**** BEGIN LOGGING AT Tue Aug 22 03:00:00 2017 Aug 22 04:32:45 in order to change from read only file system to read-write- do i have to change kernel init params (ro) along with fstab? Aug 22 04:50:00 hello Aug 22 04:50:35 in a recipe, i want to define a variable depending on a file existence. is this possible? Aug 22 04:52:23 like in: if file exists var=a else var=b Aug 22 04:53:25 c0rnel: I'd suggest an anonymous python function for that Aug 22 04:55:36 c0rnel: https://pastebin.com/yUhWzjdT Aug 22 04:56:54 bluelightning, thank you very much. can such a code be included for example in do_configure() ? Aug 22 04:58:13 c0rnel: no, since do_configure is a shell function - however you can add a prefunc to do_configure to accomplish the same thing Aug 22 04:58:40 thank you very much, bluelightning . i'll try this Aug 22 04:58:43 i.e. do_configure[prefuncs] += 'myfunc' then python myfunc() { ... } Aug 22 04:59:03 ah, i see. thanks Aug 22 05:19:16 all those variables described in mega manual, where are they defined? i have a recipe using SRC_ARCHIVE_FILE but i have no idea where this is coming from :/ Aug 22 05:19:50 and i don't find it in mega manual Aug 22 05:20:07 <_nobody_> Where I can find the documentation how to use DNF in YOCTO (Pyro)? Thank you very much for the tips! Aug 22 05:20:53 <_nobody_> I know that somewhere on the target I need to configure repositories. Aug 22 05:21:05 <_nobody_> https://docs.fedoraproject.org/en-US/Fedora/24/html/System_Administrators_Guide/sec-Configuring_DNF_and_DNF_Repositories.html Aug 22 05:26:56 don't knoiw what are you trying to do but dnf repo files are same as yum repo files Aug 22 05:30:54 <_nobody_> It is not 1:1. I do agree with you, present state. In Pyro 2.3 DNF is added instead of smart. Aug 22 05:32:12 <_nobody_> And, as matter of fact, I do NOT see on the Pyro target /etc/yum.repos.d/ directory. Aug 22 05:32:44 <_nobody_> The Server repos ((link to them) should go in this directory! Aug 22 05:46:25 <_nobody_> [to YOCTO architects] Even I'll go further! YOCTO should itself maintain the latest PYRO prebuilt .rpm packages, and these should be publicly available! Aug 22 05:52:00 in fedora this is created by fedora-repos package Aug 22 05:52:21 but i assume that for testing purposes you can first create it manually Aug 22 05:53:53 guys, if a variable appears in the recipe but does not shu-up in bitbake -e, can i safely conclude is not actually used? Aug 22 05:53:59 show Aug 22 05:58:17 <_nobody_> c0rnel, I know all about Fedora. YOCTO is somehow different. I can add the package in F26 I do not have. This is why I advertise YOCTO servers similar as Fedoras ones. Aug 22 05:59:22 _nobody_, ok, then Aug 22 06:32:44 Hi, Does anyone has succeeded in getting Qt compiled statically in yocto? qtbase keeps failing with the installed-vs-shipped error on the /usr/lib/qt5/plugins folder where all the .a and .prl files are Aug 22 06:37:47 <_nobody_> Timon, why statically? Aug 22 06:39:52 To keep the rootfs small and to make startup faster Aug 22 06:40:11 <_nobody_> There is a problem with that, do you know? Aug 22 06:40:34 no, i am not aware of that. What is the problem? Aug 22 06:42:08 <_nobody_> This is how I understand it> licence issue. QT sits on the top of eglibc, which is LGPL2.1. To keep your QT code private, you MUST link dynamically. IMHO. Aug 22 06:44:44 Well i have a device integration license from Qt. Already contacted Qt support. but they only support the default images that they ship Aug 22 06:45:24 <_nobody_> Pls, read Lesser GPL licence. Aug 22 06:54:40 So in your opinion, you may not build qt statically even if you have the right licencing to build qt statically Aug 22 06:58:19 <_nobody_> I just advise you to be cautious. So to prevent somebody later on to ask these licence questions. This is all. Aug 22 06:59:34 ah oke. Thanks for your advice. I will contact legal to check it. Aug 22 07:00:27 but lets say, all the licensing is ok. Does anyone succeeded building Qt statically in Yocto? Aug 22 08:21:19 Guest47991: normally ".a" gets packaged into -staticdev (if you've enabled static libraries). The default packaging rule probably does not match /usr/lib/qt5/plugins/*.a though Aug 22 08:23:08 so you might have to add FILES_-staticdev + = "/path/to/*.a" Aug 22 08:23:52 but if this is the case then nobody seems to be building qt statically... Aug 22 08:24:27 tried to to that with the line "FILES_${PN}-staticdev += "/usr/lib/qt5/plugins/*" but then xmlpatterns is failing because it cannot find qconnmanbearer Aug 22 08:25:38 "plugins/*" sounds wrong, try "plugins/*.a" Aug 22 08:29:58 and ${libdir} instead of literal "/usr/lib" is a good idea Aug 22 08:35:58 jku: trying both suggestions now :) Aug 22 08:48:04 plugin/*.a is not working. There are only subfolders in the plugin folder Aug 22 08:55:25 if i check the LIBS from the xmlpatterns makefile it checks the recipe-sysroot-native/usr/bin/qt5/plugins/bearer/ But that folder doesn`t exists. in recipe-sysroot-native/usr/lib/qt5/plugins/bearer are .so files Aug 22 08:56:04 in recipe-sysroot/usr/lib/qt5/plugins/bearer are the .a and .prl files. What does the linker need. I assume the linker needs the .a and .prl files because i want it to build statically Aug 22 09:13:06 i have built a yocto project resulting in a zImage that I can run on my board. Now I want to change my linux kernel config, and then rebuild to have a zImage that reflect thouse changes Aug 22 09:13:11 how do I do that? Aug 22 09:28:47 hiwk: there are various options Aug 22 09:29:39 hiwk: bitbake -c menuconfig and generate a new config, then replace defconfig in the recipe Aug 22 09:30:08 hiwk: or create additional .cfg files to modify the existing defconfig Aug 22 09:30:31 hiwk: or work using a custom git repo Aug 22 09:32:04 hiwk: http://www.yoctoproject.org/docs/2.2/kernel-manual/kernel-manual.html Aug 22 10:11:41 mckoan: `bitbake -c menuconfig` just says "Nothing to do. " Aug 22 10:14:35 hiwk: try `bitbake -c menuconfig virtual/kernel` Aug 22 10:15:25 Please let me know following testsuite is used test selinux in yocto ? Aug 22 10:15:25 https://github.com/SELinuxProject/selinux-testsuite Aug 22 10:18:32 mckoan: I'm not running a virtual box. How can I figure out which kernel target (?) is being used? Aug 22 10:19:24 mckoan: 'bitbake linux-imx -c menuconfig' seems to work (guessed from bitbake-layers show-recipes | grep linux) Aug 22 10:39:49 mckoan: runnung menuconfig for linux-imx, copying the rsulting file to ../sources/meta-axiomtek/recipes-kernel/linux/linux-imx-3.14.52/rsb10x/defconfig , and then redo bitbake axl-image-base seems to do the trick Aug 22 10:40:09 mckoan: thanks! it was the "replace defconfig" step I hadn't guessed Aug 22 11:12:09 I am trying to add an unstable version of an application (DEFAULT_PREFERENCE = "-1") to my image by using PREFERRED_VERSION_myapp = "1.2.3". but bitbake still installs the stable version. what could I be missing here? Aug 22 11:12:53 `bitbake-layers show-recipes` shows both versions Aug 22 11:25:14 hm, `bitbake -s` only shows the stable one? Aug 22 11:36:54 doesn't PREFERRED_VERSION_ work in a bb-file? Aug 22 11:56:50 nrossi: I'm a bit confused with regards to dts-files. meta-xilinx has dts files (recipes-bsp), but those are only used for the kernel dtb, but not u-boot. In order to get my dts-to u-boot(-xlnx) I had to write a patch to add it to the source-tree and to the Makefile. Wouldn't it make sense to push the recipes-bsp dts to u-boot somehow during build? So the dts can be modified in a single place? Aug 22 11:59:22 nrossi: To avoid confusion, I've been creating my own layer that's based on meta-xilinx, but I use u-boot-xlnx. I think forking my own version of u-boot-xlnx doesn't make a lot of sense, so what I need is to either patch it, or get my recipes-bsp device-tree into it somehow. Aug 22 12:00:21 hiwk: ;-) Aug 22 12:08:58 nrossi: Or perhaps I should doing a bit of both: Patch the u-boot Makefile (arch/arm/dts/Makefile) and then copy the dtb, similar to how ps7_init_gpl.* is handled in u-boot-spl-zynq-init.inc? Aug 22 12:30:12 JoiF: your last message is the way I had anticipated implementing u-boot dts input support. But nothing needs it at the moment? and you should be able to handle it with changes to the DEFAULT_DEVICETREE config of u-boot Aug 22 12:36:50 if I define a recipe for an system image and I want to specify the version of an application contained in the image - where do I put this version information? Aug 22 12:40:25 I know there is a `conf/layer.conf` file where I can add `PREFERRED_VERSION_myapp = "1.2.3"`. but this effects the whole layer, not just the recipe for the image Aug 22 12:41:54 brrm: I guess typically that would be a distro decision, not image. so e.g. poky has examples in meta-poky/conf/distro/poky.conf Aug 22 12:46:08 speaking of distro configs... RP: is poky-bleeding somehow useful as an example or should I remove it (and the referneces in documentation)? Aug 22 12:46:26 because the versions in there make it more like poky-ancient... Aug 22 12:46:58 PREFERRED_VERSION_gtk+ ?= "2.13.3" :) Aug 22 12:48:56 jku: I keep meaning to turn it into something useful again. Its not useful today Aug 22 12:53:16 nrossi: Yes. So what I'm currently doing (to u-boot) with patches is: adding defconfig for my board (where I sed the DEFAULT_DEVICE_TREE), adding a configuration header (include/configs/zynq_[myboard].h), adding my dts file and lastly adding my dts file (or more exactly, adding a reference to the resulting dtb in the dtb-$(CONFIG_ARCH_ZYNQ) list). Aug 22 12:54:29 nrossi: For my board, I'm happy with doing all of those things with patches, except for "patching in" the dts-file itself. it makes modifying the dts really cumbersome.. Aug 22 12:55:27 JoiF: sure, also looks like arm needs something like this in its dts makefile :) http://git.denx.de/?p=u-boot.git;a=blob;f=arch/microblaze/dts/Makefile;h=f80d8fd8506db1c4e3ff3cdb83c257dba968bab0;hb=HEAD#l5 Aug 22 12:56:55 nrossi: I'll figure something out .. I just needed a second opinion on whether it made sense to copy my layer/recipes-bsp-dts file into u-boot or if I should stick to patches for all of it. Aug 22 12:58:07 JoiF: for the dts it makes sense to pass this through, i've been tempted to figure out a better way of doing is so as to pass it to the kernel as well so it can be packed with the fit image stuff Aug 22 12:59:13 JoiF: and to avoid having the device-tree recipe do the same compile steps as the kernel Aug 22 12:59:33 nrossi: I see Aug 22 13:02:08 JoiF: theres also the whole EXT_DTB flow of u-boot to consider as well http://git.denx.de/?p=u-boot.git;a=blob;f=dts/Makefile;h=3a93dafb5164ed6e8e4318071cd6612350b09b27;hb=HEAD Aug 22 13:06:07 nrossi: Interesting. Aug 22 13:07:10 JoiF: so if you come up with a nice solution that works for you would be great to see it :), send patches ;) Aug 22 13:07:26 nrossi: haha, I will if I do. :) Aug 22 13:19:17 Hi, Does anyone has succeeded in getting Qt compiled statically in yocto? Aug 22 13:35:00 Guest47991: so I assume the last attempt didn't work if you're asking again. what goes wrong? Aug 22 13:36:09 also, are you sure static linking is worth the effort for you? Not saying it can't be but I'm wondering what you want to achieve... Aug 22 13:39:10 No. keeps getting the same errors. while compiling qtxmlpatterns the linker cannot find -lqconnmanbearer -lqgenericbearer -qnmbearer Aug 22 13:39:24 and yeah, i am almost at the point i will skip the whole static compiling Aug 22 13:40:06 i had a small rootfs without qt libraries of 70mb. with qt libraries my rootfs is 200mb Aug 22 13:40:46 so your goal was to only ship the parts of qt you use? Aug 22 13:44:02 yeah. and i figured i dont need to ship any libraries if i compile it statically Aug 22 13:46:36 well, the ones you use will still be part of your app... I would first take a good look at the packages qt5 produces and make sure you don't accidentally get any you don't need on the image Aug 22 13:49:12 yeah i am trying now to add the packages one by one. if the application doesn`t start i check which lib it needs to run. etc Aug 22 13:50:06 also: since your problem was with some plugins: I don't know how plugins in QT work but if they are always dynamically loaded then you just can't link those statically... Aug 22 13:54:52 As far is i know and read on the qt help. The static flag is all the libraries that can handle the static linking. if it doesnt support it. it will loaded dynamically Aug 22 13:54:58 but thanks for your help! Aug 22 14:52:09 Hi, why does do_fetch want to connect to remote git server also when the commit needed is already in local download cache? Aug 22 14:56:26 hi Aug 22 14:57:30 I'm trying to add a custom machine with custom kernel and config. Everything compiles find but it fails to create rootfs: No package packagegroup-core-boot available. Aug 22 14:58:41 the strange thing is that I can find this file: "tmp/deploy/rpm/my_machine/packagegroup-core-boot-1.0-r17.my_machine.rpm" Aug 22 14:59:03 does someone knows what is wrong in my config? Aug 22 14:59:30 is there something to add somewhere to find packages? Aug 22 16:40:31 can you try bitbake packagegroup-core-boot Aug 22 16:41:24 khem: package is already built. Aug 22 16:41:44 problem is solved, I removed upcase letters in machine filename Aug 22 16:43:38 yes OE/opkg follows debian naming lexicon Aug 22 16:49:19 https://www.debian.org/doc/debian-policy/ Aug 22 16:55:26 you can disable -package- naming for the debian style.. (this is what renames say 'glibc' to 'libc6'..) Aug 22 18:28:49 hello, does anyone know how to change the config.txt file during the build process using yocto? Aug 22 18:38:56 zarzar_: no idea what config.txt even is Aug 22 18:40:11 kergoth there is a config.txt file that hassettings like enable_uart=1 for raspberry pi 3, most of the time it is changed at runtime, i want to change at build time Aug 22 18:40:40 kergoth i found it once in the yocto sources directory bit i did not note where it was\ Aug 22 18:42:36 Would anybody happen to kow how to add the zgrep binary to the gzip package? Aug 22 18:43:36 So far I have a bbappend with a do_install_append() that copies that file to the same dir as the other files that do make it into the image, but that's not working. Aug 22 18:47:11 Is there a way to invoke the bitbake command with a particular BBVERSION? Aug 22 18:56:35 hello, does anyone know how to change the config.txt file during the build process using yocto? the config.txt file usually changed at runtime in boot Aug 22 18:56:39 not in uboot Aug 22 19:02:26 * xthunderheartx waves Aug 22 19:03:23 test Aug 22 19:08:42 overdamped: what do you mean? Aug 22 19:17:39 kergoth: I have a recipe with 3 BBVERSIONS. I would like to build it with bitbake Aug 22 19:18:06 a.k.a. something akin to bitbake "recipe-name(=myversion)" Aug 22 19:19:03 *3 BBVERSIONS should read 3 versions within BBVERSIONS Aug 22 19:21:43 But I'm probably thinking about this wrong. I want to try and shim 3 different images into a single image recipe .bb. I have a feeling a bbclass is probably the easier way to do this. Aug 22 19:22:21 i'm using yocto for custom rpi build, how is the config.txt file generated, config.txt file is usually changed at runtime, the file i am referring to is this: http://elinux.org/R-Pi_configuration_file Aug 22 19:23:16 zarzar: Try using fgrep or find to locate that file? Aug 22 19:25:18 hi najuk the file doesn't exist until after the build, i am trying to change it as part of build Aug 22 19:25:26 majuk* Aug 22 19:31:38 overdamped: the BBVERSIONS mechanism is deprecated, and as far as i know isn't used anywhere, by anyone Aug 22 19:31:52 if you just mean how you specify which version of a recipe to use, see the PREFERRED_VERSION mechanism, read the yocto project docs Aug 22 19:38:26 kergoth: Thanks for the pointer! Aug 22 20:11:43 Hrmm... so I am trying to move my Yocto build environment over to an unRaid server's NFS. Mounted the NFS to my ~/yoctobuild/ directory, I got all the sources pulled, setup the build directory, but OE-core is complaining it can't find the build/tmp directory. Aug 22 20:12:04 https://pastebin.com/6FawD6WC Aug 22 20:13:41 The way the error is phrased, it seems like it's looking for that path using the NFS as the root which, yea, that isn't valid. Aug 22 20:13:52 Any comments or questions appreciated. Aug 22 20:15:44 lol, it even wrote the log file for that error to the tmp directory. Aug 22 20:18:41 Alright then, mailing list tells me having the tmp dir on an nfs won't work. That... is not optimal for me. Wieners. Aug 22 20:53:04 what is python-native? Aug 22 20:53:26 python version that runs on the host to assist with somethign during the build Aug 22 20:53:34 'python' would be the target version of python Aug 22 20:54:24 "python-native" versus "python" in openembedded-core Aug 22 20:55:13 as I said, python-native is the version to run on the host, python to run on the target.. Aug 22 20:55:36 compilation options and runtime environments are different. There is (no name) this is target, then -cross, -native, -nativesdk Aug 22 20:55:43 https://layers.openembedded.org/layerindex/recipe/194/ Aug 22 20:55:55 https://layers.openembedded.org/layerindex/recipe/191/ Aug 22 20:55:57 each of these would have a slightly differnt purpose.. -cross for cross compiling, -native to run on the host, and -nativesdk to run on the host w/in the SDK Aug 22 20:56:18 this is what I just explained.. Aug 22 20:56:19 my god why is there such uncontrolled proliferation of options Aug 22 20:56:39 why would i cross compile python Aug 22 20:56:51 its interpreted isn't it? Aug 22 20:57:00 so you can run it on the target, how else would you get it onto the target? Aug 22 20:57:10 or do you mean 'python-cross'. there is no such thing Aug 22 20:57:18 what? dude what are you talking about\ Aug 22 20:57:40 to my knowledge python is not compiled Aug 22 20:57:43 If you want to run python scripts on your target system, you need to cross compile 'python' for the target Aug 22 20:57:49 otherwise how can you run it? Aug 22 20:57:50 what? Aug 22 20:57:56 its a script dude\ Aug 22 20:58:00 your host (x86) binaries are not going to magically run on the target architecture Aug 22 20:58:03 you don't compile scripts Aug 22 20:58:11 something must execute the scripts.. Aug 22 20:58:20 the python interpreter\ Aug 22 20:58:33 not compiled, not cross-compiled Aug 22 20:58:37 interpreted Aug 22 20:58:52 The recipe 'python' -is- the python 2.7 interpreter Aug 22 20:58:53 python interpreter reads and interprets the file Aug 22 20:58:56 as well as modules and other items Aug 22 20:59:09 you need the interpreter to be compiled for the target system and architecture you will run it on Aug 22 20:59:34 the python -recipe- is the source code for python interpreter, modules and related items. Aug 22 20:59:47 ok so do i need python or python-native to run python Aug 22 20:59:52 python-native -recipe- is the soruce code for the python interpreter, modules and related items -- with instructions for building it to run on your host Aug 22 21:00:08 Where do you want to -run- python? host, or target? Aug 22 21:00:20 on custom rpi3 Aug 22 21:00:32 then 'python' that is the target version (once constructed) Aug 22 21:00:40 as I said 'python-native' is to run on your BUILD host.. Aug 22 21:01:14 you may need to do this in order to constrct the target version -- as python (target) may have dependencies on build environment that can only be satisfied by python-native (on your host) Aug 22 21:02:01 sounds like you don't know exactly Aug 22 21:02:06 no extension -- code that runs on the 'target'. -native code that runs on the (build) host... Aug 22 21:02:12 but thanks for the help Aug 22 21:02:24 things that end in -cross, executes on the (build) host, but produces code for the 'target'. Aug 22 21:02:41 stop answering fray Aug 22 21:02:59 In the case of 'python' for the target, you may need python-native to run on the HOST for the build environment, you will need a cross compiler (-cross) to build the software for the target.. and you will end up with a target set of binaries and erlated Aug 22 21:12:27 Hi, I'm wondering about this: in the case of a git SRC_URI, with BB_FETCH_PREMIRRORONLY= "1", is it normal that the upstream SRC_URI is still checked, silently, for extra commits? Aug 22 21:13:18 I have a git tarball in the premirror... and new commits in my upstream... I was expecting the fetch to fail. Aug 22 21:13:19 in the normal case it is to make sure that the system is in sync.. but with premirroronly I would not have expected that.. Aug 22 21:13:22 it sounds like a bug **** ENDING LOGGING AT Wed Aug 23 03:00:00 2017