IRC logs for #baserock for Wednesday, 2014-11-12

radiofreei'm at odds with this unified patch series00:29
radiofreeon one hand, apparently, the current release has the same issue with weston-clickdot (but only under weston-ivi-shell, it works fine in desktop-shell)00:30
radiofreebut there's an issue with weston-simple-egl in *both* shells (that is, if you abort it brings down everything graphical)00:31
radiofreeso it's a combination of kernel/driver upgrades and (maybe?) userspace upgrades00:31
radiofreei'll resubmit the patch series with the 3.17 kernel that at least works as well as the genivi kernel, but will allow us to unify 00:32
radiofreei'll test a bit more tomorrow, but i think that's the best option for a release on friday00:33
*** sambishop [] has quit [Ping timeout: 245 seconds]01:20
*** benbrown_ [] has quit [Ping timeout: 245 seconds]03:37
*** benbrown_ [] has joined #baserock03:37
*** SotK [] has quit [Ping timeout: 240 seconds]03:37
*** SotK [] has joined #baserock03:37
*** aananth [~caananth@] has joined #baserock06:48
paulsher1oodpedroalvarez: you're a star :)07:35
abdulssam2, deployment of wandboard image is completed08:13
*** wdutch [] has joined #baserock08:13
paulsher1oodabdul: w00t! :)08:27
*** sambishop [] has joined #baserock08:41
*** tiagogomes [] has joined #baserock08:45
*** mariaderidder [] has joined #baserock08:59
*** tiagogomes [] has quit [Ping timeout: 250 seconds]09:06
paulsher1oodwould anyone know why deploy might hang on 'populating "orig" subvolume'?09:12
paulsher1ood25 minutes so far09:12
franredwhich cluster are you deploying?09:13
*** aananth [~caananth@] has quit [Remote host closed the connection]09:14
*** tiagogomes [] has joined #baserock09:21
*** jonathanmaw [] has joined #baserock09:26
pedroalvarezpaulsher1ood: 25 minutes  is too long, has it finished?09:30
*** mauricemoss_ [] has joined #baserock09:33
*** franred [] has quit [Remote host closed the connection]09:45
*** franred [] has joined #baserock09:50
paulsher1oodi've repeated this several times. i fear something has changed09:50
*** aananth [~caananth@] has joined #baserock09:51
pedroalvarezthe way that this works is:09:53
pedroalvarez1) It "clones" doing a btrfs snapshot of the orig version of the default system09:53
pedroalvarez2) It rsync-s the new system to the created snapshot09:53
pedroalvarezmaybe rsync is going slow due that everything has changes09:53
paulsher1oodi doubt it's that, tbh.09:54
aananthI am following "".  My question is w.r.t command "ssh git@trove group adduser local-config-admins adminuser". I don't see any username "git" in /etc/passwd/ or /etc/shadow/09:54
paulsher1oodbut i'll prove my theory soon09:54
aananthHow do I verify if the deployment is correct?09:54
pedroalvarezpaulsher1ood: yeah, but is my first theory09:54
Kinnisonaananth: If there's no 'git' in /etc/passwd then it sounds like the Trove ansible setup hasn't run09:54
Kinnisonpedroalvarez: ^^^09:54
pedroalvarezaananth: Kinnison is right, can you paste your /var/log/ansible file?09:55
aananthOk, let me do that...09:56
pedroalvarezpaulsher1ood: your theory is?09:56
paulsher1oodplease hold :)09:57
aananthpedroalvarez: Please find the contents in
*** Krin [] has joined #baserock09:58
pedroalvarezaananth: I think your trove may not have internet connection. Is that right?09:59
pedroalvarezaananth: can you `ping` ?09:59
aananthYes it is behind the proxy.10:00
pedroalvarezaananth: have you ever managed to put a baserock machine running behind that proxy? In that case, can you do the same for this trove?10:01
pedroalvarezaananth: the command that is failing to run is `ssh-keyscan localhost`10:03
aananthpedroalvarez:  Yes, for devel-Virtual Machine. W.r.t ping, I can ping the ip address of baserock, but not the text.
* pedroalvarez ponders a trove deployment troubleshooting wiki page10:03
aananthLet me do similar set up on trove as well.10:04
pedroalvarezaananth: aah! I understand. How do you fixed it? adding to resolv.conf?10:04
aananthpedroalvarez: I have not fixed it, but just found the ip address of and typed it manually to check if the internet connections exist.10:05
pedroalvarezaananth: yeah, I meant, how did you fix it for other machines.10:06
aananthpedroalvarez: we install a tool named "corkscrew", we configure it and also configure git, ssh to execute corkscrew.10:08
aananthpedroalvarez: In order to add configs for proxy, shouldn't I add my username as the first step using "ssh git@trove group adduser local-config-admins adminuser"?10:09
aananthAre you saying if the proxy issue is resolved, I will be able to do the first step?10:10
pedroalvarezNow I understand the real problem  here10:12
franredjjardon, is your patch-series: "Add weston system containing latest stock weston" ready for reviews?10:13
pedroalvarezIs not possible to configure a proxy before the trove has internet connection, since the trove uses the internet to configure itself10:13
jjardonfranred: it is10:14
aananthpedroalvarez: Ok, thanks. Let me first setup corkscrew and configure ssh to tunnel through our proxy first.10:14
franredjjardon, thanks, we will review it, sorry for the delay10:15
pedroalvarezaananth: can we try another thing? can you add "nameserver" to your /etc/resolv.conf file,  then `systemctl restart trove-setup`, and then send me the /var/log/ansible file again?10:16
aananthpedroalvarez: OK. Let me do the above first.10:26
aananthBut I have almost done the corkscrew setup as well. Let me execute by adding nameserver first.10:26
aananthpedroalvarez: Here is the output -- (this log was after I did a reboot)10:30
pedroalvarezaananth: I thinkg the log is incomplete, maybe it was still running10:33
pedroalvarezI think this time it worked and you can begin with the steps to configure the proxy10:33
* pedroalvarez wonders why he can't reuse the cache that is on
aananthpedroalvarez: I could see I make progress, but I get "Unknown command: useradd" error. Any error in "ssh -i git@vtsc-trove useradd caananth "Aananth C N" """?10:43
pedroalvarezaananth: I believe that "useradd" is not the right command here10:45
pedroalvarezaananth: can you point me to the instructions that you are following/10:45
Kinnisonif you're adding a user to trove, then 1. the command has a space in (i.e. user add) and 2. it's username email realname10:46
Kinnisonso ssh -i git@vtsc-trove user add caananth Aananth C N10:46
Kinnisonmight work better :-)10:47
aananthpedroalvarez: Sorry, I am wrong. I took wrong notes from Sam's training. It was based on, but the command is from my notes.10:49
aananthKinnison: thanks for correcting me. It worked.10:49
Kinnisonaananth: glad I could help10:50
paulsher1oodpedroalvarez: my theory was that latest morph changes might have broken something. however i think i've disproved it.... which brings me to the glibc change, i fear. still investigating10:54
pedroalvarezmorph changes can't explain this behaviour10:54
persiaCould it be a disk space issue?  Performing significant changes to my systems caused me to move from a default 4GB volume to a default 8GB volume10:55
paulsher1oodgood point. i'll try a smaller system10:57
paulsher1ood(as deployment target)10:57
paulsher1oodpedroalvarez: could the glibc thing have increased our sizes significantly?10:57
pedroalvarezI doubt  it :/  Is not a bigger thing, is just a newer version10:58
KinnisonBut it means *everything* got recompiled10:59
Kinnisonand got new version symbols10:59
Kinnisonso every binary is different10:59
pedroalvarezthat was my first theory10:59
*** locallycompact [] has quit [Read error: Connection reset by peer]11:04
aananthpedroalvarez: command "ssh git@vtsc-trove whoami" works perfectly! But the command "git clone ssh://git@vtsc-trove/local-config/lorries lorry-config" fails! Error Message: "[vtsc-trove] CRIT: Repository local-config/lorries does not exist". Any clue why?11:06
Kinnisonyou need your trove ID in there11:06
Kinnisone.g. git clone ssh://git@vtsc-trove/TROVEID/local-config/lorries lorry-config11:06
radiofreepedroalvarez: so the gpu stuff11:07
Kinnisonwhere TROVEID is whatever you configured the trove with11:07
pedroalvarezradiofree: any news?11:07
radiofreethe main issue is crashing clients in *stock* weston (desktop-shell), which is a regression, and important to fix11:07
radiofreehowever i believe that to be a userspace issue11:07
radiofreethe crashing clickdot in weston-ivi-shell is of no concern for a genivi release11:07
radiofreethat fancy shell isn't part of genivi, it's just a reference11:07
pedroalvarezradiofree: right, does the mock navigation works?11:08
radiofreei'll test some egl clients in wayland-ivi-extension (the actual important genivi bit) and make sure everything works there, then send for review again with my comments11:08
radiofreehowever i think we might need to downgrade cairo or something...11:08
radiofree*but* there's one last kernel/nouveau combo for me to try first :)11:09
pedroalvarezradiofree: thanks!11:09
pedroalvarezis this going to be 3.17 or 3.18?11:09
radiofreerc4 to keep paulsher1ood happy11:10
aananthKinnison: my trove id is "vtsc-trove",  I tried "git clone ssh://git@vtsc-trove/vtsc-trove/local-config/lorries lorry-config", I get "CRIT: This repository is not for you". Any clues11:11
KinnisonSounds like your user isn't in the local-config groups11:12
Kinnisoncan you run ssh git@vtsc-trove whoami11:12
Kinnisonand pop the result in a pastebin for us?11:12
paulsher1oodradiofree:  :-)11:14
paulsher1oodpedroalvarez: implication of this would be we update to rc4 for all our default definitions, is that ok?11:15
paulsher1ood(or maybe not :-) )11:15
aananthKinnison: Here it is Also I found that the local-config directory is in "/home/git/repos/vtsc-trove/local-config" path inside trove.11:15
KinnisonYeah, that whoami output doesn't give any group memberships11:15
KinnisonWith the key which is trove-admin, could you please run: group adduser local-config-writers bsp_ts11:15
Kinnisonthen try the clone again11:15
Kinnisonaananth: when I say 'run' I mean via ssh git@trove11:16
* paulsher1ood hits the deploy hang with base-system too, so probably not size-related11:20
pedroalvarezpaulsher1ood: does your system have space enough?11:21
aananthKinnison: I get "CRIT: You may not perform site administration". I tried with "-i" and without it for the ssh command.11:22
paulsher1oodpedroalvarez: how do i find that out?11:22
pedroalvarezpaulsher1ood: can you leave it running at some point today and check if it finishes?11:22
pedroalvarezpaulsher1ood: `df`11:22
paulsher1oodpedroalvarez: i've left it overnight in the first instance. am not repeating that11:22
Kinnisonaananth: with -i -- what does whoami give you?11:22
pedroalvarezpaulsher1ood: oh! i didn't know that, I thought you only waited for 25 minutes11:23
paulsher1oodpedroalvarez: on a working version this step takes 1 minute11:23
aananthKinnison: Same as what I had pasted in Let me check if I am using wrong file...11:24
pedroalvarezaananth: I think you are :)11:25
paulsher1oodpedroalvarez: oh, interesting. i have run out of space. you're right. now the question is, why? :-)11:25
franredpaulsher1ood, have a look at /tmp/failed11:25
pedroalvarezhe is out of context :)11:26
aananthKinnison: I think I am using the right file, it gives me the same output as in pastebin.11:26
franredpedroalvarez, yeah :S11:26
Kinnisonaananth: It's likely picking up the identity file from ~/.ssh11:27
pedroalvarezpaulsher1ood: ugh,, don't remove anything in /tmp please11:27
pedroalvarezpaulsher1ood: or you will have to flash again11:27
rdale_i have a line 'ref: 5.4' in stratum and i am getting this error: ERROR: Ref 5.4 for qtwebkit in stratum strata/qt5-tools-qtwebkit-wayland.morph is not a string11:28
paulsher1oodpedroalvarez: this is my vm, not jetson11:28
aananthKinnison, pedroalvarez: No, I have the trove-admin pub key in my local folder and I think it is right. May be I will redo the whole excercise.11:28
rdale_and i can't see what is wrong with '5.4'11:28
Kinnisonaananth: Do you have the private key for trove-admin too?  Kinda needed :-)11:28
paulsher1oodrdale_: put quotes round it11:28
paulsher1oodknown bug11:28
Kinnisonaananth: I typically try and have multiple users (real UNIX users) for multi-key scenarios11:29
rdale_ah ok - a number is something that starts with a digit, but v5.4 would be a string11:29
pedroalvarezpaulsher1ood: right, first `umount  /tmp/*`11:29
paulsher1oodwithin the vm?11:30
paulsher1oodbefore doing anything, any idea what has filled this up?11:30
pedroalvarezno clue, but I want you to do that first because you may want to free space in /tmp thinking that is full, and you have your filesystem mounted there11:31
KinnisonIt's likely a failed deploy left mounts around11:32
paulsher1oodumount: /tmp/ssh-tFOukP19y5: not mounted11:32
paulsher1oodnow i have some space11:32
pedroalvarezpaulsher1ood: yeah, that's normal given my command, but now your `df` shouldn have things like: "/dev/sda               4194304   3819308      4300 100% /tmp/tmp.WBzb0V"11:33
Kinnisonthat looks better11:34
Kinnisonnow clean up your TEST11:34
Kinnisonthen check df again11:34
paulsher1oodi have no TEST11:34
KinnisonBut there's an implication that anything larger than about 900M won't deploy safely11:34
paulsher1oodso i don't have to mount anything else? i should be ok to rerun a deploy?11:36
pedroalvarezaananth: I can help you to recover the admin pub key. it should be in "/etc/trove/" in your trove11:36
aananthKinnison: Yes, I have private key also.11:36
pedroalvarezoh, yeah you need the private11:37
Kinnisonaananth: well, once you've worked out the ssh runes to get the right private/public keypair working, add your bsp_ts user to the local-config-writers group and you should be golden to proceed11:37
* Kinnison unfortunately isn't sure what ssh runes to recommend :-(11:37
pedroalvarezaananth: ignore my comment about recovering the ssh key11:38
aananthpedroalvarez: Ok. Kinnison: I am bit confused :-), should I use private key of trove admin to add user and group?11:41
aananthOk :-) I messed it. I hope if I retry with private key again, it should work?11:42
KinnisonIt should11:43
KinnisonThe user looks to have been added okay, so you just need to do the 'group useradd local-config-writers bsp_ts'11:43
paulsher1oodpedroalvarez: after clearing that space, and trying again only to deploy a base-system, still hangs12:13
pedroalvarezand is your system full again?12:13
paulsher1oodpedroalvarez: yes12:14
persiapaulsher1ood: Do you have enough systems in your environment that you can delete some with system-version-manager?12:15
pedroalvarezpersia: nope12:15
pedroalvarezHe only has one I believe12:15
paulsher1oodthis is the weird thing. i've been successfully doing this for a long time, getting up to 3 coexisting systems. here i have one, am trying to deploy a second, and hanging12:16
persiaAnd a large one, hrm.  Yes, that requires larger storage to deal with the massive delta glibc introduces12:16
paulsher1oodso glibc does increase our sizes significantly?12:17
paulsher1oodare we doing deltas? i thought we weren't12:17
persiapaulsher1ood: It has to do with the size of the delta: the more similar the systems, the more btrfs can reuse stuff with a CoW snapshot, so you don't actually need all the underlying storage (or e;se I misunderstand how the upgrade process works)12:17
pedroalvarezpersia: you are right, that's why he can have even 3 systems in the same vm with only 4G12:18
persiapaulsher1ood: We're not precomputing deltas, but because of the use of CoW snapshots and rsync, most of the benefits of deltas are captured.12:19
paulsher1oodah, ok12:21
pedroalvarezI councur with "4g may not be enought to upgrade to the latest version of definitions"12:21
paulsher1oodin which case, i'm not alone with this. anyone trying to redeploy to self after the glibc change is unlikely to succeed?12:21
pedroalvarezpaulsher1ood: I have a system  here with 59% of its space used, I'm going to try to upgrade it12:23
paulsher1oodcan/should deploy check availalbe space before going into a hang?12:24
pedroalvarezis hard to know how many space we need12:24
paulsher1oodis there any stuff i can clear out to make room?12:26
* paulsher1ood wonders what's clogged his / partition now....
richard_mawperhaps some failed upgrades left in your root subvoleume12:28
paulsher1oodwhat would they look like, please?12:28
paulsher1oodis there some pattern i can rm ?12:29
richard_mawI don't know it from memory, but if you mount /dev/sda to /mnt, you can poke around int here12:30
richard_mawand you'll want to `btrfs subvolume delete` instead of plain rm12:30
richard_mawsince a) it'll be quicker and b) won't require allocations to remove the files12:30
pedroalvarezI think that even the failed upgrades should appear in `system-version-manager`12:30
paulsher1oodthat sounds a bit beyond my skill level, tbh12:30
richard_mawunfortunately it's difficult to have our installs able to clean up on failure if they get stuck12:31
persiaI just generated a new VM with larger /dev/sda when I had this issue.12:31
richard_mawwe might be able to do something like monitor the rsync process for not changing which files it has open in /proc/$PID/fd, and kill it if they haven't changed in a couple of minutes12:32
*** Krin [] has quit [Read error: Connection reset by peer]12:33
paulsher1oodi think it's a topic for deeper thought - would be a shame for this kind of thing to happen on a fleet of targets12:33
*** Krin [] has joined #baserock12:34
aananthOk, Thanks to all, it was a great learning, I captured them in I will retry tomorrow.12:35
paulsher1oodaananth: you've been busy :-)12:36
*** zoli_ [] has joined #baserock12:46
*** zoli_ [] has quit [Changing host]12:46
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock12:46
rdale_is there a standard way of dealing with 'Configure' scripts generated by 'metaconfig'?12:55
pedroalvarezoh really? 12:58
rjekWhy does that pastebin seem to choose which language to highlight as at random?12:58
richard_mawrdale_: I've never heard of metaconfig, sorry.12:58
paulsher1oodpedroalvarez: ouch!12:59
* radiofree is flashing a jetson board from another jetson board13:10
* paulsher1ood has done that before :-)13:11
radiofreeusing dd?13:11
paulsher1oodsing your flashing script, i think13:11
radiofreeoh right, i didn't think that would have worked!13:12
radiofreethought it had intel binaries13:12
Krinhmm, ok so if i was going to use zookeeper to essentialy watch the baserock chunks for any update to a systems installed chunks, would it be possible to then update the chunk without restarting the system? so zookeeper would essentialy be acting as a notification of pushes to the relevant git repos?13:12
KinnisonI'm unsure that's a viable approach.  Nominally you could have a zookeeper attached thing watch for newly deployable content and notify a system, but in current upgrade models it'd then involve a build/deploy/reboot cycle to get it in use13:13
Krinhmm, and the major challenge would be getting zookeeper into git, as the file system that is getting updated needs to be zookeeper aware, as zookeeper works by the file that is being updated notifying all it's watchers that it has changed to stop the need for polling13:14
paulsher1oodradiofree: actually maybe it didn't work :-) i can barely remember when i last had two jetsons side by side...13:15
richard_mawI have my suspicions we could construct a parallel mount tree, pivot_root and umount2(path, MNT_DETACH) to get an atomic filesystem update to a running system, but I'd need to prototype it to prove whether it's possible.13:15
paulsher1oodrichard_maw: that sounds exciting ! :)13:16
rjekrichard_maw: As a reboot-free upgrade?13:16
richard_mawfrom my understanding of the kernel, this ought to work, but I've never heard of anyone trying this before, so it might nto work13:16
richard_mawrjek: aye13:16
Krinis there an automatic notification of updates to baserock chunks already in place? or is it manually maintained by the users?13:17
rjekrichard_maw: You'd probably still need to stop all the services because the file system's vanished from under them (even if GCed), which is only slightly better than a reboot.  (Although it might be important if you've bought Enterpriseā„¢ hardware that takes 20 minutes to POST)13:17
richard_mawor you have clever services that notice when their libraries change and pivot themselves into a new process13:18
richard_mawirssi has /upgrade, and systemd has `systemctl daemon-reexec` to handle this13:18
Kinnisonrichard_maw: presumably you'd also want a variant where if the kernel changes, it kexec's the new system instead13:18
Kinnisonno idea where that apostrophe came from :-(13:18
rjekYour keyboard!13:18
richard_mawyeah, though the difference between kexecing and rebooting is usually only important on BIG IRON ENTERPRISE MACHINES13:19
Kinnisonnaughty naughty keyboard.  You shall have to be punished.13:19
Kinnisonrichard_maw: true13:19
richard_mawKinnison: HIT IT HARDER13:19
Kinnisonrichard_maw: although, even my laptop is such that at least 60% of my boot time is before grub gets a look-see13:19
KrinKinnison, i was more tempted to reference pratchet and say "your finger you fool"13:20
* Krin wonders if anyone gets the reference or he just sounds insulting now13:20
* Krin is going to take the sudden silence in channel to mean the latter13:23
pedroalvarezi need some help with this problem w13:26
pedroalvarezI'll ask for help later, now lunch13:26
pedroalvarezI was going to talk about this error:
pedroalvarezthis is building gstreamer with new bison13:28
persiaI'd likea a more verbose log for that13:28
franredpersia, it is a gstreamer 1.2 compilation and it fails when try to run configure (looks like the perl does not understand the letters as characters to check the versions?)13:30
franredversions of bison*13:31
paulsher1oodcan/should we move gstreamer forward? 1.2 seems a long time ago13:31
pedroalvarezthis is the full log:
* pedroalvarez lunches13:32
pedroalvarezI will try to upgrade gstreamer to see if that solves the issue13:33
paulsher1oodin other news, i've managed to self-deploy the latest devel-with-glibc to a recent devel vm, so the space issue is probably specific to my original vm13:46
paulsher1oodthanks for the help on diagnosing that13:47
pedroalvarezjjardon radiofree: would an upgrade of gstreamer cause problems I'm weston/wayland? 13:47
jjardonpedroalvarez: dont think upgrade will fix the issue: i thin franred is rigth and the version checker for bison doest expect letters13:55
paulsher1oodthis reminds me of a previous experience where 'repo' expected only numbers in git's version14:08
paulsher1ood(ie android repo)14:08
paulsher1oodSotK: can you get your prototype running in public somewhere? would be nice to see it in action.... maybe we have some spare capacity at dc14:16
SotKpaulsher1ood: I can try, who do I ask to obtain an account at dc?14:20
paulsher1oodSotK: you shouldn't need an account i think - maybe pedroalvarez can help you14:21
* paulsher1ood spoke too soon - managed to kill his new vm entirely14:23
jmacsHmm, morph produces a lot of strace output14:28
*** cosm_ [] has joined #baserock14:32
*** cosm [] has quit [Ping timeout: 255 seconds]14:36
*** cosm_ [] has quit [Ping timeout: 245 seconds]14:41
pedroalvarezI need some kind of help with this problem  of bison and gstreamer14:52
pedroalvarezI wonder now how this has ever worked with previous versions of bison14:53
pedroalvarezok, is because bison is giving things that are not in [0-9] in its version number14:56
richard_mawpedroalvarez: create a .tarball-version file in pre-configure-commands?14:57
pedroalvarezof bison you mean?14:57
pedroalvarezthat fix makes me sad14:59
paulsher1oodthere must be another solution, surely?15:03
paulsher1oodwhat's reacting badly to the version number?15:03
pedroalvarezI can also rebase my commit until its first 4 chars of the sha1 doesn't contains [a-f]...15:03
pedroalvarez(joking, of course)15:03
pedroalvarezgstreamer, when checking the version of bison, fails:
paulsher1oodpatch to fix gstreamer version checking?15:04
paulsher1oodor ignore version checking? :)15:04
pedroalvarezactually they don't need any version, just need bison15:05
pedroalvarezI don't know who wants to compare it with a version15:05
pedroalvarezto me this sounds like an autoconf bug15:05
pedroalvarezor bison for adding the sha1 to its version number15:06
franredpedroalvarez, sometimes you need to use bigger versions of certain packages to compile your package (functions.... not sure if the term is API compatible)15:07
franredso you need to check that your system contains the package and its version is correct15:08
pedroalvarezfranred: It doesn't look like that in gstreamer:
rdale_in the git log of our lorried bison, jannis has made a commit about getting the version from the NEWS file and putting in into .tarball-version - what does that do?15:09
pedroalvarezhe dropped it later15:11
pedroalvarezI just want to note, that it has been working by chance before my upgrade to a newer version15:12
pedroalvarezI tempted to add that patch again 15:17
jmacsDoes morph filter the output from configure scripts so it only shows things starting with ERROR? Is there a quick way to disable that?15:21
straycatI don't really remember15:25
straycatBut I did add a --build-log-on-stdout option15:25
straycatit's really an internal option for use by distbuild, but it might be useful for you in this case15:26
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]15:26
franredjjardon, radiofree, do we need the command-line parser for gstreamer in baserock? pedroalvarez can --disable-parse will not check bison nor flex15:32
pedroalvarezis kind of a shame that we have to disable things because of that15:32
franredjmacs, you could also check the errors in morph.log15:32
franredpedroalvarez, have you check the gstreamer bug tracker to check if there are more people with that error?15:33
pedroalvarezjmacs: I don't think we filter anything 15:33
pedroalvarezfranred: nope15:33
jjardoncommand line parser is important: I alot of time you need to create test pipelines using the command line (with gst-launch, for example)15:33
franredpedroalvarez, maybe it is worth to check by something like:  AG_GST_BISON_CHECK errors 15:34
franredjjardon, I feared that was the answer15:34
franredjmacs, a very nice way to check what happens when your compilation fails is do a chroot into the staging area (but if it fails it will be in "failed" rather than in "staging" - error message wise)15:36
franredso you can run the commands manually in the status as the system is15:37
jmacsI'm not talking about compilation but configuration scripts15:37
jmacsI have just put "ERROR:" in front of all my debugging printouts15:39
franredjmacs, do you have run with --verbose?15:39
jmacsNope, I could try that15:39
* paulsher1ood is sad to note that upgrading to the glibc version is also hanging on jetson 15:49
jmacsOK, looks like morph doesn't filter output at all. No idea what was going on there. Thanks for the help anyway.15:50
*** genii [~quassel@ubuntu/member/genii] has joined #baserock15:58
*** zoli_ [] has joined #baserock16:02
*** zoli_ [] has quit [Changing host]16:02
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock16:02
persiapaulsher1ood: Generally speaking, you want to have sufficient free space on the target for the entire new system on upgrade.  There are some space savings because of btrfs CoW and some network savings because of rsync, but relying on these is a recipe for pain.16:07
paulsher1oodpersia: i'm using default sizes. so this means that users following down this same path will all suffer the same fate, i'm afraid16:14
paulsher1oodie anyone downloading current devel, then building current definitions and deploying to self16:15
persiaYes.  I believe the default size recommendation was made before upgrades worked this way.16:15
* paulsher1ood would like folks to consider using to test buld and deploy of changes before they hit master16:16
mauricemoss_radiofree, I want to flash u-boot as a coreboot payload to be able to boot any kind of Kernel.16:19
radiofreesorry that sounds very androidy, i have no experience with that16:19
radiofreeyou could try asking in #tegra16:20
pedroalvarezwhy is autoreconf failing here?
paulsher1oodis this on latest systemd?16:23
pedroalvarezi'm trying to build bison, from  source. but out of morph16:24
pedroalvarezrunning the same commands that morph runs, but seems I'm missing something16:25
pedroalvarezok, I was being silly16:27
persiapedroalvarez: Anytime you run perl and get that message, any text you generate is best considered undefined (as your locale may be far from C).  Try rerunning in C, nd see if that works.16:27
paulsher1oodwe should probably fix perl locale properly?16:29
persiaIT's not a perl problem.  It's that there are no locales installed by default.16:30
pedroalvarezI want to be in the locales discussion, but I don't have the time now16:30
radiofreepedroalvarez: the unified patches are ready for review again :)16:30
persiaI was looking at that some time ago, but got busy with $work and an injury.  I'll try again (unless someone beats me to it)16:30
radiofreethere are still some issues, my patches for weston not to use glesv2 should fix the bulk of them though16:31
radiofreehowever, EGLWLMockNavigation in ivi-controller works just fine16:32
radiofreewhich i suppose is what genivi are after16:32
* radiofree will quickly test a 3.17 kernel though16:32
paulsher1oodpedroalvarez: was there another step for clearing space in after the 'umount /tmp/*' - if not i'm going to have to recreate my vm, and reflash my jetson...16:33
pedroalvarezpaulsher1ood: there wasn't, unmounting /tmp/* was for safety reasons16:34
paulsher1oodpedroalvarez: so no way back from ?16:34
pedroalvarezdoesn't system-version-manager shows anything this time?16:36
paulsher1oodah. ok. thank you16:36
radiofreepaulsher1ood: if you run btrfs filesystem max /16:38
radiofreeyou should get another 6GB16:38
paulsher1oodradiofree: btrfs filesystem: unknown token 'max'16:39
radiofreeoh well i got that command wrong16:39
radiofreebtrfs filesystem resize max /16:40
*** jonathanmaw [] has quit [Quit: Leaving]16:40
paulsher1oodcool, radiofree !!16:40
paulsher1oodshould that be on the wiki?16:40
pedroalvarezok, the delta that we are creating in bison to change the submodules url is causing this version number change16:40
pedroalvarezdisabling the check of bison in gstreamer doesn't sound like an option, because other components may do the same check16:41
pedroalvarezI think that generating .tarball-version is a good option here16:42
* paulsher1ood can't comment, doesn't understand the issues fully16:42
KinnisonCan we do that usefully in the morphology?16:42
pedroalvarezwe used to do it:
radiofreepaulsher1ood: if you change the ROOTFS_SIZE=15032385536 on line 28 of the flashing script, when you run that command you end up with a 14G /16:43
Kinnisonpedroalvarez: perhaps we should again :-)16:43
radiofreeshould probably do that16:43
pedroalvarezwriting .tarball-version doesn't seem to solve the issue16:50
radiofreei think i'll resubmit those patches with the 3.17 kernel, less issues16:51
radiofreespoke too soon16:52
pedroalvarezOk, I think I'm going to remove the check in gstreamer, I don't think I can easily fix the bison problem16:53
Kinnisonnot even inserting tarball version?16:53
radiofreei saw some scooters in china town on sunday for
radiofreethey appear to be a "just-eat" equivalent for chinese students16:55
radiofreemost of the restaurants in china town are on there, which is nice16:55
radiofreethis is totally the wrong channel for that, sorry16:56
rjekFood is a nice distraction from time to time.16:57
* pedroalvarez finds the problem with .tarball-version16:58
*** tiagogomes [] has quit [Quit: Leaving]17:00
pedroalvarezfor the curious, this is the patch I think I need to fix the version of bison:
*** wdutch [] has quit [Quit: Quit]17:03
pedroalvarezhow likely is this patch going to be merged?17:04
paulsher1ood+1 from me :)17:07
paulsher1ood(a one line change in a morph file? how much resistance could there possibly be?)17:07
paulsher1oodpedroalvarez: ^^17:07
pedroalvarezit was a way to check if there were reviewers here :)17:08
paulsher1oodpedroalvarez: if you are confident it works, i would venture a +2 to see how others react :)17:08
* paulsher1ood considers himself reasonably competent to review .morph file changes. less so morph code17:09
pedroalvarezon line change is easy to review, but it doesn't have to be ok17:09
paulsher1oodpedroalvarez: would you prefer me to build it?17:10
* paulsher1ood has managed to recover from the space issues now17:11
persiaSome one-line changes in morph files are very hard to review: for example if one just changes a SHA1.17:13
paulsher1oodpersia: true17:15
ratmice________pedroalvarez: in theory the "autoconf way" is not to check versions but features, so gstreamer-common/m4/gst-parse.m4 should check if the '<>' feature works instead of the version :/ not sure exactly what that entails though17:15
* paulsher1ood has separated out into its own page17:17
paulsher1oodi suggest we might notify the list that it might be necessary to increase space before self-upgrading to anything after the glibc change17:18
pedroalvarezratmice________: agree, I should take a look at gstreamer and see what can I do17:18
franredpedroalvarez, +1 if you add the comment about why we use this workaround17:20
pedroalvarezfair enough, thanks franred 17:20
pedroalvarezfranred: is this ok?
franredpedroalvarez, it is alright, although I would mark it as a workaround/hack in some way17:27
pedroalvarezmerged, thanks!17:30
franredpedroalvarez, it is late but...
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]17:32
* pedroalvarez re-populates the cache17:33
* franred hides... it is the same thing17:34
paulsher1oodpedroalvarez: shouldn't we just mail the list and recommend the mason cache-server?17:35
paulsher1ood(unless that has to migrate soon)_17:35
persiaI think we should give that a name, and modify morph to default to it, *and* mail the list telling folk not using current morph to consider using the new cache server.17:35
paulsher1ood(would save a lot of copying)17:35
persiaDoes morph support redirects?  Another option would be to have g.b.o redirect cache requests to the working cache server17:36
franredpersia, modify morph to default = add the line in the wiki when it says create morph.conf17:37
pedroalvarezI've sent an email to remember how to reuse the cache from there17:38
persiafranred: No, rather modify morph so that if you don't have a line in morph.conf it goes there.17:38
persiaThe current line is not mandatory17:38
paulsher1ood+1. but then there are many other mods to morph that would be nice too. for example am i right that it downloads every component to cache if it can before downloading the rootfs if it can?17:41
pedroalvarezpaulsher1ood: we should look at that bug17:41
paulsher1oodradiofree: i've built and booted your unified kernel... it doesn't seem entirely happy though :/17:43
radiofreewhat do you mean?17:44
radiofreeclear and concise bug report that ;)17:44
franredpersia, so redirect to the cache server rather than g.b.o, yep, agreed17:45
paulsher1oodfirst reboot, lots of errors, then it hung. on hitting reset, it seems happier.17:45
paulsher1oodusing screen, so i seem to lose all backscroll, otherwise i'd cutnpaster it17:45
radiofreepaulsher1ood: ctrl+a : scrollback 100017:46
paulsher1oodradiofree: i'm on mac17:46
*** Krin [] has quit [Remote host closed the connection]17:47
paulsher1oodi may choose to disregard what i saw ... subsequent boots look better17:47
radiofreeit's still gnu screen though?17:47
pedroalvarezmaybe is Command + a 17:47
paulsher1oodradiofree: it is. so having set that, how do i actually scroll back?17:49
radiofreeit's incredibly intuitive paulsher1ood, Command +a :[17:49
jmacsCommand+a ESC then use pageup/pagedown17:50
* paulsher1ood fails17:53
*** mariaderidder [] has quit [Quit: Ex-Chat]17:58
pedroalvarezI'll test james patches tomorrow morning just to verify that everything works in the release, and then do the release :)17:59
radiofreepedroalvarez: needs the weston patches as well!18:01
radiofreei'd personally like to upgrade the weston-ivi-shell and wayland-ivi-extensions18:01
pedroalvarezjavier's patches?18:01
radiofreeoh no, i don't like them18:01
radiofreeweston-ivi-shell not weston18:01
radiofreebut it's touch-and-go whether or not the wayland-ivi-extensions stuff we've done will be merged in time for tomorrow!18:02
radiofreehopefully will find out in the morning and send a patch to upgrade those18:02
rdale_if i add a configure line to a build morph, does it add to any existing config lines for that sort of project, or does it replace any default config commands?18:21
paulsher1oodrdale_: ^^18:23
rdale_ah ok thanks18:23
paulsher1oodradiofree: i spoke too soon....18:31
paulsher1oodcan't get back to my trusty devel image18:31
radiofreepaulsher1ood: is there an sd card in?18:32
radiofreei've been seeing this as well, maybe it's a problem with doing the btrfs max command?18:32
radiofreeif you reboot it a few times it should work18:32
radiofreefailing that18:33
radiofreeinterrupt u-boot and type18:33
radiofreesysboot mmc 0 btrfs 0x90000000 extlinux.conf18:33
paulsher1oodError reading config file18:33
radiofreeoh dear, did you run a cycle script or are you just rebooting?18:34
radiofreebtrls mmc 018:34
radiofreedo you can any output from that18:34
paulsher1oodi had run cycle. was trying to boot back18:34
paulsher1ood<DIR>       systems18:34
paulsher1ood<DIR>       state18:34
paulsher1ood<FILE>      extlinux.conf18:34
radiofreei believe your extlinux.conf my be corrupted then18:35
paulsher1oodseems so :/18:35
paulsher1oodoh well18:35
radiofreethere might be a way to salvage it18:35
radiofreeums 0 mmc 018:35
radiofreeplug the jetson board into your laptop (using the usb port you use to flash it)18:36
radiofreeshould come up as a usb device on your laptop18:36
radiofreepaste the contents of /path/to/baserock/extlinux.conf somewhere18:36
radiofreethese btrfs patches are incredibly flakey though, i'll take a look at them tomorrow morning18:36
paulsher1oodi think this is too complex for me today... it's beer o'clock18:37
richard_mawin other news, I tested the principle by pivot_rooting into a parallel mount tree, and it appeared to be workable, next up in that experiment is probably to extend system-version-manager, but that'll have to wait until I have time to look at it18:39
radiofreemaybe we could have some fall back in u-boot if the extlinux.conf fails18:40
radiofreejust boot the default subvolume18:40
rdale_isn't that the same problem i had the with extlinux.conf file with uboot, and i did a 'lsbtrfs' command or similar in uboot and it magically fixed it?18:45
*** zoli_ [] has joined #baserock18:45
*** zoli_ [] has quit [Changing host]18:45
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock18:45
rdale_or radiofree did a magic lsbtrfs i should say18:45
radiofreeclose, btrls18:48
radiofreethat helps in the cases where u-boot can't read anything at all18:48
radiofreelooks like it's loading the config and failing because it's wrong18:48
rdale_i see but this is different then?18:48
*** franred [] has quit [Quit: Leaving]18:48
radiofreerdale_: however yeah, a few btrls' to warm up the fs might help18:48
radiofreeone of the failures for the config loading is it it can't load /systems/foo/kernel etc...18:49
radiofreeactually for me just then it was failing because the sd card was in :\18:51
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]21:16
*** zoli_ [] has joined #baserock21:32
*** zoli_ [] has quit [Changing host]21:32
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock21:32
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]21:33
*** thecorconian [] has joined #baserock23:05
pedroalvarezlooks like linux-pam needs to be upgraded too23:23
pedroalvarez which I believe is fixed with this patch:
pedroalvarezI wonder why we are lorrying from tarball23:28

Generated by 2.15.3 by Marius Gedminas - find it at!