/srv/irclogs.ubuntu.com/2013/12/12/#ubuntu-mir.txt

=== thomi_ is now known as thomi
=== chihchun_afk is now known as chihchun
=== shuduo_afk is now known as shuduo
=== duflu_ is now known as duflu
=== marrusl is now known as marrusl_afk
=== marrusl_afk is now known as marrusl
=== shuduo is now known as shuduo_afk
=== shuduo_afk is now known as shuduo
=== shuduo is now known as shuduo_afk
=== shuduo_afk is now known as shuduo
=== tvoss|food is now known as tvoss
* duflu suspects we'll be landing things by hand again :/06:37
RAOFWow. I think I may have found a case where g++'s error message is better than clang's.07:02
RAOFHuh. Tell a lie.07:05
RAOFBah!07:08
RAOFC++11, why does “GLContext(GLContext const&) = delete;” delete all subclass' move constructors?07:08
RAOFAh, ok.08:00
RAOFSo, nested worked on android by accident *anyway*.08:00
RAOF(It selects an EGLConfig supporting EGL_WINDOW_BIT and then proceeds to create a pbuffer surface with that config)08:03
=== shuduo is now known as shuduo_afk
RAOFMan, we have *far* too many EGLContext wrappers.08:26
dufluCode duplication... to be sure to be sure08:30
alf_RAOF: kdub wanted to deduplicate, but it seems you got there first08:40
RAOFYup.08:41
=== shuduo_afk is now known as shuduo
anpokduflu: i meant the methods of DefaultServerConfiguation are spread accross multiple cpp files09:16
dufluanpok: It's not a class or a method. I think you're confusing default_configuration with DefaultServerConfiguration. They're unrelated09:17
anpokoh .. too many files named default_configuration.cpp09:20
alf_anpok: That's so that we can keep implementation details internal to the components that provide them. Each component implements part of the DefaultServerConfiguration in the default_configuration.cpp file located in its source directory.09:22
alan_gduflu: did we figure out what causes the CI test failures?09:33
duflualan_g: No, haven't looked very deeply. They commenced failing again as soon as I said they were passing09:33
duflualan_g: Just got a full pass a second ago!09:34
alan_gThere's a lesson in that! ;)09:34
dufluThis is way too erratic and unreliable09:34
alan_g+1009:34
tvossdidrocks, can you help alan_g and duflu track down the test flakiness?^09:35
didrockstvoss: isn't that the CI team role or the QA team one?09:50
didrockstvoss: FYI, all known flaky tests are summarized in my last emails09:51
didrocksif you see more failure, it's either:09:51
didrocks- working with the other upstreams to get their new revealed flakyness fixed09:51
didrocks- or fix a real bug09:51
didrocks(too many bugs are excused as flakyness, where they are flaky behavior in reality)09:52
tvossdidrocks, fair, just trying to identify a POC09:52
alan_gdidrocks: I was discussing it with ChrisGagnon yesterday - we're getting inconsistent (and not locally reproducible) results on the phone hardware that's just been switched on.09:53
didrocksalan_g: will be interesting to have all debug infos for the other teams to fix it (can be autopilot, can be the app or can be a real issue)09:54
didrocksanyway, I think it's QA/CI team that you want09:54
didrocksto help identifying the cause09:55
alan_gdidrocks: I guess you've been too helpful to tvoss in the past. ;)09:56
didrocksahah, no, he still has credits on my help card :p just I think I won't be able to help reliably (especially with all the meetings I'm in)09:57
didrocksand I think some people need to fulfill and own their role now ;)09:57
* duflu is paranoid and assumes he has not credits left09:58
dufluAlthough I don't think we need didrocks help on this issue yet09:58
alan_ggreyback: are you subscribed to mir-devel and the "Shell communication channel: simple, half-assed or fully-arsed?" discussion10:08
* alan_g wonders why his keyboard just switched to US layout10:08
greybackalan_g: I am. I see your thread, I'll add my 2cents shortly10:08
RAOFEOD for me.10:11
alan_gGoodbye RAOF, have a nice evening10:11
tvossRAOF, o/10:15
* duflu isn't sure using one's full arse is in fact better than half thereof10:36
dufluAnd on that thought, good night10:36
=== shuduo is now known as shuduo_afk
=== shuduo_afk is now known as shuduo
=== shuduo is now known as shuduo_afk
=== dandrader is now known as dandrader|afk
=== alan_g is now known as alan_g|lunch
=== dandrader|afk is now known as dandrader
=== alan_g|lunch is now known as alan_g
=== dandrader is now known as dandrader|lunch
=== alan_g is now known as alan_g|tea
mterrykgunn, when might we see a new mir drop? (resending, because I think IRC timed out on me)15:32
seb128mterry, one day you need to get a stable internet connection ;-)15:33
kgunnmterry: i'd like to say next week15:33
seb128mterry, or an IRC proxy15:33
mterryseb128, naw, anything important is logged  :)15:33
seb128;-)15:33
mterrykgunn, :(  ok15:33
kgunnmterry: we've got wonky integration tests right now on the dev branch that alf is trying to sort thru15:34
kgunnas soon as those stabilize we can update15:34
kgunnmterry: is there a feature on dev you really need pulled thru ?15:34
mterrykgunn, yeah, this is the bug racarr fixed at the sprint, blocking nested mode15:35
mterrykgunn, I can wait15:35
mterrykgunn, just wanted to see it land15:35
kgunnmterry: well...me too15:35
kgunnmterry:  i was ready to promote it this week...but we started seeing flaky ci15:35
kgunnmterry: what's bothering me...we've been chasing for a few days...and its nothing easy...we're starting to think stuff like...toolchain deltas or low level arm diff between calxeda vs n415:36
kgunncalxeda being the big arm server where the build is done15:37
kgunnso its native...but not to the same arm core15:37
kgunnand its the classic...only the auto ci will fail...no local attempts fail15:38
mterrykgunn, yikes15:38
=== alan_g|tea is now known as alan_g
=== dandrader|lunch is now known as dandrader
=== dandrader is now known as dandrader|afk
=== dandrader|afk is now known as dandrader
mterrygreyback, so I've heard that scenegraph will make animating stuff in USC easier.  Is that because I'll be more easily able to use Qt stuff or what?17:07
=== chihchun is now known as chihchun_afk
greybackmterry: essentially with QML, any application surface will be added to the QML context as an Item. So you can animate it with the usual QML animation system17:09
mterrygreyback, hmm, OK.  And the USC would be able to do the same with session surfaces17:10
greybackmterry: as an example, see http://bazaar.launchpad.net/~gerboland/+junk/qpa-mirserver/view/head:/qml-demo-shell/qml-demo-shell.qml17:10
greybackmterry: when a new surface appears, its x coordinate is animated in. When it is destroyed, it scales to zero.17:10
greybackmterry: yep17:11
greybackmterry:  to USC, the unity8 user session is just a surface17:11
mterrygreyback, awesome.  I want to play with that.  What is rough ETA for landing?17:12
alan_ggreyback: once you start using hardware overlays that won't be true17:13
greybackalan_g: sure, once hardware overlays are happening, QML won't be drawing the surface (hardware will), but it will still manage the surface17:14
greybackmterry: oh January, maybe later. It's in demo condition still. I'm still learning the ins and outs.17:15
alan_ggreyback: usc is in the path to hardware - so the unity8 session becomes a set of surface17:15
=== pete-woods1 is now known as pete-woods
greybackalan_g: sure, but unity8 will decide what surface(s) to be overlaid, send that to USC which will either hardware draw it, or fallback to GL. But that still requires unity8 to manage the surface properties17:17
=== alesage is now known as alesage|brb
mterrygreyback, OK, so not baked enough for me to try and use?17:17
=== alesage|brb is now known as alesage
greybackmterry: let me send you some bits and pieces so you can play.17:18
=== yofel_ is now known as yofel
=== alan_g is now known as alan_g|EOD
kgunnracarr: btw, i meant to tell you earlier... mterry was chatting with me yesterday and was indicating that21:02
kgunnthe whole seperate queue for differing events was going to be more critical sooner that later21:02
kgunnbecause it wasn't just an annoyance of that stale frame showing on screen-on21:02
kgunnmterry cmiiaw...but you're effectively  needing to render prior to the screen coming on21:04
kgunnmterry: and i can't remember...was there some other reason it was going to be critical ?21:05
mterrykgunn, just that it was an ugly enough regression when splitting the greeter that it would likely block landing21:08
mterrykgunn, the problem being that Mir freezes the greeter early in its startup, so it can't queue up what the screen will look like for when the screen does come back on21:08
anpok_hm my connection is getting bad again21:37
=== anpok_ is now known as anpok|afk
kdubkgunn, ping23:07
kdubif its not your eod23:07
kgunnyo23:07
kgunnkdub:23:07
kdubi've wandered above mir in looking for that bug...23:08
kgunnkdub: ok...do we need gerry to pick it up ?23:09
kdubmight be something I want to look at with him23:09
kdubbecause I see in unity-mir the sequence coming off of dbus is on, off, on, on23:10
kduband the double on causes us problems23:10
kduband from powerd's perspective, i just see  on, off, on, off repeated23:10
kdubi don't know what dbus can or cannot do with message ordering23:10
kgunngreyback: you really on ?23:14
greybackkgunn: yes, but only because I forgot to turn off IRC :)23:14
greybackkgunn: what's up?23:15
kgunngreyback: oh, sorry...but if you're tempted ^23:15
kgunnwhat kdub is looking at....2 "on" msgs23:15
kgunncausing screen to stay off23:15
kdubgreyback, could dbus send out of order messages?23:15
greybackkdub: dbus guarantees order of messages23:16
greybackwell as long as there's 1 source & 1 destination23:16
kdubcould it send a message twice?23:17
greybackkdub: so you get "on off on on" all over dbus? From powerd?23:17
kdubyes, from powerd23:17
greybackkdub: nope, never heard of that happening23:17
kdubbut looking through powerd, looks like it should never do that23:17
kdubi'll keep poking around23:17
greybackif you've no luck, mail me before you EOD, and I'll take it up in the morning23:18
kgunngreyback: thank you23:18
kdubalright, thanks23:19
greybackkdub: it's strange as I thought this line in unity-mir would prevent it asking Mir for a double-on:23:20
greyback        if (displayConfigOutput.power_mode != newPowerMode) {23:20
greybackdbusscreen.cpp:6923:20
kdubthat would stop the blank/unblank, but not the compositor stop/start23:21
greybackoh23:34
greybackhmm23:34

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