**** BEGIN LOGGING AT Fri Nov 13 02:59:57 2020 Nov 13 06:47:48 hi, we added a second display ( oled black and wite) to our yocto target how do i get an image to it ? ( it is listed as fb1 and i'm using weston) Nov 13 07:32:45 Hi everyone, quick question. Is it possible to make bitbake execute a script when it's done building? I would like to be notified when my build is done. Does bitbake have anything that could work for this? Nov 13 07:44:42 >yo dudX Nov 13 07:54:23 hi LetoThe2nd Nov 13 07:58:23 Sibert: see IMAGE_POSTPROCESS_COMMAND or task do_image_complete Nov 13 08:06:57 alright thanks mckoan Nov 13 08:19:09 Hi! I need to add the Jinja2 template engine to my SDK for code generation. But if I do this TOOLCHAIN_HOST_TASK_append = " nativesdk-python3-jinja2" bitbake can't find the package. but if I do TOOLCHAIN_HOST_TASK_append = " python3-jinja2" I'm getting " - package python3-jinja2-2.10-r0.cortexa9t2hf_neon does not have a compatible architecture" How Nov 13 08:19:09 do I solve this? Thanks! Nov 13 08:28:42 Plinnie: have you tried "python3-jinja2-native"? thats the package name to be used when needing to run it on the host. Nov 13 08:32:01 No, what is the difference between nativesdk-* and *-native? However, I get Nothing RPROVIDES 'python3-jinja2-native' . Nov 13 08:32:54 I know the package does exist. It is in my oe distribution (as part of Petalinux 2019.1 by the way) Nov 13 08:32:58 nativesdk-* is a task, *-native is a package Nov 13 08:33:24 Oh, but then I probably need to add it as a dependency, not as a task Nov 13 08:33:46 * LetoThe2nd is just guessing, almost no experience there too. Nov 13 08:34:41 ok.. well, I've been working with this for some time, but it is quite a study to really understand it. It still feels more like hacking then mastering :-) Nov 13 08:35:52 the name of the variable might be misleading, but the documentation pretty clearly states that package names shall be added to TOOLCHIN_HOST_TASK: https://www.yoctoproject.org/docs/current/sdk-manual/sdk-manual.html#sdk-adding-individual-packages Nov 13 08:37:07 ok, well, unfortunately no success. I've now tried adding it as a dependency to the recipe which actually uses it (DEPENDS += "grpc grpc-native python3-jinja2 python3-jinja2-native").. but it still does not seem to be able to find it. Nov 13 08:37:32 it is like the package does not export a native version Nov 13 08:38:24 (it has no problems with python3-jinja2) Nov 13 08:38:25 it has BBCLASSESTEND native, hence it should work. what if you bitbake the -native version directly? Nov 13 08:38:49 maybe my version of OE is too old. Nov 13 08:39:38 ... which is? Nov 13 08:40:57 I'm not sure... it is provided by Xilinx. I will try to look it up. Anyway, attempted to build it directly gives the same error. "Nothing PROVIDES 'python3-jinja2-native' " Nov 13 08:41:31 that sounds fishy. let me check. Nov 13 08:42:29 apparently it is yocto 2.6.1 .. Nov 13 08:47:09 Hi guys... I have to add meta layer which contains a lot of .bbappends which cannot be found in other layers... I can then add a dummy recipe in my custom meta layer and .bbappends don't complain anymore Nov 13 08:47:34 is this ok solution or is there any better? Nov 13 08:47:57 wooosaiii: then you are either mismatching versiosn, the layer has a dependency that you do not have, or it is outright buggy. pick your case :) Nov 13 08:48:14 layer has a dependency I don't have :D Nov 13 08:48:31 wooosaiii: then why don't you add it? Nov 13 08:49:13 LetoThe2nd: I will try it... Nov 13 08:50:06 but I don't want to use it in the end i.e I have to find a solution for not using it in the end Nov 13 08:51:03 wooosaiii: essentially it boils down to two alternatives: 1) the layer depends on another specific layer. then you have to add it too. (and the layer you already have is buggy because it does not list the dependency) or 2) the layer optionally can use things from another layer, but does not actually depend on it. then the layer you already have is buggy because it does not mask out the bbapends. Nov 13 08:51:11 so either way, the layer is buggy. Nov 13 08:51:18 let me guess, BSP? ;-) Nov 13 08:51:53 So, I could probably simply extend python3-jinja2 using bb-extend and add the native class, right? Nov 13 08:52:22 LetoThe2nd: long story short... I want to remove meta-arago and use plain poky Nov 13 08:53:13 LetoThe2nd: however there are a lot of bbappends in meta-phytec that require this layer... Nov 13 08:54:24 LetoThe2nd ok, anyway, thanks for your help. Time for coffee Nov 13 08:54:55 Plinnie: i can confirm that it works nicely for dunfell. 2.6 is thud, which is EOL, plus the jinja recipes are massively different there and being carried in meta-python still. so you might try and backport from dunfell poky, but essentially you're out of support. Nov 13 08:55:51 wooosaiii: long answer short: if they sold you something that depends on something you don't want to use, either ask them to sell you a version without the dependency, or hack up the solution yourself. Nov 13 08:56:36 wooosaiii: i personally refuse to do support for BSP induced problem that go beyond the absolute minimum (like, how is 'bitbake' spelled, or comparatively) Nov 13 08:58:51 LetoThe2nd: OK... but lets say it has nothing to do with BSP... IMHO... I have to options: 1) add dummy recipes in my custom meta layer 2) do BBMASK for those bbappends that fail Nov 13 08:59:18 LetoThe2nd: right or? Nov 13 09:03:29 * LetoThe2nd will not "lets say", because i have given all the hints i am willing to do in that case. Nov 13 09:03:56 it might sound rough, but i'm picky there. they sold you something, they got money for it. i do not. Nov 13 09:05:05 LetoThe2nd: OK :) I respect that... Nov 13 09:07:21 wooosaiii: BBMASK things away. I do that for all of our BSP layers which contain things I don't want or need. I end BBMASKing away 90% of things and keep just bootloader and kernel and few userspace pieces like GPU libs... Nov 13 09:08:08 mcfrisk: thanks for advice Nov 13 09:08:24 i totally understand that you're the scapegoat here. but we have so many vendors handing out bsps that cause problems. if we happily help and fix all that stuff without it being taken back to them, why should they even try and change? see. go back to their support, and request what fits your needs. or figure out how to handle the stuff for yourself. the hints to masking are already there. and obviously, others have a Nov 13 09:08:24 different point of view than i have - and thats completely legit too. Nov 13 09:09:59 LetoThe2nd: yeah.. will have to do it on my own :) Nov 13 09:10:09 wooosaiii: have fun! Nov 13 09:17:49 I'm trying to add a file to my rootfs (/lib/firmware/). I've used 'install' in do_install_append but it's not there on my target Nov 13 09:17:59 also added FILES_${PN} += "${datadir}/lib/firmware" in my recipe Nov 13 09:18:43 What am I doing wrong? Nov 13 09:18:53 Sibert: its not datadir :) Nov 13 09:19:31 then the internet lied to me Nov 13 09:19:35 what a shocker :P Nov 13 09:19:55 hm wait. Nov 13 09:21:55 nah, nvm. datadir points to /usr/share in the target. Nov 13 09:22:35 ah so it is wrong Nov 13 09:22:39 Hi! I have a repository with submodules and I need to add a patch for a submodule. What is the correct way to add it in the recipe? Nov 13 09:23:12 Sibert: yup, *think* it should be D. but not super certain. Nov 13 09:24:09 okay Nov 13 09:24:13 thanks Nov 13 09:24:35 yocto is so freaking hard Nov 13 09:25:41 LetoThe2nd yes, I was planning on upgrading Petalinux, but it is a can of worms. I would practically need reintegrate all my recipes. Last time I tried grpc failed horribly. Nov 13 09:26:39 And upgrading yocto inside of Petalinux is probably a very bad idea. Nov 13 09:28:23 Sibert: can't find a good example ad hoc, sorry. Nov 13 09:28:46 Plinnie: well... Nov 13 09:30:56 I know I will have to at some point, but for now I'm trying to 'fix' things :-) Nov 13 09:31:07 Sibert: essentiall its https://www.yoctoproject.org/docs/3.1/ref-manual/ref-manual.html#var-FILES and meta/conf/bitbake.conf Nov 13 09:32:07 thanks I'll look into it Nov 13 09:32:54 Sibert: https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/conf/bitbake.conf#n46 Nov 13 09:33:19 so ${libdir}/firmware if i'm not completely mistaken. Nov 13 09:33:54 alright I'll try that Nov 13 09:37:12 more liekly nonarch_libdir actually if it's really for firwmares Nov 13 09:37:35 sorry, nonarch_base_libdir Nov 13 09:37:53 otherwise it stops working when you use multilib Nov 13 09:37:53 :) Nov 13 09:42:27 hm this is given build errors Nov 13 09:42:47 giving* Nov 13 09:42:47 Sibert: what has and what are the errors? Nov 13 09:43:00 recipe: Nov 13 09:43:01 FILES_${PN} += "${libdir}/firmware"do_install_append() { install -d ${D}/${libdir} install -d ${D}/${libdir}/firmware install -m 0777 ${WORKDIR}/toplevel_counter.bin ${D}/${libdir}/firmware/toplevel_counter2.bin} Nov 13 09:43:09 hm formatting is fucked Nov 13 09:43:20 linux-tegra-4.9.140+gitAUTOINC+a58470bb0f-r0 do_package: QA Issue: linux-tegra: Files/directories were installed but not shipped in any package: Nov 13 09:43:35 Please set FILES such that these items are packaged. Alternatively if they are unneeded, avoid installing them or delete them within do_install.linux-tegra: 4 installed and not shipped files. [installed-vs-shipped] Nov 13 09:43:56 it's trying to install /usr, which is not right either Nov 13 09:45:04 Sibert: I see it's a do_install_append so check your do_install :) Nov 13 09:46:59 I'm appending to my actual linux kernel Nov 13 09:47:02 might that be the issue? Nov 13 09:47:14 I'm appending linux-tegra_4.9 Nov 13 09:49:17 first, replace libdir with nonarch_base_libdir ;) Nov 13 09:49:49 otherwise, we need to have the logs and recipe to help you Nov 13 09:50:16 nah, me doesn't Nov 13 09:50:24 i will not help, only annoy and sabotage! Nov 13 09:52:45 ndec: re the dunfell docs... I don't really know which way we should go, I cannot decently review the first (the big one) patch without actually doing the work you did.. Nov 13 09:52:53 Do we just take it and then do a pass on it again Nov 13 09:52:55 ? Nov 13 09:53:12 (a pass = a review on the actual output, like I did so far with gatesgarth) Nov 13 09:55:45 and very stupid question but... do we actually need to migrate dunfell to sphinx too? why? Nov 13 10:01:32 qschulz: for bitbake it's a bit simpler than yocto-docs. i know the first commit is 'unreviewable'. but at least i recorded the commands to regenerate it. then comparing the log history of 1.46 and 1.48 is straight forward, and we can clearly see which commits to 'fixup'. Nov 13 10:03:04 for yocto-docs, it's a bit more tedious.. since there are docs changes in dunfell branch (docbook) which don't exist in master (sphinx) (mostly all the commits about each release). then there are changes in master (docbook) that may or may not apply to dunfell (which we need to keep them or 'revert' them). and then there are plenty of docs changes in sphinx that may or may not apply to dunfell Nov 13 10:04:47 qschulz: i have an initial WIP branch for yocto-docs, just pushed it here: http://git.yoctoproject.org/cgit/cgit.cgi/yocto-docs/log/?h=sphinx Nov 13 10:05:17 what's left to do is to 'undo' the changes brought by the sphinx migration which don't apply to dunfell. Nov 13 10:09:14 https://www.irccloud.com/pastebin/fWv6jvrU/ Nov 13 10:09:34 qschulz: ^^ this is the output of my analysis for yocto-docs.. Nov 13 10:20:33 LetoThe2nd Good news, I managed to get it working. I've created python3-jinja2_%.bbappend and python3-markupsafe_%.bbappend (dependency), and both files contain a single line 'BBCLASSEXTEND = "native"' Nov 13 10:20:50 Its not very nice, but it works Nov 13 10:24:49 Plinnie: \o/ Nov 13 10:29:05 it was your hint that it needed a -native package which helped point me in the right direction, so thanks. Nov 13 10:29:46 :) Nov 13 10:30:14 i love giving hints, just no handholding. glad if somebody knows and can handle this. Nov 13 10:39:35 how can we make a recipe to inherit a bbclass conditionally ? Nov 13 10:43:59 yann: inherit ${@ 'myclass' if d.getVar('MEH') else ''}? Nov 13 10:44:12 or something along those lines, there are a few examples in some layers (meta-qt5? poky?) Nov 13 10:46:22 oh right, I was so sure that one would not work I did not even check, thx! Nov 13 10:46:44 hehe Nov 13 10:47:11 yann: We use such a mechanism in thud, didn't work on krogoth though, so might depend on your version of Yocto Nov 13 10:47:20 standard software "this is so crazy it cannot work -> check -> doesn't work" Nov 13 10:47:48 yocto: "this is so crazy it cannot work -> check just for fun -> It WORKS?!?!?" Nov 13 10:49:03 eh eh :) Nov 13 10:50:11 actually I use it with require and not inherit, but is worth a try anyway Nov 13 10:51:11 its just one of those days: 1) go down do mfg floor, check box is powered on. 2) go back up to office. 3) doesn't work, go back to mfg floor and pick up box 3) go back up to office, set up box 4) take box back down to mfg floor 5) go back up to office 6) realize that you forgot to power on... Nov 13 10:52:01 LetoThe2nd: TGIF Nov 13 10:52:24 * LetoThe2nd starts another round trip to the mfg floor, then. Nov 13 10:53:19 LetoThe2nd: home gym... but at work :p Nov 13 10:57:37 yes, my external brain #2 (a.k.a. hipster smartwatch) confirms that i've already exceeded expectations in terms of stairs i shall climb. Nov 13 11:04:00 can somebody please kick me so i finally get the livecoding announcement done? Nov 13 11:04:24 with pleasure Nov 13 11:04:41 * rburton kicks LetoThe2nd Nov 13 11:05:15 rburton: thx Nov 13 11:19:01 ok, I was too fast. Even though it compiles, there still is nothing in the native part of the SDK.... ;( X) Nov 13 11:20:23 even though it can find the python3-jinja2-natvie and nativesdk-python3-jinja2 packages it doesn't actually seem to do anything. I must be missing something. Nov 13 11:55:39 did you actually add them to the sdk? Nov 13 12:09:30 rburton I'm still a bit of a noob in this subject. I have an old Yocto version (part of Petalinux) which I still need to use. Now the original recipes do not have the line BBCLASSEXTEND = "native nativesdk". I created bbappend files for those recipes and added this. I also added TOOLCHAIN_HOST_TASK_append = " nativesdk-python3-jinja2" Nov 13 12:09:54 Is there anything else I should do to make this work? Nov 13 12:09:57 that last bit should have worked Nov 13 12:10:15 where did you put the TOOLCHAIN_HOST_TASK> Nov 13 12:10:54 its inside petalinuxbsp.conf . There I've also added some other tasks, which do end up in the sdk Nov 13 12:11:19 i'd say that should have worked then Nov 13 12:11:21 (TOOLCHAIN_HOST_TASK_append = " nativesdk-protobuf-compiler"TOOLCHAIN_HOST_TASK_append = " nativesdk-grpc-dev") Nov 13 12:11:43 yes, I will do a mrproper and retry Nov 13 12:11:50 maybe something is stale Nov 13 12:12:10 try using oe-pkgdata-util to list the contents of nativesdk-python3-jinja to be sure it wasn't a stupid packaging mistake Nov 13 12:12:40 ok, I'll try. never used that before Nov 13 12:22:59 So I've got a layer containing foo-image and bar-image -- both images require package baz, but in foo-image, custom overrides should be applied, i.e. custom build parameters and some changes in file system. What's the best solution? Nov 13 12:30:47 hmm.. this seems to be a bit more complicated in Petalinux than with plain yocto Nov 13 12:31:06 jmiehe: you want package baz but compiled differently depending on which image is built? Nov 13 12:34:30 anyone done a yocto bitbake on a 5950x yet? :D would be interested to know build times... Nov 13 12:42:22 I don't seem to be able to get oe-pkgdata-util to work.. this must be petalinux thing Nov 13 12:44:06 https://github.com/shr-project/test-oe-build-time - curious to know how the 5950x stacks up against 3960x or 3970x Nov 13 12:44:18 "ERROR: Unable to connect to bitbake server, or start one" when I try to do anything but invoke help Nov 13 12:46:25 qschulz: yes, while baz might also be an official package like nginx. Nov 13 12:46:31 Petalinux adds too much 'magic' to the distribution. I wish they would have taken a bit more pragmatic approach. Nov 13 12:47:07 jmiehe: that can't work, image recipes can't impact other recipes. Only conf files can impact recipes (i.e., machine, distro, layer, local configuration files) Nov 13 12:48:23 i know Nov 13 12:48:36 yes. So can I define my own "foo-baz" recipe which basically extends "baz" recipe but adds in custom stuff? Nov 13 12:48:46 qschulz: @ Nov 13 12:51:49 jmiehe: yup, that's the other way to do it Nov 13 12:53:26 jmiehe: just have a second recipe (named differently) that inherits the first one Nov 13 12:55:56 s/inherits/requires/ Nov 13 13:38:08 hello everybody, quick question: I have a recipe X that inherits "deploy" and deploys a file F, then I have a recipe Y that uses file F during compilation. Now I want Y to re-run "do_compile" task when the F has changed. Is this possible? Nov 13 13:39:22 robin92pl: You'll need Y to depend on X Nov 13 13:39:26 robin92pl: if you've set the dependency correctly, it'll just work Nov 13 13:39:56 do_task[depends] += "recipeX:do_deploy" in recipe Y Nov 13 13:41:11 do_compile[depends] += "recipeX:do_deploy" in recipe Y even Nov 13 13:42:30 @paulbarker, @qschulz: I did that and sometimes do_compile is not triggered even though F has changed. I also tried with DEPENDS_append for a package-wise dependency but it also failed. I think I will checkout the latest Yocto and try to reproduce the problem. Thanks for quick response. Nov 13 13:44:54 robin92pl: how is F "changed"? Nov 13 13:47:01 qschulz: I run `cleanall` task for X and then run `bitbake Y`; fiile F has creation timestamp in the first line so it is different than before running "cleanall" Nov 13 13:48:13 robin92pl: which version of Yocto are you using? Nov 13 13:48:45 qschulz: quite old unfortunately - 2.4 Nov 13 13:48:52 robin92pl: DEPENDS = "X" in recipe Y is incorrect and you can have a race Nov 13 13:48:57 (just FYI) Nov 13 13:50:48 robin92pl: where do you install your file F in your deploy task in recipe X? Nov 13 13:50:51 what's the path? Nov 13 13:51:37 qschulz: as for the race, I suspect it may be when only DEPENDS is used but not when both DEPENDS and varflag, isn't it? I deploy to ${DEPLOYDIR} Nov 13 13:52:43 when you've the line I sent, it should just work yeah Nov 13 13:52:48 how did you add your deploy task? Nov 13 13:53:38 could you send both recipes maybe? Nov 13 13:54:59 qschulz: not really, though, I will send simplified recipes when I manage to reproduce the issue on 100% my recipes. Right now I am working on bbappends for a recipes provided by our chip provider. Nov 13 13:56:32 robin92pl: check for typos but outside of this I'm clueless at the moment without more code :/ Nov 13 13:57:31 qschulz: yeah, so am I; more than a solution I was looking for a confirmation that my approach was correct and it *should* work; now I can start debugging why it fails Nov 13 14:08:51 if I am gonna patch a config file with sed, at what stage should it be done?. based on the image I am using I set a var, that var is used to fill in the place holder in the config file. Nov 13 14:09:36 the config file is a part of my own recipe Nov 13 14:10:05 Ad0: depends when your file is being used? Nov 13 14:10:12 I've seen a few seds in do_configure_prepend Nov 13 14:10:35 the file is just put on a location for use by a systemd service after boot Nov 13 14:10:55 Ad0: If you're checking a var set in the image you'll need to patch the file in the image recipe itself. No other recipe knows which image you're building Nov 13 14:11:20 maybe I should set the var elsewhere Nov 13 14:11:38 Ad0: conf file or in the recipe itself (manually that is) Nov 13 14:12:06 qschulz, hm? :) Nov 13 14:12:47 Ad0: for the variable... if you want to set a variable that needs to be used in a recipe, either you set it in the recipe, or in a configuration file (distro, machine, layer, local...) Nov 13 14:13:08 ok Nov 13 14:13:28 so there is no way to define variables in an image recipe that is exported for global use in other recipes Nov 13 14:13:52 Ad0: no because an image recipe is a recipe Nov 13 14:14:02 and you can't share variables between recipes Nov 13 14:14:05 ok so it's in that scope Nov 13 14:14:13 a common scenario are some settings like host / url depending on production / dev Nov 13 14:14:31 so I should rather make a distro for production / dev? Nov 13 14:16:22 I could echo it to a file and then use the file in a later stage in the life cycle lol Nov 13 14:18:57 I asked earlier some days ago what the best route was for having a prod / dev setup, a) have prod / dev iamges with different packages and image settings, b) have different distros and have conditional includes based on the distro in one image file Nov 13 14:19:24 maybe there is a C too :) Nov 13 14:22:13 Ad0: you already know everything then :) Nov 13 14:22:25 lol Nov 13 14:22:33 I dunno Nov 13 14:22:42 what would you do personally :) Nov 13 14:23:01 if the 2 differences are image settings like readonly rootfs etc and packages, and that variable Nov 13 14:23:40 I have 2 machines, and 2-3 targets like prod, dev, QA Nov 13 14:24:01 right now it's 1 distro, 2 machines, image recipes for prod and dev Nov 13 14:25:04 maybe I should have divided up in distros because then the OS name can be different too... Nov 13 14:25:44 Ad0: FWIW, that's the exact situation we wrote https://github.com/garmin/whisk to handle Nov 13 14:26:00 I think it's possible to do something similar with kas Nov 13 14:26:40 interesting Nov 13 14:26:49 Ad0: Well, that *and* our different products use different version of Yocto, but we want them all in the same source tree :) Nov 13 14:26:54 it looks like a job to make it work with whisk Nov 13 14:27:40 i sorta made my init do different things to manage bbappends and all that Nov 13 14:28:50 Ya; that kinda thing is pretty common. Whisk handles that by allowing different products to pull in different layers Nov 13 14:29:41 But in general we've moved away from "product layers" and use variables defined in the product conf file instead.... they layers were... abused... sometimes :/ Nov 13 14:30:02 hehe Nov 13 14:30:07 thanks I will bookmark that for later Nov 13 14:30:19 seems like a bit of a job to integrate that and to redo everything Nov 13 14:30:29 Ad0: Ya, that's fair Nov 13 14:31:09 * JPEW needs to write a quickstart guide for it... Nov 13 14:31:15 is it yours? Nov 13 14:31:37 Ya. It was an internal tool that I got published Nov 13 14:31:48 And I wrote it Nov 13 14:31:51 nice! seems fairly new too Nov 13 14:32:20 I will def look into it Nov 13 14:32:42 We've been using it internally longer.... we're still transitioning all our products to use it, but we currently have about 15 using it Nov 13 14:33:20 nice Nov 13 14:36:13 so if you were in my shoes and whisk didn't exist, what would you do? :) Nov 13 14:40:37 Ad0: If I *had* to do it stock.... I would either A) use TEMPLATECONF to make pre-defined local.conf.sample/bblayer.conf.sample templates for each different build you want Nov 13 14:41:07 thanks Nov 13 14:41:14 Ad0: or, make pre-canned site.conf or auto.conf files for each build that users can copy or symlink into the conf/ directory Nov 13 14:41:39 never heard of those files Nov 13 14:42:01 Ya, so it turns our bitbake looks for local.conf, site.conf, and auto.conf in the conf/ directory Nov 13 14:42:07 hehe ok Nov 13 14:42:19 running bitbake-selftest shows several tests not finding bitbake, in run_bitbakecmd(), Would it be making expectations not encapsulated in the test-running logic ? Nov 13 14:43:43 Ad0: I'd probably prefer the second option (linking in a site.conf) honestly; it's a lot easier to get a consistent build. The template stuff works really well to initial setup an environment, but IMHO it's not the greatest at tracking changes Nov 13 14:44:14 thanks! Nov 13 14:44:40 different distros and everything then Nov 13 14:45:21 Depends on how much you want to change.... if it's just a few minor things I'd probably put them in the per-build site.conf Nov 13 14:46:39 We have just a few distros that define broad policies: for example one distro for headless devices that has no graphics support, and another for devices with a screen that enables all the graphical goodies Nov 13 14:47:18 well the difference i have are 2 machines, production/dev/qa builds (os name, packages in image, readonly fs on prod), patching of files based on production/dev Nov 13 14:47:42 patching meaning just sed'ing variable into a config file Nov 13 14:48:23 Hi, I have a recipe that install -g 1002 where 1002 is present in files/group (but not in /etc/group). I get an error with dunfell from sstatesig.py: KeyError: 'getgrgid(): gid not found: 1002. Can it be a dependency missing in my recipe? Nov 13 14:48:31 Ah, ya. Once you get multiple axis of configuration it gets more difficult (hence, why we wrote whisk :) ) Nov 13 14:48:59 Hi, I've got a problem getting php and apache2 to run under yocto (dunfell) it keeps failing with php-native not being able to find the httpd.conf for apache2-native Nov 13 14:49:33 stenscott: IIRC you need to actually create the group somewhere. Just having it in files/group is insufficient Nov 13 14:50:09 stenscott: files/group is a template for how to create a given group (when using static IDs); it's not copied verbatium Nov 13 14:50:23 JPEW, my thought was to just divide it up in distros now. because a development is sort of is own distro since it has a bunch of development distro "features" like debugging tools Nov 13 14:50:35 Ad0: Ya, that's fair Nov 13 14:50:52 a bonus is to being able to have different OS / Distro names Nov 13 14:51:03 I'll check out whisk too Nov 13 14:51:15 so you basically have your own layer in there. and whisk manages all the other layers? Nov 13 14:51:35 Ad0: Ya. If you want a functional example of using whisk, you can check out: https://github.com/JPEWdev/yocto-doom-demo Nov 13 14:51:45 thanks! Nov 13 14:52:11 seems like the submodules are checked in Nov 13 14:52:21 JPEW: thanks, I thought it was enough with USERADDEXTENSION = "useradd-staticids" in conf Nov 13 14:52:31 stenscott: Nope Nov 13 14:52:50 Ad0: Ya, I use submodules, but whisk doesn't care what you use as long as they are present Nov 13 14:52:55 It used to work before dunfell, but I guess that was just luck Nov 13 14:53:14 JPEW, ok ! Nov 13 14:54:11 Ok, so I'm the only one stupid enough to try to get PHP on my embedded system it seems (I don't blame you) Nov 13 15:15:39 JPEW: Users/groups are actually added in other recipes for the static ids with USERADD_PACKAGES, but the recipe that changes the file permissions doesn't seem to have access to those ids Nov 13 15:23:52 stenscott: Ya... does it DEPEND on the other recipe? Nov 13 15:30:35 JPEW: No, that may be the issue. I'll try modifying the DEPENDs. Nov 13 15:30:54 Is there any way to get my build to not include a recipe (ie. php-native ) when I have `IMAGE_INSTALL_remove += "php-native"` and `PACKAGE_EXCLUDE += "php-native"` in my image file? Nov 13 15:34:15 Paul41: You can use BBMASK to hide recipes from bitbake Nov 13 15:35:42 JPEW: I think that solved my issue. Thanks again! Nov 13 15:36:11 One of our apps is using yarn, so we have written a yarn fetcher, which is kinda working bug kludgy (derives from GitSM, from which it inherits donestamp handling, so git's donestamp is wrongly taken for the whole). I'd like to set all of this straight so it gets into proper shape for submission, and I'm looking for advice. Nov 13 15:37:33 I've looked at how npmsw does, and the basic idea looks nice, except the yarn.lock to reference (the equivalent of npm-shrinkwrap.json) is part of the git repo Nov 13 15:39:37 are users of npmsw making a copy of the npm-shrinkwrap.json file for this ? Nov 13 15:59:31 well, things are made harder with yarn, since downloading the versions needs not only the yarn.lock file, but the full package.json which may include other files, which basically requires the full repo checked out just to fetch deps, sigh. Nov 13 16:04:41 stenscott, I forgot about that one, thank you I might try that Nov 13 16:06:30 Nope, doesn't help. Problem is I need php, but php-native keeps failing to install because of apache2 configuration is missing Nov 13 16:11:30 Is there a way to configure a package one way for -native and one way for target? Nov 13 16:11:40 *another way for target Nov 13 16:16:45 Paul41: You should be able to use _class-target and _class-native overrides IIRC Nov 13 16:17:35 Turns out in my .bbappend I can `BBCLASSEXTEND_remove = " native"` But apparently I'm not allowed to Nov 13 16:17:53 ERROR: Nothing PROVIDES 'php-native' Nov 13 16:17:55 :( Nov 13 16:22:47 Is there no way to change configuration for the -native package? Nov 13 16:23:10 Paul41: paulbarker told you above how to do it Nov 13 16:26:10 can you try his way and see if it helps? Nov 13 16:27:33 OHH! Sorry I didn't get that it was `_class-target` because of webirc doing markup, will try that now! Nov 13 16:37:37 Well at least it compiled and did the image this time, thank you so much for the help Nov 13 17:01:35 have the Summit videos been posted? Nov 13 17:20:12 qschulz: I managed to reproduced the problem on clean dunfell release; the minimal recipes and my terminal session have been uploaded to the Google Drive; feel free to check them out: https://drive.google.com/file/d/1cruql38-QOZ1i2mNNjfaWYdDoy7gtwg0/view Nov 13 17:53:06 dreyna: ndec: have the Summit videos been posted? Nov 13 17:53:23 tlwoerner: no! Nov 13 18:38:59 ndec, dreyna: do you need help processing the videos? I am taking time off of the US thanksgiving week. Nov 13 18:48:53 denix: does bbb wic file write MLO and u-boot to raw sectors or load from FAT partition Nov 13 18:49:33 khem: bbb loads them from fat Nov 13 18:51:33 denix: I am trying to format eMMC from a SD boot and I put u-boot and MLO in FAT partition on eMMC and eject SD, it seems to load it but with one caveat, I also have uEnv.txt to customize uboot env a little. it gets used when booting from SD but not when booting from eMMC Nov 13 18:52:16 denix: I am trying to use kernel + dtb which is stored on FAT instead of one from roorfs/boot directory Nov 13 18:54:07 denix: my uEnv.txt is http://sprunge.us/a7Wc9n Nov 13 18:54:34 armpit: thanks, that would really help! as you guessed that's the reason why i haven't done it yet.. so i have 2 videos of 7/8 hours each, they need to be split by session/presentation.. i can give one to you, and do the other one? Nov 13 18:55:14 khem: https://groups.google.com/g/beagleboard/c/1S9of2vmaLM?pli=1 Nov 13 18:57:54 khem: uEnv.txt is expected from SD card, not eMMC Nov 13 18:58:46 denix: so whats the magic for doing it in eMMC as well ? should I need to patch u-boot env directly ? Nov 13 19:01:01 khem: can you saveenv? Nov 13 19:01:26 denix: its installer image I am trying for Nov 13 19:01:35 so it should work out of box Nov 13 19:02:19 khem: sounds like you need to patch the default env then... Nov 13 19:02:48 denix: thanks was trying to avoid patching u-boot but I guess its fine Nov 13 19:10:25 denix: is include/configs/am335x_evm.h right file to alter DEFAULT_LINUX_BOOT_ENV Nov 13 19:11:36 khem: yes Nov 13 19:37:46 denix: if I run envboot on u-boot CLI it loads uEnv.txt from emmc too Nov 13 23:20:23 JPEW: not sure why but that systemd patch doesn't apply Nov 14 01:39:53 zeddii: can you branch meta-cloud-services for gatesgarth please ? Nov 14 01:40:06 it's not ready yet :D Nov 14 01:40:13 oh bummer Nov 14 01:40:29 neither is meta-virt. I hope early next week. Nov 14 01:40:41 rpi4 kernel is failing to compile on linux-yocto-dev 5.10-rc so I am reverting... Nov 14 01:40:45 ah ok Nov 14 01:41:57 I finally figured out my dns woes so I want to get k8s running on the rpi4 (yp-dev-day_virtualization) Nov 14 01:42:07 nice. Nov 14 01:42:29 my stupid ISP puts 4 nameservers in /etc/resolv.conf Nov 14 01:42:39 the wind river BSP guy (who's name has escaped me), will be sending 5.10 rpi curated patches FWIW. Nov 14 01:42:51 and then all the DHCP devices blindly follow suit... Nov 14 01:43:03 oh good Nov 14 01:43:15 Ming Li or something like that Nov 14 01:43:30 I'm still not able to get my single node master / agent k3s to work, I wasn't able to look at it again today. So I'm going to try once more on Monday before decided if I just boot k3s to the next release. Nov 14 01:43:46 yes. he's very good with the changes. He contacted me about three weeks ago to confirm the plan. Nov 14 01:44:00 yeah... I haven't tried k3s either... Nov 14 01:44:28 yah. wait on that. when I'm done, there will be a step by step README and packagegroups, kernel configs, etc, that should make it much easier. Nov 14 01:44:31 I keep meaning to build it Nov 14 01:44:46 nice Nov 14 01:44:53 I just can't get the agent to register. complains about either a certificate or username/password, or something similar. but it is quite close. Nov 14 01:45:07 I uprev'd k3s to the latest, hoping for some magic. Nov 14 01:45:12 alas, no. same problem :D Nov 14 01:45:17 because of Tekton, I am finally learning k8s Nov 14 01:45:25 sigh Nov 14 01:45:38 quite honestly, I find some of their concepts a bit obtuse. Nov 14 01:45:38 so much promise, so much pain Nov 14 01:45:44 indeed Nov 14 01:45:55 it's really, really, really hard to figure out what has gone wrong and get the right logs, etc. Nov 14 01:46:12 We now have another open source project that has a very steep learning curve but is very flexible and capable... Nov 14 01:46:14 and heaven forbid you figure out that containerd/runc, etc, were all down there doing the lifting Nov 14 01:46:25 strange nobody bitches about the ramp up constantly... Nov 14 01:46:28 that's an apt description. Nov 14 01:46:42 the docs are scattered / terrible for it too. Nov 14 01:46:51 but again, at least I don't see all the bitching. Nov 14 01:46:59 I guess you need google standing behind you. Nov 14 01:47:10 tgamblin and I basically figure the biggest service we are performing is gathering all those scattered things in one place ;) Nov 14 01:47:22 I agree, 100% Nov 14 01:47:41 the info is out there, collecting it coherently is the challenge. That's basically what I'm doing with k3s. Nov 14 01:47:53 I beat my head on cri-o and calico on debian-10 Nov 14 01:48:01 cri-o is fine, but calico... Nov 14 01:48:17 I should have just used the default bridge networking... stupid human tricks Nov 14 01:48:30 heheh. there's a reason why the CNCF ecosytem chart looks like a star field when viewed at 100% Nov 14 01:48:31 or stuck with flannel Nov 14 01:48:37 ha Nov 14 01:48:52 CNI and networking, yet more magic, buried way at the bottom. Nov 14 01:49:05 I struggled how to integrate and split it in meta-virt for k3s. Nov 14 01:49:12 I've figured out dozens of ways to make it not work Nov 14 01:49:21 lol. I was going to say the same thing. Nov 14 01:49:42 unreachable urls are my specialty Nov 14 01:49:48 Luckily I've seen enough patterns now to fix things (except calico) Nov 14 01:50:33 but which tool to use ? kubectl? tkn? dmesg? which log is the magic one that has the special sauce? roll the dice Nov 14 01:50:57 journalctl -xe > f; grep "wtf" f Nov 14 01:51:11 system really makes seeing the logs for some things hard Nov 14 01:51:12 | tee /dev/null Nov 14 01:51:21 yup Nov 14 01:51:28 there used to just be stuff in /var/log .. alas. Nov 14 01:51:55 and with Tekton, sometimes it's the dashboard that actually has the log you want to see (then you learn how to get the info from tkn) Nov 14 01:52:30 hah. 7 layers to get to the error message you can google :D Nov 14 01:52:35 oh, thank you so much for cloud-init Nov 14 01:52:52 I haven't used it yet, but that's on deck for the weekend Nov 14 01:53:10 it was a bear to get building / generating service files. I smoke tested it, it starts up. but it has a configuration learning curve to make it right for what your image wants. Nov 14 01:53:36 along with every other tool in this space :D Nov 14 01:53:50 I could see a cloud-init.bbclass or something like that for recipes / images to configure it, but that was overkill until I had a few examples of using it. Nov 14 01:54:18 yeah. I'll be capturing some patterns soon enough. Nov 14 01:54:27 yah. it can configure a LOT of stuff. I had to throw out everything I had done for openstack. so this is new, shiny and unconfigured. Nov 14 01:54:36 lol Nov 14 01:54:44 heck, just using it to set the hostname is a victory. Nov 14 01:54:53 \o/ Nov 14 01:54:54 meaning, that's surprisingly hard to do in an image recipe Nov 14 01:55:14 and that can make them unique on startup, etc. that's what I did in openstack, but they changed the config, so I deleted it for now. Nov 14 01:55:32 yep... and I've got that shiny ubuntu-20.04 image running on kvm to finally login to :) Nov 14 01:56:12 still steaming .debs in the morning dew. Nov 14 01:57:54 make[1]: *** No rule to make target 'overlays/at86rf233.dtbo'. Stop. Nov 14 01:57:58 oh well Nov 14 01:58:08 time to power down the droids I guess Nov 14 01:58:17 have a great weekend zeddii Nov 14 01:58:19 yep. call it a weekend, and pick up the bits on Monday. Nov 14 01:58:21 likewise! **** ENDING LOGGING AT Sat Nov 14 02:59:56 2020