**** BEGIN LOGGING AT Fri Feb 14 02:59:59 2020 Feb 14 09:09:07 tlwoerner: yes, thats what I found out - ATAGS is getting deprecated, but what is not clear to me if its still possible to pass parameters from the bootloader in device tree mode, so that bootloader params could be appended to whatever the dtb defines, still digging Feb 14 13:51:43 Jin^eLD: wow! unbelievable. all the atag/cmdline/dtb stuff is gone from 5.5! i thought you were just looking in the wrong place Feb 14 13:52:06 i guess it's supposed to be efi from here on? Feb 14 13:52:58 decades of SBCs swept aside for the non-existant arm servers we've all been *hearing* about (but not actually seen) Feb 14 14:16:48 Jin^eLD, tlwoerner: with DT, you can pass the kernel args via /chosen/bootargs Feb 14 14:27:34 shoragan: ah, thank you (it took a moment to realize understand your comment) Feb 14 14:27:53 Jin^eLD: so you'll need separate DTs (or separate DT overlays) when doing A/B switching? Feb 14 14:46:26 tlwoerner: I was hoping that I would not, actually it would be enough for me to be able to pass root=/dev/mmcblk1p${active_root} and then swupdate sets active_root variable in the u-boot env from userspace Feb 14 14:46:50 then I could sysboot or bootm the chosen partition Feb 14 14:50:05 I tried only with sysboot so far, there bootargs were ignored although kernel config suggest it should work Feb 14 14:50:16 so maybe bootargs work only in combination with uboots bootm command Feb 14 14:50:27 but I need to figure out where to load the kernel first Feb 14 15:48:02 it's an orange plastic/rubber mat that is water proof, and glued to the subfloor and tiles on top.. Feb 14 15:48:08 oops Feb 14 16:36:22 lol Feb 14 17:24:56 tlwoerner: I think I figured it out, you need to boot with bootz, then bootargs from u-boot are being taken, if you use sysboot then they are ignored Feb 14 19:18:32 hmm thats weird, the resulting wic image always has one partition more than I define in the wks file Feb 14 19:26:12 seems if type "rootfs" is used, then weird things happen Feb 14 19:43:27 is it possible to have an IMAGE_FSTYPES per recipe? or is IMAGE_FTYPES always global for everything? Feb 14 19:53:24 if I use wic's rootfs plugin then the result is simply not what I want, but then again I can not use --rawcopy to simply dump the ext4 rootfs into the partition because for rawcopy stuff should have been built earlier by another receipe I guess Feb 14 19:53:44 at least I was not able to tell wic to dump the ext4 into the wic from within the same recipe Feb 14 20:07:41 hey! if a recipe A DEPENDS on another recipe B, the files installed by recipe B will end up in recipe A's sysroot, right? I have a situation where I see that recipe B is in recipe A's "sysroot providers", but I don't see the files? how does bitbake select what files go into the sysroot of a recipe? Feb 14 20:14:45 I'm trying to move the build of a pyro-based distro to a different server. all layers are at the same hash, but write_deploy_manifest fails because it can't find build/tmp-musl/deploy/licenses/systemd-boot/recipeinfo Feb 14 20:16:07 it never built systemd before, I tried overriding it with EFI_PROVIDER = "grub-efi" ... and now the same fails but for grub-efi Feb 14 20:16:17 any hint on what could I be missing? Feb 14 20:17:45 who usually pulls EFI_PROVIDER in? Feb 14 20:32:56 BlastuR: only some of A's files will end up in B's sysroot. e.g. the -dev files would, but binary files wouldn't. so it depends on what ends up in A's packages Feb 14 20:33:28 (or vica versa) Feb 14 20:38:38 mnemoc: my guess is that years of accumulated cruft had simply put the right file(s) in the right place(s) on your old build server. a completely clean build on the old buildserver would probably fail too Feb 14 20:40:45 ow :( Feb 14 20:51:06 tlwoerner: so it depends on the packages split? Feb 14 20:52:29 BlastuR: yes, a sysroot wouldn't be populated with cross-binaries intended for the target, just header files and stuff like that Feb 14 20:52:50 tlwoerner: where's the logic that decides what packages splits go into the sysroot? Feb 14 20:55:05 BlastuR: i can guarantee it'll be somewhere in openembedded-core/meta/classes, not 100% sure off the top of my head Feb 14 20:55:30 maybe staging.bbclass Feb 14 22:03:38 tlwoerner: so, seem to be decided by SYSROOT_DIRS variable in staging.bbclass. Funny, it defaults to ${includedir}, but even if my recipe installs stuff to ${D}${includedir}, and those files are present in the ${WORKDIR}/image directory, they still don't get included in the sysroot :( Feb 14 22:04:27 BlastuR: on master? Feb 14 22:04:58 on warrior Feb 14 22:05:16 (2019-04) Feb 14 22:08:07 i suspect it has something to do with the fact that my recipe where the sysroot isn't populating properly is a kernel recipe (includes linux-yocto.inc), which in turns inherit kernel.bbclass. In that class, there is a lot of funky stuff going on with the PN, FILES,.. essentially rewriting the package name to always be "kernel" and moving all FILES from the original ${PN} into that space instead. I bet Feb 14 22:08:10 that's why the recipe sysroot doesn't contain anything.. Feb 14 22:14:21 Anyways, bedtime! thanks for your help tlwoerner ! **** ENDING LOGGING AT Sat Feb 15 02:59:57 2020