/srv/irclogs.ubuntu.com/2013/09/20/#ubuntu-desktop.txt

Laneyxnox: "once"? what about UIF?08:02
Laneyp.s. hi08:02
=== tkamppeter_ is now known as tkamppeter
darkxstHey Laney09:25
asacanyone can look at ==== unity ==== stack changes here ... http://people.canonical.com/~platform/cu2d/results09:40
asacand tell me if that changes anything in the ui?09:40
asacthose are staged and just wanted to punt them in the archive... just double checking because of the freezes09:40
Laneyhi darkxst09:42
Laneyasac: I guess #ubuntu-unity would know more09:42
asacok thought you were the integration experts :)09:42
Laneywell, certainly not /me/, at least for that stuff :P09:43
Laneybut the CI guys should be in unity anyway09:43
darkxstLaney, so T will move to 3.10?09:50
Laneydarkxst: I'd expect so09:50
Laneymight have similar time problems as we had this cycle though09:50
darkxstok, right now there is big problem with the xrandr stuff being moved into mutter09:50
darkxstprobably it will need to be copied into Unity09:50
darkxstthe changes to gnome-desktop will be incredibly messy to revert for unity only09:51
Laneyoh, what did they do?09:52
darkxstall the display config stuff is now a dbus api in mutter09:52
Laneyah, guess this is some abstraction over X/Wayland09:52
darkxstyeh mutter abstracts all that away, so gnome-desktop neednt have any idea whether running X or wayland09:53
LaneyI suppose unity (or something) wants to implement this interface too then09:53
darkxstyup09:53
Laneyfun09:53
darkxstit would be cleanest in Unity I guess09:54
LaneyI suppose file a bug at least09:55
Laneyseems like it'll lead to a lot of duplication09:56
darkxstLaney, I'm guessing Mir doesnt use xrandr etc?09:57
darkxstalthough Xmir might?09:57
Laneyxmir will need to, sure09:58
=== MacSlow is now known as MacSlow|lunch
davmor2Daft Question time.  Empathy, with google moving to it's own system, microsoft moving to sykpe, and facebook only working some of the time, isn't it just a poor irc client now?11:15
jpds_davmor2: Have some empathy.11:17
davmor2jpds_: :)11:18
Mirvogra: can you check/ack content-hub packaging changes? http://pastebin.ubuntu.com/6132295/11:41
Mirvcomes down to adding glib/gsetings/libnih dependenies11:42
ograMirv, shipit11:58
Mirvogra: ok11:58
=== MacSlow|lunch is now known as MacSlow
=== pstolowski is now known as pstolowski|lunch
pittiGood morning12:26
desrtpitti: hello!12:27
pittiit's a desrt! hey mate, how are you?12:27
RAOFdesrt, pitti: Good morning.12:27
desrtstill down in the delta?12:27
pittihey desrt12:28
pittidesrt: yes, last conf day; will fly back tomorrow12:28
desrtpop quiz: pitti and RAOF both wish you good morning.  which one uses the word 'mate' in his greeting?12:28
desrtRAOF: hi :)12:28
desrtpitti: what kind of swanky events did they have this year?12:28
* desrt sort of regrets not coming12:29
pittidesrt: we had a "Mardi Gras" parade on Wednesday evening, right through the city (with police escort)12:31
pittidesrt: right into the "House of Blues" where they had an awesome band (and buffet and bar, of course)12:31
desrtfigures :p12:31
desrtlinuxcon events are getting ridiculous :p12:32
desrtwas it fun? :)12:33
pittidesrt: yes, much12:34
desrtlast year's campfire-on-the-beach thing was pretty sweet too12:34
pittidesrt: I've hung out in bars with life music three times this week now, it's just an amazing place for that12:34
desrtya... we have some pretty good jazz clubs in toronto with live music every night... but i just can't imagine what it's like down there12:35
desrtprobably completely off the charts12:35
=== pstolowski|lunch is now known as pstolowski
xnoxLaney: i think it all landed on time.13:10
=== m_conley_away is now known as m_conley
attentemterry, hi13:41
mterryattente, hello13:41
attentemterry, do you have time to do a review of https://code.launchpad.net/~attente/unity-greeter/indicator-keyboard/+merge/179057 ?13:42
Laneyxnox: I don't think so, because it's now UIF and there is no upload13:42
xnoxLaney: right, sorry. ubuntu-themes upload happened in time for UIF, but not wallpapers.13:43
Laneyho hum!13:44
Laneykenvandine: do you have default wallpaper news? ;-)13:44
kenvandineLaney, no... i was just looking to see if he emailed me again13:45
* kenvandine pings 13:45
Laneyroxor13:45
kenvandinenot on irc... jounih emailed me yesterday saying it was ready and what format i needed it in13:45
kenvandinei told him and asked him to email it to me asap... and nothing13:46
mterryattente, does that need an FFe?13:49
attentemterry, too be honest, i'm not sure13:50
attenteshould i file one anyways?13:50
mterryattente, and maybe we should convert any values of "ug-keyboard" we see into "keyboard" and drop our custom ug-keyboard code, if it's not useful anymore13:50
mterryattente, probably?13:50
attentemterry, sure13:51
=== csslayer_ is now known as csslayer
kenvandineLaney, i got the wallpaper and updated it14:40
kenvandineLaney, but... the package fails to build with --fail-missing, which was already there14:40
kenvandinemissing the translations14:41
Laneythose are new14:41
Laneyit's conceivable there could be problems there :-)14:41
kenvandineoh... the translations are new?14:41
Laneyya14:41
Laneywell, I just turned it on in LP14:41
kenvandinei see14:41
Laneyso they'll be newly exported to the branch now14:41
kenvandinehttps://code.launchpad.net/~ken-vandine/ubuntu-wallpapers/13_10/+merge/18681314:46
kenvandineLaney, ^^14:46
kenvandinemind giving that a review?14:46
Laneyhah14:46
LaneyI was getting ready to blame glib for a bug14:46
Laneybut it turned out to be me :)14:46
kenvandinehaha14:46
kenvandinethere are no bugs in glib14:46
kenvandine:)14:46
Laneyjbicha: who is ubuntu-docs now?14:47
Laneywoo, proper space listing in u-s-s14:49
Laneydesrt: what a GREAT API!14:49
desrtLaney: can i see the patch?14:50
Laneynot yet14:50
desrtah.  you're working on it?14:51
Laneyyeah, need to report when it's in progress14:51
* desrt is happy to see the fruits of his labours appreciated so rapidly14:51
LaneyI don't think the way I used the async api from c++ is very nice14:51
desrtah.  nice.  you're gonna do the progress?14:51
Laneyjust in progress/not14:51
desrtah.  fair enough.14:51
* desrt likes to see the counters go up as a form of progress reporting14:52
Laneycould do that I suppose14:52
Laneybut it'd be annoying to make that work with the bar thing14:52
desrtya... seb said he didn't want progress reporting this way14:52
jbichaLaney: https://code.launchpad.net/~ubuntu-core-doc/ubuntu-docs/saucy14:52
desrti only added it because they want to use this API in nautilus and sushi now14:53
Laneyfair enough14:53
Laneyso I made a struct to pass as user_data to the callback function which contains the 'this' pointer, and a counter so we know when all of the operations are finished to emit the signal which updates the UI14:54
Laneyseems janky14:54
desrtyou're doing it for all of the user's documents/downloads/photos/videos/etc. folders?14:58
Laneyvideos/audio/pictures14:58
desrtif you want to report those as a single consolidated size, that's pretty much what you have to do, i figure14:58
Laneynah, individually14:59
Laneyhttps://wiki.ubuntu.com/AboutThisDevice#phone-storage14:59
desrtinteresting.15:00
desrtso do you show progress for the whole thing until you know all the answers?15:00
desrtor do you fill in each category as you find it?15:00
LaneyI'm doing it all at once15:00
desrti assume you also dispatch the async ops at the same time15:01
Laneyya15:01
desrt(which is a great idea from a performance standpoint btw)15:01
desrtyou're pretty much doing it the correct way, then15:01
LaneyI just keep a counter and when they're all in then fire off the event15:01
Laneywhich makes QML fetch the values again15:01
Laneyok, thanks15:02
Laneyyou can see the reality in a little while15:02
desrti'll be happy to take a look.  gimme a ping.15:02
tedgdesrt, Do you know of an easy way to go from PID to DBus bus name?15:13
desrttedg: no.15:13
tedgI mean, clearly I can get the list and go through getting PIDs, but I was looking for something more elegant.15:13
desrtmostly because a given PID could have multiple bus names15:13
tedgHmm, okay.15:13
tedgTrue, I'd take an array :-)15:14
desrt(and not just multiple well-known names.... but there could be multiple DBus libraries involved)15:14
desrttedg: i think you'd pretty much need to iterate over all the names on the bus, asking for PIDs and do the reverse mapping15:14
desrttedg: the good news is that it's only two roundtrips to do that...15:14
tedgdesrt, ?  GetConnectionUnixProcessID only takes a string?15:15
desrttedg: dispatch all of them at the same time15:15
tedgdesrt, Seems I'd have to call it for each.15:15
tedgOh, yes.15:15
desrtthe art of dbus: look for ways to decrease roundtrips15:15
tedgI was thinking reducing messages.15:15
desrtya.  you're stuck on that point, unfortunately15:15
tedgStill n messages.15:15
desrtmessages don't matter as much as context switches15:16
desrtand in this case you're in luck, because you'll end up with fewer context switches than messages _and_ because you're talking to the bus daemon, so the number of switches is already halved15:16
tedgMulti-core FTW!  ;-)15:16
desrtie: it's not that bad...15:16
tedgNaw, it's not.  But it's a little brute force.  Was hoping there was a lookup I didn't know about.15:17
desrtyou could maybe propose a new API15:17
desrtwhat are you trying to do with it?15:17
tedgUnfortunately this is a "done by Tuesday" type of thing.15:17
desrtlol15:17
tedgNot time for that in v1.015:17
desrtlooking up bus names for PIDs seems slightly suspect...15:18
tedgIt's okay, according the apparmor I'm "trusted" ;-)15:18
* desrt needs to have a chat with apparmor15:18
desrt (( "psst... don't you know this guy works for the NSA?" ))15:18
tedgIt's clear Ubuntu needs to switch to SELinux so I'll be trusted again!15:19
desrttedg: i mean it seems suspect from a code-smell standpoint15:19
tedgdesrt, I agree, and I think we should add API for it.15:20
tedgdesrt, But that'll obviously take longer than I have right now.15:20
desrttedg: no... i mean the desire to map PIDs to busnames is suspect15:20
tedgdesrt, But it can be hidden from users.15:20
desrtwhat are you trying to do?15:20
tedgdesrt, Basically implement the org.freedeskop.Application calling on secondary activation for confined apps.15:21
tedgdesrt, But I only know PIDs15:21
desrtwhy not just send a message to the well-known name?15:21
tedgConfined apps can't have well known names.15:22
desrtO_o15:22
ochosirobert_ancell: we tried to debug the restart-menuitem-missing bug in lightdm a bit more, it seems to return "challange" instead of "yes"15:22
desrtokay.  i don't want to know :)15:22
ochosirobert_ancell: when querying "gdbus call --system --dest org.freedesktop.login1 --object-path  /org/freedesktop/login1 --method org.freedesktop.login1.Manager.CanReboot"15:22
robert_ancellochosi, huh, mine returns yes15:23
robert_ancellochosi, it might be a PolicyKit issue15:23
ochosirobert_ancell: that call was run when logged out15:23
ochosirobert_ancell: when logged in, it also returns "yes" for me15:24
desrttedg: how are you going to deal with processes that have multiple DBus connections?  introspection?15:25
tedgdesrt, Haven't decided.  Trying to decide how evil it would be to just send the message to all and have them reject.15:26
desrttedg: this is what i was talking about when i mentioned code smell :)15:26
tedgdesrt, I think that this is probably something we could also add to the dbus daemon, a way to query the filters of a another process.  Then we'd get "introspection" at least enough to avoid many of these cases.  Though, definitely need to think about it more.15:28
desrttedg: i think you should allow each confined app to own exactly one well-known bus name15:28
desrtie: its ID15:28
desrtway easier to avoid these types of hacks this way15:29
tedgdesrt, We can't because we don't want confined apps to see other confined apps on the system.  Also, which connection gets to be "the" connection?15:29
desrttedg: whichever one claims the name first15:29
desrtsame as normal dbus rules15:29
desrtexcept that you'd reject name ownership requests for anything but the one name that is allowed15:30
tedgBut then for instance my HUD might not work because it's gdbus because my qtdbus connection was faster.15:30
desrttedg: dbus doesn't work like that....15:30
desrtthe connection that wants to have the well-known name has to request it15:30
tedgSure, but if both of them need to be "the" connection.  Then they'll both request it.15:31
desrtno...15:31
desrtin the gapplication case if you have qtdbus coming up as well, only the gdbus connection will try to grab the app's well-known name15:31
desrtthe qtdbus connection won't -- it will remain as having only the unique name15:31
desrtit doesn't matter which one comes up first15:31
desrtthis is already happening in (probably) dozens of apps running on your system today15:31
tedgSure, but let's say you're using GApplication and libubuntufoobar,which happens to want to be on "the connection" too.15:31
tedgThe problem is identity.15:32
tedgThey all want to be on "the" connection then.15:32
desrttedg: well-known names are added to the connection _after_ it is opened (and maybe after it has been used for a while, by other parts of the process)15:32
desrtso it doesn't matter who brings up which connections, or by which library15:32
desrtthe name won't be acquired until GApplication tries to grab it15:32
tedgI understand the dbus part here.  The issue is that both libubuntufoo and libubuntubar want to be on the connection with the well known name.15:33
desrttedg: why?15:33
tedgIf libubuntufoo uses a different dbus implementation than libubuntubar they can't both be there.15:33
desrttedg: that's already a problem in existing dbus15:33
tedgCorrect.  It is.15:33
desrtand it's not a problem that we've ever really .... had a problem with15:33
tedgSo that's why having a wellknown name per-app doesn't help.15:34
desrttedg: unless you're doing _exceptionally_ strange things with well-known names, this just isn't a problem15:34
tedgWell, there are work arounds in thing like libdee to deal with it...15:34
desrttedg: not afaik?15:34
desrtlibdee does fall into what i would consider an exceptionally strange use of well-known names, but it is pretty elegant, actually15:35
tedgLook at how libdee registers well known names... crazy.15:35
desrtlibdee shares a wellknown name between multiple processes as a token15:35
tedgYes.  I know.  And it gets a lot of robustness there as well.15:35
desrtindeed.  very elegant15:35
desrtbut certainly not the sort of thing that GApplication is doing15:35
desrtand if you're forbidding well-known names entirely, you're going to break the dee case anyway15:36
desrt(which may be appropriate.  i'm not commenting on that)15:36
tedgYup.  So anyway.  We're blocking well known names.  And I have to work around that in this case. :-)15:36
tedgTrying to make that not suck as much as possible.15:37
desrtit's going to suck lots :)15:37
desrtit's also going to be super-racy15:37
desrtbut i guess that's the design choice you made....15:37
tedgWe should a NoSQL-like thing where we upload javascript snipets to the dbus-daemon to execute to make things less racy.15:38
* tedg solves all the problems15:38
tedg;-)15:38
desrtwe already have name activation.  it's pretty much the perfect tool to take care of what you're trying to take care of.15:38
* desrt doesn't really feel like discussing this further15:39
=== Ursinha-afk is now known as Ursinha
desrtLaney: your use of a uint pointer is very strange here17:54
desrtalso: it's evil how the list of 4 directories is coded in one place and the number (4) is elsewhere17:55
desrtrather you should have 'uint outstanding;' and do outstanding++ each time you issue a new async17:55
desrtthen outstanding-- when each comes in17:55
attentemterry, i filed the FFe, do we basically just wait for someone from the release team to look at it?17:58
mterryattente, yeah.  I suppose I can review your branch in the mean time17:58
attentemterry, ok, thanks for your help17:58
mterryattente, you need to handle upgrade path in case someone set gsettings and ug-keyboard is still there17:59
mterryattente, and we no longer grab layouts from LightDM.  How does indicator-keyboard grab its list of available layouts?18:00
attentemterry, indicator-keyboard goes to accountsservice for the user's input sources18:02
attentefor the upgrade path, i-keyboard is also taking the user's old keyboard layout settings and moving them to the correct gsettings key18:03
attenteit shouldn't be possible for the user to have both ug-keyboard and i-keyboard together at the same time, if that's what you mean?18:04
mterryattente, sorry, when I said upgrade I meant with the unity-greeter patch  (i.e. please make "ug-keyboard" -> "keyboard"18:04
mterryattente, someone might end up with ug-keyboard in their indicator list still18:04
attentemterry, oh, ok, i misunderstood you, sorry18:04
attentei'll fix it18:04
mterryattente, menubar.vala:237.5-237.33: warning: method `MenuBar.cmp_layout' never used18:06
mterryattente, and add it to Recommends18:10
attentemterry, ok18:10
Laneydesrt: that is a good idea18:12
LaneyI knew it was evil but somehow didn't think of that18:12
desrtLaney: also not sure why you used a pointer there at all18:19
Laneyinstead of a class variable?18:19
=== Ursinha is now known as Ursinha-afk
desrtLaney: no... it should ... i'm confused :)18:25
desrtLaney: why didn't you just use 'uint' instead of 'uint*'?18:26
Laneydesrt: it goes out of scope, or do you mean why not use uint in the struct? because I want to only emit the signal once so I'm sharing the counter18:34
* Laney is off for now, maybe back later :-)18:34
Laneyhave good weekends18:35
=== Ursinha-afk is now known as Ursinha
=== m_conley is now known as m_conley_away
=== m_conley_away is now known as m_conley
=== m_conley is now known as m_conley_away
=== m_conley_away is now known as m_conley
=== m_conley is now known as m_conley_away

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!