**** BEGIN LOGGING AT Fri Nov 06 02:59:57 2020 Nov 06 04:17:29 question about state cache: does it get invalidated if I run: MACHINE= bitbake ? Nov 06 04:19:40 I'm building the same set of recipes on two "MACHINE" types, but I'm noticing that the second "MACHINE" is running a ton of "do_fetch", "do_unpack", "do_populate_lic". (NOTE: before starting either of these bitbake invocations I have removed my "downloads" directory, but left the state cache in tact) Nov 06 04:31:02 maybe I need to use separate build dirs for each MACHINE type? Or possibly separate state caches? Nov 06 05:26:15 meaning, I wouldn't have expected to have to redownload a bunch of stuff if I retained my state cache... Nov 06 06:39:21 wberrier: sstate can be shared among machines you dont need to keep them separate Nov 06 06:55:16 Good morning, I've got some QA warnings during building an image. I'm not sure if they should concern me. TI (which provides the configuration) says I should not worry about this :/ Nov 06 06:55:34 For example: external-arm-toolchain-2019.12: external-arm-toolchain: /sbin/ldconfig is owned by uid 1000, which is the same as the user running bitbake. This may be due to host contamination [host-user-contaminated] Nov 06 07:48:17 yo dudX Nov 06 07:53:46 alejandrohs yes, I can SSH to localhost from inside qemu but can't from qemu to my computer and from computer to qemu Nov 06 07:56:43 DanmerZ: are you using slirp? is this a really run of the mill linux on your host? Nov 06 07:58:01 LetoThe2nd runqemu qemux86-64 example-image nographic slirp Nov 06 07:58:17 DanmerZ: slirp is the problem. Nov 06 07:58:29 DanmerZ: remove it and use the tun/tap interface Nov 06 07:59:04 LetoThe2nd great, thank you Nov 06 08:00:09 IIRC there was a patch that fixes the port forwarding with slirp on the ML lately, but can't find it at the moment Nov 06 09:19:40 Hi! My yocto shows the message "Started telephony service" during bootup. I could systemctl mask that service to switch it off, but is it maybe a package I need to remove from the poky distro? Nov 06 09:20:56 My distro just includes conf/distro/poky.conf but I believe it does not add packages for telephony stuff Nov 06 09:22:06 fbre: distros don't have packages. images do. Nov 06 09:23:33 hmm, OK, my built Image is core-image-minimal which I tweaked a bit by using my own machine and distro Nov 06 09:23:37 fbre: ofono, turn off the 3g distro feature Nov 06 09:24:23 rburton: does that trigger some expansion in a packagegroup? Nov 06 09:25:05 is ofono part of core-image-minimal? Nov 06 09:28:13 Is it DISTRO_FEATURES_remove = " 3g " ? Nov 06 09:28:31 _remove is bad. Nov 06 09:28:40 or DISTRO_FEATURES_remove = " ofono " ? Nov 06 09:28:54 https://theyoctojester.info/session_14/main.html see note by chris Nov 06 09:29:06 better just set the features directly as needed. Nov 06 09:34:36 3G_REMOVE ?= ā€œ3gā€ Nov 06 09:34:37 3G_remove = ā€œ${3G_REMOVE}ā€ Nov 06 09:34:37 This way? Nov 06 09:35:37 *sigh* Nov 06 09:35:58 copied from Chris' note Nov 06 09:36:04 what the problem with just writing the DISTRO_FEATURES = "..." line in your distro just as you want it to be? Nov 06 09:41:46 Since my .conf file just includes conf/distro/poky.conf and actually calls DISTRO_FEATURES_append = " myadditionalstuff ". Isn't it better to run the strategy to extend/shrink a base system than to rebuild it from scratch? It's a principle of tweaking than to reinvent. Maybe that's a philosophy question but I'm not an expert and try to Nov 06 09:41:46 aovid changing a running system Nov 06 09:42:02 Can somebody share his SSD wear levels after some period of regularly building Yocto? I have concerns of build time on HDD .vs. costs of SSD. Nov 06 09:42:40 I searched and found somewhere opinion that SSD dies after 0.5..1 years of building yocto daily Nov 06 09:44:05 fbre: generally i'd agree, but for things like DISTRO_FEATURES i've found just "nailing it down" the way i want it to be is working better. Nov 06 09:44:32 amerlyq: well, "opinion" Nov 06 09:45:09 LetoThe2nd: OK, I trust your experience. Thanx a lot (y) Nov 06 09:45:34 amerlyq: it depends on the ssd so much, as well as the actual count of builds. one image build per day? 24hrs constantly fully loaded? that makes a huge difference. Nov 06 09:46:11 Hi, i have a question concerning the recipe-sysroots. I try to build a recipe with a python binding extension. When building it tries to link against ssl, which leads to the error: recipe-sysroot-native/usr/lib/libssl.so: file not recognized: file format not recognized. Is "recipe-sysroot-native" the correct one? Or shouldnt it link against recipe-sysroot ? Nov 06 09:46:14 fbre: there's nothing bad with puling poky in, if it works for your use case. i would just always reset/overwrite DISTRO_FEATURES to a known state that matches my needs below the include. Nov 06 09:47:02 fl0v0: if the thing you want to build (the binding) shall go onto the target, then -native is wrong. Nov 06 09:47:31 Hey, quick question about psplash - Is there interest in a patch that adds support to disable the progress bar as a runtime option? :) Nov 06 09:47:44 LetoThe2nd: yeah it should go on the target. Thought so, thanks for confirming Nov 06 09:48:23 LetoThe2nd: Now i have to find out why it does use the wrong one :/ Nov 06 09:48:28 LetoThe2nd: I wonder a bit why 3g is part of a base image. It feels like very special to me since most of the embedded world is without mobile phone stuff Nov 06 09:48:58 fbre: see, thats why i don't trust the default distro features. QED. Nov 06 09:49:29 fl0v0: almost certainly some part of the build script not using the environment properly but sticking to its call directory. Nov 06 09:52:40 wertigon: define "interest". if you've got something, then send a rfc patch and see what happens. :) Nov 06 09:53:47 LetoThe2nd: basically two cases 1) Jenkins with clean nightly build + 40 builds daily from sscache; 2) DevPC, only sscache, around 30 incremental rebuilds daily with full image packing. Is it too much for SSD? Is it not? The main concern is Yocto writing too many very small files -- each resulting in rewrite of 1MB block of SSD, which totals to astronomic write load. Nov 06 09:56:58 amerlyq: i think it mostly boils down to the real amount of data being written. the dwpd rating should give you a good idea there. i can say that most devs here build on ssds, and i haven't heard cases of unexpected burn-outs there. Nov 06 09:57:28 LetoThe2nd: Well, our use case; if our payload app ever crashes, we want to show the psplash with an error message but no loading bar, I have a patch ready but unsure if I should try to upstream. :) Nov 06 09:57:44 amerlyq: plus, if you're on a real setup, i'd expect raid controller+cache to do away with most of the smallish block amplification anyways. Nov 06 09:58:25 wertigon: it sounds like a legit usecase. i see nothing wrong with showing a rfc patch and stating that in the explanation. Nov 06 09:59:15 LetoThe2nd: Ok, let me just pretty it up and send then Nov 06 09:59:49 LetoThe2nd: how about general tip from https://wiki.yoctoproject.org/wiki/Build_Performance ? Nov 06 09:59:55 It's pretty good already, just want to double check whitespace etc Nov 06 09:59:55 > Put the build directory on its own disk. This is good practice in its own right since the build system has a tendency to wear disks heavily. Nov 06 10:00:18 and "heavily" it's like ~4 months in 2014-2015... Nov 06 10:01:40 amerlyq: that page is about 8 years old, i would outright ignore it. Nov 06 10:02:43 we will soon transition our jenkins setup to an all-ssd one, just for the sake of pure performance Nov 06 10:03:59 Hi, Hopefully this is an easy question. If I run a step after build history has been updated to produce a report what's the best way to do this? Nov 06 10:05:37 dmation: i would make a fancybuildhistory.bbclass, which inherits buildhistory and adds what you want. then use the fancy one. Nov 06 10:10:47 LetoThe2nd, that seems like a nice, simple solution. Do you think something like buildhistory_commit_append() work in this case? Nov 06 10:11:13 dmation: i don't know. try and find out. Nov 06 10:11:58 Will do, thanks for the lead Nov 06 10:13:18 have fun! Nov 06 10:25:12 rburton: yay for another pythonsci Nov 06 10:28:44 hi, is there a way to create mixedcase group names with the useradd.bbclass? I have a piece of precompiled legacy software, which should be integrated an the usergroups are hardcoded in the binaries -.- Nov 06 10:29:10 like HiImAGroupName Nov 06 10:44:02 I am trying to extract SDK extended bitbake myimage -c populate_sdk_ext and it fails with "do_sdk_depends: The file /usr/lib/libcollectdclient.so.1 is installed by both collectd-plugins and collectd, aborting" Nov 06 10:44:44 Is it something wrong with the image itself? Nov 06 10:51:10 DanmerZ: myimage bilds fine? Nov 06 10:51:58 yes Nov 06 10:53:19 then i would start digging intt the recipes, sounds like one bundles the lib into the -dev package accidentially Nov 06 11:56:14 hello, can anyone provide me with a definitive answer whether it is possible to boot an aarch64 Linux kernel using a 32-bit (ARMv7) compiled mainline u-boot? Nov 06 12:51:31 eduardas: somebody in #armlinux certainly can Nov 06 12:52:03 eduardas: all the cool kids are there, we're just the boring neckbeards that compile stuff Nov 06 13:03:31 LetoThe2nd: ouch, that hurt :D Nov 06 13:04:20 LetoThe2nd: I still kind of expect a lot of people here to be more involved... If you only do bitbake recipes for a living and don't really have to do u-boot/kernel development, I envy you :) Nov 06 13:05:49 eduardas: it just depends. but your question is really soemthing they know instantly there. Nov 06 13:06:44 LetoThe2nd: I asked it yesterday on both u-boot and armlinux... did not get a reply within about half a workday Nov 06 13:06:57 eduardas: seriously? surprising. Nov 06 13:07:05 don't know how annoying I should get :D Nov 06 13:07:57 there is a u-boot contributor I could nag, but I'm trying to be polite Nov 06 13:08:28 and not to abuse people's time too much Nov 06 13:08:30 Isn't it faster to build and check it yourself in that case? Nov 06 13:08:57 amerlyq: I did not get it to work, but my case was vendoruboot + mainline Nov 06 13:09:30 amerlyq: I don't really have a system on hand that supports both mainline u-boot + mainline Linux Nov 06 13:13:11 If you already tried, your best bet to find ready recipe from someone, or debug this yourself the hard way. AFAIK it must work if your CPU can work in compat mode with older instruction set of ARMv7, and kernel must be able to migrate into upper RAM after being loaded by uboot, but it has so much things which may go wrong, that you basically have only two above choices. Go nag them, or shoot Nov 06 13:14:55 But with vendoruboot even if you get working recipe, you 50/50 won't be able to succeed with it anyway. So... jtag till second coming? Nov 06 15:26:30 good afternoon Nov 06 15:28:54 RP: I think the patch I sent on bitbake-devel, has a bug on it. I use self.mc in the base class, but it's not defined, would it be fine to move the `mc` inside the base class NoCache? Nov 06 15:33:05 roussinm: possibly, I'd need to look at that code in more detail... Nov 06 15:46:03 I'm using mdev in my distro and modprobing a kernel module. does mdev have any effect on what modprobe does and what creates and removes /dev/my-device-[0-1] every time I run modprobe (-r)? Nov 06 15:46:50 I'm having a really hard time setting permissions on the device node so regular users can use the device too :( Nov 06 16:37:11 khem: thanks, I'll try a few more configurations/iterations to try avoiding downloading stuff during the build (after have deleted the downloads dir) Nov 06 17:21:29 hi, when I'm using ${IMAGE_ROOTFS} inside image.wks should I rename it to image.wks.in? Nov 06 17:22:08 is it *.in extension to expand variables inside wks ? Nov 06 17:24:35 I can't find any documentation on this topic Nov 06 17:26:59 dev1990: yes, that's correct Nov 06 17:29:53 ok thanks Nov 06 21:13:56 zeddii: around? cloud-init you just imported fails with installed-vs-shipped QA Nov 06 21:14:19 really ? works fine here. which arch ? I built for arm* and x86* Nov 06 21:14:24 https://paste.ubuntu.com/p/8RCDwYD2NN/ Nov 06 21:14:36 I guess every arch in builds with systemd Nov 06 21:14:43 nope Nov 06 21:14:49 I actually tested in on target, with systemd Nov 06 21:15:17 this was rpi4 world build Nov 06 21:16:18 https://pastebin.com/TeXrW0ir Nov 06 21:16:41 I have the same log from qemuarm64. I'll see if I can make it happen here, maybe my local distro is masking something. Nov 06 21:28:59 FILES_cloud-init-systemd=" /usr/lib/systemd/*" Nov 06 21:29:35 export nonarch_base_libdir="/usr/lib" Nov 06 21:29:54 openembedded-core/meta/conf/bitbake.conf:38 Nov 06 21:29:54 # "${nonarch_base_libdir}/systemd" Nov 06 21:30:03 maybe you're using something like usrmerge? Nov 06 21:33:38 only if it's on in poky, I have an integration distro that just inherits that and builds, no significant tweaks. Nov 06 21:34:01 I'll come up with a fix though, that should be using variables, the original cloud-init was super old. Nov 06 21:37:31 ok, thanks Nov 06 22:58:23 khem: thanks for the help. Turns out I just needed separate DEPLOY_DIR for each MACHINE. Then it doesn't re-download all the source. Nov 06 22:59:00 wberrier: hmm it should not need anything actually Nov 06 22:59:49 RP: something is broken in master-next/bitbake I am getting do_populate_lic failures in random recipes but master works ok Nov 07 00:16:25 khem: hmm, I just queued for testing, not looked yet Nov 07 00:17:20 you will see them I got atleast 10-12 in my image build Nov 07 00:23:51 khem: my test build exploded too. Not sure which patch it is though :/ Nov 07 00:27:14 khem: maybe one of the bitbake ones... Nov 07 00:27:51 right its somthing between those 4 commits on bitbake Nov 07 00:28:02 khem: reverted one of them, trying that Nov 07 01:46:52 Hello, does anyone have experience with qualcomm/code-aurora based yocto BSPs? Nov 07 01:49:43 I'm working on a project where we are designing a product with a board based on a reference design that QC has provided a yocto BSP for. I'm wondering if it makes sense to create our own bsp and inherit from QC's layer (and also how to do that, ideally initially identical) or if we should add a machine (or distro?) to their bsp layer (and how to go about doing that) Nov 07 01:50:07 I've been reading developer guides, but I figured I'd ask in here if anyone has done it before in case someone has an opinion on approaches Nov 07 02:10:54 A simpler question. If I have a meta-oem-bsp layer, and I want to create a meta-mine-bsp layer that builds the exact same thing as meta-oem-bsp but with one package override, how would I go about doing that? (Assuming this gives me the ability to override everything else in the bsp with packages) **** ENDING LOGGING AT Sat Nov 07 03:00:32 2020