**** BEGIN LOGGING AT Wed Apr 11 03:00:04 2018 Apr 11 05:20:29 Hi All Apr 11 05:21:28 Anyone has use i.max6 BSP for linux? Apr 11 06:41:59 Ashok if you are talking about imx6 yes Apr 11 06:43:08 Hello, I want to remove /usr/include/c++ headers from my target (space constraint), how can I do it ? Apr 11 06:48:28 kbo, what image do you use? Apr 11 06:49:41 I use a custom image recipe that inherits the core-image Apr 11 06:52:35 kbo, maybe you have some kind of "IMAGE_FEATURES" with "dev-pkgs" included Apr 11 06:55:32 kbo, https://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html#ref-features-image Apr 11 06:56:51 Aur313, I have EXTRA_IMAGE_FEATURES ?= "debug-tweaks" Apr 11 06:56:59 in my local.conf Apr 11 07:03:04 yes Aur123...it's about imax6 Apr 11 08:00:26 hello, Is there a way to call a image recipe from another recipe ? (I did try depends but it doesnt work, require image/.. works but it creates a new tar again) Apr 11 08:21:16 Is there a pkg_postinst_${PN} available for Image recipe ? Apr 11 08:39:53 What way is the best to change the password of the root user ? Apr 11 08:43:57 osten: EXTRA_USERS_PARAMS = "\usermod -p '\$1\$17daNVyp\$i7uiAI8yZhHPbpbyomFLk1' root; " Apr 11 08:49:40 New news from stackoverflow: Yocto do_kernel_metadata failed with custom BSP Apr 11 08:52:38 prabhakarlad: Would i need to inherit extrausers for using it ? Apr 11 09:04:04 osten: Yes Indeed INHERIT += "extrausers" Apr 11 09:08:50 Hi all. Apr 11 09:09:19 I'm pretty sure I'm in the wrong place, but I'm hoping someone might be able to steer me to right place! Apr 11 09:10:31 I'm trying to sort out a problem with an imx6 and its galcore driver. IT works when built within Yocto, as a bitbaked recipe, but I get nowhere when just using the toolchain and building stand-alone. Is there anywhere anyone knows that can help me with that? Apr 11 09:19:46 New news from stackoverflow: Enabling systemd in yocto image removes bash Apr 11 09:51:16 prabhakarlad: Ty, do you also happen to know how i should change the /etc/network/interface file ? Just with a recipe witch override the file or ? Apr 11 09:54:58 osten: you can use ROOTFS_PREPROCESS_COMMAND Apr 11 09:59:41 prabhakarlad: If i just copy it during the preproccess, wouldn't it just be overwritten when the default network recipe copies the default interface file ? Apr 11 10:19:56 New news from stackoverflow: Adding bash to a yocto image Apr 11 10:22:21 hey, guys Apr 11 10:27:28 osten: IMAGE_POSTPROCESS_COMMAND should do it when everything is done and its about to tar Apr 11 10:29:34 guys, I have a recipe and would like to make a package will be compiled every time regardless of whether iw was compiled or not Apr 11 10:29:49 here is the draft of my recipe: http://ix.io/17wb/ Apr 11 10:30:00 what do I have to change there ^^^? Apr 11 11:06:25 anybody got an example of an out-of-kernel-tree kernel module build which installs its own header into /usr/include/linux ? Apr 11 11:06:40 I've got a recipe which creates the -dev package, but whenever I populate_sdk it doesn't turn up Apr 11 11:07:38 (and the header is present in the -dev rpm) Apr 11 11:12:49 fetchall cmd not running Apr 11 12:58:42 Is there a correct approach for recipes that install python packages? Seemingly no matter how I setup my FILES definition, I end up with packaging errors that two or more recipes try to install /usr/lib/python2.7, for example. The error only occurs if I try to deploy RPM packages. Apr 11 13:04:18 Hi, I want to add files to packages-split/${PN}-dev directory, but with no luck Apr 11 13:04:31 tgoodwin, I blame fray for all rpm issues Apr 11 13:04:44 I have added to my recipe FILES_${PN}-dev += "/usr/include/imx" Apr 11 13:05:04 do_install_append() { oe_runmake headers_install INSTALL_HDR_PATH=${D}/usr/include/imx ARCH=$ARCH} Apr 11 13:05:24 sysroot_stage_all_append() { sysroot_stage_dir "${D}/usr/include/imx" "${SYSROOT_DESTDIR}/usr/include/imx"} Apr 11 13:06:32 anyone have some ideas ? Apr 11 13:12:58 Crofton|work, lol, alright. For now I can get around it picking one of the others, but with the RPM one being the 'default' in the local.conf template, I had guessed I would see more forum posts about other people hitting this issue unless I'm flat doing something wrong. Apr 11 13:13:00 :) Apr 11 13:14:17 Yeah, I'm not an rpm user, and despite the default, I susepct most heavy users use opkg :) Apr 11 13:59:20 I have a strange issue on rocko: If I use rsync from target to another target, ssh fails "ssh: Could not resolve hostname xxx.local: Inappopriate ioctl for device", yet running ssh using the same URL from the command line works fine. Any ideas on how to debug this? I'm openssh Apr 11 14:18:57 Might there be some enviroment settings that rsync doesn't pass on to ssh, that it use when being called directly? Apr 11 14:19:21 It applies to .local domain, i.e. Avahi mdns lookups Apr 11 14:20:44 New news from stackoverflow: How to add the standard library to arm-fslc-linux-gnueabi-gcc in yocto meta-toolchain fatal error: stdio.h: No such file or directory Apr 11 14:40:22 Hi, I'm trying to disable service from SysVinit like dnsmasq, I have look at dnsmasq recipe and it seems that it extend from update-rc.d wich add the service Apr 11 14:41:07 What would be the clean way to disable that inherit ? I guess it's not to modify the recipe itself right ? Apr 11 14:45:58 Hi! I'm using Yocto and a distro I've based off poky. Is this the right place to ask questions? Apr 11 14:51:52 I'd like to know how Yocto manages module autoloading. I've got an image autoloading a driver fine. However, I've replaced the driver with one built outside of Yocto (though using its toolchain). I've rebooted the system, rerun depmod and the module modprobes fine. Apr 11 14:52:01 It does not, however, autoload. Which seems weird. Apr 11 14:52:50 So I don't know how Yocto is making the system autoload the module. At the moment my thoughts are that I'm using a busybox depmod on the running image and perhaps Yocto does something different when it builds the system offline? Apr 11 14:57:34 I think the dns resolver in rocko might depend on stdin(!!!) for avahi .local lookups. If I do "ssh machine.local" from a yocto system, it succeeds. However if I do "ssh machine.local 0<&-" which closes stdin, it fails with "ssh: Could not resolve hostname machine.local: Inappropriate ioctl for device" Apr 11 14:57:46 Anyone knows how to add kernel'headers into sdk ? Apr 11 15:01:27 Anyone else here with a running yocto (rocko) system with open(ssh) and avahi running? It would be nice if one could test "ssh hostname.local 0<&-" to see if it fails. (Actually not interested if you can get login or not.) Apr 11 15:23:34 sveinse, I know the automatic testing is done using ssh for coms to the target. not sure if avahi is running Apr 11 15:23:58 so its tested with every stable build Apr 11 15:32:06 how to pack kernel and device-tree into jffs2 and rootfs into ubi? I have so far rootfs in ubi, but have no clue how to make a kernelfs with the same run. Apr 11 15:34:48 armpit: ssh to the target or from target? Apr 11 15:34:58 This occurs from target, but to target works fine Apr 11 15:35:01 ssh to the target Apr 11 15:35:21 I'm building vanilla rocko for qemu x64 now to see if it suffers the same issue Apr 11 15:35:45 which core-image-minimal ? Apr 11 15:36:24 * armpit will have to give it a try Apr 11 15:37:39 svenn: sounds like you are talking about fitimage Apr 11 15:37:42 svenn: https://patchwork.openembedded.org/patch/92797/ Apr 11 15:39:05 armpit: I'm building core-image-x11, but I suppose the minial could work Apr 11 15:41:21 sveinse, k Apr 11 15:42:20 darn, it works fine on qmeu Apr 11 15:44:18 Are there any generic arm targets that can easily be used with qemu from yocto? Apr 11 15:46:06 A question on the side: What option in yocto includes rpcbind and rpc.statd ? Or rather excludes... Apr 11 15:47:37 stephano: looks very much like what I want as we bundle kernel and device-tree on a a jffs2 image to burn to a different partition than the rootfs Apr 11 15:48:33 svenn: for those partitions have a look at wic: https://www.yoctoproject.org/docs/latest/mega-manual/mega-manual.html#creating-partitioned-images-using-wic Apr 11 15:51:33 ok, reading up on that now Apr 11 16:14:59 hmm, BBFILES_DYNAMIC is rather high overhead, having to add each layer area one by one Apr 11 16:18:55 I'm trying to rid my production image of rpcbind and rpc.statd, but it seems they are included due to a depchain from packagegroup-core-full-cmdline, which eventually pulls in rpcbind. Is there a (clean) way to refuse installing a singluar package from the system, or do I need to take the work to duplicate the packagegroup hierarchy? Apr 11 16:20:00 if it's pulled in via RRECOMMENDS rather than RDEPENDS you can use BAD_RECOMMENDATIONS. iirc bluelightning added a feature to explicitly remove a package after the install too, but if it's via RDEPENDS it'd end up uninstalling packagegroups too Apr 11 16:20:20 absolute worst case you could find the spot where it's pulled in and remove it there, i.e. with _remove Apr 11 16:21:52 Its pulled in via RDEPENDS unfortunately: poky/meta/recipes-extended/packagegroups/packagegroup-core-full-cmdline.bb Apr 11 16:23:57 then you can always alter that via _remove, either in local.conf or a bbappend Apr 11 16:24:14 unless you want to retain the existing unmodified packagegroup, in which case you'd need to create a replacement Apr 11 16:52:49 this name resolution issue is so weird! Apr 11 17:07:48 Is there an easy way to convert a python 'requirements.txt' directly to yocto layers? is there a layer that handles running a pip install cross-compile automagically? How do I pull in all the python requirements, preferably without having to manually enumerate each dependency the hard way. Apr 11 17:09:06 aarcane: The pythonic way of handling py dependendencies is via setuptools and pip. Apr 11 17:09:29 Not completely sure how that fits into the world of Yocto thou Apr 11 17:09:30 sveinse, yeah, pip install -r requirements.txt Apr 11 17:09:45 when running stock raspbian it's super simple Apr 11 17:10:51 I'm doing a lot of py-coding, but I haven't had the oppertunity to get any of that into a yocto image yet Apr 11 17:11:14 I have a python component that is the majority of my application, so I would like to be able to simply pack it up and have it do the equivalent of running 'pip install -r requirements.txt' before building the image. I'm sure it's not too uncommon of a requirement... Apr 11 17:12:33 In a Yocto image, why would one want pam? What is the use case where you need it? Apr 11 17:13:04 sveinse, uh... I don't need it, but I can think of a few cases Apr 11 17:13:18 like when you're making ldap connected thin terminals Apr 11 17:13:46 aarcane: when? right. We're connecting only via a preshared ssh certificate. Would one need pam for that? Apr 11 17:14:20 Uh.. Not to establish an outbound connection, no... Apr 11 17:15:06 aarcane: inbound? Is pam used in the mechanism which makes ssh check ~/.ssh/* and so on? Apr 11 17:17:08 pam is one of those base package things... someone might know better than me, but you almost never want to be without pam. YOu don't specifically need it, but so many other things are expecting it to be there by default Apr 11 17:18:38 just curious since 'pam' is a DISTRO_FEATURES option, so it is apparently workable without it Apr 11 17:19:02 sveinse, well, read the documentation about that feature Apr 11 17:19:28 sveinse, though truth be told, I've found that there is very minimal documentation for yocto... Apr 11 17:20:04 sveinse, a lot of "Here's how you make an image", and very little "Here's how to make an image do these common tasks" or "Here's how to configure these common packages" Apr 11 17:21:11 aarcane: I hear ya. I'm making the same struggles Apr 11 17:25:33 Truth to be told, I'm unsure if Yocto is viable in our organization. I'm the only one working on this, and tasks are time consuming. Like now, where I "just" need to remove three services which has open ports to the wild, and it turns out to be buried after multiple recursive dependencies. So it simply takes too much time. Apr 11 17:26:16 But on the other hand, there is no real alternative. Well there is, but they too has complications and complexities. So it's Yocto for now Apr 11 17:27:24 I am to investigate if running android is an alternative. Apparently there is better access to devs there Apr 11 17:27:27 aarcane: most yocto builds dont' use pam. it can be opted into, but most distros don't Apr 11 17:28:03 sveinse: to be fair, any linux distro would have similar issues if you want to remove a package that's pulled in via package dependencies indirectly Apr 11 17:28:10 debian, etc Apr 11 17:29:26 kergoth: yes. I used to lead a ubuntu based image a couple of years back. Built the build system from scratch. Not unlike bitbake in concept actually. Wasn't free of hazzle to put it that way :P Apr 11 17:31:05 but progress /is/ slower on oe/yocto. There are many implicit or buried rules, so you have to find them. The layering of information makes navigation complex Apr 11 17:33:13 just out of curiosity. info on working with yocto is unclear or hard to find, not there? Apr 11 17:33:57 Neither qemux86_64 nor qemuarm on rocko exhibits the dns lookup issue I'm seeing on my target Apr 11 17:34:28 benvantende[m]: sorry, what do you mean? Apr 11 17:35:23 benvantende[m], I'm going to say 'yes'. There is a lot of documentation I haven't found yet... but I cna't say it doesn't exist Apr 11 17:36:53 Ahh just not easy to find. Common problem. Thanks for the feedback Apr 11 17:37:57 benvantende[m], try google searching "Yocto networking" and see how long it takes to find a good, concisive, semi-relevant guide that explains the various layers and packages and how to configure them Apr 11 17:38:35 aarcane: check Apr 11 17:39:44 One example now: I don't want busybox version of netstat. What do I do to not get that? Apr 11 17:51:03 the convention is to just add the real package to your image Apr 11 17:51:11 check your linux host, see what package owns netstat, add to IMAGE_INSTALL, done Apr 11 17:51:29 we use update-alternatives for busybox symlinks, the real versions will be used in preference to busybox versions Apr 11 17:53:08 kergoth: which reminds me, in lack of a yocto package database, I've built world for a qemu machine to build a library of packages in which I can search to find what recipe it belongs to Apr 11 17:57:52 I see that my target which troubles with dns lookups is running named (for some reason). Gotta find out why and if that can be the culprit Apr 11 17:58:14 sveinse, what's in resolv.conf ? Apr 11 17:59:18 my dns servers on lan Apr 11 17:59:37 not your problem Apr 11 18:00:09 use wireshark or tcpdump to trace your dns resolution Apr 11 18:01:17 you don't really need to build world to know what binary packages will be emitted, bitbake already knows this via PACKAGES & PACKAGES_DYNAMIC. not sure we provide a build time tool to access that data, though, since i stopped working on the bb tool. oe-pkgdata-util can search an dlist packages, but only after the build Apr 11 18:01:22 s/an dlist/and list/ Apr 11 18:01:26 hmm Apr 11 18:02:03 aarcane: not that simple unfortunately. It's not for dns, its for mdns (avahi). And secondly it happens locally: if I run ssh mymachine.local from the command-line it works fine, but if I do "ssh mymachine.local 0<&-" it fails with "ssh: Could not resolve hostname mymachine.local: Inappropriate ioctl for device". Apr 11 18:02:20 Hence it's the closing of stdin on the dns resolver (!) which croaks Apr 11 18:02:36 sounds like a bug in ssh to me. i'm assuming you tested with ping, host, nslookup, etc? Apr 11 18:02:55 also verified the configuration in /etc/nsswitch.conf? Apr 11 18:03:03 specifically where mdns or mdns_minimal are listed Apr 11 18:03:24 ugh.. is https://layers.openembedded.org/layerindex/branch/master/layer/meta-web-kiosk/ broken or am I missing something? Apr 11 18:04:55 kergoth: it is plausible it's a ssh bug, haven't been able to recreate with any other dns-resolving apps Apr 11 18:05:39 Does yocto support generation of "vmdk" images (used by vmware/virtualbox)? I tried doing a bitbake with IMAGE_FSTYPES += " vmdk"; but it doesn't work. Apr 11 18:05:55 and did you compare to a different ssh implementation? Apr 11 18:05:59 dropbox and openssh are both available Apr 11 18:06:26 dropbear? Apr 11 18:06:44 kergoth: nsswitch.conf has mdns under "hosts: files myhostname mdns4_minimal [NOTFOUND=return] dns mdns" Apr 11 18:06:56 yeah, dropbear, thanks Apr 11 18:06:58 This is openssh Apr 11 18:08:18 it sounds like ssh is just misreporting the error. it tries to resolve, and while doing that it also needs to do an ioctl on stdin for some unrelated reason, which fails, sets errno, and openssh reports that as a failure on the resolution Apr 11 18:08:29 * kergoth shrugs Apr 11 18:10:09 hah, if I put mdns first in the nsswitch.conf it resolves. On to something here... Thanks, kergoth Apr 11 18:11:41 np Apr 11 18:12:56 and rsync is back on track. perfect. Now I need to read up on the consequences for rearranging the order in nsswitch.conf Apr 11 18:14:33 if 'files' isn't first, /etc/hosts won't be able to override dns, myhostname, or mdns, etc. i don't know the difference between mdns4_minimal and mdns, though Apr 11 18:17:17 I think I have found the culprit: On the failing system I have "hosts: files myhostname mdns4_minimal [NOTFOUND=return] dns mdns" while on a fresh qemu rocko system I get "hosts: files mdns4_minimal [NOTFOUND=return] myhostname dns mdns". If I move 'myhostname' then it resolves correctly Apr 11 18:17:49 So something has reordered the sequence here. Apr 11 18:19:18 tough to say, if myhostname and avahi both alter nsswitch.conf with sed, it could just be the package install order affecting the result Apr 11 18:19:26 seems clear it's myhostname resulting in the issue, the real question is why myhostname is failing or has anything to do with stdin whatsoever Apr 11 18:21:04 Is is failing on stdin, I've traced that. But if this is after ssh has failed resolution and wants to do something with stdin is plausible Apr 11 18:21:48 it doesn't make sense that having myhostname before mdns4_minimal would prevent it from resolving. if it doesn't resolve via myhostname, it moves on and tries the next resolver Apr 11 18:22:04 hmm Apr 11 18:22:26 https://pastebin.ca/4014082 Really helpful error message Apr 11 18:23:22 Hmm, looks like paul added some more methods and commands to tinfoil to deal with the server split, i might be able to fix some of my bb tool commands to work again without too much difficulty.. course some of it is no longer as important given oe-depends-dot and oe-pkgdata-util current functionality, but some is Apr 11 18:31:19 I'm not able to get it to fail on vanilla systems by rearranging myhostname the "wrong" way. But it is consistently fixing my issue on target. Very strange Apr 11 18:32:34 and for the record, I'm looking up other .local hosts, not my own hostname Apr 11 18:34:33 Yeah, something's broken in meta-web-kiosk and specifically vala... https://pastebin.ca/4014085 Apr 11 18:34:36 i wonder if the myhostname version is different, and also what, if any, recommendation for nsswitch.conf the myhostname project itself advises Apr 11 18:36:01 on my host ubuntu myhostname is last of all Apr 11 18:39:23 huh, that's interesting Apr 11 18:45:08 Allright, gotta go. Your input has been invaluable, kergoth, thanks a lot Apr 11 18:45:18 np, good luck Apr 11 18:52:08 well, I got vala and midori to *think* about building... Apr 11 18:52:15 Let's see if they compile Apr 11 18:53:03 aarcane: probably best to use epiphany instead of midori these days, web-kiosk hasn't been touched for ages Apr 11 18:53:24 rburton, I've noticed that, but there's no documentation on epiphany Apr 11 18:53:47 or rather, there's no documentation on building a web-kiosk using epiphany Apr 11 18:55:33 aarcane: changing midori to epiphany in the image will be a big part of it, then change the init script Apr 11 21:14:48 okay, so a package (webkitgtk) failed to build because OOM killer. I expanded the memory and restarted the build, but some files were corrupted... so I tried removing the folder 'rm -rf tmp/work/cortexa7t2hf-neon-vfpv4-poky-linux-gnueabi/webkitgtk/' Apr 11 21:15:42 But now when I try to continue, bitbake webkitgtk, it gives me some error about cmake not found and doesn't try to re-download or re-extract or anything webkitgtk Apr 11 21:32:57 bitbake -c cleanall webkitgtk was what finally made it servicable Apr 11 21:58:11 aarcane: the easy fix is bitbake webkitgtk -C unpack Apr 11 21:58:33 which forces a clean unpack and then builds Apr 11 21:58:47 bitbake webkitgtk -cclean would have worked Apr 11 21:59:27 just deleting bits of tmp didn't work because bitbake didn't know you did that, so it thought everything was still in there Apr 11 22:59:53 how do I do the equivalent of bitbake -j 1? Apr 11 23:01:55 aarcane: set BB_NUMBER_THREADS to 1 and PARALLEL_MAKE to "" Apr 11 23:10:01 yeah, if there are four different threads trying to use all 6 GB of ram, it's never going to finish linking :P **** ENDING LOGGING AT Thu Apr 12 03:00:05 2018