**** BEGIN LOGGING AT Thu Apr 23 02:59:58 2020 Apr 23 05:43:50 New news from stackoverflow: How to inherit a class globally in the image recipe - yocto Apr 23 07:52:43 hi; i am currently building a yocto recipe for azure walinuxagent (using python3-setuptools). somehow when building it always misses the "_crypt" module when importing "crypt" which should all be part of the the python core libraries Apr 23 07:53:46 does somebody have a clue how to fix this issue? Apr 23 07:58:20 drkhsh: I had some issue with it (with paramiko I think) and problem was, that it was not supported for python release I used Apr 23 07:59:51 mattovsky: hmm maybe; i will check that out! thx Apr 23 08:37:29 Hello everybody Apr 23 08:52:31 Today I'm struggling with SDK generation: I would like to add to my SDK some coding tools, like clang-tidy (and ccache, cppcheck, doxygen, ...); I need to run them on my host computer, mainly to analyze code that will be crosscompiled. It's correct to add nativesdk-clang (as example) to MACHINE_EXTRA_INSTALL_SDK_HOST? Apr 23 08:54:29 mous16: Yes, but do note that meta-clang doesn't ship clang-tidy by default Apr 23 08:56:00 mous16, TOOLCHAIN_HOST_TASK iirc. Havent heard of MACHINE_EXTRA_INSTALL_SDK_HOST before Apr 23 09:03:51 kroon: sorry, the variable I use is introduced by boot2qt (which I'm working with); it ends up adding specified packages to RDEPENDS of a recipe listed in TOOLCHAIN_HOST_TASK, so I think it does the trick. Apr 23 09:07:21 neverpanic: it appears that clang-tidy, with all other clang tools, is added to sysroots/x86_64-pokysdk-linux/usr/bin/, but have some relocation problems so I couldn't run it. I think I ended up in this trap: https://github.com/kraj/meta-clang/issues/119 Apr 23 09:17:26 Hi all, I'm fighting with the OOM-killer in the `populate_sdk` phase. It kills that `tar` process, pbly. due to xz memory usage. I use ipk packages and already tried limiting PARALLEL_MAKE=" -j1", but that didn't help. How do I reduce the memory usage of the `tar` process in the `populate_sdk` command? Apr 23 09:21:41 problame, git grep XZ_ in oe-core, there are some variables you can tune Apr 23 09:26:06 problame: How much RAM do you have? Apr 23 09:52:26 For local development the default xz args OE uses are quite costly. I use gz or zstd lately to speed builds up during development. Apr 23 09:58:24 kroon: sry for the late reply: I found kroon: Apr 23 09:58:37 XZ_DEFAULTS ?= "--memlimit=50% --threads=${@oe.utils.cpu_count()}" Apr 23 09:59:11 shouldn't that be : XZ_DEFAULTS ?= "--memlimit=${@"{}".format(50/oe.utils.cpu_count()}% --threads=${@oe.utils.cpu_count()}" Apr 23 10:05:43 problame, that wouldn't look reasonable to me Apr 23 10:06:21 i'd expect the memlimit to apply to the whole process Apr 23 10:14:44 New news from stackoverflow: Yocto recipe for perl lib Device::SerialPort Apr 23 11:03:10 kroon: ok, I played around with several configurations for XZ_DEFAUTLS in local.conf Apr 23 11:03:15 kroon: even XZ_DEFAULTS = "--memlimit=1G --threads=1" doesn't work Apr 23 11:03:22 the machine has 32GiB of memory Apr 23 11:04:56 this only happened after I upgraded the project from sumo to zeus Apr 23 11:05:03 problame: With that much RAM you should be able to produce an SDK with no problems. Have you added many extra packages to the SDK? Apr 23 11:05:25 paulbarker: the entirety of qt Apr 23 11:05:48 paulbarker: but that shouldn't matter, xz acts as a stream processor, i.e. it doesn't need to load the entire image into RAM Apr 23 11:06:22 (the failing build phase is archive_sdk) Apr 23 11:06:25 problame: Agreed, the final SDK archive may end up large but the memory usage during the build should be sane Apr 23 11:06:53 If you remove qt from the SDK and try again does that build ok? Apr 23 11:07:13 paulbarker: yeah, but i need a qt sdk Apr 23 11:08:43 huh, when I get a devshell with `bitbake THE_IMAGE_RECIPE -c devshell` and run printenv, XZ_DEFAULTS doesn't show up Apr 23 11:09:01 problame: Ok, so it works with a smaller set of SDK packages but then fails when qt packages are added Apr 23 11:10:21 ... where should I override XZ_DEFAULTS? I do it in local.conf ATM, but since it doesn't show up in printenv in the devshell, maybe these changes don't have any effect? Apr 23 11:21:14 problame, you can check its getting applied by inspecting the environment of the image recipe, -e Apr 23 11:24:59 problame, I hit this before Apr 23 11:25:12 it was this patch: https://git.openembedded.org/openembedded-core/commit/meta/recipes-devtools/opkg-utils?id=369c43cca4f00272e0f9eb8b939f38f00dc43cc3 Apr 23 11:25:25 well, it is still there... Apr 23 11:25:58 Hey everybody, Apr 23 11:27:39 I read that the kernel configuration is stored on the target at /proc/config.gz Apr 23 11:28:08 I don't have that file on my image. Do I have to set a flag to include it into the image? Apr 23 11:29:14 emirus: You'll need the right kernel config option enabled for that, google should be able to find it I can't remember the exact name of the option Apr 23 11:30:21 emrius ^^^ Apr 23 11:31:05 ant__: well that looks interesting, I'm actually using OPKGBUILDCMD = "opkg-build -Z gzip" Apr 23 11:31:38 paulbarker thanks! I think I found the flag: https://cateee.net/lkddb/web-lkddb/IKCONFIG_PROC.html Apr 23 11:34:45 problame: if you revert the patch you get the old behavior,like in sumo Apr 23 11:35:41 which means slower packaging of kernel modules i.e., otherwise you can try changing it like compressorargs="--threads 0 -O 3" Apr 23 11:35:48 bbl Apr 23 11:38:32 ant__: I just checked the opkg-build in build/tmp/sysroots-components/x86_64/opkg-utils-native/usr/bin/opkg-build , and it doesn't contain that patch Apr 23 11:40:33 ant__: my poky checkout is zeus-22.0.2 Apr 23 11:47:22 ant__: I think it's 5ef5ece35bffa76b08d038ed0e3ebbdcbaf4b11c in poky/meta/classes/populate_sdk_base.bbclass Apr 23 11:47:32 > d.setVar('SDK_ARCHIVE_CMD', 'cd ${SDK_OUTPUT}/${SDKPATH}; tar ${SDKTAROPTS} -cf - . | xz -T 0 -9 > ${SDKDEPLOYDIR}/${TOOLCHAIN_OUTPUTNAME}.${SDK_ARCHIVE_TYPE}') Apr 23 11:47:38 I'm getting the warning: `- was registered as shlib provider for mylibrary.so, changing it to - because it was built later` - even there is no mylibrary.so in the whole working directory. What can it be? Apr 23 11:47:40 => hard-coded xz args Apr 23 11:49:36 problame, looks plausible. its been adjusted in master Apr 23 12:00:48 kroon: could you point me to a commit? Apr 23 12:01:23 kroon: I am just trying https://gist.github.com/problame/c71e7abbdc3f8870e06ae2f68230c2b4 Apr 23 12:33:21 problame, https://github.com/openembedded/openembedded-core/commit/25533a48a19e3e6eb9f96be51b1373830b5bb9f0 Apr 23 12:33:32 problame, basically what your patch does I think Apr 23 12:48:02 kroon: lol, yeah, extended my patch a little bit, basically with the exact same var names Apr 23 12:53:54 RP: are there plans to support Ubuntu 20.04 e.g. in a dunfell point release ? Apr 23 13:08:40 dl9pf: probably. A question for sakoman when he's around Apr 23 13:09:01 dl9pf: some depends on how much is broken too Apr 23 13:16:08 Right. Tnx! Apr 23 13:21:37 it works fine for me with dunfell and for zeus I've noticed only one issue in qemu-native (patch sent) Apr 23 13:32:41 hi, Like a BBAppend is there something like extending a machine conf? Apr 23 13:40:39 Do you keep the Yocto layer definition for your proprietary software in the same repository as the software itself? If so, how do you reference the software in the recipe's SRC_URI? `SRC_URI=file:///../../src` doesn't work, and putting an absolute path there isn't a great idea either Apr 23 14:09:03 problame: externalsrc Apr 23 14:09:16 qschulz: thx, just found it Apr 23 14:10:00 problame: or git subpath? Apr 23 14:10:57 problame: depends on why you've both in the same repo. BTW, make sure your build directory is not a subdir of your EXTERNALSRC... bite us hard a few months back :) Apr 23 14:11:18 qschulz: thanks for the hint! Apr 23 14:11:42 git subpath didn't work, and is suboptimal, since I actually want devs to be able to compile the package from a dirty git dir state Apr 23 14:11:46 havok101: no there isn't, though you can create your own machine configuration file and inherit from the original one Apr 23 14:12:44 problame: then externalsrc is really the way to go... But depending on the original need for that, a yocto SDK might have been a better idea. Though.. admittedly, I don't like externalsrc and I have no experience in sdk :D Apr 23 14:13:21 problame: also, devtool/devshell could have been another solution :) Apr 23 14:13:36 qschulz: s/dev/CI system/ although admittedly that should always be using a clean git repo Apr 23 14:14:26 problame: that's... hum surprising usecase for CI :D but you do you :) Apr 23 14:16:39 qschulz: so can I inherit from one in another layer and basically override certain things? Apr 23 14:17:54 qschulz: especially for devshell, I think externalsrc is the best choice Apr 23 14:18:18 (can edit the source (not the workdir) right from the devshell then) Apr 23 14:18:19 havok101: sure. I chose the wrong word though, it should be require and not inherit (inherit is for classes) Apr 23 14:19:43 havok101: and require takes a path relative to the root of *any* layer Apr 23 14:20:01 so you don't need to explicit from where yout ake the file Apr 23 14:23:30 qschulz: cool i'm going to play around with that, thanks Apr 23 14:29:54 Open Source Summit + Embedded Linux Conference North America 2020 Apr 23 14:29:54 is now a Virtual Experience, happening June 29 - July 2 - Apr 23 14:29:54 AND THE SCHEDULE IS NOW LIVE. Apr 23 14:37:33 that's https://events.linuxfoundation.org/2020/04/23/the-linux-foundation-announces-the-open-source-summit-embedded-linux-conference-north-america-2020-program-of-sessions-and-its-pivot-to-a-virtual-event/ Apr 23 14:42:27 armpit: have fun with the virtual yocto BoF Apr 23 14:59:54 drinking corona of course! Apr 23 15:26:56 qschulz: re update on externalsrc: doesn't work with `inherit qmake5` in do_install Apr 23 15:27:24 apparently qmake5_base_do_install uses ${WORKDIR} but should actually use ${B} Apr 23 15:37:53 Hello. I am getting the following warnings: WARNING: /home/.../poky/meta-intel/recipes-kernel/linux/linux-intel_4.19.bb:do_kernel_configme is tainted from a forced runWARNING: /home/.../poky/meta-intel/recipes-kernel/linux/linux-intel_4.19.bb:do_compile is tainted from a forced run Apr 23 15:38:07 Any idea how to get rid of this warning? Apr 23 15:40:35 problame: look if it isn't already be fixed in master, else please send a patch or at least report a bug :) Apr 23 15:41:07 riz1985: bitbake linux-intel -c cleansstate for sure but I don't know if there's a more optimal way Apr 23 15:41:23 Thanks! Apr 23 15:47:28 RP: I meant to disuss this last week https://bugzilla.yoctoproject.org/show_bug.cgi?id=13690 Apr 23 15:47:29 Bug 13690: normal, Medium+, 3.1 M4, timothy.t.orling, IN PROGRESS IMPLEMENTATION , runqemu with slirp, forwarded host ports are not available on host Apr 23 15:47:37 * moto-timo slept through two alarms Apr 23 15:48:23 RP: Paul has a change to the test to use SSH with SLIRP, which is the one thing I think was of value in that thread Apr 23 15:48:49 RP: but not sure if that is going to cause breakage or grief? Apr 23 15:49:09 RP: can oeselftest exclude a case if it doesn't have SSH? Apr 23 15:51:46 moto-timo: We have tests which depend on the ssh connection test passing Apr 23 15:52:21 RP: right Apr 23 15:52:54 RP: maybe two test cases? one with SSH one without? Apr 23 15:53:51 moto-timo: that could work Apr 23 17:09:34 Intel's Clear Linux To Divest From The Desktop, Focus On Server + Cloud Workloads Apr 23 17:09:39 https://www.phoronix.com/scan.php?page=news_item&px=Clear-Linux-Divesting-Desktop Apr 23 18:01:32 Just a quick sentiment grab, everyone bullish on balena? Apr 23 18:02:42 evaluating balena vs DevicePlane vs Mender Apr 23 20:04:25 dedline: depends on your needs Apr 23 20:05:15 kanavin_home: the standard bar is 2 yrs it seems for initiatives :) has clearlinux has their 2yr run Apr 23 20:05:56 so there is a recipe linux-variscite_4.14.98.bb and i'm appending it with linux-variscite_4.14.98.bbappend. But it refuses to those as appended when I run bitbake-layers show-appends linux-variscite Apr 23 20:07:16 havok101: interesting, can you do bitbake -e on the recipe and check changes from your bbappends are in the output ? Apr 23 20:07:44 khem: two years ago there was an internal initiative at intel, called one linux, which aimed to replace all the 'shitty distros' (implying, yocto-based) with clear linux for all use cases. This seems to have now failed as many said it would. The damage is that most of the yocto folks (incl. yours truly) left the company over this. Apr 23 20:09:41 khem: nothing, i have a patch in there and different revision, nothing shows Apr 23 20:14:06 kehm: ok.. got it, i needed to put it under the recipes-kernel/linux/ directory. earlier it was just recipes-kernel/. I didn't realize it was that strict. I feel like I put my qt append where ever and it works Apr 23 20:52:57 kanavin_home: short attention span I think oh well Apr 23 21:34:59 khem, have you ever benchmarked Clean Linux vs. an optimized OE-based distro? Apr 23 22:26:12 ant__:no not personally, but I know the technologies they are using Apr 23 22:28:21 of all distros I know clear linux gets benchmarked by phoronix along with other major distros like ubuntu/debian/fedora etc. which is quite interesting perhaps phoronix guys are very interested in clear linux for some reason Apr 23 22:29:12 since I dont see them benchmarking archlinux or manjaro which have way larger userbase Apr 23 22:42:35 khem, a friend of mine did it vs. Gentoo on an older i7. Not so much diff Apr 23 22:42:48 but is not a 'scientific' test Apr 23 22:44:24 what is funny is that 'desktop' still means old-design x86 PC hw with many secs needed just for BIOS Apr 23 22:44:45 so kernel boot time is not all... Apr 23 22:46:12 khem, by the way I am looking at that meta-initramfs-image U-A issues Apr 23 22:46:29 I think I tracked it down to acl pulling util-linux Apr 23 22:46:57 has nodistro always had acl? must check Apr 23 22:48:26 armpit, ^ there are the klibc binaries with same name installed. I think you did not have acl in your DISTRO_FEATURES Apr 23 22:58:57 acl is default in nodistro Apr 23 23:01:30 I suspected... Apr 23 23:03:06 I can't see how util-linux is called otherwise... Apr 23 23:03:07 "acl.do_build" -> "util-linux.do_package_write_ipk" Apr 23 23:17:43 khem, anyway it is an OR: klibc-utils or util-linux. There isn't any U-A for the former Apr 23 23:18:02 I don't see any sense in having all that stuff mixed together Apr 23 23:18:29 maybe for quick build-test of all proggies in meta-initramfs? Apr 23 23:18:38 armpit, ^? Apr 23 23:21:56 khem, it is the only buggy recipe of the layer Apr 23 23:23:18 it was not in the world-builds being EXCLUDED_FROM_WORLD Apr 23 23:28:08 see: I'm sending a patch at least fixing it...still not graspin Apr 23 23:52:52 k Apr 23 23:53:18 khem, patch coming, I see culprit is dracut Apr 23 23:53:36 but I let it not knowing image purposes Apr 24 00:10:30 khem, next patch will be removal. we had an initramfs framework in oe-classic, maybe revamp it.. Apr 24 00:10:36 gn Apr 24 00:56:14 There seems to be problem with the layer index. The version number for rsyslog and other packages is incorrect: https://layers.openembedded.org/layerindex/branch/master/recipes/?q=rsyslog Apr 24 00:57:36 also seen: https://layers.openembedded.org/layerindex/recipe/845/ where the 'Bug Tracker' link is wrong but the 'Recipe File' link is correct. Apr 24 01:01:31 seems to be related to the VER and SRCREV from libnfc: https://layers.openembedded.org/layerindex/recipe/97731/ Apr 24 01:04:59 * vmeson reports ^^^ by email to bluelightning, armpit Apr 24 02:47:56 I wonder if that may be related to the db move? **** ENDING LOGGING AT Fri Apr 24 02:59:57 2020