**** BEGIN LOGGING AT Tue Nov 05 02:59:58 2019 Nov 05 07:59:05 good morning Nov 05 08:05:58 Hello! I'm using fitImage on my FPGA board. Because the device is running under very low cpu frequency. Also I don't care about security problems. That's saying, I don't want any hash algorithm to verify my Image. Could anyone give me any suggestion? Nov 05 08:11:43 zhangzhi: unless you specify anything, i don't think the fitimage would be signed or such Nov 05 08:15:47 letothe2nd, from poky/meta/classes/kernel-fitimage.bbclass, fitimage_emit_section_config() always generate " hash@1 { algo = "${conf_csum}"; };" Nov 05 08:17:35 U-boot takes very long time to check sha256 of each image Nov 05 08:18:16 why not select something really simple like crc? Nov 05 08:19:06 i can't look it up right now, but on our 350mhz board uboot checks a 30mbyte fitimage in like 1 second Nov 05 08:19:55 Anyone else having problems accessing https://www.yoctoproject.org/ due to certificate problems ? Nov 05 08:21:30 actually, I'm not only running it on my FPGA board (which use 20MHz cpu), also running it on zebu environment. It's not acceptable to take too much time to check image. Nov 05 08:22:05 zhangzhi: ah ok. just had a quick peek. we're using a custom class to generate the fit image, and inject a handcrafted config. sorry, can't hand this out. Nov 05 08:23:32 I think, maybe we should introduce algo="none" to our fitimage hash method config? Nov 05 08:23:50 only hint i can give is, we basically ripped out the image generation from kernel-fitimage, packed it in a custom class and replaced the config generation with a direct injection Nov 05 08:24:48 and that our custom config does not specify a hash/signature at all Nov 05 08:25:53 ok, got it. thanks! Nov 05 08:34:12 RP: ah yes, BBPATH, not BBLAYERS for classes "priority". It's just that the usual default for BBPATH in new layers is to append to it (at least that's what I've seen usually). So playing with prepending BBPATH is okay-ish for class "ordering"? Nov 05 08:36:39 letothe2nd: no sorry, I have IRC only at work ATM.. No didn't seem Khem unfortunately. I didn't eat with yocto folks on Friday lunch and couldn't see him at breaks and after the OEDEM track (I was in the other room) Nov 05 08:39:38 marler8997: IIRC, RDEPENDS for native packages should be working a little better since one or two releases ago. I moved recently on thud and AFAICT, that still does not work nicely (but it's already EOL so try warrior or zeus) Nov 05 08:42:06 RP: letothe2nd: (sorry to highlight you, I don't know the correct people to contact) ssl certificate expired on yoctoproject.org today Nov 05 08:42:25 qschulz: just pinged you because khem was around yesterday evening Nov 05 08:42:50 halstead: got this : (sorry to highlight you, I don't know the correct people to contact) ssl certificate expired on yoctoproject.org today Nov 05 08:43:06 but he'll not be around until in a few hours, i think Nov 05 08:43:59 letothe2nd: *facepalm* I think I'll never have a chance to catch him at the same time. I might write a mail so I don't flood the chan with a context and it's more async-friendly than IRC Nov 05 08:45:24 letothe2nd: thanks for the ping though, appreciated :) Nov 05 08:52:26 qschulz: we need halstead Nov 05 09:32:43 how do you go about tweaking the package sources without the whole push stuff to repo and clean and rebuild cycle? Nov 05 09:33:22 varjag: devtool modify Nov 05 09:34:05 varjag: basically thats just comfortable magic for externalsrc etc, but it works pretty well Nov 05 09:34:11 I need to patch a systemd service file of a package. what is the best way of doing this? Nov 05 09:34:33 oh.. had no idea devtool is a thing Nov 05 09:34:34 thanks! Nov 05 09:34:45 rcrudo: patch, in terms of? Nov 05 09:35:41 letothe2nd: I want to change redis.service file, adding an extra line e.g.:ExecStartPost Nov 05 09:37:06 rcrudo: depends a bit on your exact setup, but the two classic ways would be to either make a .patch and add it to SRC_URI, or make the manipulation manually in do_install, somehing echo, cat, sed Nov 05 09:40:04 depending on the exact usecase though, a second service file that has an After=redis.service also is an option. Nov 05 09:40:10 letothe2nd: I've tried the first option but it can't find the patch file. maybe because it is not part of the source code? Nov 05 09:40:48 rcrudo: no you just have to put it in a directory next to the recipe that is either called files or the recipe name Nov 05 09:41:01 rcrudo: just look at some layer, there are plenty of examples Nov 05 09:42:23 letothe2nd: sorry, that wasn't a precise explanation. I meant the target patch. I get the error: "No file to patch. Skipping patch." Nov 05 09:42:59 rcrudo: so the original redis.service is not even in the source package but comes from somewhere else? Nov 05 09:44:02 letothe2nd: is this one: https://github.com/openembedded/meta-openembedded/tree/master/meta-oe/recipes-extended/redis. it service files is not part of the source code but an extra file Nov 05 09:46:14 rcrudo: in that case, make an append that sets the FILEEXTRASPATH with prepending and just brings your modifed service file Nov 05 09:59:53 letothe2nd: where can I find an example? Nov 05 10:02:02 rcrudo: its a little more, but quite close: https://github.com/Pelagicore/meta-pelux/tree/master/recipes-swupdate/swupdate Nov 05 10:04:28 Hi. Just want to share my experience with the current yocto documentation: Following this guide https://www.yoctoproject.org/docs/3.0/brief-yoctoprojectqs/brief-yoctoprojectqs.html (particular checkout tags/yocto-3.0) resulted in error at package "icu" (https://pastebin.com/im2277qS) Nov 05 10:05:01 However, if i switch to current HEAD of "zeus", building package icu works. Nov 05 10:05:25 ThomasD13: can you file it on bugzilla, please? Nov 05 10:05:51 ThomasD13: so this feedback doesn't get lost, and yes, there is really somebody looking into it. Nov 05 10:05:52 However, my git console skillz are not the best, so I don't know which changeset resolves the error for me Nov 05 10:06:15 doesn't need a bug as its fixed in zeus head Nov 05 10:06:26 python3: -dev should depend on distutils Nov 05 10:06:36 rburton: ah, great! Nov 05 10:06:40 ThomasD13: ignore me then please! Nov 05 10:06:48 letothe2nd okay. Nov 05 10:07:10 ThomasD13: recommend you either just build zeus head, or cherry-pick that fix if you don't want to sit on an arbitrary sha Nov 05 10:07:12 rburton how do find the resolving commit so fast?? Nov 05 10:07:52 ThomasD13: becuase there's <20 commits between 3.0 and zeus head, the error was 'distutils', and i wrote that patch Nov 05 10:08:01 rburton: here's your chance to make somebody believe you are superhuman Nov 05 10:08:06 ah, you already gave it away Nov 05 10:08:26 tbh a bit surprised that fixes it directly, i thought the fix was installing python3-distutils on the host. Nov 05 10:08:42 it must be doing something fun with native python Nov 05 10:08:52 lol. Okay, thanks rburton for your work! Maybe someone should introduce a new tag and refer to the official documentation Nov 05 10:09:04 ThomasD13: there will be a new point release soonish Nov 05 10:09:12 alright, thanks guys! Nov 05 10:25:02 rburton: is this related? https://stackoverflow.com/questions/58703657/yocto-bitbake-core-image-sato Nov 05 10:25:26 no Nov 05 10:25:49 kay Nov 05 10:49:46 is it just me or is the certificate used by https://www.yoctoproject.org/ expired? Nov 05 10:50:05 T_UNIX: already reported Nov 05 10:50:22 once $ADMIN is wake, will hopefully take care Nov 05 10:50:27 :-D Nov 05 10:53:30 letothe2nd: I hope they can sync the expiration time to a PST-compatible one :D At the same time good excuse for europeans to not work... i'm torn Nov 05 11:02:28 kanavin: I fail to understand what find-critical-path is for (and I'm having a hard time compiling it on Ubuntu 18.04). Care to explain a little bit? Nov 05 11:21:20 qschulz, it is to determine which tasks in your build are on the critical path, e.g. if each of these tasks takes longer to complete, then the whole build will take longer Nov 05 11:22:25 it gives you a starting point where you can optimise the build times, either by trying to make these tasks run quicker, or break dependency chains between tasks on the critical path Nov 05 11:23:15 there's an article https://en.wikipedia.org/wiki/Critical_path_method Nov 05 11:38:42 kanavin: nice find with vnc! Nov 05 11:38:50 kanavin: we'll get the workers switched over... Nov 05 12:26:27 kanavin: cool! thanks for taking the time to explain Nov 05 12:31:08 Are you aware of that the SSL certificate for www.yoctoproject.org has expired? Nov 05 12:31:25 Azoff: yup, thanks :) Nov 05 12:31:37 Ok. Nov 05 12:31:46 waiting for admin to wake up :) Nov 05 12:31:48 Then I climb back into the shadows again ;) Nov 05 12:37:02 RP: cheers :) also it was rather awesome to run qemu inside qemu, and retain kvm acceleration even in the innermost guest OS :) Nov 05 12:44:28 kanavin: that is rather neat :) Nov 05 12:58:21 RP: it's also not the first time Debian fails to suggest a preferred option from many alternatives Nov 05 12:58:43 they seem to have a tendency to package everything under the sun and then leave the users to decide for themselves Nov 05 13:11:59 Hmm, I think we have a pseudo syscall issue on f30 Nov 05 13:15:47 kanavin: we mainly use the older vnc as we're still using relics of the original autobuilder :/ Nov 05 13:16:13 seebs: would pseudo have a problem with statx() calls? Nov 05 13:20:31 * RP is bemused by the arguments about statx() time precision and lots of "I'm not a physicist but" comments Nov 05 13:22:25 RP, link ? Nov 05 13:24:42 kroon: https://lwn.net/Articles/707602/ Nov 05 13:25:10 * RP *is* a physicist Nov 05 13:26:46 RP: something about Linux just draws the physics types in Nov 05 13:54:36 adelcast: i seem to have hit an edge case where opkg-build does sort dirs in data.tar.gz in the wrong order, resulting in non-reproducible builds Nov 05 13:55:48 i don't know what triggers it exactly yet, i'm trying to reproduce ipk generation from git and from a (manually, via makefile) derived source release tarball Nov 05 13:56:33 semantically, opkg-build is pointed to the same target dir, and the same tar is used (same machine, just different paths Nov 05 13:57:18 Piraty: same opkg-build cmdline? Nov 05 13:57:28 i debugged a bit, the tsortargs="--sort=name" seem to bit in both cases, Nov 05 13:57:38 * Piraty double-checks Nov 05 13:57:56 kroon: You mentioned reproducible ext4 images, are you still looking at that? Nov 05 13:59:08 ah that reminds me Nov 05 13:59:10 adelcast: came Nov 05 13:59:18 *same Nov 05 13:59:32 adelcast: basically "opkg-build -o root -g root ./build/foo" Nov 05 13:59:54 RP: JPEW: ross/pod is my WIP for the pod2man issue Nov 05 14:00:04 iirc it didn't quite wipe out 100% of the issues yet Nov 05 14:00:16 adelcast: "tar -c build/foo | md5sum" in both places is equal Nov 05 14:00:37 i think the remaining instances were recipes using the host perl pod modules directly Nov 05 14:01:03 which is annoying and suggests a do_package postprocess might be better Nov 05 14:01:47 adelcast: i mean $(tar --sort=name -c build/foo| md5sum ) is equal in both places, but ineed i found they're different if i omit --sort Nov 05 14:02:06 rburton: That is annoying Nov 05 14:03:55 maybe file_list somehow messes it up with tar? Nov 05 14:04:01 How can I tell if my bbappend is actually getting included as source when a recipe gets built? Nov 05 14:04:37 ecdhe: bitbake-layers has a subcommand to do that Nov 05 14:04:41 I've got a .bbappend that's supposed to add some kernel config options, but I clean and build the kernel and the config options don't take effect Nov 05 14:05:35 adelcast: i guess i can fix my own source tarball derivation to use --sort=name, but that might worth a double-check. maybe i get to it Nov 05 14:06:13 rburton: great, I can see that my bbappend file is in `bitbake-layers show-appends' Nov 05 14:07:50 rburton: thanks! Nov 05 14:08:16 Hi just a short question: What is the proper way to add a init.d script to an image? Just manally deploy the script and create symbolic links from rc.X to it? Nov 05 14:08:51 ThomasD13: write a recipe to deploy the init script, using the proper classes to create the links etc Nov 05 14:17:38 rburton: that is annoying :( Nov 05 14:18:04 rburton: how common is perl pod usage? Nov 05 14:22:21 RP: very common in perl recipes Nov 05 14:29:22 RP, I've never heard of `statx`, so probably. Nov 05 14:32:56 seebs: I'm pretty sure this is why pseudo is breaking on fedora30, ls uses it Nov 05 14:38:12 RP: from inside tinfoil can i tell bitbake to start builds Nov 05 14:38:26 rburton: yes, build_target Nov 05 14:38:29 woot Nov 05 14:38:31 rburton: was debugging that the other day Nov 05 15:03:18 seebs: I'm now quite sure statx() is the problem. Is this something you'd be able to look at or should I aim to see if I can break it? :) Nov 05 15:04:12 ggahhhh why would they do this to me why why why Nov 05 15:04:19 i can probably fix it, looks like my host has it Nov 05 15:04:37 maybe Nov 05 15:04:55 it looks like my host has no glibc wrapper for it. can you check the man page on fedora30 and see whether that's still true? Nov 05 15:05:20 what. why. Nov 05 15:05:27 Note: for performance and simplicity reasons, different fields in the statx structure may con‐ Nov 05 15:05:27 tain state information from different moments during the execution of the system call. For Nov 05 15:05:27 example, if stx_mode or stx_uid is changed by another process by calling chmod(2) or chown(2), Nov 05 15:05:27 stat() might return the old stx_mode together with the new stx_uid, or the old stx_uid together Nov 05 15:05:27 with the new stx_mode. Nov 05 15:05:44 okay i would like to state for the record that this call should never have existed, it should not exist in this form, and no one should be using it. Nov 05 15:05:47 this is *insane* Nov 05 15:06:23 seebs: I looked at the glibc source and there is a wrapper for it, and a generic fallback function Nov 05 15:07:28 seebs: so it looks like you need a system with recent glibc and coreutils which f30 seems to have Nov 05 15:08:06 seebs: and yes, that does seem rather crazy but I can see why they'd try this for "performance" Nov 05 15:09:20 hmm. i'm a bit unsure what semantics we even *want*. i think probably just "use one of our existing stats, also use statx, combine them" Nov 05 15:10:03 seebs: I was thinking just blast our existing stat data into the statx return value? Nov 05 15:11:22 Hmm. you know, I haven't got a way to *test* this handy right now -- if you wanna do a test implementation I can review it. Looking more closely, we might have another loose end I wasn't thinking about, BTW. Nov 05 15:11:42 seebs: its a bit like a filesystem uncertainty principle. You can know its mode or its owner but not both at the same time with certainty :/ Nov 05 15:11:47 *at now have an AT_NO_AUTOMOUNT that we weren't aware of, although I think I mostly checked specifically for AT_SYMLINK_NOFOLLOW. Nov 05 15:12:37 seebs: I'd hope NO_AUTOMOUNT isn't particularly relevant to us Nov 05 15:12:59 seebs: I can test relatively easily FWIW, its just the best implementation I'm slightly struggling with Nov 05 15:17:32 seebs, I have an Ubuntu 18.04 you an log into.. if you need access just let me know Nov 05 15:18:46 (48 thread , 128 GB machine.. so it should be able to iterate pretty quickly) Nov 05 15:21:14 fray: is it build box comparison day again? Nov 05 15:22:10 * letothe2nd still needs to write up the new livecoding dates and topics Nov 05 15:24:15 the box is old.. 2014.. Nov 05 15:24:18 so nothing too great Nov 05 15:24:26 but it's available.. Nov 05 15:24:34 seebs: we can get you access to the f30 machine we see the issues on, just not right away as michael is flying atm Nov 05 15:24:50 fray: just pulling your leg :) Nov 05 15:25:04 ...now that 192 thread machine ... that was work bragging.. ) Nov 05 15:26:17 hrhr Nov 05 15:30:28 I have ubuntu 18.04, which still claims to have no glibc wrappers. Nov 05 15:30:31 for statx, at least. Nov 05 15:37:36 ohh ok Nov 05 15:37:49 I can setup an F30 container for you if you want Nov 05 15:40:40 Hello! I am trying Yocto for the first time in a while. I am using Fedora 31. I use Poky 3.0 as explained in https://www.yoctoproject.org/docs/3.0/brief-yoctoprojectqs/brief-yoctoprojectqs.html. I can not source oe-init-build-env properly: Nov 05 15:40:52 Error: The bitbake directory (/home/julien/poky/on) does not exist! Please ensure a copy of bitbake exists at this location or specify an alternative path on the command line Nov 05 15:41:44 sounds like you passed 'on' as the second argument to oe-init-build-env Nov 05 15:41:55 hence /poky/on Nov 05 15:43:36 I did not... I also tried source oe-init-build-env bitbake Nov 05 15:44:02 that is wrong.. Nov 05 15:44:08 no, its really source oe-init-build-env without anything else Nov 05 15:44:11 bitbake should not be the argument as it can't be the build directory Nov 05 15:44:28 ./oe.... ../build ../bitbake :) Nov 05 15:44:29 AS letothe2nd said, do that first.. it will create a new 'build' directory Nov 05 15:44:33 jgrossholtz_: go checkout your livecoding sessions where the whole process can be seen in action :) Nov 05 15:44:43 s/your/our/ Nov 05 15:44:45 nesting bitbake inside anothe rgit repo is silly :) Nov 05 15:45:57 BTW, yoctoproject.org certificate expired on 11/4/2019 : Error code: SEC_ERROR_EXPIRED_CERTIFICATE . Who can renew it? Nov 05 15:45:57 jgrossholtz_: https://www.youtube.com/watch?v=EfKLrSxA_H8 Nov 05 15:45:57 vmeson: we know and already pinged halstead Nov 05 15:45:57 letothe2nd: thx Nov 05 15:45:57 halstead: so here's a lot of people screaming for hot new certificates! Nov 05 15:46:07 Ok thanks Nov 05 16:00:26 YPTM: Randy joined. Nov 05 16:00:58 YPTM: Richard joined Nov 05 16:01:05 YPTM: Joshua Watt, Tim Mertz, and Dustin Bain here Nov 05 16:03:35 YPTM: Denys is here Nov 05 16:03:50 YPTM: Scott joined Nov 05 16:04:03 YPTM: sjolley is here Nov 05 16:04:23 YPTML dreyna is here Nov 05 16:04:32 YPTM fray is here Nov 05 16:05:38 YPTM: Trevor joined Nov 05 16:05:56 YPTM: Jon joined Nov 05 16:05:58 Propsed 3.1 dates Nov 05 16:06:03 https://docs.google.com/document/d/1UKZIGe88-eq3-pOPtkAvFAegbQDzhy_f4ye64yjnABc/edit Nov 05 16:06:32 YPTM: letothe2nd is not around but listening to metal Nov 05 16:06:35 Nov 05 16:09:08 YPTM: jsmoeller joined Nov 05 16:30:16 What is the best way to see what configs the kernel build is using and how it is ordering and applying the changes? So far I haven't hand much luck with logs and debug output. Nov 05 16:31:35 bitbake-layers show-appends |grep kernel ? Nov 05 16:32:16 it would show the layers that append to the kernel and the ordering would be the layers priority Nov 05 16:32:18 I guess Nov 05 16:33:58 keyterina, I checked that and confirmed my append is on the list. Nov 05 16:37:18 sorry, that's all I know, I only use a defconfig and assume the order is the layers' Nov 05 16:40:07 d_thomas: is it ONLY a full defconfig you're replacing or are you doing something else? If you want to know in which order the files are taken, check with bitbake -e virtual/kernel and look for FILEPATHS (or ~), this is going to give you all the paths that are checked Nov 05 16:40:18 I think WORKDIR/temp/log.do_fetch should do the same Nov 05 16:41:12 I was attempting to append/override part of defconfig with my own configuration Nov 05 16:41:37 Right now I was just trying to change the LOGLEVELs Nov 05 16:42:59 log.do_fetch is showing what I would expect Nov 05 16:43:00 are you using config fragments then? Nov 05 16:44:12 kayterina, good question. I suspect that's what I want but I might not be using them Nov 05 16:45:44 a,ok, I don't either, search the mega manual while you wait for schultz :P Nov 05 16:46:26 just knowing what to search for sometimes is very helpful Nov 05 16:47:11 d_thomas: also, the certificate expired, they know it Nov 05 16:47:25 Link to all YP Summit presentations: https://wiki.yoctoproject.org/wiki/YP_Summit_Lyon_2019 Nov 05 16:47:56 ha, no worries. I won't say anything about it Nov 05 16:48:09 NET::ERR_CERT_DATE_INVALID error on YP website Nov 05 16:48:18 looks like it just expired Nov 05 16:48:48 anyone else seeing it? Nov 05 16:48:49 yes. expired recently Nov 05 16:48:59 When Michael gets online he'll fix it Nov 05 16:50:11 who let him take any days off? ;-) Nov 05 16:50:22 on plane Nov 05 16:51:29 BuT tHeRe Is WiFi In PlAnEs NoW Nov 05 16:52:18 d_thomas: Please explain exactly what you're trying to do and what you've done so far that does not work because I'm quite confused at the moment Nov 05 16:53:02 config fragments aren't available for all recipes Nov 05 16:53:09 are you building linux-yocto? Nov 05 16:54:45 qschulz I'm working with meta-atmel and building the linux-at91 kernel. I'm trying to apply the configuration fragment from http://book.yoctoprojectbook.com/page/code-chapter-9 just as an example (although I'm also setting LOGLEVELS) Nov 05 16:55:40 d_thomas: only full defconfig supported Nov 05 16:56:06 Oh, okay. Where does one see that? Nov 05 16:56:08 what you're looking at is config fragment which are supported in linux-yocto Nov 05 16:56:45 d_thomas: assume default is full defconfig Nov 05 16:56:51 https://github.com/linux4sam/meta-atmel/blob/warrior/recipes-kernel/linux/linux-at91_4.19.bb#L20 Nov 05 16:57:34 qschulz, I see, that's what that means Nov 05 16:58:27 for config fragments it's in https://git.yoctoproject.org/cgit.cgi/poky/tree/meta/classes/kernel-yocto.bbclass#n27 Nov 05 16:58:35 this class is inherited by linux-yocto Nov 05 17:02:59 qschulz Wow, that's good to know. I'll go the full config route then. Thanks! Nov 05 17:04:03 The "Embedded Systems with the Yocto Project" book lead me to believe the config fragments were standard Nov 05 17:05:12 they are standard for any kernel that uses the linux-yocto class.. Nov 05 17:05:28 (and we suggest even non linux-yocto kernels use the config fragments as it makes it MUCH easier to support kernels..) Nov 05 17:05:43 Definitely Nov 05 17:28:22 a few others do support them. linux-qoriq and a few others. there's an open yocto bug to make it stanard Nov 05 17:28:24 standard, even Nov 05 17:30:01 there is documentation on how to enable it for all kernels.. it's very easy. basically inherit the class, and make sure that kernel doesn't overwrite the constructed defconfig.. (as well as define a KMETA parameter if I remember right) Nov 05 17:30:32 last time I looke dit was about 5 or so lines of setup and then some basic validation.. (and of course creation of config fragments if what you want wasn't previously known by the system) Nov 05 17:54:30 hey guys! I'm building a Krogoth-dated eSDK. I have this error when I try to install the result: "ERROR: Uninative selected but not configured, please set UNINATIVE_URL". No idea how to make it happy :-( Nov 05 17:58:41 within the target installation dir of eSDK I see ${BSPDIR} that spells just like that :-) Nov 05 18:00:09 I guess, I'd better get the Krogoth's tip then Nov 05 18:07:12 ... and it doesn't really work ;-( Nov 05 18:58:21 fray: You are very right about how easy it was to enable configuration fragments. I just had to include kernel-yocto and set KMACHINE in my bbappend file and it looks like it's working Nov 05 19:00:54 hate to beat a horse that might already be dead, but is someone going to fix the ssl certificates? Nov 05 19:02:05 mischief: its a known issue and will be addressed as soon as possible Nov 05 19:46:04 jonmason: excellent Nov 05 19:49:22 i hereby declare that for every time i have to excuse for the certificates, halstead owes me a beer. Nov 05 19:50:49 letothe2nd: I got 0 beers on Friday, and that's after you didn't show up at the original place for dinner Nov 05 19:50:59 you owe me all the beers! Nov 05 19:51:34 jonmason: you run off too quick so i couldn't redirect you and the new place had many beers. Nov 05 19:51:46 not my bad :) Nov 05 19:51:50 and yet you got me 0 Nov 05 19:51:52 ;-) Nov 05 19:51:59 it worked out Nov 05 19:52:15 hrhrhr Nov 05 19:52:31 * letothe2nd hi5es jonmason for beer places. Nov 05 19:52:55 you can make it up in dublin Nov 05 19:54:32 then why didn't you just grab one from the oedem gift table? Nov 05 20:08:16 He was being responsible Nov 05 20:08:19 not like me Nov 05 20:09:38 letothe2nd: I don't drink on the job, unlike other people Nov 05 20:10:01 I'm not sure where the other 3 went Nov 05 20:12:12 I only had one Nov 05 20:13:16 i know where they went :) Nov 05 20:15:20 jonmason: i do drink on the job. at least when it suits the situation. Nov 05 21:01:53 * RP thinks the cert may be fixed Nov 05 21:03:43 you are correct Nov 05 21:03:45 thanks folks Nov 05 21:36:28 I added a defconfig to my kernel build. I see the defconfig in the /tmp/work directory, but the .config file in the same directory is picking up different settings. Is there an order of operations to how config settings are applied? Nov 05 21:37:33 Note that the .config file is picking up config fragments from a bbappend file. It's only the defconfig specified in the bb file that are not being applied Nov 05 21:42:47 the generated defconfig is USUALLY done in a task called configme which is dependend on by 'configure' Nov 05 21:43:05 I've see some private kernels do things like cp defconfig in 'configure' which replaces the kernel-cache based .config Nov 05 21:44:52 I wondered about that, so I dumped the do_kernel_configme() function with the "-e" option. It's not doing that directly, but perhaps indirectly it is Nov 05 21:45:34 unfortunately this is where I've used it before, and understand the basics, but I've not dug into exactly what is happening.. Nov 05 21:47:37 The person asking this question (https://stackoverflow.com/questions/48441494/kbuild-defconfig-kmachine-defconfig-file-does-not-work-as-expected) posted an interesting work around, but I'd rather not go that route if the problem is just user error on my part Nov 05 21:48:54 I'm looking to see where it's copied in.. but I suspect it's in some helper we don't see in the code Nov 05 21:50:25 I think it's do_kernel_configcheck that actually moves the stuff to .config Nov 05 21:50:53 zeddii -- ping Nov 05 21:51:01 zeddii would know the answer BTW.. not sure if he's around Nov 05 22:10:55 Hmm... if I"m picking up do_configure from https://git.yoctoproject.org/cgit.cgi/poky/tree/meta/classes/kernel.bbclass, then I'm not sure how this works Nov 05 22:23:39 Actually, I can see how it works now and everything seems to be in place. Something must be going wrong with the KERNEL_CONFIG_COMMAND Nov 05 23:07:20 3 Nov 05 23:25:35 Is there a particular recipe that is known to be quick to build? Nov 05 23:26:23 As in, light on dependencies and short compile time? Nov 05 23:27:02 Just looking for something to test a local wrapper for bitbake with Nov 05 23:28:16 busybox ? Nov 05 23:28:39 I would suggest to look into oe_selftest might be more relevant Nov 05 23:28:57 will do, thanks Nov 05 23:54:05 is it just me or is lz4-native missing from the kernel classes? Nov 05 23:54:52 got a quick question of preference...if you have a recipe that just installs a bash script, would you set the source directory to S and install from B? Nov 05 23:56:37 or would not not set S at all? Nov 05 23:57:26 or maybe you would just set B and not set S? Nov 05 23:59:34 i think lz4 initramfs is just broken :( Nov 06 01:14:06 marler8997: install it directly from {WORKDIR} Nov 06 02:10:55 Hello, if I modified do_install() but didn't touch do_compile() at all, will do_compile() also be processed? Nov 06 02:26:16 from my local test, do_compile() won't be called. but sometimes it's called. I have no idea why Nov 06 02:38:00 d_thomas: you need to set KCONFIG_MODE to “alldefconfig” Nov 06 02:38:26 if it isn’t a *full* .config Nov 06 02:38:40 but I’d have to see the defconfig you are using to know which it actually is. **** ENDING LOGGING AT Wed Nov 06 03:00:06 2019