**** BEGIN LOGGING AT Tue Oct 21 02:59:59 2014 Oct 21 11:32:43 is there some recommended best practise or example on how to build a self-contained kernel-initramfs cmbination? Oct 21 11:34:21 i've found the INITRAMFS_IMAGE and INITRAMFS_IMAGE_BUNDLE variables, but as far as i get it, those are more meant for the boot stage of a full blown rootfs Oct 21 11:34:28 LetoThe2nd use the one from your os or buildsystem Oct 21 11:35:14 woglinde: in what sense? Oct 21 11:35:38 LetoThe2nd if you use debian or oe use the the system they provide Oct 21 11:35:54 and edit stuff afterwards in case of debian Oct 21 11:36:03 depends all on your use case Oct 21 11:36:17 I would rather find hard to start a initramfs from scratch Oct 21 11:36:48 woglinde: as i'm in #oe, i#m basically looking for the oe way of doing things, e.g. i want to have everything neatly automatized and reproductible. RTFM is fine, i just need a pointer to a fitting FM ;) Oct 21 11:37:12 ups okay, sorry Oct 21 11:37:19 I thought we were on beagle Oct 21 11:37:28 shouldn't in the most trivial case the initramfs jsut be a cpio.gz'ed rootfs? Oct 21 11:37:50 I only customized the yocto initramfs which install an image to a partition Oct 21 11:38:07 woglinde: ah i already wondered on the strange kind of answers ;) Oct 21 11:38:11 LetoThe2nd yes cpio.gz the directory Oct 21 11:38:52 I think you even can do it with the IMAGE_TYPES in local.conf Oct 21 11:39:24 so your produced image will be available with cpio too Oct 21 11:39:43 besides ext4 or ubi or whatever Oct 21 11:39:53 so i'd still need to manually do the bake-into kernel step, right? Oct 21 11:40:16 hmm, there is a way Oct 21 11:40:36 ah sorry Oct 21 11:40:53 INITRAMFS_IMAGE_BUNDLE/INITRAMFS_IMAGE are actually used in the "build" process Oct 21 11:42:08 LetoThe2nd: one interesting bit is, that if you bake an image into the kernel, then it must only contain gplv2 compatible things. Oct 21 11:42:47 IIRC there was some way of just appending it, that was kosher for even proprietary Oct 21 11:43:03 bencoh: so use those, and then rather use virtual/kernel as the main target/image? Oct 21 11:43:09 tbr: good point, indeed. Oct 21 11:44:17 LetoThe2nd: something like that, yes Oct 21 11:44:36 bencoh: ok, thanks Oct 21 11:45:43 IIRC the old way of baking into the kernel went out of fashion around 2.4 or at least wasn't very popular when 2.6 came around Oct 21 11:47:09 basically we're just investigating ways to create a small-selfcontained rescue blob Oct 21 12:02:51 *nod* been there Oct 21 12:03:09 actually, I might have been there using oe-classic Oct 21 12:03:28 * tbr should look at the old archos stuff, somedayâ„¢ Oct 21 13:43:33 RP, is tomz still the wic guy? Oct 21 13:46:58 Crofton|work: yes Oct 21 13:47:40 hmm, I wonder if he reads oe-core list :) Oct 21 13:47:52 I am trying to debug a problem Oct 21 14:11:57 is there more magic to an initramfs than just bundling a cpio.gz'ed rootfs? it seems to always run into Kernel panic - not syncing: /dev/console is missing or not a character device! Oct 21 14:21:23 the kernel includes devtmpfs though. Oct 21 14:28:50 shouldn't USE_DEVFS cause enough base preparation? Oct 21 14:37:13 (it obviously doesn't, /dev is completely empty) Oct 21 14:53:37 ok, am i really misunderstanding this? USE_DEVFS="1" is set, as well as IMAGE_DEVICE_TABLES, but /dev stays completely empty Oct 21 14:59:51 Crofton|work: tom just joined #yocto Oct 21 14:59:52 LetoThe2nd: do you have CONFIG_DEVTMPFS_MOUNT as well? Oct 21 15:00:10 otherwise it's not mounted automatically Oct 21 15:08:35 ndec: sure, but google says this does not work for initramfs Oct 21 15:10:58 LetoThe2nd: hmm. really. Oct 21 15:11:31 so i'm basically wondering why /dev is still empty when i inspect the cpio.gz Oct 21 15:17:07 and VIRTUAL-RUNTIME_dev_manager = "" indeed gets rid of udev and stuff, but still no dev nodes Oct 21 15:18:33 JaMa, thanks Oct 21 15:19:18 so i'm really looking for the point why the build just won't populate /dev. is there some arcane black magic needed? Oct 21 15:42:58 and even more fun, the initramfs as suggested by core-image-minimal-initramfs.bb doesn't even contain a /dev/ directory Oct 21 15:43:11 * LetoThe2nd headdesks Oct 21 15:52:33 it even gets weirder... now i'Ve found the initramfs-module-udev package which even is supposed to create /dev/ an mount devtmpfs into it. and manually inspecting the cpio.gz tells me the script is there! Oct 21 19:27:56 woglinde! Oct 21 19:28:57 hi mario-goulart Oct 21 19:29:06 Hi Oct 21 19:29:32 woglinde: does "build/tmp/sysroots/x86_64-linux/usr/bin/ecj-initial: line 4: java-initial: command not found" look familiar when building jamvm-native? Oct 21 19:30:23 I have this in my local.conf: http://paste.debian.net/128014/ Oct 21 19:31:58 hm no it should build cacao-initial Oct 21 19:32:00 first Oct 21 19:32:10 and there fore have java-initial Oct 21 19:32:17 but its not all sstate safe Oct 21 19:32:25 woglinde: I must say I didn't have all that in my local.conf when I first built openjkd-7 Oct 21 19:32:42 Oh, so I should probably nuke my sstate cache Oct 21 19:32:54 if you can bear it Oct 21 19:33:42 I suppose it will be faster that attempting to debug the issue... Oct 21 19:34:50 Hmmm. I could also "bitbake -c cleansstate" for all recipes in meta-java Oct 21 19:36:34 yes Oct 21 19:36:54 or just bitbake -c cleansstate cacao-initial Oct 21 19:37:38 Alright. Thanks for the hints. **** ENDING LOGGING AT Wed Oct 22 03:00:00 2014