**** BEGIN LOGGING AT Tue Jan 21 02:59:56 2020 Jan 21 07:21:36 New news from stackoverflow: Nvidia Jetson Nano with docker Jan 21 07:32:01 Hello guys, after Yocto update to Zeus branch, i cannot build anymore a cutom binary (based on autotools) because of opencv version is 4.0.1 and all libraries has been moved to /usr/include/opencv4/opencv2 ... , any idea ? Jan 21 07:33:17 PinkSnake: like i said, fix package Jan 21 07:33:23 pkg-config is your friend Jan 21 07:35:38 Yes, I have already set DEPENDS=" opencv", but the configure step still stuck "No package 'opencv' found Jan 21 07:38:06 again, look at configure.ac and Makefile.am to make sure pkg-config is evaluated properly Jan 21 07:38:53 I have to fix PKG_CHECK_MODULES first isn't it ? Jan 21 07:39:07 if the package is not found no chance to build Jan 21 07:40:35 then the next step is to find out if opencv changed something there. Jan 21 07:41:19 the big change is opencv 3.4 -> to 4.0.1 and all prefix pathes change Jan 21 07:41:38 but maybe they also changed the pkgconfig nmae? Jan 21 07:43:22 start a devshell for your package maybe and start poking pkgconfig there. Jan 21 07:47:00 poll here: https://twitter.com/YoctoThe/status/1219515995288633344 Jan 21 07:47:26 (will repeat a couple of times over the day. you have been warned) Jan 21 07:51:20 Why can't I select them all ? :c Jan 21 07:51:36 kriive: no idea, ask twitter. Jan 21 07:51:37 New news from stackoverflow: How to compile c++ files using Makefile Jan 21 07:52:41 grrrr twitter Jan 21 08:06:12 can i vote somewhere else but twitter :P Jan 21 08:07:57 stuom1: you can just fix bugs and see what happens! Jan 21 08:09:54 I fixed one bug already, now I'm waiting for my prize Jan 21 08:10:34 stuom1: as you can see, were in the process of deciding on which prize you'll get :) Jan 21 08:28:26 Guys I don't really get the difference between bitbake core-image-minimal -c populate_sdk_ext and bitbake meta-toolchain-qt5 Jan 21 08:29:17 Wouldn't the first also include Qt stuff? Like qmake, libs... Jan 21 08:32:04 Ok, already found the answer by myself, should fine tune my google search next time before calling it quit Jan 21 08:44:33 LetoThe2nd: on a personal level, I don't care too much about recognition (c.f. your tweet). But I was writting the "new release of Linux kernel, here are our contributions, we did Xth in contributions in commits and Yth in lines as a company" in my previous company Jan 21 08:45:03 LetoThe2nd: and I guess you can have kind of a free ad for Yocto with that kind of articles every ~6 months Jan 21 08:46:00 Companies usually like to boast their contributions. Attract people interested in upstreaming stuff and show "we're among the good guys" Jan 21 08:57:45 qschulz: click "contrib ranking" then, because thats exactly what it means Jan 21 08:58:20 qschulz: but twitter limits the selections to 25 characters, so i had to be really brief. Jan 21 09:06:12 LetoThe2nd: I just meant that it wouldn't encourage me to contribute on a personal level. Company level is a different thing, especially the consulting companies. Jan 21 09:06:20 LetoThe2nd: I clicked before I asked :) Jan 21 09:06:28 s/asked/told/ Jan 21 09:07:38 qschulz: :) Jan 21 09:10:48 RP: coming from the Linux kernel community, what I'm missing for the few patches I contributed is the lack of review (we don't have to be as nitpick-y as in the kernel though :) ) and the one thing that annoys me most is I don't know if a patch was merged or not if I don't check myself in the git repo. Kernel maintainers *usually* (not all, not always) send a prepared mail saying the patch is now merged. Jan 21 09:12:55 and I think the project is struggling because upgrading Yocto is not as critical as upgrading the kernel and people are ready to take the maintenance burden on themselves since upgrades are so rare, it's also usually easier to maintain them compared to the kernel since there's no rebasing involved, just quickly testing. Jan 21 09:13:51 (just giving my 2¢, YP is technically the 3rd community I'm joining after the kernel and u-boot, so all new to those things, not a lot of experience, etc.) Jan 21 09:15:54 qschulz: interesting that you're mentioning u-boot and the kernel. my minor contributions to the kernel and u-boot were just the reverse. never got an acceptance mail from the kernel. got them from u-boot, however very late as the patches were super small and not urgent. Jan 21 09:18:04 LetoThe2nd: which subsystem for the kernel patches? Jan 21 09:18:14 qschulz: some board support, minor drm Jan 21 09:18:28 qschulz: but all in all, really minor. i mostly have drive-by fixes to submit. Jan 21 09:18:47 Then maybe I was lucky enough to fall in the "nice" subsystems :) Jan 21 09:20:30 forget about the comparison then. let's say that I appreciate knowing if my patches are taken or not, and I appreciate the reviews because it is in some forms knowledge sharing. Jan 21 09:21:01 qschulz: thats a good conclusion, yes. Jan 21 09:22:02 New news from stackoverflow: Yocto: check final content of MACHINE_FEATURES Jan 21 09:38:04 LetoThe2nd pkg-config failed but headers has been deployed in /tmp/work/cortexa9t2hf-neon-poky-linux-gnueabi/bin/1.0-r0/recipe-sysroot/ o.0 ? Jan 21 09:38:26 PinkSnake: then find out why pkg-config fails. thats the root of eveil for you, i'd say. Jan 21 09:52:07 New news from stackoverflow: Yocto sato image is running extremely slow, is there a way to improve its performance? Jan 21 10:38:56 ndec: howdy! jsut a short question, what do you use for trimming the videos? Jan 21 10:46:54 qschulz: Hi. It has been a while, but I've tried your suggestion to add "SECCOMP_FILTER" to the kernel. I tried to search it with "/" and found this: Jan 21 10:47:02 Symbol: HAVE_ARCH_SECCOMP_FILTER [=n] │ │ Type : boolean Jan 21 10:47:03 │ │ Defined at arch/Kconfig:325 │ │ Selected by: ARM [=y] && AEABI [=y] && !OABI_COMPAT [=y] Jan 21 10:47:18 But I can't find it on the menu options. Can you help me? Jan 21 10:47:41 It says: `Defined at arch/Kconfig:325` - Where is it? Jan 21 10:47:52 well... in arch/Kconfig file :D? Jan 21 10:47:57 line 325? Jan 21 10:48:16 but it just means that your architecture does not support SECCOMP_FILTER apparenlty Jan 21 10:48:17 Without using "menu"? Jan 21 10:48:44 qschulz: Why does not support? Jan 21 10:48:46 the only unmet condition is !OABI_COMPAT [=y] Jan 21 10:49:06 which means..? Jan 21 10:49:24 well, OABI_COMPAT is y but should be no so that HAVE_ARCH_SECCOMP_FILTER is yes Jan 21 10:49:54 https://elixir.bootlin.com/linux/latest/source/arch/Kconfig#L438 Jan 21 10:50:40 https://lore.kernel.org/patchwork/patch/418217/ Jan 21 10:51:02 disable OABI support if possible Jan 21 10:51:04 afk Jan 21 10:51:51 thanks. I will search how Jan 21 10:52:46 maybe you can disable it from menuconfig as well Jan 21 10:53:15 where `arm/kconfig` file is stored? Jan 21 10:53:38 arch* Jan 21 10:53:56 nacknick: in the sources of the kernel Jan 21 10:54:28 found OABI and disabled Jan 21 10:54:40 what next? Jan 21 10:55:45 need to change *kconfig* as well? or disabling OABI it's enough...? Jan 21 10:56:50 In this file *build/tmp/work-shared/raspberrypi3/kernel-source/arch/Kconfig* at line 325: Jan 21 10:57:01 config HAVE_ARCH_SECCOMP_FILTER bool help An arch should select this symbol if it provides all of these things: - syscall_get_arch() - syscall_get_arguments() - syscall_rollback() - syscall_set_return_value() - SIGSYS siginfo_t support - secure_computing is called from a ptrace_event()-safe context - secure_computing return value is Jan 21 10:57:01 checked and a return value of -1 results in the system call being skipped immediately. - seccomp syscall wired up Jan 21 10:57:17 if it is the correct file... Jan 21 11:32:19 qschulz: OK. It's configured automatically by the boolean conditions Jan 21 12:00:34 nacknick: exactly, so you would need to disable OABI to be able to have HAVE_ARCH_SECCOMP_FILTER to yes and then be able to select SECCOMPT_FILTER Jan 21 12:01:09 if that is all, you would need CONFIG_OABI=n AND CONFIG_SECCOMP_FILTER in the .scc file (provided these are the only things needed) Jan 21 12:16:46 Hello, i have a question. Extensible SDK built for my image has (sdk_install)/sysroots/x86_64-pokysdk-linux/usr/bin/python3.real binary which is not relocated. The problem is that the directory (sdk_install)/sysroots/x86_64-pokysdk-linux/usr/bin is first in PATH and devtool command is not working. I compared it to eSDK for core-image-minimal and this directory has a lot less files and no python3 there. And then proper (sdk_install)/buildt Jan 21 12:16:46 ools/sysroots/x86_64-pokysdk-linux/usr/bin/python3 is used. What can be reason that my eSDK has python(and some more) binaries in (sdk_install)/sysroots/x86_64-pokysdk-linux sysroot? Jan 21 12:18:38 Randal: maybe something gets magically pulled in? how does your image and therefore the esdk relate to core-image-minimal? Jan 21 12:21:47 qschulz: It's working! Thank you so much! Jan 21 12:27:19 nacknick: glad it works :) Jan 21 12:30:27 LetoThe2nd: my image inherits from core-image, and that is the only common to core-image-minimal. I have own packages installed by IMAGE_INSTALL but there is no package related to python3. Jan 21 12:33:31 Randal: but nothing specifically added to the sdk through TOOLCHAIN_TASK or whatever? Jan 21 12:38:31 LetoThe2nd: no, I jest grepped all nonstandard layers with TOOLCHAIN_HOST and nothing is there Jan 21 12:39:26 Randal: if i had to guess then something in the things you added to IMAGE_INSTALL is not clean. i'd remove all, add one by one and see which makes it break. Jan 21 12:42:50 LetoThe2nd: that is what I started to do, we will see, thanks. Jan 21 12:53:51 kanavin: thanks, that round of patches was a bit smoother :) Jan 21 12:54:28 kanavin: I didn't take some of them either I wanted to think more. Not sure removing is a good thing Jan 21 12:54:34 (the ptests) Jan 21 12:57:41 here we go again, everybody: take your vote! https://twitter.com/YoctoThe/status/1219515995288633344 Jan 21 12:57:56 RP, right, but on the other hand keeping @expectedfail isn't great either, as ptests keep regressing and no one notices Jan 21 12:58:30 RP: kanavin: Referring to lttng-tools? Jan 21 12:58:52 tgamblin, yes Jan 21 13:03:30 kanavin: I'd mush rather teach the QA code if something was a regression Jan 21 13:03:58 kanavin: i.e. have a "must not regress" list somewhere Jan 21 13:04:29 kanavin: realistically some aren't going to 100% pass and I don't want that to be the bar Jan 21 13:04:42 RP, then we might split ptest into ptest-expectedfail and ptest-mustnotfail maybe? Jan 21 13:05:12 kanavin: maybe, not sure how that would look but it could work Jan 21 13:05:45 RP, is the output from ptest (e.g. what failed) processed further? Jan 21 13:06:36 kanavin: Ah. I've been looking at that too. We're seeing a hang on lttng-tools ptest after PASS: tools/rotation/test_schedule_api 26 - Listing returned 0 rotation schedules after removal Jan 21 13:11:21 This is the (highly-condensed) version of the log reported by one of our testers https://pastebin.com/vZy0Uvde Jan 21 13:28:09 kanavin: yes, by resulttool **** BEGIN LOGGING AT Tue Jan 21 13:31:00 2020 Jan 21 13:32:56 kanavin: See https://autobuilder.yoctoproject.org/typhoon/#/builders/83/builds/651 "Send QA Email" at the end Jan 21 13:33:17 kanavin: run by the script in yocto-autobuilder-helepr Jan 21 13:34:02 RP, right, so the 'allowed to fail' check can happen at that step? Jan 21 13:34:43 kanavin: yes, that code is meant to do regression analysis by comparing against previous builds Jan 21 13:34:57 kanavin: sadly its not quite working and never has :/ Jan 21 13:35:07 The previous builds results are in git Jan 21 13:42:13 I'm trying to bake multiconfig with "bitbake multiconfig::image-name multiconfig:qemuarm64:image-name" but it still starts building both with the MACHINE defined in local.conf Jan 21 13:42:25 and not in qemuarm64.conf Jan 21 13:42:35 what could I be doing wrong? Jan 21 13:44:12 stuom1: how are you setting MACHINE? Jan 21 13:44:22 "?=" or "=" ? Jan 21 13:44:27 ?= in both Jan 21 13:45:19 I'm trying to build openjdk-8 in my build, but I'm erroring out with "g++: error: unrecognized command line option ‘-fmacro-prefix-map=.../build/tmp-glibc/work/core2-64-wrs-linux/openjre-8/172b11-r0=/usr/src/debug/openjre-8/172b11-r0’" Jan 21 13:46:25 -fmacro-prefix-map appears to be a gcc version 8+ preproccesor feature that I don't have using gcc 7 Jan 21 13:46:34 stuom1: there is your problem. Try "=" Jan 21 13:47:06 I unfortunately cannot upgrade gcc. Is there any way of building without the flag? Jan 21 13:47:15 thanks RP, it is working now! Jan 21 13:47:16 Nathan22: why is it using the native compiler for something which should use the target compiler? Jan 21 13:47:27 Fantastic question Jan 21 13:57:18 In multiconfig .conf file, should I only set variables that change and the rest comes from local.conf? Jan 21 14:03:40 Do I need to force/clean with bitbake after applying a patch with git patch ? Jan 21 14:13:38 Nathan22: It shouldn't be necessary if you mean applying a git patch to one of your layers Jan 21 14:14:45 Damn. Yeah that's what I meant, didn't fix the issue regardless. Thanks anyway though Jan 21 14:52:12 I am using TOOLCHAIN_TARGET_TASK to add pkgA-staticdev to the SDK, but I get both pkgA-staticdev and pkgA (Which contains a target binary that I don't want in the SDK) in the SDK sysroot, is there a way to only get pkgA-staticdev in the SDK? Jan 21 15:10:32 hpsy: you'd probably need to break the dependency: RDEPENDS_${PN}-staticdev_pn-XXX = "" Jan 21 15:26:37 @hpsy: it it a standard SDK or an extensible SDK? Jan 21 15:28:45 @hpsy: in case it's a standard SDK you might get away with it via package splitting. How did you get pkgA-staticdev in there? Jan 21 15:37:32 Is there any way to specify that I want to use systemd in a image other that setting it in local/conf? I'd like to put it in my image recipe Jan 21 15:38:03 conf/local.conf* Jan 21 15:38:44 kriive: nah, it belongs into the DISTRO Jan 21 15:39:12 kriive: https://twitter.com/YoctoThe/status/1217166071519744000 Jan 21 15:39:28 that's not actually true. a distro can enable systemd, sysvinit, or both, if both, the image can select which it wants to actually install Jan 21 15:39:34 it's not a common usage, admittedly :) Jan 21 15:41:16 kergoth: yup. it just doesn't work the other way round - image defining it for all the rest of the build Jan 21 15:41:44 kriive: i totally encourage you to look at the live coding session on MACHINE, DISTRO, IMAGE :) Jan 21 15:41:53 it's true. they did ask how to specify if they want to use it in the image, though, not necessarily enable its support globally Jan 21 15:42:06 agreed, very important to grasp those and how they fit together Jan 21 15:43:01 LetoThe2nd: thank you so much! I will catch up with those videos after I manage to boot this crappy piece of metal also know as up board Jan 21 15:43:30 btw i love my upboard Jan 21 15:43:31 kriive: believe me, the other way round it would be a more efficient way to invest 2 or 3 hrs. Jan 21 15:44:08 ok, i'm going to follow your suggestion Jan 21 15:47:58 mcfrisk: http://lists.openembedded.org/pipermail/openembedded-architecture/2020-January/001776.html Jan 21 15:52:29 kriive: just think about how long it took you to try and ask the systemd question. Jan 21 15:55:32 LetoThe2nd: too much, just 5 minutes in the video and I understood much more than two hours table head banging Jan 21 15:56:25 kriive: thats why i keep on saying, what the videos. Jan 21 15:58:04 YPTM - armin is on Jan 21 16:00:20 YPTM - Scott is on Jan 21 16:02:07 vmeson: https://autobuilder.yocto.io/pub/non-release/20200120-11/testresults/testresult-report.txt Jan 21 16:02:15 vmeson: https://autobuilder.yocto.io/pub/non-release/ Jan 21 16:02:24 vmeson, https://wiki.yoctoproject.org/wiki/BuildLog Jan 21 16:03:05 vmeson: https://autobuilder.yocto.io/pub/non-release/20200119-10/testresults/testresult-report.txt has a fail Jan 21 16:03:34 YPTM - Jan-Simon is on Jan 21 16:07:09 RobertBerger: It is a standard SDK, I added pkgA-staticdev to the SDK explicitly using the TOOLCHAIN_TARGET_TASK, do you mean to make the main package (pkgA) empty? Jan 21 16:07:38 yep Jan 21 16:08:21 @hpsy: if you inspect your packages what is inside? Jan 21 16:09:34 https://www.yoctoproject.org/ shows 3.0.1 as latest stable, but link on https://www.yoctoproject.org/software-overview/downloads/ is 3.0.0 only Jan 21 16:09:35 @hpsy: oe-pkgdata-util can help. If the stuff you don't want is not in the -staticdev package someone pulls it in and you can figure out why Jan 21 16:10:52 vmeson: https://bugzilla.yoctoproject.org/show_bug.cgi?id=13664 Jan 21 16:10:54 Bug 13664: normal, Medium+, 3.1 M2, hongxu.jia, NEW , _signal_clean_handler failure in oe-selftest Jan 21 16:12:04 @hpsy: I would first check with oe-pkgdata-util the content of the packages and if pkgA is indeed what you don't want see who pulls it in. Quick and dirty you could try to make it empty via package split. (and allow empty packages to be added) Jan 21 16:12:31 and one more time: please take your vote! https://twitter.com/YoctoThe/status/1219515995288633344 Jan 21 16:13:31 https://bugzilla.yoctoproject.org/show_bug.cgi?id=13604 Jan 21 16:13:32 Bug 13604: normal, High, 3.1 M2, paul.eggleton, NEW , [master-next] Distrodata.test_maintainers fails Jan 21 16:16:45 there we go, got my windows terminal settings into my dotfiles finally Jan 21 16:17:37 @RobertBerger: pkgA contains a binary (Which I don't want in the SDK), while pkgA-staticdev contains a library that I need in the SDK (I got the content information by looking at packages-split in the work directory), both packages get pulled into the SDK when I add ' TOOLCHAIN_TARGET_TASK += "pkgA-staticdev" ' , I will try to inspect the packages with oe-pkgdata-util Jan 21 16:21:41 @hpsy: OK just make sure that it's really like that;) Wait maybe you should try something else. Normally if you do TOOLCHAIN_TARGET_TASK_append = " pkgA-staticdev" it does not automatically pull in pkgA. So maybe you have some dependency you should not have. Jan 21 16:22:37 @hpsy: what are the dependencies for pkgA-staticdev? Jan 21 16:23:27 New news from stackoverflow: Why doesn't Yocto generate the tarball when BB_GENERATE_MIRROR_TARBALLS is set? Jan 21 16:34:32 RP: https://www.yoctoproject.org/ shows 3.0.1 as latest stable, but link on https://www.yoctoproject.org/software-overview/downloads/ is 3.0.0 only Jan 21 16:38:11 @RobertBerger: Nothing depends on it at the moment Jan 21 16:39:08 @hpsy: IMAGE_INSTALL maybe pulls it in? Jan 21 16:41:17 @RobertBerger: No Jan 21 16:43:32 @hpsy: Jut to summarize: In your local.conf or so you add TOOLCHAIN_TARGET_TASK_append = " pkgA-staticdev". It ends up in the SDK. It should not end up in the rootfs. But also pkgA magically appears in the SDK and also the rootfs? Jan 21 16:44:58 RP: it looks like it is as you said, adding RDEPENDS_${PN}-staticdev = "" to the recipe removes pkgA from the Jan 21 16:45:04 SDK Jan 21 16:45:21 Fedora badge system: https://badges.fedoraproject.org/ -- doesn't appeal to me but it seems to be popular in Fedora. Jan 21 16:46:10 @RobertBerger: not in the rootfs, only in the SDK Jan 21 16:49:11 @hpsy: Well there is a runtime dependency in the recipe you just removed. This means that the -staticdev package now does not depend on anything. Jan 21 16:50:36 @hpsy bitbake pkgA -e | grep ^RDEPENDS or so should tell you more Jan 21 16:51:35 @hpsy: without emptying out RDEPENDS_${PN}-staticdev what does it contain? Jan 21 16:53:16 RP: but I don't get the "_pn-XXX" part of your answer (RDEPENDS_${PN}-staticdev_pn-XXX = "") should it be pn-pkgA, is this an override of some kind? Jan 21 16:56:55 @RobertBerger: Yeah, I don't know why I didn't use bitbake -e, I always use it for these kind of issues. It looks like an implicit dependency that bitbake adds it goes pkgA-staticdev -> pkgA-dev -> pkgA Jan 21 16:56:55 # $RDEPENDS_pkgA-staticdev [2 operations] Jan 21 16:56:55 # set /ws/poky/meta/conf/bitbake.conf:319 Jan 21 16:56:55 # "${PN}-dev (= ${EXTENDPKGV})" Jan 21 16:56:55 # rename from RDEPENDS_${PN}-staticdev data.py:117 [expandKeys] Jan 21 16:56:56 # "${PN}-dev (= ${EXTENDPKGV})" Jan 21 16:56:56 # pre-expansion value: Jan 21 16:56:57 # "${PN}-dev (= ${EXTENDPKGV})" Jan 21 16:56:57 RDEPENDS_pkgA-staticdev="pkgA-dev (= 1.0-r0)" Jan 21 16:57:37 @hpsy: mystery solved ;) Jan 21 16:58:39 @RobertBerger: Yes, Thank you :) Jan 21 16:59:11 on the YP tech call, someone mentioned AGL developer interviews. Could you post a link, my google foo isn't working. Jan 21 16:59:17 @hpsy: you're welcome! Jan 21 16:59:19 vmeson: https://www.automotivelinux.org/blog/2019/11/26/developer-showcase-barker Jan 21 16:59:32 Just trying to find contact details for the advocacy team Jan 21 17:00:00 YPTC: minutes: https://docs.google.com/document/d/1ly8nyhO14kDNnFcW2QskANXW3ZT7QwKC5wWVDg9dDH4 Jan 21 17:00:07 paulbarker: thanks. Jan 21 17:00:37 YPTM- is over Jan 21 17:02:23 Looks like I need to join yocto-advocacy to email the list Jan 21 17:03:30 Is there someone on the advocacy team I can hand off the developer interviews idea to? Jan 21 17:03:33 hi, i pushed a few patches to bitbake. the ones for removing unused imports and handling singleton comparisons properly got merged, but my deprecated methods removal didnt. i got a response for the ones that got merged as they needed amending, but i think my deprecated methods removal didnt because the commit message wasnt properly formatted, but Jan 21 17:03:33 it did work. is it ok to send it again with the message correctly formatted, or is it not wanted? Jan 21 17:23:40 New news from stackoverflow: Is it possible to create a docker image from a yocto generated image? Jan 21 18:40:40 Hi.. I have a package which is licensed as MIT/X11 but does not provide a license file. I know that it's possible to select the license file and checksum from files included within yocto/bitbake, but can't seem to find an example or the proper search terms. Jan 21 18:43:07 fullstop: How do you know it's MIT/X11 licensed if they don't provide a license file? Jan 21 18:43:21 fullstop: See, e.g., systemd-conf in OE-Core. It uses LIC_FILES_CHKSUM = "file://${COREBAS}/meta/COPYING.MIT;md5=3da9cfbcb788c80a0384361b4de20420" Jan 21 18:43:22 That sounds like you should raise an upstream issue Jan 21 18:44:13 fullstop: But I do agree with paulbarker Jan 21 18:44:41 paulbarker: The last upstream commit was 7 years ago. Jan 21 18:45:00 (And "${COREBAS}" above should of course be "${COREBASE}".) Jan 21 18:45:12 paulbarker: the "rockspec" guy (It's a Lua thing) specifies the license as MIT/X11 Jan 21 18:46:01 wait, there is a copyright file which looks similar to MIT Jan 21 19:40:38 final reminder for (my) today: https://twitter.com/YoctoThe/status/1219515995288633344 Jan 21 19:40:41 take your vote! Jan 21 20:10:40 Who is responsible for the initial hddimg boot menu? Initramfs? Because there is a bug in the installation script: reboot utility is not found Jan 21 21:10:14 trying here: I have been trying to create a recipe to remove /etc/rc5.d/S20hwclock.sh Jan 21 21:10:32 how is something like this usually accomplished? Jan 21 21:24:52 alesan: Find the recipe which creates that file and either modify it with a bbappend or don't install the offending package into your image Jan 21 21:25:06 Already gone haha Jan 21 21:27:26 paulbarker. too slow! ;) Jan 21 22:52:19 when should i use require and when should i add a new MACHINE_FEATURE? is there any major difference between the two methods? Jan 21 22:54:33 i suppose MACHINE_FEATURE is more appropriate when it needs to be read in multiple recipes? Jan 21 23:56:47 hey there, as far as i understood from the documentation it should be sufficient to add a defconfig file to the SRC_URI just like normal patches wo get it activated Jan 21 23:57:30 i did so, but when i run '-c menuconfig' i see only the patches applied, but not activated (which they should be through the defconfig). anything else i am missing here? Jan 22 00:18:08 also, im confused by how kernel-module-split and RPROVIDES works Jan 22 00:19:19 if recipe foo-modules.bb creates kernel-module-foo and i add kernel-module-foo to RDEPENDS, how does bitbake know that foo-modules.bb makes kernel-module-foo if foo-modules.bb wasn't also added to RDEPENDS? Jan 22 00:23:32 mischief: i am not sure, but as far as i understood, 'require' and 'MACHINE_FEATURE' are not achiving the same thing! Jan 22 00:24:31 require is just an explicit way of including settings fomr another file into your recipe. with the difference to a simple 'include' that it would fail in case the file to be included is not found! Jan 22 00:25:29 creich: i have a simple case - i want to MACHINE_ESSENTIAL_EXTRA_RDEPENDS += "kernel-module-foo" and KERNEL_MODULE_AUTOLOAD += "foo" Jan 22 00:25:59 so e.g. if you write some recipe and split it in some kind of generic part like your_recipe.inc and some version specific part like your_recipe_0.1.0.bb you can make the .inc a requirement within the bb file Jan 22 00:26:25 should i put that in a .inc and require it, or should i also wrap it in ${@bb.utils.contains('MACHINE_FEATURES', 'foomod',...)}" and set MACHINE_FEATURES += "foomod"? Jan 22 00:27:01 i still feel that you kind of mix up two concepts here Jan 22 00:27:27 wether or not you put it in an .inc is part of your development process. Jan 22 00:28:07 since i am not sure how to handle kernel_modules, i can not really help with the other thing, sorry Jan 22 01:00:46 ok, now i figured out that my defconfig file would be used when i create a subfolder 'poky' since it seems like bitbake looks there first. only putting it into the folder i added with FILESEXTRAPATHS_prepend did not help! Jan 22 01:01:21 so here are two questions. 1) how can it be, that my added path is not searched first, since i put _prepend to it? Jan 22 01:02:26 2) is there any way to get the folders listed that bitbake will search? i just used another bug-log to figure the hierarchy and then created the new folder Jan 22 01:02:32 thx in advance :) Jan 22 02:59:19 perhaps show the code which is failing can give better idea Jan 22 03:00:16 poky is DISTRO so any value that is DISTRO will be appended to path and so will be MACHINE, if you add poky dir then you are effectively saying this defconfig is only applicable when building for poky distro **** ENDING LOGGING AT Wed Jan 22 03:01:07 2020