[08:57] Saviq: so i found why the dash tests were failing [08:57] and it sucks :D [08:57] tsdgeos, hit me [08:57] basically the precission of some widths goes crazy and since we're clicking on 0x0 after the million transformations of pos into pos into pos [08:57] we end up clicking on 0, -0,0000somthing [08:58] i.e. out of the area [08:58] .. [08:58] doesn't our mouseClick default to the middle these days? [08:58] yeah [08:58] nope [08:58] it does with http://paste.ubuntu.com/9697260/ [08:58] well, yeah, that's what I meant [08:58] and i'll be proposing the change upstream too [08:58] yeah, makes total sense [08:59] so we don't need the overrides of mouseClick on UnityTestCase [08:59] it's actually a bug in somewhere [08:59] right [08:59] since printing widths [08:59] some has with 640 [08:59] and the child has 639.999999999999 [08:59] which is what starts all the crap [08:59] but to be honest it's not like a user will notice [08:59] because if he's clicking *so* into the border [09:00] he'll expect it to fail if the click doesn't register [09:00] so i'd say click on the middle and carry on [09:06] +1 [09:08] tsdgeos, are you proposing that change? I just rebuilt our vivid silo but could just as well wait for that [09:08] Saviq: i am if you give me 10 mins [09:08] kk [09:08] i'm also changing all those that had width/2, height/2 to have nothing [09:08] looks much cleaner [09:08] when reading [09:08] agreed [09:10] Saviq: and if you can get someone to review https://code.launchpad.net/~aacid/unity8/enterSpreadTest/+merge/245928 we'll get rid of the other flacky test [09:10] mzanetti: ↑ ? [09:10] ack [09:11] tsdgeos, mzanetti, already done [09:20] Saviq: https://code.launchpad.net/~aacid/unity8/mouseClickMiddle/+merge/245932 [09:20] tsdgeos, o/ [09:22] hmmm [09:22] wot, lots of tests failing now [09:22] give me a sec :D [09:23] tsdgeos, yeah, fails here, too [09:23] i think i copy pasted something wrong [09:24] tsdgeos, but actually, trunk fails for me too, at least in PreviewActions [09:25] hmm no [09:25] ah, [09:25] make install [09:25] or well, make [09:28] oh lol [09:28] somehow i dropped the ! from [09:28] if (qtest_events.mouseClick(item, x, y, button, modifiers, delay)) [09:45] Saviq: ok, should be fine now [09:45] tsdgeos, tx, testing [09:59] tsdgeos, now, are you venturing into the AP test flakiness at all? [10:04] Saviq: i'll give it a go yeah [10:04] thanks [10:08] mzanetti, btw, did LP notify you about bug #1408819 ? [10:08] Error: Launchpad bug 1408819 could not be found [10:08] https://bugs.launchpad.net/ubuntu-rtm/+source/unity8/+bug/1408819 [10:08] nope [10:08] mzanetti, crash when tapping on the top quicklist item in rtm [10:08] mzanetti, has there been changes to launcher backend in vivid that are not in rtm? [10:08] ack, fixing [10:08] not that I know of === karni is now known as karni-away [10:09] investigating [10:09] mzanetti, yeah, that's what I thought, so not even sure we're to blame, hopefully the stacktrace will help [10:09] it basically crashes in url-dispatcher [10:09] if it's 100 reproducible it should be clear soon [10:10] in utf8 encoding/decoding... but not every time ¿? [10:10] it is [10:10] but only rtm [10:10] can get you a set of dbgsym packages if you want [10:11] let me see [10:12] url = 0x281ceb0 "tu.calculator/calculator/current-useA" [10:12] mzanetti, looks suspicious ↑ [10:12] it does :) [10:13] but it doesn't happen if you just normally click the icon? [10:13] (battery won't let me turn on my rtm device yet) [10:14] mzanetti, yeah, does not [10:14] mzanetti, and actually for me it only happens the second time I try [10:14] hmm... for me it doesn't happen at all :/ [10:17] nope... not happening [10:17] meh [10:21] hah. got it reproduced on krillin. not on mako with rtm though... strange [10:25] Saviq: i've updated the MR a bit [10:26] mzanetti, right, yeah, it seems to be krillin specific indeed [10:27] heh, /me just understood what's happening... my touchpad does kinetic scrolling *internally* [10:28] Saviq: ? [10:28] which means that if I scroll in one window and move with the mouse (not touchpad) or something changes under the cursor... the scrolling happens on whatever is currently under the cursor [10:28] oh yeah... have that too [10:29] esp. alt+tab goes crazy :D [10:29] right, like WTF [10:30] it actually generates events for *multiple seconds* after I stopped touching it [10:30] ýep [10:30] who thought of that ;/ [10:30] depends on how strong you flick [10:31] wouldn't think Apple did that [10:31] hmm... tbh I haven't had this in KDE with same hardware === karni-away is now known as karni [10:37] mzanetti, doubt that, you must've just not noticed it, it's definitely hw (well, driver) [10:39] hmm... hard to believe too... because atm I scale my qtcreator font like twice a day because of this [10:39] it might be that the driver didn't do it back then [10:40] yeah, it's probably a change between utopic and vivid or something [10:40] happened already in utopic for me [10:40] not sure about trusty - but that's the time I switched to unity [10:43] mzanetti, when https://ci-train.ubuntu.com/job/ubuntu-landing-024-1-build/29/console completes, it'll be ready for testing [10:43] ack [10:45] Why can't I call a public Q_INVOKABLE method from an object returned by ListModel.get(i) like someListModel.get(i).method(), but only directly from the object itself myObject.method(), which was added to the ListModel with someListModel.append(myObject)? [10:46] I can easily access and change regular/public properties via someListModel.get(i).property [10:50] Saviq, mzanetti, tsdgeos: ^ any idea? [10:51] MacSlow: are you registering the object metatye? [10:51] metatype [10:52] tsdgeos, in my mock's plugin I do the usual qmlRegisterType(...) if that's what you mean [10:54] can't think of a reason really [10:57] tsdgeos, I don't use Q_DECLARE_METATYPE in my mock and thus follow the real implementation, which also does not use it for the Notification object/class. [11:00] making this qmltest work with the mocked Notification is a real pain [11:00] MacSlow, prolly best to push some code [11:04] tsdgeos, Saviq: just pushed the current state... if some one can help out... I'm basically lost and don't know why it's not working [11:04] that's the branch lp:~macslow/unity8/swipe-dismiss-snap-decisions [11:04] mzanetti, got my launcher stuck 2 px on screen when hidden, is it something you think can happen with current code or was due to mterry's refactor for the edge demo? [11:05] Cimi: hmm... haven't seen it before [11:05] might be his branch then [11:06] mzanetti, http://paste.ubuntu.com/9697734/ [11:07] mzanetti, unless you can quickly spot an issue here, I will ask him [11:07] looking [11:07] MacSlow, tests/mocks/Unity/Notifications/MockNotificationTypes.h:41: Warning: Property declaration rawActions has no READ accessor function or associated MEMBER variable. The property will be invalid. [11:08] Cimi: hmm... might well be this, yes, but can't tell where. line 71/72 from the paste look suspicious as it influences the move delta below [11:09] MacSlow, tryNotifications doesn't work btw [11:09] mzanetti, ok, will ask him [11:09] thanks [11:09] Saviq, I know... that's what I'm trying to fix [11:10] Saviq, the split between rawActions and actions is because of the way actions have to be passed in via the qmltest [11:11] MacSlow, you're using a ListModel, that's why [11:12] MacSlow, ListModel really isn't geared towards storing objects, it basically is a proxy to properties [11:13] mzanetti, shall we use 0 in the test comparison? https://code.launchpad.net/~mzanetti/unity8/fix-left-edge-on-spread/+merge/243400 [11:13] getRaw isn't even documented [11:15] Saviq, getRaw() is needed/used because QML can't deal with QSharedPointers directly [11:15] MacSlow, I know, but ListModel does not have a getRaw method [11:15] at least not one that would do what you expect it to [11:16] Saviq, so I've to change from ListModel to ListView?! [11:16] MacSlow, no, you need a custom mock Model [11:17] Saviq, *sigh* yet more mocking... all that code just to get close() to work [11:17] MacSlow, ListModel is a very dumb thing [11:17] * MacSlow goes in a corner and cries [11:18] Cimi: whatever :) [11:18] MacSlow, as you're just reusing the same Notification objects, you can probably just reference them directly [11:20] * davmor2 throws wet sponges at macslow to hide his tears......don't let the code see it is destroying your soul ;) [11:20] MacSlow, but I imagine sooner or later you'll want your mock model to be smarter, so you'll need an actual mock model at some point anyway [11:20] Holas [11:21] davmor2, :) [11:22] Saviq, I wouldn't mind the amount of work, if the cost-benefit relation was reasonable... but in this particular case (for the swipe-to-dismis qmltest) it's appallingly bad [11:23] MacSlow, that's why I'm saying - go for a shortcut [11:24] MacSlow, you have static Notification objects on which the .close() method is called [11:24] aah wait [11:24] now I get it [11:24] it's the real implementation that requires getRaw to work [11:24] yes [11:24] MacSlow, still, put all the Notification objects in a list [11:24] MacSlow, and in getRaw [11:25] MacSlow, iterate over that, and find the one with the right nid [11:25] instead of trying to pull it out of the ListMoel [11:25] Model, even [11:25] tsdgeos, hum, mouseClickMiddle got a segfault [11:26] yeah, but it's old code i guess something got weird [11:26] let the new code run [11:26] ah [11:26] ok let's see [11:27] actually i'm not sure it's a segfault [11:27] when we get segfaults the log says "Segmentation Fault" usually [11:27] in this case it's "killed" [11:27] actually not segfault [11:27] wth [11:27] no idea who killed it [11:27] there was a test failure too [11:27] yep [11:27] that's what the other change fixes [11:27] but yeah, let's see the new run [11:27] kk [11:27] MacSlow, see what I mean? [11:27] we're clicking exactly in the middle of the two buttons of the categories [11:28] and once in a while it goes to "the other" button [11:28] that was a valid 0,0 that should not have been blindly replaced :D [11:28] :) [11:28] Saviq, hm... not really sure... first I'd like to actually see the notifications being added in the tryNotification part [11:29] Saviq, getting the methods (close() and invokeAciton()) to work would come second then [11:29] MacSlow, your getRaw implementation depends on ListModel to give the model up, but it's not possible [11:29] MacSlow, so put all the Notification { } declarations you have in a list [11:29] or an object so you can reference them by name and not by id [11:30] Saviq, ok [11:30] and then in getRaw you just go: return dataDict[mockModel.get(id).nid] [11:30] assuming that your keys will be nid [11:30] Saviq, I see [11:31] or well, nid is integer, so a list would work, too [11:44] re [11:45] Cimi: ok, pushed (sorry, my internet connection broke down for the last 10 minutes) [11:45] Saviq: ok, added some debug stuff. the line that crashes is: [11:45] QDesktopServices::openUrl(getUrlForAppId(appId)); [11:46] whereas the passed argument is always: "appid://com.ubuntu.calculator/calculator/current-user-version" [11:49] mzanetti, so it looks like we need to push the bug down to qtubuntu? [11:49] I'd say so... [11:49] intersting thing though: [11:50] I installed url-dispatcher-tools and tried to do "url-dispatcher appid://com.ubuntu.calculator/calculator/current-user-version" [11:50] now I can't repro any more [11:53] right [11:53] still, please comment on your findings and bump to qtubuntu, maybe add a url-dispatcher task to boot [11:58] Saviq: done. why is this private? === MacSlow is now known as MacSlow|lunch [12:22] dednick: could you please merge qtmir trunk into 1355173.trust-prompt-suspend [12:53] mzanetti: busy? Could I add a review to your plate: https://code.launchpad.net/~gerboland/qtmir/acquire-wakelock/+merge/245942 [13:15] greyback: ack [13:15] mzanetti: thanks! [13:23] mzanetti, coredump [13:25] Saviq: ? === MacSlow|lunch is now known as MacSlow [13:25] mzanetti, coredumps potentially include private data [13:25] ah [13:25] I lost the connection to my question [13:25] mzanetti, so bugs are created private by default [13:25] ack [13:27] mzanetti, I usually un-private them after they're retraced and the coredump removed, but the retracer doesn't seem to be keen on getting there [13:28] ah actually it's done now [13:28] yep, a minute ago [13:29] more like over an hour, but yeah [13:29] (private→public is manual) [13:56] greyback, deeep sleep [13:56] :D oopsie [13:57] greyback, and missing c in checkDashFocusDoesNotAquireWakeLock [13:58] thank you typobot [13:58] nw [13:58] can't we run autopilot tests on the destktop anymore? [13:58] i get to run them fine on the phone [13:58] but desktop [13:58] PYTHONPATH=../tests/autopilot autopilot run unity8.indicators.tests.test_indicators.IndicatorPageTitleMatchesWidgetTestCase.test_indicator_page_title_matches_widget [13:58] tsdgeos, https://code.launchpad.net/~saviq/unity8/no-sigstop-main/+merge/245758 [13:58] Ran 14 tests in 0.007s [13:58] OK [13:58] ah that makes sense [13:58] tsdgeos, ah those [13:58] tsdgeos, those never ran on X11 [13:59] really? [13:59] tsdgeos, they only have phone scenarios [13:59] oh [14:03] tsdgeos, FWIW, those are integration tests, so shoulnd't even live in unity8 [14:14] mterry, did you see my comment? [14:14] oh yes [14:14] just saw mail [14:16] :) [14:17] it's amazing how often these autopilot tests fail in CI [14:17] but running on my phone that is exactly the same setup as CI [14:17] can't reproduce :/ [14:24] tsdgeos: I bet it is not the exact same setup though [14:25] davmor2: well i'd like to know what CI does other than flashing and running autopilot that may make it break [14:25] tsdgeos: how about if some background process is using lots of CPU (e.g. apport/whoopsie) - maybe that destabilizes the AP test run? [14:26] greyback: but why those would be running in CI? [14:26] tsdgeos: dunno, just hypothesizing [14:26] sure :) [14:27] tsdgeos: network setup, initial flashing, password and developer mode, the fact that it's behind a restrictive firewall, no cellular except through mocking etc etc [14:27] Cimi, hey, got an update on bug #1363400 ? [14:27] bug 1363400 in ubuntu-system-settings (Ubuntu RTM) "[wizard] allows to "Continue" without connecting to network" [High,Triaged] https://launchpad.net/bugs/1363400 [14:28] dednick, and for you - update on bug #1372061 ? [14:28] bug 1372061 in unity8 (Ubuntu) "SMS notification: time format not translatable" [High,In progress] https://launchpad.net/bugs/1372061 [14:29] Saviq: what happens if you are not at home when you setup the wifi and you don't want to connect? [14:29] davmor2, you skip ;) [14:30] Saviq: and is that not what the continue does is you don't select a network? [14:30] davmor2, continue != skip [14:31] davmor2, but the actual issue is that you get a popup for the password, but still can press continue [14:32] Saviq: fair enough :) Skip is an odd term when there is a continue button too, yes I want to continue, I don't want to skip the phone setup [14:32] davmor2, you're on a page to set up your wifi, not the whole phone [14:33] Saviq: :) [14:54] dednick: where can I find the messages.py mentioned in the comments here? https://code.launchpad.net/~nick-dedekind/unity8/lp1385331.led/+merge/241417 [14:57] tsdgeos: do you know? ^ [14:57] pfff [14:57] it was in a pastebin [14:57] :D [14:57] and in my phone but i think i deleted it [14:57] let me check [14:59] mzanetti: i *think* it's this one http://paste.ubuntu.com/9530101/ [15:00] I'll just run it. what can possibly go wrong :D [15:00] you need some extra python stuff though [15:00] well the imports [15:00] python will complain if you don't have them [15:02] right... it's python2 and we only have 3 on the phone it seems [15:03] mzanetti, 2to3 should help here [15:03] oh, nice [15:03] it wasn't much though, already adjusted [15:03] still a useful hint :) [15:08] I fail to find this one: cannot import name MessagingMenu, introspection typelib not found [15:08] gir1.2-messagingmenu-1.0 [15:08] i think [15:08] mzanetti: ↑ [15:12] thanks! working now [15:16] mzanetti: https://dl.dropboxusercontent.com/u/85539674/messaging_menu.py [15:16] ah. bit late! [15:16] :) [15:19] dednick, if you missed my question, any update on bug #1372061 ? [15:19] bug 1372061 in unity8 (Ubuntu) "SMS notification: time format not translatable" [High,In progress] https://launchpad.net/bugs/1372061 [15:20] Saviq: woops. forgot about that one :/ code is ready. testing did not go well. [15:20] dednick, any case, lower prio than the prompt one [15:21] dednick, let's get back to it later [15:21] Saviq: since the translations are moving to the sdk now i had to fudge them in to test it, but couldnt get it to pick up. [15:22] but technically it should work! think it's just my knowledge of the translation system failing me. [15:23] dednick, I could help if you linked a branch to the bug :) [15:24] Saviq: give me a sec. [15:24] sure [15:24] Saviq, I cannot fix it on my side... [15:25] Saviq, requires changes in network manager side I think [15:25] Cimi, is anyone looking at it? [15:26] Saviq, nope [15:26] Saviq, we can chase again Wellark but he said is not doable... [15:26] in the timeframe [15:27] Wellark, Cimi, can you please update the bug with any status/findings [15:27] Saviq: https://code.launchpad.net/~nick-dedekind/ubuntu-ui-toolkit/lp1378821.time-translation [15:27] https://code.launchpad.net/~nick-dedekind/ubuntu-settings-components/lp1378821.time-translation [15:27] dednick, thanks [15:27] https://code.launchpad.net/~nick-dedekind/unity8/lp1378821.time-translation [15:28] Saviq: the ubuntu-settings-components is just a fix for text alignment. since it's a bit screwed when things get translated these days [15:28] kk [15:34] Cimi, I already fixed the launcher-2px thing, you can continue that review if you like [15:39] dednick: note, menus may have been bolted into indicators, so that HUD would index both? [15:39] mterry, cool thx! [15:41] greyback: not sure about that. but it doesn't seem like the right approach to me. [15:41] dednick: was just a thought. But agreed it's a weird design [15:43] greyback: yeah, i mean inicators are just essentially menus so i guess it makes some logical sense. in the world of the "global menu" anyway. [15:45] dednick: true [16:52] elopio: hey [16:52] mzanetti: hello. [16:52] I'm testing a silo with your branch and I'm getting this error: http://paste.ubuntu.com/9699519/ [16:52] am I doing something wrong? [16:52] seeing that it passed in jenkins [16:54] mzanetti: let me see the code. [16:54] mzanetti: probably you should do autopilot3 run. [16:54] but unity8 helpers still need to be compatible with python2, so I probably forgot that part. [16:54] elopio: oh, that looks better [16:55] elopio: what does that mean? you want to fix it or should we merge it (the test passes with python3)? [16:56] mzanetti: I just need to add the stupid super parameters that py2 needs. Give me a minute. [16:56] that's a good catch, and I better fix it before you merge it. [16:56] elopio: wait a second [16:56] Saviq: this is the only issue I found with the silo ^^ [16:56] wanna do another rebuild? [16:57] elopio: or well, if we break something we have to fix it anyways... [16:59] elopio: ok... Saviq seems to be away. I'll do the rebuilding then. feel free to fix&push [16:59] mzanetti, elopio I'd rather land it and re-merge next week, nothing's using it, it's a new helper etc. [16:59] ah, there he is :) [16:59] ack. works for me [16:59] mzanetti: Saviq: ok. works for me too. [16:59] I'll push it to a new branch. [17:00] mzanetti: Saviq: I've already pushed it to my branch. Should I revert it, or it won't cause any problems? [17:00] I don't think it will. the silos keep a copy [17:00] elopio, it's fine, leave it there [17:00] elopio, we'll just do another MP from the same branch [17:00] elopio, or well, land the same MP again, actually [17:01] Saviq: ok then. It's ready. [17:01] mzanetti: thanks for checking it. [17:02] Saviq: done. spreadsheet updated [17:02] mzanetti, thanks [17:02] elopio: no worries [17:18] mzanetti, for monday https://code.launchpad.net/~mzanetti/unity8/fix-left-edge-on-spread/+merge/243400 [17:19] mzanetti, the enterSpread test fails [17:19] Cimi: ack, thanks === alan_g is now known as alan_g|EOW [18:52] greyback, still around? [22:57] Saviq: back now [22:58] greyback, nothing pressing, when you want to just tell me if I'm talking bullshit in https://code.launchpad.net/~saviq/qtmir/openurl-keep-bytearray/+merge/246017 [22:58] Saviq: ok [22:59] Saviq: that makes the crash go away? [23:00] greyback, couldn't reproduce with that, and debugs as per https://bugs.launchpad.net/ubuntu-rtm/+source/qtmir/+bug/1408819/comments/10 confirm it's ok [23:00] Launchpad bug 1408819 in qtmir (Ubuntu RTM) "unity8 crashed with SIGABRT in g_assertion_message() when launching calculator from quicklist" [Critical,In progress] [23:01] greyback, and it kinda makes sense, the QByteArray from encoded() can be deleted as soon as you take the constData() out of it [23:01] greyback, qtubuntu seems to have avoided that by just putting url.encoded().constData() directly in the call to urld [23:01] so there's no way the QBA would be deleted before it returns [23:02] Saviq: I see your point. Was pondering if the temporary QByreArray could really be deleted by then - and probably yeah, since it's a temporary [23:03] yeah exactly [23:04] now it'd be interesting to find out a way to unit-test this... [23:11] unit testing C++ object lifetimes? Have fun with that