**** BEGIN LOGGING AT Wed May 17 03:00:04 2017 May 17 05:28:44 Since I update to the latest poky git:master my custom kernel builds are failing with: "kgit-s2q: command not found" May 17 05:36:01 zzeroo, normally that would be in recipe-sysroot-native/usr/bin/kgit-s2q of your kernel build May 17 05:40:12 No isn't there.... I had tryed to include "DEPENDS = "kern-tools-native"' but that dos'nt work, too. https://gist.github.com/95a2d93ae0fe0b09fa9f544bc2981b5f May 17 05:43:43 you shouldn't need to add a depends on kernel tools, if you inherit/use the base kernel stuff. May 17 05:46:41 paulg: I know this was just a test. I fight since two hours on that May 17 06:00:11 zzeroo, yeah I get how fighting with infrastructural stuff can be frustrating. Have you seen this already? May 17 06:00:16 https://www.yoctoproject.org/docs/1.6.1/kernel-dev/kernel-dev.html May 17 06:03:30 paulg: I use this https://www.yoctoproject.org/docs/latest/kernel-dev/kernel-dev.html May 17 06:04:44 ok, yes - that is a more up to date version. May 17 06:04:48 But the whole documtentation dosn't contain on 'kgit' ^^ May 17 06:04:56 But the whole documtentation dosn't contain one 'kgit' ^^ May 17 06:07:02 not a solution, but an approach -- you could maybe start with a bbappend instead of a custom bb; then you should have all the tools / depends at your fingertips May 17 06:08:09 A bbappend for a vanilla linux kernel? which May 17 06:10:43 paulg: but i'll try. This is one thing i hav'nt yeat May 17 06:17:25 Is there anyone here who is using the "testimage" features, also called "automated-runtime-testing" in the dev-manual? I'm trying to use it using the systemd-boot "backend" but it seems like it's not really working out-of-the-box so I'm wondering if there's any current users or if the feature might have been broken at some point. May 17 06:19:35 Someone know why there is nothing in my mkefisik.wks file when I : wic create mkefidisk -e core-image-x11 ? May 17 06:22:35 LetoThe2nd, : Hi, do you have knowledge about the wic create command ? May 17 06:50:06 I added this to my mkefidisk.wks file but I still can't boot part / --source rootfs --ondisk sdb --fstype=ext4 --label platform --align 1024 --size 10240 May 17 06:50:15 someone can help me ? May 17 06:52:40 ChrysD, : Are you there ? May 17 06:53:11 hamdyaea : Yeah May 17 06:53:33 hamdyaea : Are you aware that i'm completely a newbie in yocto ? ahahah May 17 06:53:59 ChrysD, : No I don't know it May 17 06:54:16 hamdyaea : I can try to help but still i can't guarantee any result. May 17 06:54:22 ChrysD, : I am very close to what I want now May 17 06:55:08 hamdyaea : So what's the problem now ? May 17 06:56:21 ChrysD, : I create the image with the command : wic create mkedifisk -e core-image-x11 and the system don't boot anymore when I flash the .direct. May 17 07:21:24 hamdyaea : I sent you a message at the top May 17 07:21:38 you should see my name at the top of the chat May 17 07:29:44 The command to build the target OS Image SDK (toolchain and sysroots) is ? Searched the docs, but clearly I'm missing something May 17 07:41:23 @Smitty_ http://www.yoctoproject.org/docs/2.3/sdk-manual/sdk-manual.html#sdk-building-an-sdk-installer ? May 17 08:14:18 ed2, : I have a question about the : mkefidisk.wks can you help me ? May 17 08:14:38 hamdyaea: sure May 17 08:15:32 ed2, : The last time I have created an image with wic create mkedifisk -e core-image-x11, there was many things already in the mkedifisk.wsk file May 17 08:15:48 ed2, : Now there Is nothing May 17 08:16:12 Sorry, I dropped my connection before I could copy that link. Could you post it again, please May 17 08:16:20 ed2, : I just added : part / --source rootfs --ondisk sdb --fstype=ext4 --label platform --align 1024 --size 10240$ May 17 08:16:37 ed2, : But I don't know why the last time there was many things May 17 08:17:03 hamdyaea: could it be that you've changed it accidentally or changed the branch? May 17 08:17:19 hamdyaea: is gid diff shows something for this file? May 17 08:17:24 git diff May 17 08:20:02 ed2, : https://pastebin.com/StYG927s there are many diffenrence but I didn't change the branch May 17 08:21:23 hamdyaea: looks like you've changed it. "git checkout scripts/lib/wic/canned-wks/mkefidisk.wks" will restore it back May 17 08:21:34 hamdyaea: however, your last change will be lost May 17 08:21:52 hamdyaea: I mean the last like you've added May 17 08:22:46 ed2 : I am in the jethro branch I have to do : git checkout jethro scripts/lib/wic/canned-wks/mkefidisk.wks ? May 17 08:26:05 nope, just what i showed May 17 08:26:37 it will revert that file to HEAD, which is jethro in your case May 17 08:27:18 ed2, : OK I see, I have a question again : -part /boot --source bootimg-efi --sourceparams="loader=grub-efi" May 17 08:27:39 ed2, : How can I modify it if I don't want the efi ? May 17 08:28:27 hamdyaea: you can look at directdisk.wks. I thought this is what you've used to produce bootable image, no? May 17 08:30:05 ed2, : OK I see, so I can do a : wic create directdisk -e core-image-x11 ? May 17 08:30:12 yes May 17 08:30:29 ed2, : Thanks you saved many many hours :) May 17 08:30:30 or wic create -e core-image-x11 May 17 08:30:38 ed2, : of search May 17 08:31:44 hamdyaea: anyway, you'd probably benefit from reading the doc: http://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html#creating-partitioned-images May 17 08:32:04 and reading wic help kickstart too :) May 17 08:34:49 ed2, : OK i will read it May 17 08:36:35 hamdyaea: I'm actually surprised that your hardware doesn't support EFI boot. Could it be that you've missed it somehow when you looked for it in BIOS settings? May 17 08:39:01 ed2, : I looked everywhere there is no efi option in the bios and when I try an efi image it don't works.. May 17 08:39:12 Can someone please post the link to the creation of an SDK (toolchain and sysroots) for an OS Image. I cannot find it May 17 08:39:23 ed2, : So it don't work and I don't see the advantage of the efi May 17 09:05:09 @Smitty_ sure http://www.yoctoproject.org/docs/2.3/sdk-manual/sdk-manual.html#sdk-building-an-sdk-installer ? May 17 09:40:56 Don't undertsnad why my conneciton keeps dropping !! May 17 09:41:56 Anyway, thanks for the link. That confirms my understanding, it's an option to the target image build, rather than a separate target May 17 09:43:44 I'm not sure I undertsand the section about static libraries. "By default, this toolchain does not build static binaries. " How does that make any sense ? May 17 09:44:59 this is not because of GPL? May 17 09:46:26 To me to make much more sense for an SDK to include static libs by default, and an OS Image to *not* include those libs. After all, how many people are going to be doing development *inside* their running OS Image ? Yet the whole point of the SDK is to do development - and static libraries are definitely needed there. May 17 09:47:46 And the instructions for addressing this issue is to also include the static libs in the OS Image, which seems unnecessary, even undesirable. May 17 09:48:50 Smitty_ : Maybe i'm in another world as i'm a newbie, but because of avoiding lot of issue due to GPL requirement, does it something like... As by default you doesn't build static binaries, and the fact that you have to enable, make you more responsible in case of violation of GPL? May 17 09:49:39 I don't follow your argument. How does a library being static change anything to do with licencing ? May 17 09:57:38 So, again, I ask this question, since the answer I received yesterday was confusing. What's the approved/recommended way to include the static libraries in the SDK only ? May 17 09:58:58 RDEPENDS_${PN}-dev = "" ? May 17 10:03:19 If the library is built using GPL components and is not GPL itself, then the applies as if you were building a statically linked executable. May 17 10:03:30 If the library is built using GPL components and is not GPL itself, then the same applies as if you were building a statically linked executable. May 17 10:06:12 And all of this is the reason why copyright, patent and (perhaps) trademark laws should all be taken out back and shot. May 17 10:06:33 IMHO May 17 10:07:23 JoiF : Maybe i'm not understanding, but are you admitting what i'm seeing ? My english is quite bad. May 17 10:07:48 Admitting? May 17 10:07:54 agreeing or whatever May 17 10:08:05 Smitty_: is your question "how do i put static libraries into a SDK i've built?" May 17 10:08:05 ChrysD: Probably May 17 10:09:44 ChrysD: Neither of us are probably offering any help to poor Smitty_ (on my part because I don't know a helpful/practical answer to his question), but I think we're both talking about the same thing, yes. May 17 10:10:28 JoiF : Yeah it's what I was talking about to be sure that i understand thanks. May 17 10:12:02 ChrysD: And then I went on (completely off-topic for this channel) to express my opinions on particularly copyright and patent law. ;) May 17 10:13:19 Smitty_: because if it is, the answer is SDKIMAGE_FEATURES_append = " staticdev-pkgs" in either your local.conf for just you, your distro conf as a product-wide default, or an image recipe for per-sdk control May 17 10:14:01 Smitty_: note that poky doesn't build static libraries generally so hopefully you've started constructing your own distro May 17 11:06:09 Please clarify "...poky doesn't build static libraries generally..." - ~~Shirley~~ that's down to the individual recipes. I mean, there's loads of absolutely standard utilities that are composed of libraries. May 17 11:07:00 Smitty_: it defaults to only building dynamic shared libraries, not static ones May 17 11:07:46 I don't understand how that makes any sense, for the reason I just stated. Plenty of standard utilities are made up of static libs May 17 11:07:49 not dynamic May 17 11:08:28 I'm not sure we're even talking about the same thing here May 17 11:08:39 Smitty_: poky configures autotools to pass —disable-static so vast amounts of the system won't be built statically May 17 11:08:51 WHAT !!! May 17 11:09:01 chill, it's an option. turn it off. May 17 11:09:22 remember poky is an example. make your own distro. if you want static libraries, don't turn them off. May 17 11:09:35 FWIW the system does not need or use those static libraries, we've had no reports of problems due to that change May 17 11:09:45 and as rburton says, if it bothers you, enable them May 17 11:09:57 Yeah, maybe this is one of the things I don't understand correctly yet. I keep thinking Poky is a core framework May 17 11:10:12 poky is an example of how to build a distribution using oe-core and bitbake May 17 11:11:48 what is "oe-core" then ? Is it the set of layers that make up the core-image-sato (or minimal) ? May 17 11:12:00 oe-core is basically the meta/ directory in the poky tree May 17 11:12:13 http://git.openembedded.org/openembedded-core/tree/ <— oe-core May 17 11:12:47 as i said yesterday, poky is oe-core + bitbake + meta-poky squashed into a single repo May 17 11:13:10 But, doesn't pretty much everyone use that as a starting point, then maybe add or possibly remove unwanted components from it ? In this sense, it looks very much like a framework May 17 11:13:21 what is that? May 17 11:13:35 poky May 17 11:13:38 some do May 17 11:13:43 some just take bitbake and oe-core directly May 17 11:14:09 if you're building your own distro then you might as well just use oe-core+bitbake directly May 17 11:14:35 but there's hardly anything else in the poky tree May 17 11:14:40 correct May 17 11:14:49 RP: since you are around, you might be able to answer this question i had regarding RSS -> Is it an intended feature of RSS that any task dependency (even a dependency on say foo:do_deploy) causes the recipe-sysroot population of that dependent task's sysroot components? May 17 11:14:58 Smitty_: the only bits in poky that are not in oe-core and bitbake are the pieces you'll be replacing anyway May 17 11:15:36 nrossi: in general, yes but there is logic which limits that dependency tree May 17 11:15:55 nrossi: e.g. dependencies of a target recipe on native recipes are not followed May 17 11:16:11 nrossi: its the same logic as the "build from sstate" situation would use May 17 11:16:48 nrossi: see setscene_depvalid() in sstate.bbclass May 17 11:17:51 is there a description of the layers in poky: meta, meta-poky, meta skeleton, meta-yocto, meta-yocto-bsp, meta-selftest and why they are there ? May 17 11:18:17 RP: ok, thanks. Just wanted to make sure I was understanding the logic correctly. I am creating a "native" like extend class for a tertiary platform cpu, and it is being populated into the sysroot. Unfortunately I don't see an easy way to prevent that would modifying staging.bbclass like is done for native/multilib/etc. May 17 11:20:08 nrossi: might you be better using multiconfig? May 17 11:20:36 nrossi: I know we are missing inter multiconfig dependencies right now but that shouldn't be hard to fix May 17 11:20:53 RP: yes that what is need from multiconfig to have it working May 17 11:21:25 RP: but this is all targeted for meta-xilinx's pyro release, but would be great to use multiconfig in the future :) May 17 11:22:08 nrossi: I'd love to have the patch to sort that. It shouldn't be hard as we put all the foundations in place... May 17 11:22:30 RP: i've just worked around the sysroot population issue for now by setting the extension class to use a /usr/ May 17 11:27:14 rburton: I'm still missing something. The project that gets downloaded when following http://www.yoctoproject.org/docs/2.3/mega-manual/mega-manual.html#building-images is "poky", which is essentially all the meta data responsible for instructing *some set of tools* to construct an OS image. Some of those tools are of course required as the target OS toolchain, I'm not interested in them. I want to know what the other set of tools May 17 11:28:22 RP: also just curious, with regards to the staging populating. In your opinion should the image recipes populate things like u-boot, qemu-helper-native, etc. (stuff in EXTRA_IMAGEDEPENDS)? May 17 11:29:14 It seems like some of those tools may be specific to the Poky OS image, and so are provided by the poky project I download. Then there's other tools that are bitbake, but, strangely also downloaded when I retrieve poky. (BTW, that's very confusing) May 17 11:29:24 Smitty_: the quick start has you download poky as its quicker/easier than bitbake+oe-core and ensuring you get matching versions May 17 11:29:48 this really isn't complicated. poky is a few projects merged into a single git repo for convenience. May 17 11:30:06 It's not complicated when you understand it May 17 11:30:30 nrossi: populate in what sense? you mean the do_populate_sysroot tasks from u-boot and so on? May 17 11:30:42 Smitty_: poky is bitbake + oe-core + meta-poky, in a single git repo, for convenience. May 17 11:30:50 use it or not, your choice. May 17 11:30:53 some do, some don't May 17 11:30:56 The clouds are lifting May 17 11:32:03 Did I miss that day in school where that was discussed ? (where is that explained in the documentation ?) May 17 11:32:13 RP: yes, whilst u-boot is mostly non-problematic, qemu-helper-native can be in that you can only have one set of qemu binaries despite the usefulness of RSS where you could have two different qemu-helper-native's which provide different sysroots for qemuboot May 17 11:32:29 Smitty_: the quick start definitely skips over it as the purpose of the quick start is to be quick May 17 11:32:53 ok, so I'm not as bad as I feared May 17 11:33:38 nrossi: it was originally intentional as we had a lot of trouble actually getting qemu binaries where we could find them to boot images May 17 11:34:01 nrossi: I say originally as I'm unsure where we ended up running them from in the end May 17 11:34:19 What do you mean "ensuring you get matching versions" - matching versions of what ? what needs to match ? May 17 11:34:51 Smitty_: bitbake has an API which oe-core uses, so if you use new oe-core with two year old bitbake, the APIs don't match May 17 11:34:52 RP: from what i can tell it was because of this that qemu was being populated into the images -native sysroot, and that was what qemuboot was pointing at May 17 11:35:07 I'm not sure if I say is correct, but you download poky because everything come with same version and everything work... so you don't sometimes " use" poky, but you download the " poky directory". May 17 11:35:44 bitbake "has an API" ? isn't it a collection of commandline tools ? May 17 11:36:00 RP: but with the recent changes, qemuboot now points to the sysroot of qemu-helper-native May 17 11:36:13 nrossi: right, I thought it had changed May 17 11:37:07 Smitty_: bitbake parses the metadata. The format of the metadata is effectively an API and that format can change over time like any API May 17 11:37:25 Smitty_ : You have learn yoct at school ? May 17 11:37:33 Smitty_: also there's a python library which the metadata has access to May 17 11:37:33 RP: previously with the single sysroot, i had a custom QEMU installing into a /usr/bin/qemu-foo/ directory so that both could exist. But now with RSS it would be nice to remove this hack :) May 17 11:37:35 Smitty_ : I mean, oe-core bitbake ? May 17 11:38:24 nrossi: I'm open to ideas... May 17 11:40:22 RP: im not entirely sure, but i would imagine something along the lines of excluding do_deploy tasks from sysroot population, and simply setup qemu-helper-native so that is "deploys" its native sysroot (aka nop deploy task) May 17 11:41:20 nrossi: do_deploy tasks are excluded as the sysroot only handles do_populate_sysroot tasks. You mean skip anything which takes a dependency chain through a do_deploy task? May 17 11:41:49 RP: yes, thats what i mean May 17 11:44:00 nrossi: hmm, maybe May 17 11:56:59 That problem exists in loads of other environments, and is solved usually by having the tool not honor (warn or error) on instructions it doesn't understand. It's still the developers problem to get a matching tool and API as expressed in the meta-data May 17 11:58:41 the failure case is an error that the required API version isn't high enough May 17 11:59:14 right, so the tool (bitbake) just says, sorry, I can't understand this, get the right version of bitbake you dummy ! May 17 12:00:14 I didn't realise that Yocto was trying to do all that May 17 12:03:54 ANd, I'm not sure why you would want to try - can someone explain that ? I mean, is there some compelling reason that it's not left to the developer to get the right version of the tools ? May 17 12:05:49 Smitty_: accidental use of mismatching versions: eg bitbake adds new api that oe-core then uses , if you upgrade oe-core but not bitbake then it will faiu May 17 12:05:51 fail May 17 12:06:43 rburton : that's why we download poky to get sure that everything work, not because we want poky itself right? May 17 12:07:01 it saves you the need to track two repos yeah May 17 12:07:15 but generally speaking you'll get a big error if you need to upgrade bitbake May 17 12:07:16 ie http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=66bd8b82472ee6dc91a715ac91410dc5a826f2a3 May 17 12:07:35 Sure, so they get delivered together - that makes sense. But, why deliver them with poky ? Because poky also relies upon specific version of the API ? May 17 12:08:03 because oe-core and bitbake are separate repositories and they only get delivered together in poky May 17 12:08:42 How does that make sense, since their versioning is so tightly coupled, why are they seperate ? May 17 12:08:47 guys, how to resolve QA warnings? I have several after building mariadb: http://dpaste.com/0V4FRTK May 17 12:09:07 Smitty_: because bitbake is a generic build system, you don't need to use oe-core May 17 12:10:57 OK, so if you want to use oe-core, you need the right version of bitbake, and if you want to use yocto, you then need the right version of bitbake+oe-core. May 17 12:11:10 yocto isnt software you can download May 17 12:11:22 the yocto project is oe-core + bitbake + lots of other bits of software May 17 12:11:23 yocto is the meta-data May 17 12:11:26 right ? May 17 12:11:34 no, oe-core is the metadata May 17 12:11:45 well, some of the metadata. the core metadata, thus the name. May 17 12:11:50 Smitty_ : Yocto is just thename of the project. May 17 12:11:56 which project ? May 17 12:12:01 the yocto project May 17 12:12:01 Smitty_ : Yocto Project ahah May 17 12:12:10 ChrysD: ^5 May 17 12:12:47 the project to produce bitbake and oe-core in a manner which can be used to produce an OS Image (or sdk, or whatever) and the example is poky ? May 17 12:12:49 Smitty_ : Yocto it's only the name of the project that is : thecombinaison of tools and humans ressources beyond the mecanism that we use in everyday life May 17 12:12:59 Smitty_: yes May 17 12:13:32 It's like a bunch of little all ladies making pancakes. They call themselves "Lil'Ol'Ladies" but old ladies is not their product.. pancakes is their product.. :p May 17 12:13:46 s/all/old/ May 17 12:13:55 JoiF : thank you. ahahah May 17 12:14:39 And Open-Embedded is also the same ? Like it's the name of the project besides oe-core? May 17 12:14:43 Smitty_: https://www.slideshare.net/rossburton/the-yocto-project May 17 12:15:18 So, in fact, if I'm a parasitic business, trying to produce a product that requires a full linux distro, and also uses some weird hardware - in fact, I don't care about Yocto, I am really only interested in bitbake and oe-core, and I might have a look at the poky example to get a feel for how to use them May 17 12:17:55 Smitty_ : What you mean by " you don't care about yocto ".d. May 17 12:21:09 Does the open embedded project still exist and yocto project follow the evolution of oe-core, or now open embedded project doesn't exist anymore and it's the yocto wrokgroup that is in charge of upgrading oe-core? May 17 12:26:37 OK, so, once I download poky, I see a bunch of "layers" each in its own directory. How do I know which layer depends upon which ? Can bitbake show me that ? May 17 12:27:16 Smitty_ : What you mean by which layers depends upon which ? The purpose of layers is to be not dependant. May 17 12:28:34 then in what way are they layers ? May 17 12:30:00 For example, if I want to build the core-image-minimal, doesn;t that require that some other layer like meta get processed first ? May 17 12:30:23 then meta-skeleton, then meta-yocto (or whatever ? May 17 12:31:37 Smitty_ : you need that : https://www.google.fr/search?q=yocto+layers&client=ubuntu&hs=8e0&channel=fs&source=lnms&tbm=isch&sa=X&ved=0ahUKEwjR19e79_bTAhVPZ1AKHc4NBRwQ_AUICigB&biw=1301&bih=671#imgrc=voOjRiJQPtAARM: May 17 12:34:32 right, that's what I thought, so there is a dependence. I can't build a layer I add in for my app that I want to sit on top of core image minimal until I build whichever layer is responsible for the toolchain (oe-core, I guess ???)..Furthermore, I probably want other libraries that are not present until another layer, like meta-poky is built. So, in what way is there not a dependency ? May 17 12:34:40 Smitty_: the layer README should state the layers it depends on. The actual dependencies will be from individual recipes depending on recipes in the other layer May 17 12:35:11 Smitty_: meta-skeleton is an example, meta-yocto-bsp are some sample BSPs so unrelated to the image. May 17 12:35:23 also each layer.conf can express layer dependnecies May 17 12:35:35 Smitty_ : I prefer that image http://www.aosabook.org/images/yocto/aosa2.jpg May 17 12:35:43 eg LAYERDEPENDS_meta-python = "core openembedded-layer" May 17 12:36:32 rburton: do we use that for something? May 17 12:36:35 I'm getting there.....v e r y s l o w l y ..... May 17 12:37:03 jku: bitbake erros out if they're not satisfied May 17 12:38:08 oh right just found it May 17 12:39:01 So, back to yesterdays question. I want to add a static library to core-image-minimal. In fact I really only want it in core-image-minimal' SDK. So, I create a layer for my library. using something like a template (meta-skeleton???) as starting point. May 17 12:39:27 use devtool to make a recipe, then add the package to your sdk May 17 12:40:04 rburton : RDEPENDS_${PN}-dev shuold help there and specify a package which include the static library ? May 17 12:40:48 ChrysD: more logical to use the right variable to add the static library pakage to the sdk May 17 12:40:54 SDKTARGETSOMETHINGOROTHER May 17 12:41:07 why do I use devtool ? is copying meta-skeleton and editing it not a good idea ? May 17 12:42:04 Smitty_ : Because some people in the yocto project spend our doing that tool ? lool May 17 12:42:11 Smitty_ : Because some people in the yocto project spend hour doing that tool ? lool May 17 12:42:34 And I want a layer , not just a recipe , don't I ? May 17 12:43:34 because I don't want to change the existing layers, which I am quite happy with. I should just add a layer on top of them May 17 12:44:05 Smitty_ : Yeah but you have also some tools to do a layer May 17 12:45:11 Sorry, I don't understand - "to do a layer" May 17 12:45:19 you mean create a new layer ? May 17 12:45:21 Smitty_ : to create May 17 12:45:43 and I don;t copy skeleton why ? what's it for ? May 17 12:46:02 Smitty_ : For example skeleton show you how to do a recipe of a kernel module May 17 12:46:12 Smitty_ : How to do a recipe for a service May 17 12:46:24 Aaaaaaarrrrrrrrrgggggghhhhhhh May 17 12:46:44 you mean like skel the kernel module example ? May 17 12:46:49 Hey, i'm trying to build the 3.4 kernel for my yocto build but it says it can't find the kernel even though it exists. Does anyone know why it doesn't work? http://git.yoctoproject.org/cgit.cgi/poky/tree/meta/recipes-kernel/linux/linux-yocto_3.4.bb?id=28fb81ab56fda2c2ef5b7c8f84610a35efcdaba6 May 17 12:47:46 Smitty_ : in meta-skeleton/recipes-skeleton/service you ahve an example of recipe for a service May 17 12:47:51 from the famous book by gkh ? May 17 12:48:54 I think the modul is scull ;) May 17 12:48:54 Smitty_ : I don't remember but I think meta-yocto-bsp is also an example of bsp layer May 17 12:49:33 yeah, that's kind of obvious from the name, unlike skeleton which sounds like a "bare-bones' generic layer for *something* May 17 12:49:59 Smitty_ : meta-skeleton is a bunch of lot recipes examples for various thinks May 17 12:50:08 Hi, how do I rebuild a single recipe in such a way that do_install_append() defined inside the recipe is run again? May 17 12:51:06 A couple of sanity checks inside do_populate_sysroot in meta-qt5 fails and I've added a few sed commands to the do_install_append func, but they don't seem to have any effect May 17 12:51:48 I even tried to do echo something ?? /home/user/hello.txt but that file isn't created May 17 12:52:00 Just to check if the func was executing May 17 12:53:09 Err echo something >> hello.txt May 17 12:53:14 BaloneyGeek|work: I think you can also just clean that package, and then rebuild May 17 12:53:29 bitbake -c clean qtbase && bitbake qtbase? May 17 12:53:38 BaloneyGeek|work: Something like bitbake -c clean and then rebuild May 17 12:53:47 It didn't have any effect the last time but let's try again. May 17 12:53:52 BaloneyGeek|work: i think that would work yea May 17 13:11:20 rj_: Nope, that failed May 17 13:12:03 echo ${D} >> /home/builder/helo.txt <-- that line's in do_install_append May 17 13:12:19 BaloneyGeek|work: Aww, i expected it to work May 17 13:12:20 I don't see that file being created either May 17 13:13:11 howabout something like... touch /home/user/test May 17 13:14:37 Hang on, clean and make will take a while :-/ May 17 13:15:49 In the meantime, does this sed command look okay? May 17 13:15:52 sed -i -E "s@-L(.*)/opt/vc/lib@-L/opt/vc/lib@g" ${D}/${libdir}/pkgconfig/Qt5*.pc May 17 13:16:25 It's supposed to basically fix -L/home/builder/.../opt/vc/lib to -L/opt/vc/lib May 17 13:16:43 I honestly don't know that much about sed, as i am quite new to linux/yocto May 17 13:16:49 @rj_ PREFERRED_PROVIDER_virtual/kernel is correctly set ? May 17 13:17:18 PinkSnake: In the machine.conf? May 17 13:17:28 @rj_ yep May 17 13:17:41 @rj_ to select the kernel kernel May 17 13:18:04 PinkSnake: it's not there at all, i only have: PREFERRED_VERSION_linux-yocto_genericx86-64 ?= "3.4%" May 17 13:18:06 @rj_ and PREFERRED_VERSION_virtual/kernel May 17 13:18:41 @rj_ maybe check in board configuraiton maybe an other kernel is selected May 17 13:18:56 PinkSnake: Ah, its in the .inc: PREFERRED_PROVIDER_virtual/kernel ?= "linux-yocto" May 17 13:19:07 Should i change that to "3.4%" ? May 17 13:19:29 @rj_ maybe try 3.4 but normaly 3.4% is ok May 17 13:19:42 no no May 17 13:19:48 you have to set like that : May 17 13:19:57 PREFERRED_PROVIDER_virtual/kernel ?= "linux-yocto" May 17 13:20:08 Yea, i have that atm May 17 13:20:12 PREFERRED_VERSION_virtual/kernel = "3.4" May 17 13:20:34 provider and version are used to select the correct kernel recipe May 17 13:21:01 aaah okay ill try May 17 13:26:25 rj_: Nope, same error, file not touched May 17 13:32:14 BaloneyGeek|work: That's weird... it should work, since i have the same method only in do_configure_append May 17 13:39:18 Any other ideas? May 17 13:41:23 @BaloneyGeek|work you to add the file inside the rootfs target ? May 17 13:41:33 you want* May 17 13:42:11 PinkSnake: No, I was trying to test if the do_install_append function was being executed at all May 17 13:43:10 What I want to do is execute this line before do_populate_sysroot is ru May 17 13:43:11 sed -i -E "s@-L(.*)/opt/vc/lib@-L/opt/vc/lib@g" ${D}/${libdir}/pkgconfig/Qt5*.pc May 17 13:43:45 So I put this command into do_install_append inside qtbase_git.bb (the function was already there) May 17 13:43:51 But it doesn't seem to have any effect May 17 13:44:30 @BaloneyGeek|work do_install_append is inside recipe.bb ? May 17 13:46:39 PinkSnake: yes May 17 13:47:41 BaloneyGeek, what's the parens for ? May 17 13:48:00 parens? May 17 13:48:15 (.*) May 17 13:48:19 that part May 17 13:48:24 @BaloneyGeek|work do_install_append is not only in .bbappend ? May 17 13:48:53 The idea is to convert -L/home/builder/.../opt/vc/lib to -L/opt/vc/lib May 17 13:49:04 I independently verified that the sed command is correct May 17 13:49:22 PinkSnake: it's in the .bb file. There's no .bbappend May 17 13:49:53 OK, so you are not asking any more for support for the sed itself ? May 17 13:50:13 @BaloneyGeek|work don't modifie recipe directly should use .bbappend May 17 13:50:14 Smitty_: No, because I'm pretty confident that the sed command is correct May 17 13:50:18 OK May 17 13:50:33 Unless ${D}/${libdir}/pkgconfig/Qt5*.pc is wrong? May 17 13:51:14 PinkSnake: I'd need to have my own layer for that right? May 17 13:51:20 Seems overkill for one sed line May 17 13:53:17 OK, still trying to figure out how a layer is added when *not* added by me to build/conf/somefile.conf - I mean, clearly something is adding the layers to the build, what is it ? May 17 13:53:57 Smitty_: Layers aren't added automatically right? May 17 13:54:45 That's right, as far as I have been told. It's not like simply the presence of a subdirectory of my project source directory that looks like a layer is enough to get it added May 17 13:54:48 I tried to add a new function called do_populate_sysroot_prepend and do the stuff there if that worked, but no joy there either May 17 13:55:15 Smitty_: indeed, i always had to manually add it to build/conf/bblayers.conf May 17 13:55:31 Smitty_: if its not in bblayers.conf it won't be read at all May 17 13:55:39 BaloneyGeek|work: How about do_configure_append() ? May 17 13:56:00 Smitty_: Are you confusing a "layer" with a recipe that's included in a layer? May 17 13:57:17 a "layer" normally has a "conf/layer.conf" associated with it May 17 14:00:11 Yeah, I am slowly groking this stuff. So, we for some reason have a file still called bblayers.conf.sample that is copied to MYBUILD_DIR/conf/bblayers.conf I *must* get used to this idea that these files are *generated* dynamically by a standard process May 17 14:00:19 @BaloneyGeek|work yes but it's yocto friendly May 17 14:00:26 are not generated May 17 14:01:45 PinkSnake: Yeah but would there be other solutions in case I didn't want the overhead of a new layer right now? May 17 14:02:52 BBLAYERS_NON_REMOVABLE doesn;t appear in mega manual. What is it ? May 17 14:04:40 Smitty_: if i say it doesn't matter and you can ignore it, will you just ignore it? May 17 14:04:48 just pretend its not there May 17 14:05:18 close your eyes May 17 14:05:59 It's kind of weird, whoever created our bblayers.conf file left it there, but duplicated meta and meta-yocto in BBLAYERS. Is it actually required ? May 17 14:06:28 the NON_REMOVABLE assignment is for toaster May 17 14:07:17 i think there's a better way of doing what it does but thats for later May 17 14:07:36 yeah, another tool I know nothing about - really, what is toaster ? (please tell me there's a good reason for that name) May 17 14:08:12 @BaloneyGeek|work it's up to you but in my opinion it's better to use .bbappend, for example I created a .bbappend on the kernel recipe juste to save .config file ... May 17 14:09:22 Smitty_: It's for creating a pretty picture such as this one: https://www.jof.guru/temppics/non-remove.png This was Hob, now deprecated, toaster is a web-based GUI thingee for messing about. May 17 14:09:51 Sorry, I am a detail guy, I really struggle to move past stuff that seems wrong (and my definition of wrong includes anything that looks duplicated) May 17 14:11:15 OK, so there's some gui tool that's trying to sanity check on editing a config file (which is wrong, since there is already a tool that has to parse the config file, it should do the sanity check) May 17 14:11:45 Smitty_: its not sanity checking but please just don't rathole into what toaster is May 17 14:11:47 Not exactly. Toaster is a GUI for configuring May 17 14:12:05 right, I don't care. May 17 14:13:08 Errr.. You asked.. :) May 17 14:22:34 So, oe-init-build-env calls oe-setup-builddir which wants a file (oddly it wants it already in the build directory) that indicates which layers to be built. If it doesn't find it it looks for the bblayers.conf.sample file, and moves that into the build directory (under conf). I am struggling to understand the mindset that devised this. I really don't get the idea that I should ever pre-populate *anything* inside a build directory May 17 14:22:54 ...or put files in my source directory that just get copied into the build directory May 17 14:23:05 as configuration May 17 14:23:20 why not just read the configuration out of the source directory ? May 17 14:26:54 i'm strugging to wonder why you're still worrying about this. you need to configure the layers that are available. it can't auto-detect because it has no way of knowing where you've put them, and mandating a layer placement policy would be silly and annoy people who routinely add/remove layers. May 17 14:27:06 please just accept thats what oe-init-build-env does, and move on. May 17 14:31:06 lol, he's still going? Lawd. May 17 14:31:17 It's a question of comprehending how to do what seems like a simple task - add a layer to a project so that another developer can retrieve the project and *just build it*. without having to edit build/conf/bblayers.conf and add in my layer - which if I follow every piece of documentation on how to add a layer, is what he would have to do. May 17 14:32:30 Smitty_: as discussed yesterday it's either 1) document to edit bblayers.conf, or 2) provide your own bblayers.conf.sample. May 17 14:32:53 Smitty_, maybe it's easier if you see how Yocto was built around OE May 17 14:33:00 pls check out https://www.openembedded.org/wiki/OE-Core_Standalone_Setup May 17 14:33:29 Smitty_, the idea was, we must not scare Corp. developers :) May 17 14:33:47 WHat does that mean ? May 17 14:34:06 it means you can just start with bitbake + oe-core May 17 14:34:31 we went through this earlier today May 17 14:34:40 ah, sorry then May 17 14:34:43 np May 17 14:34:45 Smitty_ : What's bad about bblayers.conf? May 17 14:39:05 Smitty_ : I mean, what's bad with the idea of telling which layer you would like to the specific build you are doing ? May 17 14:39:25 Okay I'm a fully laden ocean going idiot. I was editing the file in the wrong git clone. May 17 14:39:39 Works now May 17 14:39:55 Smitty_ : And what about a script you can make in other to source " automatically" and download " automatically" the good git repo and editing "automatically" the bblayers.conf for the developper you would like to give May 17 14:41:18 ChrysD: no need for that, the git repo can provide a correct sample May 17 14:41:23 (which is what poky does) May 17 14:42:26 rburton : what you mean? May 17 14:42:51 oe-core and meta-poky provide bblayer.conf.sample files, which are different May 17 14:43:08 You know, sometimes it's the little things that cause confusion. Just call the file bblayers.conf (remove the .sample) Enough said. Sorry for being a pain. May 17 14:43:37 Smitty_: why? that would be wrong. its not bblayers.conf mainly because if you used it as-is it would fail to parse May 17 14:44:08 what do you mean ? Why would it fail to parse ? May 17 14:44:12 bblayers.conf.template i can see as being a sightly more descriptive filename May 17 14:44:36 BBLAYERS ?= " \ May 17 14:44:36 ##OEROOT##/meta \ May 17 14:44:36 " May 17 14:44:52 yeah, that's what mine looks like May 17 14:44:57 thats the sample May 17 14:45:06 the ##OEROOT## gets replaced for you May 17 14:45:11 to be a real path May 17 14:45:18 then its not a template but a working file May 17 14:45:36 Exactly, so, call it bblayers.conf May 17 14:45:46 it's the confuigraiton file for the layers May 17 14:45:46 but it won't parse with ##OEROOT## in! May 17 14:45:59 it will when you run setupdir May 17 14:46:11 which - let's face it - is mandatory May 17 14:46:13 which is when the template gets copied to bblayers.conf May 17 14:46:17 and is no longer a template May 17 14:46:29 copied *and modified* May 17 14:47:12 so what ? by then it's in the build directory, it's not something I want to control at that point. May 17 14:47:44 I want to control it before it gets used by the build system, not after May 17 14:48:06 and oe-init.XXX is part of the build system, May 17 14:51:15 Don't you see why that is confusing ? May 17 15:04:24 does someone here the `meta-rust-bin` layer? May 17 15:04:37 +know May 17 15:06:24 I ask because of this bug i discovered yesterday https://github.com/rust-embedded/meta-rust-bin/issues/19 May 17 15:52:40 Smitty_: Look at this, as an example: http://freescale.github.io/#download the section "Download the source code" May 17 15:56:34 So they're using repo to get all kinds of stuff. One of them is this repo: https://github.com/Freescale/fsl-community-bsp-base May 17 15:57:03 Which contains both a setup-environment script as well as conf/bblayers.conf May 17 15:58:14 So if you need your coworker to "just build it", you could set up a "repo" repository that fetches everything and does what he needs May 17 16:01:46 also, repo is just a wrapper for git May 17 16:05:21 aye, repo is one of many multiple git repo management tools. repo is one of the most popular, others use git submodules, myrepos, or use combo-layer or git-subtree to maintain a monolithic repo May 17 16:41:33 Exactly. Hence the "example" :p May 17 16:54:59 Does anyone know where packagegroup-core-ssh-openssh is defined? There are a bunch of these packagegroups in meta/classes/core-image.bbclass that I can't seem to find the RDEPENDS for May 17 16:59:39 meta/recipes-core/packagegroups/packagegroup-core-ssh-openssh.bb May 17 16:59:43 how can I configure the kernel? is bitbake linux-xlnx -c menuconfig the right command to display the kernel menuconfig panel? May 17 17:01:55 hello, when adding a recipe with Proprietary license, bitbake complains about not finding it in the collected licenses; isn't proprietary license supposed to be already defined ? May 17 17:07:22 top22: IIRC that exception works only with CLOSED May 17 17:24:53 JaMa: poky/meta/files/common-licenses/ contains Proprietary but not CLOSED May 17 17:25:30 strange May 17 18:22:16 joshuagl: Doh, that makes sense. Thanks May 17 18:31:46 Is it possible to define PREFERRED_VERSIONs of packages in image recipes? It doesn't seem to be working for me and I'm wondering if it's intentional or not May 17 18:33:53 no.. PREFERRED_VERSION is a system wide setting.. it affects what is -built- (which then affects what is installed)... May 17 18:34:00 image is 'too late', in that everything is already built May 17 18:35:07 Ahhh I see, thanks! May 17 19:17:51 FWIW, I saw the thing about pseudo and /etc/group but I haven't had time to look closely. Will try to get a look at it "soon". May 17 19:21:18 lol, then you are ahead of me.. I must have overlooked it May 17 19:21:47 ... oh, i see, they just sent it to me directly, not oe-core May 17 19:22:18 long story short, the fakepass stuff has static char foo[PSEUDO_PWD_MAX], which is 4096, because who would need more than 4096 characters for a line in /etc/group? May 17 19:22:25 answer, someone who's making a group with >350 members May 17 19:24:10 seebs: How are you managing your user recipes and their dependencies? May 17 19:24:54 I'm currently working on a user management related problem May 17 19:25:02 I am not! May 17 19:25:03 ahh.. a 'real bug', but an odd one to have that many folks in a group May 17 19:25:13 ah, misunderstood May 17 19:25:30 I don't actually use Yocto anymore, it's just that pseudo is my baby and I want to keep working on it. :P May 17 19:25:44 ok May 17 19:28:07 Well then, anyone else have pointers? I want to create a bunch of groups (essentially to allow a non-root user access to what would normally be privileged hardware), and then have a single user get assigned to the correct groups based on some configuration variables May 17 19:29:06 I currently have a recipe per group, but managing the dependencies for the main user is a pain May 17 19:44:09 Anyone experienced problems with openssl recognizing certs after build? I am drowning. May 17 20:13:58 hello May 17 20:15:16 I take it that plenty of people come through here to ask technical questions, but i'm new to the whole embedded linux field and I am receiving some errors building my setup May 17 20:15:45 and considering the error is coming from standard GNU software (GDBM) i was wondering if anyone knew how to help me May 17 20:28:09 twallis: if you have a particular problem or question, ask it, or post the error May 17 20:28:14 asking if you can ask is not useful May 17 20:28:39 fair enough May 17 20:28:43 chmod: gdbm.h: new permissions are r-xrwxrwx, not r-xr-xr-x May 17 20:28:43 | Makefile:168: recipe for target 'gdbm.h' failed May 17 20:29:10 i'm receiving this error while attempting to 'do_compile' gdbm_1.8.3.bb May 17 20:29:26 what version? May 17 20:29:35 could be a pseudo issue, seebs, ever see anything like that? May 17 20:30:03 no, this is a fresh yocto build May 17 20:30:30 I did retry the bitbake command and it proceeded after but it noted that this job was in error May 17 20:32:06 https://pastebin.com/wg1ix4LD May 17 20:32:09 ^^error log May 17 20:32:57 twallis: is it master or pyro or morty? May 17 20:33:23 krogoth May 17 20:38:30 thank you denix for clarifying, apparently 'what version' was unclear regarding what it was referring to May 17 20:41:28 ah, well gdbm version 1.8.3, yocto krogoth release, and if it counts, I'm trying to build the TI processor SDK version 3.03.04 May 17 20:47:07 hmm. that's a very odd set of permissions, to put it mildly. May 17 20:48:06 could that have been something to do with my build environment prior to calling bitbake? May 17 20:48:47 bitbake's pretty good about setting its own permissions. May 17 20:49:03 We did have an issue at one point involving explicit chmod calls on hard-linked trees, but I don't think that would have produced this. May 17 20:49:40 was the problem something similar to this? May 17 20:53:33 no, it was that a second build of the same package would add user-write permissions to the files. May 17 20:53:47 hmm May 17 20:55:45 any idea what would cause something like I'm having though? I'm reading back through the logs and trying to find when/why chmod is configuring the file the way it is May 17 20:56:26 but this is the first time I've built a linux image like this and I'm not getting very far finding that May 17 20:56:27 twallis: can you see where that chmod is coming from? May 17 20:58:00 chmod -w gdbm.h May 17 20:58:06 chmod: gdbm.h: new permissions are r-xrwxrwx, not r-xr-xr-x May 17 20:58:36 so, chmod -w only affects owner permissions, not all May 17 20:59:40 from the manual - If no references are specified it defaults to “all” but modifies only the permissions allowed by the umask. May 17 21:00:07 denix isn't it based on the umask? May 17 21:00:14 (w/o the 'who' part?) May 17 21:00:19 tahts why I always do ugo-w May 17 21:00:22 or similar May 17 21:00:29 fray: per the manual I cited above, yes May 17 21:00:35 it's getting called in a makefile, as -w yeah. hmm, I wonder if I ran the build scripts as root if that would make a difference. i'm running as a sudoer, but not directly as root May 17 21:00:52 so, if gdbm wants to properly change permissions, it shouldn't use chmod -w May 17 21:00:57 yup May 17 21:01:19 Right now I'm trying to find the makefile that was called so I can change that chmod call May 17 21:01:35 you should not be building as root at all, ever May 17 21:01:42 okay May 17 21:02:35 twallis: what's your filesystem you are running this on? May 17 21:03:39 just a straight ubuntu 16.04 lts May 17 21:03:57 i am however running the build on an external hard drive May 17 21:04:07 the drive was clean when I started this build May 17 21:04:19 that could have something to do with file permissions, actually May 17 21:04:32 is it fat by any chance? May 17 21:04:46 that would explain all rwx permissions for new files... May 17 21:05:15 ntfs May 17 21:05:25 not much better :) May 17 21:05:31 ya.. ya fat/ntfs doesn't have real permissions.. May 17 21:05:35 what should it be formatted as? May 17 21:05:37 i'd be amazed if that was anywhere near working May 17 21:05:39 (well ntfs 'kind of might) May 17 21:05:42 you need to use real unix fs May 17 21:06:04 so the actual mounted hard drive, not an external May 17 21:06:36 can be external, just need to be ext3/ext4/xfs/btrfs, etc May 17 21:06:52 ah May 17 21:07:02 I'll give that a shot May 17 21:07:30 thank you very much! May 17 21:07:40 if you don't have control over your hard disk partitions, you can also do a loop back (will be slower) but it's a workaround for those who are prevented by IT policy from having a working system May 17 21:07:42 twallis: so, mkfs.ext4 would be fast fix, since you said the harddrive was empty May 17 21:08:21 was, yeah. all it has on it is the build environment, so that's not too hard to replicate after a format **** ENDING LOGGING AT Thu May 18 03:00:02 2017