**** BEGIN LOGGING AT Wed Sep 11 03:00:21 2019 Sep 11 04:39:32 RP: i see the mistake i made :(, must have lost that line when rebasing. I also see the issue with resultutils, already got a fix for that. And it looks like the rpath fix did not take, going to investigate locally more before i send a v2 of those changes. Sep 11 04:43:35 Let's say that I have package A which contains a file "file_a", that is installed manually with a custom do_install() function in the . Sep 11 04:44:21 bb file. If I create a .bbappend for this recipe in my own layer, which also contains a "file_a", will it install the "file_a" from my layer to the final image? Sep 11 04:47:59 iceaway, yes you can prepend to FILESEXTRAPATHS in your .bbappend Sep 11 04:48:39 iceaway, lots of .bbappend's do something like "FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:" Sep 11 05:28:22 kroon: cool, great! Sep 11 05:45:12 Hello. Can someone explain what the "1:" part means in the 3rd field of the following line from an image manifest? Sep 11 05:45:16 libpixman-1-0 armv7at2hf-neon 1:0.34.0-r0.0 Sep 11 05:45:36 The actual package name does not have this extra part. Sep 11 06:13:25 deuteron: https://www.yoctoproject.org/docs/2.4.2/ref-manual/ref-manual.html#var-PE Sep 11 06:17:30 silviof: Thanks! Sep 11 06:23:00 nrossi: issue with resultutils? Sep 11 06:23:12 nrossi: sadly it does look like the rpath fix didn't work :( Sep 11 06:31:01 LetoThe2nd: I haven't followed the stream, so can't tell with full context. Messing up is fine, it shows everybody can and maybe you even made the same mistakes as others did or will do. If you can follow-up with fixes and ideally explanations, that would be even more useful than sessions where everything goes smooth. My 2¢ :) Sep 11 06:31:10 nrossi: I'm going to be away for a couple of days, rburton should be able to rerun builds though Sep 11 06:36:36 qschulz: yeah , usually it is like that for my streams, no worries. Sep 11 06:38:39 RP: the issue with resultutil was me missing that there was two sections of code that extracted logs Sep 11 06:39:12 RP: have a good time with your days off :) Sep 11 06:40:50 hi, short question.. how do I add a custom library to the native sysroot, used during package compilation Sep 11 06:43:02 e.g. I have some software I would like to trace with lttng-ust. In order to build the software, lttng-ust headers/libs need to be present in the native sysroot Sep 11 06:45:42 Dracos-Carazza: DEPENDS = "lttng-ust"? Sep 11 06:48:15 worth a try :-) Sep 11 06:48:34 I'm still a beginner with respect to yocto/bitbake Sep 11 08:06:28 New news from stackoverflow: How to use variable(PV) from one .bb file to another bbappend file, when both files are in the samemetalyer? Sep 11 08:14:19 I have two yocto projects. The first project is my real build which goes on the emmc and is working. The second project is to flash the emmc via the sd card. Is there a way i can use a fixed path for the source location of the first projects sdimg. e.g. file:///home/buildserver/mainbuild-yocto/images/main-image-hub-nanopi.sdimg ? Sep 11 08:31:09 RP, a follow up question related to overrides, is it true in general that during variable expansion, all overrides are applied before all _append/_prepend/_remove are applied ? Sep 11 08:36:33 New news from stackoverflow: How to use variable PV from one .bb file to another bbappend file, when both files are in the samemetalyer? Sep 11 08:45:07 In .bb reciepes are the functions bash? Sep 11 08:54:14 In sh you would probably do ‘value=cat config.txt’ and bash ‘value=$( bentech, I wouldn't count on it Sep 11 08:55:11 bentech, easiest is just to test Sep 11 08:57:01 bentech, see https://www.yoctoproject.org/docs/latest/bitbake-user-manual/bitbake-user-manual.html#shell-functions Sep 11 08:58:45 Ah thanks, when i was searching i was looking for 'enviorment’ Sep 11 09:25:50 hmm, I have poky master based build stuck on a build machine. cooker process is trying to read from pipe which is just timing out. Any ideas what could be causing this? Sep 11 09:29:01 ah, never mind. Some java processes are stuck doing nothing in do_compile stage... Sep 11 10:14:32 <__angelo> hi, i have a BLK_DEV_LOOP [=m] in the defconfig, but for some reason, on final ext4 produced by sumo there is no such loadable module, nor built in. If any idea, welcome. Sep 11 10:16:46 __angelo: IMAGE_INSTALL += "kernel-module-whatsitcalled" Sep 11 10:17:06 kernel modules are only installed upon explicit request Sep 11 10:21:32 or add kernel-modules to the image, which pulls in *every* module built Sep 11 10:32:25 is anyone else seeing bitbake master leaving processes hanging around on every call? Sep 11 10:40:37 kergoth: vscode-remote is bloody awesome Sep 11 10:40:40 that is all Sep 11 10:41:02 once i realised that none of my extensions are in the remote bit, which is odd but fine, it's bloody awesome Sep 11 11:18:16 rburton: ++ Sep 11 11:21:42 rburton: since those arrived, almost my entire dev work is done through that, either by ssh or devcontainers. Sep 11 11:22:20 especially devcontainers are nice for software developmetn. no more cluttering your host with all those toolchains and related stuffs. Sep 11 11:24:16 i've always used ssh to the build machine but used to have netatalk on that so i can access the file system though my mac laptop Sep 11 11:25:23 i've had the "pleasure" to use a macbook for about two years. never ever again. Sep 11 11:26:28 i know all the world loves them. somehow their mindset jsut doesn't match mine Sep 11 11:31:28 Hi all, some here knows where some users share best practices about Yocto workflow ( I mean how to manage custom firmware versioning ) ? :) Sep 11 11:32:32 didn't we discuss that only a couple of days back already? Sep 11 11:34:34 Mmmh no mistake about my identity ? I have talked about Docker few days ago but that all (in my memories o.0 ) Sep 11 11:35:04 and btw thank, everything seem works with a shared DL_DIR and SSTATE_CACHE ;) Sep 11 11:35:24 no, i'm not sure at all. so, do you have any specific questions besides the usual ways of using repo, using kas or git submodules? Sep 11 11:37:06 No no my question is more about recipe's version, I would like to know if some ways are better to manage efficiently custom firmware :) Sep 11 11:37:17 <__angelo> LetoThe2nd, rburton in my image recipe i have MACHINE_ESSENTIAL_EXTRA_RRECOMMENDS += "kernel-modules" but only certain are added, not all Sep 11 11:37:34 PinkSnake: "custom firmware", as in? Sep 11 11:38:11 __angelo: that variable should be set by the machine, it's not for images to use Sep 11 11:38:19 in images, just extend IMAGE_INSTALL Sep 11 11:38:47 the usual recipe-local/conf-global applies :) Sep 11 11:39:18 MACHINE_ESSENTIAL_EXTRA_RRECOMMENDS is used by a packagegroup so overriding it in the image doesn't change anything Sep 11 11:40:58 LetoThe2nd In fact I have to manage yocto in order to build images for custom boards (10 max). Each board must embed some custom applications, and I don't know how to manage efficiently this app and i looking for some best practices Sep 11 11:46:19 <__angelo> LetoThe2nd, rburton thanks. Added kernel-modules to IMAGE_INSTALL_append, building ... :) Sep 11 11:47:36 PinkSnake: i don't get the question, sorry. are you asking about how to do configuration management for your custom application? or about bsp layers? feature layers? Sep 11 11:50:17 LetoThe2nd sorry I do not feel very comfortable with Yocto terminology ... It's more about custom application management ? because we can set preferred_version in conf/local.conf, distro.conf, machine.conf and i'm not able to find a doc about that ... Sep 11 11:52:05 PinkSnake: PREFERRED_VERSION is a way o select a specific version of a recipe. it implies that for this recipe a variety of version incarnations in your layers exist. Sep 11 11:52:35 PinkSnake: example: selecting kernel versions. Sep 11 11:54:01 PinkSnake: PREFERRED_PROVIDER on the other hand would switch a virtual target between several recipes that provide the desired functionality Sep 11 11:54:45 LetoThe2nd Ok thank you i think a got it :) Sep 11 11:58:36 PinkSnake: depending on your usecase those might or might not suit your needs, hard to tell for us. Sep 11 11:58:49 @let Sep 11 11:59:09 <__angelo> LetoThe2nd, rburton it worked. Strangely before only "net" part of modules was installed. But not important. Sep 11 11:59:22 <__angelo> thanks again Sep 11 11:59:48 __angelo: probably something set a dependency or install on some module in the net namespace, thats why they were pulled in. Sep 11 12:00:24 <__angelo> LetoThe2nd, aah, iptables :) Sep 11 12:00:38 <__angelo> great, thanks guys Sep 11 12:01:03 LetoThe2nd I understand that there is many ways to manage efficiently but I was hoping to find a "best workflow" :) Sep 11 12:01:56 PinkSnake: the "best workflow" depends. thats what i'm trying to say. Sep 11 12:02:44 LetoThe2nd Yes yes it's clear ;) No magic way to keep meta clean Sep 11 12:05:30 PinkSnake: are all the machines running the same software? Sep 11 12:07:10 New news from stackoverflow: Application and kernel modules deployment on Yocto Linux images Sep 11 12:09:20 rburton No... that's my problem, I have created a basic image based on custom distro to keep the same common base, after that i have created several image to add softwares. I have also remove MACHINE from conf/local.conf in order to path it as argument. I'm able to build different images with different softwares. But not sure that's a good way. Sep 11 12:09:20 There is no doc about workflow management because it's really user specific i mean Sep 11 12:12:27 PinkSnake: i mean is the difference between machines just a change of image Sep 11 12:13:40 rburton Yes bc I have to run same software on several boards, and some boards have some custom app added Sep 11 12:13:48 so if its all the same machine and distro, just a bunch of different images, then where's the problem? Sep 11 12:15:19 there is no problem i'm just looking for advice to keep close Yocto philo (y) Sep 11 12:15:56 nothing in local.conf, as much as possible in machine/distro/images Sep 11 12:16:05 local.conf should simply set machine and distro Sep 11 12:17:02 and other stuff like DL_DIR etc Sep 11 12:17:13 if you have a company-wide policy then site.conf is good for those Sep 11 12:17:22 rburton that's kind of advice i was talking about. thank you :) Sep 11 12:17:51 local.conf should be reserved for the few lines to set machine/distro, and then *the stuff you want to change right now* Sep 11 12:21:17 finally got scipy cross compiled. \o/ Sep 11 12:21:30 make that recipe public !!! Sep 11 12:21:41 tbh you could probably auction the recipe Sep 11 12:21:50 (y) Sep 11 12:22:32 i'll donate a beer. Sep 11 12:23:16 (service to the community, i actually don't care about scipy) Sep 11 12:24:37 hello everybody Sep 11 12:26:57 I have a question regarding kernel recipe. I hope that somebody will be able to help me. Sep 11 12:27:20 rburton: I'll clean it up and see what I can do to publish it, shouldn't be a problem. Sep 11 12:27:52 monkeyman79: jsut ask you question, and if somebody knows you'll get an answer. Sep 11 12:34:14 I need to add custom driver sources to yocto kernel, but it can't be out of tree module. I now, that the proper way is to add meta-mylayer/recipes-kernel/linux/linux-yocto/linux-xyz.bbappend and patches + config files, which I did initially and tested that it works, but I ended up with 3 megs patch file. Sep 11 12:35:27 I wonder if there is any way to put just separate .c source files into the files directory and make the build just copy it do specific location instead of patching the kernel sources Sep 11 12:35:46 it would be much easier to maintain for me Sep 11 12:35:56 monkeyman79: i would rather suggest to make up a recipe that refers to your custom kernel repository Sep 11 12:37:16 LetoThe2nd: I wanted to avoid that, I don't think that would be worth the effort Sep 11 12:37:17 New news from stackoverflow: Can i use cuda folder from meta-tegra master branch in thud meta-tegra branch? Sep 11 12:37:21 monkeyman79: just refer to the files in SRC_URI and set subdir to unpack it into the right place in the tree Sep 11 12:37:49 because that pathc file obviously has to come from somewhere, and usually that is you kernel tree where you jsut format-squash out a patch. so if that kernel tree already exists, there's nothing wrong with referencing it directly. Sep 11 12:38:11 SRC_URI+="file://mymodule.c;subdir=${S}/drivers/foo" Sep 11 12:38:37 (a proper tree where you can work on the file instead of having the source in the layer is better, though) Sep 11 12:38:57 rburton: (y) thanks, I'll try that Sep 11 12:46:27 hey rburton :) https://github.com/rossburton/customdistro thanks for this one! Sep 11 12:53:39 rburton: BTW, you @ ELCE? Sep 11 13:16:18 Hello. I'm updating my poky rocko with swupdate, but for some unknown reason the kernel is not updated from 4.9.59 to 4.9.80, only the modules .Does anyboy have an idea ? Sep 11 13:17:31 is the kernel loaded from the filesystem, or is it somewhere else in flash? Sep 11 13:18:08 it's on a rpi3, I think it's loaded from the filesystem. Sep 11 13:18:31 monkeyman79: you'll anyway need to edit the Makefile and maybe even the Kconfig for your driver to be compiled though. How is your new driver a 3MB patch if you know you only need one .c file? Sep 11 13:18:53 monkeyman79: and why is out-of-tree driver out of the question? Sep 11 13:23:27 Crofton|work: yep I find the uImage in /boot which is on the sdcard. Sep 11 13:23:27 jwwww: I assume you are using default SD card foermat then the kerneel is not in rootfs partition but on boot partition, and your updater might only be updating rootfs partition and not boot partition Sep 11 13:23:51 khem: yes you may be right. Sep 11 13:27:19 khem: yes you are right, it's because in sw-update's config there is no section about that boot partition. Sep 11 13:28:04 thanks a lot for the tip ! Sep 11 13:34:47 oh no.. layers.openembedded.org isn't responsive. Sep 11 13:38:25 LetoThe2nd: sadly not Sep 11 13:39:17 sad. Sep 11 14:16:40 rburton: I just sent out v2 patches to fix up issues with the previous ones. The numbers are out of sequence but the subject should match up. When you get a chance could you please update master-next and kick off an autobuilder run Sep 11 14:29:41 nrossi: fired (I'm not here really) Sep 11 14:29:45 (thanks) Sep 11 14:30:08 RP: thanks :) Sep 11 15:07:42 New news from stackoverflow: cannot su to new user added in yocto build [closed] Sep 11 15:26:40 RP: go away! Sep 11 15:37:47 New news from stackoverflow: u-boot bbappend causes file system corruption Sep 11 15:52:26 evening, I'm having errors while compiling swupdate in poky-rocko.It used to compile fine but I enabled feature ubifs and compilation fail with error : | cc1: fatal error: include/generated/autoconf.h: No such file or directory .the full log is there http://dpaste.com/3DEYGB4 can somebody help please ? Sep 11 18:02:13 this might not be the best place to ask this, and maybe I should find a cmake channel.. but how do I stop cmake from passing rpath options to the linker? Sep 11 18:03:31 I have another NoGood Python package which also uses cmake to build some c++ code. It cross compiles just fine but the linking step includes explicit rpath directives targeting stuff in the recipe-sysroot-native directory which obviously does not work. Sep 11 18:03:53 I used set(CMAKE_SKIP_RPATH true), but it seems to be ignored? Sep 11 19:00:38 rburton, you around? Sep 11 19:14:38 khem, RP: were there specific reasons or any discussions our ARM32 toolchain doesn't have '-gnueabihf' (with "hf") suffix when "callconvention-hard" is set in tunes? Sep 11 19:36:37 adelcast, late for him :) Sep 11 19:38:37 cmake is infuriating at times Sep 11 19:52:55 Crofton|work: yeah, don't blame him, =) Sep 11 20:39:42 denix: yes it was discussed and IIRC I send some patches as well but our HF/SF story existed before it was coined in distro world Sep 11 20:40:44 secondly, we wanted a unified toolchain for arm regardless of float ABI and let gcc pick the right multilib Sep 11 20:42:15 since then a lot has changed and now I See that gnueabihf is used quite a lot by packages to decide features etc. Sep 11 20:43:05 in some places we patch it so that we can check for compiler cmdline but thats a battle Sep 11 20:52:47 khem: I see, thanks. are there any archives of the discussions for reference? Sep 11 20:56:01 * armpit brains are funny things. read gnueabihf as Gandalf make the sentence more understandable ; ) Sep 11 20:56:56 armpit: lol Sep 11 20:57:29 denix: dont have links handy Sep 11 20:57:42 it could have been IRC discussion too like this one Sep 11 21:03:47 khem: ok, thanks Sep 11 22:10:28 I ran into a new series of problems with usrmerge and multilibs. Sep 11 22:11:13 For some reason the QA check fires with usrmerge only, but the same check passes without usrmerge. Sep 11 22:11:22 For example building lib32-systemd Sep 11 22:11:29 ERROR: lib32-systemd-1_239-r0 do_package_qa: QA Issue: /usr/lib/kernel/install.d/90-loaderentry.install contained in package lib32-systemd-kernel-install requires /bin/bash, but no providers found in RDEPENDS_lib32-systemd-kernel-install? [file-rdeps] Sep 11 22:11:57 I don't think the right answer is to add an RDEPENDS for "bash" to everything. Sep 11 22:12:28 The file it trips up on always passes if you build without usrmerge. Sep 11 22:12:43 I just don't know where to look to find/fix/understand the problem. :-) Sep 11 22:43:27 adelcast: was doing dad stuff, now bed. email? Sep 11 23:18:09 jwessel: https://bugzilla.yoctoproject.org/show_bug.cgi?id=9217 Sep 11 23:18:10 Bug 9217: normal, Medium, 3.99, Martin.Jansa, RESOLVED FIXED, Many unsolveable QA warnings from build-deps and file-rdeps **** ENDING LOGGING AT Thu Sep 12 03:02:41 2019