/srv/irclogs.ubuntu.com/2016/02/16/#snappy.txt

wiglewormanyone online03:09
=== ruenoak1 is now known as ruenoak
=== chihchun is now known as chihchun_afk
elopiowigleworm: more or less. About to leave05:17
elopiopitti: I think I found my problem. I'm creating a keypair for scalingstack, and then I need to pass the private key to ssh -i. But the nova script tries to connect to ssh without specifying the private key, so it times out.05:19
elopioI have solved it for now putting the private key in ~/.ssh/config.05:19
elopionow I need tons of squid exceptions, but it's running the whole suite.05:20
pittielopio: right, you need to upload your ssh key to nova with keypair-create06:21
pittielopio: if you only have one key in "nova keypair-list", then the setup script will use that, otherwise you have to add a --keyname option to chose the one you want06:22
=== chihchun_afk is now known as chihchun
dholbachgood morning07:37
didrockshey dholbach!07:40
dholbachsalut didrocks07:41
=== kickinz1|afk is now known as kickinz1
fgimenezgood morning08:12
ysionneauany idea why I'm not able to ping anything from snappy? name resolution works, and I was able to snappy install nethack-armhf.ogra so network seems approximately ok08:46
ysionneauis there some telnet client that I can install?08:46
didrocksogra_: hey! stupid question, but how is the grub.cfg taken from the gadget snap in 16.04? I don't see any bindmount or anything to it09:16
longsleepHey snappy, so what is the best strategy to convert a snap for 15.04 to one for 16.04?09:37
JamesTaitGood morning all; happy Tuesday, and happy Innovation Day! 😃09:40
popeyhELLO!09:52
popeyDo I need to jump through hoops to make wifi work on the Pi2 snappy images or will it "just work"?09:52
didrockspopey: depends on the version :)09:53
didrockspopey: which one are you trying?09:53
popeythe one from mvo's home09:54
popeybuilt 4/2/1609:54
didrocksok, so 16.04, you just need to create the wlan0 file then09:54
didrocksuno momento09:54
didrockspopey: http://www.marinus.nu/2015/02/enabling-wifi-on-snappy-ubuntu-core.html09:55
didrockspopey: from "Create a config file for your wlan0 interface:09:55
didrocks"09:55
didrocks(ignore the above)09:55
didrockspopey: I don't have a wifi module, so didn't test myself09:55
popeyhm, okay09:57
* popey plugs into ethernet for now (I assume that works OOTB)?09:57
didrockspopey: yeah, it does09:59
didrockspopey: I even share my laptop wifi connection through ethernet here09:59
popeydidrocks: on wired, does it do dhcp by default or do I need to plug a physical kb/screen in before I can use it?10:01
popey(I don't see it listed on my dhcp server)10:01
didrockspopey: it does dhcp client by default10:02
popeyhmmm10:02
didrockspopey: tried rebooting (unsure it has hotplug support though, even if it's auto eth0)?10:03
popeyhad to plug it into a telly and ifconfig to find out the ip10:05
popeyfor some reason it's not showing up on my dhcp server, but it has an ip10:05
popeymost strange10:05
popeyubuntu@localhost:~$10:05
popey\o/10:05
didrocksgood! :)10:05
didrocks(and phew, "just" have to blame your dhcp server :p)10:05
popeythis command requires root access. Please re-run using 'sudo'10:06
popeythat should have a capital letter at the start10:06
popey#justsaying10:06
popey 😃10:06
popeyugh, all messages in errors.go are badly capitalised. My eyes!10:10
didrockspopey: yeah, I was looking at the same. I'm unsure this is on purpose (like combining with "Error:") or something else10:10
popeyit can't be intentional, or if it is, it's inconsistent10:11
popey"this command requires root access. Please re-run using 'sudo'" - "this" lowercase "t", "Please" uppercase "P".10:12
didrockswell, you have a full stop in between10:12
didrocksthe consistency is in all sentences are starting in this file with no capital10:13
didrockslike if they wanted to prepend "Error: {}"10:13
didrocksso the result would be: "Error: this command requires root access. Please re-run using 'sudo'"10:13
didrocks(which is consistent)10:13
popeyokay. still looks odd (and is inconsistent with other things like the kernel). I'll not worry about that  ðŸ˜ƒ10:17
popey:( failed in my first step in the guide. trying to "snappy install docker hello-world"10:19
popeydocker failed to install: can not open /tmp/docker963014042: cannot open snap: unknown header: "!<arch>\ndebian-binar"10:19
didrockspopey: yeah, that's the current issue with 16.04…10:20
didrockspopey: it's all moving right now, and the store isn't yet snap-all compatible (it shows you old snaps that are still ar-based)10:20
popeyah, okay.10:21
popeyLiving life on the edge :)10:22
ogra_didrocks, i think u-d-f puts it in place10:32
didrocksogra_: ah, so it's a copy when generating the image, but what if you update the gadget snap thus?10:32
popeyWhat's the replacement for "snappy search"?10:33
ogra_popey, snap find10:33
ogra_didrocks, you have to ask mvo if he added cod to copy it then (i havent seen any)10:34
ogra_*code10:34
popeythat's not listed in the --help :(10:34
popeyand snappy find doesn't work10:34
didrockspopey: *snap*10:34
popeyoh, snap?10:34
didrocksnew snappy name from what I know10:34
ogra_yeah, because we like to be intuitive ;)10:34
didrockseven if I never saw anything written down10:34
popeyso snappy -> apt-get, snap -> click/pkcon?10:35
popeykinda10:35
didrocksno, I think snappy will move to snap, right?10:35
didrocksand commands will transition one after another?10:36
didrocks(ogra_ probably knows more than I here, I just saw this command being mentioned once)10:36
ogra_i have no idea ... it might be documented in some google doc though10:37
* ogra_ also only picked it up from IRC conversations10:37
ogra_i assume we want to put all operations you can do directly to a snap into that command ... though looking at the --help output "find" somehow doesnt fit there10:38
didrockszyga: you probably know more than us there then ^10:39
didrockszyga: snap vs snappy command (will they eventually merge into one?)10:39
popeydholbach: https://developer.ubuntu.com/en/snappy/build-apps/get-started/ says i should install snappy-tools to get snapcraft, I had to manually install snapcraft itself, it wasn't pulled in by snappy-tools10:40
popeydholbach: want me to file a bug?10:40
zygapopey: hey10:47
zygaer10:47
zygadidrocks: hey :)10:47
zygadidrocks: snap is the new version of snappy command10:47
zygadidrocks: snap works over the REST api and talks to snapd10:47
zygadidrocks: snappy is the old command that has everything inside10:47
zygadidrocks: snappy will be removed before 16.0410:48
zygadidrocks: and we'll just have one10:48
didrocksI was right \o/10:48
didrocksthanks for confirming zyga :)10:48
ogra_ah, good to see a proper summary, thanks zyga10:49
zygaogra_: lots of things are in snap today10:49
zygaogra_: but daily builds are broken so we don't see that10:49
ogra_?10:49
ogra_daily builds of what ?10:49
zygaogra_: and daily builds are broken because of launchpad git model choking on gpg signed git commits10:49
zygaogra_: of snappy itself10:49
* ogra_ didnt see any build failures10:49
zygaogra_: mvo found this yesterday or last week10:50
ogra_oh, it fails before it even attempts to build ?10:50
zygaogra_: we have a backup plan to just do manual git pull github; git push launchpad10:50
zygaogra_: yes, the import is stuck now10:50
ogra_ah, that explains why i see no failures :)10:50
zygaogra_: I'm sure we'll untangle this as the sprint develop10:50
ogra_right10:50
popeywhen's that?10:50
ogra_this week10:51
zygapopey: well, this week10:51
popeyheh10:51
zygapopey: ask mvo for details10:51
popeyno need, just wondered :)10:51
zygapopey: we could do a plan-B in 30 minutes or so10:51
* popey sets expectations accordingly :)10:51
zygapopey: but I wanted to sync with mvo to see who's going to do that10:51
* ogra_ actually prefers outdated but working over landed but broken (and nobody there to fix it) state10:52
zygaogra_: daily is broken too :)10:53
zygaogra_: but differently, no executable seems to run today10:53
ogra_i can use it here10:53
zygaogra_: as in real master, not the outdated build from lp10:53
ogra_right, so just leave LP as is until everyone is back ;)10:53
zygaogra_: do you have any updates on dragon kernel/gadget?10:57
ogra_zyga, done, i just need to roll an image and push it up10:58
zygaogra_: nice, I can offer simple testing on my board to iron our works-for-me issues :)10:58
ogra_it doesnt resize yet, and systemd spills a few rfkill messages on boot ... the rest should be all fixed10:59
ogra_(i ripped resize out completely for now so it doesnt break on big SD cards anymore)10:59
ogra_sudo ./ubuntu-device-flash --verbose core rolling --channel edge -o dragon-all-snap-test.img --gadget canonical-dragon.canonical --kernel canonical-dragon-linux --os ubuntu-core.canonical11:00
zygaogra_: yeah I've seen those11:00
zygaogra_: oh that a nice touch, thanks11:00
ogra_(with mvo's u-d-f binary)11:00
zygaogra_: so essentially, ubuntu-image dragon :-)11:00
* zyga rebuilds11:00
zygaogra_: building now, I'll let you know when it boot11:02
ogra_good11:02
zygaflashing11:08
zygaI should write the flash tool11:12
zygaall I want is a freeling progress bar11:12
zygaogra_: for the first time I can also measure power usage of my dragon board, if you ever want to experiment with some tweaks to the kernel, I can give that a go11:14
ogra_zyga, there is godd, that has a progressbar11:16
zygaoh, thanks, I'll give it a try next11:16
* ogra_ doesnt want no shitty progressbar ... i just want it to flash 10x as fast instead :P11:17
zygaogra_: booting now11:20
zygaogra_: flash over usb3.011:20
zygaogra_: on a speedy card11:20
ogra_i do ... still takes ages11:22
davmor2ogra_: b.....b....but without progress bars you wouldn't have a break to go make coffee11:22
ogra_davmor2, true indeed :)11:22
zygaogra_: boot done11:23
zygaogra_: is the mac address stable now?11:23
ogra_yep11:23
zygaogra_: perfect, let's update my mac tables11:23
zygaogra_: wifi operates too11:23
ogra_i still need to make the code a bit more generic, but the function is fine11:23
zygaogra_: so +1 from me, works like a charm11:23
* zyga posted https://github.com/ubuntu-core/snappy/pull/490 and https://github.com/ubuntu-core/snappy/pull/48911:58
popeyIs there any prospect of snapcraft 2.x being backported to trusty?12:00
popeythe ppa still has snapcraft 1.x12:00
ogra_no, 2.x is solely for xenial12:02
ogra_use a xenial chroot, container or use the classic dimension on your xenial snappy install12:03
popeythats what i am doing12:03
popeybut my laptop is 14.0412:03
popeyso i have a difference between them12:03
ogra_between the chroots ?12:03
popeywill create a xenial chroot on my laptop, ta12:03
ogra_ah12:03
popeyno, between 14.04 on my laptop and xenial on my pi12:03
ogra_yeah12:03
popeynvm, it's cool :)12:03
ogra_well, there is no cross building ... so after all you want to use the classic dimension on your pi to get armhf packages12:04
popeyyeah, i was building on my laptop then copied the yaml over to the pi in classic and found the syntax differed12:06
popeyof both the yaml and snapcraft itself12:06
popeywhich is what made me look at the versions12:06
popeywoot, made a snap12:06
ogra_yay12:07
ogra_ooooh !12:07
* ogra_ found his next snap 12:07
ogra_https://github.com/coolwanglu/BrowserHack12:07
ogra_(i already packaged nethack, but this is even cooler)12:09
popeybah, can't find some perl libs12:12
ogra_add them to stage-packages12:13
popeyis it a comma separated list? stage-packages: [foo,bar,baz] ?12:15
ogra_http://paste.ubuntu.com/15090658/12:16
ogra_thats a snapcraft 2.1 file though ... the security stuff is different now12:16
popeyok12:16
* popey will play12:17
kyrofaGood morning12:20
* zyga finds classic dimension fantastic and uses it for daily work12:27
* ogra_ misses it on arm64 :P12:27
didrockshey kyrofa!12:44
kyrofaHey didrocks, how's your week so far?12:45
didrockskyrofa: it's quite good, thanks! Working on the developer website, but can't wait to be able to dive into the more technical aspect of snappy once 16.04 settles down :)12:45
didrocks(and be able to work more closely with you guys)12:46
kyrofadidrocks, awesome! Well we look forward to that, too :)12:46
kyrofadidrocks, how many times have you looked at a Snappy question on, say, stack overflow, become completely confused, only to realize that it's Google's snappy compression lib?12:48
noizerHi the system variable for the snaps path?13:23
didrockskyrofa: ah, that didn't happen yet (sorry, was on a meeting) :p13:35
didrockskyrofa: for me, it was more snappy ubuntu core vs ubuntu core ;)13:36
didrocks(like our old named ubuntu core)13:36
kyrofadidrocks, ah, right13:36
kyrofadidrocks, well, keep an eye out if your subscriptions span more than askubuntu13:37
didrockskyrofa: will do, even just for the sake of fun :p13:37
kyrofadidrocks, :D13:37
dholbachpopey, I'll file a bug13:46
ogra_kyrofa, pfft ... compression lib ....13:51
ogra_ogra@styx:~/Devel/mycroft$ apt-cache show snappy|grep Description13:51
ogra_Description-en: Powerful media player with a minimalistic interface13:51
ogra_:P13:51
kyrofaogra_, wait... :P13:51
ogra_hmm, we dont ship hdparm in the rootfs ... i wonder if we should13:56
popeydholbach: okay14:00
popeyhm, adding the perl-modules package to my stage-packages: has no effect, it doesn't get bundled into the snap14:02
ogra_in a copy plugin ?14:02
popeyah14:04
ogra_(not sure how stage-packages behave in other plugins, i always ever used them in that context )14:05
popeyis the "files" section documented somewhere?14:06
ogra_well, "source: destination"14:07
ogra_(relative paths)14:07
popeydoes it not assume same path14:11
popey?14:11
ogra_only for the toplevel14:13
ogra_aha, found the issue with rfkill .... systemd wants a dir /var/lib/systemd/rfkill and wants that writable14:15
ogra_werid though, i thought we had added that to /etc/system-image/writable-paths before ... but seems we didnt14:17
zygaogra_: nice14:20
zygaogra_: that's going to fix rfkill for all systems, right?14:20
ogra_well, it is going to quieten the systemd error .... not sure thats enough to fix the low level14:22
=== chihchun is now known as chihchun_afk
elopiopitti: yes, I create the keypair and pass it with --keyname. The thing is that if the keypair I create doesn't have the same private key as the entry in ~/.ssh/config there's no way to tell nova which private key it should use.14:44
elopioogra_: happy birthday!!!!14:44
ogra_elopio, thanks !14:44
=== chihchun_afk is now known as chihchun
popeyOooh! Congratulations ogra_ for clinging to this rock for another solar orbit. Well done you!14:45
ogra_haha, thansk :)14:46
ogra_*thanks even14:46
pittihey ogra_, happy birthday! *hug*14:47
ogra_thanks !14:47
kyrofaAh, happy birthday ogra_!14:49
ogra_thanks !14:49
asacppisati: so i chatted to the linaro qemu guy14:49
asacppisati: he said we could try to build the u-boot as the bios image14:49
asacto get a qemu that boots with u-boot14:49
asac15:48 < pm215> build it to expect to start at address 0 with the usual ARM interrupt/exception vector table14:50
asac15:48 < pm215> ie as if you were going to burn the thing into rom14:50
asacppisati: ^^ means anything?14:51
ogra_sound like "dd it to the start of your img file"14:51
elopiofgimenez: I might be 10 minutes late to our meeting. I'm running to get back on time. bbs.14:51
asacogra_: right, but guess thats not all14:53
asacprobably needs more tweaking of init values etc.?14:53
ogra_depends14:53
ogra_hmm ... i wonder if all-snaps supports me sideloading ubuntu-core on a running image (i.e. manual upgrade)14:54
ogra_Installing ubuntu-core_16.04.0-4.arm64_arm64.snap14:56
ogra_ubuntu-core_16.04.0-4.arm64_arm64.snap failed to install: package "ubuntu-core" is already installed with origin "canonical" your origin is "sideload"14:56
ogra_ah, sad14:56
ogra_would have been nice for testing14:56
ppisatiasac: never faffled with uboot internals, so i don't know14:57
ppisatiasac: qemu versatile / vexpress?14:57
fgimenezelopio, ok, no problem14:59
asacppisati: he says we should use "virt" unless we have special requirements14:59
asacnot sure what virt means15:00
Dikkekiphi there, i'm new to the whole docker infrastructure. but i'm looking for a way to make a docker webserver. but i see there are a lot off options out there. what is to be preferred?15:03
Dikkekipcoreos15:03
Dikkekipvs ubuntu15:03
Dikkekipkubernetes vs shipyard15:04
Dikkekipwho wins?15:04
ogra_are you sure you are in the right channel ?15:05
Dikkekipnope15:05
ogra_:)15:05
Dikkekipbut i'm asking the guys who use ubuntu15:05
ogra_sounds like you want to be in #cloud-battle :)15:05
davmor2ogra_: man it seem like only yesterday you were a year younger, Congrats and Happy Birthday dude :)15:08
ogra_davmor2, thanks a lot dude !15:08
asacpitti: so i see a huge number of .service files in /lib/systemd/system/15:59
asacare all those run ?15:59
didrocksasac: you need to look at the *.requires/ or *.wants/ ones16:00
asacdidrocks: so in /lib/systemd/system is the superset of all you can use16:01
didrocksasac: right!16:01
asacand then in .wants and requires is the actual stuff linked?16:01
asaclike mods-available and -enabled for apache?16:01
asackk16:01
didrockssystems ones that you can't disable are in /lib/systemd/system/*.{wants,requires}16:01
didrocksthe ones you are enabling/disabling manually are in /etc/systemd/system/16:01
didrocks(and you can also creates services there)16:02
asacyeah. so i guess we could remove all those that we dont want/require on a snappy system from /lib/system/  and noone would notice?16:02
didrocksofc, you have the socket activated ones in addition to this (and timer, but we don't use them much)16:02
asace.g. cruft?16:02
didrocksasac: yeah, if they are enabled on a system level (in /lib) and we want to mask them, we can symlink with the same file name to /dev/null16:02
didrocks(again, same with path/socket/timer activation)16:03
asaci am not sure16:03
asaci thought everythign not .wants or .requires is not used at all16:03
asacfrom /lib/systemd/system/16:03
asaccan we not remove all of those then?16:03
didrocksnormal services not in .wants or .requires are not activated the regular way (and will never be activated normally)16:04
didrocksunless a socket/path/timer activates them16:04
didrocksasac: if you have docker install, look at /lib/systemd/system/docker.socket16:04
didrocksyou see PartOf=docker.service16:04
didrocksso docker.service can be activated by the docker.socket16:05
didrocksand /etc/systemd/system/sockets.target.wants/ has a link to docker.socket16:05
didrocks(so the socket service is activated via the sockets.target, which in turn, can activate docker.service when it receives something in the socket)16:06
didrockson*16:06
asacdidrocks: do socket activated services get killed if the socket client closes or dies?16:07
didrocksasac: with PartOf= yeah16:07
didrocks(it's double linked IIRC)16:07
didrocksbut the service should exit after a timeout anyway16:08
didrocksit's the upstream recommendation16:08
didrockssome like docker doesn't do it IIRC though)16:08
asacdidrocks: so systemd passes you the socket fd on startup rigth? i read it can also pass you many sockets at the same time16:09
asachow do you know which is for what purpose?16:09
didrocksasac: libsystemd pass an array, let me check back16:10
didrocksasac: yep: http://0pointer.de/blog/projects/socket-activation.html16:11
didrocksn = sd_listen_fds(0);16:11
didrocksso you get the number of fds (generally 1)16:11
asacright i read that doc16:11
asacand it tells it can pass you multiple16:11
asacbut cant figure how someone might then guess which socket is what16:11
asacso think > 1 is just not working16:11
asaclike in the exampleL:16:12
asac        fprintf(stderr, "Too many file descriptors received.\n");16:12
didrockswell, it's just because he expected one, what is blocking you?16:12
asacwonder why it even tries to pass multiple ones... i think there must be some determinism defined what goes to which array bucket16:12
asaci just try to understand that feature :)16:12
asacbecause the doc says youy can get multiple ones16:12
asacbut it doesnt tell me how to ensrue that i alway sget the FD i want at position 916:13
didrocksso, what happens, imagine a service which can take multiple sockets as input/output16:13
didrocksoh, it does16:13
didrocksfd = SD_LISTEN_FDS_START + 816:13
didrocksfor instance :)16:13
didrocksto get position 916:13
didrocksI'm unsure the name of the socket is determinist though, you just get a fd16:13
didrocksso between 2 boots, it may changes as being in position 0 or position 5, or position 8 (if you have 9 of them)16:14
asaci know i can just ask for socket 816:14
asacbut i cant see how that socket 8 will always be the socket i expect it to be16:14
asaclike i can imagine client A talks to socket 1 and client B talks to socket 2 ... how does systemd ensure that its not suddenly the other way around?16:15
didrocksI think it's part of the protocol rather16:15
asacguess some annotations that can be looked up16:15
asacwould help16:15
didrockslike announcing their name16:15
asacpart of the protocol?16:15
didrocksthe protocol you set between your client and services16:16
asacok so you say all need to be of same protocol?16:16
didrocksservice*16:16
didrocksat least in the first bytes16:16
didrocks(that's just a guess)16:16
asacok so restriction is that all sockets are on same port for INET and hence have the same server16:16
didrocksasac: TBH, for systemd-fskd, I just use one socket16:16
asacthat makes more sense16:16
didrocksthen this socket is getting client connection request16:16
asacright. i see16:16
didrocksand depending on this, I create more connections :)16:16
asacits for being able to buffer if multiple requests come in16:17
asacnot to enable activation through different protocols16:17
didrocksso all clients use the same socket to initiate a request16:17
asacyeah16:17
didrocksthen, their are connected through their own sockets that have been given back16:17
didrocks(by the service)16:17
asacdidrocks: and you said that once all sockets are closed the service gets killed by systemd?16:18
didrocksno, sorry, I was unclear, it's when the .socket systemd service is stopped (manually)16:18
didrocksthanks to PartOf=16:18
asacah16:18
asacso there is no deactivation16:19
didrocksI implemented a service timeout when having 0 connection16:19
didrocksand that's what is recommended16:19
asaci guess socket activated services wont get restarted thouygh until there is a client again?16:19
asace.g. its the servers responsibilty to just die if all clients are gone16:19
asacand if a new cient comes systemd will activate again?16:19
didrocksexactly, another client connexion to the socket reactive the service16:19
didrocksright16:20
didrocks(and this is working well)16:20
didrocksalso, the service status is correctly reflected16:20
asacwhat is the service status if i exit a process that got socket activated?16:20
didrocksdepending on the exit return code16:21
didrocksso, basically 0, if you don't set to restart it in the .service file, you are considered as being oneshot16:22
asacdidrocks: does systemd have an API you can use to create nbew services etc. or is the interface "put a unit file there and run reload" ?16:22
didrocksand so, everything is good16:22
didrocksasac: yeah, you need to run systemctl daemon-reload16:23
asacok so no api16:23
didrocksthat way systemd is aware of the new unit files16:23
didrockswell, systemctl is using systemd API16:23
didrocksso, there is one16:23
didrocksunsure how stable it is though16:23
asacright. nothing anyone ever said would be something for use not through CLI16:24
* didrocks needs to run to the bank16:24
didrocksyeah16:24
didrockssee you tomorrow guys :)16:24
asacthansk for giving me inducation didrocks :)16:24
didrocksyw asac ;)16:24
kyrofaelopio, excellent fix on the catkin thing-- I know exactly what the issue is16:26
kyrofaelopio, and I'm sad I didn't catch it in review, I'm sorry :(16:26
elopiokyrofa: well, nobody of us caught it. Tell me more...16:28
kyrofaelopio, so initially the env() function set the PYTHONPATH explicitly using the in-snap python version, and then added more items to it via ROS's setup.sh. However, in the recent refactor for the python version globbing, that explicit set line was moved AFTER the ROS setup.sh call, which wiped out the ROS environment :P16:30
kyrofaelopio, I have no idea how this ever passed. And indeed, I see the failure now on my dev machine. Really odd16:30
kyrofaelopio, anyway, so your change keeps the ROS environment, which makes it work16:31
kyrofaelopio, I suspect now that my working tree wasn't up to date16:32
elopiokyrofa: is there a nicer way to solve it?16:34
kyrofaelopio, no I think your solution is perfect16:34
elopiolucky shot.16:34
kyrofaelopio, all skill ;)16:35
elopioah, yeah, that too ;)16:36
elopiokyrofa: hey, the problem with this autopkgtest job in jenkins is that it will take like an hour to run.16:36
elopiodo you think that's too much?16:36
kyrofaelopio, per PR?16:36
elopioyes.16:37
kyrofaelopio, that's certainly long, and I can see it being frustrating if we really want to crank something out, but can we honestly say that ensuring quality takes too much time?16:37
pcocaHi everyone, Is there any way to deploy snaps to many devices at the same time? (with snappy remote or other tool) or needs to be done with the store?16:38
elopiokyrofa: we can make the rule to wait for it only on the changelog PR. Ignore it for other pull requests we want to merge faster.16:39
elopioand I think I can run the integration and the examples tests in parallel, somehow. I'll look into it.16:39
sergiusenspcoca, snappy remote with a "store" name should just work16:39
pcocasergiusens, then all the devices pointing at that store will get the snap?16:40
sergiusenspcoca, real features for multi deploys will come with landscape integrating with the rest api16:40
sergiusenspcoca, yes to your question16:41
pcocasergiusens, thanks! :)16:41
kyrofaelopio, honestly most PRs sit for an hour or so anyway, I'm fine with it. sergiusens may have an opinion (morning sergiusens!)16:42
sergiusensmorning16:43
sergiusensI don't mind PRs taking an hour16:43
sergiusensbut do we have to use the main upstream or will it work in our forks?16:43
kyrofasergiusens, good question16:44
kyrofaOh elopio, I accepted your PR but I suppose the ROS test should have been enabled in debian/tests/exampletests16:57
elopiokyrofa: maybe. Let me first confirm it runs in jenkins.16:57
elopiowe are still missing some squid rules16:57
kyrofaelopio, alright16:57
jerryGogra: can you answer a question about coredumps ?16:58
kyrofajerryG, were you ever able to enable them?16:58
ogra_jerryG, not sure, try it :)16:59
kyrofajerryG, https://github.com/ubuntu-core/snapcraft/blob/master/docs/debug.md if you haven't seen it16:59
jerryGkyrofa: ogra_: I'm having trouble getting coredumps to work.... it's creating folders... but it's not creating any coredump files17:00
jerryGkyrofa: thanks. that's what I was using as a guide17:00
ogra_jerryG, well, we dont do anything special despite using the kernel defaults ... perhaps you need to tweak something17:01
kyrofajerryG, I just did it the other day doing that (though using "snaps" instead of "apps" for all-snaps)17:01
jerryGkyrofa: so /home/ubuntu/snaps ?17:01
ogra_there is also nothing in userspace that should touch the defaults either17:01
kyrofajerryG, indeed. Were you using apps?17:01
kyrofajerryG, that's a bug in the documentation17:02
jerryGkyrofa: ok thanks.  ill try that17:02
zygakyrofa: hey, have you seen pad.lv/1545786 by any chance?17:02
kyrofa(carry-over from 15.04)17:02
zygakyrofa: elopio tells me you touched the launcher lately17:02
zygakyrofa: I'm trying to see what's causing that error to happen17:02
kyrofazyga, no I haven't, thanks for the heads up!17:03
zygakyrofa: I'm running raspberry pi with latest snappy master spliced into an image I built yesterday17:03
zygakyrofa: nothing I install runs17:03
zygakyrofa: I cannot find any matches on the error message so far17:03
kyrofazyga, is the user data directory owned by root or anything?17:03
kyrofazyga, it's an errno17:04
zygakyrofa: which directory is that? I did see root owned directories in /var/lib/snap17:05
zygakyrofa: older (prebuilt) were root:ubuntu17:05
zygakyrofa: post-install changed were root:root17:05
zygakyrofa: chgrp on that doens't help17:06
kyrofazyga, no, user data directory ($SNAP_USER_DATA)17:06
zygakyrofa: no, those are all ubuntu:ubuntu17:07
kyrofazyga, if you can walk me through how to get on edge I can look into this. I tried the other day but it seemed that channel-changing was broken17:07
zygakyrofa: (in fact, nothing in $HOME is owned by root)17:07
zygakyrofa: I built an image using my small tool: https://github.com/zyga/devtools17:08
zygakyrofa: https://github.com/zyga/devtools/blob/master/ubuntu-image17:08
zygakyrofa: essentially ubuntu-device-flash from mvo and some defaults for pi217:08
kyrofazyga, alright, let me check it out :)17:08
zygakyrofa: cool, I'll try to strace the error in the meantime17:09
* zyga has a bunch of straces to look at17:11
zygakyrofa: got it:17:13
zygaopen("/", O_RDONLY|O_DIRECTORY|O_NOFOLLOW|O_CLOEXEC) = -1 EACCES (Permission denied)17:13
zygawrite(2, "failed to create user data direc"..., 36) = 3617:13
zygawrite(2, ". errmsg: Permission denied\n", 28) = 2817:13
zygakyrofa: that is ubuntu-core-launcher17:13
zygaelopio: ^^17:13
kyrofazyga, huh... why would that be permission denied?17:13
zygajdstrand: ^^17:13
zygatyhicks: ^^17:14
zygano idea17:14
ogra_/ is readonly ...17:14
kyrofaogra_, it's not trying to write to it. It's trying to get the fd17:14
* ogra_ guesses thats just fallout 17:14
ogra_oh17:14
zygaogra_: no, I don't think you'd get EACCES for O_RDONLY open on /17:14
ogra_is it ?17:14
kyrofaogra_, it uses that fd with a series of openat/mkdirat calls17:14
jdstrandzyga: read on '/' isn't allowed by the policy17:15
ogra_and it doesnt use the actual path ?17:15
kyrofajdstrand, but it's done before the app is confined17:15
zygaogra_: I guess using openat means it has to get the fd for / first17:15
zygajdstrand, tyhicks: for context, we're trying to understand https://bugs.launchpad.net/snappy/+bug/154578617:16
ubottuLaunchpad bug 1545786 in Snappy "error trying to execute snap binary" [Critical,Triaged]17:16
jdstrandkyrofa: did '/' get wrong permissions? is the mount wrong?17:16
ogra_it is readonly17:16
kyrofajdstrand, good question. zyga ?17:16
ogra_since it is a squashfs17:16
jdstrand'failed to create user data directory. errmsg: Permission denied'17:16
ogra_but that shouldnt get you a permission denied i guess17:16
jdstrand'user data directory'17:17
zygaubuntu@localhost:~$ mount | grep '/ '17:17
zyga/dev/loop0 on / type squashfs (ro,relatime)17:17
zygais a loop-mounted squashfs from the os snap17:17
zyga(well, it is the os snap)17:17
ogra_right17:17
zygajdstrand: I straced that to ubuntu-core-launcher17:17
jdstrand'ubuntu@localhost:~$ hello-world.echo'17:17
jdstrandzyga: yes17:17
jdstrandit tries to create the user data directory17:17
jdstrandas of recent commits (I guess they landed)17:18
zygajdstrand: hello-world.echo behaves the same way (just installed it and tried the command)17:18
jdstrandtyhicks: might be able to provide some more insight on openat, but the security team is currently working on a critical update17:18
jdstrandzyga: I get that. what I'm saying is that if run as a user, it is possible that $HOME/snaps is already there with root permissions. that is a different bug17:19
jdstrandso checks the perms of that17:19
zygajdstrand: $HOME/snaps and inside has nothing owned by root (checked now)17:20
jdstrandok, good (it isn't that bug)17:20
zygajdstrand: everything is ug+rwX,o=rX17:20
jdstrandzyga: I would guess it has something to do with the mount options17:21
zygajdstrand: mount optionon /writable or / ?17:21
zyga*optioins17:21
zyga*options17:21
jdstrandI'm not sure17:22
jdstrandare there apparmor denials in syslog?17:22
jdstrandthe launcher runs under its own profile. I wonder if it needs an update with the recent changes17:22
tyhicksI can't think of what would cause that failed open()17:22
tyhicksthat is really odd17:22
jdstrandzyga: grep DEN /var/log/syslog17:23
kyrofatyhicks, indeed17:23
tyhicksah, I guess it could be the ubuntu-core-launcher AppArmor profile blocking it17:23
zygahttp://pastebin.ubuntu.com/15094280/17:23
zygajdstrand: ^^17:24
jdstrandok, there you go17:24
tyhicksyeah17:24
jdstrandlet me look at the rest of the profile and I'll commit something to the tree17:24
zygajdstrand: thanks, if there is something I can tweak in my image in the end to unblock on this that would help me a lot; if not I'll just wait for the bugfix to be available and if needed, update/rebuild the image17:25
kyrofajdstrand, tyhicks I'm sorry guys, I'm not sure how I didn't discover that in testing17:25
jdstrandzyga: so, to work around this, you'll have to add '/ r,' to the profile17:25
zygajdstrand: I assume the new os snap is needed to fix this?17:25
zygajdstrand: ah, easy enough, I'll try17:25
jdstrandzyga: well, you can't tweak in the image since it is read only, but we can test what is needed without that17:25
jdstrandzyga: it is17:25
zygahmmm, right17:25
jdstrandzyga: do this: cp /etc/apparmor.d/usr.bin.ubuntu-core-launcher /tmp17:26
jdstrandzyga: then modify /tmp/usr.bin.ubuntu-core-launcher to have:17:26
jdstrandr,17:26
jdstranderr17:26
jdstrand / r,17:26
zygadone17:27
jdstrandthen do:17:27
zygabind mount over?17:27
jdstrandsudo apparmor_parser -r /tmp/usr.bin.ubuntu-core-launcher17:27
jdstrandno don't worry about that17:27
zygaah17:27
zygaright, this is in-kernel17:27
jdstrandwe can load the profile into the kernel from anywhere17:27
kyrofajdstrand, ah, and that profile is path-specific? If the launcher was run from elsewhere it wouldn't be applied?17:27
jdstrandkyrofa: in the profile there is '/usr/bin/ubuntu-core-launcher ... {'17:28
kyrofajdstrand, argh, that was my testing gap17:28
jdstrandkyrofa: that gives what we call binary attach17:28
jdstrandkyrofa: I can actually test this here17:29
zygajdstrand: done, still same error, inspecting new set of strace logs17:29
jdstrandzyga: if you'd like to go back to sprinting, I'll assign this to myself17:29
jdstrandzyga: I suspect there are other similar accesses that are required17:29
jdstrandso I can work through those17:30
zygaopenat(3, "home", O_RDONLY|O_DIRECTORY|O_NOFOLLOW|O_CLOEXEC) = -1 EACCES (Permission denied)17:30
jdstrandyep17:30
zygait seems to need to go all the way down to $HOME/snaps17:30
jdstrandI'll take it from here17:30
* jdstrand nods17:30
zygajdstrand: actually, I'm not sprinting (no remote sprinting this time)17:30
zygajdstrand: so I can focus on this issue to unblock myself from skill security work17:30
zygajdstrand: I'll make a coffee and be back soon,17:31
kyrofazyga, trying to use ubuntu-image for the pc I get "expected 3 partitons but found 0"17:31
jdstrandzyga: to unblock yourself, in addition to '/ r,', add '/**/ r,'17:31
jdstrandzyga: I'm going to do something more specific, but that should unblock you17:31
zygakyrofa: oh? weird, maybe you are affected by kpartx bug? (I assume that's at the very end when it builds the image)17:32
zygakyrofa: I'm running xenial daily and it works okay17:32
kyrofajdstrand, I suppose /, /home/ and /root17:32
zygajdstrand: perfect, thanks17:32
zygakyrofa: and $HOME/snaps too in the end17:32
jdstrandkyrofa: yes, but I wanted to sprinkle some 'owner' matches in there too17:32
* zyga starts to grok all the security parts involved even if they are still complex17:32
kyrofajdstrand, ah right17:33
kyrofazyga, hmm... I'm on xenial, though I haven't updated for a while17:33
elopiojdstrand: kyrofa: zyga: we are working on a process to release ubuntu-core to edge, and run the full suite there before moving it to other channels. That should find this type of errors earlier.17:33
kyrofaelopio, very cool!17:34
jdstrandelopio: sounds excellent :)17:34
zygagreat elopio17:35
=== kickinz1 is now known as kickinz1|afk
zygajdstrand: the workaround works, thank you :)17:37
jdstrandzyga: cool :)17:38
zygawooooot, my demo works :)17:38
zygavery close to i2c through skills17:38
zygano longer running as root :D17:38
zygawhee :D17:38
elopiozyga: \o/17:44
sergiusensutlemming, https://bugs.launchpad.net/cloud-init/+bug/154623017:49
ubottuLaunchpad bug 1546230 in cloud-init "change the default to the nocloud data source" [Undecided,New]17:49
jdstrandzyga: what are you testing this on? I don't seem to have any updates for my os snap in kvm17:53
kyrofajdstrand, edge17:56
=== chihchun is now known as chihchun_afk
jdstrandzyga: hmm, I'm on core/rolling with ubuntu-core 16.04.0-11. how did you install 'edge'18:06
popeyIf I am using snapcraft and want to wrap my executable thing in a shell script to add environment variables like setting PERL5LIB to SNAP_APP_DATA_PATH etc, how do I tell snapcraft to "install" my shell script?18:10
kyrofajdstrand:18:10
kyrofakyrofa> zyga, if you can walk me through how to get on edge I can look into this. I tried the other day but it seemed that channel-changing was broken18:10
kyrofa<zyga> kyrofa: (in fact, nothing in $HOME is owned by root)18:10
kyrofa<zyga> kyrofa: I built an image using my small tool: https://github.com/zyga/devtools18:10
kyrofa<zyga> kyrofa: https://github.com/zyga/devtools/blob/master/ubuntu-image18:10
kyrofa<zyga> kyrofa: essentially ubuntu-device-flash from mvo and some defaults for pi218:10
zygajdstrand: using ^^18:10
zygajdstrand: it's magic18:10
zyganobody knows how to use ubuntu-device-flash :D18:11
kyrofapopey, using the copy plugin18:11
popeyah18:11
kyrofazyga, indeed! That tool got super confusing18:12
zygaI'm sure we'll see real ubuntu-image sometime soon but before that happens, I've just do what I do18:12
zygadone, well anyway18:13
* zyga sleepy18:13
kyrofazyga, seems my kpartx is up-to-date... so I don't know what's going on with my lack of image creation abilities18:14
jdstrandzyga: well, there are so many different versions of udf :)18:14
ogra_there is only one true one !18:14
ogra_(the latest local binary on people.u.c from the person that implemented the last feature indeed !!!)18:15
* kyrofa hears ogra_ now: "The one that creates dragonboard images!"18:15
kyrofaogra_, hahaha18:15
zygajdstrand: yeah, I'm using the build mvo makes18:15
ogra_:)18:16
zygakyrofa: hmm, if you want I can make an image for you18:16
zygakyrofa: but otherwise no idea18:16
kyrofazyga, it sounds like jdstrand is on it, but I may take you up on that offer next time I need to test something on edge if that's okay ;)18:17
kyrofazyga, hopefully the channel-switching bug will be fixed at some point as well18:17
jdstrandI'm running ubuntu-image now. I'm expecting I'm fine18:17
* zyga wonders if the security team is busy with http://people.canonical.com/~ubuntu-security/cve/2015/CVE-2015-7547.html18:18
ubottu** RESERVED ** This candidate has been reserved by an organization or individual that will use it when announcing a new security problem.  When the candidate has been publicized, the details for this candidate will be provided. (http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2015-7547)18:18
jdstrandits fun that the kernel snap is like 2/3 bigger than the os snap18:18
jdstrandzyga: yes18:19
ogra_jdstrand, all that firmware ...18:19
zygaah, man, security is hard work! thanks for working on this18:19
ogra_kernel and modules alone would be smaller18:19
jdstrandI'll say "you're welcome" on behalf of the security team (/me hasn't yet done his part for that update)18:20
jdstrandzyga (and kyrofa): ok, have an edge image working. thanks18:28
kyrofajdstrand, ah, so it worked for you? /me pouts18:29
jdstrandkyrofa: all I did was ./ubuntu-image18:30
jdstrandthen it asked me some stuff and out popped a pc.img that worked18:30
kyrofazyga, I'm running it in an lxc. Think that would break things?18:30
jdstrandkyrofa: I've occasionally found that rebooting helps18:30
jdstrandkyrofa: eg, if you've used udf a few times in a row, etc18:31
jdstrandI think I opened a bug on that18:31
zygakyrofa: perhaps yes18:31
zygakyrofa: actually qemu is pretty nice18:31
* jdstrand did not run it under lxc18:31
zygakyrofa: just run xenial in qemu to build the image18:31
ogra_yay18:42
* ogra_ has a working resize script for the dragonboard ... now i just need to pull that into the initrd18:43
ogra_and since i'll have to touch the initrd package for that i'll make asac happy tomorrow and simply also add the code for a generic binary build :)18:43
ogra_zyga, do you have random USB devices you could try on the dragonboard to see if they work (i tried the three USB ethernet adapters i have here but lack more stuff)18:46
ogra_would be good to know they load their respective modules at least and dont throw errors in dmesg18:46
zygaogra_: yes, sure18:50
ogra_thanks18:50
zygaogra_: trying now18:50
zygaogra_: tried wifi dongle, eth dongle18:57
zygaogra_: (eth dongle flaky but that's the dongle)18:57
zygaogra_: trying more18:57
ogra_both load modules and all ?18:57
zygaogra_: yes (both work as network interfaces)18:58
jdstrandzyga: fyi, uploaded ubuntu-core-launcher to xenial and the image ppa for the fix18:58
ogra_awesome !18:58
zygaogra_: just FYI, I still see this while bootinh18:58
ogra_zyga, "this" ?19:00
zyga[FAILED] Failed to start Load/Save RF Kill Switch Status.19:00
zyga(had to reboot to see it, damn serial)19:00
zygaminicom over screen over classic mode pi over ssh over ssh19:01
zygadon't ask ;)19:01
zygahttps://xkcd.com/908/19:01
zygaapplies19:01
ogra_zyga, yeah, that wont be fixed before tomorrow ... needs the fix uploaded first (i havent yet) ... then a cdimage build ... and then manually snapping that and uploading to the store19:01
ogra_(the rfkill stuff)19:02
ogra_i was planning to do a full set of ubuntu-core snap updates tomorrow anyway19:04
ogra_also to test the arm64 update (mvo claimed it should work, but nobody has tested it yet :) )19:05
zygacool19:07
zygaI'll gladly update tomorrow19:07
ogra_it should even do that itself :)19:07
ogra_(unless you disabled autoupgrade)19:08
sergiusenskyrofa, elopio how are things progressing on your side?19:30
kyrofasergiusens, not too bad here. How's the sprint so far?19:32
svijogra_: häppy börsday!19:37
sergiusenskyrofa, good; looking into ubuntu image more than anything else19:40
kyrofasergiusens, ah good, that will be super useful :)19:41
sergiusensand trying to wrap up cleanbuild a bit during spare time19:41
kyrofasergiusens, you have spare time? Not 8-8 meetings eh?19:41
mvoogra_: sure it will work ;)19:43
sergiusenskyrofa, 8-8 meetings; spare time is me getting distracted or during the night ;-)19:45
kyrofasergiusens, hahaha19:45
kyrofaogra_, were you ever able to fix the dragonboard partition-resize-eating-itself issue?19:48
popeyHow long to snap reviews take in the store?23:08
* popey is used to clicks being reviewed in about a minute23:08
zygapopey: I may be wrong but reviews might be manual until squashfs is supported by review scripts23:12
popeyah okay23:14
popeywell, I successfully uploaded my first snap to the store, so that's a milestone for me  ðŸ˜ƒ23:14
=== cprov_ is now known as cprov
=== ev_ is now known as ev
=== tedg_ is now known as tedg
=== alecu_ is now known as alecu
=== Trevinho_ is now known as Trevinho
=== moul_ is now known as moul
=== pitti is now known as Guest82926
=== stgraber_ is now known as stgraber
sergiusensstgraber, hey about https://bugs.launchpad.net/snappy/+bug/1543764 do you have implementation details on how you plan to do this?23:57
ubottuLaunchpad bug 1543764 in Snappy "snappy classic must use officially supported lxd images from simplestream; not unsupported ones from linux-containers.org" [High,New]23:57

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