**** BEGIN LOGGING AT Wed Aug 08 03:00:02 2018 Aug 08 10:00:50 Hi All, Aug 08 10:01:33 How can I force bitbake to fetch git repo. everytime I compile code Aug 08 10:02:16 > bitbake should always fetch git repo. and build it. Aug 08 10:09:18 hello Aug 08 10:09:44 I'm having trouble with mariadb / mysql in yocto - any kind soul online ready to help? Aug 08 10:45:35 Hi, I have a recipe that appends some files and different versions of these files are kept in subdirs named like different machine overrides Aug 08 10:46:23 Now I have the problem, that the rebuild isn't triggered when I switch to an other machine Aug 08 10:47:36 webreformer: You can do "bitbake -C fetch -f " Aug 08 10:51:40 frsc : Thanks Aug 08 10:55:46 Do I need to follow below steps everytime I make some changes in source code, Aug 08 10:55:47 - push it to respective git repository then Aug 08 10:55:47 - fetch it and then Aug 08 10:55:47 - compile using bitbake Aug 08 10:56:26 Is there any shortcut way for development ? Aug 08 11:43:41 webreformer: SRCREV = "${AUTOREV}" should always use the latest commit Aug 08 11:43:50 only recommended for local development Aug 08 11:44:11 no need to use -C fetch -f, just bitbake normally Aug 08 11:47:24 webreformer: while working on the source code you should probably use devtool Aug 08 11:48:49 webreformer: e.g. devtool modify copies your sources to a working directory Aug 08 11:49:40 webreformer: you can change the sources and recompile without needing to fetch et.c Aug 08 11:51:06 webreformer: after finishing you have to bring the changes back to your original code base or create patches Aug 08 11:55:26 ok Aug 08 12:31:11 Hi, I have a bbappend that appends some files to SRC_URI and different versions of these files are kept in subdirs named like different machine overrides Aug 08 12:31:50 Depending on the MACHINE set the recipe picks up the files from the correct subdir and builds the package Aug 08 12:33:10 But I need to force the rebuild after MACHINE has changed. Otherwise the old package for the previous MACHINE will be kept. Aug 08 12:34:02 Anyone has an idea how I can fix the rebuild behaviour? Aug 08 12:50:00 make the package arch MACHINE-specific Aug 08 12:50:05 frsc: ^ Aug 08 12:54:24 rburton: that sounds reasonable, but PACKAGE_ARCH is already set to "${MACHINE_ARCH}" Aug 08 12:54:34 then changing MACHINE should be causing a rebuild Aug 08 12:57:16 rburton: I just checked once more and it doesn't Aug 08 13:04:36 rburton: I can see that PACKAGE_ARCH changes correctly to the name of the selected MACHINE, but no rebuild happens Aug 08 13:16:39 Does anyone know how to force u-boot-fw-utils to use the correct board configuration? Aug 08 13:20:47 New news from stackoverflow: How to disable Warning error flag in yocto recipe Aug 08 13:21:41 rburton: Any idea how to debug such an issue? Aug 08 13:41:25 Hi all, is anyone here familiar with the Vivante driver and its relationship with Qt? I'm getting into a circular argument with it and mesa. Aug 08 13:43:41 Everything seems to have contradicting PROVIDES, I think this is because the Vivante recipe will needs some mesa, as it doesn't do everything. I can't get it to play ball with EGLFS Qt though. Aug 08 13:47:17 I recently see lots of file-rdeps issues on master, anybody seen this too? (ERROR: dbus-1.12.8-r0 do_package_qa: QA Issue: /usr/libexec/dbus-daemon-launch-helper contained in package dbus requires libdbus-1.so.3(LIBDBUS_PRIVATE_1.12.8), but no providers found in RDEPENDS_dbus? [file-rdeps]) Aug 08 13:49:41 JoeR: Are you on i.MX6 and are you using meta-freescale? Aug 08 13:50:12 Yes Aug 08 13:51:22 Setting PREFERRED_PROVIDER_virtual/egl to imx-gpu-viv just upsets other things further down the line, and on and on. Aug 08 13:53:00 I'm using meta-freescale on sumo, this is an upgrade of a working system from Krogoth to Sumo. So I need to change what I have, but I'm not quite sure how. My image includes linux-imx, kernel-modules, kernel-module-imx-gpu-viv, imx-gpu-viv Aug 08 13:53:43 And qtbase, which should be running on EGLFS, preferably backed by the vivante driver. Aug 08 13:55:36 JoeR: Hm, the last time I was using this setup was on morty. I haven't looked into sumo so far. Aug 08 14:01:09 frsc: We don't move forwards between releases of Yocto that often. So it seemed like now was the time to get on the new release. Perhaps it's a jump too far! Aug 08 14:07:30 JoeR: We also just move to a later branch rarely. The last jump we did was from morty to rocko, but on rocko we are using a upstream kernel now without the Vivante drivers, etc. Aug 08 14:19:09 what is the PROVIDES name of the rootfs, so that i can add it to a recipe? Aug 08 14:20:19 name fo the image recipe Aug 08 14:20:31 unless you add an explicit provide Aug 08 14:36:43 Does anyone know how to force u-boot-fw-utils to use the correct board configuration? Aug 08 15:20:03 hi, I have a strange issue with a recipe fetching from git : I'm using PRSERV_HOST = "localhost:0" and PV = "gitr${SRCPV}" so I expect the version numbering to be handled by the prserv. If I add a SRCREV_pn-mysw = "1ad58b2488ce91ce4d0c4aae4189108b0281aae6", I see in buildhistory PV = gitrAUTOINC+1ad58b2488 and PKGV = gitr0+1ad58b2488 . Aug 08 15:20:29 if I then fix the hash to 6bd8208a0754a23b5437cb34a361fc22d7dc0c3b, I see -PKGV = gitr0+1ad58b2488 +PKGV = gitr0+6bd8208a07 +PKGR = r0.0 which is expected Aug 08 15:21:08 but then if I come back to the hash 1ad58b2488ce91ce4d0c4aae4189108b0281aae6, I see -PKGV = gitr2+6bd8208a07 -PKGR = r0.1 +PKGV = gitr0+1ad58b2488 and thus I have the package numbering going backward. Aug 08 15:21:32 Do you have any idea on how to fix this ? Aug 08 15:36:51 eric__31: most likely it pulled the previous package and packagedata from sstate. Aug 08 15:36:57 * armpit tests patchtest Aug 08 15:37:32 kergoth : is that expected even if I clean the sstate cache between both run ? Aug 08 15:37:43 using sstate-cache-management.sh -d --cache-dir=sstate-cache -y Aug 08 15:51:19 New news from stackoverflow: menuconfig window pop-up issue in linux Aug 08 17:11:22 What is the best way to disable or mask a systemd service from starting at boot? I don't want to remove the service just stop it from starting at boot time. Should I create a bbappend file for the service recipe and use the SYSTEMD_AUTO_ENABLE var? Aug 08 17:27:21 is the phytec bsp developer here by any chance? Aug 08 17:28:35 kism: you can ask your question here if its related to yocto/oe Aug 08 17:35:10 after building and flashing to hardware. My device ip address is set to a default static ip that is hardcoded somewhere because everytime I reset it's always the same one. Now I can temporarily change it with "ifconfig eth0 192.168.x.xx" with success moving forward but if the hardware is turned off and on again it defaults to the original one I talk about in the beginning. There is a network settings file in /env/network/eth0 but when i c Aug 08 17:35:43 ip doesnt change. so it appears this file is not what im looking for or there might be another file somewhere overwriting this one. Aug 08 17:39:12 any help appreciated. If there are any other files that set up the boot up ip address I am missing? Aug 08 17:41:44 i attempted to devtool modify -x swupdate and get this error: https://paste.fedoraproject.org/paste/THoR6fjN0tfU-3f9RIowsg Aug 08 17:41:53 i don't understand why. Aug 08 18:05:04 any ideas? Aug 08 18:06:31 nm Aug 08 18:27:10 is there an image type that will generate a gunziped (gz) ext4? Aug 08 18:27:25 i'm looking at image_types.bbclass and i don't see that. Aug 08 18:29:08 just put ext4.gz in IMAGE_FSTYPES Aug 08 18:29:17 gz is a generic image compression type, you can append it to any other type Aug 08 18:29:27 see COMPRESSIONTYPES or CONVERSIONTYPES depending on your oe/yocto version Aug 08 18:29:43 that's what i thought, but i'm not seeing it in the deploy Aug 08 18:34:44 this should work for my hw-test-image.bb file: IMAGE_FSTYPES[hw-test-image] += "ext4.gz" ? Aug 08 18:36:23 any thoughts to my above issue of not being able to save a new ip addr for reboot? Aug 08 18:37:04 IMAGE_FSTYPES is a list Aug 08 18:37:10 IMAGE_FSTYPES[hw-test-image] won't do anything at all Aug 08 18:37:15 that's setting a flag that isn't used Aug 08 18:37:23 IMAGE_FSTYPES += "ext4.gz" Aug 08 18:38:23 it's confusing to me when flags are used and when they aren't.. Aug 08 18:40:21 flags are only used in *specific* cases Aug 08 18:40:28 do not use a flag unless you know it's what you should be using Aug 08 18:41:22 originally flags were intended as a way to add metadata about the metadata. flags control how tasks are run (dependencies, etc), how signatures are calculated, etc. in certain cases they're overloaded as a generic dictionary, but that's fairly rare Aug 08 18:41:49 ok, my bad on flags... lol Aug 08 18:41:58 i.e. the only difference between a variable and a function from bitbake's perspective is the presence of a 'func' flag set to '1'. defining a function with foo () {} just sets key=value and sets the func flag for you Aug 08 18:42:20 kism: so you want to save the ip address you have set across reboots? Aug 08 18:43:30 i would grep for your initial ipaddress and see how/where they're setting it. Aug 08 18:45:44 kergoth: thanks for the correction. Aug 08 18:46:07 kism: is your system using systemd or sysvinit ? Aug 08 18:46:16 np, was hoping to clarify the usage Aug 08 18:46:23 yes, you did. Aug 08 18:47:33 the datastore is really just a dictionary/hash table with additional dictionaries/hash tables about particular keys (flags) and the ability to clone a copy-on-write child of itself (this is how the global configuration data goes into a reicpe's metadata) Aug 08 18:48:16 kergoth: are you showing off? :) Aug 08 18:48:25 i didn't understand 2/3rds of that.. Aug 08 18:48:46 hmm, was hoping to make it seem less mysterious, but i think it backfired Aug 08 18:49:10 :) np, thanks for trying. Aug 08 18:49:46 the copy-on-write thing is an implementation detail, you can think of it like inheritence. a child instance can access all the data from its parent, just happens to avoid actually copying anything until it has to. probably should have left that out :) Aug 08 18:49:50 when you say "the datastore", are you referring to the "d" object that is globally available to the classes? Aug 08 18:49:56 yeah, precisely Aug 08 18:51:30 i had (i think) a bit of a revelation about yocto/bitbake recently. Aug 08 18:51:43 more of a basic way to understand it. Aug 08 18:53:00 think of a matrix of rows and columns. the columns are all the "standard" tasks (do_configure, do_compile, etc.). the rows are recipes which configure/populate those tasks. Aug 08 18:54:19 kergoth: do you have any experience with u-boot-fw-utils? Aug 08 19:01:50 hi, I get this error when building on Arch Linux with python 3.7 / 2.7.15 Aug 08 19:01:51 https://paste.pound-python.org/show/qdpdYzHSpTlq8Bn64HdO/ Aug 08 19:02:19 I know Arch Linux isn't an officially supported host OS, but I am not sure if this is related to python 3.7 Aug 08 19:02:29 so might as well just give a heads up Aug 08 19:05:29 hmm, no Aug 08 19:05:36 isn't related to python version Aug 08 19:05:47 i had relative paths in conf/bblayers.conf Aug 08 19:57:38 do variables in a recipe are available in shell function? Aug 08 20:07:20 xtron: they are expanded by bitbake before the shell function is executed, but if you need them in shell environment (e.g. for some shell script called inside do_compile) then you need to export them Aug 08 20:11:32 khem: systemd Aug 08 20:12:41 yates: I want to set a new ip addr that is saved in memory so when i reboot it uses that one instead yes, The current one I can't see in my local network. Aug 08 20:13:21 sorry for delay I just got back from lunch. Aug 08 20:14:24 I'm completely new and working towards getting a working prototype for internship Aug 08 20:18:44 JaMa: I want to modify them in recipe from shell function, can I do this in shell function + using python d.getvar and d.setvar Aug 08 21:19:31 How can I make my install image boot into a shell instead of running init? Trying to debug an issue on new hardware, kernel can't recognize my emmc so I want to poke around in a shell in the initrd/initramfs (which is it?) Aug 08 21:53:56 Im really liking this new sstate summary Aug 08 21:54:25 emg: pass the init parameter to the kernel, set it as bin/bash Aug 08 21:54:45 emg: or use poky-tiny which does a little more than just that Aug 08 21:55:32 nice to see an sstate summary become default, been using buildstats-summary.bbclass for that for yeras Aug 08 22:01:29 How can I disable systemd-timesyncd service? I am running ntpd instead. Aug 08 22:16:41 packageconfig Aug 08 23:04:37 aehs29: I added init=/bin/sh but that didn't work, I think because of the initrd? Aug 08 23:22:31 emg: it really depends on what youre using but it looks like you should change your init script on initrd to drop to shell Aug 08 23:23:00 emg: basically you want to avoid switching root Aug 08 23:23:01 aehs29: where do I find that? (new to yocto, guy who normally does this isn't here) Aug 08 23:23:41 emg: initrdscripts recipe, init-live.sh might be what youre looking for Aug 08 23:25:15 aehs29: thank you Aug 09 00:00:23 khem, glibc 2.28 test run failed. Aug 09 00:00:45 checking version of make... 3.82, bad, is this the host issue ross talked about ? **** ENDING LOGGING AT Thu Aug 09 02:59:59 2018