**** BEGIN LOGGING AT Mon Feb 08 03:01:12 2021 Feb 08 07:37:49 good morning Feb 08 07:50:00 yo dudX Feb 08 07:52:46 hi LetoThe2nd Feb 08 08:20:11 The following question was asked yesterday by someone and I am curious to know as well: The link http://www.airspayce.com/mikem/bcm2835/bcm2835-${PV}.tar.gz is invalid. Why is it still there? Feb 08 08:20:37 he meant the link in some recipe Feb 08 08:20:53 swissaccent12: probably because nobody bothered to send a patch for it, then? Feb 08 08:21:18 plus, "in some recipe" is not exactly a good place to go looking for it. Feb 08 08:21:20 too old I see, it is for the rpi1 if I'm not mistaken Feb 08 08:23:44 swissaccent12: there's nothing wrong with building for the rpi1 :) Feb 08 08:24:36 respect for the oldies ;D Feb 08 09:36:11 well I guess we'll never know which recipe they were talking about Feb 08 09:36:23 * LetoThe2nd doesn't feel like grepping too. Feb 08 09:38:01 zeddii: jonesv[m]: on aarch32, all device trees are in the same directory (arch/arm/boot/dts) as opposed to aarch64. So I wouldn't be surprised if there's something broken in qcom vendor kernel Feb 08 09:38:46 I might be wrong but IIRC, Yocto builds the kernel in a different directory, which is probably not tested by Qualcomm Feb 08 09:41:10 qschulz vocabulary question: "vendor kernel" means my downstream kernel, right? It's an old 3.18.31 fork Feb 08 09:41:54 jonesv[m]: "fork" in the name => vendor kernel Feb 08 09:42:03 name/sentence Feb 08 09:43:16 qschulz so do you think that I should just move the dts out of dts/qcom? I can try that... Feb 08 09:54:32 Hello, i have a question regarding the recent CVE on sudo that has been published. Is there any patch available for yocto rocko ? Thanks in advance for any info Feb 08 10:14:30 Stimfarmer: rocko is EOL for a long time already, so short answer, no. Feb 08 10:14:51 You'll have to backport it yourself or find someone who has done it already and agrees to share it with you Feb 08 10:19:34 qschulz thanks for the answer. Feb 08 10:19:34 Backport is a tough task since it requires an upgrade of the glibc for rocko. Feb 08 10:19:35 I managed to backport to zeus but not rocko. Feb 08 10:26:45 simple fix idea: stop shipping sudo! \o/ Feb 08 10:47:48 Hi, I'm using google repo for my project and I want to work with local git source code repositories instead of a remote. I need something similar to externalsrc but I want yocto to track git repo. Could someone help with my issue? Feb 08 10:49:51 tema25: you mean that you want to create a recipe that builds something, where the "something" is using repo? Feb 08 10:50:01 tema25: externalsrc is for recipes, google repo is for system setup Feb 08 10:55:15 tema25: also define "I want yocto to track git repo". What exactly do you want it to do? what's the use case? Feb 08 10:57:06 Here is my use case: I check out all source code with google repo and rebuild the image with current uncommitted changes in my app repo during development. I don't want to change recipe hash every time I do rebuild or do a release. During the release, I create pinned manifest with all source code. IDK what fetcher I need for my case. Feb 08 10:57:14 > also define "I want yocto to track git repo". What exactly do you want it to do? what's the use case? Feb 08 10:57:23 I need SRCPV from git repo Feb 08 10:59:55 tema25: i don't get it. what are you you using repo for? the complete, yocto based build? or your application? Feb 08 11:03:07 Everything is checkout with google repo including yocto layers and my source code. Feb 08 11:07:06 tema25: reimplement externalsrc with the correct SRCPV read from the git repo... though it does not make much sense as if your git tree is dirty, the SRCPV will still be the same that clean git repo, so it's not really useful Feb 08 11:07:09 tema25: sounds like you are essentially mixing up things. Feb 08 11:07:10 IMO at least Feb 08 11:08:08 tema25: you can also use devtool modify instead of externalsrc if you just want a way to modify sources quickly before committing Feb 08 11:08:21 (though, under the hood, devtool uses externalsrc) Feb 08 11:09:00 and if you want auto bump of the recipe during dev, you can use SRCREV = "${AUTOREV}" Feb 08 11:09:29 (which will take the latest version available in your git repo for each build) Feb 08 11:14:19 LetoThe2nd: damn too fast for me Feb 08 11:14:30 qschulz: BAM! Feb 08 11:15:32 and Martin too, don't remember what's their nick Feb 08 11:16:43 I want to develop and rebuild without pushing to remote. It's not convenient thought. What are the best practices to work with yocto when I do local kernel development or local app development? Feb 08 11:16:44 I need a proper way to provide developers to test their changes locally without pushing anything to the remote. Current developers do not want to modify recipes. Feb 08 11:17:23 tema25: devtool Feb 08 11:18:09 their on-going debug or dev session is finished, they push to the remote Feb 08 11:18:32 otherwise maybe extended SDK? (eSDK) never used it so can't say Feb 08 11:19:11 if your people are doing kernel dev, they probably know how to flash their device too, use tftp and all that stuff which does not require a complete Yocto rebuild and is MUCH faster than debugging with Yocto Feb 08 11:19:26 I'd say kernel dev with Yocto is not suitable Feb 08 11:19:32 too much overhead Feb 08 11:19:44 local app => devtool or esdk I think Feb 08 11:20:17 devtool looks like a hack for me. I worked with buildroot before and it's possible to use the local site. SDK is good when there is no kernel development. Feb 08 11:20:39 qschulz: exception: when kernel dev means sorting config fragments or such Feb 08 11:21:04 LetoThe2nd: that's part of Yocto for me, so debugging Yocto recipes. Feb 08 11:21:14 could someone provide tutorial how to add custom fetcher? Feb 08 11:21:25 qschulz: for me too, but hey, people might think different. Feb 08 11:21:46 tema25: 1) look at existing fetcher 2) copy 3) hack 4) profit. Feb 08 11:22:36 tema25: but i'd also say that you are probably on the wrong track, and that re-thinking the process should be step #1 Feb 08 11:23:21 LetoThe2nd: wasn't there a talk about ci/cd and dev in Yocto at ELCE 2019? Feb 08 11:23:26 by paul maybe? Feb 08 11:23:32 Then how to avoid manual work without fixing hash commits in recipes during release? Feb 08 11:23:44 qschulz: there was always a talk about ci/cd Feb 08 11:23:49 tema25: AUTOREV Feb 08 11:24:07 AUTOREV is bad practice for releases. Feb 08 11:24:15 tema25: and "release" basically means "fixing the hash" Feb 08 11:24:16 it's okay for dev but not releases. Feb 08 11:24:26 tema25: yeah i misread. Feb 08 11:24:58 devtool finish should update the rev, AFAIK Feb 08 11:25:46 Pretty sure AUH supports this somehow? might need to tag different git repos though Feb 08 11:26:06 s/different/all Feb 08 11:26:36 qschulz: no more AUHbombing please. Feb 08 11:27:17 Trying to get a systemd timer to work, and I'm stumped - anyone got experience with this? Do I need to register it in a recipe of some kind? Feb 08 11:27:19 LetoThe2nd: I used AUH a few times without sending mail bombs :) one just needs to read the doc and conf file :) Feb 08 11:28:24 tema25: the issue with using an externalsrc-like approach is how you're able to differentiate a dev build from a release build Feb 08 11:28:43 how to make sure that it is really vanilla from the package version Feb 08 11:29:29 you could do this by extending externalsrc to add something when the git repo is dirty. Feb 08 11:29:54 and it would mean versioning and version bumping would be the burden of your `repo` tool and conf file, not yocto Feb 08 11:31:02 note that if you want to set a variable in a recipe's task/function that applies to other tasks/functions (even implicit ones), the variable setting logic has to be inside the pyth onanonymous function Feb 08 11:34:42 https://pastebin.com/KnYtbT92 <-- My current attempt at a timed service - anything I'm missing? Feb 08 11:36:25 Is there a SYSTEMD_TIMER_${PN} perhaps...? So confused :( Feb 08 11:38:10 wertigon: its very well possible that those timers need some special sauce that is not there yet because nobody needed it. at least i have never noticed any specific support for them. Feb 08 12:05:18 LetoThe2nd: Ok :/ Let me see what systemd says about getting timers up and running Feb 08 12:05:51 wertigon: i would try to make it work in target first and then replicate that through recipes. Feb 08 12:06:22 hmm, apparently, systemctl enable $NAME.timer Feb 08 12:07:40 LetoThe2nd: Ok, I try that Feb 08 12:17:34 Hmm, apparently it required addendum of RequiredBy Feb 08 12:17:48 And WantedBy Feb 08 12:20:29 Oh, and question #2; if I want to build a kernel tool for a separate package, can I somehow piggyback on the kernel source code? Feb 08 12:21:34 wertigon: "piggyback"? Feb 08 12:21:56 Well, basically, I don't want to download two full versions of the kernel source code Feb 08 12:22:26 But neither do I want the package in all my images Feb 08 12:23:04 wertigon: ah. well IIRC the kernel is special in terms of its sources being available in a shared workspace, but you'd have to look for an example recipe yourself that makes use of it. Feb 08 12:24:37 can confirm, it's using a work-shared area Feb 08 12:25:03 I would take inspiration from kernel modules recipe because those need the kernel sources/headers to build Feb 08 12:36:37 qschulz: kernel modules recipe, is that poky/meta/recipes-kernel/kmod/kmod_git.bb? Feb 08 12:41:31 wertigon: most of it is probably abstracted in module.bbclass Feb 08 12:50:04 Ah :) Feb 08 13:03:22 beginner question: by default TMPDIR is commented out inside local.conf, why is the temporary folder needed anyway and what are the advantages of uncommenting this? Feb 08 13:04:34 koan85: it is commented to use the default. if you uncommenti it you have to specify a different directory Feb 08 13:06:15 aha gotcha, thanks Feb 08 13:08:37 koan85: and it is needed for about everything, because this is where all things are assembled in the end. Feb 08 13:09:53 but this directory cannot be reused am I correct (using checksums)? (per the comments in that file) Feb 08 13:10:35 koan85: TMPDIR is unique per build. SSTATE is meant for reuse Feb 08 13:11:59 clear Feb 08 13:12:05 thanks Feb 08 13:14:04 then it would be nice to be able to set a variable to automatically remove this directory once the build finishes Feb 08 13:14:55 koan85: then you wouldn't be able to seize any build artifacts, which is kind of unproductive, right? and for everything else, we have rm_work Feb 08 13:17:21 sorry meant the temporary files, because this directory holds the temporary files (unique per build so in my eyes unneccessary if I run a build again since it can't be reused) and build outputs (which is what I need) Feb 08 13:17:24 but you're right Feb 08 13:18:34 think I got enough information for now, amazing that this channel exists, thanks guys! Feb 08 13:19:52 this is about exactly what the rm_work thing does. Feb 08 13:20:32 (y) Feb 08 13:56:21 Hi All, after building a recipe here https://github.com/dorindabassey/Poky/tree/my-yocto/meta/recipes-devtools/debuginfod-config Feb 08 13:56:52 i was trying to deploy target for the recipe, so after executing runqemu to start QEMU and boot the image....i try to deploy the new recipe like this but am getting the following error Feb 08 13:57:47 https://www.irccloud.com/pastebin/24xqWwZb/ Feb 08 13:59:12 i'll appreciate any help Feb 08 14:02:35 dorinda: how did you launch qemu? Feb 08 14:04:10 127.0.0.53 sounds fishy for sure. Feb 08 14:04:45 qschulz: didn't you have something worked out for that use case? Feb 08 14:09:28 after sourcing env script i ran $runqemu qemux86-64 Feb 08 14:10:37 dorinda: and how did you obtain that ip address? usually the qemus should have 192.168.7.2 Feb 08 14:11:27 LetoThe2nd: hmm :( i also tried 192.168.100 from list of listening ports Feb 08 14:11:53 sorry 192.168.0.100 Feb 08 14:12:26 dorinda: you have to look inside the qemu, not by guessing on your host. hat does "ip a" there give you? put into a pastebin if you need somebody to look at it. Feb 08 14:16:05 oh oh i see the qemu ip there thanks, but am still getting connection refused error as above Feb 08 14:19:45 LetoThe2nd: mmmm, but by default it's in noslirp mode IIRC, which requires root privileges? Feb 08 14:20:23 dorinda: you had to give your user's password when starting runqemu right? Feb 08 14:21:07 Hi, is there a way to build a static strace, which i can move to the target if i need it? Feb 08 14:21:52 qschulz: yes password and user as root. Feb 08 14:22:36 dorinda: what's qemu's ip addr and your dev machine's netmask/ip addr? Feb 08 14:24:53 qemu ip=192.168.7.2::192.168.7.1:255.255.255.0 Feb 08 14:29:40 fwiw, I used the slirp option Feb 08 14:29:46 runqemu qemuarm64 slirp nographic qemuparams="-nic user,hostfwd=tcp::5555-:22" Feb 08 14:29:52 ssh -o "UserKnownHostsFile=/dev/null" -o "StrictHostKeyChecking=no" -p 5555 root@localhost Feb 08 14:30:17 my machine ip=127.0.0.1 192.168.7.1 192.168.0.100 netmask 255.255.255.0 Feb 08 14:30:42 dorinda: mmmm, k so it probably should just work. Can you ping? Feb 08 14:30:49 both ways Feb 08 14:31:15 well, if it works one way it should work the other way except if iptables is wrongly configured but I doubt it :) Feb 08 14:34:00 okay, let me try that now Feb 08 14:35:01 it might just be that your image does not have an ssh server running Feb 08 14:48:33 qschulz: so i ran ssh -o "UserKnownHostsFile=/dev/null" -o "StrictHostKeyChecking=no" -p 5555 root@localhost but got error: connection reset by peer Feb 08 14:54:48 and you ran runqemu with the exact command I gave you? Feb 08 14:55:55 dorinda: next try is `ps | grep ssh` (might be `ps aux`) or `ps | grep dropbear` Feb 08 14:56:18 yes i did, but with qemux86-64 Feb 08 14:57:17 ok, in qemu now, run `netstat -lauten` and check if something is listening on port 22 Feb 08 14:57:45 (does it listen to all interfaces? 0.0.0.0 or at least 192.168.7.2?) Feb 08 15:01:05 ps aux | grep ssh on my machine returned Feb 08 15:01:15 https://www.irccloud.com/pastebin/ZZigVUka/ Feb 08 15:02:14 dorinda: not on your machine, in qemu :) Feb 08 15:02:17 sorry, wasn't clear Feb 08 15:02:44 in qemu am getting Feb 08 15:02:53 https://www.irccloud.com/pastebin/PFDFbPIc/ Feb 08 15:04:16 hi guys, is there any reason why cmake files are removed in llvm recipe? Feb 08 15:04:16 https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/recipes-devtools/llvm/llvm_git.bb#n137 Feb 08 15:04:16 I'm building a recipe and fails because of: Feb 08 15:04:17 ``` Feb 08 15:04:17 | Could not find a package configuration file provided by "LLVM" with any of Feb 08 15:04:18 | the following names: Feb 08 15:04:18 | Feb 08 15:04:19 | LLVMConfig.cmake Feb 08 15:04:19 | llvm-config.cmake Feb 08 15:04:19 dorinda: just ps alone then Feb 08 15:04:20 ``` Feb 08 15:04:46 but when i tried to run telnet localhost 22 am getting telnet: can't connect to remote host (127.0.0.1): Connection refused Feb 08 15:06:09 dorinda:1) did you ping your running qemu from your machine? Feb 08 15:07:07 dorinda: 2) `ps | grep ssh` and `ps | grep dropbear` in running qemu. Is something returned by those commands? Feb 08 15:07:31 dorinda: 3) `netstat -lauten`, anything listening on port 22? Feb 08 15:09:31 dorinda: 4) if 2 and 3 returns nothing, add `EXTRA_IMAGE_FEATURES_append = " ssh-server-dropbear"` to your local.conf and try again (rebuild, restart qemu) Feb 08 15:10:06 1) yes i pinged qemu ip from my machine Feb 08 15:14:16 2) returns Feb 08 15:14:25 https://www.irccloud.com/pastebin/ybzoWadA/ Feb 08 15:18:25 3) netstat -lauten didn't return any list of active internet connections, about to try step 4) Feb 08 15:27:12 dorinda: c.f. https://docs.yoctoproject.org/dev-manual/common-tasks.html#customizing-images-using-custom-image-features-and-extra-image-features Feb 08 15:28:36 Ok, here is a sample recipe that we're trying to fit. What we want to accomplish; ensure SDK has bpf tool from kernel repository installed. Feb 08 15:28:51 https://pastebin.com/D2eiUS6r Feb 08 15:29:25 Is there an easier way? Biggest problem is that bpf requires special treatment in makefile. Feb 08 15:34:45 The above works, btw, just feels like walking upstream for water Feb 08 16:07:40 Hi, avahi-daemon cannot be started. it internally uses libdaemon and that tries to create an Exclusive lock on /run/avahi-daemon//pid and this returns with EACCES Feb 08 16:07:44 Any suggestions? Feb 08 16:10:21 GeneralStupid: Make sure you have read and write access to that file. Feb 08 16:11:29 wertigon: i have. I can create and write to a pid file as the avahi user Feb 08 16:11:32 Also, see if the file is created on a read-only filesystem Feb 08 16:11:47 wertigon: /run is a tmp file system Feb 08 16:12:06 GeneralStupid: Mounted how? Feb 08 16:12:27 Only other thing I can think of is that something is holding the file in that case. Feb 08 16:12:46 Are you running avahi as the avahi user? Feb 08 16:13:02 wertigon: tmpfs on /run type tmpfs (rw,nosuid,nodev,mode=755) Feb 08 16:13:26 wertigon: i tried both. It automatically drops root rights. But i tried with --no-drop-root Feb 08 16:13:55 Ok, good. Then avahi is probably running as something else Feb 08 16:14:02 Check for typos :) Feb 08 16:14:19 Its telling me that its switching to the (correct) rights of the avahi user Feb 08 16:14:39 How can i check if another program has a lock on that files? Feb 08 16:14:48 Probably a case of misspelled groups then Feb 08 16:15:04 see if it's incorrectly typed as avahl or something like that Feb 08 16:15:48 checked that, looks ok Feb 08 16:16:32 Hmmm... Then you probably have something already creating the lock file or the double slashes could cause problems Feb 08 16:16:47 e.g. should be /pid instead of //pid Feb 08 16:17:31 wertigon: ok, but this is hardcoder :/ Feb 08 16:26:42 wertigon: i just installed lslocks. I have no locks currently Feb 08 16:32:16 GeneralStupid: And what happens if you start the service? systemctl start avahi Feb 08 16:34:54 wertigon: https://dpaste.org/xVMG Feb 08 16:35:00 my docker volume woes seem to continue unabated, this is a real weird one right? https://gist.github.com/jordemort/fa9330f2369788d57d234b886fe42a5c Feb 08 16:35:01 Exception: subprocess.CalledProcessError: Command '['rm', '-rf', '/work/build/tmp/work/raspberrypi4_64-bantha-linux/bantha-image/1.0-r0/oe-rootfs-repo/rpm']' died with . Feb 08 16:35:03 rm gets SIGABRT? not even a permission denied Feb 08 16:35:27 It could be that the service boots before filesystem is up and running Feb 08 16:35:57 jordemort: https://wiki.yoctoproject.org/wiki/Pseudo_Abort Feb 08 16:36:34 RP: yeah i've seen that wiki page a lot lately and i have no idea what's happening still Feb 08 16:36:57 i'm runing crops/poky, i've got a volume mounted on /work/build, nothing else touches that volume Feb 08 16:37:04 wertigon: but if this would be the case, shouldnt it possible to start it now? Feb 08 16:37:23 i'll have a successful build and then the very next build will fail with weird filesystem permission stuff Feb 08 16:37:29 GeneralStupid: I have no idea then :/ It's probably something incredibly obvious but I have too little experience here to be much use, sorry Feb 08 16:37:56 GeneralStupid: That was my thinking too Feb 08 16:38:00 jordemort: its a sign that pseudo file tracking isn't working properly. Have you looked at what pseudo.log says? Feb 08 16:38:03 wertigon: Ok, thanks a lot for your help. But this is really frustrating Feb 08 16:38:21 jordemort: is this master or gatesgath/dunfell? Feb 08 16:38:30 RP: dunfell Feb 08 16:38:37 getting pseudo.log now Feb 08 16:38:43 GeneralStupid: Yeah, tell me about it, I have a similar problem right now with a systemd service on dunfell... Feb 08 16:39:49 Got the timer to work finally, but not the service connected to it XD fun times tomorrow! Feb 08 16:40:15 Time to walk off work, might check in tomorrow too! GLHF! Feb 08 16:40:39 RP: pseudo.log says a LOT, mostly "inode mismatch" Feb 08 16:41:13 let me upload it Feb 08 16:45:37 RP: pseudo.log https://gist.github.com/jordemort/94a860bb79ff924743e6616500bdcc6f Feb 08 16:46:32 i bet this is because i'm using a named volume instead of a direct host path as a volume mount and docker is doing something stupid between container runs to trash it Feb 08 16:47:40 jordemort: the inode numbers don't look stable so pseudo is getting rather upset Feb 08 16:49:02 jordemort: pseudo won't work properly if inode numbering isn't stable Feb 08 16:49:33 we've seen issues w/ docker, but never been able to track it down Feb 08 16:49:39 RP: makes sense Feb 08 16:51:08 jordemort: basically that log is saying "file has this inode number but now it has that inode number, what happened?" Feb 08 16:51:30 * RP suspects docker happened Feb 08 16:51:53 /khanify docker Feb 08 16:54:27 how bad would it be to just not persist that between runs and only persist downloads and sstate? Feb 08 16:55:34 jordemort: not too bad IMO Feb 08 16:57:28 jordemort: There are a few caveats when building in a container Feb 08 16:59:11 jordemort: The one that really bit us WRT to pseudo is that the kernel will SIGKILL all child processes when the container exits, which doesn't allow pseudo to dump the in memory database: See the "cleanup.py" question in https://github.com/garmin/pyrex#faq Feb 08 16:59:57 But, FWIW we also don't use volumes and instead bind-mount in the working directory, so it could be a problem with volumes also Feb 08 17:00:35 JPEW: thank you, i think you just handed me the key to fixing it <3 Feb 08 17:09:14 i wonder how hard it'd be to replace pseudo with uidmap Feb 08 17:09:17 probably pretty fscking hard Feb 08 17:09:46 pseudo is a lot more then just uid mapping Feb 08 17:55:06 Has anyone seen the use of the AC_PROG_LIBTOOL macro in a configure.ac break a package? With it included, I get linking errors during do_compile (missing symbols). Without the macro, the package is fine. I've verified it's passing --with-libtool-sysroot pointed to recipe-sysroot, and the resulting linker flags appear to be the same in both cases. Feb 08 17:57:20 I'm trying to get QEMU to provide an alsa device for testing audio. I added alsa to the packageconfig of qemu-system-native , but that did not help. any idea? Feb 08 18:42:44 I'm a bit confused. I've got recipe, and a append file, which I would like to apply. The 5.1.5. "Using .bbapend Files" should clarify how to apply them, but even after reading the paragraph I'm not completly sure. Feb 08 18:42:55 How exactly do I apply the bbappend? Feb 08 18:48:54 a bbappend file in an included layer are automatically applied Feb 08 18:50:59 fray: when I run bitbake-layers show-appends, it says (skipped) Feb 08 18:52:03 malinus: does "bitbake " work? or is it skipped there too? Feb 08 18:52:18 that might explain why it is being skipped Feb 08 18:52:30 Does it say the package is skipped or the append is skipped? Appends are applied based on the configuration in the layer itself. (i.e. https://git.openembedded.org/meta-openembedded/tree/meta-oe/conf/layer.conf#n15 ) Feb 08 18:53:09 Assuming the item is available in the layer and picked up my the BBFILES.. then it's likely the recipe is being skipped for another reason (incompatible, etc) Feb 08 18:53:17 ohhh the layer is skipped Feb 08 18:53:31 thanks, that explain is, doh Feb 08 19:10:17 I don't get it. "runqemu qemux86-64 audio kvm" produces something that shows me alsa devices in the VM. and yet, I hear nothing, and pavucontrol on the host shows that QEMU is not actually opening any PCM output. Feb 08 19:11:14 is audio in qemu broken in gatesgarth? Feb 08 19:24:39 Hello, I have been trying for a few days now to build yocto with x11 support with core-image-x11 image, but having difficulties with atk library. I have installed both gobject-introspection and  libgirepository1.0-dev (host system is Ubuntu 18.04), but I still get configure: error: gobject-introspection-1.0 is not installed. Can someone please Feb 08 19:24:40 help? Thanks. Feb 08 21:24:35 arg, this prefix map issue makes no freaking sense. driving me nuts Feb 08 21:50:47 Hello, anyone got this error?| pseudo_client.h:17:65: error: ‘_STAT_VER’ undeclared (first use in this function) Feb 08 21:51:15 ERROR: Task (virtual:native:/home/miha/workspace/poky/meta/recipes-devtools/pseudo/pseudo_git.bb:do_compile) failed with exit code '1' Feb 08 21:51:50 I saw a reference recently about that, but I didn't catch what host OS they were using (gatesgarth ws the YP version though) Feb 08 21:51:52 getting it on dunfell branch Feb 08 21:52:22 i am using arch linux Feb 08 21:52:24 which host oS? Feb 08 21:52:34 arch Feb 08 21:52:44 Hmm.. not sure, I don't know much about arch sorry.. Feb 08 21:53:05 the error BTW is that it's using a variable (likely from glibc) that isn't there anymore.. Feb 08 21:53:13 most likely arch being arch and breaking things first. did they get a new glibc recently? Feb 08 21:53:19 Your host glibc version (2.33) is newer than that in uninative (2.32). Disabling uninative so that sstate is not corrupted. Feb 08 21:53:22 :) yes Feb 08 21:54:05 we recommend that people using arch either 1) do builds in a container of a known good distro or 2) send patches Feb 08 21:55:46 hmm, i wonder if devtool/recipetool/etc should be obeying the bitbake log config the way the knotty ui does, or keep doing their own thing Feb 08 21:57:48 hm, another kind of pseudo failure: Feb 08 21:57:49 path mismatch [1 link]: ino 16777305 db '/home/builder/.gnupg/private-keys-v1.d' req '/home/builder/.passphrase'. Feb 08 21:58:04 that's understandable because i recreate those files every container run Feb 08 21:58:16 but i don't really want pseudo tracking them, do i need to figure out how to exclude them? Feb 08 21:58:36 or do i need to find a better way to inject gpg keys? Feb 08 21:58:38 Add /home/builder to your exclusion path Feb 08 21:59:18 fray: that is PSEUDO_IGNORE_PATHS? local.conf a fine place for it? Feb 08 21:59:27 Not sure.. sorry Feb 08 21:59:41 (but it should be that, you'll likely have to _append) Feb 08 22:04:11 I need myrecipe-native to be deployed before WIC creation. Now I am using EXTRA_IMAGEDEPENDS += "myrecipe-native", but it doesn't work. How can I deploy some recipe before WIC? Feb 08 22:05:45 * rburton wonders why pseudo has opt-out path tracking not opt-in Feb 08 22:06:05 we could tell it to opt in to $B $D and a few more, and be done, surely Feb 08 22:06:27 because we don't know the paths that can affect teh build.. but we do know paths that don't.. Feb 08 22:06:32 Seebs would be able to expand on it Feb 08 22:06:42 eh it might not be bad to be overly paranoid about sneaky contamination from the host Feb 08 22:06:55 i'd appreciate this failure, if it was something i wasn't expecting to change :D Feb 08 22:07:30 Huh, that's a fascinating thought. I suppose the question would be, in part, detecting contamination. It might be interesting to scan the database for paths that aren't in $B/$D/etc Feb 08 22:23:43 Hmmm. Not sure that went through. Let me try that again -  I've been using an older branch of Poky to build a genericx86-64 image, and I updated the code to using gatesgarth. I can't figure out how to get the install option to show up in Grub. So I can't get anything other than the live system option (booting off of a USB drive.) Any idea how to Feb 08 22:23:43 get that back in gatesgarth? Feb 08 22:29:45 fray: more cautious than optimistic but `PSEUDO_IGNORE_PATHS_append = "/home/builder"` in local.conf seems to be working out, thanks Feb 08 22:33:04 I think it's in path format so..:/home/builder (I think), but I'm not sure.. Feb 08 22:33:09 look at what the default value is first Feb 08 22:37:42 ah looks like it's comma-separated Feb 08 22:47:27 how can I make my image depend on some native task DEPLOY Feb 08 22:48:29 kergoth: they do things a bit differently :/ Feb 08 22:49:07 seebs, jordemort: when sorting out PSEUDO_IGNORE_PATHS, I did scan the databases for exactly that, files which shouldn't be there... Feb 08 22:56:00 jonmason: we're going to get warnings in meta-arm from http://git.yoctoproject.org/cgit.cgi/poky/commit/?h=master-next&id=737fabf68606a3293780b63f96cfeb142e7901c5 Feb 08 23:00:57 hi. i made a `standard` sdk but there is no make program in it. why is that? Feb 08 23:01:09 how can I force to do_deploy of some recipe before image.wic build? Feb 08 23:01:32 mischief: its assumed make from your system would be ok. You could always add nativesdk-make to the sdk if you wanted it... Feb 08 23:01:58 RP: aha. Feb 08 23:02:16 well, it added nativesdk cmake and meson so i thought the lack of make was curious. Feb 08 23:02:38 i installed the sdk inside a ubuntu:18.04 container, which of course lacks anything terribly useful by default Feb 08 23:03:50 mischief: we patch meson and cmake to work better for cross use iirc, make should be standard Feb 08 23:17:22 RP: if you knew how much we have to drag our guys kicking and screaming to upgrade their u-boot and kernel sources... Feb 08 23:29:33 how can I restart image creation? https://pastebin.com/2RxdN8i1 Feb 08 23:41:25 hi qschulz back again :) so i added the `EXTRA_IMAGE_FEATURES_append = " ssh-server-dropbear"` to local.conf but got some errors, which i think was probably because some ssh-server-dropbear dependencies were missing, so instead i added `EXTRA_IMAGE_FEATURES_append = " ssh-server-openssh"`, built the image again and ran QEMU, this time `netstat -lauten` returned LISTENING port 22. And you were right about my image not Feb 08 23:41:26 having an ssh server running, had to set it up and now am able to deploy target for the recipe. Thank you :) Feb 08 23:55:15 hm.. when i run ./poky-glibc-x86_64-core-image-minimal-armv7vet2hf-neon-qemuarm-toolchain-ext-3.1.5.sh it seems to hang at `Initialising tasks`. can't tell if it's doing something or just stalled forever Feb 09 00:10:57 jonmason: I have some idea :( Feb 09 00:28:44 ive tried with the sdk on two different hosts (debian, ubuntu, and container ubuntu) but they all just hang. is anyone else having this problem? Feb 09 02:01:22 aha.. i let it sit for a while. WARNING: Error contacting Hash Equivalence Server typhoon.yocto.io:8686: [Errno 110] Connection timed out **** ENDING LOGGING AT Tue Feb 09 03:00:29 2021