=== salem_ is now known as _salem === _salem is now known as salem_ === salem_ is now known as _salem [07:28] morning [07:28] elo [07:28] Cimi, can we go with a flat white highlight for now? that's the last thing missing in alt nav currently [07:32] Saviq, only in the bad case of weird images and such [07:32] Saviq, otherwise we want what we did [07:32] Saviq, I'd still get the average pixel color and shade it instead white [07:32] Cimi, we can't do that [07:32] Cimi, we can't sample images in QML [07:33] Cimi, the solution is the shader, but I don't know how to do it, you don't know how to do it, and only you are going to notice, so let's stop wasting time [07:33] Saviq, you can use a shadereffectsource of width and height 1pxof the image then stretch it? === tvoss is now known as tvoss|test === tvoss|test is now known as tvoss [07:34] but I see the issue with the color, right [07:34] Cimi, how's that gonna help [07:34] it is not indeed [07:40] Saviq, scope settings? [07:40] Cimi, did you reply to all the comments? [07:40] apart one [07:40] Cimi, in both revisions? [07:40] s/revisions/diffs/ [07:41] Cimi, no you didn't [07:42] Cimi, there's a bunch of unanswered comments in the r1122 into r1160 diff [07:42] Cimi, and another bunch in r1134 into r1185 [07:43] Cimi, it's difficult to find what you did change or not in the big diffs, please reply to all the inline comments [07:46] Saviq, inline diff is still a mess indeed [07:48] Saviq, commented [09:06] seb128: hello [09:06] mzanetti, hey [09:06] seb128: I'd need a new schema for the launcher favorites... the AS stuff as we do currently doesn't work out [09:06] seb128: and can't use the existing unity7 for some incompatibility reason [09:07] mzanetti, oh? which ones? [09:07] seb128: any suggestion on what would be a good one? [09:07] seb128: incompatibility, well, appid:// for one [09:07] unity7 would drop those entries again when started on a unity8 config [09:07] same the other way round... [09:07] k [09:07] do you need a gsettings or a.s schemas or both? [09:08] gsettings [09:08] as too eventually I guess [09:08] but that would be just a one way sync to update the greeter's launcher. not for actual persistent config [09:08] k [09:09] Saviq: mentioned something like /com/ubuntu/touch/ becoming popular? [09:09] err... add seb128: in front and drop the ":" after Saviq ^ [09:09] mzanetti, no, don't use "touch", we are tried to not use that [09:09] good [09:09] ack [09:09] +1 [09:11] mzanetti, we currently have the unity stuff under com.canonical.unity ... not sure if we want to keep that [09:11] well, that's the thing... unity7 uses that too for the launcher [09:11] I guess we could use a different key in there though? [09:11] we could have com.canonical.unity.Launcher [09:11] right [09:11] you could create "items" or "config" instead of "favorites" [09:11] why don't we update unity7? [09:12] to honor appid:// things as well? [09:12] might be an option too I guess [09:12] well, at least to not kick them [09:12] right [09:13] but its additional efforts In the long run I guess [09:13] also unity8 would need to deal with the other keys of unity7 which we don't need/want [09:13] like unity://running-apps [09:17] Saviq: your input? [09:17] the options are: using another key (which would mean sticking with the schema file in lp:unity I guess) [09:17] or spending the efforts to make unity7 and unity8 play along nicely [09:18] or come up with another path, where so far we didn't have good ideas [09:18] we can't spend the time on the latter, we need a custom key (at least temporarily) [09:18] I tend to agree [09:18] so I'll add the "items" key in the existing schema, ok? [09:19] k [09:19] larsu: but hey, since you're around, QGSettings crashes if trying to read/write an invalid key [09:19] larsu: while the doc says it would print a warning [09:24] mzanetti: are you using it from qml or c++? [09:24] larsu: c++ [09:24] * asac posted a gdb output for bug 1360593 [09:24] bug 1360593 in unity8 (Ubuntu) "unity8 freezes randomly" [Critical,New] https://launchpad.net/bugs/1360593 [09:24] Saviq: ^^ anything else you want from this state? [09:25] mzanetti: it only prints a warning in qml [09:25] larsu: not sure I understand :) [09:26] mzanetti: qml is much more dynamic and it is convention to not crash on errors there [09:26] mzanetti: not so much for c++. You should never put unknown values into qgsettings::get [09:26] larsu: I'd say nothing should crash, regardless of the language [09:26] asac, pkill -SIGSEGV unity8 [09:26] s/values/keys [09:26] Saviq: without debugger attached? [09:26] mzanetti: you should not use c++ then... [09:26] e.g. just like this? [09:26] asac, without [09:27] larsu: no... thats just silly talk [09:27] asac, this should end up in a .crash file for unity8 [09:27] mzanetti: why? [09:27] larsu: because it just increases the risk of mistakes [09:27] Saviq: it didnt create a new one for unity8 ... only for unity8-dash i got a crash now [09:28] -rw-r----- 1 phablet whoopsie 1395979 Aug 26 11:27 _usr_bin_unity8-dash.32011.crash [09:28] larsu: if its starting with Q its not supposed to crash :) [09:28] -rw-r----- 1 phablet whoopsie 9041951 Aug 25 14:09 _usr_bin_unity8.32011.crash [09:28] Saviq: guess the -dash is fine? [09:28] * asac finds it interesting that it has the same PID :) [09:28] asac, the -dash is just fallout, uninteresting [09:28] mzanetti: not really. What do you propose should happen when you give ::get() a non-existant key? [09:28] QVariant() [09:28] asac, that's UID, not PID ;) [09:28] Saviq: no new crash file produced [09:28] mzanetti: do you want to check _every_ invocation for empty variants? [09:28] Saviq: oh :) [09:28] hehe [09:28] asac, yeah, fallout of you killing unity8 I mean [09:29] asac, that's why unity8-dash crashed [09:29] Saviq: right, but i didnt get an updated unity8 one [09:29] larsu: on my end? no... why would I? [09:29] asac, yeah, so nothing more you can give us [09:29] too bad [09:29] guess next time i should delete the directory first [09:29] larsu: if the key doesn't exist, I'll get a QVariant(), which means there are no default items in the launcher (in this example) [09:29] mzanetti: because you might get an empty one. Now you always get one of the corrrect type [09:29] asac, shouldn't have been needed, if .uploaded was there apport would overwrite [09:29] mzanetti: no, this is not how it works [09:30] mzanetti: the key always exists. It it's empty, you get an empty list [09:30] larsu: this is how it works throughout the Qt libraries I'd say [09:30] Saviq: there only is .upload [09:30] mzanetti: I don't understand. A qt library would return an empty variant instead of an empty list to signify "no items"!? [09:30] guess i had a crash that wasnt uploaded yet, so its it didnt overwrite? [09:30] asac, yeah [09:30] larsu: you always have bool contains() [09:31] asac, apport bailed out to not overwrite an old report === vrruiz_ is now known as rvr [09:31] * asac checks out why whoospie claims i am offline [09:31] larsu: if you want to know if its there, use that ^^ [09:31] how awful [09:31] * asac will remember for nexttime [09:31] larsu: otherwise, if you don't care, just use it, you'll get a value or not [09:31] mzanetti: I still don't understand. Where is this bool contains()? [09:31] it's not on qvariant... [09:31] larsu: no, QSettings, QList... [09:32] larsu: additionally get methods look more like get(key, default), where you'd get the default if key is not in there [09:32] larsu: like QVariant get(cosnt QString &key, const QVariant &default) [09:32] mzanetti: ah, got it. My point is, the case you're describing doesn't exist. You always know at compile-time which keys exist and which don't [09:32] but *never* crash [09:32] mzanetti: a contains() doesn't make any sense [09:33] why not... of course it does [09:33] mzanetti: give me an example. [09:33] you have a schema file, which key would you want to look up? [09:34] well, take the launcher [09:34] that's like asking whether qvariant has a method called "contains". You know it's not there, why bother asking [09:34] I don't know.. well, I maybe do in the gsettings case... but still its very unQool to crash/abort [09:35] the alternative is worse [09:35] in which you'd need to check every invocation needlessly [09:35] no, you don't [09:35] yes, you do... [09:35] no, the alternative is return a proper default value [09:35] which is what sane APIs do [09:35] ^ [09:35] nor tear down an app from a library [09:35] that's evil [09:36] tsdgeos: the proper default value would be an empty variant [09:36] yes [09:36] which you'd need to check for emptiness [09:36] no you don't [09:36] you just use it [09:37] and then your program behaves weirdly... [09:37] or, if you really need it to contain something, then your get() call supplies the default you expect if key is not there [09:37] but it doesn't crash! [09:37] you still don't get my point [09:37] the key is _never_ not there [09:37] I guess I do get your point... but I just don't agree with it [09:38] data Maybe : Just a | Nothing [09:38] but ok... so if this is what its meant to be... I thought it would be a bug [09:39] do you check the return value of malloc() for NULL? Do you check whether qt's moc files where generated correctly? [09:39] this is the same kind of thing [09:39] * larsu thinks qt should abort() when QObject::connect() fails [09:40] but it doesn't afaik [09:40] the library should not [09:40] anpok: why not? [09:40] it should invoke some sort of error handling [09:40] and let the user decide what to do about [09:40] that's what it does. It prints errors which never get fixed [09:40] throw and clean up properly [09:40] the user can't decide anything [09:40] this is a programmer error [09:40] or write a funny number to *0 [09:41] larsu: well, it returns a bool [09:41] mzanetti: nobody checks for that... [09:41] because that would be madness [09:41] larsu: I do when its needed [09:41] no. crashing is madness [09:41] mzanetti: are there dynamic signals/slots? [09:41] larsu: there are [09:41] yes [09:41] oh, neat [09:41] there are no dynamic gsettings keys, so we don't need that ;) [09:41] and also the objects might be dynamic [09:42] but still we shouldn't crash [09:42] larsu: so you prefer a user not being able to use an application at all to it maybe sometimes misbehaving because a corner case connect failed? [09:42] nah [09:42] tsdgeos: of course [09:42] larsu: your users must love you :D [09:42] honestly i prefer a 99% working app than a 0% working app [09:43] +1 [09:43] tsdgeos: have you seen the amounts of warnings in ~/.cache/upstart? Nobody ever fixes those because they're not visible enough [09:43] tsdgeos: turns out, this is rarely a problem in practice [09:43] these things tend to get noticed quickly [09:43] especially with tests [09:46] useless discussion anyway, it's not like anyone's going to change their minds [09:46] let's have it with a beer next time we meet [09:46] * larsu is looking forward to that ;) [09:48] oh my [09:49] Cimi, can you merge trunk int scope settings? [09:49] Saviq, k [09:49] Cimi, or actually [09:49] Saviq, currently adding semicolons everywhere xD [09:50] Cimi, oh yeah, do merge trunk, I need to merge it everywhere else, too... [09:50] will do [09:52] larsu: but in any case, no matter which one is the more "correct" thing to do, if it starts with Q, a user of it doesn't expect it to abort - ever [09:53] Saviq, for the settingsmodel mock, I used the mock of the backend [09:56] Cimi, I know, I saw that [09:57] Saviq, so if it's wrong, we should probably also fix it there :) [09:57] Cimi, it's not wrong [09:57] Cimi, it's just different style [09:57] Cimi, we're closer to Qt style, they're not [09:58] mzanetti: ya, fair enough. I might change it based on that argument. [09:58] Cimi, but yeah, I know it's a stupid thing to complain about, which is why I won't enforce it [09:58] Cimi, not before I get astyle to do what we want [09:59] tsdgeos, looks like you forgot a prereq [10:00] Saviq: F5 [10:00] tsdgeos, ;) [10:00] tsdgeos, you can't pull emails out from my inbox this way though :P [10:01] Saviq: yeah well, the code should not be there, it's not a missing prereq, just a wrong initial branch [10:01] tsdgeos, ah ok === davmor2_ is now known as davmor2 [10:11] Saviq, the meh in my onLoadedChanged: if (loaded) genericScopeView.scope = scopes.getScope(2); [10:11] Saviq, is for, remove that or just "meh" [10:18] Cimi, I'd rather you didn't check it in with that branch, but meh :P [10:26] Saviq, I did not understand "You should have a test that verifies initial value when != 0." [10:26] Cimi, same that a comment I did above [10:26] Cimi, your initial value for the selector is 1 [10:27] Cimi, you need to verify that the selector indeed displays the option with index 1 [10:27] Cimi, in the mock I mean, it's 1, so "Second" should be displayed when you open the panel [10:27] Cimi, but it doesn't seem to be the case [10:27] nope [10:28] I didn't change the selected index indeed [10:31] Saviq, not sure where the value is stored though [10:31] Cimi, wdym? in the "value" member of the Data struct in the settings model [10:31] Saviq, I did that in Component.onCompleted and onModelChanged [10:31] Saviq, undefined [10:32] ListItem.ItemSelector does not have count [10:33] I confirm widgetData.value seems null [10:33] solve it [10:33] ok [10:33] but write a test first [10:41] Saviq, onSelectedIndexChanged of ListItem.ItemSelector seems to get called [10:41] while loading the component [10:41] thus triggering the updated signal [10:41] Cimi, to 0 I assume? [10:42] to 1 [10:42] no to 0 indeed [10:43] something, probably internal, trigers that [10:53] Saviq, any idea when shall I set the selectedIndex? [10:54] Cimi, talk to #sdk === MacSlow is now known as MacSlow|lunch === cking_ is now known as cking [11:16] Holas [11:16] Cimi, ok, I pushed #CCFFFFFF into alt nav, please have a look if we want to change that value, but let's leave it flat for now [11:24] Saviq, is this only in case of an image right? [11:24] Cimi, no, it's also in case of flat or gradient background [11:24] Saviq, no then [11:25] Saviq, the current solution was better [11:25] Cimi, but I don't want to dig out the values from the background if I can help it [11:25] Cimi, the current solution is still there [11:25] so I don't understand [11:25] the code we did was taking the bg color [11:25] Cimi, alt nav just introduces a background where there wasn't one before [11:26] Cimi, the navigation buttons were transparent (and that's what looks fine in header customizations) [11:26] I see [11:26] Cimi, now they have a solid, gradient or image background [11:26] so it will look quite ugly [11:26] ok I have one more temporary idea [11:26] more pressure for me doing the shader [11:36] Saviq, I don't want to compromise the visuals we had for the 5% someone uses a weird image as bg [11:36] Cimi, it's not exactly like that, but I've a fix [11:44] Saviq, this seems to work http://paste.ubuntu.com/8149278/ [11:45] Cimi, you don't need the onInitialValueChanged, and it's a loop btw [11:45] Cimi, in onSelectedIndexChanged you want to reset initialValue [11:47] Cimi, ok pushed a fix to alt nav [11:47] Cimi, just check out tryDash [11:47] Saviq, onSelectedIndexChanged, initialValue is still 0 [11:48] Saviq, so I need onInitialValueChanged [11:48] Cimi, then set it before you set the model in onLoaded [11:48] Cimi, it will have to be there already [11:49] Cimi, if the binding causes the delay, don't use a binding, initialValue should never change in that case anyway, you got it in a "static" object [11:58] Saviq, so I should now add initialValue for all the settings? [11:58] I am not sure this will work though [11:58] Cimi, no [11:58] feels racy [11:58] asking for troubles [11:59] Cimi, in this case if (hasOwnProperty('initialValue')) makes sense [11:59] Cimi, in the factory [11:59] Cimi, and just set the initial value *before* you set the data [11:59] Cimi, that's not racy then, because initialValue is *set* before selectedIndex can change, because there's no model === alan_g is now known as alan_g|lunch === Saviq_ is now known as Saviq [12:05] any idea why ubuntu-ui-toolkit's tests might fail on Jenkins but not for me? It seems that it can't load the image provider: https://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-utopic-amd64/3388/console [12:05] mzanetti, can I ask you for https://code.launchpad.net/~saviq/unity8/scope-favoriting/+merge/232109 please [12:05] in tst_headActions.qml [12:06] larsu, best talk in #sdk on canonical's irc or #ubuntu-app-devel [12:06] k, thanks [12:11] dednick, hey, anything I can help re: prompts inception? [12:13] Saviq: hey. i just replied to kevins email. It's still in progress. [12:13] dednick, ok, lemme know if you need anything [12:13] I'm about to start testing what i've odne [12:14] and lol on getting an email from your manager with http://memegenerator.net/instance/53651480 [12:14] Saviq: hehe, yeah. [12:16] Saviq, here we go http://paste.ubuntu.com/8149484/ [12:16] oh yay. unity-api has changed... === _salem is now known as salem_ [12:19] Cimi, aaand why the tmpValue? [12:19] Cimi, why can't you just do selectedIndex = initivalValue; initialValue = -1? [12:24] Saviq, because changing selectedIndex will trigger another onSelectedIndexChanged? [12:25] Cimi, ok, right [12:25] Cimi, this is all nasty and loopy, but yeah I think there's nothing better [12:25] Saviq, I was not sure onPropertyChanged ends before processing other events [12:26] Cimi, no, you're right, you need to make sure that the loop will be broken at next iteration [12:27] Cimi, and the way to do that is to make sure the condition doesn't let it through [12:27] yeah [12:27] Cimi, one thing you could do is only trigger updated() if value actually changed [12:28] Cimi, otherwise you'll end up emitting updated() on startup with the initial value [12:28] Saviq, there's a return though [12:28] but yeah maybe [12:28] you're right [12:28] Cimi, not maybe [12:28] Cimi, selectedIndex = tmpValue [12:28] will cause another run through onSelectedIndexChanged [12:29] and will emit updated(selectedIndex) [12:29] even though it's equal to model.value still [12:29] Saviq, it does not get emitted if is 0 [12:29] but if it's 1... let me think [12:30] Cimi, say on startup initialValue is 1 [12:30] Cimi, you go: [12:30] selectedIndexChanged(0) [12:30] because the model got set [12:30] Saviq, channel #Cimi for live debugging :D [12:30] ahah [12:30] inside you go selectedIndex = tmpValue [12:30] which emits selectedIndexChanged(1) === MacSlow|lunch is now known as MacSlow [12:30] and then you end up emitting updated(1) [12:31] because it's neither 0 or initialIndex is >= 0 [12:31] if selectedIndex != widgetData.value then [12:31] so return never happens [12:31] yes [12:31] Cimi, but that you might do in the factory [12:31] Cimi, otherwise you'd need it in all the widgets, not worth it [12:52] tsdgeos, when rebulding ci, make sure to bump revision [12:52] yeah :/ [12:52] sorries [12:53] nw, just restarted with the new r [12:54] was trying to get all approved MR to be also approved by CI [12:54] yeah, good idea [12:54] I'll have another look through once the current set completes === alan_g|lunch is now known as alan_g [12:59] actually started them already === salem_ is now known as _salem [13:52] Hey guys, got a question, not unity related but I think devs here will answer it best -- I want to refactor out some code to a .js file, but it requires access to import Telegram 0.1 -- is there a way to do that? (basically, I need visibility of an enum exposed from C++ via qml plugin) [13:54] don't think that's "correct" [13:54] i mean if you need QML stuff then it's not JS anymore, no? [14:00] lifecycle got merged! \o/ [14:00] zbenjamin, Did the URL dispatcher branch work for you? [14:01] tedg: hey, i did not realize it is finished [14:01] dednick, time to rebase your stuff. I'm pretty sure there will be big conflicts [14:02] zbenjamin, Ah, I mentioned it to Saviq last night. He must be hiding it to show his immense power. ;-) [14:02] tedg: most likely ;) [14:02] zbenjamin, Yes, I expect it to work, but wanted to make sure it was all the same names, etc. [14:02] zbenjamin, you're even subscribed to the bug :) [14:02] zbenjamin, bug #1361349 [14:02] bug 1361349 in url-dispatcher (Ubuntu) "Should open scope:// URIs through unity8-dash" [High,In progress] https://launchpad.net/bugs/1361349 [14:03] * zbenjamin hides [14:03] Again Saviq shows strength as he lords over #ubuntu-unity! [14:04] lol [14:04] tedg: is there some archive i can pull the package from ? or do i need to build it myself? [14:05] dandrader: ya. i've been busy with it for awhile now :) [14:06] zbenjamin, I *think* that Jenkins will build it in a few, but it's also quick to build. [14:06] tedg: not if you need to setup a builder chroot before ,) [14:07] Heh, just build it on the device. [14:07] true [14:13] dednick, I feell your pain. had to do it before [14:13] Saviq, I pushed without the initialValue test [14:16] Cimi, String(factory.source) not + [14:16] ok [14:17] Saviq, also, the change on the loading of the settingspage is by your request [14:17] Saviq, when you asked me to wait onLoaded for open = true [14:17] Cimi, yeah I know [14:17] Cimi, it even makes sense probably [14:18] mterry: hey [14:18] mzanetti, hello! [14:18] Cimi, although we could only delay a fade in or something [14:18] Cimi, but it's fine for now [14:18] mterry: I've got bad news for you :) [14:18] tedg: do i need to create a url-dispatcher file with that patch at all? [14:18] tedg: or will it just work [14:18] mterry: in my efforts to rework the launcher plugin, I stripped out all the accountsservice stuff [14:18] mterry: but there's also good news for you [14:19] mzanetti, :( [14:19] mterry: you'll get a separate launcher plugin that only reads stuff from accountsservice [14:19] Cimi, the initial value test should be easy [14:19] mterry: this mix of dconf and AS was keeping to bring up issues [14:19] mterry: so the idea is this: [14:20] Saviq, it is not [14:20] mterry: the launcher plugin loaded in the session only works on dconf, has the interfaces for count emblems and whatnot [14:20] Saviq, because here the model is static [14:20] Saviq, so at launch it doesn't trigger the onSelectedIndexChanged [14:20] that is triggered with the mocks [14:21] mterry: it will also live-sync *all* its state changes (including recent apps, non-pinned ones) over to AS somehow in a future commit [14:21] * tsdgeos shakes fist against UbuntuShape [14:21] Cimi, just set the model in init [14:21] somehow it's eating the sourceChanged signal of Image [14:21] Cimi, instead of binding it [14:21] mterry: and the greeter just loads another launcher plguin that only reads that AS state [14:21] mterry: basically being a read-only copy of the session one [14:21] mterry: so the question is: does this work for you or am I missing a use case? [14:21] Saviq, ok [14:21] tsdgeos: I reported a bug for that long time ago [14:22]