**** BEGIN LOGGING AT Tue Jun 24 02:59:59 2014 Jun 24 04:45:29 oh yeah, gentoo x86/32bit does work fine with poky/meta-oe latest Jun 24 04:45:35 *as build host Jun 24 04:45:58 what does it take to get it "validated" ? Jun 24 07:37:00 hi there Jun 24 07:38:10 I am modifying my u-boot, which is on a github, and I would like to push my modifications, without fetching each time. I read the development manual, in particular this section : Jun 24 07:38:10 http://www.yoctoproject.org/docs/1.6/dev-manual/dev-manual.html#modifying-temporary-source-code Jun 24 07:38:40 but if I understand correctly this is intended to create patches Jun 24 07:39:24 I tried to use the adress like when you git clone from github : git@github.com:MyName/Myproject Jun 24 07:39:36 but bitbake does not want to fetch it Jun 24 08:04:47 gm Jun 24 08:07:58 nerdboy: which version of gentoo would you validate? Jun 24 08:08:19 nerdboy: that is the problem with gentoo fwiw, the lack of a version we can point at :/ Jun 24 08:31:13 it's a rolling build Jun 24 08:31:29 i could validate stable and testing i suppose... Jun 24 08:32:03 * nerdboy running full ~ for a long time Jun 24 08:33:11 any non-hardened profile on x86 should be fine Jun 24 08:34:09 is there an actual validation protocol? Jun 24 08:34:41 nerdboy: we use the autobuilders as the test Jun 24 08:35:04 nerdboy: so basically the test runs the autobuilder uses are the validation Jun 24 08:35:19 it's basically a matter of installing the build deps Jun 24 08:36:11 nerdboy: the rolling nature of it is just not something we can easily validate and we had to draw the line somewhere Jun 24 08:36:44 once we release, it keeps rolling too which means that new changes introduce bugs and mean we have to push point releases harder Jun 24 08:37:08 there is a base release version... Jun 24 08:38:26 Linux-3.13.10-gentoo-x86_64-AMD_A8-4500M_APU_with_Radeon-tm-_HD_Graphics-with-gentoo-2.2 Jun 24 08:38:52 that gentoo-2.2 part is an actual release version Jun 24 08:39:15 "base system version" i suppose Jun 24 08:41:01 nerdboy: didn't you have the sanity check for make_3.82 ? Jun 24 08:42:39 ? Jun 24 08:42:49 Gentoo ships a 'buggy' make. Today Jun 24 08:43:05 not a problem here Jun 24 08:43:07 I have had a Gentoo rig until last month Jun 24 08:43:55 if you completely remove the files under build/conf you'll see it triggering the QA for make version Jun 24 08:43:56 sys-devel/make-4.0-r1:0 is current Jun 24 08:44:22 i thought 3.82 was patched Jun 24 08:44:39 not stable Jun 24 08:44:50 ah Jun 24 08:44:55 kinda weird... Jun 24 08:45:03 http://packages.gentoo.org/package/sys-devel/make Jun 24 08:45:15 ~ Jun 24 08:45:37 nerdboy: that make produces valid images ;) Jun 24 08:45:49 ..but somehow fails th etest Jun 24 08:51:50 you're talking make-3.82-r4 ? Jun 24 08:53:04 i get some qa warnings, but nothing to do with make/gentoo Jun 24 08:53:14 same stuff on debian/ubuntu Jun 24 08:56:22 I dont have the system here, just remove that file from build/conf. Just keep local.conf and bblayers.conf. Jun 24 08:56:58 keep that file! Sorry I don't remember its name Jun 24 08:57:13 smthg about validation iirc Jun 24 08:57:31 it is generated on bitbake first run Jun 24 08:57:45 well, actually sourcing tye env Iguess Jun 24 09:14:24 good morning Jun 24 09:18:06 hi, I'm trying to get my system to poweroff on acpi button events. I tried using acpid but it seems to require an acpi interface from procfs which is removed from the linux kernel Jun 24 09:18:12 http://www.gossamer-threads.com/lists/linux/kernel/1590074 Jun 24 09:18:29 what is the current way to handle acpi button events then? Jun 24 09:30:33 good night Jun 24 09:32:43 gn Jun 24 12:24:35 hey I'm having quite a difficult time updating my recipes to daisy due to an issue with the git fetcher. It's claiming that a revision for a git repo doesn't exist when it very clearly does Jun 24 12:25:09 I'm not even providing a SRCREV, just a tag in the SRC_URI, and getting this line: Jun 24 12:25:13 ERROR: Function failed: Fetcher failure for URL: 'git://git@bitbucket.org/devonit/qjsonrpc.git;protocol=ssh;tag=v1.0.2'. Unable to fetch URL from any source. Jun 24 12:25:29 any ideas on what's going on there? Jun 24 12:33:49 mbroadst: try git clone https://bitbucket.org/devonit/qjsonrpc.git in a tmp ditrctory Jun 24 12:34:33 mckoan: I seem to have isolated the problem to the fact that "git branch --contains " returns absolutely nothing Jun 24 12:34:55 it must be that the version was somehow tagged incorrectly, even though it does provide the right code when I check that out Jun 24 12:36:15 if I clone the code locally, checkout that tag, then run a git reflog show --all, that's the only place the commit shows up with this entry: Jun 24 12:36:16 adefe89 HEAD@{0}: checkout: moving from master to v1.0.2 Jun 24 13:12:53 this appears to be a bug with the git fetcher, it can't handle git tags that were made off of other git tags.. is there a way to disable that "git branch --contains" check? Jun 24 13:14:35 or, in other words, there is no branch that contains the revision I'm providing, however the tag and revision do exist. so currently this will always fail with the git fetcher Jun 24 13:16:39 oh looks like nobranch=1 might do it Jun 24 13:37:17 huzzah Jun 24 13:46:58 hello Jun 24 13:47:33 is there a way to force a recipe to be always rebuilt? Jun 24 13:47:43 zeddii: I'm using 3.15 headers. Do you plan to upgrade straight to 3.16? Jun 24 13:49:08 by example to follow a firmware version Jun 24 13:50:49 ant_work, yep. I'm working on 3.16 now, I pretty much always skip one version after a release :) Jun 24 13:51:03 the -dev kernel will jump shortly. Jun 24 13:51:14 the crazy mtd-utils guy did add 2 ioctl's Jun 24 13:51:41 ubiblock fwiw Jun 24 13:52:04 sounds like fun. I'm working on some infrastructure and haven't hit runtime testing yet. Jun 24 13:52:52 mtd-utils_1.5.1 fails with klibc/musl and has patches for uclibc. Ships its newer out-of-sync headers :/ Jun 24 13:53:12 it's a case it compiles with eglibc? Jun 24 13:55:37 it built here, but I only booted 3.16-rc2 on qemux86 and had left the 3.14 headers in place to only change one variable at a time. I'll spawn some new builds. Jun 24 13:56:52 I patched HEAD and now compiles with klibc Jun 24 13:57:03 but 3.15 ubi-user.h is needed Jun 24 14:10:26 ant_work: you trying musl build starting from Khem's branch? Jun 24 14:10:39 I plan to Jun 24 14:10:55 atm I've spotted sure stoppers Jun 24 14:12:30 likewise: basically where they use you have issues Jun 24 14:19:22 ant_work: ok, I was planning to give it a spin soon Jun 24 14:38:31 hi. i wonder: is there any established method for safe, "murphy compatible" operating system updating on embedded hardware? based on yocto of course... Jun 24 14:39:38 i have a method in my head, but i'm not sure how to apply that on yocto ... so i thought someone else must surely have thought about this before me ... ? Jun 24 14:40:36 my method: one partition with a minimal os that updates and/or boots the real operating system Jun 24 14:41:27 the real operating system is placed on one of two possible partitions. if update fails or the updated operating system fails to boot, the system falls back on the previous partition. Jun 24 14:56:55 YPTM: Armin is on Jun 24 14:57:19 YPTM: Conference Details: Company: WIND RIVER SYS Ready-Access Number: 8007302996/9139049836 Access Code: 2705751 Jun 24 14:57:28 YPTM: Stephen has joined Jun 24 14:59:15 ionte: this is quite a common design pattern - there are some mechanisms in existing bootloaders etc. for this kind of A/B firmware support. Jun 24 14:59:32 * nerdboy loves waking up to a meeting... Jun 24 14:59:37 ionte: completely murphy compatible is hard though. Jun 24 15:00:01 YPTM: melissa is on the call. Jun 24 15:00:35 ionte: u-boot supports boot counters within persistent memory (or any register) to keep track of failing boot attempt, redundant env partition and other stuff, but you need to tweak it to your product req. Jun 24 15:01:23 YPTM: tom z on Jun 24 15:01:41 YPTM jefro is on the call Jun 24 15:02:00 Herb Kuta is on the call Jun 24 15:02:05 YPTM: Saul is on Jun 24 15:02:19 YPTM: Michael here. Jun 24 15:02:35 YPTM: Richard joined Jun 24 15:03:01 YPTM: Mark is here Jun 24 15:03:26 YPTM: Bruce is on the call Jun 24 15:04:16 YPTM: Matthew on the call Jun 24 15:04:49 I'm a noob. What's YPTM? Jun 24 15:05:02 Yocto Project Technical Meeting.. Jun 24 15:05:02 Yocto Project Technical Meeting Jun 24 15:05:07 ty Jun 24 15:06:37 YPTM : nitin is on the call Jun 24 15:08:08 YPTM: I'm fine w/ every other week.. mail and irc discussions seem to be going well Jun 24 15:10:06 That is RPM for MPW? Jun 24 15:10:12 YPTM: These meetings seem to get a bit more lively as we near release dates. Jun 24 15:10:27 YPTM: Cristian is in the call Jun 24 15:10:30 sjolley RPM for meta-mingw Jun 24 15:10:40 Thanks Jun 24 15:10:43 IPK seems to work fine Jun 24 15:14:46 RP - which host OS/kernel version is that machine using? Jun 24 15:16:28 YPTM is over. Jun 24 15:17:02 Here is that script for anyone looking http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/scripts/contrib/build-perf-test.sh Jun 24 15:17:47 I'd add that this mainly affected the fedora19 machine, the ubuntu 12.04 didn't show the behaviour Jun 24 15:19:58 RP, We currently only have Fedora 20 in the build pool. Fedora 19 is still supported though. Would you like a Fedora 19 machine added to the autobuilder? Jun 24 15:21:57 halstead: that probably won't make much difference. I just don't understand why the performance is drifting over time like this on that machine in particular Jun 24 15:22:15 * halstead nods. Jun 24 15:24:02 Is the sstate dir being cleaned? Maybe something simple like traversing the sstate dir with 1000s of files repeatedly could cause such a slow down. Jun 24 15:29:54 vmeson, Looking at the script I see sstate is frequently wiped. Jun 24 15:31:19 those tests don't store sstate. the only thing kept around is buildstats Jun 24 15:32:18 ok, capturing a build with ltt/systemtap/sysdig/... now and in a week would be interesting. Jun 24 15:43:57 I'm getting an exception "cannot concatenate 'str' and 'NoneType' objects" related to IMAGE_CMD in do_rootfs, sound familiar to anyone? Jun 24 15:46:02 only one result on google, and it's a pastebin :) Jun 24 15:46:28 it's definitely the same problem: http://pastebin.com/s20DpBG1, so I know someone's seen it before Jun 24 15:59:01 something is fishy... Jun 24 15:59:20 does IMAGE_CMD look sane? Jun 24 15:59:38 * nerdboy guessing it's a custom image Jun 24 16:00:22 IMAGE_CMD in bitbake.conf is: IMAGE_CMD = "_NO_DEFINED_IMAGE_TYPES_" Jun 24 16:00:24 yeah it's a custom image Jun 24 16:00:48 we're building an ext3 image, and our old configuration has an append for IMAGE_CMD_ext3 Jun 24 16:00:54 set an IMAGE_TYPES in local.conf Jun 24 16:01:16 better use an ext3 one... Jun 24 16:03:40 let's try that.. Jun 24 16:04:35 we defined the IMAGE_FSTYPES = "vmdk" previously, that seemed to suffice Jun 24 17:21:15 nerdboy: well it moved forward but still errors out.. we're basing our custom image on: http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-core/images/build-appliance-image_8.0.bb Jun 24 17:21:57 I added a IMAGE_TYPES = "ext3" to our conf, but now its failing on boot directdisk because the *.ext3 file doesn't exist Jun 24 17:30:35 does anyone here build the build appliance with any regularity? Jun 24 17:32:48 unfortunately the buildbot's localconf gets chopped off Jun 24 17:36:01 My filesystem ran out of space during a bitbake. Is it safe to free space up and simply continue, or do I need to clean up the tasks that were running at the time? Jun 24 18:06:21 cycle Jun 24 18:06:32 /fail Jun 24 18:07:01 mbroadst: i think you want IMAGE_FSTYPES no? Jun 24 18:08:33 nerdboy: looks like we had a replacement image-vmdk.bbclass and boot-directdisk.bbclass laying around in our layer (to add swap to the directdisk mostly) that was causing it all Jun 24 18:08:36 heh Jun 24 18:08:42 #funwithbitbake Jun 24 18:20:37 Hey everyone Jun 24 18:20:51 Can anyone help me with a bitbake recipe problem? Jun 24 18:32:17 nerdboy: really appreciate the help btw Jun 24 18:35:01 When I grow up I create every other day a new license. Jun 24 18:35:48 openssl_tpm_engine uses an OpenSSL license from 1998-2007 that is different to the one in Yocto. Jun 24 20:58:01 anyone know why run-postinsts.service wouldn't be run on the first bootup of my image? it shows up in systemctl units as running, but it should be a single shot and then disable itself Jun 24 21:00:31 in my case it's sysvinit, but it seems like pkg.postinst isn't getting run on first boot anymore either Jun 24 21:00:53 * nerdboy needs to check for a configure script Jun 24 21:01:39 I've found a post on the mailinglist indicating that it's EITHER run when building the rootfs OR on firstboot but not both Jun 24 21:02:00 I'm wondering if there's a rogue recipe that's forcing it to run before firstboot and therefore stopping it from running? Jun 24 21:20:06 what is state of the art for building sd cards from OE? Jun 24 21:54:32 nerdboy: it seems to be because run-postinsts wants to look for /etc/*-postinst, when really all I need done is opkg-cl configure Jun 24 21:54:37 strange Jun 25 00:02:04 Crofton: state(s) of the art are all different Jun 25 00:02:43 meta-rpi, meta-beagle, meta-fsl all seem to have their own .bbclass for that Jun 25 00:03:17 be nice to consolidate that functionality and use config params for each machine... Jun 25 00:04:29 eg, latest kernel/u-boot has the same sdcard partitions and u-boot.imx for many iMX boards Jun 25 00:05:02 nerdboy, that is what I am thinking Jun 25 00:05:04 *same u-boot image install, different build config obviously... Jun 25 00:05:31 I need to directly make some sd cards and do not feel like making another wheel Jun 25 00:05:38 or at least copying something crappy :) Jun 25 00:06:50 well, the meta-fsl example seems in transition Jun 25 00:07:31 nothing in poky/oe i've seen yet uses the latest patches/kernel/u-boot as rcn documents in LinuxOnArm Jun 25 00:07:34 'wic' seems promising, but i think mostly has bits for x86 atm **** ENDING LOGGING AT Wed Jun 25 02:59:58 2014