/srv/irclogs.ubuntu.com/2014/01/08/#ubuntu-mir.txt

RAOFYou know, I wish I could have Mir on one monitor and X on another...00:48
kdubkgunn, that palatable workaround for the n7 works on all my devices, so I posted it for review01:35
=== duflu_ is now known as duflu
kgunnkdub thanks!02:43
=== chihchun_afk is now known as chihchun
RAOFCome for the feature addition, stay for the “hmm, that could really use a refactor” :/03:38
dufluRAOF: You're not selling it03:45
SarvattRAOF: do you work for intel now?04:05
RAOFSarvatt: No, but I'm making sure my work-habits will be compatible if I do :)04:06
Sarvattwho needs to backport anything to a stable release? :P04:06
Sarvattlol04:06
RAOFOh, right. Our cross build support is broken in the presence of pkg-config. Huzzah.04:40
RAOFAh.04:52
RAOFWhich is why we reimplement it for all the libralies.04:52
dufluSpeaking of Intel, keithp is in town and apparently giving a talk on zero-copy compositing this week. Pity I'm not registered, come to think of it05:14
* duflu -> lunch05:15
RAOFIt'll be on the intarwebs.05:18
Sarvatthavent heard about recordings, only livestreams :(05:26
Sarvattotherwise why am i up at 12:30 watching anholts lca talk on http://timvideos.us/05:29
RAOFSarvatt: Oh, is anholt on now?05:53
SarvattRAOF: http://timvideos.us/gentilli05:54
* duflu wonders with so mean livestreams how much work will get done today06:19
duflu(by himself)06:19
duflu-mean +many06:20
* duflu also wonders why "mean" is in muscle memory06:20
dufluUgh, 7ms round-trip on N7 :(07:37
dufluAnd seemingly even longer latency between compositor schedule and composition starting07:38
dufluI blame the kernel 3.1 we have on the N707:38
RAOFWhere does the 7ms round trip come from?07:38
dufluRAOF: mirping07:38
dufluDoing the smallest, lightest, operation I know of07:39
RAOFAlso, damn annoying stupid partial-armhf-mostly-working-stupid-annoynig-annoyance.07:39
RAOF7msec seems huge.07:42
dufluRAOF: It's not terrible. Certainly better than the kernel of only a few years ago. And if we're smart about pipelining then it won't be a big issue07:46
dufluAlthough I suspect we can do significantly better than we do now07:47
dufluBecause it takes at least 2 context switches for an input event to reach the client, and then at least 2 more for the resulting change to reach the screen07:48
RAOFOk. What would it take to requrie an *actual* armhf chroot to cross-build Mir?08:02
anpokduflu: /o\08:28
dufluanpok: ???08:29
anpoka nothing .. just reading through and older geometry::PixelFormat-removal MP08:29
anpoks/and/an08:29
dufluanpok: Well, one type is better than N>108:31
dufluEven if that one type is suboptimal in some use cases08:31
dufluAlthough I don't think that's true right now08:31
anpokah back then those types were two separate enum..08:34
=== chihchun is now known as chihchun_afk
=== chihchun_afk is now known as chihchun
=== chihchun is now known as chihchun_afk
=== dandrader is now known as dandrader|afk
=== dandrader|afk is now known as dandrader
=== alan_g is now known as alan_g|afk
=== alan_g|afk is now known as alan_g|lunch
=== alan_g|lunch is now known as alan_g
=== dandrader is now known as dandrader|lunch
didrockskgunn: Mir releasing as we speak FYI :)15:31
kgunndidrocks: thanks much! a nice coordinated effort once more :)15:31
didrockskgunn: yeah! ;)15:31
=== alan_g is now known as alan_g|tea
=== alan_g|tea is now known as alan_g
* ogra_ hugs didrocks 16:05
ogra_(and hopes it wont take 5 weeks anymore in the future with the new CI infra to get a commit show up in the image)16:06
* didrocks hugs ogra_ back16:07
didrocksogra_: well, as soon as we finished the pilot programm, I'm sure the Mir components will be in it16:07
didrocks(and so platform-api as well)16:07
ogra_yay16:07
ogra_bug 1258056 had its actual commit into devel on dec. 6th ... its just hilarious the process takes so long16:08
ubot5bug 1258056 in Mir "nested mir on android fails on galaxy nexus" [High,Fix released] https://launchpad.net/bugs/125805616:08
ogra_such a fix shouldnt take more than a week or two16:08
=== dandrader|lunch is now known as dandrader
mterry_ricmm, so image 119 will have new Mir!  That means the libhybris patch should be the last thing blocking nested mode17:22
ogra_mterry_, didnt the hybris fixes land long ago ?17:49
ogra_i thought the Mir stuff was the last bit17:50
* ogra_ is actually waiting with your branch in hands since monday to finally land it :)17:50
ogra_(tapping my feet)17:51
mterry_ogra_, tell me about it17:57
ogra_heh17:57
mterry_ogra_, to my knowledge, the hybris fix is still waiting to land in trusty.  Not sure why exactly17:57
ogra_hmpf17:57
mterry_ricmm, what's the blocker to landing the hybris fix?17:57
ogra_do you know who handles that ?17:58
mterry_ogra_, ricmm17:58
mterry_ogra_, he had a fix at the sprint in London17:58
ogra_ah17:58
=== alan_g is now known as alan_g|EOD
ricmmmterry_: manpower18:17
ricmmmterry_: today by EOD18:17
ricmmI dodnt have a fix i had the idea of what was wrong, ultimately the fix is a bit different18:18
ricmmjust havent had time to sit down with it, ill do that today18:18
ricmmso ogra can land it all tomorrow18:18
* ogra_ hugs ricmm 18:19
ricmmcan you tell him that when he returns?18:19
ricmmi need to grab some foods18:19
ogra_sure18:19
ricmm(:18:19
ricmmthanks18:19
ogra_mterry_, in case you missed it, ric said he will have the code ready for me so i can land it tomorrow18:38
ogra_(he asked me to forward that to you)18:38
mterry_ogra_, thanks!18:38
mterry_ogra_, I did miss it18:38
ogra_yeah, you seem on and off18:39
mterry_ogra_, yah  :(18:39
ogra_freezing bits in the pipe ? :)18:39
mterry_ogra_, I am trying to install a modified final image zip manually onto my mako, but my changes don't seem to be getting through once I flash.  Do you know any trick to it?  Like, is there an md5sum somewhere I need to update?18:43
ogra_hmm, no, flashing the zip should be enough18:44
ogra_how did you modify it ?18:44
mterry_ogra_, I added some debugging output to a qml file18:46
mterry_ogra_, inside the zip, there was a tar.gz, and I modified the files in that18:46
ogra_byond that its the normal rootfs ?18:46
mterry_ogra_, yeah18:47
ogra_hmm, i think that should work, though not sure, sisnce we use gpg signatures nowadays ... but i'm not sure we do on that level, stgraber should know18:47
ogra_(i would just modify the file on disk though)18:48
mterry_ogra_, problem is I'm trying to debug first boot18:48
mterry_ogra_, and I don't know a better way to get in there.  Do you?18:48
ogra_not really18:49
mterry_Well, I'll put this down for now and test my nested branch with newest Mir stack18:52
ogra_ask stgraber how to better modify the zip18:53
ogra_he surely knows18:53
mterry_ogra_, OK18:53
kgunnkdub: you gonna address alan_g's comment on arbitrary casting of ints to pointers comment on your fix for the snapshot flickers ?21:16
kgunnor should i just top approve?21:17
kdubshould have been addressed21:19
kdubr1317 in that branch21:20
kgunnkdub: ack....my bad (my bad eyes missed it)21:41
anpokdo we have a new ci problem? the last two mp failed for the same unit test22:21
RAOFrobotfuel: What would I need to do to get an armhf schroot on the CI hardware?22:25
kgunnanpok: can you repro at all locally ? (e.g. repeat run an insane # of times)22:25
robotfuelRAOF: The builder?22:25
RAOFrobotfuel: That'd do, I think.22:25
robotfuelRAOF: can you link to the failed job?22:26
RAOFrobotfuel: The basic problem is that setup-partial-armhf-chroot has fundamental limitations that I'm running into.22:26
anpokkgunn: not today.. the n10 refused to charge, I put it on a different wall charger.. currently shows only the battery icon22:26
robotfuelRAOF: we are building arm natively22:27
RAOFrobotfuel: Oh, I don't want access to the builder. I want the android tests to be run in a *real* armhf chroot, rather that the bailing wire and twine that we currently use.22:27
robotfuelRAOF: the android tests are being run on the phones22:27
RAOFrobotfuel: So if I removed cross-compile-chroot.sh from the tree that *wouldn't* break CI?22:27
kgunnanpok: so, there was one bad mako when alf_ was looking into ci troubles before break...when that was taken out of cycle22:28
* RAOF doesn't intend to remove cross-compile-chroot.sh, but boy would it be nice to thorourghly break it.22:28
kgunnit self corrected....it'd be interesting to see if its the same unit running22:28
kgunnthose ci tests22:28
kdubRAOF, i wouldnt object to removing it22:28
robotfuelRAOF: we are building on Calxeda and then running the tests on the phones22:28
RAOFrobotfuel: https://jenkins.qa.ubuntu.com/job/mir-android-trusty-i386-build/565/console suggests otherwise?22:29
kgunnanpok: are you talking about failure on this jenkins run https://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/226/console22:31
anpokyes22:31
robotfuelRAOF: that seems to be new, that is a pbuilder22:31
anpokone was kdubs screenshot fix, the other one was the removal of a android lib in the mir source..22:31
RAOFkdub: I wouldn't actually remove it, but I would replace it with something that uses a real armhf chroot, so we don't have to do all the munging. (And adding a udev dependency dramatically increases the amount of munging required)22:31
anpokso both on the .. android gpu test.. let me look again22:32
anpokndroidGPUDisplay.gpu_display_ok_with_gles22:32
anpokC++ exception with description "error posting with fb device" thrown in the test body.22:32
anpokif I can trust the battery indicator... it is now fully charged..22:34
anpokand starts22:34
anpokwill try22:34
kgunnanpok: just fyi...you can find the specific device name in the console output22:34
kgunnin this case it was device22:35
kgunnmako-04cb53b59854653422:35
robotfuelhttps://jenkins.qa.ubuntu.com/job/mir-mediumtests-trusty-touch/206/console shows that an arm(calxeda) server built the packages22:35
robotfuelRAOF: ^22:35
kgunnalf_:  curious if this is out special friend mako-04cb53b59854653422:35
kdubit just looks flaky to me22:35
kdubtested on my device here, seems it doesnt get to gtest_repeat=1022:36
robotfuelkdub:  it failed on mako-0090f741e3d141bc22:37
robotfueloops kgunn^22:37
RAOFrobotfuel: So, I'm trying to fix all the failures of https://code.launchpad.net/~raof/mir/mockable-udev/+merge/196824/comments/466002 . There is an actual problem there, which I can fix, but I can't fix the first failure in a sane way without requiring a real armhf chroot for that test.22:37
kgunnanpok: on your comment about "pick just one"...i'm confused...what do you mean ?22:38
kgunnanpok: ah...nvmd22:39
kgunnone currently22:39
kgunngot it22:39
robotfuelRAOF: Ah I was looking at the latest failure my bad22:40
robotfuelRAOF: I am not sure about the pbuilder, fginther can get you access22:40
kdubso it looks like something is failing, i can reproduce about 1/10 times on my device with that test case22:41
RAOFrobotfuel: I don't need access (unless it's to persistently change the machine's configuration); I need a change in the build environment.22:41
RAOFrobotfuel: Or, I guess, for us to drop that test run? Is it covering anything not covered by the other tests?22:41
kdubwell, with gtest_repeat it fails, but thats almost expected... let me see how its running on the device...22:42
robotfuelRAOF: that has to be changed by the CI team, I don't know if they are using that pbuilder for other builds.22:43
robotfuelfginther: ping ^22:43
fgintherrobotfuel, RAOF, I was trying to follow, but I don't know exactly what the issue is22:44
RAOFActually, if we've got a bunch of Caldexa nodes, why are we even doing that armhf build on i386?22:44
fgintherRAOF, which build?22:45
fgintherdo you have a link to a job?22:45
RAOFhttp://jenkins.qa.ubuntu.com/job/mir-android-trusty-i386-build/565/console22:45
RAOF(From https://code.launchpad.net/~raof/mir/mockable-udev/+merge/196824/comments/466002)22:45
robotfuelfginther: the mir-android-trusty-i386-build22:46
fgintherRAOF, that's an intentional cross compile job (setup ages ago by mmrazik). If you go to the parent job, you'll see that there is a separate job to run on a calxeda node: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-trusty-armhf-ci/344/22:47
robotfuelRAOF: it looks like vera++ needs to be added to the debian control as a build requirement for it to pass on mir-android-trusty-i386-build22:47
RAOFrobotfuel: And libudev; but setup-partial-armhf-chroot.sh can't work properly with libudev.22:48
robotfuelRAOF: unless we can ignore the error "vera++ not available - disabling make target style_check"22:48
RAOFWe do ignore that warning, yes. :)22:48
RAOF(Because libudev the first library we use that lives in /lib rather than /usr/lib)22:49
RAOFSo, if we're already building and running the tests on android-armhf, is there any value in an explicit cross-compile test run?22:49
robotfuelRAOF: if we are telling people they can use an armhf pbuilder to build armhf builds in the documentation it is worth having. Other than that no.22:50
fgintherRAOF, that would be up to the mir team. If the job is no longer needed, we can remove it.22:50
anpokkdub: how do you run the test cases on the device? mir/partial-armhf-chroot/usr/src/gmock/gtest/include/gtest/internal/gtest-port.h:1340:: pthread_mutex_lock(&mutex_)failed with error 2222:51
RAOFrobotfuel: I'll fix up the cross-compile-chroot.sh to setup and use an armhf schroot, so that'll still work.22:51
anpokthats what I got with current devel.. a few weeks older binaries dont behave like that22:52
kdubanpok, i'm not seeing that with what I just built22:52
anpokthats a failure in gtest22:52
anpoki am obviously doing something horribly wrong22:53
RAOFfginther: Ok. I'll poke mir-devel@ to see if anyone is attached to that particular build.22:53
kdubanpok, if you're cross compiling, try compiling with the script22:53
kduband if it is with the script, try regenerating the partial chroot the script makes22:54
kdubif your cross compile toolchain and the stl on the device get of sync, i've seen that cause problems before22:56
anpokah, thank you.23:00

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