**** BEGIN LOGGING AT Tue Jan 30 03:00:02 2018 Jan 30 06:41:05 Yocto/Bitbake gurus needed. Simple question about Bitbake dependency DAG resolution. In short: does Bitbake support conditional DAG dependency tree build in the way similar to CMake/BYPRODUCT and Ninja/restat commands? Jan 30 06:44:27 I am currently reading Bitbake manual and found references to task input checksum, which defines whether task is need to be executed. So I am actually trying to find similar behavior for the task output. So the task calculates at runtime whether other tasks which depend on it will be actually triggered. Jan 30 06:46:58 In CMake this usually results in total number of tasks yet to be executed initially calculated. And during build time when task is dropping the output (i.e. when output is generating the same checksum as before) overall amount of tasks is instantly decreased without running them. Jan 30 06:47:08 Daniel__: bitbake-devel list might be a better place for it Jan 30 06:49:03 Thanks. Will ask there. Was expecting that such issue is well known and maybe somebody will point me to the exact documentation page. Jan 30 06:51:54 Daniel__: it's just that yocto chat is a hit and miss, people in europe are still sleeping and probably won't scroll up to answer 'expired' questions after waking up Jan 30 06:52:16 it's something of 'as time permits' venue for us Jan 30 06:55:26 That explains silence. I am in PST time zone. Will try to ask that again tomorrow at morning if that does not violate chat rules. Jan 30 08:33:08 hello, I have a package kernel-image-uimage-4.9.11+gddaf07227a3e_4.9.11-r0_fod_som.ipk that contains the kernel uImage called uImage-4.9.11+gddaf07227a3e How can I rename it to simply uImage? Jan 30 08:33:27 because that is the name that I wish to use in my u-boot environment Jan 30 08:34:20 as far as I can tell the package generation is related to kernel.bbclass, but it is quite confusing on how it is actually created and included in the final image Jan 30 08:37:43 eduardas_m: the default kernel will package the uImage symlink in the 'kernel' package Jan 30 08:55:53 nrossi: not with my BSP, provided by Variscite, it leaves the kernel package empty as the kernel is actually put into a separate partition that does not involve package management Jan 30 08:56:47 I am actually deviating from what Variscite does and want the kernel in my main partition in the boot directory Jan 30 08:57:59 eduardas_m: then your best bet would be revert that part of the kernel recipe back to default with a .bbappend or make your bbappend populate the symlink that you want Jan 30 08:58:54 nrossi: I wonder - u-boot actually follows symlinks for kernel images? Haven't tried that Jan 30 08:59:26 eduardas_m: it will if your boot partition is ext, if its fat, then your boot partition wont allow symlinks or hardlinks Jan 30 09:00:19 nrossi: good to know, thank you Jan 30 09:00:22 eduardas_m: if your using fat, then you might want to consider a postinst for the package that populates the symlink Jan 30 09:00:39 nrossi: I am using ext4 actually Jan 30 09:00:57 eduardas_m: yay, fat is a pain ;) Jan 30 09:04:03 Is it possible to pack whole rootfs ( / ) into rpm packet in Yocto? Jan 30 09:06:51 wooosaii: I would expect that to be possible only if you are just taking bitbake and do the entire BSP from scratch. Not with a usual Yocto release. Jan 30 09:07:59 wooosaii: also I wonder what your usecase would be... if you want relatively reliable rootfs updates, I recommend using SWUpdate Jan 30 09:09:31 eduardas_m, my use case: boot with initrd image, prepare root partition and install rootfs (rpm) to previously made partition Jan 30 09:10:10 eduardas_m, after that reboot and boot to that root partition Jan 30 09:10:55 wooosaii: me and my colleague actually do something similar, but without package management Jan 30 09:11:46 wooosaii: boot into initrd image (recovery mode), prepare root partition and install rootfs from SWU archive (actually cpio) Jan 30 09:12:01 then change u-boot environment to boot into that partition Jan 30 09:12:38 eduardas_m, I know swupdate, we do it in some other project... but this project requirements actually require packet management (rpm + yum) to install rootfs Jan 30 09:12:52 eduardas_m, yes we modify u-boot afterwards... Jan 30 09:13:26 wooosaii: why would you want package management to install rootfs? Jan 30 09:13:56 eduardas_m, client actually wants it... Jan 30 09:14:22 eduardas_m, it would make a lot more sens to use partition image and flash it directly Jan 30 09:14:47 eduardas_m, but they want yum cuz they are accustom to it... :) Jan 30 09:14:55 wooosaii: exactly... or just a tar.gz would do Jan 30 09:15:27 wooosaii: it's not enough to just give them yum inside that rootfs? Jan 30 09:16:27 eduardas_m, I did port yum already... so I have everything prepared now... except the rootfs.rpm packet :D Jan 30 09:17:35 yocto already builds all rpms for me... so I wonder if it is possible to actually pack rootfs into rpm... Jan 30 09:18:12 wooosaii: custom bbclass? Jan 30 09:18:41 wooosaii: if you know Python and bitbake internals - I suppose so... Jan 30 09:18:51 eduardas_m, something like this I guess: http://git.yoctoproject.org/cgit.cgi/poky/plain/meta/classes/rootfs_rpm.bbclass?h=laverne Jan 30 10:33:02 How do I use rootfs_rpm.bbclass Jan 30 10:33:16 I tried INHERIT += "rootfs_rpm" in local.conf Jan 30 10:33:32 does not work... Jan 30 10:38:51 wooosaii: You simply set PACKAGE_CLASSES to what you want, images are built from packages Jan 30 10:39:27 RP, my PACKAGE_CLASSES ?= "package_rpm" Jan 30 10:39:40 but no rootfs.rpm is built? Jan 30 10:40:04 wooosaii: we don't typically turn images into rpms Jan 30 10:40:32 RP, why is then there rootfs_rpm.bbclass Jan 30 10:40:33 ? Jan 30 10:40:39 wooosaii: IMAGE_FSTYPES sets the form it turns the rootfs into but we don't have an rpm version Jan 30 10:40:54 wooosaii: it generates a rootfs *using* rpms Jan 30 10:41:38 RP, I thought it will produce rootfs.rpm... Jan 30 10:41:49 wooosaii: I can assure you it doesn't Jan 30 10:42:14 RP, :) ok... Jan 30 10:42:24 RP, but how can I do it anyways? Jan 30 10:42:38 I want rootfs.rpm... Jan 30 10:42:54 wooosaii: I'd suggest taking the tarball output of the rootfs and converting it to the form you think you want Jan 30 10:43:21 probably it would have to be a fstype of its own, if it has to be in the buildprocess Jan 30 10:43:33 otherwise, external prostprocessing Jan 30 10:43:38 wooosaii: if you can come up with a command, you can then turn it into a new image fstype Jan 30 10:44:36 I already have some custom_image_types... but thought I would not need to do it also for the .rpm :( Jan 30 12:08:55 I'm getting QDBusConnection: name 'org.freedesktop.UDisks2' had owner '' but we thought it was ':1.10' when trying to use the UDisks2 d-bus API on my embedded Linux system. My Qt GUI application is running as root. I am using the d-bus permissions for UDisks2 from the default meta-oe recipe for Rocko. Jan 30 12:09:09 here are the contents of my d-bus configuration for UDisks2: https://pastebin.com/UcAdmV5v Jan 30 12:10:13 If anyone is familiar with such issues, could you please point out to me what I am doing wrong? Jan 30 12:15:11 might have more luck on the dbus channel Jan 30 12:19:51 rburton: actually, was not aware there was one, thanks... still I wonder how often UDisks2 is used in embedded systems?... my colleague has prototyped some stuff with it for usb key mounting on the desktop and it works fine... I suggested it some time ago instead of handcrafted scripts because I saw that mer project uses it... if you happen to know a better solution for embedded Linux in this area, that would be good to hear Jan 30 12:39:50 rburton: Hi is there some problem with your "gettext: beat library detection into shape"? I'm just asking why it wasn't merged with last 2 batches from mut Jan 30 12:40:22 good question, probably because when RP goes to look at mut he skips things with lots of reverts ;) Jan 30 12:40:39 i'll clean it up now and get it posted, we're discussing whether to fire a m2 shortly anyway Jan 30 12:40:48 ok, thanks Jan 30 12:41:09 rburton: I've been meaning to ask you Jan 30 12:41:09 I'm of course fine with dropping my change and revert of it in your queue :) Jan 30 12:43:04 so whats the opinions of having a device-tree.bbclass in meta/? (that is essentially what device-tree.bb does in meta-xilinx) it seems users are interested in building device trees that are not from the kernel. Just curious if its worth upstreaming it and making some useful documentation around it Jan 30 12:44:09 nrossi: I'm ok with the principle if there is a clean class for it Jan 30 12:45:01 RP: anything that pops out which is clearly undesirable from the recipe here: http://git.yoctoproject.org/cgit/cgit.cgi/meta-xilinx/tree/meta-xilinx-bsp/recipes-bsp/device-tree/device-tree.bb Jan 30 12:46:20 nrossi: no Jan 30 12:46:33 rburton: that kernel revert - is there a reason for it Jan 30 12:47:15 RP: ok, I will sort out a bbclass and some documentation and post it for review :) Jan 30 12:47:52 nrossi: sounds good thanks Jan 30 12:57:09 RP: reverted it late one night as there was disusson on the list... Jan 30 12:57:13 forgot to go back Jan 30 12:58:50 rburton: fair enough, I must admit I had that flagged but I have a lot of things flagged Jan 30 13:30:04 Hey, i'm trying to run an arm build of core-image-sato under qemu and I'm having trouble getting graphics to work Jan 30 13:30:48 I'm using the crops docker image to get my environment setup and can easily boot the image with then nographic option but sdl doesnt work and publicvnc never starts Jan 30 13:30:56 Have I missed something obvious here? Jan 30 13:40:25 * zeddii is tring to figure out how a libc-headers version bump went in without my ack Jan 30 13:40:43 RP: I'll send a patch to drop myself as the maintainer for linux-libc-headers and just focus on the kernel Jan 30 13:44:07 zeddii: Sorry, I think it went into mut and I didn't notice it was that recipe :/ Jan 30 13:44:40 I just rebased my 4.15 work on top of it, so I suppose it won't last long. and I'll drink some coffee and chill :P Jan 30 13:44:42 rburton: ^^^ Jan 30 13:45:24 zeddii: 4.15 sounds good. I've been trying to get patches moving a bit faster but things have moved too quickly :( Jan 30 13:45:34 * RP is also distracted with the autobuilder Jan 30 13:45:39 I just get twitchy when something so core updates without a log of testing, and matching newer kernels, etc. Jan 30 13:46:07 since I know, if something fails to boot or work with a syscall (in particular with the retpoline foo) .. I know who gets the bugzilla ;) :D Jan 30 13:46:44 zeddii: we don't tend to see a lot of problems with the kernel headers at least... Jan 30 13:47:12 agreed. but with these asm/gcc/glibc changes for the enter/exit, I was taking a close look. I'll just continue to test on top of my 4.15, not exactly a challenging rebase. Jan 30 13:47:34 zeddii: not for a person with your patch handling foo :) Jan 30 13:47:41 zeddii: thanks and sorry Jan 30 13:48:31 meh. no worries. you have bigger fish to fry. Jan 30 13:50:09 zeddii: do you know if Kevin is working on updates for meta-yocto-bsp? Jan 30 13:50:30 zeddii: I delayed M2 until we had patches, just wondering if I should wait on that Jan 30 13:54:44 RP: I'll ping him now, but won't hear until tonight. he's been very on top of it, and I'm seeing other patches from him .. so hopefully he saw all the bumps. Jan 30 13:54:54 * zeddii is an idiot who forgets to cc' him a lot. Jan 30 13:55:06 zeddii: thanks Jan 30 14:25:11 Has anyone ran into this error | configure: error: in `/home/user/build/poky/build/tmp/work/x86_64-linux/glib-networking-native/2.50.0-r0/build': Jan 30 14:25:11 | configure: error: "Package 'libidn2', required by 'gnutls', not found" Jan 30 14:41:33 hi, I am really confused. Why are my .so files not going into a generated -dev package? Jan 30 14:42:01 lpapp: because FILES_* says so? Jan 30 14:42:48 rburton: I have FILES_${PN}lib = "${libdir}" Jan 30 14:42:55 so you would think that is the cause, however Jan 30 14:43:23 bitbake -e foo | grep ^PACKAGES -> shows: foo-dev earlier than foolib in the output Jan 30 14:43:47 so should foo-dev not pick up the plain .soS in this case prior to foolib even though FILES_${PN}lib = "${libdir}" is specified? Jan 30 14:44:00 if not, what should I do? FILES_* needs to go really that specific? Jan 30 14:44:05 That would be uncomfortable. Jan 30 14:44:12 well, what is FILES_$PN-dev set to Jan 30 14:44:17 nothing Jan 30 14:44:24 do I have to? Jan 30 14:44:30 using -e, what is it set to Jan 30 14:44:38 not what does the recipe set it to Jan 30 14:44:55 as i'm not at your machine reading your recipes, i can't tell what classes you've got Jan 30 14:45:22 it only says "/usr/include" Jan 30 14:45:31 not sure why Jan 30 14:46:25 luckily -e can tell you, as bitbake.conf should be setting FILES_${PN}-dev to include /usr/lib/lib*.so Jan 30 14:47:02 so what to check next? I am really unsure Jan 30 14:47:27 FILES_SOLIBSDEV = "" was specified, however I removed that Jan 30 14:47:33 is there any other flag that could affect it? Jan 30 14:47:34 right well there you go Jan 30 14:47:37 you broke library packaging Jan 30 14:47:42 oh-oh! Jan 30 14:47:45 buggy recipe, very buggy Jan 30 14:47:47 oh god Jan 30 14:47:54 someone put FILES_${PN}-dev = "${includedir}" in there Jan 30 14:48:00 that should be removed and everything will be fine? Jan 30 14:48:09 by default it picks that includedir up and what you said? Jan 30 14:48:10 sounds like you need to burn the recipe and start again Jan 30 14:48:28 just look in bitbake.conf, FILES_$PN-dev is in there Jan 30 14:50:13 rburton: cheers Jan 30 15:19:24 * kergoth yawns Jan 30 15:41:19 I want to change a file in rootfs < /etc/... > where I can change this so it wont cleanup during "cleanall" command and appear in rootfs in modified form Jan 30 15:42:01 xtron: modify the recipe that pulls in the file accordingly. maybe through a bbappend Jan 30 15:43:10 is there a clean way to install a disabled `systemd.service` using `package-a` and enable that service using `package-b`? Jan 30 15:44:19 LetoThe2nd: I think once fetch it remains somewhere in the < tmp/ > directory even I < cleanall core-image > it should be available, my question is where it is, recipe is not fetching it again and again Jan 30 15:46:23 xtron: well that depends on where recipe actually fetches it. if the recipe has reason to think that it doesn't need to get it again as it is not supposed to have changed.. then, yes. Jan 30 15:47:44 I've seen `SYSTEMD_AUTO_ENABLE` but I'm wondering whether it's allowed to enable third party services across package-borders Jan 30 15:50:07 I guess `SYSTEMD_PACKAGES` might work Jan 30 16:17:08 it doesn't. `SYSTEMD_PACKAGES_other_pn` is not available Jan 30 16:30:53 RP: what does it take to have a BSP layer on git.yoctoproject.org (i.e. instead of github)? Jan 30 16:31:51 abelloni: depends on the layer and if its community oriented or commercially driven Jan 30 16:35:38 let say it will support community boards but will be supported by a Linux Foundation member Jan 30 16:47:05 abelloni: that is a grey area :/ Jan 30 16:52:03 T_UNIX: a recipe can't control variables in other recipes Jan 30 16:55:26 rburton: that's why i hoped it would simply inject SYSTEMD_PACKAGES as dependencies and create a task to enable a unit. Jan 30 16:57:21 rburton: what's the supposed way of enabling another package's unit? simply add some kind of `package_postinst(){ systemctl enable foo.service}`? Jan 30 17:02:20 T_UNIX: write a bbappend to control the recipe directly Jan 30 17:02:39 a postinst to enable the service would work, but is a bit yuck Jan 30 17:03:23 I'm trying to provide a dev-env that contains a `demo.service` unit for the developers to enable at will Jan 30 17:03:50 however, if an `expo-image` is generated, it shall be enabled by default Jan 30 17:05:46 s;it;`demo.service`; Jan 30 17:07:46 T_UNIX: rootfs post hook to enable it in the image Jan 30 17:11:12 rburton: thanks for the hint :) Jan 30 17:11:14 Our CM consists of a settings-file and a small addition to local.conf. I wonder if it would be possible to tap into the py bitbake parser and have a common file for the CM and for the local.conf? Anyone tried something like that? Jan 30 17:14:06 Perhaps the most naïve approach is to run bitbake -e and then parse the output, but that feels like overkill Jan 30 17:23:33 sveinse: nothing wrong with -e to get the actual configuration thats being built Jan 30 17:25:45 New news from stackoverflow: fetching from local directory Jan 30 17:26:42 rburton: I ment injecting other vars unused by bitbake into local.conf intended for pre-bitbake setup-stuff. Such as the collection of layers that this configuration shall checkout and use and their version Jan 30 17:48:41 sveinse: it's actually surprisingly easy to write a standalone python script that parses bitbake format files. if you don't need to parse bitbake.conf, and just want to parse a random file, you wouldn't even need tinfoil Jan 30 17:50:16 kergoth: do you know what py module I need to load to just parse a random file? Jan 30 17:51:02 bb.parse Jan 30 17:52:08 d = bb.data.init(); bb.parse.init_parser(d); d = bb.parse.handle(filename, d) Jan 30 17:52:14 iirc that should do it Jan 30 17:53:46 hi all Jan 30 17:54:36 i just build my first yocto image, but i dont know which file to choose from the /images/ folder to copy to sc card Jan 30 17:55:58 can anybody help me? Jan 30 18:01:02 octo_: which platform is it ? for rpi it would be file ending in rpi-sdimg Jan 30 18:03:30 it is for beaglebone green wireless Jan 30 18:03:43 i used the meta-ti layer. didnt know if this is right Jan 30 18:04:05 kergoth: perfect, I'll try that, thanks Jan 30 18:04:11 np Jan 30 20:07:54 kergoth: oh that is easy. trivial enough to wrap into the fuzzy lop python integration too Jan 30 20:10:36 kergoth: ok need to set a variable so it can find classes too Jan 30 20:11:16 kergoth: if i set LAYERDIR and then parse layer.conf, will that do the right thing? Jan 30 20:11:25 been meaning to throw afl at the parser for a while Jan 30 20:11:33 and doing it via bitbake is waaay to slow Jan 30 20:13:12 it should. you'll want to .expandVar('LAYERDIR') to force immediate expansion on it the way bitbake does, most likely, dpeending on what you're doing Jan 30 20:13:50 File "/home/ross/Yocto/poky/meta/classes/siteinfo.bbclass", line 158, in __anon_167__home_ross_Yocto_poky_meta_classes_siteinfo_bbclass throws an exception Jan 30 20:13:51 hmm Jan 30 20:14:54 anyway, dog walk time Jan 30 20:19:46 ah right its bailing as it doesn't know what machine Jan 30 20:19:56 okay find i'll parse bitbake.conf too Jan 30 20:20:51 if you're going to do that, might as well let tinfoil do it for you :) bblayers+bitbake.conf is basically the minimum the cooker will parse at this point Jan 30 20:21:01 course it's probably slightly slower, due to spawning the server.. Jan 30 20:22:05 the server was the problem, afl wants to parallelise the processes Jan 30 20:22:20 ah Jan 30 20:22:29 afl? Jan 30 20:22:36 american fuzzy lop Jan 30 20:22:43 fuzzer with python hooks Jan 30 20:23:51 ah Jan 30 20:32:59 anyone uses ansible for image roll out / update on yocto devices? Jan 31 00:42:49 I have a line like "addtask foo before do_install" Jan 31 00:43:22 Is it possible to only run that line depending on the value of "d.getVar('RELEASE')"? Jan 31 00:50:48 brianm_: a bit messy but bb.build.addtask in a python block would work I think Jan 31 01:11:00 clsulliv: Couldn't quite figure out the arguments, but I got something similar working: `"if d.getVar('RELEASE'): d.appendVarFlag('do_install', 'depends', ' ${PN}:foo')` Jan 31 01:21:31 Does eventhandler honor COMPATIBLE_MACHINE? Jan 31 01:22:19 I have a event handler registered in a recipe which should be run only for specific machine. But, the event handler is getting triggered for any machine Jan 31 01:23:07 Does anyone know how to make sure my eventhandler does not get triggered for machines outside COMPATIBLE_MACHINES? Jan 31 01:57:27 New news from stackoverflow: AM3703 based chip not boot **** ENDING LOGGING AT Wed Jan 31 03:00:02 2018