**** BEGIN LOGGING AT Fri Jan 19 03:00:01 2018 Jan 19 05:33:15 New news from stackoverflow: Yocto: Adding glibc libraries to the rootfs Jan 19 07:40:23 hello Jan 19 07:40:52 i have apackage that produces generally 642-bit binaries with only one exception file which is 32-bit Jan 19 07:41:28 the build fails with ERROR: QA issu: Architecture did not match. How do i fix this? Jan 19 07:41:58 the task is package_qa() Jan 19 08:01:16 let's rephrase: how do i inform yocto that one binary is of different architecture, so that it no longer fails in this "fatal" error? Jan 19 08:02:08 cornel: if its actually intendet and correct, then setting some QA skip should do the trick Jan 19 08:02:28 LetoThe2nd, it's correct and intended. Thank you very much! Jan 19 08:40:33 hmmm, INSANE_SKIP_${PN} = "arch" seems to be ignored Jan 19 08:40:59 LetoThe2nd, isn't this the right way to skip 'arch' QA on this package? Jan 19 08:43:55 probably something specific to this project ... Jan 19 08:53:09 hello guys Jan 19 08:53:27 i am trying to create an ext3 image in yocto 1.7.3 with capabilities Jan 19 08:53:56 i set the capabilities in post_install scripts and they are present in the location where packages are installed Jan 19 08:54:10 however they are present in the final ext3 image Jan 19 08:54:17 are not * Jan 19 08:55:40 pkg_postinst_${PN}() { PSEUDO_UNLOAD=1 setcap CAP_SYS_NICE,CAP_DAC_OVERRIDE+ep $D$bindir/some_binary } Jan 19 09:33:55 New news from stackoverflow: Yocto recipe to manage Go dependencies (solved) Jan 19 10:08:54 Hey, anyone using Intersil TW9912 ADC on I.MX6 4.9.x kernel ? Jan 19 10:11:33 rburton: the solution was to use arm-poky-linux-gnueabi-libtool Jan 19 11:04:06 New news from stackoverflow: Yocto recipe to manage Go dependencies Jan 19 12:34:19 New news from stackoverflow: Yocto recipe to manage Go dependencies with dep tool Jan 19 13:09:50 is there somewhere more documentation on UPSTREAM_CHECK_GITTAGREGEX ? Jan 19 13:10:19 jaeckel: you can ask here Jan 19 13:10:29 jaeckel: I know more than most about it ;) Jan 19 13:12:50 I'm trying to set the version of a library I want to build from git and I thought probably I need to somehow modify UPSTREAM_CHECK_GITTAGREGEX so it can correctly parse my version schema Jan 19 13:13:27 which is basically vMAJ.MIN.PATCH with optional '-rcX' added Jan 19 13:15:54 but my output package version is still called AUTOINC+commitid Jan 19 13:18:52 jaeckel: that's only for checking what the latest upstream version Jan 19 13:19:10 jaeckel: to set the version built, you need to set PV Jan 19 13:19:49 jaeckel: and to specify the particular git commit id, you set SRCREV - PV is not matched against tags or anything, as tags are unstable Jan 19 13:24:12 kanavin: and I thought when I set PV to ${SRCPV} and SRCREV to "master" I could probably easily tag all releases and get them properly named ;) Jan 19 13:25:57 jaeckel: if you only ever plan to use tagged commits, you can skip SRCPV stuff altogther, and just set PV to match the tag, and SRCREV to the commit id of it Jan 19 13:26:38 jaeckel: if you plan to use commits in the middle of development, it's a bit more tricky, and involves SRCPV (oe-core has plenty of examples for both) Jan 19 13:27:19 jaeckel: the idea is reproducibility, tags can move around or disappear, but commits are forever Jan 19 13:27:26 kanavin: is there a way to "always track the latest master" without having to adapt the recipe? Jan 19 13:27:52 kanavin: and still have the output called by the tag that is currently on master Jan 19 13:27:57 jaeckel: no Jan 19 13:28:28 it would be a support disaster as someone's build failure would be impossible to properly reproduce Jan 19 13:29:28 jaeckel: you can however use externalsrc for local development, and do git pull on that tree Jan 19 13:31:49 kanavin: okay, so by setting SRCREV = "master" and PV = "" I would have the same behavior, right? Jan 19 13:32:39 jaeckel: maybe but I wouldn't accept anything like this into a layer Jan 19 13:36:48 kanavin: yeah, I also wouldn't propose it like that, but allowing this still leads to the same support disaster :D Jan 19 13:43:05 if you want to track master then you can use AUTOREV Jan 19 13:43:17 but you'll regret it at some point) Jan 19 13:43:33 kanavin: I also think that SRCREV should be set to the commit id, is there a way to use the tag on that commit id as version? Jan 19 13:44:15 jaeckel: you can probably hook a bit of py to run git describe to get a nice name but if you're tracking head then what if there isn't a tag... Jan 19 13:44:32 the only thing I want is to prevent to have more that one thing to update :) Jan 19 13:44:41 s/that/than Jan 19 13:44:59 all you need to update is SRCREV (new revision) and if you care, PV. Jan 19 13:45:13 you could not care and just have the PV contain a git sha Jan 19 13:46:01 I'd like to have something like AUTOPV which uses the tag as PV, which is why I landed at UPSTREAM_CHECK_GITTAGREGEX :D I feel like turning around Jan 19 13:46:49 jaeckel: sounds like you want gitpkgv.bbclass from meta-oe Jan 19 13:46:54 # - GITPKGVTAG which is the output of 'git describe' allowing for Jan 19 13:46:54 # automatic versioning Jan 19 13:47:13 etc Jan 19 13:47:15 the class has docs Jan 19 13:48:04 ey, nice! that could be it Jan 19 14:00:10 SSL isn't working on our rocko image, and I think its related to libgnutls-openssl27 missing from target. Should I depend on this package, or is there a better name for it? I'm sceptical to link to version numbers, yet there is libgnutls-openssl it seems Jan 19 15:01:20 If an image includes more than one shell interpreter (for example bash and busybox) how can I select which will be used? (that is, what /bin/sh will point to) Jan 19 15:23:51 ttllkk: update-alternatives controls it Jan 19 15:24:49 how is that controlled from Yocto, though? Jan 19 15:25:29 ttllkk: the recipes inherit the update-alternatives class and set priorities for their provider Jan 19 15:28:53 thanks, I will have a look Jan 19 15:46:12 * kergoth yawns Jan 19 15:49:29 anyone here use NetworkManager? Jan 19 15:49:56 "Identity not validated against passwd-file" Jan 19 15:50:03 asking for a freind, really :) Jan 19 15:58:08 sure Jan 19 15:58:15 * zeddii nods wisely Jan 19 16:09:18 rburton: that worked! thanks Jan 19 16:11:08 what's the proven approach to develop in yocto when you have a monorepo with all your different projects in there? does that work with devtool or is there another best practice? Jan 19 16:11:58 jaeckel: write separate recipes for each of the projects, the repo will be fetched only once and shared between them Jan 19 16:25:17 so I have a repo with projects a and b in their respectively named folder, how do I reference that folder in the recipe? and what will devtool resp. devshell do with it? Jan 19 16:29:11 I'd be happy to just get pointers in the documentation, but I couldn't find them yet... Jan 19 17:35:24 New news from stackoverflow: Qt WebEngine 5.6 Recipe do_install Jan 19 18:05:32 New news from stackoverflow: Can u-boot verify a signed hardware? Jan 19 18:28:10 is there a way to (re)build just the .dtb files instead of the whole linux kernel? Jan 19 18:33:34 yates: dont think so Jan 19 18:34:04 the dtb build functions are appended to normal configure/compile/install functions of kernel Jan 19 18:36:55 i guess there are dts[i] dependencies on, e.g., .h files that might get generated/patched in the kernel build process? Jan 19 18:37:14 right Jan 19 18:37:41 in general it appends to build process if dtbs are supposed to be part of kernel Jan 19 18:52:11 What is the easiest way to convince my yocto build to use mdev instead of udev? Jan 19 18:55:16 spierepf: VIRTUAL-RUNTIME_dev_manager = "busybox-mdev" Jan 19 18:55:31 read conf/local.conf.sample.extended Jan 19 19:42:13 i'm having a strange problem: building the kernel is not generating any z... files in tmp/deploy/... Jan 19 19:42:23 this has been working for months. Jan 19 19:42:30 now today it isn't working.. Jan 19 19:43:46 is there a way to clear out cache and/or do other sort of non-typical initialization? Jan 19 20:02:49 no zImage anywhere in my build directory. Jan 19 20:33:29 so im trying to build rocko for the pine64. I've got arm-linux-gnueabihf-gcc installed on ubuntu 16.04, however halfway through the bitbake process, it fails where it can find arm-linux-gnueabihf-gcc. Jan 19 20:33:57 someone mentioned earlier that it was because it wasn't in the local path, but i don't see where I can define that. Jan 19 20:36:01 New news from stackoverflow: Pip install on Yocto failing || How to initialize Bluetooth in a startup script with Yocto Poky Linux || What can cause the following error unable to initialize libusb: -99 Cripes Jan 19 21:25:08 I can't seem to make the fetcher to fetch a repo from gitlab that is public. Jan 19 21:25:26 I got it to work a few days ago when the repo was private. Jan 19 21:26:11 The URI looks like this: git://git@gitlab.com/dposey/dbus-cxx.git;protocol=ssh Jan 19 21:28:00 When I fetch I get: Permission denied (publickey). Jan 19 21:29:37 Now gitlab docmentation says: "Public projects can be cloned *without any* authentication Jan 19 21:30:56 So do I need to tell the fetcher *not* to use my key somehow? Jan 19 21:31:32 How does it fail authentication when none is required? Jan 19 21:31:56 how do you change a kernel menu config in yocto? Jan 19 21:34:16 xthunderheartx: you need to use a https URL not an ssh URL Jan 19 21:36:10 yates: bitbake virtual/kernel -c menuconfig Jan 19 21:37:38 Ok so you only need token if you are pushing via https? Jan 19 21:38:57 I'll give it a whirl. Thx Paul Jan 19 21:39:00 yates: delete your tmp/ dir and rebuild Jan 19 21:44:24 where is the modified .config stored? Jan 19 21:48:31 khem: you mean do a "bitbake linux-variscite" (my normal kernel build)? Jan 19 21:48:51 where did yocto store the new configuration? Jan 19 21:49:47 is it in a patch somewhere? Jan 19 21:52:19 khem: that did not work. Jan 19 21:52:59 after rebuilding and reflashing the new kernel, "zgrep CONFIG_RTC_DRV_ISL1208 /proc/config.gz" still reports the "is not set" Jan 19 21:54:05 khem: was i supposed to run "bitbak virtual/kernel"? Jan 19 21:55:24 hello? Jan 19 21:55:51 yates, fwiw, i ran the same command and got garbage out while building morty Jan 19 21:55:59 i don't understand how this "virtual/kernel" thing works.. Jan 19 21:56:40 and im not sure how to add the arm toolchain to the bitbake path Jan 19 21:56:53 the morty version works just fine, but rocko doesn't see it. Jan 19 21:59:15 i'm using morty Jan 19 21:59:29 ok Jan 19 21:59:35 so you and i are in the same boat Jan 19 22:00:07 yes Jan 19 22:00:34 https://stackoverflow.com/questions/42543390/trying-to-use-linux-menuconfig-with-yocto-morty Jan 19 22:00:57 let me try that after this rocko build fails Jan 19 22:01:28 wasutton3: i was just looking at this: https://community.nxp.com/thread/360368 Jan 19 22:10:40 if youre getting that oe-gnome terminal error I think you can just set OE_TERMINAL to screen as a workaround Jan 19 22:10:46 yates: wasutton3 ^^ Jan 19 22:12:09 https://pastebin.com/raw/HhKfuxVa Jan 19 22:12:12 thats my error Jan 19 22:14:17 yates: the new config file will be saved on your kernel build directory, e.g. build/tmp/work/qemux86-poky-linux/linux-yocto/4.12.18+gitAUTOINC+b66a4f9730_558fe84d69-r0/linux-qemux86-standard-build/ Jan 19 22:14:44 so you can build morty with the 4.12 kernel? Jan 19 22:15:05 replace the paths according to your MACHINE and version Jan 19 22:15:13 there you should see a .config file Jan 19 22:15:45 wasutton3: I have no idea why youre getting the expand error Jan 19 22:21:00 hmmm could it be a python2/python3 mapping error? Jan 19 22:23:45 aehs29, could it be because im building rocko and morty on the same box? Jan 19 22:24:37 aehs29: ok fine. but doesn't that mean it's "nonvolatile"? Jan 19 22:42:50 aehs29: ok fine. but doesn't that mean it's "nonvolatile"? ???????? !!!!!! Jan 19 22:43:16 i mean it will get lost the next time i bitbake the kernel Jan 19 22:43:34 yocto is a pain in the ass Jan 19 22:44:06 BSP kernels generally are not tied to YP release versions Jan 19 22:44:20 i imagine its not bad once its set up yates Jan 19 22:44:43 but getting the build env set up right is a REAL PITA if you have non-standard hardware Jan 19 22:44:47 then its like gentoo Jan 19 22:44:57 so how do i make this new .config file "non-volatile"?!? Jan 19 22:45:13 yates: there are few ways Jan 19 22:45:19 update it in the git repo that supplies the kernel? Jan 19 22:45:35 yates: you should use config fragments Jan 19 22:45:41 yates: probably Jan 19 22:45:52 make savedefconfig and copy the resulting defconfig to kernel sources, then edit the kernel recipe to use it Jan 19 22:45:54 yates: take a look at the kernel dev manual and you'll find whats right for you Jan 19 22:46:12 or second option is to use config fragments if you care for subset of options Jan 19 22:46:30 yates: http://www.yoctoproject.org/docs/current/kernel-manual/kernel-manual.html Jan 19 22:46:38 yates: what khem said ^^ Jan 19 22:48:45 virtual/kernel is an alias which will work irrespective of kernel recipe name. This is to abstract out knowing exact kernel recipe name, sometimes there are multiple recipes providing kernel for a given board and the names are specific to boards etc. e.g. linux-raspberrypi linux-mainline etc. Jan 19 22:48:59 khem: yes, all we need is that one driver configuration enabled, so how do config fragments work? where are they documented? Jan 19 22:49:19 yates: which BSP are you using ? Jan 19 22:49:28 it depends on if BSP layer has implemented them Jan 19 22:49:49 if people derive their kernel out of linux-yocto them it comes for free Jan 19 22:50:49 khem: i don't know - where is the bazillions of files and options would that be specified (BSP layer)? Jan 19 22:50:50 usually you define it in a file e.g. usb.cfg and then add file://usb.cfg to SRC_URI and store usb.cfg along with other patches in metadata Jan 19 22:51:03 ..where in the.. Jan 19 22:51:13 content of usb.cfg would be CONFIG_FOO=y/n/m Jan 19 22:51:34 yates: which board are you building for Jan 19 22:52:08 a variscite dart 6 ul (imx6 ull) Jan 19 22:53:11 ah Jan 19 22:54:27 usb.c does not have relation to the usb bus, does it? Jan 19 22:54:35 usb.cfg, that is Jan 19 22:55:25 no that was just an example you can have many options in file and file could be called anything.cfg Jan 19 22:56:53 so https://github.com/varigit/meta-variscite-fslc is the BSP layer for you Jan 19 22:58:18 they dont seem to implement the config fragment parts in kernel recipe :( Jan 19 22:58:52 you need something like https://github.com/kraj/meta-variscite-fslc/commit/1f33a83190683e6ee68bb9b79af490bd7eeb6fa8 Jan 19 23:00:23 actually, I don't see the point in config fragments because they make it difficult to know what the final config looks like Jan 19 23:01:58 you might, some others might not. e.g. if someone is writing a driver alone Jan 19 23:02:28 he would like to use the defaults from upstream but then ensure to enable/disable options for the driver he/she is interested in Jan 19 23:03:14 and then they are not actually enabled because they depend on whatever config option that is not enabled Jan 19 23:04:30 and they get confused :) Jan 20 02:22:18 Hello, Has anyone been able to build static qemu-native? **** ENDING LOGGING AT Sat Jan 20 03:00:01 2018