**** BEGIN LOGGING AT Tue Nov 24 02:59:57 2020 Nov 24 03:59:13 hi all, any obvious reason why in a base-files_%.bbappend a do_install_append_my_machine does not get picked up? Nov 24 03:59:48 a do_install_append does Nov 24 04:00:07 and in other recipes do_install_append_my_machine works fine? **** BEGIN LOGGING AT Tue Nov 24 07:44:48 2020 **** BEGIN LOGGING AT Tue Nov 24 07:47:40 2020 Nov 24 07:56:55 yo dudX Nov 24 08:37:44 hi everyone, could I fix the colors? https://imgur.com/hYmmNvh.png minicom has -c arg enabled so I guess I need a different PAGER or something like that 🤔 Nov 24 09:44:26 Hello everybody, I need to change the device offset in `/etc/fw_env.config` for u-boot. Is there an environment veriable to do that or to I need to modify the actual file while bitbakeing? Nov 24 09:46:26 emrius: usually you modify/patch uboot Nov 24 09:46:47 LetoThe2nd: Thanks! Ok, will do Nov 24 09:49:05 emrius: wondering if it's not in libubootenv now though Nov 24 09:49:19 check with oe-pkgdata-util find-path /etc/fw_env.config Nov 24 09:50:32 qschulz: good point, its always worth checking. Nov 24 09:51:19 qschulz: Oh, nice!! that comes handy Nov 24 09:53:15 qschulz: I didn't know about that tool! That's a killer feature! Nov 24 09:57:38 now, somebody mention oe-pkgdata-browser..... Nov 24 09:58:23 LetoThe2nd: do you have a session online about these tools? Nov 24 09:59:02 nope, not explicitly Nov 24 10:00:19 Anyone knows from which package I should be able to get libqtquick2plugin.so? I've been told it's part of qtdeclarative but couldn't find it in any of the packages built by that recipe. Note... still on qt5.6.4 (krogoth branch) and cannot upgrade (GPLv3...) Nov 24 10:01:29 JaMa: or khem maybe? ^ Nov 24 10:05:21 could you give me a hint of how to deal with systemd colors? because I'm getting the color codes in minicom 😞 Nov 24 10:05:31 and apparently colors cannot be disable in systemctl output Nov 24 10:07:27 wyre: I use picocom and use sysv so can't help :/ Nov 24 10:19:12 qschulz, by default warrior have been built with systemd, which is strange because pyro haven't been Nov 24 10:19:44 I think this is also a build env var in the building process 🤔 Nov 24 10:21:08 anybody know if bitbake servers are undergoing any maintenance? linux-yocto-tiny-5.8.18 has been stuck for a few hours Nov 24 10:21:44 s/have/has/ Nov 24 10:22:26 amitk_: bitbake servers? linux-yocto-tiny? Nov 24 10:28:13 amitk_: fetch, presumably? control-c and try again? Nov 24 10:30:56 LetoThe2nd: core-image-minimal Nov 24 10:31:49 rburton: tried Ctrl-C and restart a couple of time on two different machines and it gets stuck on the same package Nov 24 10:32:06 amitk_: it would be really helpful if you tried to formulate a meaningful sentence that we can understand... :/ Nov 24 10:32:09 but what is it *doing* Nov 24 10:32:27 * LetoThe2nd just built core-image-minimal with cold download caches just fine a couple of hours ago Nov 24 10:36:43 rburton: having a beer and listening to metal. Nov 24 10:37:24 halstead: git.yoctoproject.org isn't responding Nov 24 10:38:45 rburton: LetoThe2nd: Apologies. :-/ Let's try again. I tried this: https://pastebin.com/bKnr1r9k Nov 24 10:39:14 rburton, Checking into it now. Nov 24 10:39:20 amitk_: the git server is down, so its probably trying to hit the mirrors at that point Nov 24 10:39:35 and I'm stuck at bitbake trying to fetch linux-yocto-tiny-5.8.18.... for a few hours now Nov 24 10:39:47 thanks rburton Nov 24 10:39:55 ah, the *fetch* was the missing keyword. Nov 24 10:40:37 amitk_: hint: the term bitbake-server actually refers to a process in th build, so if that were under some form of maintenance it would both strange and caused by yourself :) Nov 24 10:41:53 LetoThe2nd: thanks for the hint. I should've said git server Nov 24 10:42:01 halstead: hello! git.yoctoproject.org server is down if you didn't already know about it :) Nov 24 10:42:15 amitk_: :) Nov 24 10:44:28 Thanks qschulz. I just started looking into the problem a moment ago. Nov 24 10:49:07 halstead: I also just noticed, my autobuilder keeled over :/ Nov 24 10:50:53 RP, Unfortunate. I'm forcing a reboot of the VM. I do have a read only mirror ready I can switch to if this doesn't come back up shortly. Nov 24 10:51:08 RP: https://youtu.be/ta-Z_psXODw Nov 24 10:52:46 rburton, qschulz, RP, git services are functional again. Nov 24 10:53:19 amitk: test again now :) Nov 24 10:53:23 halstead: thx :) Nov 24 10:54:00 thanks for the quick response, qschulz, LetoThe2nd, rburton. Seems to work now. Nov 24 10:54:41 \o/ Nov 24 10:55:26 halstead: thanks! Nov 24 10:56:47 Happy to get it back up RP. I'm not sure about the cause yet. I see repack processes causing a lot of load. Nov 24 10:57:07 This server will be replaced in the near future. Nov 24 10:57:29 halstead: I was just thinking about the changes planned and wondering if this was related or not Nov 24 10:57:32 qschulz, and what about colors in therminal? Nov 24 10:57:59 halstead: I'm sorry you're up at this time but having it back is extremely helpful! Nov 24 10:59:24 RP, The mirroring is incremental at this point and shouldn't cause trouble. Appears the lots of requests at the same time as running git repack on linux-yocto overloaded this VM. Nov 24 11:03:46 halstead: go to bed now though! Nov 24 11:04:06 * RP seconds rburton Nov 24 11:04:20 we need that 'remote power switch' thing we discussed earlier Nov 24 11:04:49 rburton: we are transitioning some of this to other people at the LF FWIW Nov 24 11:05:00 ah Nov 24 11:05:25 can I override variables set in a .class in a different layer? Nov 24 11:05:37 without completely overwriting the bbclass? Nov 24 11:05:47 are they global variables? Nov 24 11:06:26 cengiz_io: in short, yes, usually Nov 24 11:06:29 Okay rburton & RP. I've looked at the logs and I don't see anything alarming at the moment. Nov 24 11:06:42 * halstead heads back to bed. Nov 24 11:06:43 cengiz_io: depends whether the class is globally inherited or recipe specific Nov 24 11:06:50 halstead: g'night Nov 24 11:06:54 halstead: good plan, sleep well Nov 24 11:07:13 RP it comes from AGL, I inherit it indirectly Nov 24 11:07:42 and the variables inside have ??= so I guess if my layer's priority is high, these can be overridden by me Nov 24 11:07:58 halstead: night night Nov 24 11:09:56 cengiz_io: a distro or recipe can happily override variables in classes if they use ?= or ??= Nov 24 11:10:42 where shoud I put those vars? local.conf? Nov 24 11:14:23 qschulz, when I use --no-pager in systemd color codes are processed so I guess it has something to do with less (which is the pager that's being used) Nov 24 11:14:57 cengiz_io: do you want to modify the variable for all recipes inheriting this class or not? Nov 24 11:33:50 qschulz yep. because the variables are used in many recipes which I don't have control of Nov 24 11:34:12 https://gerrit.automotivelinux.org/gerrit/gitweb?p=AGL/meta-agl.git;a=blob;f=meta-agl-profile-graphical/classes/agl-graphical.bbclass;h=65106db181c634bdf47adcbe7d46215c390f6a15;hb=HEAD to be exact Nov 24 11:34:51 qschulz (tty7 is hardcoded but my board has display on tty1) Nov 24 11:36:57 I've put them into my local.conf. let's see if it propagates Nov 24 11:42:59 how can I check the dependencies for the BSP layer? Nov 24 11:48:15 Hi, I saw a mechanism in yocto that custom kernel recipes provide configuration changes in ".cfg" files, which are used automatically to create a composed "defconfig" for the kernel to be build. I figured out this behaviour seems to come from the yocto-kernel.bbclass in the poky/meta layer. The function defined there doesn't seem to be used elsewhere. Do you know of any similar mechanism for other recipes, to create a combined configuratio Nov 24 11:48:45 I could write something similar like this myself for sure, but I'm quite sure something like this does exist somewhere in Yocto.. Nov 24 11:50:16 dleppich: just inherit yocto-kernel? Nov 24 11:50:31 wyre: you have to be a bit more explicit Nov 24 11:50:54 wyre: easiest way is to have only openembedded-core and you BSP layer and compile until you don't have an error anymore Nov 24 11:51:39 cengiz_io: you shouldn't put stuff in your local.conf :) Nov 24 11:52:18 qschulz what is the canonical way? Nov 24 11:52:59 considering how small and easy to maintain this bbclass is, just override the bbclass with yours Nov 24 11:52:59 otherwise, if you really only have one machine Nov 24 11:52:59 what if the class gets updated on upstream? Nov 24 11:52:59 I would put it into its conf file Nov 24 11:52:59 cengiz_io: that's the maintenance burden Nov 24 11:52:59 machine conf Nov 24 11:52:59 I would like to use a similar mechanism for barebox. I want to have a .cfg with some settings changed and add this file to my SRC_URI. I want this file to be processed to create a config file with my changes applied, before building barebox Nov 24 11:53:02 which you anyway have because you have to override a variable as a whole Nov 24 11:53:37 qschulz, I need also a distribution layer, don't I? Nov 24 11:53:39 cengiz_io: I'm pretty sure the conf files are parsed before the classes so you can just use the same loose assignments Nov 24 11:53:59 ok thanks. I shouldn't use = then Nov 24 11:54:02 wyre: poky can give you one, otherwise probably nodistro or maybe your vendor is giving you one already Nov 24 11:54:10 So I want to use this config split stuff, which is done in yocto-kernel, in other use cases, which are not kernel-related at all. Nov 24 11:54:12 cengiz_io: you can but I think it's unnecessary Nov 24 11:54:39 dleppich: ah... then I don't think it's supported in a class because it's probably specific to the sw you want to build Nov 24 11:54:51 the merging of config fragments is done by scripts in the kernel Nov 24 11:55:11 and I doubt the scripts are the same for other pieces of SW Nov 24 11:55:20 qschulz: okay, I though a generic version of this merging of config fragments does exist somewhere in yocto maybe :) Nov 24 11:56:07 dleppich: maybe but I doubt it, since it cannot be generic as it depends entirely on the merging script used by the SW which to my knowledge isn't generic nor a separate project Nov 24 11:56:22 Because I saw this kind of configuration multiple times already, in different software Nov 24 11:56:39 if you manage to make it generic, by all means, send a patch to Yocto :) Nov 24 11:57:13 qschulz what about my layer's priority? is 20 a good value? Nov 24 11:57:31 dleppich: I think you have this behavior in busybox too Nov 24 11:58:00 cengiz_io: I mean... It all depends what you need this priority for? Nov 24 11:58:08 the priority isn't used for conf files or bbclasses Nov 24 11:58:08 ok Nov 24 11:58:08 cengiz_io: BBPATH is what's used for those files' "priority" Nov 24 11:58:08 qschulz let's say I need to override a file of a recipe from another layer Nov 24 11:58:27 thus I create the same path, and put my file there Nov 24 11:58:51 (specifically a splash screen image) Nov 24 11:59:13 while do_unpack, doesn't it take it by the priority? Nov 24 12:07:20 qschulz, https://imgur.com/0bUsAwk.png why is not finding the machine? Nov 24 12:07:31 qschulz: Thanks. I had a look into what busybox does and indeed there are minor differences. Nov 24 12:08:25 When I first thought about this kind of task, my inner code duplication alarm was triggered. But I guess it is fine in this case Nov 24 12:11:16 this makes no sense, the layer is included in bblayers.conf and micreogea.conf is inside of meta-engicam/conf/machine folder Nov 24 12:11:20 why is not fetching it? Nov 24 12:11:35 wyre: What's in layer.conf for that layer? Nov 24 12:11:54 Also please use a pastebin rather than sharing an image, we can't search in the text in an image Nov 24 12:12:22 paulbarker, http://ix.io/2FiB Nov 24 12:15:11 wyre: If you set MACHINE back to let's say `qemux86`, can you start a build? Nov 24 12:15:36 You can Ctrl+C once it's started and running tasks and share the header printed by bitbake which should say what layers are included Nov 24 12:16:53 cengiz_io: priority is not in play, you need to create a bbappend for it and prepend to FILESEXTRAPATHS Nov 24 12:17:34 wyre: Also check the environment variables set in your shell. I've seen silly issues like this when I've moved from one build directory to another and forgot to init the build environment again Nov 24 12:17:37 is there anywhere with instructions on how to get grub working without efi? finally come to the conclusion that syslinux is crap Nov 24 12:19:53 paulbarker, I've initialized the build environment several times Nov 24 12:20:05 qschulz ok thanks. Nov 24 12:20:12 I'm going to try with quemux86 Nov 24 12:20:33 qschulz btw is there a log that shows all included files of an image? `require` directives? Nov 24 12:20:53 qschulz I think my `require` doesn't do its job recursively. Nov 24 12:20:54 maybe. Nov 24 12:24:24 where is a short documentation on bb.utils.contains? Nov 24 12:49:52 LetoThe2nd: http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/bitbake/lib/bb/utils.py#n970 Nov 24 12:50:25 That's the only docs I know of for that but I may be missing somewhere else, "contains" is hard to grep for Nov 24 12:50:38 paulbarker: yeah thats the obvious one. nothing more "user friendly", i guess? Nov 24 12:50:49 paulbarker, it seems to be something in the build folder, I've created another one and it's building 🤔 Nov 24 12:53:13 cengiz_io: do you override a file defined in SRC_URI or a file that is require/include in the recuipe? Nov 24 12:53:56 qschulz to be clear, this is a completely different question. "btw is there a log that shows all included files of an image? `require` directives?" Nov 24 12:54:23 qschulz overriding a splash logo is done, thanks to your help Nov 24 12:54:24 @LetoThe2nd: http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/bitbake/lib/bb/utils.py#n970 Oida! Nov 24 12:54:38 RP: If I want to check the test matrix should I be looking at the `yocto-autobuilder-helper` repo? Or is it documented somewhere else Nov 24 12:54:47 paul was faster ;) Nov 24 12:55:25 cengiz_io: what exactly is your issue or what do you want to check? Nov 24 12:56:24 RobertBerger: OIDA! Nov 24 12:57:35 RP: I've also noticed that `opensuseleap-15.1` is listed in SANITY_TESTED_DISTROS but not 15.2 which was released in July Nov 24 12:59:25 paulbarker: config.json in -helper Nov 24 12:59:48 paulbarker: we're forgotten to update. Master has fedora33 now too Nov 24 13:04:04 qschulz, what's exactly the distro for? I mean, ... I'd like to build a minimal system, I'm not sure what I need this fsl-bsp-release Nov 24 13:04:18 could I build the system with nodistro and don't use this layer? Nov 24 13:08:57 qschulz nevermind. I'd better ask it on #automotive Nov 24 13:10:26 RP: I see there's an opensuse 15.2 worker listed on https://autobuilder.yoctoproject.org/typhoon/#/workers Nov 24 13:11:13 The fedora 33 worker looks to be paused Nov 24 13:11:31 I'm happy to send a patch to update poky.conf if those are both now tested distros on the autobuilder Nov 24 13:15:33 paulbarker: f33 was paused as we've had issues with it but those issues should be resolved now, I will unpause when the changes in maser-next merge Nov 24 13:21:44 RP: Ok, I can send a patch to update SANITY_TESTED_DISTROS for master Nov 24 13:22:57 paulbarker: sounds good thanks Nov 24 13:24:14 wyre: which layer are you talking about? Nov 24 13:24:42 wyre: also, please have a look at the Youtube tutorials available on the YoctoProject youtube channel, there are some you'll probably find interest in :) Nov 24 13:27:26 qschulz, ty, I'm reading the "Embedded Linux System with the Yocto Project" but I'll take into account your suggestion 😄 Nov 24 14:46:59 question to the general audience, hope that someone could lead me into a right direction: my image installs a single package from linux-firmware (namely: linux-firmware-imx-sdma-imx7d), it is pulled via MACHINE_EXTRA_RRECOMMENDS. It all goes well, and rootfs does contain only that firmware requested. When however, I do set 'dev-pkgs' in IMAGE_FEATURES - I get *all* firmware files installed into rootfs, even though they were not requested. Nov 24 14:47:43 Looking at the linux-firmware recipe, I saw that linux-firmware-dev package is indeed present because PACKAGES are set with =+ Nov 24 14:49:26 if i try to "remove" -dev by explicitly saying PACKAGES = "${PN} " and have 'dev-pkgs' still enabled - I do not get any firmware package installed from linux-firmware (linux-firmware-imx-sdma-imx7d is totally missing). Nov 24 14:51:05 it looks to me as once 'dev-pkgs' is there - you get *all* firmware files as if you request to install linux-firmware "meta" package, and removing -dev gives nothing at all. Nov 24 14:53:56 ah, the question: does it makes sense to have -dev for firmware-only package, and why if 'dev-pkgs' is set and -dev is not omitted - main package (or package from split) is not getting installed? Nov 24 14:54:27 s/is not omitted/is omitted/ Nov 24 14:58:15 zandrey: ${PN}-dev RDEPENDS on ${PN} by default Nov 24 14:58:30 linux-firmware actually RRECOMMENDS **all** firmware packages Nov 24 14:58:38 http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-kernel/linux-firmware/linux-firmware_20201022.bb?h=master#n988 Nov 24 14:58:42 so, it's correct Nov 24 14:59:24 one way would be to either remove the RDEPENDS from -dev to normal package (but meh...) Nov 24 15:00:48 anyone else seeing: "libgomp: Thread creation failed: Resource temporarily unavailable"? Nov 24 15:04:27 i can work around this problem if i set the "threads" and "parallel" explicitly Nov 24 15:16:30 qschulz: thanks! this chain i've traced down already. but that does not explain why if i say i want just one specific package from split, and "meta" package does not have associated -dev - split package is not getting installed. Nov 24 15:16:45 does it relates to RRECOMMENDS in this case? Nov 24 15:17:40 meaning: no ${PN}-dev means - no ${PN}, so no ${PN}- Nov 24 15:35:37 yes to your last sentence Nov 24 15:36:03 oh... just re-read and forgot to finish the different things you could do :p Nov 24 15:36:42 you could also add NO_RECOMMENDATIONS = "1" to a conf file or your image recipe but that also mneans absolutely no recommendations Nov 24 15:39:37 qschulz: got it. thanks for your advise on the options i have at hands. Nov 24 15:43:20 i'd have a look at how to ease those requirements though, since i still do not get why if -dev is not there (omitted or absent) - ${PN} is not getting installed if it is listed in recommendations. this might fire back at some allarch packages, since there are examples there where PACKAGES are defined with =, effectively excluding -dev from being created. If then 'dev-pkgs' are defined by the user - some allarch would disappear from the rootfs, leaving user Nov 24 15:43:20 to figure out how this had happened. Nov 24 15:43:37 got 2 go now, thanks for help though! (y) Nov 24 15:44:42 time to find a drink and get comfy, we'll be live in 15 minutes at: https://www.twitch.tv/theyoctojester Nov 24 16:12:36 zeddii: https://github.com/moto-timo/meta-virtualization/tree/timo/k3s-wip including 1.19.4+k3s1 and k9s 0.24.1 Nov 24 16:13:00 zeddii: I think the root cause of CrashLoopBackOff is in containerd, but still troubleshooting Nov 24 16:36:46 LetoThe2nd: missing s in bb.util Nov 24 16:36:59 I have more commits to push to the wip branch eventually, but didn't get much done yesterday. Nov 24 16:37:16 moto-timo, perphaps, but I ran containerd images manually, so it is at least smoke-test sane Nov 24 16:45:26 LetoThe2nd: I also like oe-pkgdata-util lookup-recipe because sometimes the packagename is not derived from the recipe name Nov 24 16:45:51 fray: no it's pretty old, we have it in thud Nov 24 16:45:59 maybe even in krogoth Nov 24 16:46:17 I knew there was a library called that -- but I didn't remember this interface.. Nov 24 16:47:40 ya, it must be browser I'm thinking of.. Nov 24 16:47:55 fray: didn't have the chance to play with this yet :p Nov 24 16:48:21 ya, I havn't either.. saw it in the Dunfell or Gatesgarth release notes.. Nov 24 16:48:28 LetoThe2nd: joined mid stream :( completely forgot about it Nov 24 16:48:53 qschulz: no problem, glad you've been around - and helpful as ever! Nov 24 16:51:08 LetoThe2nd: Looking forward to watching the recording :) Nov 24 16:51:34 LetoThe2nd: good idea for multiconfig, it's used more and more it seems (or at least more and more people have questions about it) Nov 24 17:05:07 zeddii: I'm still also seeing a lot of thrash creating the veth devices Nov 24 17:05:45 zeddii: I'm building an image without k3s and am going to attempt to install using their scripts... although perhaps that is just madness Nov 24 17:05:53 already did that. Nov 24 17:05:56 same result for me. Nov 24 17:06:26 core-image-full-cmdline + I think a couple of extra things, had enough to run their scripts. Nov 24 17:06:33 it was either curl or wget I added, IIRC. Nov 24 17:07:09 zeddii: ok... so much for that idea Nov 24 17:07:31 might work differently for you. I though the same thing, rule out our image and utilities. Nov 24 17:07:50 zeddii: maybe go back to 1.18.x ? was 1.19.x the beginning of the troubles? Nov 24 17:08:06 1.18 didn't work either, but for different reasons. Nov 24 17:08:25 authentications stuff, but I don't remember the crash backoff Nov 24 17:08:30 zeddii: PITA because I can't get any logs, the symlink to /var/volatile is invalid Nov 24 17:08:37 yup! Nov 24 17:09:01 which is what I was saying on the mailing list thread. We need to fixup the logs, but I didn't get the right image features yet, or am somehow messing it up. Nov 24 17:09:02 and I don't exactly know what the 'journalctl -u k3s' is telling me Nov 24 17:09:10 yeah Nov 24 17:09:11 yah, kinda garbage in that. Nov 24 17:09:28 I asked for the image definition that the so-called 'working results' are from :D Nov 24 17:09:34 and google search is useless... the symptoms point to stupid things like selinux (not) Nov 24 17:09:50 and didn't get anything, but it just shows that I need to persist on this and get my packagegroups merged. Nov 24 17:10:04 so that these are not just buildable, but still garbage dumps. Nov 24 17:10:11 at least I am doing the happy dance to have k9s building and working Nov 24 17:10:45 +1 Nov 24 17:10:51 saves a lot of 'kubectl describe pod foo -n kube-system' typing Nov 24 17:11:04 hah. I'm in the thousands on that! Nov 24 17:11:18 and on my builder that takes a while to come back. Nov 24 17:13:39 I was debating about opening an issue on k3s github, but then I go in circles trying to figure out what logs they might want to see Nov 24 17:14:04 and then have them say "that's a containerd thing" or "that's a cni thing" Nov 24 17:14:15 in my experience, they just ask a couple of questions, hear that it isn't a known distro and stop answering it, or close it. Nov 24 17:14:28 which is probably sane on their part Nov 24 17:14:41 yah. I can't throw any stones on that front. I'd do the same. Nov 24 17:14:54 I love the bots that come around and close old issues because they are old. Nov 24 17:14:58 We need a bot to do that! Nov 24 17:15:10 (that would get rid of half my bugzilla's) Nov 24 17:15:21 lol... instead of the bug triage "bump it" every two years? Nov 24 17:21:23 moto-timo: be nice if it was in the recipes... k9s Nov 24 17:23:10 OutBackDingo: I just need to sort out why protobuf keeps not being able to be removed... file permissions issue Nov 24 17:23:26 OutBackDingo: I will absolutely be contributing it to meta-virt Nov 24 17:24:04 ERROR: Command '['rm', '-rf', '/srv/build/ttorling/build-ypdd-virt/build/tmp-k3s-host/work/core2-64-poky-linux/k9s/0.24.1+gitAUTOINC+68981ff500-r0/build/pkg']' returned non-zero exit status 1. Nov 24 17:30:10 qschulz: its in qtdeclarative-qmlplugins Nov 24 17:30:42 Guest25560: thx, it turns out you need opengl enabled to have qtdelcarative compile it Nov 24 17:30:57 Guest25560: and we don't want to enable it, so we just got rid of this dependency :) Nov 24 17:31:01 thats right Nov 24 17:31:17 are you using eglfs ? Nov 24 17:46:01 khem: I don't want GPU support :) Nov 24 17:49:43 opengl distro_feature does not necessarily mean you need dedicated GPU Nov 24 17:50:42 correct.. that is a setting within mesa, weston or... Nov 24 17:54:39 khem: don't want the DISTRO_FEATURES :) Nov 24 17:55:25 we are size constrained quite a lot and this DISTRO_FEATURES is usually pulling a lot Nov 24 17:55:40 (through "dynamic" PACKAGECONFIGs) Nov 24 17:55:45 IIRC Nov 24 17:56:34 I've found that things that say they want opengl are usually not appropriate for constrained systems.. :) (i.e. they have heavy graphic needs) Nov 24 19:37:44 mc in yocto always says "7-bit ASCII mode". Why? Nov 24 21:42:52 dv|2: perhaps missing UTF-8 charsets ? Nov 24 21:44:06 khem, systemd-localed is there and localectl reports System Locale: LANG=en_US.UTF-8.terminal works with utf8 Nov 24 21:45:22 mc won't set utf8 in "Display bits". it always changed back to 7-bit ascii on startup Nov 24 21:49:02 interesting does mc have some locale bits parted into its own packages that are needed on target I wonder Nov 24 21:51:00 khem, it detects terminal as 7-bit. I think, it is not about the locale files... Nov 24 21:53:42 it is related, can you try `export LC_ALL=en_US.UTF-8 ` and then launch mc Nov 24 21:55:41 khem. It detected UTF-8! But the problem in pseudographics stays the same Nov 24 22:03:03 dv|2: might be terminfo related instead of locale Nov 24 22:03:37 smurray, how can I detect that? Nov 24 22:06:36 dv|2: from a quick look at the recipe it RDEPENDS on ncurses-terminfo-base, you might need ncurses-terminfo as well if you're using a minimal image Nov 24 22:07:24 dv|2: the other thing I've seen is that for things like screen or tmux, you might need to tweak things to get a terminfo that matches the default TERM setting Nov 24 22:07:58 dv|2: worth seeing what TERM is, and perhaps trying e.g. "TERM=vt100 mc" to see if it works Nov 24 22:09:26 smurray, ncurses-terminfo-base is there. TERM was xterm, like on my PC. with TERM=vt100 it gets black and white, pseudographics stay corrupted Nov 24 22:10:19 I am running it in Putty on top of bare SSH, no screen, no tmux Nov 24 22:10:43 dv|2: try TERM=xterm-color Nov 24 22:12:24 smurray, oh, I just got MC pseudographics work with "export NCURSES_NO_UTF8_ACS=1". found recipe there: https://serverfault.com/questions/475925/how-to-fix-putty-showing-garbled-characters/705558#705558 Nov 24 22:12:47 dv|2: yeah, I was about to say, likely to be a putty issue Nov 24 22:13:00 smurray, xterm-color gave nothing... Nov 24 22:13:34 dv|2: that's one of the terminfos that's definitely in ncurses-terminfo-base Nov 24 22:13:58 smurray, putty issue... Definitely, I have running it OK with any Linux or other terms. But PuTTY corrupts pseudographics. Nov 24 22:14:28 dv|2: if you're on win10, the WSL terminal is supposed to be decent Nov 24 22:15:25 smurray, I'm Linux guy. what is WSL? Nov 24 22:15:47 dv|2: windows subsystem for linux, MS's new Linux support stuff Nov 24 22:16:15 dv|2: I'm a Linux guy too, but that's the new hotness for Windows user's doing Linux Nov 24 22:16:21 *users Nov 24 22:17:26 smurray, oh! my customers are mostly using simple Putty... I have to fix that in my devices, not on they side. OK. What is the conclusion about NCURSES_NO_UTF8_ACS=1? what can be fixed? Nov 24 22:17:49 smurray, users, thanks Nov 24 22:18:36 dv|2: you could hack ncurses in your tree to make that a default, though I suspect there's a non-zero chance it might break behavior for some other terminal Nov 24 22:19:05 dv|2: or bake in something via /etc/environment or /etc/profile.d/foo.sh, etc. Nov 24 22:19:51 dv|2: or stick in a mc wrapper. If Putty sets something in the env you can detect, that'd be somewhat innocuous Nov 24 22:22:17 smurray, thank you. I've thought about profile.d... don't want to hack ncurses, nor add more to mc wrapper Nov 24 22:23:47 dv|2: just need to bbappend the mc recipe and move the binary to e.g. mc.real and the install wrapper as mc, seems perhaps the least pain to me Nov 24 22:26:08 dv|2: look at create_wrapper() in utils.bbclass Nov 24 22:26:26 even better ;) Nov 24 22:27:18 smurray, too long way :) I'd better add two variables export into my default profile.d Nov 24 22:27:48 heh Nov 24 22:28:40 smurray, thank you for your help! Looks suspicious that On any Linux system I have no this bug with ncurses... Nov 24 22:31:31 dv|2: when I did last use putty about 10 years ago, it had problems like this back then as well, iirc Nov 24 22:31:41 smurray, https://linux.die.net/man/3/ncurses says about terminfo the the help... probably my system have no some terminfo packets... Nov 24 22:32:26 dv|2: look in /usr/share/terminfo, you'll see what's available Nov 24 22:34:35 smurray, xterm is in /usr/share/terminfo and in /etc/terminfo/ too (from ncurses-terminfo-base) Nov 24 22:35:07 dv|2: if it works fine from a Linux machine and not from putty, that is quite suggestive as to where the problem lies Nov 24 22:36:42 smurray, ncurses doc says, it tests for terminal capabilities somehow. =1 means "do not check"! Nov 24 22:37:23 smurray, I suspect the problem is probably in wrong information of the terminfo Nov 24 22:38:03 dv|2: xterm may not be right for putty, it might not something different/simpler Nov 24 22:38:09 err, might need Nov 24 22:38:46 smurray, xterm in PuTTY works for other Linuxes... Nov 24 22:39:19 dv|2: I dunno, break out strace and see what it's doing Nov 24 22:39:52 dv|2: like I said, I've not used it in 10 years, and I'm also not a mc user, so I've exhausted what I can offer Nov 24 22:41:53 smurray, oh, probably I know what the problem is! I have not "ncurses-terminfo-base" on my PC's Linux. But for device I have both: ncurses-terminfo-base and ncurses-terminfo. MC opens the first one and doesn't go to complete definition at ncurses-terminfo... Nov 24 22:43:03 dv|2: ncurses-terminfo-base is a OE packaging thing, I wouldn't necessarily expect a desktop distro to reflect that packaging Nov 24 22:43:47 smurray, it is in MC dependancy Nov 24 22:44:04 e.g. on fedora the packages are ncurses-base and ncurses-term Nov 24 23:29:39 hi, i'm hitting a problem with rpm's debugedit, building against a newer elfutils resolves it ... but even if i PREFERRED_VERSION_elfutils-native too many things rebuild ... not quite sure how to work around this Nov 24 23:35:48 looks like with rpm-native changing every rpm will rebuild ... **** ENDING LOGGING AT Wed Nov 25 02:59:57 2020