14:36:58 #startmeeting 14:36:58 Meeting started Thu Aug 18 14:36:58 2011 UTC. The chair is joanie. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:36:58 Useful Commands: #action #agreed #help #info #idea #link #topic. 14:37:16 Agenda is in the topic as usual 14:37:26 #topic Collection Interface 14:37:57 #info Aline was working on documentation for AtspiCollection and had questions and doubts. 14:38:14 #info Aline asked Mike and Joanie about them, and Mike and Joanie were not sure either. 14:38:32 #info Joanie started looking through the code to see if she could help explain the interface to Aline. 14:39:07 #info Upon seeing what all was present, and reviewing what she needed to do in Orca, Joanie asked the following question on the devel list: https://mail.gnome.org/archives/gnome-accessibility-devel/2011-August/msg00000.html 14:39:33 #info Only Ariel answered and his answers were not complete and still left questions and doubts. 14:39:51 #info We're going to be implementing AtkCollection, and Joanie thinks this mess needs sorting out. 14:40:05 Okay, sorry to present via #info but it does make the minutes easier. 14:40:18 I'm opening the floor to discussion, ideas, etc. on the entire mess 14:40:21 * joanie sits 14:40:22 no prob 14:42:24 So... No opinions? 14:42:42 I mean, should I just change the spec of collection interface? 14:42:58 because I'm brilliant and all should do as I say? (joke, joke, joke) 14:43:28 I think we need some sort of input on the matter. Other than my asking what the heck all that stuff is for. ;-) 14:43:45 I guess that if the new one means some improvement, this makes sense 14:44:04 :) Sorry, I do not know much about the topic.... so I would rather listen about it. Bu I am looking to the directions cited 14:44:07 An opinion is: KISS is good: I just got rid of over 1500 lines of dasher code on Friday, so simplify interfaces. Orca is a good consumer. If you haven't needed these functions, then it sounds as though they superfluous. That's just an "opinion" though - I have never used collections! 14:44:09 some of the stuff in there, getting objects by chuncks looks like they were useful performance hacks for the PCs of the time 14:44:27 Hi all. Sorry for being late. 14:44:45 mgorse: hey. No worries 14:45:09 hi mgorse: we're talking about the collection interface 14:45:32 prlw1: Yeah, the problem is that no one has ever used collections. :-) 14:46:30 Do we know if the collection interface is used anywhere ? 14:46:30 AlanBell: Yeah, by the way (to all) I forgot to mention that at the Linux Foundation Open Accessibility Workgroup meeting the other day, which mgorse and I attend, the conclusion from the IA2 folks was similar: It was done for performance reasons. 14:46:36 jjmarin: In Orca. 14:46:45 what is the harm in leaving them? 14:46:58 AlanBell: I'm not saying remove all of collection. 14:47:07 I'm saying remove the superfluous crap 14:47:12 and make the interface more clear 14:47:16 AND the reason being: 14:47:24 We need to implement AtkCollection 14:47:34 and we want AT-SPI2 and ATK to be in sync 14:48:02 so if AT-SPI2 is filled with confusing, convoluted stuff (and possible hackish garbage) 14:48:07 which we don't need 14:48:28 Should we really be wasting finite resources to implement a less-than-ideal AtkCollection? 14:48:59 And then turning around and asking toolkit developers to then implement it? 14:49:17 Instead I think we should figure out what the interface should look like, both AT-SPI2 and ATK. 14:49:26 make sense? 14:49:31 does to me 14:49:35 (i.e. that this isn't change for change's sake) 14:49:43 If collection is just a performance hack, we can reconsider the performance problem and find another approach that it easier to get AT-SPI2 and ATK in sync 14:50:01 jjmarin: +1 14:50:16 But for a number of reasons that's not really ideal 14:50:30 And the most critical one is.... Well, lemme be lazy and #info it. 14:50:52 on the other hand, do we have any idea hoa many actual users of those collections still use old style computers? 14:51:12 s/hoa/how 14:51:21 #info The driving factor for why we need to do AtkCollection sooner rather than later is because LibreOffice needs such a solution in place because they refuse to expose full documents and spreadsheets to ATs. 14:52:04 machado: I wouldn't necessary say "old style computers" as much as I would say "general accessibility performance issues" (which might still persist) 14:52:24 Anyhoo, regarding the libreoffice stuff, there are libreoffice developers **waiting for us** 14:52:36 joanie: agreed 14:52:51 to implement AtkCollection or something that more or less looks an awful lot like collection, but in AtkDocument 14:52:57 so... We have to move forward. 14:53:08 anyhoo, if API were here, he'd kill me for the deep dive 14:53:12 so I'm going to conclude 14:53:56 #action Team members are encouraged to comment on the mailing list thread. 14:54:09 I'll summarize the discussion manually when I do the minutes. 14:54:27 the next topic is aline's who isn't here yet 14:54:28 so... 14:54:43 #topic Should we back out the fix for Bug 656004? 14:54:43 04Bug https://bugzilla.gnome.org/show_bug.cgi?id=656004 normal, Normal, ---, liyuan, UNCONFIRMED, GNOME-Shell gets frozen on selection events using Orca 14:54:55 mgorse: I'm starting to think we should. 14:55:17 But I don't know how much of this is pygobject version stuff 14:55:24 * joanie gives floor to mgorse 14:57:08 joanie: sorry--I'm not really sure what is going on right now. I need to investigate--should have done that a while ago... 14:57:29 mgorse: okay 14:57:36 My guess is that some versions of pygobject crash if we use a GPtrArray, and some crash if we use a GArray 14:58:01 and, looking at the git log, it looks as though the newest pygobject might fix the GArray issue 14:58:17 and there are other cases in libatspi that I missed (in collection, for instance) that use a GArray 14:58:39 so crash on my system. I tried updating pygobject last night on my system, but then I ran into other problems, since I needed to update gobject-introspection as well. 14:58:51 mgorse: could you #info what should be recorded in the minutes? 14:58:57 (please) 14:59:31 Anyway, if the newest version of pygobject works without the change, at least, then I don't mind backing it out, although probably need to check with API at least, since I think we're in API freeze 14:59:40 okay, I'll #info 15:00:22 #info mgorse will investigate whether there is a crash without the fix for 656004 with the newest pygobject 15:00:52 #info since it apparently *causes* a crash for some people, possibly depending on the version of pygobject used 15:01:06 #info and now the API is inconsistent, since collections still use GArrays 15:01:13 (So, at-spi2-core correctly changed from returning a GArray of AtspiAccessible pointers, to a GPtrArray, and that causes pygobject to croak?) 15:02:20 prlw1: I changed it because pygobject was crashing when a GArray was used, because it interpreted the elements at flat objects rather than pointers, or at least some version of pygobject does. And now that change is apparently causing a crash for some versions of pygobject. 15:03:02 sounds more like a pyobject problem? 15:03:40 prlw1: And different people are using different versions of pygobject depending on their development environment 15:04:11 it's honestly (to me) quite disturbing because 2.28.6 is still filled with breakage 15:04:29 and *tons* of changes have and are still taking place in unstable (2.29.x) 15:04:40 (they're not using the traditional GNOME version numbers) 15:04:49 and we have the .90 release next week 15:05:03 and people saying, so, is Orca going to have introspection done? 15:05:33 anyway, sorry for that mini rant 15:05:40 anything else on this topic? 15:05:59 I'm checking the state of wXwigets 15:06:02 joanie: Is API away? I should probably talk with him if he's here--not sure if I need to consult the release team if I'm going to back out that change 15:06:29 sorry ... another topic 15:06:34 mgorse: yeah, he had a conflict 15:06:48 and also an oneiric update gone south 15:07:04 okay. I'll find him if he's on later in the week. In the meantime I'll investigate some 15:07:05 but I *think* he'll be around tomorrow 15:07:12 mgorse: Thank you!!! 15:07:40 the change looks correct though - you back it out, when will it be safe to "back it back in"? 15:07:57 (need an extra "if" in previous) 15:08:51 prlw1: I'm not sure, but what I'm thinking is this, fwiw: 15:09:25 The change was made at least in part due to an issue I stumbled across whilst working on Orca's introspection port (right mgorse?) 15:09:41 and that most people aren't in need of that fix 15:09:49 joanie: My guess is that you updated pygobject and then started running into it 15:09:56 mgorse: correct 15:10:18 because I learned in #python a while back that many things I need to do this conversion are only in pygobject master 15:10:23 or at the time were 15:10:28 so 2.29.1 15:10:37 What we have now is inconsistent, because I missed some things, so I'd like to either back it out or change the others, but either would be an API break 15:10:52 Anyhoo, my point is this guys: 15:11:04 mgorse: ah... 15:11:08 Different versions of pygobject are so freaking different 15:11:26 and pygobject is still quite unstable (imho) this late in the release cycle 15:11:54 that my answer to people who ask about my commiting Orca's introspection port in time for 3.2 shall be "are you f*cking insane?" 15:11:56 ;-) 15:12:07 thus, I don't need this fix now. 15:12:20 and it could be backed out if that is what mgorse and API decide is best. 15:12:55 (re apologies for re ranting.) 15:13:02 OK :-) (or change the others too ) 15:13:08 People using the newest pygobject *may* need it, though. Afaik no one commented on the bug that I filed, although it may have been delt with looking at the git log. So I need to figure out if the crash only happens with a small range of unstable pygobject versions 15:13:40 mgorse: Ok. If you could just figure out what is best and let me know, I'll proceed accordingly. 15:13:51 so... moving on? 15:13:56 (actually Joanie filed the bug and I commented on it) 15:15:09 I want to jump to action items since Aline isn't here to talk about docs. 15:15:17 #topic Action item updates 15:15:45 To be efficient, I'll call on folks. Folks state updates with #info at the front. Minutes generate themselves. 15:15:50 :-) 15:16:06 jjmarin: investigate wxWidgets' internals and how it uses Gtk+ 15:16:53 #info Yesterday Juanjo compiled the branch from the wxWidgets GSoC that is porting wxWidgets to GTK+ 3. 15:17:24 #info So ASAP Juanjo'll take a look to the code and the a11y support and he'll bring his conclusions here :-) 15:17:31 Yay! 15:17:46 And this is for eviacam, right? 15:18:04 yes 15:18:11 awesome 15:18:17 Thanks for the update! 15:18:38 eviacam is written with wxwigets 15:19:15 It will be nice to have such a solution 15:19:25 and evicam + dasher would be cool I thnk 15:19:30 and also think 15:19:45 mgorse: so you have two, but one is a case of me not removing/updating your action items. (Sorry!) 15:19:57 The first which I need to remove is the relocate checks thing 15:20:09 mgorse: but did you ever do this one: 15:20:12 File a bug regarding GAIL treeview children / children-changed signals, attach a patch, and call for testers. 15:21:33 No, I don't think I ever did that. Took me a minute to remember what it was. 15:21:52 mgorse: heh, me too. that is why I went with copy and paste. *grins* 15:21:55 no worries 15:22:20 bnitz: you have a couple. What's the latest with the testing stuff? 15:22:54 #info building 3.1.x on the latest a11y test distro 15:23:26 #info I find that LDTP and Mago do not work. 15:23:47 bnitz: I'm sorry, this is my sleep deprivation no doubt, but when you say building 3.1.x... 3.1.x of what? 15:23:57 the desktop via jhbuild? accerciser?? 15:23:59 #info modified LDTP and Mago to use gi for wnck import but still needs work to port mago to use gsettings. 15:24:22 joanie:sorry, I'll try to clarify 15:27:06 #info correction, building GEDIT 3.1.3 and other components using jhbuild on latest accessibility testing distro. 15:28:03 # Documentation of accessibility testing tools TBD 15:28:19 #info Documentation of accessibility testing tools TBD 15:28:30 That's all. 15:28:53 thanks bnitz! 15:29:00 np. 15:29:04 Uh... I've got one of my own 15:29:54 #info With respect to investigating what other resources we have/can draw upon to achieve the desktop performance testing work: Joanie has drafted a proposal for a contract. 15:30:13 #info That proposal is still being reviewed by Piñeiro. 15:30:35 #info But the amount proposed ($25,000 US) was included in the team budget 15:31:00 #info Pending Piñeiro's approval, the plan is to then submit it to the GNOME Foundation Board for approval 15:31:18 #info And if the Foundation Board approves it, we shall then put it out for bid. 15:31:21 +1 15:31:28 (done) 15:31:50 As an aside, we just don't have the resources in team to do everything we need to do 15:32:13 so Piñeiro and I are thinking at each year we budget for a major task we need done 15:32:18 and convince the Board to pay for it 15:32:28 since it benefits GNOME and its users. 15:32:39 after all, unlike most teams, we're largely volunteers. 15:32:51 anyhoo.... anything else on the action item front? 15:33:10 and if not... 2 minutes over, but. 15:33:17 #topic Miscellaneous Time 15:33:29 if anyone has anything quickish to say, the floor is yours. 15:33:32 * joanie sits 15:34:22 going once.... 15:34:28 * joanie raises the gavel 15:34:37 going twice.... 15:35:10 #endmeeting