**** BEGIN LOGGING AT Wed Jan 16 02:59:56 2019 Jan 16 07:37:09 Hi. I am trying to build Yocto Rocko core-image-full-cmdline for arm in Ubuntu 18.10 and experiencing build failures. Has anybody seen such failures? Jan 16 07:37:39 nascentmind: which failures? if its more than one or two lines, please put a log into a pastbin Jan 16 07:41:57 LetoThe2nd, https://pastebin.com/embed_js/Qb2QtrHF Jan 16 07:43:06 nascentmind: hum. can you show a little more? so we can see which recipe fails, etc? Jan 16 07:43:28 nascentmind: and, is it a rocko poky without additional layers, on an otherwise standard 18.10? Jan 16 07:45:12 LetoThe2nd, Yes it is rocko poky without any additional layers on a standard 18.10. Jan 16 07:45:41 I see patches written for this by Khem Raj but I am not sure where it is being applied and it is incomplete. Jan 16 07:45:47 nascentmind: ok, i'll kick off a build to cross-check it Jan 16 07:52:58 LetoThe2nd, This is the recipe-> https://pastebin.com/embed_js/R7T76Rn1 I have redacted my folder name with <...> Jan 16 07:53:58 LetoThe2nd, I have also tried building using CROPS docker image. It has built properly in it but I am not sure how to do runqemu for the built image? Jan 16 07:54:22 nascentmind: no idea, i'm not using crops Jan 16 07:55:45 nascentmind: just generally, while its possible to run graphical applications inside a docker container, its probably easier to just use the nographic option of runqemu in such cases. Jan 16 08:09:48 nascentmind: kicked off a build, lets see what happens. will take some time, of course. Jan 16 08:13:42 LetoThe2nd, thanks. Jan 16 08:14:15 LetoThe2nd, using nographic option it runs but I cannot input any commands. Jan 16 08:14:37 LetoThe2nd, basically I can see the cursor blinking but no inputs are taken. Jan 16 08:19:12 nascentmind: no idea, given that description. i have personally used it successfully Jan 16 08:19:55 but i invoke my docker containers directly, so some script or crops speciality might be messing with it, maybe its misssing the -it flags Jan 16 08:20:42 nascentmind: do you have a link to khems patch that you mentioned? Jan 16 08:26:52 I'm performing ETM tracing of an embedded system running a yocto-built kernel/BSP. I'm at the point where I'd like to understand the activity of running processes, such as systemd. I'm a yocto newbie; how do I configure the builds of these programs? Jan 16 08:27:24 (The kernel is no issue--I have the necessary debug symbols in the vmlinux) Jan 16 08:31:44 Hi, "bitbake -c menuconfig linux-imx" fails with mconf.c:(.text+0x8a4): undefined reference to `stdscr' Jan 16 08:32:28 This is bugreport https://bugzilla.yoctoproject.org/show_bug.cgi?id=7609 which is set on hold due missing ability to reproduce Jan 16 08:32:30 Bug 7609: normal, Medium, 1.9, saul.wold, VERIFIED WORKSFORME, menuconfig command fails due to missing symbols Jan 16 08:32:53 It is not worksforme Jan 16 08:33:01 I found the reason Jan 16 08:33:17 you need to install 2 packages to make it work Jan 16 08:33:30 libncurses5-dev and libtinfo-dev Jan 16 08:33:52 without it it will fail Jan 16 08:35:23 I don't have an account at bugzilla. So maybe someone can add my insight there Jan 16 08:35:59 I'm not allowed to comment to that 7609 Jan 16 08:36:46 Hi guys! Jan 16 08:37:17 I am still stuck with "ERROR :cairo-1.14.12-r0 do_prepare_recipe_sysroot: The file /usr/lib/libwayland-egl.so is installed by both wayland and mesa, aborting" Jan 16 08:37:24 I've checked bbappends Jan 16 08:37:43 A searched for the library on the build dir Jan 16 08:38:01 There are several references to that library there Jan 16 08:38:18 Well, instances* Jan 16 08:39:55 All meta-whatever are in origin/thud's last commit but meta-ti (which is in master last commit, due to a lack of thud branch) Jan 16 08:41:29 New news from stackoverflow: Building keras and tensor flow on yocto Jan 16 08:54:10 Morning all :) Someone here knows how to set a task (added by bbappend on recipe) as dependencies of an other task ? Of course do_postprocess[depends] += "virtual/kernel:do_task_added_by_bbappend" doesn't work, error is : depends upon non-existent task kernel:do_task_added_by_bbappend Jan 16 09:10:17 oh I was talking to a robot (yocti) (facepalm) Jan 16 09:11:16 john_connor: haha Jan 16 09:13:16 RP: who's currently taking care of old/stable? can we consider 95ca077ab871ceff46c2052f324f879a1d624ff4 to be backported to rocko? Jan 16 09:17:05 malanecora: https://github.com/agherzan/meta-raspberrypi/issues/243#issuecomment-388374552 Jan 16 09:19:28 john_connor: Yep, checked that one! That's why it seems strange to me...mesa does remove the installation of libwayland inside do_install_append() Jan 16 09:19:46 But the library installation is still under conflict Jan 16 09:20:11 LetoThe2nd, does it crash on your end too? Jan 16 09:20:23 LetoThe2nd, sorry... Fails to build. Jan 16 09:20:47 nascentmind: i could reproduce it and khems patch actually fixes that issue, but then it runs into another failure. my advice would be to stick to a stable buiold environment for now, or update poky altogether Jan 16 09:21:33 LetoThe2nd, Exactly. I have not phrased it properly. It causes failures in other places. Jan 16 09:22:05 nascentmind: no, thats even more wrong. it does not case other failures. its just that more problems are present due to the glibc changes. Jan 16 09:22:30 Furthermore, if I can build mesa in standalone mode. The crash comes when I build cairo, which thud's recipe makes little difference with the rocko's one (I'm upgrading my environment from rocko to thud) Jan 16 09:23:17 LetoThe2nd, Thud seems to work fine in 18.10 and completes the build properly. Unfortunately I will have problems with the TI meta-layer which does not have Thud branch Jan 16 09:23:19 sry, I'm to far from understanding the whole system :) Jan 16 09:23:43 nascentmind: yeah because there have been updates all over the place in thud, thats why it works Jan 16 09:23:53 *too Jan 16 09:23:57 nascentmind: meta-ti's master does not apply to thud? Jan 16 09:24:05 denix: can you comment? Jan 16 09:26:25 LetoThe2nd: I think it does, I'm currently working with Thud in all branches with the exception of meta-ti which is synced with master Jan 16 09:27:39 No issues coming from meta-ti up to this moment (I have not completed the full build because other issues arose) Jan 16 09:44:57 Is it possible to use/run uboot-mkimage that has been built by yocto from a image recipe? Jan 16 09:45:24 I need to attach a header to a cpio file and I would like to use uboot-mkimage that was built by yocto Jan 16 09:46:10 I've seen that the kernel recipes use mkimage but I can figure out what I need to add to my image recipe Jan 16 09:46:17 Or if its even possible? Jan 16 09:46:23 BlauskaerM: sure. the only question is if you need to come with the code yourself of if there is already magic for it Jan 16 09:46:46 BlauskaerM: technically it sounds like you are actually implementing a custom IMAGE_FSTYPE Jan 16 09:47:49 Might be? I have added cpio.gz to my IMAGE_FSTYPE variable and later in the image recipe, added a function that runs mkimage Jan 16 09:47:57 Dont know if that is the way to do it Jan 16 09:48:19 But I use a static version of mkimage and I would like to use the one built by yocto Jan 16 09:48:40 "come with the code yourself" - What do you mean? Jan 16 09:49:04 BlauskaerM: my $0.2 is that you need a IMAGE_FSTYPE like cpio.gz.blauskaerm that depends on cpio.gz and the uboot tools, and does the magic as needed. let me see if i can find an example Jan 16 09:49:29 Would be gold LetoThe2nd Jan 16 09:53:06 LetoThe2nd, I have not tried to apply meta-ti. Let me check that. Jan 16 09:53:28 BlauskaerM: sorry, nothing in reach :-( closest thing is https://github.com/sbabic/meta-swupdate/blob/master/classes/swupdate.bbclass but thats already pretty advanced Jan 16 09:53:56 BlauskaerM: but it gives you the ideas, hopefully Jan 16 09:54:26 Is it possible to share a variable between bbclass and recipe ? I have tried d.setVar methode and export but the variable is empty during recipe process :( Jan 16 09:55:07 LetoThe2nd: I take all advice I can get, thank you :) Jan 16 09:55:30 In shot, I should create a bbclass that describes a new "target" for IMAGE_FSTYPE? Jan 16 09:56:42 BlauskaerM: thats the general concept, yes. Jan 16 10:04:00 LetoThe2nd: Do you think I can use this as a template? Jan 16 10:04:02 https://github.com/mendersoftware/meta-mender/blob/master/meta-mender-core/classes/mender-ubimg.bbclass Jan 16 10:05:14 BlauskaerM: doesn't look bad Jan 16 10:06:11 you'll have to check for the mkimage dependency, how to set it. but otherwise the core idea is the same Jan 16 11:10:15 When a recipe depends on other/s and executes do_prepare_recipe_sysroot (do_populate_sysroot of each dependency) does it take the files from the dependency_PN_PV/recipe-sysroot or from the dependency_PN_PV/image? Jan 16 11:12:05 I suspect that the aforementioned issue with libwayland-egl has something to do with that... Jan 16 11:12:13 LetoThe2nd: armin is looking after them but I think we may have has the last rocko stable release Jan 16 11:16:09 RP: nvm, i've seen additional breakage after that patch anyways. sry for the noise. Jan 16 11:17:37 LetoThe2nd: stable series and additional breakage is always a great combination! Jan 16 11:20:01 RP: yep Jan 16 11:20:19 * LetoThe2nd ponders faxing a mailbomb to some glibc maintainer Jan 16 11:22:41 * RP congratulates LetoThe2nd on making watchlists Jan 16 11:24:37 ? Jan 16 11:45:18 Hi, folks! Jan 16 11:45:42 Is there any way to add kernel build artifacts to SDK? Jan 16 11:46:12 I mean build directory from work/machine/linux/build Jan 16 11:46:51 acrap: kernel-devsrc? Jan 16 11:47:07 kernel-devsrc consists only sources Jan 16 11:47:24 it's not enough for kernel module compiling, for example Jan 16 11:47:40 acrap: we have tests which successfully compile modules with it Jan 16 11:48:09 Maybe I am just using invalid KERNELDIR... Jan 16 11:48:18 I use /usr/src/kernel Jan 16 11:49:34 I mean crosscompiling modules with SDK Jan 16 11:50:04 It's not about compiling on target Jan 16 11:52:11 Could you clarify that your test actually does, please? Jan 16 11:52:34 It would be helpful Jan 16 11:55:18 To compile module I need, at least - some binaries like fixdep and modpost. They must be compiled for the host architecture. Build directory I mentioned consists that, but sysroot/machine/usr/src/kernel - doesn't Jan 16 12:07:44 acrap: http://git.yoctoproject.org/cgit.cgi/poky/tree/meta/lib/oeqa/runtime/cases/kernelmodule.py tests kernel module builds work on target. We don't have an SDK test but it should work similarly Jan 16 12:15:45 I sure it works on target, but the similar instructions don't work with crosscompiling on host. "make scripts prepare" just doing something wrong and I get misconfiguration that causes module build errors. Jan 16 12:18:07 I got warnings like "warning: "__LINUX_ARM_ARCH__" is not defined, evaluates to 0" Jan 16 12:18:30 and then, obviously I get errors Jan 16 12:19:40 I can't get why it's not documented... It's a common thing to use SDK for building kernel modules, isn't it? Jan 16 12:19:45 kanavin: fancy a fun python bug? :) Jan 16 12:28:40 Ok. I'll try to formulate a my question in a different way. Is there built-in feature in Yocto to add kernel artifacts, that necessary to be present on SDK to build kernel modules? Is it described somewhere? Jan 16 12:29:21 to cross-build I mean. I am not interested in compiling on target Jan 16 12:33:48 kanavin: filed a py bug for you, if you're too busy then feel free to kick it back to me. it seemed like something you'd be good at fixing though ;) Jan 16 12:44:11 Taking a look over manifest-cortexa9t2hf-neon-mx6qdl-mesa.populate_sysroot I claim that libwayland-egl.so is not being installed by the package, is there a possibility that the build system is taking manifest-cortexa9hf-neon-mx6qdl-mesa.populate_sysroot (mesa version from Rocko) instead, thus creating a flase-positive? Jan 16 12:44:18 false* Jan 16 12:45:46 This is the python fragment that throws the error https://pastebin.com/TUgwBhh4 (from cairo's recipe) Jan 16 13:01:41 Hi guys, someone here could explain the best way to share a variable between 2 recipes ? I have tried export but ( bitbake option -e doesn't show me the var) :( thx. Jan 16 13:03:57 PinkSnake: putting it into a conf file. Jan 16 13:07:11 LetoThe2nd: thx for the tip, I have already tried in .inc, there is a diff between file extension ? Jan 16 13:07:50 PinkSnake: do you mean .ini? Jan 16 13:09:23 malanecora: no no .inc like include file :) Jan 16 13:09:52 PinkSnake: .inc , as far as I understand, is a kind of "general basis" for a given set of recipes of the same PN Jan 16 13:10:37 PinkSnake: well inc is just relevant if you actually include it in the recipes Jan 16 13:10:58 its basically just inserting its contents then Jan 16 13:31:37 LetoThe2nd: malanecora thx for info guys, i'm going to take a look Jan 16 13:34:00 I just rebased and my native perl scripts started looking for modules in /usr/lib/perl5/site_perl instead of /usr/lib/perl5/perl/site_perl (where they are installed) Jan 16 13:35:33 * radsquirrel wonders if that is his bug or a regression Jan 16 13:41:13 radsquirrel: with master? perl just changed a lot Jan 16 13:41:27 RP: yep with master Jan 16 13:43:21 I picked up ~100 patches in the rebase - noticed the perl changes in there. Jan 16 13:48:04 radsquirrel: Its likely related Jan 16 13:50:34 While measuring the number of bytes written to the filesystem is easy, I would like to analyze the number of written/erased flash (NAND) pages on my embedded device. Any pointers for this? Jan 16 13:54:02 hi. does anyone know where i can find sources for linux-2.6.25.2 , preferrably from official linux hosts ? https://mirrors.edge.kernel.org/pub/linux/kernel/v2.6/ doesn't have it unfortunately Jan 16 13:56:36 This is not suposed to happen, is it? "mesa exists in sysroot, but is stale (mesa.f4c1120c4ffec527dd97ea458991107d vs. mesa.f4c1120c4ffec527dd97ea458991107d), removing." Jan 16 13:57:43 The taskhash is the same... Jan 16 14:04:49 malanecora: partially installed? The code in staging.bbclass checks for a os.path.exists(depdir + "/" + c + ".complete") as well Jan 16 14:05:36 RP: Hmmm...It shouldn't Jan 16 14:05:54 RP: I'm building it separately Jan 16 14:06:11 RP: Successfully Jan 16 14:17:55 RP: As you pointed out, in {WORKDIR}/cairo/{version}/recipe-sysroot-native/installeddeps there is a file "mesa.{hash}" that lacks ".complete" Jan 16 14:18:36 RP: And its content does makes reference to libwayland-egl.so Jan 16 14:19:39 Piraty: https://mirrors.edge.kernel.org/pub/linux/kernel/v2.6/ Jan 16 14:20:27 mckoan: unfortunately, as i wrote, this is the link iwas looking at, yet the 2.6.25.2 isn't there Jan 16 14:20:37 (attention the fourth number) Jan 16 14:21:12 What is quite strange having in consideration that mesa's manifest.populate-sysroot doesn't Jan 16 14:22:42 Something is tricking me Jan 16 14:57:41 RP: http://errors.yoctoproject.org/Errors/Details/216830/ i amseeing this error often especially on musl + kmods Jan 16 14:58:45 khem: looks like the error message is getting lost Jan 16 14:59:11 xz is dying Jan 16 14:59:28 or opkg_build maybe Jan 16 15:07:26 khem: I think stderr is going missing Jan 16 15:50:17 My suspects have been confirmed,Yocto was picking the wrong manifest Jan 16 15:50:36 In spite of building the right package version Jan 16 15:50:51 It was a false positive Jan 16 15:51:57 The build completed successfuly after remove the manifest files related to the rocko's mesa version Jan 16 15:52:49 Someone here knows how to add a recipe process after the image creation ? i don't know where I have to search this info :S Jan 16 15:53:16 Is this the expected behaviour? Jan 16 15:53:48 PinkSnake: What are you exactly trying to do? Jan 16 15:54:58 I have to pack a FIT inside a custom container, but i have to build all the image before repack... Something like post process command but for an recipe image :) Jan 16 15:55:34 sounds, again, like a custom IMAGE_FSTYPE Jan 16 15:56:46 Indeed Jan 16 15:57:25 LetoThe2nd: for sure yes, my work is based on kernel-fitimage.bbclass Jan 16 15:59:25 the trouble is in this class we can only add initrd and i want to put all my cpio.gz and bitstream so I made a custom recipe to generate FIT image, but i can't share image output name because the variable is not set during this part of build process Jan 16 15:59:52 PinkSnake: so then, introduce something like fit.custom that bundles the magic you need. do it as a class, respectively IMAGE_FSTYPE instead of bundling it into the image recipe Jan 16 16:00:22 RP: so tinfoil API has a global lock when used by multiple processes? :( I split the recipe upgrade status function to run from a process pool, and they seem to each take a turn, instead of actually running at once, so there is no speedup Jan 16 16:00:51 LetoThe2nd: thx, i have already a fit-image.bbclass ;) Jan 16 16:01:34 LetoThe2nd: The FIT image is correctly generated, the trouble may come from meta-xilinx but i'm not sure... Jan 16 16:02:58 PinkSnake: usually digging through bitbake -e tells you pretty exactly whats going on Jan 16 16:04:34 kanavin: the multiple processes would have to be server side Jan 16 16:04:51 kanavin: a command along the lines of "for each recipe call this function" Jan 16 16:05:25 kanavin: would have to be a specific tinfoil command Jan 16 16:07:06 RP: yep, that's what I am trying on the client side, "for each recipe call this function from an available process in the pool". The function creates its own instance of Tinfoil every time, but they seem to mutually exclude each other, so only one process gets to execute anything at a time. Jan 16 16:07:53 kanavin: it would need to be a something run server side which holds the lock Jan 16 16:08:13 kanavin: having more than one server against a given directory would be carnage Jan 16 16:13:24 RP: right, I guess tinfoil does not have any way to define and run 'tasks'? Jan 16 16:13:33 this was sooo much faster using them Jan 16 16:28:42 LetoThe2nd: i totaly agree but the output said: " your variable is not set " Jan 16 16:29:25 it's normal in fact because I want to use the output of image name before the variable is set Jan 16 16:30:29 kanavin: well, we can go back to using a class but I still think we could create API to tinfoil to do this without that overhead. With a class you have to collect up the data somehow :/ Jan 16 16:31:06 RP: no going back to class, that csv business was ugly :) Jan 16 16:31:39 and yeah, the only way tasks can communicate is via writing to files really, which isn't great Jan 16 16:32:28 kanavin: how about pulling the data stores out of tinfoil serially then passing them off to threads to do the tests? Jan 16 16:33:07 you might have to copy the variables the fetcher uses locally I guess :/ Jan 16 16:33:18 kanavin: I hated that csv code Jan 16 16:33:27 sadly, the 'pulling data out of tinfoil' is deeply embedded into the functions that do the upstream checking Jan 16 16:34:20 kanavin: but you could create a local data store with the values the later (slow) code needs Jan 16 16:35:30 kanavin: those functions just need a datastore, it can be a thread local instead of remote one Jan 16 16:35:44 RP: ah, so a kind of 'fake' dict-based store, instead of actual data store? Jan 16 16:36:09 kanavin: well, doesn't have to be fake, just a local non-tinfoil one with a subset of values Jan 16 16:36:27 RP: so how does one make such a store? Jan 16 16:37:02 kanavin: bb.data.init() ? :) Jan 16 16:37:13 oh Jan 16 16:37:16 (then setVar) Jan 16 16:37:39 right. I guess that could work :) Jan 16 16:38:05 I'm thinking out loud, not sure how many variables you;d have to copy in Jan 16 16:39:44 Hi Jan 16 16:52:54 Hi all, we have a common download location (/opt/yocto/downloads) for yocto, what I want is for specific recipes the download location to be something else for users is this possible ? Jan 16 17:02:29 I'm kinda stuck. I'm working on a local-only kernel patch and I want my kernel to fetch my local git repo. Isn't the syntax SRC_URI_pn-mykernel="file:///home/user/mykernel.git;branch=mykernel" ? I'm ending up with an empty ${MACHINE}/kernel-source/ dir after unpack. Any ideas anyone? Jan 16 17:04:37 The log.do_fetch are run without any errors / run.do_fetch, shows that bb.fetch2.Fetch() is used, althou it is suprisingly quiet in this instance Jan 16 17:06:13 kanavin: https://autobuilder.yoctoproject.org/typhoon/#/builders/85/builds/124 is the gl patchset on the AB Jan 16 17:06:51 RP: thanks, I am doing further work, particularly looking into egl-headless option, which would not require X, or gtk Jan 16 17:07:04 you just vnc connect to qemu and voila, in theory Jan 16 17:07:26 sveinse: I'd check SRC_URI is set to what you think it is Jan 16 17:07:41 kanavin: fair enough. I think there are some things in there like vte-native we may need to fix generally Jan 16 17:08:09 RP: ouch, a sea of red Jan 16 17:08:18 kanavin: right :/ Jan 16 17:08:52 kanavin: didn't look too far but vte-native looks like its failing everywhere Jan 16 17:09:32 yep Jan 16 17:09:40 but not here locally :-/ Jan 16 17:10:17 RP: I have. Is there a way to specify I want to use git against a local file? Because what I think is happening is that SRC_URI is interpreting the file as a file, not as a git repo and is ending up not using git. Jan 16 17:10:26 I tried setting protocol=git, but to no avail Jan 16 17:11:18 sveinse: oh, right, yes. you want git:///home/user/mykernel.git;branch=mykernel;protocol=file Jan 16 17:18:28 RP: i can reliably reproduce the problem with opkg backend Jan 16 17:18:51 RP: perfect, thank you very much. Appreciate it. -- If I ever so humbly may say so, I'm not too fond of the special SRC_URI syntax scheme. It is often confusingly similar to git URLs, but then again it isn't. When working with layers, the layer subrepos (which is not a bb thing) are configured, it is configured with git URLS, but in recipes one needs to change your mindset to the URLs. I mean this in best Jan 16 17:18:52 RP: I wonder if poky should change one of platforms to use opkgs Jan 16 17:18:57 intentions, its just confusing. Not trying to bite of the hand and so on... Jan 16 17:19:05 khem: we do build them :/ Jan 16 17:20:14 sveinse: I agree its far from perfect and its had a parallel evolution to other things in different spaces, its a tough one Jan 16 17:20:35 It is Jan 16 17:21:51 khem: is it worth putting a try:/except: around the subprocess call and put the exception into bb.warn() ? Jan 16 17:22:04 khem: I'm sure we're missing data somehow Jan 16 17:25:11 RP: and its a new error, so maybe something changed in opkg in core Jan 16 17:25:24 I think we switched compression didnt we Jan 16 17:25:46 khem: we upgraded and used their patch instead of ours Jan 16 17:26:12 khem: you should talk to adelcast Jan 16 17:26:26 hmm Jan 16 17:26:28 see http://errors.yoctoproject.org/Errors/Build/74991/ Jan 16 17:26:47 all the packages which fail in do_package_write_ipk Jan 16 17:26:50 fail at same place Jan 16 17:27:12 right, but what was the actual error? Jan 16 17:28:30 RP: you can reproduce it in poky too with opkg backend and bitbake cryptodev-module Jan 16 17:29:16 khem: just tried locally, built fine Jan 16 17:29:28 which arch ? Jan 16 17:29:50 khem: qemux86-64 Jan 16 17:29:54 hey khem, I am seeing your error log....is this the error you are talking about? http://errors.yoctoproject.org/Errors/Details/216898/ Jan 16 17:30:15 I see, yes Jan 16 17:30:17 khem: I maintain we need to fix that output so we get the real errors Jan 16 17:31:02 for some reason its not able to build the ipk for kernel module other ipks in that recipe built ok Jan 16 17:31:13 yeah I think we need more debuggable info Jan 16 17:31:39 mmm, how can I reproduce it? Jan 16 17:33:21 adelcast: I use archlinux as buildhost which has python = python3 Jan 16 17:33:57 I wonder if thats in play but then I saw it fail on ubuntu 14.04 as well for qemumips and musl Jan 16 17:34:15 RP: maybe thats another test to try for you :) use TCLIBC = "musl" Jan 16 17:34:39 RP: and use qemuarm Jan 16 17:34:46 yeah, I am on arch too... Jan 16 17:35:09 adelcast: cool, first of all congratulations on using a proper distro Jan 16 17:35:20 ah, I see, on poky and bitbake cryptodev-module Jan 16 17:35:21 hehehe Jan 16 17:36:21 secondly, to reproduce it you can use Yoe Distro as well https://github.com/YoeDistro/yoe-distro Jan 16 17:38:12 and use . ./sama5d27-som1-ek-sd-envsetup.sh && DISTRO=yoe-musl-sysvinit-wayland bitbake can-isotp Jan 16 17:39:21 RP: should I put try..catch around subprocess.check_output in meta/classes/package_ipk.bbclass line 230 Jan 16 17:39:39 it does have stderr=subprocess.STDOUT Jan 16 17:39:48 so I wonder why its eating up info Jan 16 17:40:18 ok, I am trying with Yoe disto.... Jan 16 17:40:40 adelcast: cool Jan 16 17:55:57 now i'm tempted to abandon debian for arch linux Jan 16 17:59:20 kroon: rolling distro is the future in my mind, time based releases are not scalable for todays connected world with chances of so much securities issues etc. Jan 16 17:59:58 ask people who were involved in porting the spectre and meltdown to older kernels Jan 16 18:00:15 its a very expensive and regressive excercise Jan 16 18:00:54 that money is well spent on testing and making sure that older hardware keeps performing on newer software Jan 16 18:01:06 yeah, rolling distro is nice Jan 16 18:01:35 khem: qemuarm musl built fine too Jan 16 18:01:35 khem: what makes a rolling distro? Isn't an updated distro technically rolling as new updates are being pushed? Jan 16 18:02:31 khem: them problem arises when a new version is released, then users are forked in two camps: The ones who dist-upgrade to newer version (=rolling-ish), and those who don't. Jan 16 18:02:52 Having an rolling-update only would make it hard for the latter group. Jan 16 18:05:00 yes its a mindset Jan 16 18:05:19 collect technical debt and pay off later, or pay now Jan 16 18:05:51 but when you pay now, its in small installments Jan 16 18:06:59 RP: so I introduce try except and it starts to work http://ix.io/1ytQ Jan 16 18:10:06 well no, it doesn't beging to build, but hide the existing error Jan 16 18:11:17 also, I have found arch to have way better quality than what I thought....I do system upgrades ~1 a month and they haven't been too painful (most times they are not painful at all) Jan 16 18:11:43 and when there are problems, the forums are really really good Jan 16 18:12:05 occasionally upgrades require a manual tweak, but not too bad, and usually the news on archlinux.org covers it anyway.. Jan 16 18:13:12 yeah, or via the arch-announce mailing list Jan 16 18:13:36 At least you're privileged with users/customers that are online all the time. We're operating in a marked where internet generally isn't available. This way once-a-month scheme goes south... Jan 16 18:23:03 if you have no internet connectivity thats a different usecase but fading away case in future Jan 16 18:23:11 most devices will be online and connected Jan 16 18:23:24 atleast if iot has to succeed Jan 16 18:27:55 hey all, getting an error trying to build an SDK image with "dev-pkgs" enabled Jan 16 18:28:13 usbmuxd-dev error: jsoncpp-dev-1.8.0-r2 conflicts with jsoncpp = 1.8.0-r2 Jan 16 18:28:19 needless to say this makes no sense :) Jan 16 18:30:42 ebolton: you can check the contents of these packages and their deps in build area for jsoncpp Jan 16 18:33:27 @khem: not sure how that helps....I think this is the root error, I have a recipe that DEPENDS on jsoncpp, and it generates this when I try to create the SDK image Jan 16 18:33:30 Can't install guidance-protocol-dev-git+0+0c4944dee7-r1@core2_32: Can't install jsoncpp-dev-1.8.0-r2@core2_32: no package provides jsoncpp = 1.8.0-r2 Jan 16 18:33:40 how does jsoncpp-dev not know about jsoncpp? Jan 16 18:35:39 I'm currently running a concept evaluating if running docker on an embedded imx6 arm is a viable route. And to my surprise it works! I can pull stock docker images and run them on this yocto device. Jan 16 18:35:46 also, don't mean to be terse...been a long week...I'm about to strangle bitbake Jan 16 19:12:09 khem: it reproduced, lemme dig Jan 16 19:12:31 RP: http://lists.openembedded.org/pipermail/openembedded-core/2019-January/277944.html is still waiting Jan 16 19:12:37 adelcast: superb Jan 16 19:13:37 ebolton: can you try adding ALLOW_EMPTY_${PN} = "1" to jasoncpp recipe Jan 16 19:22:14 ebolton: sounds like that recipe is bust Jan 16 19:40:54 @rburton: is a bit old...krogoth...could be @khem: will try, thx Jan 16 19:42:30 krogoth is more than a bit old ;) Jan 16 19:44:11 is amazing now long industrial hardware sticks around ;) Jan 16 19:47:23 indeed Jan 16 19:56:59 Hmm, https://github.com/openembedded/bitbake/compare/master...kergoth:yocto-12987-new seem sane? rburton? RP? i'll obviously send an RFC email, so feel free to ignore me for now, just looking to make sure i'm not crazy Jan 16 19:57:04 * kergoth gets more coffee Jan 16 20:00:36 khem, the problem is that the package name is not adhering to Debian standards: Jan 16 20:00:50 *** Error: Package name kernel-module-can-isotp-4.14.73-linux4sam_6.0 contains illegal characters, (other than [a-z0-9.+-]) Jan 16 20:01:08 It has an _, which is not allowed (https://www.debian.org/doc/debian-policy/ch-controlfields.html#s-f-source) Jan 16 20:01:27 the dpkg backend should complain too Jan 16 20:01:47 ah Jan 16 20:01:55 but why does it not show Jan 16 20:02:03 on stdout Jan 16 20:03:00 kergoth: I am on my 16th day without coffee, they say you need 21 days to shed a habit Jan 16 20:03:31 I'm torn, I hate being hooked on caffeine, but i love the taste of coffee, and decaf limits a lot of your options.. Jan 16 20:03:54 kergoth: you feel a lot better without caffaine Jan 16 20:03:59 I promise you Jan 16 20:04:06 so i'm currently focused on just limiting the amount. one coffee when i wake up, one in the afternoon, tops. ideally just the morning would do Jan 16 20:04:21 i wishi liked the taste of tea better, its caffeine hits me less hard, no jitters Jan 16 20:04:47 opkg-build is sending the message to standard error Jan 16 20:04:58 echo "*** Error: Package name $pkg contains illegal characters, (other than [a-z0-9.+-])" >&2 Jan 16 20:05:08 I realised that you need to trick the brain, so the secret is, its the process which has a significant contribution to brain wanting it rather than the substance itself Jan 16 20:05:36 kergoth: so I do the same steps except I dont prepare coffee but herbal tea :) Jan 16 20:05:49 and my mind is tricked you see Jan 16 20:05:56 adelcast: do you see it on cmdline ? Jan 16 20:06:05 I enjoy the routine of it and taste of it. the warm beverage while relaxed in the early morning, looking outside if it's light enough. tea would provide that without the caffeine though Jan 16 20:06:34 adding a proper morning routine to kick the day off with intentional slowness has been hugely beneficial in my life Jan 16 20:06:36 kergoth: yep, its the process that you need to keep intact not substance Jan 16 20:06:42 yeah, I bitbake -cpackage_write_ipk, then ran the opkg-build command manually Jan 16 20:06:43 starts without the "rushing" all the time Jan 16 20:06:46 which i like Jan 16 20:06:55 * kergoth not a fan of pointless busyness Jan 16 20:07:12 adelcast: so we still have a bug in bbclass where this message is not pumped into stdout Jan 16 20:08:06 yes, I am looking at subprocess.check_output Jan 16 20:09:19 adelcast: it seems we need to capture the output and then send it may be Jan 16 20:09:50 kergoth: I see that writing it down helps to offload Jan 16 20:10:05 * khem is at verge of starting the preaching session Jan 16 20:10:33 yep, that is it, lemme confirm... Jan 16 20:18:29 kergoth: that patch looks sane to me Jan 16 20:47:20 kergoth, khem: I cut caffeine out for about 9 months but found I do miss the tea forcing me to take breaks and I just like drinking it.Don't like coffee though Jan 16 20:48:57 https://twitter.com/whitequark/status/1085627481518952449 Jan 16 20:51:28 @khem @rburton: apologies guys, I thought jsoncpp was a recipe in poky/meta, it was in our own company layer, and the recipe was placing the headers in the base package, not the -dev package... Jan 16 20:51:40 problem solved Jan 16 20:53:26 ok Jan 16 21:29:48 I have custom BSP layer using a 32bit pentium-M with a custom recipes-kernel/linux/linux-yocto-custom_3.14.bb and defconfig. If porting all OS setup recipes to an ATOM-E3845 (keep 32 bit), is it recommended to just update recipes-kernel for latest kernel? And what is recommendation to generated a new defconfig? How do I acquire the default defconfig? Thanks for expert opions. Jan 16 21:30:43 opinions. ;) I need coffee too. Jan 16 21:31:40 I want a newer kernel since it's been locked at 3.14 due to old HW constraints. Jan 16 21:45:07 the defconfig for any kernel should be in the build dir under tmp/work....the source is kept elsewhere post-krogoth (work-shared, not work) Jan 16 21:46:00 make sure to look for hidden files if using a GUI based file browser (the name should be .config) Jan 16 21:48:58 Has anyone been able to add config fragment support to kernel recipes that don't inherit from yocto-kernel? Jan 16 21:49:13 ebolton: Thx. Do you agree with just updating the kernel recipe for this BSP layer? The system is headless (no UI). Jan 16 21:50:45 @mrk377: depends on whether or not you've modded the kernel and what drivers you need, headless def makes things less complicated because video card drivers are a nightmare Jan 16 21:51:48 The kernel was NOT modified. It was just locked to 3.14 for a vendor kernel module that had to be compiled. Jan 16 21:53:22 @mrk377: then you should be fine...you'll have to compare the configs and enable drivers you care about in the new kernel (a defconfig from 3.x almost certainly won't work unmodded in a 4.x kernel) Jan 16 21:53:33 other than that....bitbake core-image-minimal and give it a spin :) Jan 16 21:54:32 khem, I think I have a patch ready, I'll send it to the list and you can comment there? Jan 16 21:55:53 sure Jan 16 21:59:58 @cdgarren: does the kernel recipe inherit from kernel.bbclass? Jan 16 22:02:04 ebolton: Thx. I have a bitbake targets (i.e. core-image-xyx and core-image-xyz-dev). Yeah, I don't want to miss some setting/options when jumping 3.14->4.14 in the machine xyz.conf file or other new options. Jan 16 22:02:18 @cdgarren: disregard, I'm not 100% sure where the config fragment code lives Jan 16 22:03:30 @ebolton: From what I can tell, the config fragment stuff lives in kernel-yocto.bbclass Jan 16 22:04:14 @cdgarren: I couldn't find the code in krogoth, but you may be right...can you not inherit that class? Jan 16 22:04:57 @ebolton: I'm working in sumo, not sure when it was added. Inheriting it wasn't just a drop in fix, so it may need some real work to make it happen. Jan 16 22:05:28 #cdgarren: it's an OLD feature...I was using it for a 3.10 kernel in dizzy :) **** ENDING LOGGING AT Thu Jan 17 02:59:57 2019