**** BEGIN LOGGING AT Thu Sep 28 03:00:02 2017 Sep 28 05:06:39 sgw: yes I see that problem on musl+go, this is a golang issue, they check for isAlpineLinux which actually should have been isMusl, but at this point, I think there is no solution done yet Sep 28 05:07:39 khem: so your not it? otavio? I am surprised that someone has not found this yet Sep 28 07:17:21 good morning Sep 28 07:18:58 mckoan: good morning Sep 28 07:56:17 hello guys, I have a quick question Sep 28 07:56:18 I have kernel foo for my project, and I want to build two versions, a foo_normal and a foo_fallback Sep 28 07:56:51 starting from the same recipe, one maybe with a "full" defconfig, the other with a minimal one (e.g. no modules and stripped of wifi and useless stuff) Sep 28 07:57:10 what is the best way to do this? so far I had two different recipes, but we can have only one "virtual/kernel" Sep 28 07:57:19 I mean, one preferred Sep 28 08:06:33 LocutusOfBorg: using override to specify a different defconfig for your fallback, maybe? Sep 28 08:10:02 but how can I build two kernels one for each rootfs? Sep 28 08:10:28 I could use two machines, two builds but it seems to be an overkill just to have two kernels Sep 28 08:10:54 right now the main kernel is a bbappend of the default imx6 one from meta-freescale Sep 28 08:11:11 so, I can't easily have to different bbappends, I will likely need to have a full recipe for the fallback one Sep 28 08:18:36 did you see that thread ? https://lists.yoctoproject.org/pipermail/yocto/2015-August/025943.html don't know if other solution came by this time Sep 28 08:30:54 nayfe, probably nothing changed in the meanwhile... I'll split in two machines probably Sep 28 08:52:04 LocutusOfBorg: maybe wait for an expert answer. Can't you use the same kernel and undeploy modules at rootfs processing ? or use kernel fragments? Sep 28 08:52:54 having 2MB kernel blob instead of a 6MB one is interesting Sep 28 08:53:09 and kernel fragments won't help because I will still need to do *two* kernel builds Sep 28 08:53:18 so, somewhat I have to split the kernel recipe in two Sep 28 10:03:19 hello Sep 28 10:03:36 is it possible for several users to use the same downloads cache directory in the same time? Sep 28 10:04:25 cornel: they'll race over writing the final file, but they won't corrupt Sep 28 10:05:37 hm, actually, we don't write to temporary file... Sep 28 10:05:52 Hey people, I have just started using yocto. Im working on an i.mx7 device (created custom hardware) and now i want to vrify that the hardware actually works. i think i could do that by creating a linux image with a custom device tree? Sep 28 10:06:50 rburton, what we see in practice that even if only one user uses its local downloads cache, but for several simultaneous builds, it may end in fetch() failures Sep 28 10:09:32 cornel: yeah, we tell wget to write directly to the final filename. should write to a temporary file and rename on complete. if you can replicate a patch should be fairly simple and you're able to test :) download() in bitbake/lib/bb/fetch2/wget.py. Sep 28 10:11:06 DemoniacMilk: thts the basic way once you've got your bootloader up and running, yes Sep 28 10:12:30 rburton, thanks , i'll try Sep 28 10:21:25 LetoThe2nd okay thanks. I dont have a bootloader running yet. I guess I can find info on this on the NXP websites. Sep 28 10:22:31 do i have to recompile the kernel after changing the device tree? Sep 28 10:22:38 or is this separated? Sep 28 10:23:28 DemoniacMilk: if you don't need any new drivers or patching, the just recompiling the DT is enough (for same kernel versions, that is) Sep 28 10:23:30 rburton, RP: so, the binutils recipe determines the ALTERNATIVES_TARGET by itself using TARGET_PREFIX - which means it's trying to set up ld -> arm-unknown-linux-gnueabi-ld links even though the rest of this recipe has successfully made arm-linux-gnueabihf-ld Sep 28 10:23:55 the problem is now I don't think I can just set ALTERNATIVES_TARGET because do_package seems to uncondidionally set it based on TARGET_PREFIX Sep 28 10:23:58 DemoniacMilk: probably its using u-boot as the loader, so look for that. Sep 28 10:24:17 can I, er, do a replace on the python before it's parsed or something? Sep 28 10:24:43 * ramcq is actually wondering if just adding support for the right triple to Yocto wouldn't have been quicker at this point Sep 28 10:28:24 also, I've added thumb to my DEFAULTTUNE - armv7athf-vfpv3d16 Sep 28 10:28:33 but it doesn't seem to be in my CC flags Sep 28 10:29:29 ramcq: RP is on holiday right now Sep 28 10:29:40 WHAT?! VERBOTEN! Sep 28 10:29:42 :) Sep 28 10:29:49 i was shocked too Sep 28 10:30:30 LetoThe2nd yes its uboot. Ill figure out, just wanted to make sure im not following an idea that doesnt work at all Sep 28 10:30:34 thank you Sep 28 10:30:41 DemoniacMilk: have fun Sep 28 10:30:43 ramcq: i've also spent the last five years ignoring the bintuils and gcc recipes... Sep 28 10:31:17 rburton: hopefully RP is back for some booze in prague :P Sep 28 10:31:28 yeah he is only away for a week Sep 28 10:31:38 phew! Sep 28 10:35:22 rburton: ok, but if I wanted to tweak http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/recipes-devtools/binutils/binutils.inc#n61 in a .bbappend how could I do it? can I make text changes on a python thing... ? :'( Sep 28 10:38:41 no, sorry Sep 28 10:39:08 you could add another prepend in a bbappend but that will happen *before* that fragment Sep 28 10:39:47 and an append will be too late Sep 28 10:46:20 hm, I think if I use the USE_ALTERNATIVES_FOR variable I can reimplement the append then empty that var Sep 28 10:46:23 let's try that Sep 28 10:46:39 I think I am going to bitbake hell Sep 28 10:48:07 lol Sep 28 10:48:25 when RP is back I suspect he's going to rewrite some of the toolchain recipes to be more flexible Sep 28 10:48:50 ramcq: https://www.youtube.com/watch?v=xuTXycjRx4U here you go Sep 28 10:52:29 right, jobbed it. now where is libbfd going... Sep 28 10:54:49 rburton: I'm not sure flexibility is helpful here, I would actually just split the target recipts to be Quite Different Sep 28 10:54:51 *recipes Sep 28 10:55:11 also expand armhf to the right triple and fix the fall-out Sep 28 11:41:55 rburton: ok my last (I think?) issue is I want to append to FILES_${PN} but only on one machine... the obvious FILES_${PN}_machine += "foo" appends to a blank FILES_${PN}_machine so clobbers FILES_${PN} Sep 28 11:42:30 but, FILES_${PN}_machine = "${FILES_${PN}} ..." is defined in terms of itself so explodes Sep 28 11:43:43 eh, fuck it I'll use python Sep 28 11:47:00 d.getVar("FILES_binutils", True) is None Sep 28 11:47:04 good. perfect. Sep 28 11:48:14 Hi, I did set INITRAMFS_FSTYPES ?= "cpio.xz" but I still get cpio.gz after running bitbake. What am I doing wrong? Sep 28 12:11:04 DemoniacMilk: Create a custom board BSP is well explained in Linux: Embedded development book, don't know if you read it ? Sep 28 12:12:06 rburton: Hi, did you read LocutusOfBorg problem? any advice (i'm interested too) Sep 28 12:13:16 ramcq: FILES_${PN}_append_machine Sep 28 12:13:53 rburton: ... why... what... my god... I thought _append was kind of like an operator so it went afterwards? Sep 28 12:14:23 ramcq: hah trying to impose logic on bitbake assignments, you fool Sep 28 12:14:35 _append and _remove go before overrides Sep 28 12:15:38 I've actually replaced it with some python which seems to run far enough after everything else that the variable seems to have deemed fit to appear... Sep 28 12:16:03 nayfe: no i havent Sep 28 12:16:07 but good hint Sep 28 12:17:41 theres quite some of those books, any specific recommendation Sep 28 12:22:31 DemoniacMilk: unfortunately most of them are rather outdated Sep 28 12:23:48 DemoniacMilk: https://www.amazon.de/Embedded-Systems-Prentice-Software-Development/dp/0133443248/ref=sr_1_5?ie=UTF8&qid=1506601364&sr=8-5&keywords=embedded+linux seems to be rather recent (2016) and the author is known to be yocto/OE competent Sep 28 12:23:53 but i haven't read it Sep 28 12:24:18 (sry for german amazon) Sep 28 12:25:29 well german amazon is fine for me Sep 28 12:25:59 thanks Sep 28 12:26:06 Transaction check reporting conflict. This means two recipes have listed the same file in FILES, right? Sep 28 12:27:36 How should one think about directories here. I see a conflict on a directory, but both packages are probably entitled to put files into that directory, so they both need to ensure it exists... Sep 28 12:29:12 https://www.packtpub.com/application-development/linux-embedded-development this one is really complete, i read quite some books, and this one really helped me. Sep 28 12:30:23 you can read it here https://www.safaribooksonline.com/public/free-trial/ Sep 28 12:30:28 Should components list all the specific *files* they are installing and not worry about mentioning the parent directory in FILES. Is it enough? Will the parent directory be created? I'm justa asking to understand the intended policy and best practice. Sep 28 12:35:01 DemoniacMilk: again i've not read it, but authors of the packt thing that nayfe mentioned are well-known and competent too. Sep 28 12:35:36 cant find that one in book version, mostly downloads or some subscition stuff Sep 28 12:38:14 DemoniacMilk: but the really best thing is to be part of a team that does the thing for one or two board. Sep 28 12:39:19 got no one in my company that had a closer look at linux ever Sep 28 12:41:42 DemoniacMilk: well you can either hire a consultant or spend a lot of time. or something in between, attend a training to get you started. Sep 28 12:42:20 DemoniacMilk: you can /join #nayfe if you want Sep 28 12:42:27 i got like 3 weeks :D Sep 28 12:42:38 DemoniacMilk: for? Sep 28 12:42:49 verifying my hardware Sep 28 12:43:10 im not into software at all, im a hardware developer Sep 28 12:43:27 but before i can say "yep, my hardware works" i need to know if my hardwre works Sep 28 12:43:43 so im looking for an easy way to test the peripherals of my controller Sep 28 12:43:50 DemoniacMilk: well if its close to something thats known good, like some dev kit and just modified some peripherals, then 3 weeks is totally ok if there are no serious bugs. Sep 28 12:43:59 Example: http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/recipes-devtools/cmake/cmake_3.7.2.bb?h=pyro#n44 Sep 28 12:44:07 it looks a bit lazy(?) to include the whole ${SDKPATHNATIVE} - would something more explicit: "${SDKPATHNATIVE}/environment.d/cmake.sh" be better practice? (and solve a conflict in this case) Sep 28 12:44:28 then its basically 1) grab bootloader 2) add new board 3) make boot Sep 28 12:44:51 ye tahts what i would want to do Sep 28 12:44:59 but im not sure how to do that^ Sep 28 12:45:07 but if you have done some hevy modifications, like a massive change in memories, boot order, etc, then 3 weeks might be tough without prior experience Sep 28 12:45:28 for now, its just an addon for an existing board Sep 28 12:45:32 memory etc is identical Sep 28 12:45:40 but a lot of the pinout has been changed Sep 28 12:46:05 well then you certainly need to modify the bootloader Sep 28 12:46:52 in a nutshell, its looking at the source, find all points where the original board is referenced, and add the new one there too. Sep 28 12:47:42 and freescale soc pinmuxing is really ... well ... Sep 28 12:48:57 i think most of the pin muxing is dealt with when i use the pinmux tool Sep 28 12:49:48 it allows to export the device tree as well iirc Sep 28 12:49:59 i guess if i knew more about all of this, itd be rather easy from there Sep 28 12:50:10 jsut that the bootloader won't use dt Sep 28 12:51:51 DemoniacMilk: this seems to be some basic imx7 board support in u-boot https://github.com/u-boot/u-boot/blob/master/board/toradex/colibri_imx7/colibri_imx7.c Sep 28 12:52:45 DemoniacMilk: with this being the initial patch https://github.com/u-boot/u-boot/commit/ae440ab02d31179a5a4b23e7411fe1baf052e816#diff-1e69da03e73dc4f08e311a3fdd24265f Sep 28 12:53:49 maybe you need to use https://github.com/Freescale/u-boot-fslc ? Sep 28 12:54:34 havent used git before either Sep 28 12:54:38 or make or anything really :D Sep 28 12:55:22 DemoniacMilk: so your software know how for ths board actually consists of finding this very channel? Sep 28 12:55:33 then you can multiply your 3 weeks by 10 :) Sep 28 12:55:35 thats kind of it Sep 28 12:56:02 well not 3*10, but then 3 weeks sounds like very, very improbably to hit. Sep 28 12:56:02 as i said im a hardware developer Sep 28 12:56:17 DemoniacMilk: advice: go find a software developer Sep 28 12:56:33 probably a good idea Sep 28 12:56:41 but we have none in the company Sep 28 12:56:53 wont change in the next 3 weeks :D Sep 28 12:56:55 DemoniacMilk: hire one, there's a lot of consultants. Sep 28 12:57:26 well we develop the hardware for software engineers Sep 28 12:57:36 maybe they can help Sep 28 12:57:51 rburton: you have time? looking for some opinion on the above, if you have time. Sep 28 12:57:57 DemoniacMilk: to give you some comparison, i usually need a1 or 2 weeks to bring up and test a board thats closely related to something i already have and know, with some minor to medium modifications Sep 28 12:58:20 * gunnarx can't figure out if LetoThe2nd and DemoniacMilk are discussing yocto issues or just discussing... Sep 28 12:58:34 gunnarx: on what? Sep 28 12:58:44 hence my comment, it's kinda lost in the noise Sep 28 12:58:57 gunnarx: you're right, sorry. Sep 28 12:59:21 about one screen back, if you can find rburton, otherwise I'm happy to repeat Sep 28 12:59:23 ill see if ican maybe get that prcessor programmed in bare metal then and skip yocto/linux. thanks for the heads up Sep 28 12:59:41 DemoniacMilk: only one last sentence: tell your boss you need to ask for help right now. not in three weeks. Sep 28 13:00:36 gunnarx: what recipes are reporting conflicts with what files? Sep 28 13:01:51 rburton: yes, the recipe should probably use a fuller path just for clarity. Sep 28 13:01:51 in this case cmake and qtbase, but it's as I say, both probably have the right to install a file in environment.d. So I'm asking in general - 1. should FILES include directories or is it enough to just list files (are parent dirs created?) 2. How to handle if two recipes want ot install different files into the same dir. They both want to ensure the dir is there. Sep 28 13:02:05 its actually warning about environment.d itself? Sep 28 13:02:08 yes Sep 28 13:02:14 *interesting* Sep 28 13:02:21 not expected? :) Sep 28 13:02:29 different permissions or something in the rpm? obviously two packages can and do install /usr/bin :) Sep 28 13:02:50 what files are they dropping in? don't have qt here Sep 28 13:02:55 I was kind of starting to wonder if transaction check should complain about directories... Sep 28 13:03:07 rpm presumably? Sep 28 13:03:08 cmake is creating cmake.sh and qt is creating qt5.sh IIRC Sep 28 13:04:04 I'm currently trying out to have cmake reference the _file_ and not just the dir. Listing the dir seems a sloppy way of saying "just take everything that I put in this dir". Would we agree that is not best practice? Sep 28 13:04:25 trying to both fix a specific problem, and figure out policy/best practice here Sep 28 13:04:57 btw, it's working on nativesdk targets if it matters Sep 28 13:05:18 nativesdk task perhaps better worded Sep 28 13:07:23 just the directory should work though Sep 28 13:07:34 so i wonder if the permissions don't match and that makes rpm barf Sep 28 13:07:36 even if more than one component lists it? Sep 28 13:07:55 so if permissions are fine, more than one component *should* be able to list it in FILES without transaction check conflict? Sep 28 13:07:58 every time you use ${bindir} in FILES, as is default, it uses /usr/bin/ Sep 28 13:08:15 yes, my hunch is the perms are different which is why rpm is aborting. Sep 28 13:08:29 unpack the rpms manually to see the contents? Sep 28 13:09:00 agreed, but do recipes list ${bindir} or do they list ${bindir}/myfile... Sep 28 13:11:28 ok, not sure about permissions. the error msg only says there's conflict that both components want to install "environment.d" but the message might be misleading Sep 28 13:12:29 alternative question: if the explicit files are listed would rpm create the parent dir automatically or do you _have_ to list the directory before the file in FILES? Sep 28 13:13:22 FILES is a mix, its ${bindir}/*, but ${sysconfdir} Sep 28 13:13:22 parent folders are implied Sep 28 13:17:38 gunnarx: poky/scripts/rpm2cpio.sh |cpio -t -v Sep 28 13:18:15 OK, give me a sec I'm not set up to run it at the moment. Sep 28 13:19:45 doh, i just built cmake but this is nativesdk-cmake specific Sep 28 13:20:24 yeah, I did mention though :) Sep 28 13:20:28 https://github.com/meta-qt5/meta-qt5/blob/1100037b9becaaa5749602bca9d63693119c4585/recipes-qt/qt5/nativesdk-qtbase_git.bb#L71 and line 232 If it matters, they have a custom function in there... Sep 28 13:22:07 Sorry I'm not sure what commit that was, not the one we use, but it looks the same on their pyro branch https://github.com/meta-qt5/meta-qt5/blob/pyro/recipes-qt/qt5/nativesdk-qtbase_git.bb Sep 28 13:24:15 ceres Sep 28 13:24:36 (sorry, mistyped) Sep 28 13:29:31 rburton, I'll ping you again when I checked the rpms, I goofed and ran the fresh build script that starts by wiping tmp... Sep 28 13:33:01 sstate to the rescue right Sep 28 13:52:17 ya, I have the sstate fortunately :) Sep 28 13:53:02 anybody have been facing to unexpected errors using a multi user build environment with a shared SSTATE_DIR? Sep 28 13:53:17 so far looking like no diff in ownership or permissions on environment-setup.d, but let me double check that I haven't accidentally "fixed" it in some other way Sep 28 14:08:48 hi everyone. does anyone know why swupdate requires u-boot-fw-utils specifically? i'm trying to build on a platform that uses u-boot-ezynq and have dependency problems Sep 28 14:09:37 peacememories: guess because it can read and modify u-boot environments. Sep 28 14:11:00 anyone around with knowledge of the go toolchain in oe-core? I think I've made the front fall off it Sep 28 14:11:27 probably, question is... is there a solution for this? should i add the features of u-boot-fw-utils to a bbappend for u-boot-ezynq and make it provide the former? or can i somehow extend a bsp config and change the preferred u-boot and see how that goes? Sep 28 14:23:18 can i maybe stop u-boot-fw-utils from being skipped? Sep 28 14:23:45 paulbarker: you want otavio or khem Sep 28 14:23:58 rburton: Sep 28 14:24:04 https://github.com/flatpak/freedesktop-sdk-base/pull/10 Sep 28 14:24:07 this better work :P Sep 28 14:24:20 ramcq: dare i look? :) Sep 28 14:24:21 I don't quite understand the thumb handling Sep 28 14:24:46 it seems setting "thumb" tune doesn't actually pass -mthumb unless you set ARM_INSTRUCTION_WHATEVER to thumb Sep 28 14:25:55 yeah i *think* thats because setting the thumb tune just makes it available and some upstreams just fail to build with thumb enabled. you can set ARM_WHATEVER to thumb globally and you'll see some recipes unset it. Sep 28 14:26:38 but do linux distros typically enable thumb everywhere? how does gcc choose when to use thumb? Sep 28 14:31:27 the thumb tune says 'my cpu supports thumb'. I know this may be a bit strange, but there are a LOT of things that won't build with thumb mode Sep 28 14:31:48 (that list dropped for a while, but now that everyone is moving to armv8, the list is starting to grow again..) Sep 28 14:32:14 fray: i am officially not moving to armv8! Sep 28 14:32:36 ok, move to mips64.. :) Sep 28 14:33:03 hey c'mon, we're just leaving arm7tdmi (a.k. armv4) Sep 28 14:33:17 (32-bit arm still has a long life ahead of it.. but even 32-bit arm is moving the 32-bit version of armv8, I forget what it's called) Sep 28 14:33:34 * fray waits for 2038 and 32-bit timestamps.... Sep 28 14:36:48 hmm, i've found RDEPENDS_${PN}_remove, but DEPENDS_${PN}_remove doesn't seem to work? Sep 28 14:37:10 because DEPENDS_${PN} doesn't exist Sep 28 14:37:15 oh Sep 28 14:37:16 oops Sep 28 14:37:24 does DEPENDS_remove work? Sep 28 14:37:26 yes Sep 28 14:37:32 _remove works on all variables Sep 28 14:37:47 thanks Sep 28 14:38:00 Can I pass CFLAGS="-g -O0" to every recipe in my image build? Sep 28 14:38:55 I have a bug I'm tracing through a nest of shared libraries and -O2 (the common default) causes gdb to jump around too much. Sep 28 14:39:07 sr105: set DEBUG_BUILD=1 Sep 28 14:39:12 thanks! Sep 28 14:39:16 (in local.conf) Sep 28 14:43:19 where does -c menuconfig actually store its configuration?^^ Sep 28 14:43:26 armpit: which branch contains your pending morty changes? Sep 28 14:43:52 akuster/morty-next on contrib Sep 28 14:43:55 otavio: did you see the golang bug I assigned to you this morning? Sep 28 14:43:58 armpit: thanks Sep 28 14:48:09 i've found a nice utility i want to have included in my rootfs, devregs: https://github.com/FrankBau/meta-marsboard-bsp/wiki/iMX6-Register-Access Sep 28 14:48:54 instead of bitbaking the recipe directly, how do i get it to be included as part of my image recipe? Sep 28 14:49:01 add it to IMAGE_INSTALL Sep 28 14:50:00 rburton: cool. thanks. Sep 28 14:53:47 fray: so... Debian builds their GCC with --with-mode=thumb Sep 28 14:54:03 fray: does that just mean -mthumb is available, or does it enable it by default? Sep 28 14:57:36 hmm, how can i remove modifications i made via menuconfig? Sep 28 15:03:10 sgw, I am in meeting of and on today if you have any questions, there may be a delay Sep 28 15:04:58 Hello everyone, I was wondering if you can help me with this issue. Sep 28 15:04:58 Apparently nothing that needs https works from our built image. It seems that the problem is caused by the SSL_CERTIFICATE. On this topic I had a look at this conversation Sep 28 15:04:58 on the mailing list: https://lists.yoctoproject.org/pipermail/yocto/2015-July/025774.html Sep 28 15:04:58 On the openssl recipe, it says that to add the c_rehash function PKGCONFIG needs perl. I tried then adding this entry PACKAGECONFIG_append_pn-openssl = " perl" in local.conf, but still the problem isn't solved and Sep 28 15:04:59 there is no sign of the c_rehash function needed to hash the certificates.. Sep 28 15:08:09 has any of you experienced the same problem?? Sep 28 15:20:31 frank_878: did you install ca-certificates into the image? Sep 28 15:25:08 yes Sep 28 15:25:19 rburton, yes ofcourse Sep 28 15:27:10 -c cleansstate Sep 28 16:21:09 fatal: reference is not a tree: 8a93832dad621535e90aa8e1fb74ae5ba743fc3e Sep 28 16:21:09 ERROR: ptest-runner-2.1+gitAUTOINC+8a93832dad-r0 do_unpack: Function failed: base_do_unpack Sep 28 16:23:12 http://git.yoctoproject.org/cgit.cgi/ptest-runner2/commit/?id=8a93832dad621535e90aa8e1fb74ae5ba743fc3e <-- yes it is Sep 28 16:23:59 we're using a shared downloads dir on a common server, and apparnetly i don't have write access to it. but why didn't the git fetch fail earlier, rather than failing in unpack, if it wasn't able to update the repo? Sep 28 16:24:01 very odd Sep 28 16:24:50 hmm Sep 28 16:26:28 if i manually run git fetch, it errors.. so how did do_fetch succeed.. Sep 28 16:26:40 lack of error checking, i'd guess Sep 28 16:26:48 cough fetch3 cough Sep 28 16:27:01 * rburton runs for the hills Sep 28 16:27:07 shouldn't be possible, runfetchcmd() is supposed to raise an exception on any command failure.. guess we need another unit test Sep 28 16:27:16 aside: i rather hate runfetchcmd Sep 28 16:27:36 you verified git-fetch actually does set the exit code? Sep 28 16:27:48 yep, returns 255 Sep 28 16:40:48 rburton: so, want me to ste DISTRO_FEAT_NATIVE for ipv6 instead? Sep 28 16:50:14 Heyho Sep 28 16:51:13 If I manually try to install an ipk package with opkg I alway get "nothing provides libc6 >= 2.25" error. Dont know how to fix this?! Sep 28 16:53:52 it gets correctly installed with --force-depends Sep 28 16:54:06 However, I want to get rid of that error. Any ideas? Sep 28 16:55:08 Hi all, i was wondering if the following workflow is possible. I use devtool for editing/mainting code, however every once and a while i want to debug a specific program. I would like to use Eclipse for this, however, i can not find documentation on how to mix devtool and eclipse. Are there people that use this workflow? Sep 28 17:06:50 Hello. It seems like 'EXTRA_IMAGE_FEATURES += "read-only-rootfs"' sets a password for root. Any idea what the password might be? Sep 28 17:07:40 Or maye it even disables root login and makes a new user. All I can conclude is that without read-only-rootfs, I can login just fine (root with no pass), and with ro, I can't login Sep 28 17:09:51 malinus: most likely it isn't setting a password at all, but you have a EXTRA_IMAGE_FEATURES ?= farther down in your lcoal.conf which is no longer applied once you assigned it a value Sep 28 17:09:57 malinus: resultling in the default features not being included Sep 28 17:10:37 kergoth: DOH! thanks! Sep 28 17:11:47 ?= is 'set only if it's current unset' Sep 28 17:11:58 same syntax as ?= in makefiles Sep 28 17:12:37 yeah I just missed the EXTRA_IMAGE_FEATURES further down Sep 28 17:15:49 * kergoth nods Sep 28 17:15:50 np Sep 28 17:19:02 thanks again, I'm still learning my way around yocto. But it sure seems a better aproach than doing your own small debian based distro. Sep 28 17:55:05 I have to split OS layer and individual applications for update purpose and that the production systems can contain different constelations of applications. What is the best way to do so? Use OStree as pseudo overlayfs for applications? Sep 28 17:59:41 sgw: I did; I didn't yet look at the issue Sep 28 18:00:00 Is someone using package feeds with dnf? Sep 28 18:00:29 otavio: it pretty much breaks all x86-64 platforms with musl Sep 28 18:01:36 sgw: so nice! Sep 28 18:02:22 khem: you mentioned it checks for Alpine. How is it related? Sep 28 18:04:09 otavio: I did not look at the code, but I think he mentions something about maybe doing IsMusl first or instead of the Alpine check Sep 28 18:04:34 sgw: I'll dig on it later today; it is reproducible with qemu build? Sep 28 18:05:14 otavio: yes, qemux86-64, genericx86-64 both fail Sep 28 18:05:24 I found it testing intel-corei7-64 Sep 28 18:50:35 Is someone using package feeds with dnf? It is failing on me Sep 28 18:53:04 the ab exercises that, so it should work Sep 28 18:53:30 rburton: i should call dnf makecache on the box right ? Sep 28 18:53:53 pass Sep 28 20:20:10 does anyone has experience with lxc with yocto? Assume I have a ipk package and I want to isolate this application and want to isolate the used libraries. How can I accomplish this? Sep 28 23:04:02 huh, the update_font_cache intercept is what's hanging in my images. weird **** BEGIN LOGGING AT Thu Sep 28 23:16:11 2017 Sep 29 00:36:00 After I configured with bitbake -c menuconfig virtual/kernel and save, I did bitbake virtual/kernel, but after a few seconds the operation is finished without errors. However it did not seem to have recompiled my kernel. Anything wrong? Sep 29 01:24:37 Excuse me if this is inappropriate. But I am looking for a young junior dev, maybe still in college, who is a Linux enthusiast and looking to get some hands on experience and training with a Yocto build in production. There would be compensation for it. Sep 29 02:13:05 When I update to a new branch for poky or other meta, how do I make the changes to take effect when I do bitbake image ? Sep 29 02:20:18 User__: changes will take effect next time you build an image Sep 29 02:21:41 User__: after doing the bitbake -c menuconfig virtual/kernel changing something and saving the file, the config file should invalidate the shared state so when yoiu ask it to rebuild it should "know" something changed and rebuild the whole thing Sep 29 02:22:58 User__: you could do bitbake -c cleansstate virtual/kernel and then rebuild which would basically force it to rebuild, however I would investigate further the issue of saving the file, since when you rebuild it will probably wont have the changes you asked for (since it would have invaidated the shared states) Sep 29 02:23:17 User__: so, check that you are saving the file, and where you're saving it to Sep 29 02:24:19 aehs29, I just save to default, when choosing save button, it shows .config, I press enter Sep 29 02:30:15 User__: what you can do, is save that new .config file somewhere else, and use that as a defconfig Sep 29 02:31:28 User__: IIRC basically you put it on your kernel recipe directory, name it defconfig and modify your kernel recipe to add the defconfig file on SRC_URI Sep 29 02:31:53 User__: now that file will be used as your base configuration for the kernel Sep 29 02:32:13 User__: you may want to check the Yocto kernel dev manual Sep 29 02:32:36 User__: http://www.yoctoproject.org/docs/latest/kernel-dev/kernel-dev.html look for defconfig Sep 29 02:34:42 so the .config just created by menuconfig is not used to build the kernel? Sep 29 02:36:37 User__: it should be, Im just proposing a solution in case its not Sep 29 02:37:14 User__: but it should be used without you doing anything else other than saving it Sep 29 02:37:52 I see. Just confused why there is .config and defconfig Sep 29 02:38:41 User__: the .config is created from the defconfig Sep 29 02:38:54 User__: in case you specify it Sep 29 02:39:56 User__: so if the bitbake finds a defconfig it will use it as base to create .config and it will then build the kernel using the .config file it just created Sep 29 02:43:30 I just tried to bitbake the new image (after modifying with menuconfig, and issuing bitbake -c cleansstate virtual/kernel). It took a few minutes, but when I verified the images, I looks the same as before (same size, but with updated current date and time) Sep 29 02:44:18 I think the kernel was not rebuilt with the new config file? Sep 29 02:50:55 User__: it whould rebuild once you modify your .config, idk why its not doing it, if you use it as defconfig you would force the changes Sep 29 02:56:39 aehs29, How to use it as deconfig? Sep 29 02:57:07 aehs29, Is there a procedure to do to "use it as deconfig"? Sep 29 02:57:27 I mean defconfig **** ENDING LOGGING AT Fri Sep 29 03:00:01 2017