do item number 1: when we're navigating past a list boundary, we want to use an assertion for list boundary and not for the "list" role or the number of items
Matt_King: Number 2: any commands where it jumps over, we want to make the boundary assertion optional
James_Scholes: Are we saying that there shouldn't be any assertions for "role" and "number of items" in the entire plan?
Matt_King: I think so
Hadi: To me, that information is essential
James_Scholes: I think that when you navigate out of a list, I think it's fine for the AT to convey that boundary without saying the number of items, the name of the list, etc.
James_Scholes: But when you enter a list, I think that information is important
Hadi: Agreed
IsaDC: I can make a pull request and summarize the changes in the description and send it to the mailing list
Matt_King: If we find out that there are some differences in expectation when we actually review this preview, then we can talk them out
Hadi: Should I put my testing work on hold, given that the tests are in flux?
Matt_King: Yes, that seems wise