[00:08] Hey guys, it seems the android side is failing. I don't see any Ubuntu Touch UI yet either. Here's the dmesg output (if helpful): http://pastebin.com/YfThfxB2 Here's my rules file 70-beagleboneblack.rules: http://pastebin.com/. I haven't been able to get logcat working yet. Any thoughts? === chriadam|away is now known as chriadam === chihchun is now known as chihchun_afk === vying is now known as Guest77570 === duflu_ is now known as duflu === chihchun_afk is now known as chihchun [06:07] I am a Dev, is there a way to install Ubuntu touch on galaxy tab 10.1? === chihchun is now known as chihchun_afk === chihchun_afk is now known as chihchun === zequence_ is now known as zequence === chihchun is now known as chihchun_afk === chihchun_afk is now known as chihchun [08:41] good morning [09:15] ogra_, hey, we still need to seed Recommends do we? [09:29] Saviq, yup === chriadam is now known as chriadam|away [09:59] Good morning all; happy Chocolate Covered Raisins Day! :-D [10:02] bah, raising ... [10:02] *raisins [10:02] but CHOCOLATE [10:02] ++ === netcurli_ is now known as netcurli === oSoMoN_ is now known as oSoMoN === chihchun is now known as chihchun_afk [10:58] hey tsdgeos === chihchun_afk is now known as chihchun [10:59] MacSlow: hi [10:59] tsdgeos, do you have any idea why https://code.launchpad.net/~macslow/unity-notifications/mark-unsupported-examples/+merge/208113 doesn't get merged? It's approved (and top-approved). [11:00] Morning all [11:00] hey there davmor2 [11:00] MacSlow: because it needs to be train-ci-ed [11:00] i guess [11:01] tsdgeos, hm... ok Saviq ^ ? [11:02] MacSlow, noted [11:03] Saviq, I assume that's not something I need to (can) do [11:03] MacSlow, it doesn't really change much, can we wait until there's meaningful changes? [11:03] MacSlow, yes, we now accrue a set of top-ack'ed branches and land them together in the image, only when they're in the archive can they be merged to trunk [11:04] Saviq, well... I'm still trying to get my stuff in... to be able to mentally tick it off [11:04] MacSlow, ok, but it'll be low on the prio list, ok :) [11:05] Saviq, sure... in that particular case. [11:05] MacSlow, yeah, just because it doesn't change anything visible [11:06] Saviq, "documentation"-wise it does :) [11:06] MacSlow, sure [11:06] MacSlow, but who's reading that ;P [11:06] Saviq, I just thought starting the week with the simpler stuff would be good ;) [11:07] ;) [11:07] Saviq, on that regard... almost nobody is using notifications anymore ;) [11:07] MacSlow, even more, apps won't be able to anyway, only system settings effectively will [11:07] MacSlow, since we're stopping unfocused apps, and focused app shouldn't be using notifications anyway [11:08] Saviq, still have the mind-set that app could... not easy to shake that thinking off [11:09] MacSlow, :) === chihchun is now known as chihchun_afk [11:12] * MacSlow is mildly amused that qt-project.org's doc-search throws 503-errors as Guru-Meditaions at the moment [11:20] MacSlow, do notification bubbles on Touch have icons like they do on PC? I can’t find any reference to notifications at all on . === chihchun_afk is now known as chihchun [11:21] mpt, there are no special symbolic notificatin-icons (e.g. Volume-Up etc) just regular themed icons or image-files [11:21] ok, thanks [11:22] mpt, btw... I stated to document the new ones https://wiki.ubuntu.com/Touch/Notifications [11:23] mpt, but this is very low on priority right now, so it'll take some time until I can finish it [11:23] ok [11:41] Hey guys, I'm still having issues with porting. Here's the dmesg: http://pastebin.com/YfThfxB2, my rules 70-beagleboneblack.rules: http://pastebin.com/0tzu09d1, and the fstab that's created: http://pastebin.com/QCvxrL85 [11:45] hedz09, try running "lxc-start -n android -- /init" [11:45] see what that gives you [11:54] ogra_: lxc-start: failed (98) to create the command service point /var/lib/lxc/android/command lxc-start: ## lxc-start: # The container appears to be already running! lxc-start: ## lxc-start: failed to initialize the container [11:54] aha [11:54] try: lxc-console -nandroid -t0 [11:55] (might need an extra "enter" once you entered the console to get you a prompt) [11:56] ogra_: hmm, it responds back that android is not running :/ [11:57] so try: lxc-stop -n android -k [11:57] then try rthe lxc-start again [11:58] you can also try: ps ax|grep /init [11:58] ogra_: when I try to stop it it says that android is not running, and the lxc-start says the same thing. [11:58] and see if there is a /init (which is the android one) besides the /sbin/init (which is ubuntus upstart) [11:58] thats weird [12:00] ogra_: ps ax|grep /init says: /init, /sbin/init and lxc-start -n android -- /init [12:00] ok, thats fine then [12:00] i dont get why it is not properly starting the container then :/ [12:00] this android build is against the ubuntu android tree, right ? [12:01] you didnt just grab "some android for the BBB" but compiled it with all ubuntu chnages [12:03] ogra_: what I did was run phablet-dev-bootstrap to retrieve the code, put my CM device/ti/beagleboneblack contents in there, then did lunch and make [12:03] ok [12:03] that should be fine [12:03] i really dont get why it doesnt find /system [12:04] [ 5.939781] initrd: mounting /root/var/lib/lxc/android/system.img as /root/android/system [12:04] it should clearly be there [12:07] hedz09, can you chack your init.rc in your android tree to see if vold is properly turned off ? [12:08] there should be a line like: "#setprop vold.post_fs_data_done 1" [12:08] (with a hash sign in front) [12:12] ogra_: yes, it's there with the hash sign in front [12:13] hmm [12:13] hedz09, lets wait for sergiuens and rsalveti to wake up ... probably they have an idea [12:14] ogra_: Ok, sure. Is it possible there's a permission issue? Also, I don't see any Ubuntu Touch UI yet on my LCD (not sure if there should be, I'm just seeing terminal commands) [12:14] you cant, it needs the container [12:14] ogra_: ahh, gotcha. [12:15] it needs the binary 3D drivers from there [12:15] else it wont be able to display anything [12:15] oh, one last thing [12:15] mount|grep loop1 [12:15] see what that returns [12:15] /dev/loop1 on /android/system type ext4 (ro,relatime,data=ordered) [12:15] /dev/loop1 on /lib/modules type ext4 (ro,relatime,data=ordered) [12:15] /dev/loop1 on /var/lib/lxc/android/rootfs/system type ext4 (ro,relatime,data=ordered) [12:15] that is what you should see [12:18] ogra_: I'm missing the modules one [12:18] ! [12:19] hedz09, you should have all your kernel modules in the android system.img [12:19] ls -l /system/lib/modules/ [12:19] see if they are there [12:19] else make sure they get there (in android) === MacSlow is now known as MacSlow|lunch [12:20] ogra_: Aha! Nothings there [12:20] ok, you need to make sure they are there when building your android side [12:21] probably there is something modularized in your kernel that prevents the container from starting properly [12:22] ogra_: right, that is very true [12:25] * rickspencer3 powers up to image 250 on his phone [12:27] rickspencer3: good luck. hope its fantastic enough for you [12:27] asac, so I'm rocking Qt5.2, right? [12:27] is anyone else getting this when running an autopilot test on image 250? [12:27] File "/usr/lib/python2.7/dist-packages/autopilot/application/_launcher.py", line 23, in [12:27] from gi.repository import GLib, UpstartAppLaunch [12:27] ImportError: No module named gi.repository [12:27] I never saw gi.repository before [12:28] rickspencer3, so you do :) [12:28] ogra_, sweeet! [12:29] rickspencer3: ack :) [12:29] rickspencer3: do you feel it already? [12:29] now I patiently wait for cjwatson's libclick and all related changes, so I don't have to be patient waiting for apps anymore, right? [12:29] yeah [12:29] running 256 here ... apps start a lot faster ... (still 3sec or so, but not 5+ anymore) [12:29] asac, well, the first app I ran was one of my own, where I read in a 10k line text file and create a local storage database [12:30] there is still room for improvement [12:30] but the speedup is noticeable [12:30] rickspencer3: that one broke? [12:30] asac, it's not broken, just terribly coded [12:30] :) [12:30] take a course :) [12:30] at start up I do a very long operation [12:30] i heard popey gives them at times ;) [12:30] ogra_, I have a fix, just need to implement it! [12:31] ah [12:31] switch to u1db! [12:31] I just have to write a program to generate the qml file from the text file I have [12:32] * ogra_ wishes all bootcharts would look like this http://people.canonical.com/~ogra/touch-bootcharts/ubuntu-phablet-trusty-255.png [12:32] app startup> yeah, the rest is all on the toolkit/app side, I believe [12:32] yeah [12:32] and most likely also our disk IO with loop devices [12:32] unless there's something in the scope, but that should be cached [12:33] t1mp: should be fixed by https://launchpad.net/ubuntu/+source/autopilot/1.4+14.04.20140319.1-0ubuntu2 [12:33] cjwatson, well, we see weird aa-click crashes ... we were suspecting that upoad to be at fault [12:34] how could an autopilot upload that adds a couple of deps break aa-click? :) [12:34] cjwatson, apparmor.click.AppArmorException: "Could not find '/usr/share/autopilot-touch/apparmor/click.rules'" [12:35] thast the error in the .crash file [12:35] cjwatson: thanks [12:35] https://jenkins.qa.ubuntu.com/job/trusty-touch-mako-smoke-daily/175/artifact/clientlogs/ubuntu_calculator_app/_usr_bin_aa-clickhook.0.crash/*view*/ [12:35] as an example crash [12:35] ogra_: seems a stretch to connect that to that autopilot upload [12:36] well, autopilot-touch is the package shipping /usr/share/autopilot-touch/apparmor/click.rules [12:36] I understand that [12:36] But http://launchpadlibrarian.net/170373621/autopilot_1.4%2B14.04.20140319.1-0ubuntu1_1.4%2B14.04.20140319.1-0ubuntu2.diff.gz [12:36] yes [12:36] i looked that that already [12:37] sadly we dont have any test results prior to 255 ... and that exposes this crash too [12:37] (and has the autopilot change) [12:37] so its is really hard to tell when exactly the issue started [12:39] apparmor itself had its last upload before 250 was tested, so it cant be that [12:42] Saviq: so are the events that we would like to continue being processed being dispatched in the UI thread? [12:42] e.g. UI mainloop [12:42] asac, all of them that reach it, in the shell case [12:42] asac, it's not only about apps [12:43] ogra_: could it be the *previous* autopilot change? [12:43] * ogra_ checks [12:44] that was friday 3am ... we should have had it in 250 as well [12:44] http://people.canonical.com/~ogra/touch-image-stats/250.changes ... [12:45] yeah, that was in when testing 250 === _salem is now known as salem_ [12:49] hmm, and on my 256 install i clearly see '/usr/share/autopilot-touch/apparmor/click.rules' [12:49] so the file is definitely there [12:52] and i can even run the failing command manually on a freshly bootstrapped install [12:52] (and it imports the file fine) [12:54] ogra_: ok, so I put the modules in the android system.img, but the result is the same. /system/lib/modules/3.8.13-bone41 has contents, but mount | grep "loop1" still doesn't show the modules line [12:55] anything in dmesg about them ? [12:58] ogra_: only where it shows the virtua kernel memory layout: http://pastebin.com/PyChBmNW [12:59] hedz09, btw, would you mind using paste.ubuntu.com (or another pastebin) my browser goes crazy with these falsh movies all the time [12:59] *flash [13:00] not sure who at pastebin.com decided its a good idea to plaster pastes with movie ads all around them [13:01] ogra_: certainly http://paste.ubuntu.com/7146185/ (me neither) [13:03] are the phabletutils python libs also deprecated, like phablet-flash? [13:04] if yes, are there new libs we can get somewhere? [13:05] t1mp, nope, they arent === alan_g is now known as alan_g|lunch [13:06] ok [13:07] hedz09, thanks a lot ! ... and yeah, no messages about /lib/modules ... i fear we need rsalveti or sergiuens then ... i'm a bit out of ideas [13:09] ogra_: ok, sure :). When do they usually come online? [13:09] they are both on american timezones [13:09] (and might be off today, not sure) [13:10] I just ran into this bug https://bugs.launchpad.net/ubuntu/+source/phablet-tools/+bug/1284612 [13:11] Ubuntu bug 1284612 in phablet-tools (Ubuntu) "phablet-config writable-image -p foobar doesn't apt-get update" [Undecided,Confirmed] [13:11] ogra_: who maintains phablet-tools? === MacSlow|lunch is now known as MacSlow [13:11] t1mp, sergio [13:12] t1mp, we dont want to run apt-get update from any scripts on purpose [13:12] ogra_: ahh ok. Hopefully they are not off today, I'm expected to demo it tomorrow night (it's part of my final project) [13:12] that can confuse our test tools since they expect certain versions that correspond to the snapshot of the archive the image represents [13:13] ogra_: why not? now phablet-config fails which makes it useless [13:13] t1mp, it think we could add an additional switch or so [13:13] ogra_: hmm. I just tried to use phablet-config to install some autopilot tests on a fresh 250 image, and it gave me a 404 for most of the packages [13:13] but by default it should only try to install the versions that were recent when the image was built [13:14] fi the tests dont match the AP version in the image you will get breakage [13:15] ogra_: that means that the only promoted qt52 image that we have cannot be used to run app autopilot tests [13:15] you can manually run apt-get update ... not sure whats the problem with that [13:15] (but ude to the design of autopolit you might get wrong results then) [13:15] *due [13:15] or you could add an arbitrary empty PPA :-) [13:16] yeah [13:16] if args.package_dir or args.ppa: [13:16] adb.shell('apt-get update -qq') [13:16] we have a script to run all app tests (before happroving a ui-toolkit MR) on device, and the script relies on phablet-config to work [13:17] I understand now why you don't want it to update automatically [13:17] t1mp, doe that script come from the Qa team ? [13:17] *does [13:17] perhaps a switch for phablet-config to call update is indeed a good idea [13:17] ogra_: no it doesn't [13:17] they usually keep things in sync across the test infra and enduser tools [13:18] which is why we have a policy that only their tools should eb used for tests [13:18] afaik doanac has a workitem to provide such a script to the UITK team [13:18] mzanetti, ping [13:18] please talk to him [13:18] ogra_: then I spend two hours installing packages and manually starting AP tests every time I do an MR [13:18] ogra_: oh that's cool, didn't know that [13:19] doanac: hello [13:19] Saviq, regarding https://code.launchpad.net/~macslow/unity8/modal-snap-decisions/+merge/210988/comments/500423 ... [13:19] balloons: hey [13:19] fi its not done yet, it should probably get a bump in prio :) [13:20] mzanetti, Carla and I ran into a fun issue on Friday. Can you look at something? https://code.launchpad.net/~carla-sella/reminders-app/notebook-list-test/+merge/212251 [13:20] Saviq, you want to see a QSortFilterProxyModel be added to NotificationModel and have it expose that said snapDecision-count? [13:21] mzanetti, so if you grab her branch and run test_dump you'll see that the local qml files changes made for the test (added an objectname) don't exist in the tree. Meaning, it's not using the local qml files [13:21] However they were built locally and run locally. It's a bit confusing [13:21] This is a relevant snippet: http://paste.ubuntu.com/7132723/ [13:22] running the same binary the same way with autopilot launch and vis shows you the proper objectname. So, this might not be a question for you but I'll poke you anyway and see if you see anything [13:22] thansk mzanetti [13:23] balloons: didn't we add the parameter to explicitly tell which qml file to load? [13:23] are you using that? [13:25] mzanetti, yes we are using the -q. If I launch using autopilot launch with a locally built binary, and give it -q to the local qml file I see everything correctly in the dbus tree. In the test we are trying to launch the same way, but don't see the local modifications. [13:26] balloons: searching that branch for "-q" doesn't give any results [13:28] MacSlow, why do we need it in QML? [13:29] MacSlow, ah [13:29] MacSlow, yes, do, that's what I said in the comment :) [13:29] Saviq, because that's where the special "model background" is drawn or not [13:29] Saviq, s/model/modal [13:30] MacSlow, yeah, the other solution would be to count them in the backend and expose the prop [13:30] Saviq, just doing that... only wished I didn't ahve to touch the backend [13:30] MacSlow, but that's probably fine [13:30] MacSlow, well, you don't need to [13:30] mzanetti, inside __init__.py is launch_test_local. that's showing you how it launches. Regardless, since it works outside the test with AP, but not inside, it's an AP issue or something simpler than that, hah [13:30] MacSlow, we do have a SortFilterProxyModel in QML [13:31] Saviq, I've not found anything on the QML-side regarding taht.. [13:31] MacSlow, qsortfilterproxymodelqml.h in unity8 [13:31] mzanetti, it is given the local binary, -q with the local qml [13:31] Saviq, that's a special unity8-thing?! [13:31] MacSlow, it's not exposed to QML otherwise, no [13:31] Saviq, ah that explains it [13:32] MacSlow: Saviq: btw as you're referring to it, I made a uitk component based on that https://code.launchpad.net/~kalikiana/ubuntu-ui-toolkit/sortFilterModel/+merge/211945 and api doc https://docs.google.com/a/canonical.com/document/d/1B0yHmPr5yOHri3zZLtRba7eYhTR_K374zZAdED7xgy8/edit [13:32] any input/ review welcome [13:33] kalikiana, cool,stuff [13:33] s/,/ / [13:33] kalikiana, I barely know about it... not sure I can be of help yet [13:36] MacSlow: you want to use it right? the api gets twice as good for every review based on how useful it is with no regard for how it was written [13:37] kalikiana, first let me actually fix the issue mentioned in the review-comment that brought it up :) [13:38] MacSlow: sure, no rush. just review from somebody using app before it exists is oddly enough a luxury for the uitk [13:38] *api [13:40] didrocks: I think I can at least confirm that this madness with click's symlinks for the apparmor hook being thoroughly confused by different versions of applications doesn't seem to be a recent regression; I'm beginning to get a handle on it and the same logic was around even before libclick [13:40] FWIW [13:40] kalikiana, well I'm "using" the ComboButton already :) [13:40] cjwatson, you mean the carshers ? [13:40] cjwatson: thanks for confirming and keeping posted :) [13:40] yeah, unlikely thats libclick [13:41] *crashers [13:41] ogra_: I have no idea whether the crashers are related to what I'm investigating; that's not my starting point [13:41] ah, k [13:41] Kaleo: hi! I'd like to try to fix https://bugs.launchpad.net/ubuntu-ui-toolkit/+bug/1296728 because it's rather bad for us; do you have any hint of what could be possibly go wrong? [13:41] Ubuntu bug 1296728 in Ubuntu UI Toolkit "Buttons in other windows appear broken" [Undecided,New] [13:42] I'm investigating the problem sergiusens reported with the wrong versions of apps being linked into /var/lib/apparmor/clicks/ - probably also what Sam Bull reported on the list [13:42] (it is mine :) ) [13:42] ah, k [13:42] well, i'm still trying to find out why aa-click doesnt fint its profile file [13:42] (which makes *all* click tests fail) [13:43] but there is no obvious reason :/ [13:47] ogra_: well, I suppose it could be the same thing, does the relevant link in /var/lib/apparmor/clicks/ point to the right place? [13:47] ogra_: the symptom of the bug I'm investigating is that there are links there pointing to old user-installed versions in /opt/click.ubuntu.com/ when they should really be pointing to versions in /usr/share/click/preinstalled/ [13:48] i dont know, the infrastructure isnt that clever to list that anywhere [13:48] what i see is: [13:48] aa-clickhook -f --include=/usr/share/autopilot-touch/apparmor/click.rules [13:49] this is run at some point [13:49] and it claims /usr/share/autopilot-touch/apparmor/click.rules does not exist at all [13:49] which is clearly a lie [13:50] ogra_, doanac: is there a bug about that planned test script? we've been using ci@home in uitk for a while now and gotten some people curious on re-using it [13:51] kalikiana, well, when i asked about a month ago to not use self written scripts ev did put developing a script that works the same as the test infra on doanac's todo ... i thought this was long given to you [13:51] since we cant really rely on home written scripts that do their own thing for something as critical as tests [13:52] we need to be sure they all use the same setup and environment [13:52] else the testing is completely useless [13:52] ogra_: mind you the default expectation is humans running commands by hand [13:52] not sure what ci@home is [13:52] kalikiana, right [13:52] which is what we all do currently [13:53] so you're suggesting a script maintained within the team is worse than that? [13:53] except for the automated infra ... and if we want automation it needs to be identical [13:53] kalikiana, if the env isnt identical to the test infra, the tests are pretty useless, yes [13:53] ogra_: so I tried modifying the 'touch' file in the ubuntu ramdisk. I changed [ -e ${rootmnt}/android/system/lib/modules ] && mount --bind ${rootmnt}/android/system/lib/modules ${rootmnt}/lib/modules to just mount --bind ${rootmnt}/android/system/lib/modules ${rootmnt}/lib/modules, then in 'touch' I just do ls mounts but there's nothing there for modules, still. [13:53] its as good as guessing then [13:54] hedz09, but it is definitely inside your android system.img file ? [13:55] ogra_: but you do realise whatever humans do is always up to chance unlike a script that reproduces any assumptions consistently [13:55] https://code.launchpad.net/~fboucault/+junk/ciathome [13:55] in any case, it would be very useful to work together not isolated :-) [13:55] kalikiana, right, thats the reason QA was asked to provide such a script ... we discussed that at lenght with UITK people around and asac, didrocks etc involved [13:55] since they own the official testing === alan_g|lunch is now known as alan_g [13:56] ogra_: it does make sense to me that QA is involved if not owning it, but if florian hadn't made a few hours of his time we'd all be weeks behind in work done [13:56] ogra_: yes, it is. And when I log into phablet, ls /system/lib/modules/3.8.13-bone41/ has modules.dep, etc. [13:57] kalikiana, well, i really dont want to discuss this again, if you guys insist to do your own testing feel free to get approval from asac ... [13:58] kalikiana, it was discussed in a hour long discussion at the ci training with your team [13:58] ogra_: no we don't, that's why i'm asking to collaborate instead of waiting without talking [13:58] kalikiana, if QA didnt deliver the right script to you guys yet, please poke them [13:59] kalikiana, collaborate in what ? [13:59] in creating an out of bound second test tool ? [14:00] no, in creating a test tool - there's no first to begin with [14:00] kalikiana, there is [14:00] QA uses it in the testing infra we all rely on [14:00] and there was a request from your team to split that out into a tool you guys can run [14:01] which i forwarded on bzoltan's behalf to ev [14:01] and to my knowledge doanac is working on it [14:01] exactly we need something that is available. since it isn't we're forced to improvise [14:01] creating something additionally just doesnt make sense [14:03] kalikiana, so yes, please collaborate ... with doanac ;) [14:04] Kaleo as well as bzoltan know that QA is the owner of this [14:04] sure. in the same way uitk is the owner of qml components that apps may improvise if they're not available ;-) [14:05] I know both sides of the story [14:06] well, except that if one core app builds improvised stuff that doesnt work, it wnt affect the whole image and all other apps ... [14:06] yes it will. uitk tests explode if the apps do stupid things [14:06] if you get false positives in an UITK test that has the potential to kill the world [14:07] we have tons of delays due to hacks [14:07] or even unmaintained apps like notes [14:08] anyway, again, i spent multiple hours on exactly the same discussion about 1 month ago, i really dont want to have it again, there was a solution, please talk to asac, bzoltan, ev and Kaleo [14:09] kalikiana: what is ci@home? [14:10] asac: it flashes the device if it's not running the latest and then installs and runs all core apps ap tests on a given merge request [14:11] kalikiana: does it use phablet-test-run? [14:11] kalikiana: do you have a link? [14:11] asac: no, ubuntu-device-flash [14:11] t1mp: that thing now alsy runs tests? [14:11] t1mp: well, both [14:11] (not I didnt say phablet-flash :)) [14:11] asac: https://code.launchpad.net/~fboucault/+junk/ciathome [14:11] kalikiana: yes, true. === chihchun is now known as chihchun_afk [14:12] asac: yes, I am noting that now (I didn't before) :) [14:12] kalikiana: did you get 100% success on fridays build reliable with that? [14:12] asac: I did stumble on some failures that's why right now I'm doing a "harmess branch" as t1mp is also to see if the image is good [14:13] (though we did have some discussions on the failures last week, we weren't just waintg) [14:14] asac: ci@home is not meant to be a permanent solution, but it is all we have now (besides flashing, installing packages, and starting all the app tests manually) [14:14] doanac: maybe check their ciathome thing above [14:14] doanac: i remember you wanted to get a "standard tool" that would run a batch of tests with flashing etc. [14:14] similar to infra [14:16] it was expected from the start that it'd only be temporary anyway - but I think somewhere communication got stuck so that was the main reason for the above discussion [14:16] ogra_: does /var/lib/lxc/android/pre-start.sh run? I'm trying to modify the script for testing (to see how far it gets), by adding echos, but I don't see anything === greyback is now known as greyback|lunch [14:16] kalikiana, yeah, i think doanac was dragged into some important other stuff before he could finish [14:16] kalikiana: yeah, collaborate with doanac on this [14:16] ensure he embraces what you have or replaces it [14:17] hedz09, yes, it gets run by lxc-start [14:18] rsalveti, yo ... do you have any idea for hedz09 ? he tries to port to a beaglebone black, we got it to boot into the ubuntu rootfs but it seems to not be ablet to start the container ... all bits and pieces seem to be there [14:18] and i'm slightly out of ideas [14:19] init.rc complains about /system/bin/* not being found [14:22] ogra_: /system/bin/* not found probably means issues when creating the android system.img [14:23] usually you can validate if the container is working correctly by booting into android directly [14:23] if it works, then the same image can be used by the container [14:23] then all you need to do is making sure you have the needed kernel options enabled (for pid namespace and so on) [14:27] ogra_: not sure if this gives us a clue...but if I run pre-start after logging into phablet I get this: http://paste.ubuntu.com/7146569/ [14:28] hedz09, no, that wont help since you are missing variables that öxc-start sets [14:28] *lxc-start [14:28] rsalveti, well, /init starts === dandrader is now known as dandrader|afk [14:29] ogra_: right, but init is part of the android initrd [14:29] rsalveti, http://paste.ubuntu.com/7146185/ see line 434 and later [14:30] and /system is populated on the ubuntu side, so i would expect it to be available in the android side as well [14:30] depends how you're booting it [14:30] cdimage or system [14:30] system [14:30] [ 15.169616] init: cannot find '/system/bin/servicemanager', disabling 'servicemanager' [14:30] right [14:30] yeah, can't find the system partition/mount point [14:30] for all binaries [14:31] but if you look above [14:31] it properly mounts the android system.img in the initrd [14:31] 5.946209] initrd: mounting /root/var/lib/lxc/android/system.img as /root/android/system [14:32] hedz09: can you access these files from adb? [14:32] rsalveti, adb doesnt run since the container doesnt start ... he is logged in via serial tty (its a beagle) [14:33] right [14:33] but yeah, /system is populated [14:33] and looks fine [14:33] rsalveti: what ogra_ siad [14:33] (we're on this the whole weekend already) [14:35] hedz09: do you have /android/system/boot/android-ramdisk.img ? [14:36] rsalveti: yes [14:37] then you should also have a bunch of bind mounted partitions under $LXC_ROOTFS_PATH [14:37] which in theory is what the container would be able to see [14:37] ogra_: wonder if there's a new kernel option or similar that makes the container unable to see a mount point mounted by the host [14:37] $LXC_ROOTFS_PATH is /var/lib/lxc/android/rootfs [14:37] rsalveti, hmm, yeah ... that would be one explanation [14:38] i really dont see why else it would fail [14:39] this is a 3.8 based kernel [14:40] not that new either [14:40] maybe pinging stgraber === chihchun_afk is now known as chihchun [14:40] well, even the 3.0 one on the maguro still works === pete-woods is now known as pete-woods-lunch [14:41] i doubt the old kernel is the issue ... a missing config option could be though [14:41] I wonder if a newer kernel is the issue :-) [14:41] we only tested <= 3.4 [14:41] oh [14:41] heh [14:42] rsalveti: yes, the contents of /var/lib/lxc/android/rootfs is :is acct, cache, config, d, data, default.prop, dev, etc, fstab.am335xevm, init, init.am335xevm.rc, init.am335xevm.usb.rc, init.goldfish.rc, init.rc, init.trace.rc, mnt, proc, sbin, sdcard, socket, storage, sys, system, ueventd.am335xevm.rc, ueventd.goldfish.rc, ueventd.rc, usbdrive, vendor [14:42] hedz09: can you check the contents of /var/lib/lxc/android/rootfs/system ? [14:43] hedz09, chack if system has any content there [14:44] rsalveti: , ogra_ , yes it has bin, boot, xbin, etc, lib, build.prop, lost+found [14:44] cjwatson, Heh, couldn't figure out why the url-dispatcher tests were failing, libclick broke them! :-) (a good kind of break) [14:45] tedg: oh? [14:45] cjwatson, Not a big deal, just using a substitute command line for the test. Need to switch to the test click DB. [14:45] hedz09, if you call /system/bin/logcat -d ... what is the exact error ? === dandrader|afk is now known as dandrader [14:46] tedg: hm, do I need to add url-dispatcher to my list of things to libclickify, though? I hadn't noticed it was using click [14:46] tedg: or are you already doing that? [14:46]