**** BEGIN LOGGING AT Mon Nov 26 02:59:59 2018 Nov 26 09:16:30 New news from stackoverflow: Yocto Rocko bitbake process reboot on Ubuntu Nov 26 10:23:17 o Nov 26 10:23:34 I've seen bitbake causing logouts multiple times Nov 26 11:25:30 hey Yocto community, i'm having trouble adding Nodejs 10.x to a Sumo Poky image. I've tried aaronovz1's meta-nodejs fork which is the closest I could find to a maintained recipe but I keep getting build errors that I don't know how to fix. Anyone got a working recipe to share ? Nov 26 11:26:55 meego: do you actually need node 10 Nov 26 11:26:58 or would node 8 be ok Nov 26 11:27:41 hm so a recipe was upgraded and now takes three times as long to build Nov 26 11:27:55 anyone know some good tools to profile what a cmake build is up to? Nov 26 11:28:10 wondering if there's any magic to see what was built, what took time, etc Nov 26 11:31:04 cslcm: node 8 would probably be ok. I've noticed there's a node 8 recipe in meta-oe. But if I'm not mistaken that package uses bitbake to build node/npm projects. Whereas the meta-node recipe lets npm do its thing, which I would prefer for familiarity sake Nov 26 11:31:38 at build time you mean? Nov 26 11:32:25 the meta-nodejs stuff is ancient, I struggled to get it to build for days. I did manage it eventually, though. What error are you getting? Nov 26 11:38:35 cslcm: Would love to know how you managed. I'm using the pyro branch of aaronovz1's fork. I saw comments online suggesting I should use openssl 1.0 instead of 1.1 so I've added a bbpend w/ PREFERRED_VERSION_openssl = "1.0.%". Now I'm getting /poky/build/tmp/hosttools/ld: cannot find -lcrypto /home/ubuntu/poky/build/tmp/hosttools/ld: cannot find -lssl errors Nov 26 11:39:13 I didn't make that change, and I don't recognise that error Nov 26 11:40:24 i'm a little confused by what you mean by "package uses bitbake to build node/npm projects" Nov 26 11:40:50 bitbake can use npm during the build process, but once deployed it's just regular node Nov 26 11:41:00 afaik Nov 26 11:45:19 AFAIK, with meta-oe nodejs, bitbakes handles building the dependency tree, deduplicating it, fetching licences, and building checksums for each, downloading source, building native modules Nov 26 11:45:26 looking for source of this Nov 26 11:48:11 I use meta-oe nodejs and it's just regular npm on the target device, i think that just means during the yocto build process Nov 26 11:52:09 yes, that's likely Nov 26 11:52:28 so you're using vanilla meta-oe node now ? Nov 26 12:05:44 meego - yeah, though actually I copied the recipe to my own overlay and bumped it to node 10 Nov 26 12:16:57 rburton: libdnf takes longer because they rewrote it in c++ Nov 26 12:17:08 RP: are there any current issues to look into? Nov 26 12:18:34 kanavin: I'm busy trying to sort out gpg problems in master and previous releases as the tests are proving unreliable but I think we have most things under control. Things are looking happier than they were on Friday Nov 26 12:19:46 RP: I got perl-sanity to build world without issues :) Nov 26 12:20:01 RP: will try with meta-perl next, and maybe we can give it to AB as well Nov 26 12:21:30 kanavin: damn them! Nov 26 12:25:21 kanavin: have you briefly looked to see if it does anything stupidly slow we can short circuit? Nov 26 12:25:29 it got *a lot* slower to build Nov 26 12:31:05 rburton: let me run it again - it did got slower, but only midly so on my machine Nov 26 12:31:26 kanavin: perf reports says 300s slower Nov 26 12:31:34 and dnf slower too Nov 26 12:34:40 cslcm: ok sounds promising, i'll try that Nov 26 12:36:51 rburton: just building libdnf-native here, it's most certainly not 300s Nov 26 12:36:58 maybe 10 or os Nov 26 12:37:00 so Nov 26 12:40:31 Hi guys could you please tell me terminal command to get yocto version. Nov 26 12:42:03 dheeraj - bitbake --version and compare it to this list to get the yocto release version - https://wiki.yoctoproject.org/wiki/Releases Nov 26 12:48:19 Could somebody help me debug a 'fatal error'? I have a recipe that installs a .py file. The .py file starts with #!/usr/bin/python3. The build complained that, as a result in had a runtime dependency on python3, so I added RDEPENDS_${PN} = "python3". Now it says heartbeat.py contained in package iot-code requires /usr/bin/python3, but no providers found in RDEPENDS_iot-code? python3 is built and working with CORE_IMAGE_EXTRA_INSTALL += Nov 26 12:48:19 "python3" Nov 26 12:49:05 rburton: target libdnf is also reasonable here Nov 26 14:05:46 RP why did we update gnupg in rocko ? Nov 26 14:07:07 there is absolutely no info if these are bug fix only changes Nov 26 14:07:12 armpit: I was testing something in -next, we can't merge that Nov 26 14:07:33 armpit: I'm actually trying to fix gpg selftest bugs which means we need a relocatable gnupg native Nov 26 14:07:43 ah Nov 26 14:08:05 armpit: I just put the relevant patches there to talk with you Nov 26 14:08:23 armpit: doesn't even build as there are dependency version problems :( Nov 26 14:09:30 k. Nov 26 14:10:04 I had a back port request and I wanted to rebase with the right stuff Nov 26 14:12:32 ah so -next is more like a mut? Nov 26 14:14:22 armpit: well, I wanted to put those bits *somewhere* and as there was nothing in rocko-next... Nov 26 14:20:04 RP. my thud build last night has an issue checkout with mingw Nov 26 14:20:59 armpit: yes, I've broken the config in the migration, should be an easy fix Nov 26 14:21:19 what I was thinking Nov 26 14:21:27 armpit: I got it mostly fixed before you started running builds :) Nov 26 14:22:06 k. should I hold off on a rocko build till later? Nov 26 14:25:09 armpit: just pushed what I hope is the fix Nov 26 14:25:41 armpit: I did run builds for everything over the weekend including publishing artefacts Nov 26 14:28:55 thanks Nov 26 14:51:52 rburton, you missed Rburrito combination Nov 26 14:52:00 haha Nov 26 14:56:13 /ⵑ\ AΤΤN︓ Tһⅰs ⅽһɑᥒᥒel has ⅿoⅴeⅾ to irϲ．freеᥒoԁе.ᥒеt ﹟oѕⅰrⅰsⅼab /!⧹ Nov 26 15:08:37 how does the owner mapping work in bitbake? Nov 26 15:10:06 i.e. how does the bookeeping work so that wic knows which uid/gid a file in the final image is supposed to have? Nov 26 15:11:13 T_UNIX: AIUI, its "pseudo" magic :) Nov 26 15:12:26 the simplest way, I could imagine, would be a namespace and a mapping file. Nov 26 15:12:28 yes, pseudo Nov 26 15:12:35 pseudomagic is better than pseudoscient tbh Nov 26 15:12:40 pseudoscience* Nov 26 15:12:54 aka fakeroot on steroids Nov 26 15:13:05 I'd like to use wic "standalone" to create some images. Nov 26 15:13:28 run it as root, or inside your fakeroot of choice Nov 26 15:15:23 yeah. I guess so. Nov 26 15:15:36 rburton: LetoThe2nd thanks for the insight :) Nov 26 15:40:52 Quick question - what is best practice for editing the crontab for a specific user in yocto? I have a python script that I would like to run as a particular user. Ordinarily I would use crontab -e for the user, but clearly that's a runtime thing Nov 26 15:45:09 la_croix: just write the file to the right place Nov 26 15:45:32 rburton Oh, simple as that? Perfect, thank you Nov 26 15:45:54 cron doesn't care who wrote it, just that the file has the right name/format/ownership **** ENDING LOGGING AT Mon Nov 26 15:48:38 2018 **** BEGIN LOGGING AT Mon Nov 26 15:50:10 2018 Nov 26 16:08:58 RP: shall i merge master-next mingw to master? Nov 26 16:09:03 i suspect that's why the master ab run was failing Nov 26 16:17:55 rburton I tried doing that, with this recipe: http://paste.ubuntu.com/p/sbBCnC3T7p/ and I got a permission denied error... I was under the impression that the build system didn't exactly have permissions, and could do what it wanted Nov 26 16:18:24 la_croix:that's writing to the host Nov 26 16:18:36 rburton Oops... Nov 26 16:18:46 rburton: it throws a lot of warnings but probably, yes Nov 26 16:19:03 rburton: it is why master failed as I merged test code for it Nov 26 16:19:15 la_croix: also that's not the format for per-user crontabs Nov 26 16:19:33 rburton do I need to use ${D}/etc/crontab? Nov 26 16:19:47 Apparently it's the format for cronie, according to: Nov 26 16:19:49 la_croix: ${D}${sysconfdir}/crontab would be more correct Nov 26 16:20:07 https://git.yoctoproject.org/cgit.cgi/poky/plain/meta/recipes-extended/cronie/cronie/crontab Nov 26 16:20:18 rburton Ok, I'll give it a go Nov 26 16:20:24 fair enough, never really paid attention Nov 26 16:21:16 rburton Should there be a slash between. ${D} and ${sysconfdir}? Nov 26 16:21:33 la_croix: sysconfdir etc are already / prefixed Nov 26 16:21:44 Ok, thank you :) Nov 26 16:25:38 rburton : cannot create /home/ubuntu/dev/poky/build/tmp/work/cortexa7hf-neon-vfpv4-poky-linux-gnueabi/iot-cron/1.0-r0/image/etc/crontab: Directory nonexistent Am I being stupid? Nov 26 16:28:42 in a FILES_${PN}, is there a way to specify a directory and all subdirectories? Nov 26 16:28:57 like /usr/bin/wx/* ? Nov 26 16:30:14 or if you specify /usr/bin/wx, does that get all the files and subdirectories under /usr/bin/wx? Nov 26 16:43:50 la_croix: make it first Nov 26 16:44:18 you'll need to mkdir ${D}${sysconfdir}, as the recipe staging directory is empty out of the box Nov 26 16:44:23 you're not writing into an image at that point Nov 26 16:44:50 yates: if it contains a directory, it picks up all contents Nov 26 16:44:58 rburton: ok, thanks Nov 26 16:45:22 yates: eg see bitbake.conf, FILES_${PN}-doc has just ${docdir} Nov 26 16:45:45 rburton Hmm, wouldn't cronie have already done that, since I added DEPENDS = "cronie"? When I've tried building an image with cronie, but without my recipe, it creates a default crontab at /etc/crontab, so surely by the time the cronie recipe has run, this directory and file exist? Nov 26 16:46:02 where are variables associated with standard paths like bindir, docdir, etc. defined Nov 26 16:46:12 la_croix: your dependencies have nothing to do with the empty directory each recipe stages itself into Nov 26 16:46:24 i mean where are they documented? Nov 26 16:46:37 of course if cronie is writing a crontab and you're writing a crontab, then they'll conflict Nov 26 16:46:44 so i'd bbappend cronie to stop that Nov 26 16:46:57 i looked in the glossaries for the bitbake, development tasks, and project reference manuals and no go Nov 26 16:47:01 yates: bitbake.conf is fairly self-explanatory for the paths Nov 26 16:47:20 tbh the names are effectively several-decade-old standards Nov 26 16:47:38 prefix, bindir, docdir, all gnu coding standards/lfs names Nov 26 16:48:59 rburton Ok, I'll bbappend, thank you Nov 26 17:32:28 if a do_compile fails, are the files that were in use (e.g., the /usr/lib directory) available somewhere? tmp/work/..? Nov 26 17:33:29 build/tmp/sysroots/? Nov 26 17:43:34 i'm seeing the wx include files there (build-hw-test-image/tmp/sysroots/imx6ul-var-dart$/usr/include/wx-3.0/wx/app.h), but those files were not reported when i had no FILES_{PN}, only stuff in /usr/bin and /usr/lib. should i just add them anyway to FILES_${PN}? Nov 26 18:00:14 yates: by in use you mean the ones that were being used to build that package? Nov 26 18:00:58 yates: those should be on the recipe-sysroot directory inside your WORKDIR Nov 26 18:02:16 yates: the files you need for FILES_${PN} wouldnt be there yet, since do_compile failed and those are "put" there on do_install and do_package gets them later from ${D} depending on what you got on FILES_${PN} Nov 26 18:02:31 yates: but I'm still not sure if I understood your question Nov 26 18:07:36 aehs29: thanks - i think you uncovered a flaw in my thinking. rethinking... Nov 26 18:19:58 RP: I’m back working on my image type again, and I’m seeing a “permission denied” error when running the new do_image_ command. Just looking for a pointer .. are those commands run under psudeo ? It doesn’t look like it to me, but I wanted to confirm, since that might explain why I’m getting the error. Nov 26 18:20:19 aehs29: there is no recipe-sysroot dir there: https://paste.fedoraproject.org/paste/Px2gfKMjoXujptrX--MNsQ Nov 26 18:22:43 zeddii_home: can you pastebin the error output? Nov 26 18:23:54 the log.do_image, i.e. Nov 26 18:24:28 nope. this is debug code. nothing to paste. Nov 26 18:24:31 my question stands. Nov 26 18:24:38 is it run under pseudo. Nov 26 18:24:44 * zeddii_home will continue grepping Nov 26 18:25:03 zeddii_home: yes, they run under pseudo Nov 26 18:25:13 ok. so it isn’t a uid/guid zero issue. Nov 26 18:25:29 zeddii_home: I will admit the specifics of image generation are not forefront in my mind atm Nov 26 18:25:42 heheh. no worries. I’m learning as I go. just very slowly. Nov 26 18:26:45 some system call I’ve never used is erroring out. I’m going to have to abandon this approach anyway and use some simpler tools. Nov 26 18:26:54 zeddii_home: image.bbclass: d.setVarFlag(task, 'fakeroot', '1') is what does it Nov 26 18:27:12 aha! I grepped for pseudo. that’s how it excaped me. Nov 26 18:27:14 escaped even Nov 26 18:27:22 isn't it "sudo"?... Nov 26 18:27:35 zeddii_home: it had me wondering for a second too :) Nov 26 18:27:50 yates: our fakeroot emulator is called pseudo Nov 26 18:28:03 ah Nov 26 18:46:08 yates: thats because youre building from sstate, do a cleansstate for that task and you'll see the recipe-sysroot and recipe-sysroot-native directories Nov 26 18:55:20 aehs29: did you mean to do a "bitbake -c cleansstate" followed by "bitbake "? if so, that did not change the dorectpru structure - still no recipe-sysroot Nov 26 18:55:31 s/dorectpru/directory/ Nov 26 18:56:40 the date/time is updating, so i know it's being regenerated Nov 26 18:58:42 i'll keep reading the docs and try to figure out what's happening. Nov 26 19:04:18 yates: yeah thats what I meant, what exactly did you do?, can you post what taks have been run, basically the contents of the log.taskorder in the temp directory Nov 26 19:04:41 yates: perhaps theyre not there because the prepare_recipe_sysroot task has not been executed Nov 26 19:07:07 aehs29: thanks. here is the log.taskorder: https://paste.fedoraproject.org/paste/v6G86YAG5ttx1xp2lAP~mA Nov 26 19:08:29 is prepare_recipe_sysroot part of the default do_configure? Nov 26 19:08:54 yates: what release version are you using? Nov 26 19:09:02 morty Nov 26 19:10:10 i've defined my own do_configure, yyi Nov 26 19:10:11 fyi Nov 26 19:10:29 https://paste.fedoraproject.org/paste/ZGHhRxb-ALHceqlWlNSALA Nov 26 19:11:42 the error is that one of my files is not finding "#include " so i wanted to see where that file is (if it is there at all) Nov 26 19:12:30 yates: ok thats why the directories dont exist, recipe specific sysroots was added on Pyro IIRC Nov 26 19:14:33 i see. thanks for that. Nov 26 19:14:46 so where would i find the sysroot that the do_compile is using? Nov 26 19:14:53 in morty? Nov 26 19:17:12 yates: so basically, the include files should be somewhere on tmp/sysroot..., but I'd suggest adding a verbose flag to gcc (-v), before compiling, that will give you a better looking log.do_compile , which should contain lines like "include search starts here..." or something like that, that will tell you where gcc is looking for include files, then you need to find the file (on the sysroots), and lastly Nov 26 19:17:14 figure out why is gcc not looking in the directory where the include file is and either add it or fix it, depending on the "why" is it not looking there Nov 26 19:18:04 yates: there could be several things happening, e.g. the file is in the right location and gcc is not looking for it there, or maybe gcc is looking in the right places but the file isnt there Nov 26 19:22:46 ok, right. Nov 26 19:23:16 i actually think i am getting a bit closer. it's the latter of the two possibilities you stated. Nov 26 19:23:28 yates: I'd also suggest upgrading to a newer release if you can, since it has (more) support Nov 26 19:23:58 the -I option is generated via the helper wx app wx-config. it appears that is not running Nov 26 19:24:31 WX_TREE_CPPFLAGS = `wx-config --cxxflags` Nov 26 19:25:00 so i'm on my way. thanks much for your help. Nov 26 19:42:40 aehs29: thanks for the multiconfig fix! Nov 26 19:49:32 RP: np, I just wish I had noticed before Nov 26 19:50:07 RP: can you merge that on thud as well? Nov 26 19:51:02 yates: np, hope you fix it soon Nov 26 19:51:39 aehs29: yes, we need to backport that one Nov 26 19:52:18 yates: those -config binaries are the work of the devil, which is why we have a binconfig-disabled class to break them if possible Nov 26 20:18:28 rburton: to break them? why would you want to break them? Nov 26 20:23:38 yates: remove them in preference to using pkgconfig Nov 26 20:32:58 yates: because they're sometimes so broken we need to break them so things break obviously and we can patch in pkgconfig Nov 26 20:43:23 hi, looking for some help with qtwebengine ;) Nov 26 20:43:47 kaspert: what help? Nov 26 20:47:02 i have an imx6dl board. I've tested gifs animations performance, qtwebengine vs AnimatedImage QML. Qtwebengine: 3 gifs, cpu util 60%. AnimatedImage: 10gifs, cpu util 6%. Now i'm looking for a way to fix that issue. Nov 26 20:54:21 mostly related to h/w accelaration Nov 26 20:55:02 RP: the musl failure you reported is not reproducible here with bitbake -ctestimage core-image-sato is coming out to be passing all the tests for qemux86_64 Nov 26 20:55:23 RP: I am using master-next Nov 26 20:56:10 is it possible for you to give it another shot ? may be on a different OS on builder node Nov 26 20:56:19 khem: I did drop the change from -next Nov 26 20:56:34 I know Nov 26 20:56:44 khem: It also appeared twice, once on qemux86 and once on qemux86-64 Nov 26 20:56:58 http://git.openembedded.org/openembedded-core-contrib/log/?h=yoe/mut Nov 26 20:57:08 this is musl upgrade + master-next Nov 26 20:57:24 khem: Were you using kvm? Nov 26 20:57:52 gold verses binutils? Nov 26 20:58:26 RP: yes Nov 26 20:58:35 no gold no systemd Nov 26 20:58:47 kvm yes Nov 26 20:59:07 I did this test based on todays master-next Nov 26 20:59:38 and its using security flags like poky Nov 26 21:09:23 khem: I can retest it but we're missing something :/ Nov 26 21:10:10 RP: if it fails a stacktrace would be good Nov 26 21:10:29 I can try to repeoduce it with that Nov 26 21:10:51 khem: I could try and get it to save the images, see if you can reproduce with those? Nov 26 21:14:06 RP: https://pastebin.com/3DKTJ7wm, another difference is I am using opkg instead of rpm probably it should not matter here Nov 26 21:17:39 khem: ah it was core-image-sato-sdk Nov 26 21:19:30 RP: yes you reported this https://autobuilder.yoctoproject.org/typhoon/#/builders/27/builds/231/steps/7/logs/step1c Nov 26 21:19:43 and it was building core-image-sato-sdk as well Nov 26 21:20:13 khem: right, it builds sato and sato-sdk but it says it was sato-sdk that fails at the bottom Nov 26 21:36:35 any idea when Linux-yocto 4.19 is gonna be available ? it's the next LTS release ... Nov 26 21:37:33 * khem points Marex to zeddii_home Nov 26 21:37:47 RP: yes, I tested sdk image as well Nov 26 21:39:53 khem: that worked too? :/ Nov 26 21:40:04 yes Nov 26 21:41:15 bluelightning_: Paul, you around? Nov 26 21:43:36 khem: zeddii_home == bruce ? Nov 26 21:43:49 yes Nov 26 21:44:16 khem: thanks Nov 26 21:48:31 Could anybody give me a hand with python packages refusing to build? I have a recipe that builds numpy, which works fine, and now I'm trying to install a package that requires numpy - 'numba'. The recipe looks like this: http://paste.ubuntu.com/p/qvM5W5N2mc/ and the error is that it can't find numpy: https://paste.ubuntu.com/p/jfy39GyzvB/ Nov 26 21:53:55 aehs29: technically no, taken today off sick to take care of my son who is sick Nov 26 21:54:03 aehs29: feel free to email me though Nov 26 22:32:41 la_croix: perhaps you need to inherit setuptools3 Nov 26 22:33:36 khem Already doing that Nov 26 22:34:24 khem This is the full recipe: http://paste.ubuntu.com/p/qvM5W5N2mc/ and http://paste.ubuntu.com/p/8myTSTgJhZ/ Nov 26 22:42:16 what tune settings shall i select to have a x86 32bit userland but 64bit kernel? IOW i need a x86_64 toolchain and build all userland with -m32... Nov 26 22:42:33 multilib Nov 26 22:42:53 this multilib stable/supported? Nov 26 22:43:21 can't see why not Nov 26 22:45:48 derRichard: Its part of our automated testing Nov 26 22:46:20 RP: good to know :) Nov 26 22:46:59 let me see how to use it Nov 26 22:51:42 hm, not sure if i understand multilib-example.conf Nov 26 22:51:55 will it build all libs for x86_64 _and_ x86? Nov 26 23:06:00 derRichard: you end up with two namespaces, one for the normal build and one with the multilib variant, e.g. bash and lib64-bash or bash and lib32-bash Nov 26 23:06:18 so it just depends how you set it up and what you target it to build Nov 26 23:15:06 khem: I included musl in this test run so we'll see Nov 26 23:20:09 RP: thx! Nov 26 23:30:20 khem Any guesses? :/ Nov 27 00:35:14 zeddii: zeddii_home: bump regarding linux-yocto 4.19 ? Nov 27 00:35:20 bluelightning: Hi, are you attending LCA 2019 ? Nov 27 01:29:13 Anyone around? Nov 27 01:31:13 I'm having trouble with setuptools whilst building scipy Nov 27 01:35:25 scpy is a pain Nov 27 01:35:33 hav eyou found the prior art on it? Nov 27 01:36:28 Crofton Nope, I've done a lot of googling Nov 27 01:36:37 hang on Nov 27 01:37:09 It says it can't find setuptools, even though 'inherit setuptools3' is in the recipe. The only SO question I found had 'inherit setuptools3' as the answer Nov 27 01:37:27 Also, I've built 10 other python packages using the same recipe structure, and not had any problems Nov 27 01:38:18 la_croix, https://github.com/scipy/scipy/issues/8571#issuecomment-436233233 Nov 27 01:38:28 scipy is a pain Nov 27 01:38:34 much messed up shit Nov 27 01:39:19 Crofton|work I'll add that line and try building. Why is it such a mess? Nov 27 01:40:05 read the thread Nov 27 01:40:19 reminds me, need to look at tit again sometime Nov 27 01:41:07 Crofton Same problem... Have I done something stupid? http://paste.ubuntu.com/p/zHmrqyCtP5/ Nov 27 01:42:25 what am I looking at? Nov 27 01:43:44 Crofton The recipe, this is the other bit: http://paste.ubuntu.com/p/M2mPtvVkHh/ Nov 27 01:44:22 I thinkk you need to look closely Nov 27 01:44:32 I sustecpt setup.py isn't a standrad form Nov 27 01:46:13 Crofton Sorry to be stupid, but I don't really know where to start Nov 27 01:46:37 read the setup.py file Nov 27 01:46:41 scipy is painful Nov 27 01:46:55 readin ghte "recips" postsed not sure they really work :) Nov 27 01:48:17 Crofton In terms of the recipes, these are what I was given when I came on here and asked how to install python packages. ;) I've been using the same structure, just changing names, and checksums, and it has worked fine for everything apart from scipy Nov 27 01:49:27 compare setup.py in scipy to standard Nov 27 01:49:34 scipy isn't straight forward Nov 27 01:49:49 Yeah... it looks complicated Nov 27 01:50:40 Crofton Surely there's a recipe somewhere... I can't be the first person to try to install scipy on yocto :/ Nov 27 01:54:27 see link Nov 27 02:02:14 Crofton Oh, thank you Nov 27 02:20:28 Crofton I've got as far as this: https://jumpnowtek.com/yocto/Add-Fortran-support-to-a-Yocto-build.html (the recipe complained it had no fortran support), and I've followed the instructions. Now I get: No recipes found for /home/ubuntu/dev/poky/meta-custom/recipes-devtools/gcc/gcc-runtime_4.8.bbappend Nov 27 02:29:49 what is meta-custom? Nov 27 02:29:58 * Crofton isn't a poky user Nov 27 02:34:51 Crofton It's just a layer I've put all my crap in Nov 27 02:35:26 Crofton To be honest, I think this is more effort than it's worth. I'm going to modify the code and try to get away with not using scipy Nov 27 02:35:58 some stuff is also in numpy Nov 27 02:36:07 last resort you can build scipy on device Nov 27 02:36:25 not sure what gcc-runtime_4.8bbappend is Nov 27 02:36:36 but gcc version isn;t 4.8 Nov 27 02:36:47 mayb eyou have bad instructinos for turning on FORTRAN Nov 27 02:37:59 # Enabling FORTRAN Nov 27 02:37:59 # Note this is not officially supported and is just illustrated here to Nov 27 02:37:59 # show an example of how it can be done Nov 27 02:37:59 # You'll also need your fortran recipe to depend on libgfortran Nov 27 02:37:59 #FORTRAN_forcevariable = ",fortran" Nov 27 02:37:59 #RUNTIMETARGET_append_pn-gcc-runtime = " libquadmath" Nov 27 02:38:05 from local.conf.extended Nov 27 02:41:03 Crofton It's ok, I really appreciate your help, but to be honest, I just need to get an audio file into an array, and I've found a way to do it that doesn't take scipy Nov 27 02:43:15 good :) Nov 27 02:43:22 write answer Nov 27 02:43:27 er right answer Nov 27 02:49:30 :) **** ENDING LOGGING AT Tue Nov 27 03:00:01 2018