**** BEGIN LOGGING AT Fri Oct 04 03:02:52 2019 Oct 04 03:30:27 is it canonical for recipes to use tabs or spaces? Oct 04 04:09:19 mischief: Go with spaces. Some parts like python functions must use four spaces, and then it's simpler to just do that in all of the recipe. OE-core uses tabs in shell functions, but if you're starting a new layer I see no point in mixing. Oct 04 04:11:49 okay. i ask because i think i have a bitbake syntax plugin for vim which i believe is 4 spaces, but my coworker seems to have used tabs in something he wrote :( Oct 04 04:14:40 see also https://www.openembedded.org/wiki/Styleguide Oct 04 04:14:57 I guess that happens. I'd just bring it up for discussion so you can decide what to use Oct 04 04:16:40 The worst things are when people start doing style related changes without first having some kind of common guidelines. It's like a bad race condition :-) Oct 04 04:23:05 rburton: gnight Oct 04 04:23:50 rburton: I already tried reusing the sstate from yoctoproject.org but my stuff doesnt get automatically built so it doesnt exist Oct 04 04:23:59 neverpanic: I'll give that a try Oct 04 05:55:39 erbo: we're just getting started and i've been doing most of the development.. still hardly have any idea what i'm doing, but i have a working image on our hardware ;) Oct 04 06:30:33 I am wondering about the KERNEL_FEATURES variable. I know it should enable specific kernel options. And it reference a .scc file. But I can't find that file. Neither in the layer-directory nor in the build directory. Oct 04 06:31:51 I would like to know how a scc file looks, compared to a kernel configuration fragment file Oct 04 07:43:51 I am on Yocto Sumo, and need to bring in some recipe changes (bug fixes) from newer versions. I don't suppose you can patch a recipe file from a .bbappend? Oct 04 07:47:56 iceaway: you can override a lot of stuff using a bbappend. But depending on how the original recipe looks, and what you need to change, it might be tricky to fix stuff in a bbappend. Oct 04 07:49:14 If it gets too messy, you can always put the newer version of the recipe in your own layer. Oct 04 07:49:52 erbo: thanks, might just do that! Oct 04 08:57:49 How does oe.utils.conditional work? Where is it defined? Oct 04 09:19:54 FILESEXTRAPATHS is working inside a recipe .bb to override git source ? Oct 04 09:34:21 PinkSnake, if i understand correctly, not really: it changes where the local file fetcher searches for files Oct 04 09:37:29 PinkSnake: https://www.yoctoproject.org/docs/latest/mega-manual/mega-manual.html#var-FILESEXTRAPATHS Oct 04 09:38:37 kroon qschulz Thx guys! Oct 04 09:41:42 only for local files. It adds paths where the fetcher will look for local files. You can check which paths and in which order they are traversed with bitbake -e if you want and check FILEPATHS Oct 04 09:55:39 rburton, git log says:Add 0001-mesonbuild-environment.py-check-environment-for-vari.patch Oct 04 09:55:39 as particularly we set windows resource compiler through an Oct 04 09:55:39 environment variable WINDRES. This has replaced the Oct 04 09:55:39 0001-modules-windows-split-WINDRES-env-variable.patch as the code Oct 04 09:55:39 has been refactored. Oct 04 09:55:55 hm shoudl have thought of that ;) Oct 04 09:56:03 so why isn't it upstreamed :) Oct 04 09:56:12 the is_cross patch is broken sadly Oct 04 10:04:50 rburton, yeah, but I wonder why it believes the exe wrapper is needed on target Oct 04 10:05:04 because the machineinfo that early in the build is 'linux none none' Oct 04 10:05:34 the fix is actually quite simple, just need to write a test to demonstrate it for upstream Oct 04 10:05:53 rburton, right, so do I need to do anything? Oct 04 10:05:58 nope Oct 04 10:06:01 cheers :) Oct 04 10:06:06 well working out how to upstream the windres thing would be good ;) Oct 04 11:22:10 hi, is it (still) possible/allowed to share sstate-cache between simultaneous running builds? (i found an mailing list entry form 2016 where it looks like it should be fine) Oct 04 11:33:29 Hi, I building a package that requires python3-config at build time, how can I get this installed in the recipe sysroot? Oct 04 11:34:41 filt3r: I doubt that has changed, but hopefully someone involved in those parts confirm :) Oct 04 11:44:48 filt3r, yes it is Oct 04 11:45:21 thank you very much for the confirmation Oct 04 11:45:28 filt3r, what you cannnot do is remove items from the cache while a build is running (bitbake -c cleansstate, or directly) Oct 04 11:45:50 ok, ye, that makes sense Oct 04 11:55:02 hey guys, I'm trying to build a cmake based recipe that uses llvm on yocto, but I couldn't make it find the llvm cmake files. Do you know the proper way to import llvm on a cmake project? Oct 04 12:38:31 New news from stackoverflow: How to import LLVM on a cmake based bitbake recipe? Oct 04 12:47:10 yocti: I posted that question :) , but got nothing yet Oct 04 12:47:11 Klanticus: Error: "I" is not a valid command. Oct 04 12:48:09 what? Oct 04 12:51:00 Klanticus: yocti is just a bot that posts the stackoverflow questions here Oct 04 12:51:28 oh god Oct 04 12:52:50 I feel ashamed now Oct 04 12:53:27 Klanticus: no worries Oct 04 13:15:42 Klanticus: at least you didn't get into a fight with it **** BEGIN LOGGING AT Fri Oct 04 13:16:54 2019 **** BEGIN LOGGING AT Fri Oct 04 14:39:44 2019 Oct 04 14:56:24 FILES_${PN}-dbg += "${PYTHON_SITEPACKAGES_DIR}/.debug" Oct 04 14:56:42 Do I need that for recipes that build python C libraries? Oct 04 15:15:41 any idea about "nothing provides gdbserver needed by packagegroup-core-tools-debug-1.0-r3.all" Oct 04 15:19:27 pepone bitbake gdb -c cleanall and rebuild ? Oct 04 15:31:24 <__angelo> hello, trying to build a sdk (bitbake image -c populate_sdk) but getting hecking target system type... Invalid configuration `x86_64-myd-sdk-linux': machine `x86_64-myd-sdk' not recognized Oct 04 15:36:40 pepone: for any recent release you don't need to set FILES_PN-dbg at all Oct 04 15:37:03 __angelo: maybe dont use a hyphen in your os name Oct 04 15:37:24 if that fixes it we should make that error out at construction time Oct 04 15:42:12 <__angelo> rburton, mmm i am in a schroot, Oct 04 15:42:16 <__angelo> BUILD_SYS = "x86_64-linux" Oct 04 15:42:16 <__angelo> NATIVELSBSTRING = "ubuntu-16.04" Oct 04 15:42:44 i'm guessing myd-sdk is your name Oct 04 15:43:52 <__angelo> mm how can i check it ? Oct 04 15:44:06 is it not the name of your distro? Oct 04 15:44:27 <__angelo> myd is the name Oct 04 15:44:32 <__angelo> distro name Oct 04 15:44:47 PinkSnake: thanks that worked Oct 04 15:45:34 <__angelo> oh, i have set SDK_VENDOR = "-myd-sdk" Oct 04 15:47:12 <__angelo> changed to "-myd_sdk" and seems to work Oct 04 15:47:51 ok i'll file a bug Oct 04 15:48:30 <__angelo> i probably copy and pasted that name/scheme, but is possible i changed and _ to a - becouse i like hypens Oct 04 15:48:46 <__angelo> rburton, a special thanks Oct 04 16:10:56 locale -a , alwasys returns "C, POSIX" Oct 04 16:11:10 Why aren't other locales installed with core-image-minimal Oct 04 16:13:43 Seems they were build in tmp/work/cortexa8hf-neon-poky-linux-gnueabi/glibc-locale/2.29-r0/image/usr/share/i18n/locales Oct 04 16:14:49 But that is not in the rootfs of the image Oct 04 16:36:08 it's called minimal for a reason. but adjust IMAGE_LINGUAS to control which are installed in any given image Oct 04 16:42:40 I have already set it Oct 04 16:42:42 ENABLE_BINARY_LOCALE_GENERATION = "1" Oct 04 16:42:42 GLIBC_GENERATE_LOCALES = "en_US.UTF-8" Oct 04 16:42:42 IMAGE_LINGUAS = "en-us" Oct 04 17:38:27 Hi all, I am experiencing a strange crash with a QT (5.9.7) application on my embedded platform built with yocto: "The Wayland connection experienced a fatal error (No child processes)". Can anyone shed some light on this, or point me in the right direction? Oct 04 17:43:17 pepone: i wouldn't be at all surprised if minimal's .bb overrides that as a part of its goal. minimal. minimal is almost never the image you want Oct 04 17:43:27 might want to just create yourself a recipe at this point Oct 04 17:43:31 image, that is Oct 04 17:46:59 kergoth: ok I will give it a try Oct 04 18:03:15 kergoth: seen minimal images that are rather large :) Oct 04 19:44:35 we're running into problems with sstate cache and our mirror server. Does anyone know, should ever tgz file have a corresponding tgz.siginfo file? Oct 04 20:33:49 It looks like the answer might be yes Oct 04 20:34:03 I think what happened is our build server failed while uploading it's sstate cache Oct 04 20:34:23 so then there was a bunch of archive files with missing siginfo files which caused any builds using that server to fail Oct 04 20:34:46 I'm wondering if there is an upload mechanism that wouldn't leave the server in a bad state on failure to upload Oct 04 20:35:01 if I upload the siginfo first and then the archive, does anyone know if that would work? Oct 04 20:37:53 marler899712: You need both, and we ran into the same problem. We fixed it by running a script that cleans out sstate removing files missing their match Oct 04 20:39:51 So if you just have siginfo...will the build still fail? Oct 04 20:39:58 I know if you just have the archive it will Oct 04 20:48:26 marler899712: I don't remember off the top of my head, sorry... having the siginfo w/o the archive doesn't seem very useful anyway :) Oct 04 20:49:24 yeah, I just want to know if there is a way to upload files without creating a race condition where the server temporarily has invalid files that will cause a build to fail Oct 04 20:49:57 a better solution might to to patch oe-core to handle this case with a warning instead of an error Oct 04 20:50:09 marler899712: IIRC, it only looks for siginfo if the archive is found, but don't quote me on that Oct 04 20:50:21 ok, well it's worth testing out Oct 04 20:52:31 marler899712: if you have the space, rsync new version to a different directory, then mv it into place after a successful upload? Oct 04 20:55:08 smurray: I don't know if Crofton told you, but I borrowed part of your container presentation for a talk I gave at the platform security summit. It was very helpful :) Oct 04 20:55:32 smurray: that is a pretty good idea Oct 04 20:56:36 JPEW: heh, no, he didn't, but cool. I need to get back to runc tinkering one of these days, but I'm entering AGL crunch time for CES now Oct 04 20:57:25 * smurray tries to forget he has an AGL blog post to write Oct 04 20:58:10 The problem we have with rsync + sstate is multiple builds trying to upload at the same time... the sets of sstate intersect, so we need them all, but it causes issues where they overlap Oct 04 20:58:42 I'm not sure if the mv would help in that case :( Oct 04 20:59:13 yeah I was thinking about solutions like copying the old contents into the new before doing a mv, but that doesn't handle multiple uploaders Oct 04 20:59:19 you would need some sort of queue in that case Oct 04 21:00:02 marler899712: It's a harder problem than it first seems :) Oct 04 21:00:24 yeah, if we fixed yocto to handle missing files then it wouldn't matter as much Oct 04 21:00:38 perhaps need something like a WebDAV backend to plug into Oct 04 21:00:57 or maybe the hash equiv server solves this, JPEW? Oct 04 21:01:30 no, hash equivalence doesn't protect you from missing siginfo Oct 04 21:03:02 JPEW: the hash equivalence server is read only, ie multiple builds couldn't run against it and populate it at the same time? Oct 04 21:06:07 smurray: err, not sure what you mean :) multiple builds can report equivalent hashes. The HE server doesn't look at the sstate files at all, it just tracks hashes Oct 04 21:06:38 hash equiv just lets you use sstate with a different checksum than this task has, if we know it'll produce the same results. Oct 04 21:06:59 right, okay, I'll shut up Oct 04 21:07:05 ;) Oct 04 21:08:47 Anyway, there are probably a few ways of solving this. Someone should bring it up at the 3.1 planning meeting Oct 04 21:09:00 But, I have to take $CHILD to swim lessons now. TTYL Oct 04 21:09:19 raise $CHILD **** ENDING LOGGING AT Sat Oct 05 03:01:28 2019