**** BEGIN LOGGING AT Sat May 27 03:00:02 2017 May 27 11:14:57 nrossi: I guess I figured it out , but it's awful beyond comprehension May 27 11:15:12 nrossi: but I still need to set INITRAMFS_IMAGE in local.conf , I cannot set that in a recipe, can I ? May 27 11:17:24 Marex: yer you have to set that for the kernel May 27 11:54:10 nrossi: right May 27 11:54:47 nrossi: so I guess I cannot avoid bbappending the kernel recipe , right ? May 27 11:55:11 nrossi: or can I cook a recipe which sets variable for another recipe ? ie. can the image recipe set variable for the kernel recipe ? May 27 11:55:39 Marex: bbappend or setting it in your local.conf/distro/etc. May 27 12:08:00 nrossi: can I set it in machine config too ? May 27 12:09:03 Marex: yep that will work May 27 12:09:04 I guess I can May 27 12:09:21 nrossi: I still feel a little violated about it, but ... I guess that works for this specific case May 27 12:09:32 nrossi: I feel this metalayer becomes meta-xi^W-fsl-arm ... May 27 12:10:01 Marex: your use case is a great example of wic needing to support UBI ;) May 27 12:10:29 nrossi: more like a great example of matroshka effect which OE doesn't handle well May 27 12:10:37 image in kernel in image ... May 27 12:10:54 Marex: yep exactly :) May 27 12:10:59 ^-- and kernel is needed to build the first image May 27 12:11:01 geh May 27 12:11:41 Marex: I remember looking at the initramfs built-in stuff a long time ago, back for petalinux to see if it would be possible to embed the initramfs after the kernel image is built May 27 12:12:02 * Marex cringes every time he hears petalinux May 27 12:12:29 nrossi: fitImage FTW, you can embed everything into fitImage after the kernel is built :) May 27 12:12:59 Marex: thats the conclusion i came to due to ramdisk/cpio working exactly like built-in initramfs May 27 12:14:46 I dont think they work _exactly_ the same, but for the Glico distribution *cough* I think external image bundled with the kernel via fitImage is more palatable May 27 12:15:11 (two food references, yes!) May 27 12:15:48 Marex: also exactly, from the perspective of it being loaded as a ramfs filesystem they are the same. Just loaded into memory differently May 27 12:15:54 nrossi: you don't have to explicitly rebuild the kernel with an image which becomes part of it, which is dependency mess May 27 12:15:58 almost* May 27 12:16:05 jupp May 27 22:47:28 I set INITRAMFS_IMAGE_BUNDLE_ = "1" and INITRAMFS_IMAGE_ = "initramfs-image" in my local.conf May 27 22:47:56 thats enough for it to create kernel with an initramfs bundle May 27 23:05:01 khem, there are patches on ML for the removal May 27 23:05:21 of that feature, introducing kernel-initramfs May 27 23:05:53 i see May 27 23:06:01 I did nt pay attention May 27 23:06:09 how will the feature do it May 27 23:06:53 seems with an extra pass May 27 23:07:45 how to use it I mean May 27 23:08:30 IMAGE_INSTALL_append = " kernel-initramfs" May 27 23:08:52 but patch is not committed yet I think May 27 23:09:28 it's a bit what we do wit our installer recipe: a Matroska May 27 23:10:44 ok i guess I need to read it May 27 23:10:59 I am not clear as to how can I set the image to bundle May 27 23:11:06 as initramfs May 27 23:11:16 ah, he introduced a new var... May 27 23:13:18 I am busy with kernel now, hope to test and give feedback in the next days May 27 23:13:29 sounds good **** ENDING LOGGING AT Sun May 28 03:00:03 2017