**** BEGIN LOGGING AT Mon Mar 16 02:59:58 2020 Mar 16 06:56:13 ouch, sstate was hiding host tools compilation issues after switching build container from Debian 9 to 10 buster with glibc 2.28. Target build was already passing until I wiped sstate cache. Any ideas how to fix cross-localedef-native build errors like this: argp-help.c:(.text+0x2001): undefined reference to `_IO_fwide' ? Mar 16 06:56:33 and this on sumo Mar 16 06:58:28 does the pre-build buildtools tar ball fix issues like this? Mar 16 07:02:12 also a bit scary that update of build container did not invalidate sstate for the host tools like cross-localedef-native. Mar 16 07:09:17 How does kernel_do_configure() get hooked into the build process ? I can only find it being declared in kernel.bbclass, but never referenced.. Mar 16 07:12:47 kroon: check output "bitbake -e kernel_recipe", it gets called from do_configure task through some magic. Mar 16 07:14:31 mcfrisk, yup, question is where that magic is coming from Mar 16 07:17:30 kroon: now I want to know too :) Mar 16 07:18:20 mcfrisk, good :-D Mar 16 07:19:00 kroon: seems like all build tool bbclasses have same thing, cmake.bbclass has cmake_do_configure so.. Mar 16 07:22:33 even base.bbclass has a base_do_configure so maybe this magic is in bitbake itself Mar 16 07:30:28 mcfrisk, yeah sounds plausible Mar 16 07:31:23 mcfrisk, yes chapter 3.5.6 in bitbake manual Mar 16 07:38:51 kroon: yea, could not link that to bitbake side implementation though.. tried to grep various things in poky tree Mar 16 07:48:16 kroon, mcfrisk: Look at EXPORT_FUNCTIONS Mar 16 07:48:47 RP, thanks Mar 16 07:49:06 mcfrisk: the host tools change not rebuilding is expected when you contemplate the alternative. The alternative is target output being host specific :/ Mar 16 07:50:40 RP: ok, understood. I was expecting some version checks for host tools which would have triggered rebuilds. Mar 16 07:51:55 mcfrisk: Right, I understand why but for better or worse it isn't the way the system works Mar 16 07:53:36 RP: is mentioned somewhere in the documentation? Mar 16 07:54:06 need to hack the CI setups to build without caches when updates like this are applied Mar 16 07:54:43 mcfrisk: Indirectly probably. Its an implication that all native sstate has the same hash regardless of which host its built upon Mar 16 07:55:15 mcfrisk: You should not have to rebuild anything and you shouldn't see errors like that. Did you try updating uninative? Mar 16 07:56:48 mcfrisk: sorry, I was thinking runtime. For build errors you would have to test with a dummy sstate to check things do build on the new version Mar 16 07:57:23 mcfrisk: you could also use a temporary different sstate version number Mar 16 07:57:46 RP: I picked only pseudo changs to poky sumo. I will check uninative. Mar 16 07:58:04 RP: oh, how does that sstate version work? some kind of distro config PE? Mar 16 07:58:35 mcfrisk: just just a manual number so we can break "ABI" if/when we need to invalidate all sstate Mar 16 07:59:12 I've even done yocto updates without completely invalidating sstate manualy. I hope it was explicit in source every time. Mar 16 07:59:18 mcfrisk: ever wondered what the :3: in sstate file names is? :) Mar 16 07:59:37 black magic? Mar 16 07:59:41 :) Mar 16 07:59:49 mcfrisk: sstate version number Mar 16 08:00:05 mcfrisk: we never really invalidate sstate, the changes release to release are more than enough to do that Mar 16 08:00:45 yes so I hope all host tools too were recompiled due to recipe changes. Mar 16 08:01:18 so basically increment SSTATE_VERSION in e.g. distro config. Mar 16 08:01:52 mcfrisk: not sure its necessary or recommended but it is there Mar 16 08:02:29 and it's hard coded in sstate.bbclass without override possibility so maybe I need to add a patch to poky Mar 16 08:03:00 mcfrisk: right, it was really there for core ABI changes :/ Mar 16 08:04:53 SSTATE_VERSION_append = "1" in local.conf adds to it and invalidates the whole sstate cache then. Thanks RP Mar 16 08:11:23 and for the sumo build failures on Debian 10 buster, I think all needed fixes are in this series: https://patchwork.openembedded.org/patch/165261/ Mar 16 08:13:50 New news from stackoverflow: How to find out GCC -march and -mtune options on installed lubuntu? Mar 16 08:25:21 oh, or maybe I was just missing the uninative switch from bz2 to xz compression patch.. Mar 16 08:46:01 good morning Mar 16 08:48:41 morning Mar 16 09:11:53 morning Mar 16 09:14:04 New news from stackoverflow: What's the difference between IMAGE_INSTALL and CORE_IMAGE_EXTRA_INSTALL Mar 16 13:04:32 Can I have subdirectories in layer/classes directory? Mar 16 13:15:57 dev1990, yes, just set BBFILES in your layer.conf accordingly Mar 16 13:17:18 dev1990, not sure about subdirs in "classes" though Mar 16 13:17:37 my layer has only *.bb and *.bbappend files added to BBFILES, but bbclass'es still works, should I add *.bbclass ? Mar 16 13:17:40 ah ok Mar 16 13:17:45 kroon: BBFILES is only for recipes? Mar 16 13:19:06 dev1990: https://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html#creating-your-own-layer 3.create a layer confiugration file. I find the explanation confusing about BBPATH... But from the example it seems it's BBPATH/classes which is traversed? Mar 16 13:19:31 so if you had meta-me/classes/classes that might work but the question is why subdirs? Mar 16 13:21:12 maybe "inherit subdir/myclass" works Mar 16 13:21:13 qschulz: I just want to organizy stuff, I have classes that can be categorized so I just wondering to put them in subdirs. Mar 16 13:28:04 kroon: your idea works ;-) I'll probably go with it. Mar 16 13:28:36 oh that was a long shot Mar 16 13:29:35 interesting to know :) Mar 16 14:15:08 New news from stackoverflow: yocto recipe gives error -dev package contains non-symlink .so Mar 16 14:18:58 how to conditionaly include one SRC_URI or another (depend on global variable)? Mar 16 14:23:26 dv: ${@bb.utils.contains("VAR", if_true, if_false, d)}? but....... what exactly is your issue/need? Mar 16 14:25:19 qschulz, I have complex recipe with several includes and i is not convinient to have several copies of the same recipe to separate SRC_URI. Mar 16 14:26:23 so it could be easer to define variable like "bootloader_VER=DEV" in my local.conf and set SRC_URI conditionally Mar 16 14:26:55 dv: but they are effectively two different SW since you need two different SRC_URI right? Mar 16 14:27:37 dv: I mean techincally you could have two recipes with only require .inc and SRC_URI= in there) and one .inc file with all the rest. Mar 16 14:27:52 no. several groups of software share the SRC_URI, but built differently Mar 16 14:28:51 yeah that's what inline python is for Mar 16 14:29:21 dv: then isn't PACKAGECONFIG more what you want? if that is something you can do at configure step for example? Mar 16 14:29:42 qschulz, there are 5 recipes. 3 of them includes the same .inc (that include .inc with SRC) with, 2 includes only 0-level .inc with SRC_URI only Mar 16 14:33:05 dv: what exactly has to be different in SRC_URI? can't you add all the files in SRC_URI unconditionally and use a custom configure task or use PACKAGECONFIG or something? I usually don't like when a recipe's SRC_URI depends on external factors (except machine but I also don't like it that much) Mar 16 14:34:49 dv: I'm pretty sure the direct asnwer to your question is the first one I provided. But that might not be needed depending on what exactly needs to be done. Mar 16 14:40:24 qschulz, thank you for your help Mar 16 20:07:02 [RFC] what about making a form of build-appliance that, per multiconfig, incorporates the esdk of kind of target-image? Mar 16 20:24:41 Hello, I've noticed that Eclipse support has been removed since the 2.7 release. Is it possible to download the olde eclipse-plugins or to find instruction to use the SDK toolchain to build project via Eclipse ? Mar 16 20:28:37 lucile: i doubt that you will find anything better than https://git.yoctoproject.org/cgit/cgit.cgi/eclipse-yocto Mar 16 20:29:13 lucile: best bet is probably to start digging the eclipse cdt and gnuarm plugins, but its probably quite painful. Mar 16 20:34:19 LetoThe2nd: thanks, I'll look into it Mar 16 21:44:46 hi, is there some special handling when want to fetch sourec from private repo and ssh is not on port 22 but different one Mar 17 02:17:21 New news from stackoverflow: Can't not build ixgbe of meta-intel **** ENDING LOGGING AT Tue Mar 17 02:59:57 2020