**** BEGIN LOGGING AT Thu Feb 23 03:00:03 2017 Feb 23 06:35:57 hi, how can we reduce boot time in yocto ? Feb 23 06:42:48 Is there some way to reduce boot time in yocto ? Feb 23 07:03:48 anyone made a usb boot with intel cpu ? I get a blinking cursor. Feb 23 07:25:28 Is it possible to upgrade kernel in local.conf with older yocto branches (daisy for example) ? Feb 23 07:27:24 Is it possible to upgrade kernel in local.conf with older yocto branches (daisy for example) ? Feb 23 07:36:28 ranchu: in local.conf you could select a preferred version, but you still need a layer that provides the corresponding recipe. so the correct answer is: not by only using local.conf exclusively Feb 23 07:39:30 LetoThe2nd - Hi LetoThe2nd , why do I need corresponding recipe ? Isn't local.conf preferred version suffecient ? Feb 23 07:41:46 ranchu: well if you set the preferred version to something, lets say x.y, then that does not automagically tell bitbake how to make that. per preferred version, you can only select things that are known to bitbake (e.g., have a recipe Feb 23 07:43:30 Ok, I think I understand Feb 23 07:43:32 thanks Feb 23 07:44:07 by the way, is there something in filesystem that depens on kernel version, except kernel modules ? Feb 23 07:44:28 ranchu: not necessarily, but possibly. Feb 23 07:45:08 possible becuase some libraries depends on specific kernel versions (probably becuase of syscalss api) ? Feb 23 07:45:27 its more common that things rely on specific kernel features to be present, which in turn can also mean "newer then" if the features are rather recent Feb 23 07:46:21 and yocto takes cares for all these dependencies between libraries and kernel (kernel features, kernel syscalls, etc) ? Feb 23 07:48:00 LetoThe2nd - Does Yocto takes cares for all these dependencies between libraries and kernel (kernel features, kernel syscalls, etc) ? which means that it might upgrade libraries if kernel is newer ? Feb 23 07:48:54 ranchu: yocto itself cares for nothing, you probably mean OE, or poky as the reference distribution :-D Feb 23 07:51:13 yes, I mean OE, or poky... :) they should take care of these dependencies Right ? Feb 23 07:52:05 ranchu: and the build process again will not automagically update any libraries. you can prefer/request a certain version of something, and if it finds all the needed recipes for the versioned dependency chain, then those things get built. it will not by itself change/update any version, you always have to provide the specific version recipe Feb 23 07:54:33 if this is so, it sounds a real endeviour trying to upgrade kernel..... if dependencies are not automagically being taken care of... Feb 23 07:54:42 in the end thats the point of the releases. to have a coherent set of the infrastructure. Feb 23 07:54:54 if this is so, it sounds a real endeviour trying to upgrade kernel..... if dependencies are not automagically being taken care of... Feb 23 07:55:48 ranchu: not at all, because the kernel-userspace interface is well known. things become only nasty if some binary kernel magic is involved, or versioned non-opensourced drivers, or things like that. Feb 23 07:56:34 i have dora, dizzy, krogoth, morty, all on the same kernel version maintained for my projects. effort is zero. Feb 23 07:59:20 I mean that if for updating kernel it is not just a matter of changing PREFERRED_VERSION_linux-yocto , and I also need to create recipes, than it sounds quite heavy Feb 23 07:59:32 very Feb 23 08:00:29 ranchu: linux-yocto is only provided in specific versions anyways. and it is not the vanilla upstream kernel. Feb 23 08:01:18 Heya Feb 23 08:01:27 I understand, Ok. Thanks LetoThe2nd Feb 23 08:01:44 again, it depends very much on your userland. if it is well behaved, and your kernel is too, then just slap in a linux_kernel_custom with your version and source, and be done. Feb 23 08:01:50 Does anyone know how to install deb files with something like apt-get in a yocto image? Feb 23 08:03:39 maka_: what is it that you want? either you get something as a deb and want to write a recipe for it, or you want your build output to be deb formatted, and support packagemanagement at runtime? Feb 23 08:04:28 preferably i want a recipe to put it in the build while making the image. But i couldnt get that to work sadly. Feb 23 08:05:00 well then why deb? why not proper source upstream? Feb 23 08:05:07 What i'm trying now is installing the deb package after making the image with apt-get however i get some errors: Couldn't find any package by glob/regex Feb 23 08:05:19 Because the packages that were provided to me are debs ._. Feb 23 08:06:02 if thats all that you get, then you are up for some ugly unpacking and custom installation scripting in your recipe Feb 23 08:06:36 plus the fact that you will have fun in getting the arch things right Feb 23 08:06:36 I thought so... the problem is, i have no clue how to make that recipe as i am quite new to both yocto and linux Feb 23 08:07:19 maka_: basically it is this: http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#packaging-externally-produced-binaries Feb 23 08:07:34 I am however able to get some tar.bz2 packages for some of the debs, would that be easier to install? Feb 23 08:08:02 maka_: usually tarballs are easier because you don't need manual unpacking action Feb 23 08:08:29 Because i tried that aswell before, but i kept getting errors that the feed was empty or things like that Feb 23 08:09:48 maka_: then probably your install steps were faulty, and therefore created empty packages. Feb 23 08:10:35 maka_: what you describe about always means manual do_install scripts to be written. Feb 23 08:10:47 Hmm, im pretty sure thats it yea, however i have no clue how that do_install function works ._. Feb 23 08:11:11 maka_: thats why i handed you the documentation on writing recipes. Feb 23 08:11:24 i'm reading through that right now ^^ Feb 23 08:11:45 maka_: here even is a super trivial example for a do_install http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#new-recipe-single-c-file-package-hello-world Feb 23 08:14:12 the weird thing is, i also tried installing de tar package on ubuntu normally, it had some steps like ./configure and make, make modules and stuff like that. But even on ubuntu that seemed to fail Feb 23 08:14:47 * LetoThe2nd sighs. Feb 23 08:15:04 so the tarball is source, and the deb is binary probably. Feb 23 08:15:30 The tarball is indeed source, no clue about the deb though Feb 23 08:15:35 in that case look at the source, find if it is autotoolized or using some other build system, and then apply the corresponding documentation Feb 23 08:24:46 LetoThe2nd: ok, so so far it can find the file(the tar.bz2) and unpacks it, however at this part "install -m 0755 helloworld ${D}${bindir}" it has some error, ofc i changed the helloworld, but what exactly should be there? Feb 23 08:25:45 maka_: that basically copies whatever you want over to the target. Feb 23 08:25:46 https://linux.die.net/man/1/install Feb 23 08:27:06 hmm, but in that case i have to do this differently right? since there are makefiles and a ./configure Feb 23 08:30:19 hi! is it possible for two recipes to share the same work directory? Feb 23 08:31:05 I have a case where one recipe needs the generated configuration files of another recipe Feb 23 08:32:00 maka_: hm. did you actually read the documentation beyond the headline that i gave you? Feb 23 08:32:16 I did, however i dont really get it yet ._. Feb 23 08:32:42 maka_: there is examples for things that come with makefiles, and things that come autotoolized (and a ./configure is often, but not necessarily a hint that a source package is autotoolized) Feb 23 08:33:15 maka_: so what packages are this. is it something that i actually could look at, or some arcane closed whatever stuff? Feb 23 08:35:27 The one im trying to get to install now is found here: etherlab.org/download/ethercat/ethercat-1.5.2.tar.bz2 Feb 23 08:36:19 maka_: ok, so stuff with kernel modules and all. Feb 23 08:38:19 you certainly are in for a hard ride. my first suggestion would be to have a long and thorough look at what they did there, and if the layer procide something for you to recycle: https://layers.openembedded.org/layerindex/recipe/59739/ Feb 23 08:39:53 "you certainly are in for a hard ride" Not what i was hoping (._.')7 Feb 23 08:40:35 hope has nothing to do with proper code. Feb 23 08:41:00 Fair enoug Feb 23 08:41:01 h Feb 23 08:41:02 you've got a 4 year old unmaintained tarball that needs kernel modules. Feb 23 08:41:49 thats just like the definition of "pain" Feb 23 08:42:18 And thats only 1 of the packages ^^ Feb 23 08:43:06 let me guess, university project? ;-) Feb 23 08:43:30 Internship Feb 23 08:44:41 something like "here is the board, make all this stuff run on it. oh and yes, you gotta use this yocto linux, we've heard it is cool!"? Feb 23 08:45:00 or is there actually somebody competent to guide and teach you? Feb 23 08:45:05 Nopeee Feb 23 08:45:12 what nope? Feb 23 08:45:20 They heard about it, looked at it for 30 minutes and thought, thats fun, we could use this! Feb 23 08:45:31 Go find out if it works with these packages! Feb 23 08:46:11 And no one really knows how it works sadly, so i don't really have anyne to guide me... Feb 23 08:46:18 then better stop now and tell your boss "it certainly can be made work, but needs serious effort." and point out that you just don't have the necessary knowledge. Feb 23 08:46:37 Would it really be that hard to get to work yea? Feb 23 08:47:56 for someone knowing linux, cross compilation and openembedded, my estimate would be a couple of days, if there is nothing major to be patched in the source. Feb 23 08:49:01 what things do you already know, and what are you expected to learn? what is the timeframe? Feb 23 08:50:29 Well my knowledge about linux is limited to some raspberry pi stuff, i know the basics of yocto a little now, and i have about 5 months Feb 23 08:51:01 what kind of board is this? Feb 23 08:51:31 It's supposed to run on some embedded pc Feb 23 08:51:40 thats no answer Feb 23 08:51:45 something with an intel cpu, thats all i know Feb 23 08:52:00 is it a one of a kind? Feb 23 08:52:07 i don't know yet Feb 23 08:52:27 because then better just slap debain on it, manually install everything and beat onto it until it works Feb 23 08:53:12 That would be better yea Feb 23 08:53:30 OE is cool if you need to maintain things in the long run. if you need to be able to reproduce the whole system days, weeks, months or even years later if soemthing need to be changed Feb 23 08:54:02 or if you are on targets and in environments that are not supposed to have package management and all that Feb 23 08:54:28 if its just for a one of a kind test drive, better do the standard desktop thing. (my $.02) Feb 23 08:54:54 It has to work for a long time tho, its not just a test or something Feb 23 08:55:55 that sound not very well considered. Feb 23 08:57:33 i condense the situation to: "we have an intern for 5 month, no clue about OE, about linux a bit maybe, and he shall take this huge pile of partially arcane, old, and unmaintained software to deliver a reaady-to-use-product." Feb 23 08:57:57 if that sounds about right, tell your supervisor that this will not fly. Feb 23 08:58:47 I'll mention it next meeting .-. Feb 23 08:59:39 in 5 months you can get a solid grasp of cross-building software and OE. probably also make one or two packages work. Feb 23 08:59:55 more if there is someone to guide and teach, less if you are left all alone. Feb 23 09:00:28 It's the latter one sadly Feb 23 09:01:13 well then Feb 23 09:02:48 sounds like you need to compile a bit of information how things generally work, and then have some serious conversation Feb 23 09:04:45 Probably... Don't really wanna fck up internship though ._. Feb 23 09:05:18 maka_: you can also bury yourself at your desk and let things blow up the last day. Feb 23 09:06:28 Yea, fair enough bit still haha Feb 23 09:06:30 but* Feb 23 09:06:38 in my experience, it is ok the give impossible tasks to an intern. but it needs to be clear that the target is then "go as far as you can, and we'll talk regularly about the progress, problems, and things you learned." Feb 23 09:06:56 I really do hope that will be the case here Feb 23 09:06:58 like, discussing your proceeding with the supervisor once a week or so. Feb 23 09:08:53 JaMa: You around? got a query regarding the QEMU error you hit. Are you enforcing -Werror? Feb 23 09:20:17 LetoThe2nd: Either way, Thanks for the info! Feb 23 09:20:32 good luck Feb 23 09:21:43 Thanks ^^ Feb 23 09:40:53 nrossi: no I'm not enforcing Werror, see the patch I've sent Feb 23 09:53:11 JaMa: I did see it, the thing is i would have expected that the nativesdk build to see the same error since it it using glibc 2.25 but its just a warning :| Feb 23 09:54:44 JaMa: And the thing is that the default configure for QEMU doesn't enable Werror unless you are at least using it from a git repo (of which the _2.8.0.bb recipe is using a tarball) Feb 23 10:04:26 What's the prefered way to express `systemctl enable wpa_supplicant@wlan0.conf` in a recipe? Feb 23 10:35:09 hmm, patchwork seems to have missed one of my patches: Feb 23 10:35:13 http://lists.openembedded.org/pipermail/openembedded-core/2017-February/133241.html Feb 23 10:35:16 is that common? Feb 23 10:35:31 nrossi: nativesdk is probably also using patched autoconf openembedded-core/meta/recipes-devtools/autoconf/autoconf/AC_HEADER_MAJOR-port-to-glibc-2.25.patch which many hosts don't have Feb 23 10:36:11 marquiz: common in last week, but should be fixed now http://lists.openembedded.org/pipermail/openembedded-devel/2017-February/111470.html Feb 23 10:42:38 JaMa: oh, thx, that thread had avoided my radar somehow Feb 23 10:58:39 rburton: I've just sent an explanation of the AB build failure. Would you mind to pull my patchsets back? Feb 23 10:59:50 ed2: ah, ok. joshuagl, can you look at ed's latest reply to the wic thread? there's an AR for you in there :) Feb 23 11:00:55 rburton: it took me awhile to figure out that hidden MACHINE change. That's quite confusing. Feb 23 11:01:20 rburton: ed2: sure, I'll take a look. These things are usually better tracked as bug reports Feb 23 11:01:45 joshuagl: yep, I'll create a bug. Feb 23 11:02:36 thanks ed2 Feb 23 11:08:54 Heya! Feb 23 11:09:22 I was wondering, if i have some specific make commands needed for a package, how do i add those to the recipe? Feb 23 11:11:48 maka_: in do_compile() (which by default is basically "make", and can be seen in base.bbclass) Feb 23 11:12:32 And how exactly would i have to add it there? can i just type like: make nameOfFile ? Feb 23 11:18:38 maka_: yes Feb 23 11:21:03 I see in usb iso image both syslinux and grub.cfg. does it mean both are used ? Feb 23 11:21:31 hi Feb 23 11:23:27 I see in usb iso image both syslinux and grub.cfg. does it mean both are used ? Feb 23 11:26:59 rburton: It actually compiles now! However i get: i386:x86-64 architecture of input file `file/path' is incompatible with i386 output Feb 23 11:29:57 maka_: as that's not an error that our tooling produces, i can't really help Feb 23 11:30:39 Thats alright, just said i got it, trying to fix it now, something in the package should change to x64 instead of x86, at least that swha ti found Feb 23 11:30:44 thats what i found* Feb 23 13:01:46 What's the difference between hddimg and iso ? Feb 23 13:02:54 I manage to boot with rufus burn of iso, but not with linux "dd" of iso. I don't understand the difference. Feb 23 13:03:49 and hddimg is a disk image with partitions and a boot loader Feb 23 13:04:15 rburton - the sentence seemd to be cut Feb 23 13:05:03 i meant 'an hddimg' Feb 23 13:05:39 what's the difference between hddimg and iso ? Feb 23 13:05:46 iso also have bootloader I think Feb 23 13:09:39 hddimg is a disk image with a partition and boot loaders, and live boots a read-write image from the disk image Feb 23 13:09:54 and iso is a cd image Feb 23 13:12:21 Any chance i could get Gdebi package manager on yocto? Feb 23 13:12:50 maka_: sure, write a recipe Feb 23 13:22:44 rburton: I'm not seeing my version update patchbomb in patchwork, shall I resend? Feb 23 13:22:53 kanavin: second time i've been told that Feb 23 13:23:04 kanavin: no, tell lsandov instead :) Feb 23 13:23:15 i'm not relying on patchworks right now Feb 23 13:23:24 (as in my primary workflow is still mail) Feb 23 13:23:39 rburton: I think ppl have been already complaining, so no need for extra Feb 23 13:25:27 rburton: oh, and: shall we throw the dnf patchset on more than x86_64 ab? Feb 23 13:25:42 rburton: like, all of them at once, so I have no time for explorathon :D Feb 23 13:26:19 the last one was all green btw Feb 23 13:26:34 kanavin: yeah i can throw it across the hole buildsets Feb 23 13:27:05 what was the branch name? Feb 23 13:27:31 https://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/log/?h=akanavin/dnf-rpm4 Feb 23 13:27:52 kanavin: https://autobuilder.yocto.io/tgrid Feb 23 13:39:46 wget works, but firefox does not - over the same proxy.ir.intel.com :-/ Feb 23 13:39:50 something on my local setup then Feb 23 13:40:06 ah, wrong channel :) Feb 23 13:55:34 rburton: is this serious? what does it test for? https://autobuilder.yocto.io/builders/nightly-distrodata/builds/10 Feb 23 14:04:06 saul asked for that to be enabled but it doesn't seem especially useful in its present form, patches or bug reports welcome Feb 23 14:04:38 joshuagl: so it's expected to fail, and I should not be looking further? Feb 23 14:04:52 it's not expected to fail, but the output is not very useful Feb 23 14:06:02 joshuagl: what I mean is that it's not something I broke in the dnf branch that makes it fail Feb 23 14:06:50 all 10 runs have failed so that seems unlikely Feb 23 14:09:29 kanavin: I think the dnf branch may have contributed, just teasing out the data Feb 23 14:12:13 Any idea how to install hddimg or iso on virtualbox ? Feb 23 14:12:35 I try to install the images on virtualbox ? Feb 23 14:15:39 kanavin: rburton: that builder is calling "bitbake world -c distrodata" for 10 MACHINEs, then combining the generated distrodata.csv, extracting the Package and Description fields and writing those to a new distrpatch.csv, if distropatch.csv exists then the builder fails. Feb 23 14:16:33 joshuagl: pardon me, what is the purpose of -c distrodata? Feb 23 14:18:23 kanavin: do you really want to know or will you be happy with "don't worry about it" :) Feb 23 14:19:32 rburton: I somehow don't think the nightly-distrodata builder is doing what Saul wants/expects. It seems to be generating the data rrs uses? Feb 23 14:19:45 rrs doesn't use it afaik Feb 23 14:19:57 distrodata compares with other distros, rrs is upstream Feb 23 14:20:04 i'm wondering if we can ditch it Feb 23 14:21:16 the builder only cares about whether upgrades are available for recipes Feb 23 14:21:22 it ignores the data which compares with other distros Feb 23 14:22:05 that bitbake invocation generates 4 csv files: checkpkg.csv, distrocheck.csv, distrodata.csv, missinglicense.csv Feb 23 14:22:17 distrocheck seems to be the one you're thinking of rburton ? but the AB ignores that Feb 23 14:22:27 hm Feb 23 14:23:37 it would be nice to have an AB test for upstream check regressions Feb 23 14:23:44 I think joshuagl filed a bug for it Feb 23 14:24:09 me cleaning those up after people change SRC_URI is not optimal Feb 23 14:24:53 I did file a bug for it Feb 23 14:35:22 Anyone knows what is the difference between iso to hddimg ? Feb 23 14:36:27 ranchu: rburton knows, I've seen him explaining it here earlier today Feb 23 14:38:36 How do I enable the "musl" configuration used by the autobuilders? Feb 23 14:38:50 Which MACHINE, for example? Feb 23 14:39:15 qemux86 Feb 23 14:39:21 TCLIBC="musl" Feb 23 14:39:29 I tried MACHINE=intel-core-i7-64 together with require conf/distro/include/tclibc-musl.inc. Feb 23 14:39:55 So just TCLIBC="musl", no the include file? Feb 23 14:40:56 right, that's all the AB does Feb 23 14:40:56 hi rburton, can you please explain the difference ? I hope you weren't explaing it to me before :) Feb 23 14:42:28 I am doing a kernel build, but the do_kernel_metadata function make too much time. Almost an hour. Feb 23 14:43:07 If I do the same build without using yocto style recipe. The build finishes way faster. Why is that? Feb 23 14:46:07 Anyone familiar with difference between hddimg and iso ? Feb 23 14:49:18 ranchu: i was :) 'hddimg is a disk image with a partition and boot loaders, and live boots a read-write image from the disk image. iso is a cd image.' Feb 23 14:49:26 https://security.googleblog.com/2017/02/announcing-first-sha1-collision.html Feb 23 14:51:07 rburton, ok now I got your explanation. yet - I think the iso is also with bootloader. i.e. the only difference is disk to cd image. Or am I wrong ? Feb 23 14:51:32 I ask (I'm not sure, I just think I saw in iso a bootloader too) Feb 23 15:01:52 hi! just updated a build to use latest master did run a eSDK build. I noticed a change in the naming of artifacts in tmp/deploy/sdk. Some of the files there now include a reference to the date of the metadata snapshot. Wondering why this applies to "buildtools" artifacts, but not to "nativesdk-libc" ones. Even more strange SDK install scripts, manifests and testdata.json only got th '+snapshot' thing but Feb 23 15:01:58 with no date. Is this expected behaviour? (disclaimer: using some extra meta layers here, not yet tested with plain poky) Feb 23 15:07:24 kanavin: some failures on the dnf branch btw Feb 23 15:08:06 rburton: yes, was watching that anxiously :) Feb 23 15:16:17 I've added "GRUB_SERIAL ?= "console=ttyS2,115200" in /meta-hw/hardhware/hadrware.conf , yet after invoking bitbake image, grub.cfg is not updated. Feb 23 15:16:29 Do I need to invoke bitbake grub before ? Feb 23 15:18:07 no idea what meta-hw is but it sounds like some recipe doesn't have its dependencies right Feb 23 15:33:55 rburton - I rechecked, and see only one conf file in meta folders which has SYSLINUX_SERIAL_TTY, yet I don't find it in the generate syslinux.cfg Feb 23 15:56:18 How do we usually get files, I see that apt-get is not inside by default Feb 23 15:56:26 install I mean Feb 23 15:56:55 ranchu: if you have selected package_deb then just ensure that package-management is in your IMAGE_FEATURES Feb 23 15:56:58 then you'll have apt-get Feb 23 15:57:25 Ok , thanks! Feb 23 15:58:28 But is it correct to assume that it's not inside "by default" ? Feb 23 15:58:48 Ok, I assume the answer is yes. Thanks. Feb 23 16:03:44 hello, when i try bitbaking python-pysqlite i get an error saying python setup.py build execution failed Feb 23 16:03:57 anyone has any idea what could be causing it Feb 23 16:16:26 hi, i have a recipe that provides the same shared library (speex) as a recipe in poky Feb 23 16:16:53 i tried adding PROVIDES += "speex" but it still tries to build the poky recipe Feb 23 16:18:22 you have to pick which provider you want Feb 23 16:18:29 search the yocto docs for PREFERRED_PROVIDER Feb 23 16:22:08 kergoth:thanks Feb 23 16:29:34 Does anyone have any recommendations for managing multiple layers in different repos? Freescale goes with the "repo" tool but I'm not wild about it and condering looking into git submodules or subtree Feb 23 16:31:25 themikenicholson: https://github.com/Wind-River/wr-lx-setup/ is a new tool WindRiver developed Feb 23 16:36:46 joshuagl: themikenicholson: AFAIK wr-lx-setup from WindRiver uses google repo. Others successfully manage to do this with submodules (https://github.com/resin-os). IMHO subtrees are not meant for the job Feb 23 16:40:03 gizero: what is the reasoning against subtree? I havne't had a reason to play with it yet. Before yocto we used a big mono-repo. Feb 23 18:07:32 Any idea why display goes to sleep in intel image ? Feb 23 18:07:46 I have print in serial, but display is sleep. Feb 23 18:08:02 because X has a screensaver? Feb 23 18:08:25 In virtualbox the same build works without issues. Feb 23 18:08:59 I press keyboard and nothing happens. Feb 23 18:09:21 I see the first printing during boot, and then it goes to sleep (yet in serial it continue) Feb 23 18:10:36 rburton - any idea ? Feb 23 18:36:28 Do we need to do "IMAGE_INSTALL =" OR "IMAGE_INSTALL +=" Feb 23 18:43:29 i am getting this error "command 'x86_64-poky-linux-gcc' failed with exit status 1" when i try building python-pysqlite Feb 23 18:43:38 does anybody have any idea? Feb 23 18:45:06 stackoverflow says i need python-dev package but i see no recipe for that Feb 23 18:45:20 is there a way i can get python dev package Feb 23 18:54:19 does bitbake will search the internet if the recipe is missing ? Feb 23 18:54:37 oh really Feb 23 18:54:40 will take a look Feb 23 18:55:11 i looked through here https://layers.openembedded.org/layerindex/branch/master/recipes/?q=python-dev Feb 23 19:37:05 wesam: there's no need for any of that. python recipes already depend on the appropriate python bits needed to build, or the recipe is broken Feb 23 19:37:29 and it's the 'python' recipe that matters, python-dev is the name of a debian package Feb 23 19:41:38 oh really Feb 23 19:42:20 i keep getting a python setup.py build execution failed. error when i build python-pysqlite Feb 23 19:43:07 https://layers.openembedded.org/layerindex/recipe/46576/ Feb 23 19:43:15 that is the recipe i am using Feb 23 19:43:34 if you can help that will be amazing Feb 23 19:52:08 you need to stop summarizing and give us the exact error Feb 23 19:52:16 'execution failed' tells us absolutely nothing useful Feb 23 19:55:02 it gave me a huge log file. this is the main error i see "command 'x86_64-poky-linux-gcc' failed with exit status 1" Feb 23 19:56:53 also not useful. Feb 23 19:57:07 use a pastebin to post the entire log (*not* to the irc channel) Feb 23 19:58:34 where should i send it? Feb 23 19:59:35 https://www.google.com/#safe=off&q=pastebin Feb 23 19:59:53 https://en.wikipedia.org/wiki/Pastebin Feb 23 20:07:41 http://pastebin.com/MHvEZH1e Feb 23 20:08:02 there it is, didnt know about pastebin, thats incredibelly convinient Feb 23 20:08:26 wesam: pastebin is really the worst pastebin, despite it's name. Feb 23 20:12:08 oh really lol Feb 23 20:12:15 what do you recommend then Feb 23 20:24:15 Philosophical question about local.conf - what really belongs there? Feb 23 20:25:10 It seems like it should onyl be build-system-related things and it shouldn't be impacting what is getting installed in the image or what PACKAGECONFIG args are being set for a particulat package Feb 23 20:25:25 I'm getting the feeling we are abusing the heck out of our local.conf Feb 23 20:26:26 themikenicholson: personal changes only, everything else should be in distro config really Feb 23 20:26:37 or image config etc Feb 23 20:26:46 custom image recipes, custom distro config, and so on Feb 23 20:26:59 local.conf should just set distro and any tweaks that *you* want Feb 23 20:28:49 personal as in my development tweaks not shared amongst the rest of my development team? Feb 23 20:30:26 we're doing stuff like adding to IMAGE_INSTALL_append, PACKAGECONFIG_append_pn... Feb 23 20:31:33 blacklisting packages, etc. Feb 23 20:31:48 That stuff seems like it belongs in our own distro if I'm understanding docs correctly Feb 23 20:32:23 or the image for the IMAGE_INSTALL stuff Feb 23 20:42:42 Does the same yocto image recipe can create different images when invoked after some long time again ? Feb 23 20:48:43 themikenicholson: correct Feb 23 20:49:16 a minimal local.conf sets TMPDIR MACHINE and DISTRO Feb 23 20:49:29 then everything else is machine/distro/image specific Feb 23 20:49:30 sigh Feb 23 20:49:52 rburton: looks like a lot of the dnf failures are non-determenistic :( Feb 23 20:50:17 rburton: Thanks Feb 23 20:52:53 Does the same yocto image recipe can create different images when invoked after some long time again ? Feb 23 20:53:39 ranchu: what do you mean? Feb 23 20:53:47 how long it's been since you built last is irrelevent Feb 23 20:54:37 I mean if the same "bitbake " can create different rootfs, becuase package were updated in the git somewhere in the net. Feb 23 20:55:12 I mean the fetching a pckage can fetch a more update version of the package Feb 23 20:56:30 ranchu: that's only the case for recipes using SRCREV=AUTOREV. 99% of our recipes use explicitly specified versions or commit revisions, not branch head Feb 23 20:56:54 but yes, it'll rebuil dthe image *if* one of the dependent recipes tracks the head of a git branch Feb 23 20:56:57 ah, OK, I understand. Thanks Feb 23 20:57:24 just most don't. you can run buildhistory-collect-srcrevs to dump specific versions of any such recipes, in a format that you can add to a config file to lock them down, if you don't want it to change Feb 23 20:58:13 Right, Thank you. Feb 23 21:01:04 kergoth - I see for example that prefered kernel version in my target is 3.10% . does it mean that it one time can be 3.10.1 and next one 3.10.2 Right ? Feb 23 21:01:59 if you update your layers, yes Feb 23 21:02:07 if you don't want anything to change, don't pull changes from upstream Feb 23 21:02:53 Right. But if I give these layers for someone else, then he might upload other versions... Feb 23 21:03:03 rburton: would it sound crazy if I say that using kvm with qemu is causing failures with dnf? all of non-x86 arch tests completed fine Feb 23 21:03:04 download. not upload. sorry Feb 23 21:03:32 (except those where the earlier image building step failed) Feb 23 21:03:54 I need to investigate this further **** ENDING LOGGING AT Fri Feb 24 03:00:00 2017