| *** toscalix has joined #baserock | 00:09 | |
| *** toscalix has quit IRC | 00:25 | |
| *** rdale has joined #baserock | 07:36 | |
| *** paulwaters_ has joined #baserock | 07:39 | |
| paulsherwood | http://stackoverflow.com/questions/3411048/unexpected-operator-in-shell-programming | 07:39 |
|---|---|---|
| paulsherwood | i'm seeing that when ybd executes tar.check as part of a deploy | 07:39 |
| paulsherwood | the offending line is 'if [ "$UPGRADE" == "yes" ]; then' | 07:40 |
| richard_maw | yeah, should be single = | 07:49 |
| * richard_maw is surprised that snuck through, since Sam's usually better at that | 07:54 | |
| richard_maw | though it went without detection because we were running it on systems that either had bash for /bin/sh or busybox sh with all the options turned on | 07:56 |
| paulsherwood | ack, thanks. | 08:01 |
| *** jonathanmaw has joined #baserock | 08:15 | |
| *** jonathanmaw has quit IRC | 08:18 | |
| paulsherwood | https://gerrit.baserock.org/#/c/2207/ | 08:18 |
| *** jonathanmaw has joined #baserock | 08:18 | |
| pedroalvarez | tested and merged | 08:21 |
| paulsherwood | tvm | 08:21 |
| * richard_maw would have s/real/strictly posix compliant/ | 08:22 | |
| richard_maw | bash is a real /bin/sh, busybox provides a real /bin/sh | 08:22 |
| richard_maw | the fact that they also provide extra features does not make them any less real | 08:23 |
| richard_maw | it was not a problem with the shells that were previously running it, it's a problem with the script for assuming every /bin/sh has == | 08:23 |
| * richard_maw goes back to using GitLab for his current project after a brief, gleeful moment of using Gerrit | 08:24 | |
| pedroalvarez | hah | 08:25 |
| pedroalvarez | btw, upgraded yesterday to get bug/security fixes | 08:25 |
| * paulsherwood wonders what the best solution for fixing that commit message is, short of re-writing history | 08:26 | |
| paulsherwood | i guess revert, and re-apply with fixed message? | 08:26 |
| richard_maw | not really worth changing it | 08:27 |
| paulsherwood | ok | 08:27 |
| richard_maw | but git-notes are a thing | 08:27 |
| pedroalvarez | jjardon: mesa patches look good. Do you know if anybody else can test them? | 08:56 |
| *** tiagogomes has joined #baserock | 08:57 | |
| *** locallycompact has joined #baserock | 09:20 | |
| radiofree | has https on gbo always required auth to clone, or is that new? | 09:34 |
| pedroalvarez | I think that it's required since we have it working | 09:37 |
| *** toscalix has joined #baserock | 09:45 | |
| *** toscalix has quit IRC | 10:14 | |
| *** cosm has quit IRC | 11:37 | |
| jjardon | I got a kernel panic after deploying a weston system (current master definitions); does anyone have deployed and run a system built with ybd recently? | 13:11 |
| pedroalvarez | have you managed to extract some info from the kernel panic log? | 13:13 |
| pedroalvarez | is it not even booting? | 13:13 |
| jjardon | Not really :/ https://usercontent.irccloud-cdn.com/file/zSPGt4yO/Screenshot%20from%202016-07-12%2014-14-26.png | 13:15 |
| jjardon | pedroalvarez: ^ | 13:15 |
| pedroalvarez | that one sounds familiar | 13:15 |
| pedroalvarez | "attempted to kil init" | 13:16 |
| pedroalvarez | s/kil/kill/ | 13:16 |
| pedroalvarez | it's like if the init script just finishes | 13:17 |
| pedroalvarez | and probably the init script from initramfs | 13:18 |
| jjardon | let me remove initramfs from my system and try again | 13:18 |
| pedroalvarez | and probably caused by the wrong kernel args | 13:18 |
| pedroalvarez | (these are just guesses) | 13:18 |
| jjardon | pedroalvarez: similar result without the initramfs subsystem in the cluster: | 13:28 |
| jjardon | https://usercontent.irccloud-cdn.com/file/3bd2ou1U/Screenshot%20from%202016-07-12%2014-27-49.png | 13:29 |
| pedroalvarez | so, /sbin/init is failing ? | 13:30 |
| jjardon | seems so :( | 13:34 |
| pedroalvarez | hm.. | 13:37 |
| pedroalvarez | not sure if you can do some more debugging, like using init=/bin/sh and then trying to run /sbin/init | 13:38 |
| pedroalvarez | idk :/ | 13:38 |
| pedroalvarez | I could give it a try with morph, if you want | 13:38 |
| *** CTtpollard has quit IRC | 13:42 | |
| *** CTtpollard has joined #baserock | 13:44 | |
| paulsherwood | richard_maw: i started looking into your suggestion about not recursing the deterministic tar creation | 14:21 |
| paulsherwood | if your theory is correct, i think it means i'd need to bump artifact-version again :/ | 14:22 |
| paulsherwood | how confident are you in your theory? | 14:22 |
| paulsherwood | (the error in question being https://gitlab.com/baserock/definitions/builds/2310768) | 14:22 |
| richard_maw | 100% | 14:27 |
| paulsherwood | ok | 14:34 |
| paulsherwood | do i need to explicitly add the directories in such a tar? or just files? | 14:35 |
| paulsherwood | eg would http://sprunge.us/JafI be a valid artifact? | 14:36 |
| *** toscalix has joined #baserock | 14:50 | |
| richard_maw | if a tarball might end up needing to create a directory, then you need to include the directory information, since that would contain the required metadata for the directory (mode etc.) | 15:08 |
| richard_maw | there at least used to be definitions that created empty directories so it would make them with the right permissions | 15:09 |
| paulsherwood | ack, thanks. | 15:18 |
| paulsherwood | but in this approach, each dir would be added *after the files it contains? | 15:18 |
| richard_maw | yes, because if you try to set a modification time on a directory before you put files in it, when you add the files it changes the modification time to _now_ | 15:19 |
| paulsherwood | ok... eg http://paste.baserock.org/vawimafawa ? | 15:19 |
| richard_maw | possibly, you shouldn't need to chdir, if you rebased the paths right, and I'm not sure whether it iterates into empty directories, however… | 15:24 |
| richard_maw | you might be in a bind with os.walk, since if it leaves the iteration order undefined, you lose deterministic tarballs | 15:24 |
| richard_maw | since the mechanism for changing the iteration order with os.walk requires you to go topdown… | 15:25 |
| locallycompact | fs.osfs! | 15:25 |
| richard_maw | locallycompact: oh, you mean that thing which fails to handle non-unicode file paths correctly? | 15:26 |
| locallycompact | yes that one | 15:26 |
| paulsherwood | :/ | 15:26 |
| locallycompact | non-unicorn flaps | 15:26 |
| paulsherwood | i'll stop this approach, then | 15:27 |
| richard_maw | sorry I pointed you at os.walk when it turns out to be insufficient | 15:27 |
| richard_maw | closer… but not close enough | 15:27 |
| richard_maw | you need some traversal that lets you change the iteration order top-down and also process things bottom-up | 15:28 |
| * richard_maw doesn't recall seeing anything | 15:28 | |
| * richard_maw ponders whether the recent python3 directory walker might work | 15:28 | |
| richard_maw | there was a backport module for python2 | 15:29 |
| richard_maw | hm, no, that just makes iteration faster | 15:30 |
| richard_maw | here's an idea: do the os.walk top-down so you can sort the directories and files, and append those to a list, then reverse that list and use that for your bottom-up traversal | 15:41 |
| paulsherwood | ack, i'll think about it | 15:44 |
| *** toscalix has quit IRC | 15:50 | |
| *** fay_ has quit IRC | 16:12 | |
| *** bashrc has joined #baserock | 16:20 | |
| *** jonathanmaw has quit IRC | 16:53 | |
| *** bashrc has quit IRC | 16:58 | |
| *** franred_ has quit IRC | 17:13 | |
| *** locallycompact has quit IRC | 18:08 | |
| *** locallycompact has joined #baserock | 19:04 | |
| *** locallycompact has quit IRC | 19:10 | |
| *** locallycompact has joined #baserock | 19:15 | |
| *** locallycompact has quit IRC | 19:17 | |
| *** locallycompact has joined #baserock | 19:17 | |
| *** rdale has quit IRC | 19:35 | |
| *** cosm has joined #baserock | 20:06 | |
| *** locallycompact has quit IRC | 20:16 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!
