/srv/irclogs.ubuntu.com/2014/07/08/#ubuntu-mir.txt

=== chihchun is now known as chihchun_afk
=== renato_ is now known as Guest11141
duflurobert_ancell: Hey I'm still catching up after vacationing. What's the state of GTK-Mir?02:51
robert_ancellduflu, hey02:51
robert_ancellduflu, it's in the archive now. The patch is pretty basic, the main feature we're blocked on from Mir is support for multiple windows02:52
robert_ancellThe main blocker in Unity is allowing GTK+ applications to be launched without editing the .desktop files (probably using a hard-coded whitelist for now)02:52
duflurobert_ancell: Yeah that's a Unity8 thing. I'm just wondering about the packages. So standard utopic gtk3 supports Mir?02:53
robert_ancellyes02:53
dufluAwesome02:53
dufluVery nice02:53
dufluI shall have to play02:53
robert_ancellWe'll just update the debian patch from GNOME Git when we change it there02:53
duflurobert_ancell: Got a list of "pure" apps that work without recompiling?02:53
robert_ancellduflu, I've been playing with gedit, gnome-calculator, simple-scan as test cases02:54
robert_ancellJust set X-Ubuntu-Touch=true in the .desktop files to show then in Unity02:54
duflurobert_ancell: Straight from utopic archive?02:54
robert_ancelljust with the above change02:54
duflurobert_ancell: How about demo shell?02:54
robert_ancellThey run in there fine02:55
dufluHandy02:55
robert_ancellYou just need to set MIR_SOCKET and permissions appropriate02:55
dufluI didn't see any news about this milestone. But I also didn't read news during June02:55
robert_ancellfine = as well as we currently have support for02:55
robert_ancellin the public?02:55
dufluYeah02:56
robert_ancellhttp://www.omgubuntu.co.uk/2014/06/ubuntu-devs-demo-gtk-apps-running-mir-unity-802:56
robert_ancellI haven't mentioned it's all in the archive now because it's not much use until Unity 8 shows the apps by default02:57
duflurobert_ancell: What's underneath? Is it cairo on GL? cairo on software?03:07
robert_ancellduflu, cairo on software03:07
duflurobert_ancell: Oh so we have a cairo port03:07
duflu?03:07
robert_ancellduflu, no, cairo just renders to the surface memory03:07
robert_ancellit has no idea about mir03:08
duflurobert_ancell: Ah, right. Cairo just uses pointers to memory03:08
robert_ancellWe can't have cairo on GL because that requires libGL to be linked to it and nvidia libGL would use crazy amounts of memory03:08
robert_ancellcairo on GL also slower in some cases still03:08
duflurobert_ancell: Well Cairo on GL is completely unnecessary for use cases where your surfaces are pre-generated03:09
duflu(don't change much)03:09
robert_ancellyeah03:09
duflurobert_ancell: Did you encounter many gnome apps with X linkage?03:10
robert_ancellnot very many, and they've mostly been fixed because Wayland has the same issues03:11
robert_ancellWe had a couple of GTK+ plugins for Ubuntu that were assuming X, but we just disable them if not under X and they work fine03:11
dufluHeh. Nothing like supporting multiple platforms to improve code quality03:13
robert_ancelloh, you also have to unset DISPLAY because GDK tries the X11 backend before the Mir one03:16
duflurobert_ancell: OK. I'll try to get through all the un-fun tasks today and then try it03:20
dufluHow exactly am I involved in 2/3rds of all code reviews already?03:47
dufluI need to close my eyes more03:47
RAOFduflu: What would you like as an overview for RPC work?06:28
dufluRAOF: I guess just more description of what the goal/advantages are06:29
dufluIt's shinier and you'll lose 10kg06:29
duflu(tm)06:29
dufluBuy now06:29
RAOFduflu: Description updated for you :)06:35
dufluRAOF: Mostly curious. I have no immediate plans to do an in-depth review or block it06:36
RAOFAh.06:36
RAOFBasically, it makes us more awesome.06:40
RAOFBoo! I was using test_file_descriptors!06:48
RAOFAlthough all those reasons for its removal are perfectly true. The split-rpc-transport branch has some better tests for fd passing, too.06:49
dufluRAOF: "removes our current serialisation of reads/writes" ... we don't lose ordering guarantees of messages like events do we?06:50
RAOFNo. It means “writes are no longer serialised with reads”06:50
dufluRAOF: Cool. And "expose manual event dispatch" ... what's that?06:51
RAOFHanding a file descriptor out to GTK that, when it becomes readable, GTK can call mir_connection_dispatch() and have all the processing happen there.06:51
RAOFie: threading only when you want it..06:52
dufluSounds nice06:54
dufluRAOF: I was about to try benchmarking the protocol change and see if it's different... but can't build06:57
dufluMaybe later06:57
RAOFduflu: Indeed. See above, “I was using test_file_descriptors!” :)06:57
dufluI lacked the context back then06:58
dufluTime to fix dednick's tests then06:58
RAOFEh, I can just use a different message.06:59
RAOFduflu: There we go, now builds again.07:06
RAOFBAH!08:13
RAOFNext time we're sprinting, can we please have a session on profiling? :)08:13
RAOFAnyway, EOD here.08:13
anpok_mzanetti: http://paste.ubuntu.com/7764887/ please try10:35
mzanettianpok_: hey cool, thanks. will try.10:35
mzanettitrying to understand the comment...10:35
anpok_without that the system thinks that only the first pointer is pressed or released10:37
mzanettianpok_: confirming that your patch works. Thanks a bunch!11:42
anpok_cool11:49
greybackanpok_: thank you!11:53
=== alan_g is now known as alan_g|lunch
=== renato_ is now known as Guest99235
=== greyback is now known as greyback|away
=== alan_g|lunch is now known as alan_g
=== greyback|away is now known as greyback
=== alan_g is now known as alan_g|tea
dobeyAlbertA: hi!14:33
dobeyAlbertA: have you been able to make any progress on https://bugs.launchpad.net/ubuntu/+source/mir/+bug/1337481 ?14:33
ubot5Ubuntu bug 1337481 in mir (Ubuntu) "Crash in libmirclient on app exit on phone" [Undecided,Confirmed]14:33
AlbertAdobey: I was out for us holidays...but I'm back now, I will resume looking at this14:34
dobeyAlbertA: ok, thanks14:35
=== alan_g|tea is now known as alan_g
racarrMorning15:06
alan_gAlbertA: I think I've a working deadlock free observers implementation. I am cleaning up the code (too much implementation in the wrong place). But you're welcome to preview: lp:~alan-griffiths/mir/spike-deadlock-free-observers15:47
AlbertAalan_g: cool I'll check it out15:47
racarrI think we may have some memory fragmentation issues...15:51
racarrMy system got all swappy and slow15:51
racarrso I went to a VT...15:51
racarrwhere I realized I had been running15:51
racarrmir and a demo client for like15:51
racarrwell since friday15:51
racarrthen I killed Mir and everything got better15:52
racarrI wish I had thought to check some things but I had literally just gotten out of bed lol15:52
racarralan_g: on cursor-spike-phase-7 were you just talking about the whitespace? (https://code.launchpad.net/~mir-team/mir/cursor-spike-phase-7-name-some-cursors/+merge/225242)15:53
AlbertAracarr: reads like a mem leak too15:53
alan_gracarr: I was15:53
racarralan_g: We have a seperate C style?15:54
racarrI couldn't find it in the style guide...15:54
alan_gI'm talking practice not documentation15:54
racarrAlbertA: Yeah...15:54
racarralan_g: ?15:54
racarra quick peek at the files shows both are15:54
racarrused15:54
alan_gThen never mind15:55
racarr:)15:55
racarrthe wording of your comment made me paranoid that there was some strange different meaning of char const* const in C that I had forgotten15:56
racarrlol15:56
=== Guest99235 is now known as renato__
kdub_racarr, while we're on the topic of spike-715:58
kdub_I guess, once we support the custom uploading can we remove the names?15:58
greybackracarr: hey you reviewed https://code.launchpad.net/~unity-team/platform-api/devel-for-qtmircompositor/+merge/225320 - you had 2 main issues. One was that a comment (about the event struct conversion) was removed, which I put back. What was the other? I can't find it in my logs now16:00
racarrkdub_: I don't think so...where would the client get the system cursor images then?16:02
racarrkdub_: I mean we could replicate the system under X where you use an external library and a strange undocumented series of names16:02
racarrto load ARGB data for you16:02
kdub_well, I was going to say the external library, without the undocumented/strange stuff16:03
racarrWe would have to make this library though, because xcursor is strange16:04
racarrso why not have it in mir?16:04
kdub_like, it seems strange to me that mir just helps with the theming for cursors16:04
kdub_maybe I'm just asking whats the plan for how much theming mir will be responsible for16:05
kdub_like the same issue exists for window borders16:05
kdub_and stuff like that16:05
racarrmm16:05
kdub_it seems (since mir is a library) that those concerns are different enough that a theming sort of helper library would be justified16:06
racarrthe cursor theme is a little different16:06
kdub_like it could even be libmirtheming.so and live in tree16:06
racarrunlike the initial proposals...you see how its landed16:06
kdub_(talking past the immediate branch here)16:06
racarrwe dont actually expose any theming16:06
racarrits just16:06
racarra set of cursors with different meanings16:06
racarrand then we have an intree implementation that loads the default xcursor theme because thats something lots of shells may be interested in doing16:07
racarrbut the actual themeing, i.e. choosing a theme or something is basically the shells responsibility16:07
racarrI mean...I guess if we did end up with more themeing stuff espescially borders, etc (can't imagine all the way atm)16:08
racarrlibmirtheming would make sense16:08
racarrI think of the cursor names as more like the surface state enum though16:08
racarrif that makes any sense...16:08
greyback+1 for separate library from me. I don't see why core Mir should do anything more than have a method which lets a shell set an image as the cursor. Then the shell can use this helper library to get that image if it wants16:09
racarrwell 1 because USC16:09
racarrshould probably set the cursor in most configurations ...16:09
racarrI mean I guess USC could have the XCursor loader16:09
greybacknot everyone will want to use USC16:09
racarrthis isnt what the cursor names solve though...its just16:10
racarrright so if either USC or unity8 may need to load the cursors16:10
racarrthe code should be in mir -.-16:10
racarrthe names is just though16:10
kdub_right, I guess I'm talking at the next-step level16:10
racarrhow does the client request16:10
racarra "busy" (for example)16:10
racarrcursor16:10
kdub_the "spike-7" branch is okay, but I think that the next step should drive at chasing the specific cursor images out of the mir core16:10
racarrmeh, maybe16:11
racarrI mean I thought about it right because16:11
racarrits obviously ugly16:11
racarrand obviously incorrect an always will be16:11
racarri.e. there is no "This is the correct list of cursor names"16:11
kdub_like, the interested client can link to libmirtheming.so and use that, or if they are very interested in developing their own theme system, they can link to something else16:11
racarrthis doesnt prevent that though16:12
racarryou could replace the_cursor_images with something16:12
racarrthat themes as you want16:12
racarrbut you can still interpret mir_cursor_default16:12
racarras the default cursor from your theme16:12
kdub_sure, it doesn't prevent it, but it does make mir provide more16:12
racarryes16:12
racarrI guess I just figured...I mean some part of ubuntu has to provide it16:13
kdub_right, and by 'mir' i really mean 'core mir'16:13
racarrmaybe...16:14
kdub_sure, I agree... I just think its something that core mir shouldn't have (it can be in-tree under a different helpful theming stuff)16:14
racarrwhere to draw the line?16:14
kdub_like "i want to set a cursor image" or "this is my default image" seems generic16:14
racarrI mean, a shell is free to ignore surface states and implement its own system.16:14
racarrso should there be libmirshellstates16:15
kdub_I think there should be some sort of shell scratchpad, but I think that's a bit different topic16:16
racarrlol16:16
kdub_:D16:16
racarrim not really against some sort of mirtheming seperate thing I guess...just as far as I imagine it so far I think16:18
racarrI would rather just have it in mir (because its pretty harmless...) rather than have16:18
racarra new thing with a name that has to be remembered and thought about16:18
racarrlol16:18
racarrgreyback: Sorry um the16:18
racarrother thing was16:18
racarrgreyback: l42716:19
racarris a strange but I guess justified choice (danmdrader explained)16:19
racarrbut it should maybe come with a comment about how that whole block only makes sense for tablet/phone (i.e. using maximized as restored)16:19
kdub_racarr, yeah, but really, people don't want what's in mir when it gets too specific16:20
kdub_like, 'here's a compositor for you, isn't it nice?' gets met with 'no, sorry, want to do my own thing'16:20
kdub_just seems 'cursor images' is something like that16:20
greybackracarr: it's an initial state that's overwritten before used? I guess it could be U_UNKNOWN_STATE16:20
racarrkdub_: Mmm...yeah its possible...16:21
racarrill continue thinking16:21
racarrgreyback: Yeah I think so...16:21
racarrgreyback: ERr...unless windows16:22
racarrstart on maximized16:22
racarrerr16:22
racarras minimized16:22
racarrbefore they are shown16:22
racarrand line 44516:22
racarrdepends on it16:23
* kdub_ just keeps thinking of the decorations/compiz situation :P16:23
racarrI feel like its fine with a comment that maximized is only the default for tablet or phones or something16:23
greybackracarr: you're right. Code bit clunky. Comment will do16:23
racarr:) thanks16:24
racarrbesides that16:24
racarrit all looked good and left a comment on launchpad as thus16:24
racarrgosh I am so happy today16:26
racarrI bought a memory foam pillow16:26
racarrand slept the best ive slept in like a month16:26
racarrbeen having growing insomnia again16:26
greybackracarr: ok comment pushed16:27
greybackglad you're sleeping better16:27
greybackinsomnia can be horrible, I suffered it a few years ago16:27
racarrMm...I had it really bad as a kid and finally got it under control a few years ago so it was frustrating to see it return16:27
racarrI think in this case it was just back stuff though16:28
racarrand a better pillow will help a lot...maybe a new mattress16:28
racarr...better office chair -.-16:28
greybackracarr: if you're happy, please mark approved16:29
racarrgreyback: Oh hmm one more thing16:29
greybackracarr: omg an iWatch!16:29
racarrnot sure of the deal with platform-api but16:30
racarriWatch?16:30
racarrdoes it require16:30
racarran soname bump or16:30
racarrsome version must have to change right16:30
greybacksorry, bad apple "one more thing" jke16:30
racarroh lol16:30
racarrjust because its API change16:30
racarrABI16:30
racarrchange16:30
greybacksoname should bump yeah, good catch. Should match package version really16:31
racarrI think THEN we are good though16:32
AlbertAuhhh...what happened to mir trunk?16:35
AlbertAit's gone...16:35
AlbertAcamako: ^16:35
racarrlol16:35
AlbertAkgunn: ^16:35
kgunnoh no16:36
alf_mterry_: https://code.launchpad.net/~afrantzis/unity-system-compositor/grand-refactoring-first-steps/+merge/22600016:36
kgunnAlbertA: seems its still there....maybe series funny biz16:36
kgunnbut this is there16:36
kgunnhttps://code.launchpad.net/~mir-team/mir/utopic16:36
greybackracarr: pushed16:37
AlbertAok, I guess we had an alias before? lp:mir => lp:mir/utopic ?16:37
alf_mterry_: please check whether the session switching tests I introduced match the expectations we have for USC16:38
racarralf_: Sounds grand16:38
kgunnAlbertA: yeah...we need to link lp:mir-team/mir/utopic to lp:mir16:38
kgunnweird...16:38
racarr;)16:38
racarrgreyback: Looks good...approve on launchpad16:40
greybackracarr: thank you16:40
AlbertAalf_: yey tests in USC !16:40
racarrbrb more breakfast16:41
alf_AlbertA: only unit tests for one component (although central), but it's a start16:42
alf_AlbertA: do you get emails about USC MPs?16:43
* alf_ wonders if I should add mir-team to the reviewers16:43
AlbertAalf_: nope16:43
AlbertAalan_g: ok, so if I understand correctly in the deadlock free branch, observers are now a linked list of ListItems and a custom lock that allows multiple concurrent readers but only one writer at any one time...16:54
AlbertAalan_g: and allows a reader to become a writer if it's in the same thread context16:55
alan_gAlbertA: that's the idea16:55
alan_gAlso the linked list is a "grow only" structure (nodes are only deleted on destruction)16:56
AlbertAalan_g: ok, to avoid allocation overhead?17:00
alan_gAlbertA: to avoid the "current" node being removed or repurposed17:01
AlbertAalan_g: ahh17:02
alan_gAlbertA: hang on - I'm pushing a cleaner version of the code.17:02
racarrnice technique...:)17:02
alan_gAlbertA: I'm at EOD - if you want to grab the code and run with it feel free. If you don't have time I'll get back to it tomorrow.17:04
AlbertAalan_g: sure17:06
=== alan_g is now known as alan_g|EOD
camakokgunn, AlbertA, yea the alias is not right17:12
camakoany more17:12
camakokgunn, we (duflu/Alan_g/others) want lp:mir to point to devel17:13
kgunncamako: ack...i won't touch a thing....17:13
camakokgunn, and want the distro to pull from lp:~mir-team/mir/utopic17:13
camakokgunn, dunno if this is possible17:13
kgunncamako: right...i was going to say, you might need to talk to "someone"....that used to be didrocks...but he's since moved away from that job17:14
kgunnsil2100 would be the next person i would speak to...17:14
kgunnbut i thikn their machinary needs lp:mir to be trunk...(this is an ancient problem)17:14
* kgunn realizes we're overdue for our quarterly "can't we have trunk not be distro target" discussion17:17
kgunndiscussion...or inquistition17:17
camakokgunn, we should let duflu loose on 'em17:17
kgunnexactly17:18
camakokgunn, currently lp:mir is an alias to the "development focus" branch17:19
camakowhich is now mir/0.517:19
camakobut it has no corresponding branch17:19
camakoso it's difficult to set up the branches in a way that makes sense17:20
camakofor our development17:20
kgunnyeah17:20
kgunnwhich is why it was the way it was17:20
* camako thinks we'll deal with it in the next release17:22
bregmais there now power management in Mir 0.4?18:48
racarrLunnch18:50
kdub_bregma, not in mir, although iirc, that has shifted to USC recently18:55
kdub_AlbertA can probably give better dates around when that happened18:55
bregmahmm, well, it's crept in somewhere and requires I reboot the desktop every 5 minutes because there's no wake support in the Unity 8 shell.18:56
bschaefergreyback, were you able to get xmir working?18:57
greybackbschaefer: nope. I gave up and logged a bug https://bugs.launchpad.net/mir/+bug/133900118:58
ubot5Ubuntu bug 1339001 in Mir "XMir not working with nouveau" [Undecided,New]18:58
bschaefergreyback, good idea :)18:58
greybackbschaefer: things work on my intel machine, so guess it's nouveau issue18:58
bregmaahh, the power button works like on the phone, a little unexpected on a desktop system18:58
bschaefergreyback, yeah, sounds like a nouveau issue then ... strange has to be have been a recent push18:58
bschaeferbregma, like, holding the button to turn it on?18:58
bregmabschaefer, yes, just like the phone18:59
bregmaand, the power-down timeout is seems like about 30 seconds18:59
bregmaalso a little unexpected on a desktop18:59
bschaefero wow, thats a long time to hold it18:59
bschaeferyeah18:59
bschaefer(at lease on a desktop)18:59
bregmano, you just tap the power button, the screen shuts off with no activity in 30 seconds19:00
bregmaalso, the second time you press the power button it just shuts down the whole machine ☹19:01
* bschaefer does not know how phones work19:01
bschaefermy phone i have to hold the power button for ~30 seconds to shut it down19:01
bregmabschaefer, not shut down, wake up19:01
bschaeferotherwise it locks the screen19:01
bschaeferi see yeah19:01
AlbertAbregma: I updated USC to disable power key and inactivity handling with xmir19:16
AlbertAbregma: bu tthis is with unity 8 desktop? So it still uses usc wrapper to start up unity-system-compositor?19:16
bregmawhat is usc wrapper?19:17
AlbertAin touch19:17
AlbertAusc-wrapper is the one that lightdm will call to startup unity-system-compositor19:17
AlbertAfor xmir, there's another wrapper which lives in tree in unity-system-compositor19:18
bregmathat's new to me19:18
AlbertAI modified that last thursday to avoid the power key/inactivity stuff for xmir19:18
bregmawhat's with the sudden support for xmir?  who uses it?19:18
AlbertAbregma: so do you know how usc get's started for unity8 desktop?19:19
bregmalightdm starts u-s-c:  if there have been mods to have it do something special for the phone they didn't get publicized on the channels i follow19:19
AlbertAbregma: directly? no wrapper script or anything?19:20
bregmathat's the way it used to do it19:20
AlbertAthis is what will be used for ubuntu-desktop-mir:19:21
AlbertAhttp://bazaar.launchpad.net/~unity-system-compositor-team/unity-system-compositor/trunk/view/head:/debian/unity-system-compositor.sleep19:21
AlbertAhttp://bazaar.launchpad.net/~unity-system-compositor-team/unity-system-compositor/trunk/view/head:/debian/10-unity-system-compositor.conf19:21
bregmawhat is ubuntu-desktop-mir and how does it relate to the stuff currently in use in the Unity 8 desktop?19:22
AlbertAbregma: I don't know how it relates, I just know we use that for xmir19:22
AlbertAi.e. you install ubuntu-desktop-mir and you get xmir with unity719:23
bregmawhat is xmir used for?19:23
racarradventure19:23
bregmawe do not support Unity 7 running that way19:23
racarrits "available for testing"19:23
bregmaall I know is if you installed ubuntu-desktop-mir it would break OpenGL and my team would have to spend time supported people with broken systems19:26
AlbertAbregma: ok let's get back to the core issue19:26
AlbertAbregma: Is it unity8-desktop-session that has the lightdm confs?19:27
bregmaAlbertA, yes indeed19:29
AlbertAbregma: I don't see unity-system-compositor there though...19:30
bregmaAlbertA, no, because lightdm knows how to configure it automatically ... although if I can add a wrapper, I'd be happy19:32
AlbertAbregma: oh I see...well all we need is to set the cmd line arg: --disable-inactivity-policy=true19:33
bregmaevidently the ability has been there for a while, now I look through the code19:33
bregmaI shall play with this until usability is restored19:33
AlbertAbregma: well actually...usc may need more changes,19:34
AlbertAbregma: we need different policies I guess19:35
AlbertAbregma: because I assume you still need inactivity to take effect19:35
AlbertAbregma: but wakeups should occur with mouse or keyboard events, not power key19:35
bregmaAlbertA, that will require support from the Shell, and it's not on the near-term radar19:36
AlbertAbregma: ok19:36
AlbertAbregma: well ping me if I can help further...19:38
bregmasure thing19:38
racarrthe cursor renderable needs to support padding the image if the cursor buffer is larger than the image20:53
racarri.e. on some gbm where its always 64x6420:53
racarrlike mgm::Cursor does now20:53
racarrthe thing is...it should also support multiple pixel formats now.20:53
racarrI guess I am just wondering if its appropriate to link in pixman20:54
kdub_groan20:54
kdub_:)20:54
kdub_nothing against pixman :)20:54
kdub_but once we do that... why not libjpeg and  and libpng20:55
anpokhm recently saw a patch that allows fantastic 128x128 for kaveri gpus..20:57
kdub_I probably wouldn't mind libmirhelpfultheming.so linking to things like that though20:57
racarrlol20:58
racarranpok: Yes hence most. the thing is just that the cursor renderable buffer may not be the same size20:58
racarras the cursor20:58
racarrnvm im dumb. for some reason I thought I was going to have to specialize the padding code for different20:59
racarrpixel formats20:59
racarrbut20:59
racarrtransparent20:59
racarris the same everywhere...20:59
racarr*facepalm*20:59
=== xnox is now known as xnox_
=== xnox_ is now known as Eisbrecher
=== Eisbrecher is now known as Eisbrecher_xnox
=== renato_ is now known as Guest83027

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