**** BEGIN LOGGING AT Tue Apr 21 02:59:59 2020 Apr 21 03:14:58 Morning. Is there a way to re-run the Yocto `oe-selftest` without deleting the `build-st` directory? I checked the docs but there is nothing bout this. I didn't check the code yet...I thought that asking here is the first easy choice. Any help appreciated! :) Apr 21 07:04:13 New news from stackoverflow: Unable to transfer libraries from Image folder of openblas into rootfs Apr 21 07:34:19 New news from stackoverflow: Why i2cdetect 0 not returning address Apr 21 07:56:21 RP: WRT TSC minutes: FYI, why we need multilib support: third party SW vendor only wants to deliver 32b binaries (and also pre-clang-compiled with libcpp as standard lib.....) and unfortunately this SW is a non-negligible part of our whole SW stack. Apr 21 07:58:34 RP: while we never had the issue before, some clients have so many third party 32b-only pre-built binaries, one project leader decided to basically have everything in 32b instead of playing a never-ending whack-a-mole for those clients. Hence the issue. FWIW, we're still on thud, so the issue exists at least there and in master. Apr 21 07:59:00 RP: also, sorry for the oe-core patch and its associated patch in docs in two separate mails, brain was fried :/ Apr 21 08:31:23 gents, I believe I have a real stupid question with no smart answer, could someone please help (sarcasm allowed) Apr 21 08:32:02 so I have this cmake/qt project that I can configure with no problems after sourcing the sdk in a shell Apr 21 08:32:31 but as soon as I put those three lines in a script, it fails to find Qt ... wtf? Apr 21 08:32:50 source /opt/poky/environment-setup-cortexa9hf-neon-montavista-linux-gnueabi Apr 21 08:32:58 mkdir build && pushd build Apr 21 08:33:09 cmake -G "Unix Makefiles" .. Apr 21 08:34:13 why does this succeed when done interactively and why does it fail in a script? any freaking ideas? Apr 21 08:34:52 nameclash: check whats on your environent when you execute from a shell, also check which shell is it using when you are running the script Apr 21 08:34:55 qschulz: it actually needs to be in two patches as they're different repos! Apr 21 08:35:14 nameclash: bash != sh != dash != zsh != fish != busybox's sh Apr 21 08:35:15 qschulz: I know why we need multilib and we'll not be removing it, I just dream ;-) Apr 21 08:35:15 RP: yes but they could have been in the same patch series :) Apr 21 08:35:46 qschulz: its fine, I get a lot worse :) Apr 21 08:36:21 qschulz: I'm curious how you find the approach I'm testing in master-next. It is meaning more invasive changes than I'd have liked and still isn't working right :( Apr 21 08:37:29 mckoan yesterday you send me some framebuffer gui tk's Apr 21 08:37:58 RP: I dream of it too :) Apr 21 08:37:59 mckoan Qt seems to be the "best" of these. Do you have any experience with this? Apr 21 08:38:08 nameclash: either way, probably shouldnt be running make with no recipe, your build env its contaminated with your host, create a simple recipe, inherit cmake and pass EXTRA_OECMAKE to it Apr 21 08:38:49 RP: especially since we have some dirty hacks for multilib+clang for a recipe that needs to exist both with libcpp and just glibc at runtime. anyway, digressing Apr 21 08:38:54 YPTM: If I dont make it its becase I failed to wake up on time, goodnight Apr 21 08:39:08 For example with small 4 bit greyscale OLED Displays? Is there some mechanism in Linux or Qt which can prevent for burn in? Apr 21 08:40:07 alejandrohs: my objective is to run the build against a prebuilt sdk, not using bitbake... Apr 21 08:41:01 RP: I'll have a look in the next few days, carefully reading everything again and checking the patches. I suspect I won't have much to say but does not hurt to know what I'm making you folks do to our little baby. Apr 21 08:42:19 qschulz: My dilemma is whether to run with this approach or whether to add something 'clever' to bitbake which others will no doubt find interesting ways to abuse Apr 21 08:45:32 RP: I always have a hard time with "smart" things. You can't imagine how messed up "smart" displays are driver wise. My opinion is that anything that isn't crystal clear is hard to maintain and is a time-bomb for weird bugs should be avoived at all costs. Apr 21 08:47:24 ultimately only a few people will maintain that in bitbake, so it's up to you (on the principle, I haven't checked the patches/approach) but it could also make it harder to get external people to fix/contribute to it. Apr 21 08:51:41 the environment is nearly identical (only OLDPWD, SHLVL and WINDOWID differ) Apr 21 08:51:54 both are run in bash Apr 21 08:52:00 I don't get it Apr 21 08:52:42 PatrickE: I have experience with Qt and other GUI libraries Apr 21 08:54:16 mckoan do you know about OLED, small displays etc.pp.? Apr 21 09:00:38 nameclash: just to be sure, your script should be doing everything, right? like cmake is actually building your SW and is where Qt isn't found, right? You're not using your script to setup the environment and then building after you've run the script? Apr 21 09:01:28 qschulz, everything's handled completely within the script Apr 21 09:01:41 source sdk, cmake, make Apr 21 09:02:43 it also doesnt seem to be related to the project being built Apr 21 09:03:07 testing against https://github.com/apriorit/MinimalQml Apr 21 09:04:59 I believe it's related to the cmake environment setup in a rocko based sdk Apr 21 09:05:29 nameclash: so it does not look like it's a subshell issue I guess Apr 21 09:05:48 no, doesnt look like Apr 21 09:05:59 as said, environment is identical Apr 21 09:06:18 nameclash: how do you check the environment? Apr 21 09:06:37 I dumped it to a file using printenv Apr 21 09:07:17 nameclash: use "set" instead Apr 21 09:07:28 yeah I was about to do so Apr 21 09:08:33 PatrickE: Isn't OLED burn in just aging of the "pixel"? So the way to handle it is to try to not use pixels more than neccessary, and also try to use the whole screen equally so that the degradation process is similar over all pixels. Apr 21 09:15:20 qschulz: other than a bunch of shell functions missing they look pretty equal Apr 21 09:17:42 oh, wait Apr 21 09:21:26 qschulz: I think I identified the problem being in [...]/environment-setup.d/cmake.sh Apr 21 09:21:43 it aliases cmake to: Apr 21 09:21:45 alias cmake="cmake -DCMAKE_TOOLCHAIN_FILE=$OECORE_NATIVE_SYSROOT/usr/share/cmake/OEToolchainConfig.cmake" Apr 21 09:22:16 that alias doesnt seem to find its way to the sourcing script Apr 21 09:22:40 if I replace in my script Apr 21 09:22:49 cmake -G "Unix Makefiles" .. Apr 21 09:22:51 with Apr 21 09:22:57 cmake -DCMAKE_TOOLCHAIN_FILE=$OECORE_NATIVE_SYSROOT/usr/share/cmake/OEToolchainConfig.cmake -G "Unix Makefiles" .. Apr 21 09:23:04 it runs fine Apr 21 09:24:56 nameclash: any chance there is a shell variable for that (e.g. $CMAKE, like we have for $CC?) Apr 21 09:24:59 I know this was changed that in later versions of poky but didn't realize I'm actually facing a side effect of the alias version in rocko Apr 21 09:26:29 qschulz: if there was one, I would look for it in /environment-setup.d/cmake.sh but the alias is the online line in that file... Apr 21 09:27:37 I can handle that now, hopefully we'll migrate to thud soon, where at least that issue seems to be resolved Apr 21 09:28:27 cmake/qmake in thud seems to suffer from different issues there but at least this one is fixed Apr 21 09:29:28 nameclash: http://chiefsandendians.blogspot.com/2010/07/linux-scripts-and-alias.html Apr 21 09:31:08 qschulz: oh, great! was about to patch the cmake recipe... thanks Apr 21 09:32:35 nameclash: not sure it's advisable though, it makes the script rely on aliases. I don't know if it also takes aliases from the host for example, which wouldn't be a great idea :/ Apr 21 09:34:36 qschulz: yeah, you might be right... I guess patching the cmake env setup script to export a $CMAKE variable should be safer.. Apr 21 09:36:31 Hi, Is there a way to PACKAGE_INSTALL depends on machine ? Apr 21 09:38:05 rokm: do you want to include different things in an image depending on machine? If yes, there's way to do that but I don't think modifying PACKAGE_INSTALL explicitly is the way to go Apr 21 09:38:47 I have initramfs image which contains some packages, but also needs out of tree intel specific drived to be installer Apr 21 09:41:08 can you use IMAGE_INSTALL_append_ = " "? Apr 21 09:41:23 in the initramfs image recipe Apr 21 09:41:23 will try this Apr 21 09:43:46 rokm: MACHINE_[ESSENTIAL_]EXTRA_RDEPENDS? Apr 21 09:43:57 https://www.yoctoproject.org/docs/latest/mega-manual/mega-manual.html#var-MACHINE_ESSENTIAL_EXTRA_RDEPENDS Apr 21 09:44:53 rokm: which you set in your machine configuration file Apr 21 09:48:35 qschulz: in yocto manual it is written that it is not for initramfs Apr 21 09:48:59 or I wrongly read this Apr 21 09:54:00 rokm: this makes the specified packages a runtime dependency for packagegroup-core-boot, so if it is used somehow in your initramfs, it's all good Apr 21 09:54:16 it is not used Apr 21 09:54:40 but seems I have MACHINE_EXT... set Apr 21 09:54:54 in machine config file but anyway it is not taken into image Apr 21 09:55:07 meta/recipes-core/images/core-image-tiny-initramfs.bb is using it in PACKAGE_INSTALL Apr 21 09:56:09 I'd say you want packagegroup-core-boot for any image you'll ever build? Apr 21 09:57:12 for my current image it is not needed Apr 21 10:23:30 PatrickE: yes, something Apr 21 10:31:26 erbo i think in the self written driver we used on bare metal devices we switch near pixels very fast or something like this. Iam not 100% aware of this Apr 21 10:46:04 Sorry in advance for a question that maybe can be found in a manual/tutorial: If I want to run the default "do_whatever" task on my own built package - How should I do it correctly? Apr 21 10:54:39 nacknick: bitbake _your-package_ Apr 21 10:56:52 nameclash: my recipe has only do_fetch that I wrote, it downloads an archive into "downloads" folder. I want to extract the archive (guess with do_unpack?) - How should I run do_unpack if it does not run automatically with `bitbake `? Apr 21 10:58:40 it should run with bitbake your-package Apr 21 10:58:51 otherwise run bitbake -c unpack your-package Apr 21 11:02:13 you could also run: bitbake -c devshell your-package Apr 21 11:02:56 that should drop you in a shell where all environment is set up as if bitbake was running Apr 21 11:08:17 i.e. you could execute your tasks one by one interactively Apr 21 11:08:18 nameclash: I tried `bitbake -c unpack ` - I hoped to find the extracted data inside the relevant build directory - but I did not. Does it matter if the archive is tar/gz? There is any reason why `unpack` does not extract the archive? Apr 21 11:11:07 just to be pedantic, it's not "your-package" but "your-recipe" you pass to bitbake Apr 21 11:11:39 w/e Apr 21 11:13:02 can we have the recipe you're writing on some pastebin somewhere? Apr 21 11:13:58 if you read base.bbclass and IIUC, unpack only works on things defined in SRC_URI Apr 21 11:18:59 qschulz: I don't know if I read base.bbclass and IIUC - How to check it? Apr 21 11:19:20 I will give you the recipe in pastebin. just checking something Apr 21 11:20:15 IIUC = If I Understood Correctly :) https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/classes/base.bbclass#n171 Apr 21 11:21:18 Or, you can check in your recipe WORKDIR (tmp/work/*/my-recipe/*/) in temp, run.do_unpack to check what is actually done during the unpack step. Apr 21 11:24:55 is it just me or are the glibc and bitutils mirrors painfully slow at times? Apr 21 11:25:25 it's fetching @ ~100 KiB/s Apr 21 11:25:57 rokm, you can set ${YOUR_MACHINE_PACKAGE} in your_machine.conf, and add this to IMAGE_INSTALL in your custom image Apr 21 11:26:01 qschulz: nameclash: Works! Thank you Apr 21 11:26:40 rokm: like IMAGE_FSTYPES = "${INITRAMFS_FSTYPES}" Apr 21 11:27:30 check out meta-initramfs layer of meta-oe for more Apr 21 11:35:09 New news from stackoverflow: Add user to existing group in Yocto Apr 21 11:50:33 How to skip checksum for a file in SRC_URI? Apr 21 11:57:37 nacknick: don't specify one Apr 21 11:58:11 nameclash: When I don't, it screams that there is no checksum :| Apr 21 11:58:28 license file? Apr 21 11:59:37 yes Apr 21 12:00:12 you could set BB_STRICT_CHECKSUM = "0" Apr 21 12:00:24 nacknick: give the correct checksum for the license file then? it's actually a very important piece of information in your recipe Apr 21 12:00:27 `Recipe file fetches files and does not have license file information (LIC_FILES_CHKSUM) [license-checksum]` Apr 21 12:00:47 qschulz: There is no license file Apr 21 12:00:47 that'll give you a warning instead, but imho it'd be better to provide the correct chksum in you recipe Apr 21 12:00:49 not checking it is very risky and not nice to users Apr 21 12:01:28 then set LICENSE = "CLOSED" Apr 21 12:01:29 nacknick: if this a piece of code or data that you wrote and will never be redistriuted? Apr 21 12:01:31 nacknick: LIC_FILES_CHKSUM = "file://${COMMON_LICENSE_DIR}/MIT;md5=0835ade698e0bcf8506ecda2f7b4f302" Apr 21 12:01:54 qschulz: that should only be used for metadata which is MIT licensed Apr 21 12:02:08 it says "the license the the oe-core license because this is part of oe-core" Apr 21 12:02:08 rburton: yes Apr 21 12:02:24 rburton: that was an example :) There are plenty other licenses. But indeed, worth expliciting Apr 21 12:02:45 nacknick: LICENSE=CLOSED is the solution for 'properietary closed source license' Apr 21 12:03:44 nacknick: also, if there is a license in some code file, you can also say the license is in X file between lines Y and Z. But if it isn't supposed to be distributed, nameclash gave you the answer Apr 21 12:05:03 rburton: good that you're around -- remember that LONG_BIT build error when including Python.h I mentioned yesterday? Apr 21 12:06:01 turned out I was too quick with the hooray, the issue actually still remains Apr 21 12:07:22 it's currently no showstopper for me but as I', trying to understand the build system, this is something I stumbled over and have no quick answer for it without debugging the problem the hard way Apr 21 12:08:28 the recipe that fails is https://pastebin.com/ePuKggCV with compile log at https://pastebin.com/LyZ2WNwd Apr 21 12:10:01 as said, this is ran in a vanilla poky/thud environment with no custom changes other than parallel make and bb threads.. Apr 21 12:11:18 the build runs fine without BBCLASSEXTEND = "native" Apr 21 12:12:17 does anyone see any obvious d'oh! in that? Apr 21 12:16:15 native or target Apr 21 12:16:23 like does target build work Apr 21 12:16:40 target build works Apr 21 12:16:57 native fails Apr 21 12:17:47 i'd be more likely to blame the cmakelists for doing something 'clever' Apr 21 12:17:52 nameclash: very dumb question, what about target which is the same arch as your host? (some qemu machine if that exists?) Apr 21 12:18:19 actually target is qemux86 Apr 21 12:18:27 as said, it's vanilla Apr 21 12:18:45 nameclash: I was suggesting compiling for a target which has the same arch as your host Apr 21 12:19:24 isnt that the case with qemux86 ? Apr 21 12:20:18 and a little bit of google search told me it should already be what you're doing :) Apr 21 12:20:21 ah, i should pick qemu86-64 Apr 21 12:21:15 nameclash: forget last sentence, yes qemux86-64 machine if you're building on x86 64b :) Apr 21 12:22:07 rburton: CMakeLists is pretty dumb I'd say: https://github.com/nameclash/cmake-sandbox/blob/master/CMakeLists.txt Apr 21 13:19:10 hey i am facing an issue while adding IMAGE_INSTALL_append = " gdbserver" to my local.conf file. it gives a ParseError. can anyone suggest why is that. and how to solve this Apr 21 13:22:11 pastebin please Apr 21 13:23:31 sorry what? @rburton you said to me Apr 21 13:24:12 rizi: you're asking us to debug an error without any information Apr 21 13:24:55 so put in pastebin.com or similar the actual error messages Apr 21 13:27:57 rburton : the error: ERROR: Unable to start bitbake server (None)ERROR: Server log for this session (/home/rizwan/src/ara-project/build/bitbake-cookerdaemon.log):--- Starting bitbake server pid 10968 at 2020-04-21 15:14:08.009669 ---ERROR: ParseError at /home/rizwan/src/ara-project/build/conf/local.conf:144: unparsed line: 'IMAGE_INSTALL_append = Apr 21 13:27:58 “ gdbserver"' Apr 21 13:28:27 and if you delete that line it works? Apr 21 13:28:59 yes it works, but i need to add gdbserver Apr 21 13:29:24 did you actually verify that deleting that line makes it work agai Apr 21 13:29:27 n Apr 21 13:29:27 use the right quotes and it should work Apr 21 13:29:48 oh yeah, if that's a copy-paste then the first quotes are smart Apr 21 13:30:05 “ != " Apr 21 13:33:01 actually i have to add three things to image. IMAGE_GEN_DEBUGFS = "1"IMAGE_FSTYPES_DEBUGFS = "tar.bz2" Apr 21 13:33:33 and IMAGE_INSTALL_append = “ gdbserver" Apr 21 13:33:46 if you're copy-pasting that line from the manual and it had bad quotes then please do show us what bit so the docs can be fixed Apr 21 13:35:01 quotes look ok e.g. in https://www.oreilly.com/library/view/mastering-embedded-linux/9781787283282/25b6d1c4-80e9-46da-b0be-8bd333631c7d.xhtml Apr 21 13:35:08 ok i get back to you on this Apr 21 13:36:29 yes JaMa thanx for sharing Apr 21 13:41:55 hi, ive been looking at the bins in bitbake and noticed one called bitdoc, but it seems broken. i have been working on getting it working and up to date which i have done for the most part, is this something that is still used in previous versions, or desired for use now, or should it be removed? Apr 21 13:43:03 never seen that before :) Apr 21 13:43:41 Hi, I wrote this morning with question related to initramfs image and package/driver that is dependend of MACHINE Apr 21 13:44:35 I checked and "KERNEL_MODULE_AUTOLOAD": " ixgbe" and "MACHINE_ESSENTIAL_EXTRA_RDEPENDS": " ixgbe" Apr 21 13:45:13 but still when I boot initramfs image ixgbe is not available Apr 21 13:45:50 It will be available only when I add ixgbe to PACKAGE_INSTALL in my initramfs recipe Apr 21 13:46:49 rburton i have made a branch that fixes it, but if its not used or wanted anymore, do you want me to remove it. do feel its serves somewhat of a purpose Apr 21 13:46:52 `Create a set of html pages (documentation) for a bitbake.conf...` Apr 21 13:48:46 rburton: in my last company we used to pack binaries in our product installers with last modified date 10 years after the epoch, with no one knowing where the hell they came from and where the sources were... we had long discussions whether I'd be safe to delete them in the next release or if we'd freak out some server farm admin sitting under a mountain somewhere in finland... we decided to keep it Apr 21 13:49:59 rokm: shouldn't it be kernel-module-ixgbe in MACHINE_ESSENTIAL_EXTRA_RDEPENDS? Apr 21 13:50:19 I added it too Apr 21 13:50:59 then EXTRA_... contains both: ixgbe and kernel-module-ixgbe Apr 21 13:51:30 rokm: do you have MACHINE_ESSENTIAL_EXTRA_RDEPENDS or packagegroup-core-boot in your PACKAGE_INSTALL in your initramfs recipe? Apr 21 13:52:27 I have MACHINE_ESSENTIAL.. like I wrote above and it contains this ixgbe Apr 21 13:52:46 and I don't have packagegroup-core-boot Apr 21 13:53:31 rokm: do you have PACKAGE_INSTALL += "${MACHINE_ESSENTIAL_EXTRA_RDEPENDS}" in your initramfs recipe? You set MACHINE_ESSENTIAL_EXTRA_RDEPENDS in the machine configuration file as well right? Apr 21 13:54:06 yes I have MACHINE.... in machine specific folder and conf file. Apr 21 13:54:20 but I don't have PACKAGE_INSTALL like you wrote :o Apr 21 13:54:27 let me try this Apr 21 13:55:51 rburton: my yocto image was working fine until i add in local.conf file. now when i even removed those lines from local.conf file i get the same error. how to resolve this. the error is <<< on device or exceeds fs.inotify.max_user_watches?ERROR: To check max_user_watches: sysctl -n fs.inotify.max_user_watches.ERROR: To modify max_user_watches: sysctl -n -w fs.inotify.max_user_watches=.ERROR: Root privilege is required to modify max_user_watches. Apr 21 13:56:23 riz49: well... "Errno=No space left on device" Apr 21 13:58:01 qschulz: still it was running earlier with same packages. it should run atleast with same packages. Apr 21 13:58:26 how to resolve this then? should i build the image from scratch? Apr 21 13:58:43 riz49: you should make some space on your device before building again Apr 21 13:59:37 qschulz ok thanx Apr 21 13:59:38 riz49: not if you ran your build with additional packages, the build consumes space on disk and is not freed automatically just by removing the lines from you image... Apr 21 14:00:35 nameclash ah now i understood. thanx for clarifying Apr 21 14:01:02 nameclash: also.. might be some other software filling up disk space :) Apr 21 14:01:23 riz49: maybe consider also adding INHERIT += " rm_work" to your config Apr 21 14:01:56 qschulz: let's not even think about that ;) Apr 21 14:03:15 nameclash i will try that Apr 21 14:11:59 qschulz: now it is on initramfs image, thanks :) Apr 21 14:12:10 qschulz: but still autoload doesn't work Apr 21 14:12:17 I have to modprobe Apr 21 14:17:47 rokm: ok, that's a first step. Where is your KERNEL_MODULE_AUTOLOAD defined? Apr 21 14:19:25 qschulz: in the same machine conf where ..ESSENTIAL.. is defined Apr 21 14:20:05 ok, and this one you have ixgbe only and not kernel-module-ixgbe right? Apr 21 14:20:30 yes -> "KERNEL_MODULE_AUTOLOAD": " ixgbe", Apr 21 14:21:40 in machine conf and this above comes from json file after build Apr 21 14:24:33 rokm: do you have something in /etc/modules-load.d/ on the target? Apr 21 14:24:50 when I look for this driver on rootfs, it seems to be under /lib/modules/kernel../updates/driver Apr 21 14:25:24 I have this ixgbe.conf Apr 21 14:25:30 which contains ixgbe Apr 21 14:25:46 otherwise, I could suggest to have PACKAGE_INSTALL += "packagegroup-core-boot" just to check it's not in there. I'm pretty much clueless after that :/ Apr 21 14:27:16 I have my custom init Apr 21 14:27:57 I think this autoload is handled by systemd Apr 21 14:27:59 so probably It will end that I will modprobe it somewhere there Apr 21 14:28:06 I don't have systemd :) Apr 21 14:28:08 rokm: yup indeed Apr 21 14:28:21 ok Apr 21 14:28:23 we use sysv, we have to modprobe manually :) Apr 21 14:28:37 I also have sysv Apr 21 14:28:45 in my initramfs Apr 21 14:29:02 thanks for help qschulz Apr 21 14:29:05 :) Apr 21 14:29:17 rokm: my pleasure, have fun Apr 21 14:42:02 Hi, what must I type on my console to get over this error: recipe for target 'silentoldconfig' failed Apr 21 14:42:30 It happened after I switched on USB via bitbake virtual/kernel -c menuconfig Apr 21 14:43:53 bitbake core-image-minimal -c cleanstate (or also with cleanall) hasn't helped Apr 21 14:45:09 Do you have any idea? Searching the internet did not help me Apr 21 14:49:41 fbre: you cleaned the sstate-cache of your image recipe, not your kernel recipe. What is the full log of the error for the kernel? Apr 21 14:54:32 qschulz: please, look at https://www.dropbox.com/s/1dw24ukzmvlgiv1/20200421_165217.jpg?dl=0 Apr 21 14:54:35 Hi, I'm trying to build an imx8 from variscite. I keep getting this error Nothing provides virtual/egl. Apr 21 14:56:16 It says imx-gpu-viv provides virtual/egl but was skipped: incompatible machine Apr 21 14:59:04 YPTM: Scott Murray on Apr 21 14:59:47 YPTM: Jan-Simon is on Apr 21 15:01:14 I also tried (which did not help either): bitbake virtual/kernel -c cleanall Apr 21 15:01:46 and bitbake virtual/kernel -c cleanstate as well Apr 21 15:02:06 cleansstate with two s? Apr 21 15:02:59 * tlwoerner congratulates Steve Sakoman at LTS maintainer Apr 21 15:03:04 no, one s Apr 21 15:03:46 fbre: no two :) Apr 21 15:04:02 cleansstate is correct? Apr 21 15:04:20 fbre: -c listtasks (two t :) ) to check what is available for a recipe Apr 21 15:05:33 argh, I thought it means to CLEAN the STATE Apr 21 15:05:57 New news from stackoverflow: error when starting wifi interface: failed to load regulatory.db Apr 21 15:07:01 YPTM: Alejandro is on Apr 21 15:07:08 fbre: but that is a weird error... I don't see why it would fail explicitly at that point when doing the new .config... Apr 21 15:07:08 hi, does anyone do private courses? Apr 21 15:07:21 YPTM: Martin Jansa is on (muted) Apr 21 15:07:27 fbre: it's called sstate :) Apr 21 15:07:35 I have appended imx-gpu-viv_%-arch64 and added the machine i'm building to Compatible machine but it still fails Apr 21 15:08:39 havok101: the recipe does not have your machine in COMPATIBLE_MACHINE maybe? Apr 21 15:09:17 qschulz: OK, thanx so far. hmm cleansstate did not help either Apr 21 15:09:33 I think I must delete and checkout the whole stuff again Apr 21 15:10:12 so i appended the recipe and added this COMPATIBLE_MACHINE(mx8|imx8mn-var-som). Still no go Apr 21 15:10:29 Is there something else I need to be tweaking. This is based on the 4.14 kernel Apr 21 15:10:49 fbre94: and what about linux-imx instead of virtual/kernel? Are you passing the machine as well? Apr 21 15:11:43 havok101: what is the exact line you have added for the COMPATIBLE_MACHINE in your bbappend? Also please check your bbappend is taken into account by running bitbake-layers show-appends imx-gpu-viv Apr 21 15:12:25 qschulz: Do you mean core-image-minimal instead of linux-img? I don't know what linux-img mean. I guess core-image-minimal is how it is named here Apr 21 15:14:27 fbre94: core-image-minimal is an image recipe and you're trying to debug a package recipe, different things :) virtual/kernel is a virtual package, so depending how machine configuration files, distro files and/or recipe files, virtual/kernel can actually be any kernel recipe Apr 21 15:15:44 fbre94: your kernel recipe is linux-imx as per the log you sent me. Which makes sense for your machine which is an imx8 something Apr 21 15:17:00 it is very unlikely that bitbake virtual/kernel -c cleansstate actually does not fix your recipe if all you did between the moment it worked and now is using -c menuconfig and nothing else Apr 21 15:18:03 qschulz: OK, I called bitbake linux-imx -c cleansstate Next, I tried to bitbake again but it stops with that 'silentoldconfig' failed error again Apr 21 15:18:09 havok101: I suspect your bbappend isn't taken into account because I doubt imx-gpu-viv_%-arch64.bbappend works (happy to be proven otherwise :) ) Apr 21 15:18:45 I think you're right. Apr 21 15:18:58 fbre94: did it ever work before? You have made no changes whatsoever to any of the files used by the recipe (including the recipe itself or any bbappend)? Apr 21 15:20:34 yes, it has worked before. Then I just called bitbake virtual/kernel -c menuconfig loaded my kernel config, switched on the USB checkbox, saved my kernel config and then the bitbake call again which fails then as you have seen Apr 21 15:22:32 OK, guys. I must leave now. Thank you very much for your help so far anyway. I really appreciate this IRC here Apr 21 15:22:36 See you Apr 21 15:22:37 Hello everybody, I hope you are all healthy and fine. I'm struggling to start the wifi interface. Actually, it does not exists... strange. This is what I get when I run `ifconfig -a`: https://pastebin.com/sV508PPi Apr 21 15:22:39 fbre94: gl Apr 21 15:23:18 I have `DISTRO_FEATURES_append = " wifi"` in my local.conf Apr 21 15:23:35 emrius: you need wifi drivers I'd say Apr 21 15:23:44 I'm using an out of tree driver from here: https://github.com/geomatsi/meta-sunxi-contrib/tree/master/recipes-bsp/rtl8189fs-mod Apr 21 15:23:51 :) Apr 21 15:24:29 emrius: ok, do you have this kernel module available at runtime somewhere? Apr 21 15:24:41 When I run modprobe 8189fs it seems to load the driver Apr 21 15:25:02 emrius: lsmod? Apr 21 15:25:23 8189fs 1282048 -2 Apr 21 15:25:32 cfg80211 282624 -2 Apr 21 15:26:14 (that was the output of lsmod ( part of it)) Apr 21 15:26:14 nothing weird in dmesg lately? something like a missing fw? Apr 21 15:26:26 but why -2? Apr 21 15:27:09 hmm to be honest I don't know Apr 21 15:27:25 hey qschulz: i have an image built. i tried to add a package and it give me error "no space left on device". now i am trying to add space to image by adding IMAGE_ROOTFS_EXTRA_SPACE = "5242880" to local.conf file but still it is giving same error. how to solve this problem? Apr 21 15:27:38 seems to be the case for all the kernel modules. Is that bad? Apr 21 15:27:44 rizi: you are out of space on your build system not the target Apr 21 15:28:04 https://pastebin.com/RtP1Tf9L Apr 21 15:28:21 ah damn Apr 21 15:28:28 emrius: that is weird... Apr 21 15:28:47 moto-timo: we have the same issue with the arm target, isn't it some old armv5/armv6 architecture? Apr 21 15:28:56 hmm ok Apr 21 15:29:09 emrius: anything valuable in recent dmesg output? Apr 21 15:29:15 I mean when I modprobe the module it gives a pretty gerneric warning: Apr 21 15:29:17 [ 81.818203] 8189fs: loading out-of-tree module taints kernel. Apr 21 15:30:02 qschulz: i just checked my build machine has 200gb free. Apr 21 15:30:18 tlwoerner: I think many architectures have issues... interested parties sccratch the itch when they feel like it? Apr 21 15:30:28 rizi: on the partition where the tmp dir is? Apr 21 15:30:50 (tmp for Yocto tmp dir). Also, you might hit the max inode number as specificed in the error Apr 21 15:30:50 tlwoerner: and when nobody around cares enough ... (MIPS?) Apr 21 15:30:59 moto-timo: are there better emulators? Apr 21 15:31:23 tlwoerner: no data. can't give an answer. Apr 21 15:32:03 * moto-timo needs to do more research obviously Apr 21 15:32:49 as I said, half baked thought Apr 21 15:33:56 qschulz There is an issue with regulatory db when loading the driver: [ 175.470585] cfg80211: failed to load regulatory.db Apr 21 15:34:23 But the driver module loads anyways Apr 21 15:34:26 so I think my problem is bitbake-layers show-appends says imx-gpu-viv_6.2.4.p4.0-aarch54.bb (skipped) Apr 21 15:34:37 can't append if base recipe is skipped Apr 21 15:34:45 tlwoerner: we updated the qemuarm to v7 Apr 21 15:35:03 RP: ah, so that wasn't a qemu issue then Apr 21 15:35:07 emrius: IMAGE_INSTALL += "crda" and let's see after that Apr 21 15:35:21 tlwoerner: more that v7 didn't exist when qemuarm was created ;-) Apr 21 15:35:46 yeah, actually I installed that using dnf. As the error is still there I assumed it wasnt related to that. Apr 21 15:35:58 havok101: ugh, duh. Since the COMPATIBLE_MACHINE isn't set /me facepalms myself Apr 21 15:36:34 qschulz: Ah one step further: `COUNTRY environment variable not set` Apr 21 15:37:08 I'd imagine it should use -aarch64 file since it's an mx8 Apr 21 15:37:26 havok101: I don't think you can make bbappends machine specific Apr 21 15:38:01 emrius: unfortunately github is down now and hans is also not on IRC so not easy to contact Apr 21 15:38:43 ah it's back Apr 21 15:38:50 qschulz Ok, Damn github. who is hans? Apr 21 15:39:04 RP: i think moving python code from bbclasses to lib/oe will require improving the python checksumming to handle imported module files properly, at least for the oe package Apr 21 15:40:12 emrius: the one hosting this github repo. Used to do lots of work on allwinner boards. Apr 21 15:40:32 emrius: maybe you could ask folks on #linux-sunxi or contact the people from meta-sunxi Apr 21 15:40:50 Ah I see! I looked a lot into his repo to find answers. helped a lot already Apr 21 15:41:05 havok101: try with imx-gpu-viv_%.bbappend Apr 21 15:41:16 qschulz Thanks! Apr 21 15:42:40 emrius: maybe try with modprobe -v as well? Apr 21 15:44:04 I mean crda is erroring now: Failed to set regulatory domain: -7 Apr 21 15:44:13 but the driver still loads Apr 21 15:45:08 qschulz But maybe it's best to try to contact hans directly through github Apr 21 15:45:49 oh... my... god... my meta-fsl-bsp-release was not at the right commit. It has support for mx8mm but not mx8mn ..argh Apr 21 15:45:52 it works now Apr 21 15:47:08 kergoth: Right, that is one of my concerns. I did wonder about simply doing a static scan and inject a file listing the dependencies Apr 21 15:48:19 might be easiest Apr 21 15:48:22 I expect the simple case would be easy enough to do when scanning the ast, grab the import nodes, map to the files on disk, scan them.... it's uglier stuff like nested try/excepts that'd be more problematic to handle Apr 21 15:48:50 course that's assuming nothing funky in the import mechanisms, which there shouldn't be for us, but isn't true in the general case.. Apr 21 15:48:53 :) Apr 21 15:48:56 kergoth: we already have that limitation with the current scanning code? Apr 21 15:49:06 but yes, agreed Apr 21 15:49:42 kergoth: did you have any thoughts on a variable "filter" function flag? Apr 21 15:50:56 kergoth: its potentially open to so much abuse is my main worry (and performance) Apr 21 15:54:08 I believe this was the key MACHINEOVERRIDES_EXTENDER_mx8mn = "imxdrm:imxgpu:imxgpu3d" Apr 21 15:54:41 emrius: I mean, try to fix this crda error first? Apr 21 15:54:57 havok101: yup, that makes sense from the few times I had to deal with gpu on imx Apr 21 15:55:10 qschulz That is probably a good idea to rule that out. I'm on it Apr 21 15:59:40 Thanks for your help though Apr 21 16:07:38 qschulz here is an explanation why the lsmod used by column shows -2 on my target: https://unix.stackexchange.com/questions/269500/lsmod-shows-2-in-the-used-by-column Apr 21 19:08:28 Hello. I am creating to recipies, one depends on the other. The latter is not able to find some make files that the former is providing. the former's recipe is installing such cmake files in ${D}${libdir}/cmakes successfully (I can find the cmake files in the imaegs folder of the build folder of such recipe). The later one DEPENDS on the former but still it is not able to fidn the cmake files Apr 21 19:09:11 I added the dependencit also on RDEPENDS_${PN} but still the same result... am I missing something? **** ENDING LOGGING AT Wed Apr 22 02:59:57 2020