**** BEGIN LOGGING AT Wed Feb 22 03:00:02 2017 Feb 22 07:32:03 hi Feb 22 07:32:45 in case I need to create additional partitions (empty partitions), where to start from ? I am begginer with yocto Feb 22 08:11:24 Hi guys is there anyone here using yocto with a raspberry ? I need some help in order to configure boot partition Feb 22 08:34:44 I also need to understand how to create empty partitions with yocto. any idea ? Feb 22 08:36:20 ranchu: you probably mean "how do i create arbitrary, possibly empty partitions in an openembedded-stly build, using the poky distribution" :-) Feb 22 08:37:23 ranchu: in that case, your answer is: look at wic, and its corresponding configuration files. this is the tool that creates partitioned images out of the filesystems. Feb 22 08:43:14 LetoThe2nd, thanks, I made some reading about wic. but I don't understand. is it used after the build and deplyment is done , on the target which was booted with our image ? Feb 22 08:44:36 ranchu: no, it basically can take the created file/directory structure and turn it into a binary image. Feb 22 08:45:16 ok, Feb 22 09:32:40 * rburton curses Feb 22 09:33:04 a broken patch sneaked back into a branch that was running overnight on the AB Feb 22 09:42:36 anyone ? Feb 22 09:43:25 rburton: I saw your message from ~12h ago. Sent a v2 for the series which now applies cleanly to ross/mut and master. Feb 22 09:44:23 thanks nrossi Feb 22 09:53:46 ed21: https://autobuilder.yoctoproject.org/main/builders/nightly-wic/builds/623/steps/CreateWicImages/logs/stdio <— did i pull something broke into ross/mut? Feb 22 09:55:56 rburton: I'll try to find out. It looks familiar. Feb 22 09:56:25 thanks Feb 22 10:20:17 hello, I have some issue with a recipe in the meta-security layer, is this the correct channel to ask or is there a specific channel for this layer? Feb 22 10:21:34 svagionitis: try Feb 22 10:30:21 Hi, I need some help. I've updated my yocot to the lastest git master. Now my custom kernel builds fail. Error is "kgit-s2q: not found" Feb 22 10:31:30 I have created a bug for the issue that I have, https://bugzilla.yoctoproject.org/show_bug.cgi?id=11095, but to summarize here, I have some problem building the `tpm-tools` from the latest master branch of meta-security layer. It seems that it cannot locate properly the `libtspi.so` library from the `trousers` recipe. I would appreciate any help. Feb 22 10:31:32 Bug 11095: normal, Undecided, ---, saul.wold, NEW , Error building tpm-tools v1.3.9 Feb 22 10:37:55 JaMa: hi, can you try if variable expansion issue is fixed by this patch? https://gist.github.com/bartosh/b47575a869fc80dbc3995d114524540e Feb 22 10:54:44 zzeroo: you probably need a dependency on the recipe that provides that Feb 22 10:57:10 Seeing this ERROR: virtual/gettext required but not in DEPENDS for file /scratch/crash_fix_builsd/build/tmp/work/cortexa9hf-vfp-neon-mxl-linux-gnueabi/qtwebengine/5.6.2+gitAUTOINC+qtwebengine_4d61c01b71-r0/git/src/3rdparty/chromium/third_party/yasm/source/patched-yasm/configure.ac. Feb 22 10:57:34 while bitbake qtwebengine Feb 22 10:59:19 sounds like you need to add virtual/gettext to DEPENDS :) Feb 22 10:59:45 rburton: what's the difference between using variable directly and using d.getVar, e.g. WKS_FULL_PATH = "${@wks_search('${WKS_FILES}'.split(), '${WKS_SEARCH_PATH}') or ''}" and WKS_FULL_PATH = "${@wks_search('${WKS_FILES}'.split(), d.getVar('WKS_SEARCH_PATH')) or ''}" Feb 22 10:59:59 rburton: looks like it's not fully expanded in first case. Feb 22 11:00:55 ed2: i'd be incredibly surprised if ${} expanded inside python bits, use d.getVar() Feb 22 11:00:57 rburton: May I know, how and where ? Feb 22 11:01:35 hi guys. Feb 22 11:01:49 rburton: thanks. that's what I suspected. Feb 22 11:02:00 just a little quick question: how do u change the opkg urls in during build? so it points on right server from beginning? Feb 22 11:04:35 rburton: I can't reproduce that build failure on my setup. I'm using ross/mut branch. Feb 22 11:06:24 Ramose: in the qtwebengine recipe Feb 22 11:06:36 ed2: hm fun. Feb 22 11:06:44 itayo: what urls do you mean Feb 22 11:07:25 rburton: https://paste.ubuntu.com/24045900/ Feb 22 11:10:09 rburton: the commit id is different from what's shown on the ab build page though. Looks like you've changed the branch. Feb 22 11:10:25 rburton: It worked , let me resume further Feb 22 11:12:12 ed2: i just dropped the ltp patch Feb 22 11:12:18 no other changes Feb 22 11:13:31 Hello! I'm trying to install some deb packages, i followed some example from this site: https://software.intel.com/en-us/forums/2014-intel-cup-embedded-system-design-contest/topic/518037 Feb 22 11:13:58 Sadly i seem te get an error that says it cant find anything in the feed, would anyone be able to help with this? Feb 22 11:15:49 rburton: the ones that point to the package mirror Feb 22 11:15:55 rburton: in the file.... Feb 22 11:16:04 itayo: set PACKAGE_FEED_URLS Feb 22 11:16:27 itayo: erm, _URIS even Feb 22 11:16:55 PACKAGE_FEED_URIS in local.conf? Feb 22 11:16:59 yes Feb 22 11:17:02 rburton: thanks:) Feb 22 11:18:28 And could i point PACKAGE_FEED_URIS to a local file? Feb 22 11:26:42 When added to the PACKAGE_FEED_URIS i get: Debian package install does not support BAD_RECOMMENDATIONS Feb 22 11:32:09 those are unrelated Feb 22 11:32:43 (dpkg doesn't support bad recommendations, so you can't use it) Feb 22 11:42:35 ed2: tbh surprised WKS_FILES expands at all in your example Feb 22 11:43:03 ed2: using d.getVar() for both might be better Feb 22 11:47:00 rburton: it expands just fine. WKS_SEARCH_PATH also expands, but not recursively and fully. Feb 22 11:47:45 i suspect the problem is the order of expansion Feb 22 11:47:53 use getVar() to control when stuff happens Feb 22 11:53:55 ed2: yes that helps Feb 22 11:54:18 I've also noticed yesterday that using different variable name in bblayers.conf also helps Feb 22 11:54:58 so the order of expansion mentioned by ross looks likely the cause Feb 22 11:56:09 JaMa: I've sent a patch for review. Feb 22 12:10:29 Qtwebengine is not getting compiled with qt5.5 but it gets compiled with qt5.6.2 ? Feb 22 12:10:49 Is there any reason for the same ? Feb 22 12:12:21 hello Feb 22 12:12:53 why does images are in "tmp" directpory ? I mean, why is it a temporary folder ? Feb 22 12:13:49 why does images are in "tmp" directpory ? I mean, why is it a temporary folder ? Feb 22 12:20:32 ranchu: no real reason. feel free to rename TMPDIR if you want. Feb 22 12:21:36 Thx! I mean is it erased every time I call bitbake again ? or is it that calling bitbake again is like rebuild (only thing that changed) ? Feb 22 12:22:23 its only erased if you erase it Feb 22 12:23:03 i guess the name tmp came from the fact that everything inside it is generated, ie you can delete it and re-run bitbake and it will just recompile everything Feb 22 12:23:47 only recompile (without downloading again) ? Feb 22 12:25:34 I would please like to ask why we need cleansstate. Isn't bitbake will checks all packages for any change in source/makefile ? Feb 22 12:26:13 I mean in repeated build with bitbake Feb 22 12:28:54 ranchu: most of the time yes, but sometimes you do need to force a clean build Feb 22 12:29:38 Hmm, does anyone know why i can't get past "Booting kernel" when booting a genericx86 iso from usb? Feb 22 12:31:11 ... I just encounter the same issue, maka_, I think maybe it's serial/stdout issue ? what do you use ? serial or display ? Feb 22 12:31:29 I selected the graphics boot Feb 22 12:32:01 I made a few images now, and i keep getting the same thing, it instantly reboots when it hits "Booting kernel" Feb 22 12:32:29 whaich board is it ? a custom board or evm ? Feb 22 12:32:48 Im trying to boot it from usb on a normal intel laptop Feb 22 12:32:59 It used to work fine, but suddenly it doesn't Feb 22 12:34:38 However the image does work in VirtualBox ._. Feb 22 12:38:56 kanavin: any ideas why btrfs-tools-native is not buildable? bitbake btrfs-tools-native fails with ERROR: Nothing PROVIDES 'udev-native' Feb 22 12:40:13 because there's no udev-native provider Feb 22 12:40:16 there's a fix in mut :) Feb 22 12:45:52 rburton: nice :) thank you Feb 22 12:47:09 rburton: it sounds like both weston and mesa releases should happen next week... any preference for a) updates to RCs or b) sending late? Feb 22 12:47:40 jku: any idea how close the RCs are to releae? Feb 22 12:47:58 are they "we'll be shipping this" or "patches piling up already" in git Feb 22 12:48:08 my guess for weston is this week Feb 22 12:48:18 very close anyway Feb 22 12:51:04 mesa, I'm not sure but eric engestrom estimated end of next week... Feb 22 12:54:29 rburton: so maybe we decide on monday or so: weston might be out by then and the mesa 17.0.0 patches are ready so there's not much to do whatever the decision Feb 22 12:55:04 FWIW: I don't mind the .0 included in oe-core, I just thought that oe-core updates only on stable versions, so I wanted to warn that it's not the case after mesa changed versioning Feb 22 12:56:22 I cannot use recent mesa anyway, so for all my builds I'm stuck with 10.3.7 Feb 22 12:57:25 k Feb 22 12:57:32 hi Feb 22 12:57:58 is it possible to change/edit some file in output filesystem and create image again ? Feb 22 12:58:43 make the changes to the relevant recipe, and rebuild the image Feb 22 13:01:39 I understand, but is there a "quick and dirty" way just to change filesystem and then create image again. I am using "bitbake core-image-sato-sdk" Feb 22 13:02:56 not really Feb 22 13:03:29 because the moment bitbake regenerated the image you'd lose the changes Feb 22 13:05:30 I don't understand how it regenerate filesystem. Isn't there in tmp a folder which contain the filesystem with all it folders (/bin , /usr etc) ? Feb 22 13:05:36 yes Feb 22 13:05:48 Hey, i don't remember who it was, buti fixed the problem(not getting past Booting kernel) by using dd if=/path/to/image.hddimg of=/dev/usbhere Feb 22 13:05:50 what its name ? Feb 22 13:05:54 but if you changed a recipe and it wanted to rebuild the image, it would rebuild it from scratch Feb 22 13:06:16 its tmp/work/[machine]/[image]/rootfs Feb 22 13:06:33 rburton: I've sent a new gdk-pixbuf upgrade, please use that instead of the previous one -- or let me know if you want me to re-send some sets Feb 22 13:06:40 ok. I understand what you mean. Thanks Feb 22 13:06:49 jku: thanks Feb 22 13:06:58 rburton: also it has some new ugly hacks in it, just so you know :) Feb 22 13:07:17 ha Feb 22 13:07:46 sigh Feb 22 13:07:53 rburton: is the flex upgrade revert in one of your mut branches? Feb 22 13:08:15 jku: one day i'll get around to finishing my build-gtk3-from-git scripts to find this sort of nonsense Feb 22 13:08:30 JaMa: no. was the conclusion to revert? we didn't have the totally broken release. Feb 22 13:08:58 we have totaly broken flex still in oe-core Feb 22 13:08:59 and there is revert with 2 Acks on it Feb 22 13:10:58 JaMa: so there is. tagged. Feb 22 13:11:54 when changing source file of a package, we need to build that package , becuase bitbake won't detect a change in every package source code, right ? Feb 22 13:12:00 ranchu: depends what you mean by changing package source Feb 22 13:12:29 https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=2601a5c0e5304ddb0b9b1d608b5b12ce04dc3a90 https://bugs.archlinux.org/task/52256 https://lists.gnu.org/archive/html/guix-devel/2016-11/msg01221.html and few more packages even changed documentation to explicitly mention not to use 2.6.2 flex version Feb 22 13:12:58 fair enough Feb 22 13:13:37 I mean source file which is part of a package which is used by the image recipe I am using. Feb 22 13:14:13 I meant to ask if bitbake will "know" that it need to rebuild that package, or should I explicitly rebuild it, before calling image recipe. Feb 22 13:15:55 with bitbake –c compile –f Feb 22 13:16:50 ranchu: if the source is referenced directly in the src_uri of the recipe, ie SRC_URI=file://somecode.c, then it will see and rebuild Feb 22 13:16:51 if you're going into work/ and editing files, then it won't Feb 22 13:16:52 if you're doing the latter, look up devtool in the documentation Feb 22 13:17:37 thanks Feb 22 13:28:16 when do we need to do "bitbake -c patch " ? Is it for downloading source ? **** BEGIN LOGGING AT Wed Feb 22 13:31:49 2017 Feb 22 13:38:24 ranchu: dunno what that is for but bitbake -c fetch[all] is to download source atleast Feb 22 13:39:20 ranchu: according to doc : "Locates patch files and applies them to the source code" Feb 22 13:40:28 i think it might fetch first if it isnt fetched thought Feb 22 13:57:09 Does invoking bitbake might try to download upgraded files ? Feb 22 13:57:59 ranchu: if the recipes indicate a change, yes. it does not preventively check upstream servers. Feb 22 13:58:50 if I just change kernel version, will it result in trying to upgrade all packages ? Feb 22 13:59:09 thanks LetoThe2nd Feb 22 13:59:33 nope, that will just try to download the new kernel version Feb 22 13:59:42 change kernel version in local.conf Feb 22 13:59:48 OK. Thanks,. Feb 22 13:59:55 Thanks LetoThe2nd Feb 22 14:03:14 I need to apply RT patch. I see the readme in https://github.com/openembedded/openembedded-core/blob/master/meta/recipes-rt/README Feb 22 14:03:31 jku: huh, gnome-desktop-testing just failed here. does it build for you? Feb 22 14:03:38 Do I need to add COMPATIBLE_MACHINE_$MACHINE = $MACHINE AS-IS or do I need to replace $MACHINE Feb 22 14:06:25 I think I need to replace the $MACHINE as needed. Thanks. Feb 22 14:08:54 Does anyone know how to install a .deb package while building the image? Feb 22 14:09:23 maka_: a deb that is supplied to you as a binary deb and you can't rebuild it? Feb 22 14:09:56 Yea i have some .deb files that i need to install in the project, but i have no idea how to do this Feb 22 14:10:08 best way is to write a recipe that extracts the debs and re-packages them Feb 22 14:10:26 just add the deb to SRC_URI and it will get unpacked, you can then install and package it properly Feb 22 14:10:32 How exactly would i do this? Sorry i'm quite new to all of this Feb 22 14:11:17 (the problem is that a random deb will have dependencies that are valid for debian, not yocto) Feb 22 14:11:53 From what i heard these deb files do not have any dependencies Feb 22 14:12:19 literally no point in them being in a deb then :) Feb 22 14:12:37 well just do what i suggested, then you'll get a proper package with proper dependencies Feb 22 14:13:07 But how can i do that? Feb 22 14:13:26 add the deb to your SRC_URI Feb 22 14:13:32 it gets unpacked like a tarball would be Feb 22 14:13:49 install the files in do_install Feb 22 14:13:57 rburton: ah, magic - was going to suggest dpkg-deb -x :) Feb 22 14:14:15 huh /data/poky-master/tmp-glibc/work/corei7-64-poky-linux/shared-mime-info/1.8-r0/recipe-sysroot/usr/include/bits/local_lim.h:38:26: fatal error: linux/limits.h: No such file or directory Feb 22 14:14:43 How do i create that do_install? i've read alot about it but i don't really know how i would create it myself Feb 22 14:15:19 And i can't really find a proper guide or anything about those do_install, do_configure and all those functions Feb 22 14:15:41 maka_: https://wiki.yoctoproject.org/wiki/TipsAndTricks/Packaging_Prebuilt_Libraries Feb 22 14:17:12 maka_: do_install is just "put files into ${D}", how you do it is up to you. typicall, mkdir and install will be sufficient. Feb 22 14:17:53 The problem is mostly, as i said before, i'm quite new to both yocto and linux haha Feb 22 14:18:07 what would be the handiest way to determine if running under pseudo? Feb 22 14:18:18 just look at LD_PRELOAD or smth else(?) Feb 22 14:18:26 Also, thanks for that link rburton! Feb 22 14:18:34 marquiz: pretty sure there's a better variable that is set Feb 22 14:19:19 rburton: gnome-desktop-testing looks ok here Feb 22 14:19:31 rburton: i thought so, any idea what that could be? :) Feb 22 14:19:32 jku: yeah happened for shared-mime-info too :( Feb 22 14:21:03 marquiz: set [fakeroot] = 1 on do_devshell and run export in the terminal? Feb 22 14:25:46 rburton, in my deb file are 3 files, 2 tar.gz and a debian-binary file. How would i go about installing these in do_install? Feb 22 14:29:20 rburton: thanks, PSEUDO_DISABLED == "0" is the best i could find Feb 22 14:31:35 maka_: hm, i thought it would explode the tarball. you'll have to untar the right one first. Feb 22 14:31:56 (or ask whoever is giving you debs to give you a tarball like everyone else) Feb 22 14:32:27 rburton: from one of these files i could obtain a tar.gz2, would that be easier to install? Feb 22 14:33:01 yes Feb 22 14:33:04 because i tried that aswell before, and i kept getting something like: package not in feed Feb 22 14:33:51 Either that or lots of problems with the license files Feb 22 14:43:28 Where should I put a DEFAULTTUNE = "cortexa5hf-neon" directive? Should it be in the local.conf file? And if so, should it be before or after MACHINE ?= "qemuarm" Feb 22 14:44:10 how can I get the beglebone green dtb to be built? I've tried adding KERNEL_DEVICETREE += " am335x-bonegreen.dtb" to my local.conf and bitbaked my image again, but the dtb is not generated... Feb 22 14:45:39 spierepf: usually you would put it into the machine file for your specific target Feb 22 14:48:33 LetoThe2nd: I'd like to still be able to run the result using qemuarm. Is there some way to override the contents of meta/conf/machine/qemuarm.conf? Feb 22 14:49:20 LetoThe2nd: Can I create my own qemuarm.conf? Feb 22 14:50:31 What is COMMON_LICENSE_DIR ? Feb 22 14:51:31 spierepf: you can for example create your own version, like qemuarma5.conf os such. overriding mich be complicated if the tune flags clash. for a quick hack, just modify qemuarm.conf Feb 22 14:53:50 ta-daa! Feb 22 14:54:07 * kanavin is now off rburton's shame list :) Feb 22 14:54:34 kanavin: uhoh, patch bomb Feb 22 14:54:46 spierepf: just put that in local.conf Feb 22 14:55:15 urgh guile is building. i hate guile. Feb 22 14:55:18 * rburton -> coffee Feb 22 14:56:49 rburton: next version of guile should build faster Feb 22 14:57:15 six minutes on a xeon is *crazy* Feb 22 14:57:24 rburton: it has some kind of gigantic lisp compile job inside, and they finally figured they should ship pre-compiled bytecode for common architectures at least Feb 22 14:57:44 yeah half the problem is that we have to build guile-native and then guile, both of which take an age Feb 22 14:58:06 rburton: all that for *one* recipe, grub if I remember right Feb 22 14:58:46 yes Feb 22 14:58:57 hey at least your recipes are building. Feb 22 14:59:06 i wonder if i can fiddle my machine config so it doesn't use grub :) Feb 22 14:59:06 i've been beating my head against libfortran again Feb 22 14:59:30 and i can tell you, it is a mess. Feb 22 15:02:10 rburton: Does it matter where in local.conf I put the DEFAULTTUNE directive? Feb 22 15:03:34 anyone knows the difference between hddimg to iso ? Feb 22 15:13:00 how can we edit /etc files with yocto ? Feb 22 15:15:29 how can we edit /default/grub files with yocto ? Feb 22 15:38:46 <_Ben> My `packages-split` directory seems to be missing, normally it is at `build/tmp/work//python//packages-split` but that no longer exists... any ideas? Feb 22 15:42:53 _Ben: packages might have been restored from sstate cache? Feb 22 15:44:28 <_Ben> neverpanic: I did a cleansstate and rebuilt.. still nothing Feb 22 15:46:05 Hello, I have a question about an error on qemu recipe. In file: meta/recipes-devtools/qemu/qemu_2.2.0.bb, there is an SRC_URI_prepend. but it seens to miss a space at end of this line. because bitbake try to get two sources concatenated. Is that normal that is missing this space ? (I use krogoth). Thank Feb 22 15:47:13 micka: no thats a bug, iirc fixed in master so should be backported Feb 22 15:49:07 ok thank, so I can't have fix for krogoth, isn't it ? Feb 22 15:49:42 micka: you can, send a patch to the list, and it'll be added to krogoth Feb 22 15:50:40 ok Feb 22 16:51:40 Has anyone experimented with vm parameters for optimizing for yocto building? I've thrown a bunch of memory at build server, but it does not really utilize it. Feb 22 16:51:59 are you sure its not? Feb 22 16:52:13 my builds use about 3/4rs of the ram as a disk cache Feb 22 16:52:39 MemTotal: 65922568 kB Feb 22 16:52:43 Cached: 44989500 kB Feb 22 16:53:11 rburton: well, it sais 45Gb free, and 1Mb cached Feb 22 16:53:29 maybe the VM drivers behave differently as it expects the host to cache the file system? Feb 22 16:54:11 vm as in linux kernel's vm. Feb 22 16:54:33 1mb cached sounds very odd then Feb 22 16:54:44 not sure what the abbrevation is for, but not a virtual machine in this context afaik Feb 22 16:55:20 my 4gb idle nas has 3gb cached right now Feb 22 16:55:21 so, you don't do any special adaptation to the caching settings for yocto then? (This is common in databases servers) Feb 22 16:55:33 no, unlike a database server we just use normal file IO Feb 22 16:55:51 whereas a database server all the tuning is about turning off the smarts in the kernel Feb 22 16:56:20 (a database server might as well be considered a specialised OS) Feb 22 16:56:33 I convinced my boss to throw in another 64Gb memory into the blade server to aid yocto build speeds, but it doesn't really :( Feb 22 16:56:46 well if its not using any ram as a disk cache, that's your problem Feb 22 16:56:58 so look again at whatever tuning you've already done Feb 22 17:00:44 I talked to some other linux expert, and he suggested setting up building in tmpfs as a better solution. But that wouldn't work too good with yocto Feb 22 17:02:36 I've set vm.swappiness lower, as I don't want this machine to start swapping (as the swap drive is very slow) Feb 22 17:02:47 Especially if there clearly is an abundance of memory Feb 22 17:05:15 tmpfs works really well Feb 22 17:05:21 just turn on rm_work Feb 22 17:06:48 saves wear and tear on disks too Feb 22 17:07:02 which is SSD in my case Feb 22 17:07:24 if you've loads of ram, then tmpfs is the easy win Feb 22 17:07:53 put the TMPDIR in a tmpfs, and either copy out stuff you want to keep or put DEPLOYDIR somewhere else Feb 22 17:08:31 ah, great, locations can be set Feb 22 17:09:48 most of it is sstate cache utilations anyways, so you don't lose any logs or anything from it Feb 22 17:10:43 yeah Feb 22 17:12:15 allthough I'd often wished I could access the logs for a statecache object, but that is entirely another discussion Feb 22 17:16:02 If I define my own machine type (conf/machine/qemuarma5.conf) how do I invoke runqemu to use that machine type? Feb 22 17:24:46 rburton: You mentioned putting DEFAULTTUNE in local.conf. Is there a particular point in the file where I should put it, or does it matter? Feb 22 17:25:14 doesn't really matter Feb 22 17:31:52 hi Feb 22 17:32:10 I try to deploy my yocto build for intel ATOM computer Feb 22 17:32:40 I get a keyboard/mouse toggling on screen, but nothing is printed on display Feb 22 17:32:46 anyone familiar with it ? Feb 22 17:32:58 I had no errors in build. Feb 22 17:33:37 I also tried to deploy the pre-built hddimg into usb , yet the same result. https://www.yoctoproject.org/downloads/bsps/krogoth21/cherry-hill Feb 22 17:33:43 rburton: When I just add the DEFAULTTUNE directive, I get an error about "Tuning 'cortexa5hf-neon' has no defined features, and cannot be used". Which I think is because it is part of tune-cortexa5.inc which isn't included by default. Feb 22 17:33:56 rburton: where should I add the include directive? Feb 22 17:33:58 sounds like you do need a new machine Feb 22 17:34:03 * rburton doesn't understand arm arch Feb 22 17:35:08 rburton: Can you point me at some docs about defining a new machine? Feb 22 18:06:09 rburton: Can you please take a look at https://patchwork.openembedded.org/series/4820/ when you have the time? These are patches related to egg files added to path in distutils3. Not merged yet. Feb 22 18:06:31 rburton: and let me know if something else is needed. Feb 22 19:03:38 Is there an equivalent of rpm's --what-provides for the sysroot? Feb 22 19:04:03 Any way to track down which recipe/pacakge create a particular file? Feb 22 19:04:38 oe-pkgdata-utils can tell you Feb 22 19:06:01 themikenicholson: or you can use grep -R foo tmp-eglibc/sstate-control Feb 22 19:10:50 rburton: Thanks, exactly what I was looking for Feb 22 19:12:13 seems to be missing in the yocto reference manual - is there another trove of information that i should be checking before I bother the irc channel? Feb 22 19:17:38 themikenicholson: http://www.yoctoproject.org/docs/latest/ref-manual/ref-manual.html#viewing-package-information-with-oe-pkgdata-util Feb 22 19:17:48 which version of the manual are you looking at? Feb 22 19:30:55 kanavin_home: Oh, was looking at the 2.1 version Feb 22 19:31:39 in the process of moving our team to morty, still had 2.1 bookmarked Feb 22 19:33:44 themikenicholson: bookmark the above, it's 'latest' :) Feb 22 19:34:00 then you'll learn about the cool stuff as it happens :) Feb 22 20:17:20 I have a recipe that fails because configure does not support '--disable-static'. This is not mentioned in the recipe, so where is it coming from? Feb 22 20:18:06 ${EXTRA_OECONF} perhaps Feb 22 20:21:09 I try to deploy yocto to ATOM E3800, I boot with USB, yet after some printing on display it stops. Is it spme issue of display ? I tried to press shift but van't get into grub Feb 22 20:23:25 I'm looking for help me with a DEFAULTTUNE = "armcortexa5hf-neon" Feb 22 20:23:28 sveinse: conf/distro/include/no-static-libs.inc most likely Feb 22 20:30:06 kanavin_home: yeah, and then? :D (/me trying to follow the chain backwards) Feb 22 20:30:26 sveinse: grep is your friend :) Feb 22 20:31:37 sveinse: you can also use bitbake -e Feb 22 20:37:45 kanavin_home: Yes I am, but bitbake -e does not tell my why no-static-libs.inc gets pulled in in the first place Feb 22 20:41:46 Has bitbake $(filter-out)-type functionality? I'd like to keep ${EXTRA_OECONF} to the recipe's configure, but since the upstream configure don't support --disable-static, I'd like to filter it out. Feb 22 20:45:52 sveinse: it does Feb 22 20:45:58 # $EXTRA_OECONF [11 operations] Feb 22 20:45:58 #   set /home/ak/development/poky/meta/conf/bitbake.conf:498 Feb 22 20:45:59 #     "" Feb 22 20:45:59 #   _append /home/ak/development/poky/meta/conf/distro/include/no-static-libs.inc:33 Feb 22 20:45:59 #     "${DISABLE_STATIC}" Feb 22 20:47:08 ah, wait :) Feb 22 20:47:22 kanavin_home: it tells you it's set in no-static-libs.inc, right Feb 22 20:47:36 kanavin_home: but perhaps no-static-libs.inc gets included no matter what? Feb 22 20:48:41 sveinse: nope, it actually does :) Feb 22 20:48:42 #   /home/ak/development/poky/meta-poky/conf/distro/poky.conf includes: Feb 22 20:48:42 #     /home/ak/development/poky/meta-poky/conf/distro/include/poky-world-exclude.inc Feb 22 20:48:42 #     /home/ak/development/poky/meta/conf/distro/include/no-static-libs.inc Feb 22 20:48:42 #     /home/ak/development/poky/meta/conf/distro/include/yocto-uninative.inc Feb 22 20:49:02 but you could *easily* establish all of this with grep, just do it :) Feb 22 20:49:29 I did, but I don't understand the context Feb 22 20:49:41 This file is always included into the project, is that it? Feb 22 20:50:41 rburton: I can't reproduce wic build failure. I'll try again. Feb 22 20:50:48 sveinse: which file? Feb 22 20:51:04 rburton: can you merge this patchset please: http://lists.openembedded.org/pipermail/openembedded-core/2017-February/132854.html Feb 22 20:51:05 kanavin_home: ah, it is. meta-poky/conf/distro/poky.conf:108 Feb 22 20:51:55 All right, how do I cope with a upstream configure that does not like --disable-static? Feb 22 20:52:02 rburton: the failure is most probably caused by the last patchset: http://lists.openembedded.org/pipermail/openembedded-core/2017-February/133080.html Feb 22 20:52:08 a) patch the sources, to whatever avail Feb 22 20:52:23 b) filter out --disable-static to its confiure in the recipe? Feb 22 20:52:26 sveinse: EXTRA_OECONF_remove = "--disable-static" Feb 22 20:52:40 kanavin_home: good, thanks Feb 22 20:53:04 again, grep is awesome for things like this - there are several recipes that need this, and I found out by looking for 'disable-static' Feb 22 20:53:46 sveinse: ak@linux-f9zs:~/development/poky/meta> grep -ir disable-static * Feb 22 20:55:39 it's the local incarnation of 'google stackoverflow' :) Feb 22 20:56:29 kanavin_home: it is, but it's not always easy to see the context. E.g. in my repo I've got only one that use EXTRA_OECONF_remove and that is meta/recipes-extended/ed/ed_0.5.bb among some 50 matches that sets EXTRA_OECONF = "--disable-static" Feb 22 20:56:52 I like grep -rHne Feb 22 20:57:20 I found it when I knew what I was looking for Feb 22 20:57:34 sveinse: well, I'm grepping only within oe-core, and there's few results Feb 22 20:58:28 yeah, but I don't know if its poky or oe or any other fsl sources, unfortunately Feb 22 20:59:05 Another thing I've been wondering about: when to use += vs _append Feb 22 21:04:21 ranchu, linux-yocto is a kernel specific list ; you'll probably have better luck getting assistance on a generic yocto mailing list. Feb 22 21:05:34 https://www.yoctoproject.org/tools-resources/community/mailing-lists Feb 22 21:12:29 sveinse: tbh, I don't know that either, but I prefer _append, because += is too similar to =, ?= and other similar operators Feb 22 21:13:49 there's too much history of serious bugs because someone didn't notice they used the 'slightly wrong' thing Feb 22 21:28:50 hi Feb 22 21:29:21 can I expect yocto to run on different machine, by just replaing machine in local.conf ? Feb 22 21:29:45 There is many smart solutions built in to the variable system in bb, but for an outsider, there is also very much use of magic in it Feb 22 21:30:53 We consider yocto, becuase we epexct that if we replace cpu, it will be just a matter of replacing the "machine" in local.conf. Is that correct ? Feb 22 21:32:02 ranchu: yes Feb 22 21:32:28 ok. Feb 22 21:33:20 Is it correct to assume that boot time in yocto is faster than ubuntu ? Feb 22 21:33:27 I hate to be the naysayer, but it depends if all bsp for the machines are written the way they should. Feb 22 21:34:06 sveinse: yes, there's plenty of ways to make things broken, but let's assume they're not :) Feb 22 21:34:06 hi sveinse, you mean that bsp files is different in yoctoproject, right ? Feb 22 21:34:28 for each cpu, Feb 22 21:34:56 We're on custom arm HW, and it is certainly not always the case. I had to spend a few days worth of work to move our image over to another evaluation board Feb 22 21:35:36 ranchu: well, I'm no expert, but I belive the official layers are really good at this Feb 22 21:35:54 sveinse - I understand its a big efforct, yet it seem to manage to move the yocto project to another machine, right ? Feb 22 21:36:17 ranchu: boot time depends on what you place into the image you build Feb 22 21:36:36 yes, definitely. We moved our development from an embedded ubuntu type image to yocto two years ago Feb 22 21:37:11 sveine - OK, I thought you meant that you moved yocto from one machine to another. Feb 22 21:37:25 Both actually Feb 22 21:37:56 We have multiple machines/types HW, and we have moved from one type of build system to yocto Feb 22 21:37:59 so moving yocto from one machine to another is a big or small issue ? Feb 22 21:38:11 but all armv7 (cortex a9) thou Feb 22 21:38:25 from arm to intel for example Feb 22 21:38:25 ranchu: depends on the quality of the bsp layer Feb 22 21:38:53 again, my experience is that it depends. On arm, due to the system-on-chip vendors, it can vary from extremely good to not great Feb 22 21:38:54 ranchu: if both bsp layers are top quality, then you change the MACHINE setting, and rebuild, and that's all :) Feb 22 21:39:34 ranchu: what archtecture are you considering? Feb 22 21:39:47 OK, I'll gie the full story: we are in a dillema with a new intel board. not sure if use yocto or ubuntu. ubntu managed to start on board, but not yocto yet. it's just intel ATOM. not sure why. Feb 22 21:40:28 it's also have to be "realtime" (RT patch) Feb 22 21:40:56 is ubuntu less "embedded" than yocto ? Feb 22 21:42:35 yes, I think so. but with it comes availability. E.g. with yocto you need to build everything. For ubuntu, there is loads and loads of upstream packages available. Feb 22 21:43:09 our experience with running an embedded product with ubuntu, is that ubuntu is mainly adopted for desktop and server use, not so much embedded. Feb 22 21:44:19 Our product used to take 50 seconds to boot, which was acceptable. And we did a concept study to see what could be done. The conclusion was that we had to significantly patch ubuntu to be able to improve/tailor it, and then the whole motivation for using vanilla ubuntu goes away. Feb 22 21:44:26 but realtime, is a kernel issue not filesystem, right ? and boottime is mostly a matter of reducing servers/applications at startup , right ? Feb 22 21:44:27 It was then we decided to migrate to yocto Feb 22 21:44:41 now we boot in 17 secs Feb 22 21:45:22 ^ edit: 50 secs was /unacceptable/... Feb 22 21:45:42 So Isn't it possible to reduce boot time in ubuntu too ? Feb 22 21:46:13 I have ATOM with 27 seconds now with ubuntu, without doing any optimizations yet. Feb 22 21:46:32 It always is, but it is a tradeoff in how much time you want to tinker with the system to make it behave the way you want. Feb 22 21:46:57 well, things have improved with systemd I believe, so I wouldn't know for more modern ubuntus Feb 22 21:49:06 sveinse - OK, thanks for the information Feb 22 21:49:28 ranchu: you're welcome Feb 22 22:00:00 halstead: are you around? Feb 22 22:05:04 lamego: Yes but I'm in a board meeting right now. Feb 22 22:07:01 halstead: ok, just to let you know that oe-patchwork may not need the server code update. It seems to be appropriately processing patches now. We may talk about it later. Feb 22 22:07:32 lamego, I pulled changes and restarted it this morning when lsandov pinged me. Feb 22 22:07:57 halstead: that must be the reason :) thanks Feb 22 22:09:27 lamego, I'm curious if the scripts we used to get the initial patches loaded into patchwork could be useful now. Feb 22 22:12:21 halstead: are you talking about the "patchwork-update-commits" and "update-superseded" scripts? I see no reason why the couldn't Feb 22 22:13:11 lamego, Okay. It might help catch up once I get the missed patches in place. Feb 22 22:13:33 halstead: Agree Feb 22 22:14:15 lamego, Could you add your async hook to http://git.yoctoproject.org/cgit/cgit.cgi/patchwork/tree/tools ? I don't think it's tracked anywhere at the moment. Feb 22 22:14:36 halstead: sure, I will. Feb 22 22:15:02 Thanks lamego. It definitely speeds things up. :) Feb 22 22:48:45 are there any naming package rules concerning nativesdk rules? I have a meta-toolchain-lm which TOOLCHAIN_HOST_TASK += "". If I add nativesdk-packagegroup-lm-host it works fin, if I add packagesgroup-nativesdk-sp it fails for missing 'virtual/x86_64-pokysdk-linux-gcc'. It's only the filename which is different, it's the same file. Feb 22 22:49:59 Let me guess, it needs the prefix 'nativesdk-' in its name? Feb 22 22:53:51 Seems so. Renaming 'packagegroup-nativesdk-sp' to 'nativesdk-packagegroup-sp' and it works **** ENDING LOGGING AT Thu Feb 23 03:00:02 2017