/srv/irclogs.ubuntu.com/2013/03/19/#ubuntu-mir.txt

=== yofel_ is now known as yofel
robert_ancellRAOF, did you flash your nexus 4? Any gotchas?01:19
RAOFrobert_ancell: Yeah, I've flashed my nexus 4. No gotchas that I can remember.e01:19
RAOFExcept that you can't keep the screen off but the device on, so it's less interesting as a build box than I hoped :)01:20
robert_ancellRAOF, because it emits too much light?01:20
RAOFrobert_ancell: No, not really. It's aesthetically displeasing.01:21
RAOFI could totally use it as a buildbox.01:21
* robert_ancell hears the worlds smallest violin01:21
ogra_just turn the brightness to 001:21
ogra_there is a sysfs handle for it01:21
RAOFOh, nifty.01:22
ogra_(its just linux ;) )01:22
ogra_though i wouldnt recommend to build on an eMMC/SD card01:23
dufluI wonder how that compares to using the Nexus 7 as a build box (which I found quite successful)01:24
ogra_the SoC is definitely a lot more powerful01:24
dufluThough if you can cross compile from a PC, that wins01:24
ogra_but you dont have USB host mode ...01:25
ogra_well, for packages that are designed for cross building, thats true01:25
RAOFSounds like a job for sshfs!01:25
dufluWhy do you need USB host mode when you have scp/rsync? :)01:25
ogra_well, sshfs/rscny over wlan ...01:26
ogra_could use plain nfs as well, i guess that would be faster ...01:26
dufluYeah wifi's not much worse than USB 2.001:26
ogra_but if these are the chouces i'd actually take the MMC01:26
ogra_*choices01:26
ogra_though stilll, a chromebook is $250 ... and comes with  USB 3.001:27
dufluUgg. New keyboard. I'm very slow now. Please be patient.01:27
RAOFAh, *there* we go - /sys/class/backlight/lm3530/device/lm3630_backlight_on_off01:27
ogra_if you want a serious device to do native arm builds thats the way to go01:27
RAOFIs building on the MMC likely to kill the flash quickly?01:29
ogra_nah, i doubt you will manage that in the time you own the phone01:30
ogra_but its slow I/O01:30
* duflu wonders what's wrong with the existing Mir cross-compile script01:30
ogra_if your build requires at least some disk I/O it will have quite some impact01:31
ogra_duflu, in the end it will be packaged and have a bunch of dependencies ... and that means native builds01:32
ogra_while you build raw upstream code and are sure you have a cross bits in place thats different01:32
ogra_s/a/all/01:32
ogra_cross building packages with a dependency chain is trickier :)01:33
TheMusoRAOF: Do you not have a pandaboard at your disposal? Would that not be a suitable build board?02:12
dufluTheMuso: Pandas are hopelessly slow. Nexus 7 blows it out of the water and supposedly Nexus 4 is similar02:15
dufluogra_: Thanks you made me realise I can finally buy Chromebooks in Australia. But $349 :P02:16
TheMusoduflu: Rotery disk is better than flash for building no?02:26
TheMusoAt least there is that option with panda.02:26
TheMusoBut I guess you could use rotery with the Nexus7.02:26
TheMusoduflu: And out re Chromebooks in Australia.02:27
TheMusos/out/ouch/02:28
dufluTheMuso: It's true the bottleneck on Nexus 7 is slow flash. But I recall that the panda having only two cores (and slower) was a bigger issue compared to the 4-core Tegra302:28
TheMusoFair enough.02:28
TheMusoI guess we need better dev boards then. :p02:32
dufluWell if ARM servers with far-too-many-cores are now real products then that would be a good build solution, surely... ?02:34
dufluI think it was HP talking about such products in Copenhagen.02:34
dufluwww.hp.com/go/moonshot02:36
dufluOf course, that's for Ubuntu builds. Not your desk ;)02:46
robert_ancellRAOF, does installing the PPA mesa hose the phablet UI?02:52
RAOFrobert_ancell: I don't see why it would.02:52
RAOFrobert_ancell: Does it? :)02:52
robert_ancellRAOF, we'll find out soon...02:52
dufluSurely surface-flinger does not depend on Mesa library paths....02:53
dufluOr anything "Mesa"02:53
RAOFAlso, the PPA mesa shouldn't break anything that depends on Mesa *anyway*02:54
dufluTrue. My laptop still works with Ubuntu goodness02:55
dufluWhat's the _eventual_ preference for stride? Should it always represent #bytes or will it become #pixels?07:31
duflu(with respect to MirGraphicsRegion)07:32
alf_duflu: bytes07:35
RAOFBytes is probably more useful, yeah.07:36
duflualf_: Ta. And good morning07:36
* alf_ notices that we now have an Android CI job (the buildng part of it, at least)07:39
tvossalan_g, ping08:39
alan_gtvoss: ?08:39
=== mmrazik is now known as mmrazik|afk
=== mmrazik|afk is now known as mmrazik
=== mmrazik is now known as mmrazik|lunch
alf_alan_g: fyi, I am working on changing render-surfaces to use DisplayServer(Configuration)11:16
=== mmrazik|lunch is now known as mmrazik
=== tvoss is now known as tvoss|lunch
=== tvoss|lunch is now known as tvoss
alan_galf_: Excellent!12:53
=== alan_g is now known as alan_g|tea
=== alan_g|tea is now known as alan_g
racarrMew15:27
racarrMorning :)15:27
alf_racarr: Hello15:28
=== rsalveti_ is now known as rsalveti
=== Darxus_ is now known as Darxus
alan_gracarr: Hello15:50
racarralan_g: Hi15:50
racarralan_g: Btw wanted to ask you...seen any weird segfaults with15:54
racarrdroidinput::sp, droidinput::Vector and finally the android atomics?15:54
racarrI thought I was going to make my big integration test for input branch pass last night and now segfaults in android atomics15:55
racarrtried with and without android types15:55
alan_gracarr: I haven't15:55
alan_gIs this a hybris/android target?15:56
racarralan_g: No native x8615:57
alan_gNot seen it. Want me to look? (Yet.)15:59
racarralan_g: I will try and make a reduced test case later. I haven't spent so much time on it yet15:59
racarrI was just hoping it sounded familiar15:59
alan_gSorry to disappoint you16:00
racarrdeciding what to do with prepare-for-inprocess-egl-clients now16:01
racarr:) no worries16:01
racarrit is weird because the android doesn't really need16:12
racarra NativeDisplayContainer16:12
racarrso it would be possible to write the16:12
racarrmir_egl_native_display_is_valid using16:13
racarran ifdef (just mir_connection_is_valid) on android16:13
racarrseperate versions of mir_egl_native_display_is_valid16:13
racarror an android implementation (which is rather trivial)16:14
racarrbut seems kind of like a verbose way to do nothing16:14
alf_status: investigating a graphical glitch in the WIP code to change render-surfaces to use DisplayServer(Configuration)16:15
racarralf_: Any opinions on what to do with android for NativeDisplayContainer?16:17
racarrI am inclined to just move the definition of mir_toolkit::mesa_native_display_is_valid16:17
racarrto android/gbm16:17
alf_racarr: let me refresh my memory16:17
racarralf_: :) Don't feel like you have to process it immediately16:18
alf_racarr: it would be refreshing change from trying to debug a graphical glitch :)16:18
racarrAha...I know the feeling16:20
racarrthe basic issue is pretty simple, mir_egl_native_display_is_valid uses mcl::EGLNativeDisplayContainer::instance which is defined in the gbm bits16:20
racarralong with the implementation of EGLNativeDisplayContainer16:21
racarrtrying to decide if android should also implement EGLNativeDisplayContainer (seems unnecessary)16:21
racarror one of these other approaches should be used16:21
hikikohi16:23
alf_racarr: It would be a dummy implementation, so I don't see the harm in doing it this way. It would move the complication to the build system instead of the code.16:24
alf_racarr: (almost dummy implementation)16:24
racarrhmm16:27
racarrIm not sure which way to do it though (using the build system) so now I am kind of leaning16:28
racarrto just implementing the display container haha16:28
alf_racarr: IIRC, The only symbol that the rest of the code needs mcl::EGLNativeDisplayContainer::instance(), so defining it in src/client/android/android_native_display_container.cpp should do the trick16:32
racarralf_: But it wants an instance of EGLNativeDisplayContainer returned16:32
racarrso have to write a little android stub16:33
racarrthats what I just did though :)16:33
alf_racarr: right, and great :)16:33
alf_racarr: perhaps Kevin will find the abstraction useful and use it more extensively in the android backend16:34
racarrthats what I am thinking16:34
racarralan_g: prepare-for-inprocess-egl-clients is updated16:34
racarrandroid build works :)16:34
alan_gracarr: I'll look soon16:35
racarrGoing to change https://code.launchpad.net/~robertcarr/mir/happy-monday-dedupe-more-stubs/+merge/153713 to approved if no one objects soon :)16:35
=== hikiko is now known as hikiko_away
alan_gracarr: It will conflict with something I'm working on - but that's life16:41
racarrWhat are you working on?16:43
racarrhmm actually I need16:44
racarranother approve, I think no review since I merged trunk16:44
racarrI dunno if someone wawnts to review...16:44
racarrchanges at 190-211 came in during trunk merge16:44
alan_gracarr: https://code.launchpad.net/~alan-griffiths/mir/refactor-frontend-Surface-cleanup/+merge/15413816:45
alan_gracarr: looking at prepare-for-inprocess-egl-clients16:46
racarr:) thanks16:51
racarrfrntend-Surface-cleanup looks fine so far (quick read)16:51
alan_gracarr: that's what you MP would conflict with.16:51
johnjohn101any ideas on what video system will be best supported by mir (short term)?16:51
alan_gracarr: "+Cflags: -I@INCLUDEDIR@/mir" - can that be right?16:52
racarrerr. let me look16:52
racarralan_g: No its not16:53
racarrits left over from when we had like16:53
racarr err.../usr/include/mir/mir_toolkit/api.h16:54
racarrr512 :)16:54
racarrhrm16:55
racarrmy test doesn't seem16:55
racarrto fail or pass or crash or anything16:56
alan_gracarr: we need to nail down this naming issue.16:56
racarrit appears to exit unit tests normally16:56
racarralan_g: Which, sorry16:56
racarrfor headers?16:56
alan_gracarr: and their directories and packages16:56
alan_gracarr: there's a reason the first TDD step is to write a failing test.16:57
racarralan_g: It did fail until I implemented it16:58
racarrand then it failed in more surprising ways (the atomic stuff, which is now fixed)16:58
racarrbut now it just runs, seems to go through the test16:59
racarrand exit google mock normally16:59
alan_gwhat was that (atomic stuff)?16:59
racarrbut there is nothing about passing or failure or exceptions or errors or16:59
racarranything16:59
racarrjust wasn't keeping a droidinput::sp when I needed to16:59
alan_gTried putting #error in the test body?16:59
racarr?17:00
alan_gTo prove you're compiling the test and haven't lost it in a merge17:00
racarroh well I just tried that but yes17:00
racarrit is the test...17:01
racarrI can remove expectations17:01
racarrand get errors about unexpected calls etc17:02
alan_gracarr: could you be leaking the mock objects?17:02
alan_gThen their expectations are not validated17:02
racarrmm thought about that so then I tried verifying them17:03
racarrand also checking for leaks17:03
racarrthe thing is it doesn't even print OK/FAIL or whatever17:04
racarrit just exits17:04
racarrtried breaking in exit17:04
racarrbut its not informative17:04
alan_gracarr: puts("GOT HERE")?17:04
racarrit finishes the entire test17:04
racarrbody17:04
racarrand invokes the tear down17:04
alan_gracarr: that's too weird.17:05
alan_gWhat's the branch, which test?17:06
racarralan_g: lp:~robertcarr/mir/send-inputs-to-client17:07
racarrunit_tests/inut/android/test_android_input_manager.cpp17:08
racarrset_input_focus17:08
alan_galf_: do we have a useful answer for johnjohn101 @"any ideas on what video system will be best supported by mir (short term)?"17:08
alan_gracarr: I'll see what happens here17:09
racarralan_g: thanks17:10
racarrthe thing is if you comment out17:10
racarrline 23917:10
racarrthe expectations fail17:10
racarrbut it pritns things and becomes useful again and properly faild17:10
alan_gracarr: bzr denies that branch exists17:10
racarrproperly fails*.17:11
racarrerr..17:11
racarralan_g: Oh send-input-to-clients17:11
racarrnot send-inputs-to-client17:11
racarrXD17:11
alan_gWandering "s" ;)17:11
alf_johnjohn101: Do you mean what kind of GPU? Also, do you care about mobile devices or desktop?17:13
alan_gracarr: missing file?  fatal error: mir_test_doubles/mock_input_focus_selector.h: No such file or directory17:16
racarrwhoops...sorry haven't done my pre push pass on this branch yet17:17
alan_gracarr: no worries - but I need that file to test what you have17:18
racarralan_g: Pushed another rev17:19
alan_gracarr: ack17:19
alf_johnjohn101: in any case, on the desktop it's anything with a working DRM/KMS and GL (Mesa) driver (mostly tested with intel and radeon). For mobile is the nexus and galaxy nexus phones.17:20
alf_alan_g: racarr: Is anyone looking into stabilizing the VM CI build? If not, I will try to take a look tomorrow.17:21
alan_galf_: what's the problem?17:22
alan_gracarr: I see the problem17:23
racarrOh?17:23
racarr*prepares for facepalm*17:23
alan_g(But it does leak)17:23
racarroh really17:23
alan_gracarr: I don't see the solution yet17:23
alan_g;)17:23
alf_alan_g: racarr: BespokeDisplayServerTestFixture.focus_management is consistently failing in VM builds (both for CI and autolanding)17:25
racarralan_g: Oh hmm wait something related to the MockInputDispatcher leaks?17:25
racarralan_g: I have not looked yet no...didn't know it was17:25
alan_galf_: I didn't know => haven't looked into it. All yours.17:26
alf_alan_g: racarr: ok17:27
alf_all: Have a great rest of day17:27
alan_galf_: Have a great evening!17:27
racarrSee you alf_17:30
racarralan_g: Hmm wait even if I remove all the expectations I am still getting the same weird non-failure17:31
racarrbut it was definitely failing earlier...what have I done!17:31
=== mmrazik is now known as mmrazik|eod
alan_gracarr: The process is exiting (with status 1)  after doing the TearDown() step. I have a vague recollection of seeing that with an attempt to access an already dead object during destruction.17:38
racarrhmm17:40
=== kgunn is now known as kgunnAFK
racarrstatus: Stuck and hungry so breaking to cook bacon and a bagel :)17:50
racarrduring the day I sometimes think of myself as an object for transforming bacon, bread, and tea and C++ in to17:51
racarrdifferent C++17:51
ogra_poor jono17:54
ogra_but at least he has a bagel as company17:55
jonolol17:56
racarrTook me a second!17:58
alan_gracarr: I don't see what is happening. No C++ exception, no signal and _exit() is called with an uninformative stack. Going to play chess instead.17:58
ogra_funnily you wrote that while on the ubuntu phone list the mail with "[Ubuntu-phone] No bacon?" rolled in17:59
ogra_we're really community driven these days :)17:59
ogra_bacon everywhere !18:00
=== mmrazik|eod is now known as mmrazik
=== alan_g is now known as alan_g|chess
racarralan_g|chess: Enjoy :) I will figure it out18:01
racarrthanks18:01
=== kklimonda_ is now known as kklimonda
min2hello18:15
racarrnarrowed it down significantly...18:17
racarrsomething weird with droidinput::sp and gmock going on18:17
racarrsurprising! ;)18:18
racarroh or maybe not...it seems to really just be the construction of the window handle from the application handle...18:19
racarrmaybe there is something I am missing about droidinput::RefBase18:19
racarrhmm18:25
racarrI fixed it but I don't understand how...and ran in tot he next problem! InputChannel constructor throws without real fds18:25
=== kgunnAFK is now known as kgunn
racarrhmm ok later maybe a factory and mocking for droidinput::InputChannel19:05
racarrfor now the test can use a socketpair ;)19:05
=== yofel_ is now known as yofel
bcbc2I'm running mir on 13.04 with the ppa. When I add 'type=mir' to lightdm.conf and restart lightdm it runs in low graphics mode. If I then reboot I just get a black screen. I'm following the instructions here: http://unity.ubuntu.com/mir/using_mir_on_pc.html Any hints on how to debug?20:08
kgunnbcbc2: did you "ensure you have the proper custom Mesa build installed"20:18
bcbc2kgunn: it says "If you installed Mir using the packages from mir-team staging PPA ..., you should be good to go.20:21
kgunnbcbc2: it sure does...and you should....hmmm20:22
bcbc2kgunn: I tested it with those instructions for "Running MIR natively". I used mir_egltriangle and it worked okay. Does that qualify as 'some-mir-client' in those instructions?20:26
kgunnbcbc2: yes20:27
kgunnits what you think....just a little gl triangle app rendering to a mir surface20:27
bcbc2kgunn: ok... good. So are there any log files I should be checking too troubleshoot the black screen? I've been rebooting into recovery mode and removing 'type=mir' before restarting to get around it.20:30
johnjohn101is mir shipping with the next version of ubuntu ( 13.10 maybe).20:31
kgunnbcbc2: sorry...i'm not the best at desktop x stuff...20:31
kgunnbcbc2: i'll be honest...i've been on quatal....wonder if its cause your on raring?20:32
kgunnbcbc2: no specific knowledge....just thinking aloud20:32
bcbc2kgunn: ok - cool. It's not a big deal. I'll maybe play around with quantal or take a break. Thanks.20:33
kgunnbcbc2: you prompted me to update a second hand machine i had to raring....maybe i'll try those instructions and see what gives20:33
kgunnalso....i'll mention it to some of the other fellows who have more experience on x/compiz/mesa20:34
kgunnjohnjohn101: shipping is kind of loaded20:35
kgunnjohnjohn101: in general there is a goal/vision of convergence of what we achieved with the ubuntu touch/phablet20:36
kgunnif you read some of the spec20:36
johnjohn101next year is ok. i'm ready for ubuntu to be at the next level20:36
kgunnhttps://wiki.ubuntu.com/Mir20:37
kgunnjohnjohn101: sorry...digging out links i'm slow :)20:37
kgunnone more20:37
kgunnhttps://wiki.ubuntu.com/UnityNextSpec20:38
johnjohn101tx kgunn20:38
kgunnnot sure when you got disconnected20:39
kgunnbut if you look at those 2 links20:39
johnjohn101last message from you was "one more"20:39
kgunnhttps://wiki.ubuntu.com/UnityNextSpec20:39
johnjohn101sorry working and connecting to clients and it resets20:39
kgunnnp20:39
kgunnso those 2 links outline what/why and point to the how which is really in all the blueprints20:40
kgunnbut you could think...focus is to get solid on "small screen" devices first20:40
johnjohn101kind of crazy for ubuntu to switch to QT on a dime20:40
kgunnthen grow into "large screen" devices to achieve 1 code base to rule them all20:40
kgunnjohnjohn101: it may seem a little crazy...but i think Qt evolved to the point where you can get gl goodness & great ease of use/fast development (as a toolkit)20:42
=== kgunn is now known as kgunnAFK
racarr[ RUN      ] AndroidInputManagerDispatcherInterceptSetup.server_input_fd_of_focused_surface_is_sent_unfiltered_key_events21:21
racarrDevice added: id=0, name='', sources=0x0000010121:21
racarr[       OK ] AndroidInputManagerDispatcherInterceptSetup.server_input_fd_of_focused_surface_is_sent_unfiltered_key_events (4 ms)21:22
racarrYay21:22
kdub_racarr, on device :) i saw our input library booting up the input devices21:24
kdub_ *?21:25
racarrkdub_: Probably!21:30
racarrthe way it is in trunk is if you have permissions on the input device it should probe the appropriate ones and events will flow from kernel->event hub->input reader->input dispatcher->input dispatcher policy->event filters (there are none in trunk)21:30
racarrand then there will be no focused window so it will be dropped21:30
racarrthis last branch is to hook up focus and21:31
racarrTHEY GOOOOO21:31
RAOFWho wants to review client-side-buffer-age! It'll be fun!22:25
racarrRAOF: I will do it in just 10 minutes. it will be good to switch state back and forth before I do self review22:26
RAOFOh, god. Self review. BAH22:26
racarrhehe22:30
racarrRAOF: https://code.launchpad.net/~robertcarr/mir/send-input-to-clients/+merge/154221 exists and is close to its final state if you want something t look at :) will propose properly after self review round after reviewing client-side-buffer-age!22:30
racarrBut first, tea + food22:31

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