**** BEGIN LOGGING AT Thu Mar 12 02:59:58 2015 **** BEGIN LOGGING AT Thu Mar 12 05:03:11 2015 Mar 12 06:09:07 Any idea how umount deletes the mount point in distros like opensuse or ubuntu or arch? Mar 12 06:09:27 Is it done by udsiks? Mar 12 06:09:47 sujith_h, what do you mean by deletes a mount point? Mar 12 06:10:20 I've created a recipe for rtl8111 kernel mod (gigabit ethernet) where would I submit it? Mar 12 06:12:08 redengin: lets say I plug in a usb storage. And it gets automounted in /meda/ with labelid ( if it has label id). And if i execute umount /media/ then it unmounts the device mounted in /media and it deletes the folder in /media Mar 12 06:12:52 https://help.ubuntu.com/community/Autofs I believe its this Mar 12 06:16:21 redengin: Thanks. Let me try it :) Mar 12 06:39:58 is there to append .conf files? Mar 12 06:43:14 redengin: http://www.openembedded.org/wiki/How_to_submit_a_patch_to_OpenEmbedded Mar 12 06:45:11 sujith_h: you are looking for udisks if you need to do that Mar 12 06:46:03 http://cgit.openembedded.org/cgit.cgi/meta-openembedded/tree/meta-oe/recipes-support/udisks/udisks_1.0.4.bb?h=master Mar 12 06:49:29 khem_: ok. So eventually udisks rules will have to be modified so that umount command will delete the folder which was mounted with the device? Mar 12 06:53:47 yes Mar 12 06:54:04 khem_: thank you :) Mar 12 06:57:23 well you might not need to do that Mar 12 06:57:33 and put it under /media Mar 12 06:57:44 it uses /dev/disk/by-uuid/ Mar 12 07:05:56 Gh Mar 12 07:06:15 Hey Mar 12 07:08:41 In local.conf i specify to install g++ but it is not there in the root fs...doesn't the recipie for it already exist in oe? Mar 12 07:08:42 In local.conf i specify to install g++ but it is not there in the root fs...doesn't the recipie for it already exist in oe? Mar 12 07:13:14 IMAGE_FEATURES += " tools-sdk" Mar 12 07:16:17 What does it do in brief? Mar 12 07:16:43 blackfish, should install the full set of compiler tools Mar 12 07:17:12 tools-sdk - SDK (C/C++ compiler, autotools, etc.) Mar 12 07:19:54 Hey, sorry for the connection loss... So thankyou **** BEGIN LOGGING AT Thu Mar 12 07:25:16 2015 Mar 12 08:16:15 how do I add a kernel module to yocto 1.7 it says it can't find kernel-module-XXXX Mar 12 08:17:49 can't find it on the target, or during rootfs install? Mar 12 08:22:11 during rootfs install (it never gets built) Mar 12 08:22:49 I can bitbake the XXXX.bb that creates the module on its own Mar 12 08:27:11 okay, so it's not one of the packages created by the kernel recipe? Mar 12 08:29:56 i don't think i've done that before, but there should be a depend to go with that rdepend Mar 12 08:29:57 I created a rtl8168_8.0139.bb recipe that uses inherit module Mar 12 08:30:47 does your recipe have a DEPEND = virtual/kernel ? Mar 12 08:31:57 I'm using core-image-minimal and adding MACHINE_ESSENTIAL_EXTRA_RDEPENDS Mar 12 08:32:56 redpends is what tries to put it in the image, but it needs to be built/packaged first Mar 12 08:34:10 hmm, I think the sstate is messed up Mar 12 08:35:38 cleaning virtual/kernel and trying again Mar 12 08:38:08 * mr_science will shutdown so roommate can sleep... Mar 12 08:44:24 should I be able to bitbake kernel-module-rtl8168? Mar 12 08:58:56 is there a good way to check which kernel features (config) a built image has? Mar 12 08:59:25 I'm building my own and would like to see if bbappends had any effect without booting the image. :-) Mar 12 09:03:05 ok, it looks like my edits to the machine's .conf aren't getting cleared... how do I do that without a full rebuild? Mar 12 10:48:13 morning all Mar 12 10:50:14 hi there Mar 12 13:15:17 bluelightning: I'm wondering about a commit of yours in combo-layer from 2011: "d4d491 - make init set up initial component data" Mar 12 13:15:54 In it, you pass the file_filter as file list to tar. Mar 12 13:16:58 First, if file_filter contains wildcards (as in the src/*.c example), GNU tar fails and explicitly asks for clarifying whether wildcards should be used. One has too use --wildcards to enable the feature. Mar 12 13:17:40 Second, once enabled, the matching is against the entire path name. So "src/*.c" also matches "src/foo/bar/*.c", which is different from the behavior one gets when passing the same filter to "git format-patch". Mar 12 13:18:01 Is it just me or has this feature never really worked with wildcards? Mar 12 13:27:44 hi pohly Mar 12 13:31:02 I'm fairly confident it did work originally, but then that was a long time ago Mar 12 13:33:07 combo-layer has only had a limited number of users (Richard and a few others over the years), and there aren't any regression tests, so it's hard to be sure if it broke subsequently Mar 12 13:33:41 btw, reviewing your patches is on my todo list for this afternoon Mar 12 13:33:48 hello, I am trying to change the group of /etc/default/ntpdate. I add ntpdate (ntp package) to my image and append the following recipe : http://pastebin.com/vT9YP49v the acces never change, what am I doing wrong? Mar 12 13:37:16 is anyone succesfully using Eclipse with Dizzy? Mar 12 13:37:52 using pre-built toolchain or build system derived I am not having much luck Mar 12 13:38:01 * joshuagl wonders if it's generally known to be working Mar 12 13:46:15 bluelightning: good. They are still valid. Mar 12 13:46:21 Just beware that more is coming :-) Mar 12 13:47:11 I've implemented "combo-layer --history init" which imports the entire history of the components instead of sqashing them. Mar 12 13:47:33 When testing that I ran into this "tar + filter_file" issue. Mar 12 13:59:07 pohly: ah ok - that sounds good Mar 12 13:59:22 pohly: did you hit any issues with the one or two merge commits that have sneaked in over the years? Mar 12 14:00:07 My approach has no problems with that. Mar 12 14:00:22 instead of explaining here, let me finish that work and post the patches. Mar 12 14:00:39 ok, I'll wait for those then :) Mar 12 14:01:04 But in a nutshell, it's more like "git subtree", implemented using "git fetch + git filter-branch + git merge". Mar 12 14:02:26 that does sound a bit more robust Mar 12 14:03:11 does that mean you get merge commit(s) though? Mar 12 14:10:10 hey, Mar 12 14:10:31 i am trying to use opencv on yocto Mar 12 14:10:51 I want to compile cpp programs which requires open cv Mar 12 14:11:37 since i am not that good with bitbake, recipies... and.... hello world, i thought of compliling the cv code onboard Mar 12 14:13:32 these are the packages i installed: Mar 12 14:13:40 http://pastebin.com/vDC1JWxZ Mar 12 14:14:04 but when I tried cmake, this is what i get Mar 12 14:14:05 http://pastebin.com/WifpRYXV Mar 12 14:14:47 so I followed the instructions here: Mar 12 14:14:48 http://stackoverflow.com/questions/8711109/could-not-find-module-findopencv-cmake-error-in-configuration-process Mar 12 14:16:53 JaMa: ping Mar 12 14:19:01 I get this error Mar 12 14:19:09 http://pastebin.com/V7sSwP8r Mar 12 14:19:17 what should I do? Mar 12 14:19:36 if anyone has run opencv on yocto before, please help me. Mar 12 14:23:48 sujith_h: pong Mar 12 14:31:51 JaMa: I have a query with respect to Qt5. I am using Qt 5.3.2 version. When I try to launch Qt Creator and try to build a sample Qt Widget example. It works fine. I am able to compile it in the host and deploy on the target. But when it comes to test a qtquick example code, i am not able to compile it on the host. qmake throws error saying : Project ERROR: Unknown module(s) in QT: quick qml. I am using Qt 5.3.2 from meta-qt5 layer. Mar 12 14:32:37 JaMa: In the Qt Creator I am pointing the sysroot to that of the target for which is built for. Mar 12 14:34:57 JaMa: Do I need to have native recipes for qtquick and qtdeclarative to build the code on host? Mar 12 14:35:40 I'm not using Qt Creator at all, ask otavio or Andreas Müller Mar 12 14:36:02 JaMa: ok. Thanks Mar 12 14:37:48 otavio: ping Mar 12 14:38:46 sujith_h: you need to have the modules istalled Mar 12 14:39:16 otavio: Yah the modules are installed for the target Mar 12 14:39:35 sujith_h: on the SDK? Mar 12 14:41:13 otavio: well I am not using SDK. My setup comprises of core-image-sato with Qt5 recipes ( like qtbase, qtdeclarative, qtquick1 etc). Mar 12 14:41:47 sujith_h: so I don't expect it works; never did this way Mar 12 14:42:43 otavio: should I include packagegroup-qt5-qtcreator-debug recipe into the image? Will that help me? Mar 12 14:43:35 i am not sure what you are doing Mar 12 14:45:26 bluelightning: yes, fetch and filter-branch both preserve merges. Mar 12 14:46:03 pohly: what I meant was if it uses merge do we get additional merge commits beyond any present in the component repos? Mar 12 14:46:52 During "combo-layer update"? No, not at this time. That still works as before (git format-patch). Mar 12 14:47:12 Or do you mean the "master" branch history? Mar 12 14:47:19 otavio: I have the cross compiler separately installed. So I point Qt Creator to that. And i had built qtbase, qtquck1 and qtdeclarative for the target. And now I use Qt Creator and start a project by selecting Qt Quick Application and then select Qt Quick 2.2. And then try to compile that code. Mar 12 14:47:21 It basically starts with a merge commit. Mar 12 14:55:04 otavio: Can you help me understand how do you use it. So that I can try that way. Mar 12 14:55:41 pohly: I'm referring to the new --history init option Mar 12 14:56:22 pohly: I guess that works for those happy to have those in the resulting combined repo; in poky itself we prefer to avoid those but that's not a blocker just for the tool Mar 12 14:57:53 bluelightning: I guess the one initial merge commit could be avoided by importing components one after the other, but for histories with merge commits which contain conflict resolution there really is no clean way around keeping those. Mar 12 14:58:15 pohly: right, there's no way to have your cake and eat it ;) Mar 12 15:04:10 <_4urele_> is there a way to search among build packages which package install a file or directory? (today I'm using a python script which parse "rpm -qpl" on each rpm file, then I search in a dict this is quite fast, but maybe there is another way) Mar 12 15:11:39 _4urele_: yes, oe-pkgdata-util can find that information (after the recipe providing that file has reached do_package) Mar 12 15:11:55 well, the recipe producing the package that contains the file, to be precise Mar 12 15:12:27 the toaster web UI can also be used to find that information Mar 12 15:14:49 <_4urele_> thanks bluelightning Mar 12 15:16:25 systemd.bbclass does do_install[postfuncs] += "rm_sysvinit_initddir ". I would like to revert that for a number of recipes, from a class that is inherited before systemd. Is there any way to do that in a simple way? Mar 12 15:20:24 Saur: well, it's hacky but you could simply prepend the function and just exit if its not called from elsewhere Mar 12 15:20:30 Saur: otherwise not really Mar 12 15:20:41 can you explain what you're trying to achieve? Mar 12 15:20:46 bluelightning: Can you prepend to a python function? Mar 12 15:20:55 yes Mar 12 15:21:01 with python code you can Mar 12 15:21:10 Do you have an example? Mar 12 15:21:39 I don't know that we have any existing examples in the metadata, but it should be pretty quick to test ;) Mar 12 15:22:08 what I meant by "with python code" is that you can prepend python code to a python function Mar 12 15:22:24 somewhat obviously you can't prepend shell script Mar 12 15:23:02 bluelightning: What I want to achieve is that I want to disable sysvinit in DISTRO_FEATURES, but we do have quite a number of recipes that even though they use systemd still installs scripts and stuff in /etc/init.d... Mar 12 15:24:28 Hello guys, how can I develop with yocto using a stm32? Mar 12 15:24:31 Saur: hmm, I'm not following - if they "use systemd" aren't they installing units that should be used in preference to any initscripts? Mar 12 15:25:42 bluelightning: They do. But those units still use stuff that the recipes install in /etc/init.d. However, the rm_sysvinit_initddir function from systemd.bbclass removes /etc/init.d if sysvinit is not in DISTRO_FEATURES... Mar 12 15:27:22 I have to say that doesn't sound like an ideal arrangement... Mar 12 15:28:25 I guess what I'm trying to determine is is this something that systemd.bbclass in OE-Core should be trying to handle properly or is this a hack around some unusually constructed software Mar 12 15:30:20 I don't really see the problem. If a recipe installs something into /etc/init.d that its systemd units use, that's a bug and should be fixed on a case by case basis Mar 12 15:30:30 wiping /etc/init.d when sysvinit isn't in distro features seems entirely reasonable to me Mar 12 15:30:40 bluelightning: Why is that a bug? Mar 12 15:30:46 people, I'm trying to upgrade meta-oe to use Samba 4.2 , managed to make configure partially work ( after changing the buildsystem and a few other things ) Mar 12 15:30:54 now I'm having ERROR: This autoconf log indicates errors, it looked at host include and/or library paths while determining system capabilities. Mar 12 15:31:08 Rerun configure task after fixing this. The path was '/data/yocto/poky/build-qemu/tmp/work/i586-poky-linux/samba/4.2.0-r0/samba-4.2.0/source' Mar 12 15:31:26 kergoth: Why must /etc/init.d not be used by systemd units? Mar 12 15:31:27 yesterday ( I'v lost the backlog =/ ) somebody told me that this was probably an error since the path seems fine Mar 12 15:31:58 where should I look at things to try to make this work? searching on google didn't really returned anything userfull. Mar 12 15:31:59 Typically I guess the developers have simply reused the old sysvinit scripts and just start them from a unit file than from a link in /etc/rc3.d Mar 12 15:34:00 bluelightning: Yay, I got the prepend to python function to work, so now I have a solution for us. :) Mar 12 15:39:09 Saur: lazy developers putting things where they don't belong really isn't behavior we should be encouraging, imo Mar 12 15:40:44 kergoth: Maybe so, but since we are adapting an old build system to Yocto, we (in the project I'm working in) do not have the means to change this. We can recommend it to the developers and they can eventually fix it, but we need to have a workaround for now... Mar 12 15:48:55 ah, fair enough, that makes sense Mar 12 16:36:17 when cross-compiling samba I need to pass a flag to configure like this: --cross-execute='qemu-arm-static -L /usr/arm-linux-gnu' wich states the machine and libraries to execute the configure tests. What should I pass on yocto for that, or any other project that does the same that I can loock in? Mar 12 16:43:18 tomaz: EXTRA_OECONF = "" Mar 12 16:44:38 uh? empty? Mar 12 16:45:34 we handle configure tests differently, in general, by supplying the test results via the autoconf configuration cache Mar 12 16:46:17 kergoth, and if the system is not autoconf based ( in this case, waf ) Mar 12 16:47:27 I'm not asking 'fix my problem' but 'point me to the right direction', so I can fix and upstream it. :) Mar 12 17:02:06 I'm adding an external module to my build. Where do I package the udev rules file for it? Mar 12 17:46:21 tring to get rid of the distcc package, any idea why it is "in there" (my image)? Mar 12 17:51:54 T0mW: are you building for a qemu* machine? Mar 12 17:52:26 bluelightning: naw, production Mar 12 17:52:55 I think it might be in "meta/recipes-core/packagegroups/packagegroup-core-sdk.bb" Mar 12 17:53:36 right, there are only a couple of ways that it would be brought in such as that way Mar 12 17:54:00 EXTRA_IMAGE_FEATURES = "dev-pkgs tools-sdk tools-debug debug-tweaks" Mar 12 17:55:08 removed the reference in packagegroup-core-sdk, try rootfs build to see if it "goes away". Mar 12 17:55:14 :) Mar 12 17:57:15 T0mW: FYI if you do need to track these kinds of things down in future, you can use the dependency graphs produced by buildhistory, or the Toaster web UI Mar 12 17:58:31 Anybody there with stm32 experince using yocto? Mar 12 18:02:50 bluelightning: heh, find . -type f | xargs blah Mar 12 18:03:17 I'm adding an external module to my build. Where do I package the udev rules file for it? Mar 12 18:07:53 cyclist: try recipes-core/udev Mar 12 18:08:47 cyclist: I just bbappend to udev and substitute my own local.rules Mar 12 18:09:49 thanks Mar 12 18:11:34 cyclist: this would be my meta-awsom/recipes-core/udev/udev_182.bbappend http://pastebin.com/Hd53Gnkp Mar 12 18:16:14 has anyone here ever used opencv in yocto? Mar 12 18:21:53 I finally got image-core-minimal to boot on odroid-c1 \o/ Mar 12 18:22:07 my next target is xu3(-lite) Mar 12 18:24:45 thanks again Mar 12 18:27:52 What I don't understand about all of this is that the whole idea of packaging something includes everything, anddoing it this breaks all of that. Mar 12 18:28:13 er, doing it this way Mar 12 18:29:06 High granularity in our packaging gives flexibility and the ability to control what goes into your image, keeping sizes down without blindly rm -rf'ing content the way other buildsystems tend to do Mar 12 18:29:17 nothing is lost, just broken out Mar 12 18:31:16 well, the udev rule is only pertinent to this particular module, so it seems kind of broken to have to include it somewhere else Mar 12 18:31:30 open a bug Mar 12 18:31:51 I may do that :) Mar 12 18:57:59 kergoth, ahah, the thing that I was missing: inhertis native. Mar 12 18:58:04 :D Mar 12 18:58:13 samba 4.0 *almost* ready Mar 12 18:58:21 only misses everything besides do_configure() Mar 12 18:58:23 * tomaz dances Mar 12 18:58:43 does it include heimdal? :) Mar 12 19:05:24 dunno ._. Mar 12 20:04:06 is there a way to clear a single target? ( didn't found a bitbake --clear target ) or something Mar 12 20:04:24 target? Mar 12 20:04:45 a package Mar 12 20:05:19 you mean a recipe? Mar 12 20:05:25 that's covered in the yocto documentation Mar 12 20:05:26 or a collection - yes, a recipe. Mar 12 20:05:49 bitbake -c clean recipename, or -c cleansstate recipename if you want to build from scratch, or -c cleanall recipename if you want to re-download sources from remote locations Mar 12 20:06:06 see also -c listtasks Mar 12 20:06:19 also bitbake --help Mar 12 20:06:37 kergoth, I understand that it exists in the docs, the issue is that a search on google return bogus stuff if you don't really know what to look for. for instance, I searched for "yocto clear recipe build" Mar 12 20:06:42 and bogus was found. Mar 12 20:06:49 then try actually reading the docs instead Mar 12 20:06:58 or pull them up and hit ctrl-f Mar 12 20:06:59 fair enough Mar 12 20:07:24 https://www.yoctoproject.org/docs/1.7.1/mega-manual/mega-manual.html is the full set of documentation, perfect for searching Mar 12 20:08:12 thanks. I'll promisse that I'll spend a lot of time on it and only come here with questions if it's not there. Mar 12 20:30:50 JaMa, ugh. now that the do_configure passes, the heimdal fails. Mar 12 20:50:58 Is there anyway to configure EFI to autoboot grub instead of running bootx64 on EFI shell ? Mar 12 20:54:03 kergoth, do you know ^? Mar 12 21:03:41 bluelightning, do you now it ^? Mar 12 21:03:49 s/now/know **** ENDING LOGGING AT Fri Mar 13 02:59:59 2015