mozilla :: #accessibility

11 Jul 2017
10:41joanieMarcoZ: ping
10:44MarcoZjoanie: Pong!
10:44MarcoZ(Sorry had to finish a little tweet storm first.)
10:44joanieMarcoZ: hey (will look at twitter in a sec, I guess). In the meantime, I have questions :)
10:44joanieSo I was able to repro the hang in that test_valuechange test
10:44joanieand fixed it
10:44MarcoZGo right ahead!
10:44joaniebut then I had a second hang
10:45joanieand at first I figured it was my code,
10:45joaniebut it hangs without my patch
10:45joanieand better yet, only on the mac
10:45joaniein linux, with or without my patch, I'm not seeing this second failure
10:45joaniehints?
10:47MarcoZjoanie: Hmmmm, that's interesting... I know we have some disabled tests in mac where some keyboard interaction failed intermittently, or sometimes even permanently. But if something hangs on your system without your patch, but doesn't on our production servers, it's a good chance this is one of those rare cases where something doesn't want to work locally,
10:47MarcoZbut works fine otherwise. I've seen those.
10:47joanieand this is keyboard related. A down arrow press in a native select element
10:48joanieok my next question is running the failing e10s test. I didn't realize e10s tests were being run.
10:48MarcoZjoanie: Yes, I remember seeing one of those locally as well, but for some reason, they don't happen on our production systems, the tests pass there.
10:49MarcoZjoanie: Yes, those tests are being run now. E10S is being rolled out more and more, and so we need test coverage there, too.
10:51MarcoZAnd because this is E10S, it's a different test architecture. These are what we call browser tests.
10:51* joanie does a search
10:53* joanie has crap for network connectivity. brb
11:15joanieMarcoZ: and in Linux I have a bounds test fail for a combobox descendant. Only I reversed my patch and still get that failure. Which brings me to the following question: Is there a way to toss a patch at the official servers before asking for code review?
11:21MarcoZjoanie: Yes, there is, but I believe you need level 1 committer access for that. Look for the keyword try server.
11:21MarcoZI'm happy to vouch, if that's needed. haven't done these committer access things in a while...
11:22joanieokie dokie. I don't mind either way, though I feel bad when someone has to roll out a commit of mine
11:24MarcoZjoanie: Well, your contributions more than warrant you having committer access at least on level 1. And don't feel bad for backouts, that's why we have sheriffs who analyze such failures and ping the relevant patch authors, like in your case.
11:27joanieMarcoZ: having read the wiki page on TryServer, is it correct that I can set the keyword as a mere mortal non-committer?
11:27joanie(and do so without an r+)
11:29MarcoZOh, I should go re-read that page. I think you can, yes. If it doesn't say that you need committer access, then yes you definitely can.
11:29joaniepage talks about how to do it if you do have committer access
11:29joaniefrom this I'm assuming that mere mortals can set keywords to ask others with said access to shove it at the bots :)
11:30MarcoZjoanie: And yes, Try Server is designed to handle patches without an r+. In fact that's one of the reasons we have them, to be able to preview changes and results before even asking for review, if local builds don't give you the results you need or you don't have all available platforms at hand yourself.
11:31MarcoZjoanie: I think most convenient for you would be to simply request committer level 1 access for now so you can push at try yourself without any hassle.
11:31joanieMarcoZ: I'll start that process later today
11:32joaniewe're wicked behind on ARIA 1.1 milestones. Hence my contributions. ;)
11:33MarcoZjoanie: As for that bounds test: Ignore the local failure. It's another case where local builds somehow behave a bit differently than production builds.
11:33* joanie nods
11:33joanieNow that I think about it, every time a gtk+ theme changes, things wind up being about 1 pixel off from the previous release.
11:34joaniei see that with webkit
13:45joanieMarcoZ: did the request for level 1 commit access for try server. Asked you to vouch.
13:49MarcoZCool, will do that in a sec. Feel free to ping davidb or surkov if you need more vouches.
13:50joaniewill do. now to figure out the xul regressions.
13:58MarcoZjoanie: I bet these are similar to the others. Look for XULMenuListAccessible or such.
13:59joanieMarcoZ: Yeah, that's my plan. I got distracted by reported failures of test_frames.html (which passes for me).
14:13davidbMarcoZ: might be worth dogfooding with stylo on sometime soon
14:15MarcoZdavidb: OK!
14:43MarcoZdavidb: ATM noticing instant slow-down in Firefox when switching tabs after the computer has gone to sleep and I woke it up again. Reproducible each time. Need to shut down and restart Firefox to get good perf with E10S on back.
14:43davidback
15:04davidbeeejay: you joining?
15:06eeejaydavidb: yeah, be there in 2
15:08davidbsurkov: you too
15:08surkovdavidb: whats up?
15:09davidbsurkov: BS meeting
15:09surkovgood
21:16firebotNew Core - Disability Access APIs bug 1380153 filed by twsmith@mozilla.com.
21:16firebothttps://bugzil.la/1380153 NEW, nobody@mozilla.org crash near null in [@ mozilla::a11y::IDRefsIterator::IDRefsIterator]
21:41firebotNew Core - Disability Access APIs bug 1380158 filed by aklotz@mozilla.com.
21:41firebothttps://bugzil.la/1380158 ASSIGNED, aklotz@mozilla.com Use IID parameter passed to mscom::ProxyStream constructor to simplify unmarshaling code
22:24firebotNew Core - Disability Access APIs bug 1380172 filed by twsmith@mozilla.com.
22:24firebothttps://bugzil.la/1380172 NEW, nobody@mozilla.org crash near null in [@ InsertIterator::Next()]
12 Jul 2017
No messages
   
Last message: 14 days and 16 hours ago