**** BEGIN LOGGING AT Tue May 03 02:59:58 2016 May 03 06:01:33 how do I create the rootfs.tar.gz on yocto 2.0 and newer? I was using yocto 1.7 before and it came automatically when I created the image, but now it's not generated anymore. can I enable it somehow? May 03 06:01:43 now I only get image.ext4 May 03 06:03:32 rootfs.tar.gz containts the rootfs stuff packed in tar.gz format instead of complete image May 03 06:45:55 hi May 03 07:15:11 teemu: i don't think the default has changed in oe-core, but it also depends on which machine/distro you might be building for, or which additional layers you are using May 03 07:17:12 it's for imx6 so I use meta-fsl-arm - layer as well. May 03 07:17:44 the distro is the same on both so bblayers.conf and local.conf hasn't changed. May 03 07:18:02 if you mean those by 'distro' May 03 07:28:40 teemu: so it's probably something that changed in meta-fsl-arm. May 03 07:29:06 the variable that lists all images to build is IMAGE_FSTYPES, you should check its value in both cases. May 03 07:30:00 I found and added to local.conf IMAGE_TYPES += "tar.bz2" and now I have the rootfs.tar.bz2 again May 03 07:31:21 so yeah.. it seems that meta-fsl-arm has changed that. I compared to older version May 03 07:35:25 I'm still facing issues with linaro's external toolchain. The glibc-locale package apparently does not package several .so files in /usr/lib/gconv/ May 03 07:37:17 is that a known issue? May 03 07:37:51 and is it supposed to pack those files? Pack them in split packages? Or not pack them at all? May 03 07:39:25 also the gziped files in /usr/share/i18n/charmaps/ are not packaged either :-/ May 03 07:43:08 the generated split package to not contain DEBUG: LIBNAMES: pkg glibc-locale-el libs 0 bins 0 sonames [] though May 03 07:57:28 Hi guys, where should I set DISTRO_FEATURES? May 03 07:57:52 I would like have to be taken into account for all my recipes... May 03 07:58:40 I tought that the image recipe was the proper place... But apparently not. May 03 08:15:18 what's the use of those "tar cf ... | tar xf - ..." lines in libc-package.bbclass? looks like weird cp O.o May 03 09:07:32 How would I loop over all generated packages in a recipe in the presence of PACKAGES_DYNAMIC? Is there a better way than listing the subdirectories of ${WORKDIR}/packages-split? May 03 09:25:59 Hm, read the PACKAGES field from ${PKGDATA_DIR}/${PN} seems to be the way to go? May 03 09:29:17 neverpanic: in what context? May 03 09:30:06 I have a bbclass that's inherited by recipes, and I want to write a metadata file for each generated package May 03 09:30:32 I currently have a task that's run after do_packagedata, so reading ${PKGDATA_DIR}/${PN} should work May 03 09:31:01 I guess you could do that... otherwise you could do it within do_package May 03 09:31:12 oh right across recipes May 03 09:31:27 no, wait May 03 09:31:31 It's what buildhistory.bbclass does. for pkg in d.getVar('PACKAGES', True) doesn't work for packages generated using PACAKGES_DYNAMIC May 03 09:31:45 at the right point within do_package it will yes May 03 09:32:14 btw. PACKAGES_DYNAMIC doesn't generate any packages at all May 03 09:32:34 all it does is satisfy dependencies from BitBake's perspective May 03 09:33:25 there'll be some other logic which is actually adding items to PACKAGES May 03 09:33:53 Yes, I know -- it's just what I'm currently doing and it doesn't give me the ipv6 kernel module package, but maybe I should read the source of do_package then. May 03 09:37:54 bluelightning: ping May 03 09:38:38 neverpanic: for the kernel you will find the logic in the kernel / module classes not package.bbclass FYI May 03 09:38:44 sujith_h: pong May 03 09:39:44 Lets say I have a recipe with a proper version ( 1.1.1) and a git version. And I want to use only the git version and not the 1.1.1 version. So will PREFERRED_VERSION variable help me to use git recipe? May 03 09:40:23 or DEFAULT_PREFERENCE = "-1" should be used in the recipe with version 1.1.1 so that it won't be added in the build May 03 09:44:57 Generally I have seen most of the recipes with git go for DEFAULT_PREFERENCE = "-1". My intention is other way round May 03 09:45:24 or to be precise opposite of what I have seen :) May 03 09:50:16 Weird, my task runs after do_package, and do_package updates the PACKAGES variable, yet I write a file for pkg d.getVar('PACKAGES', True) and am missing a file :/ There's probably a bug somewhere in my code then. May 03 09:51:35 Hey guys... is it necessary to add an Ethernet package to the loca.conf file or will the ethernet always be there by default? My ethernet Amber light stays on and I have no internet or network on it May 03 10:01:00 cart_man: depends on what you put in your image really :) core-image-sato comes with connman, core-image-minimal/-base comes with ifupdown. as with everything, its up to you May 03 10:01:48 why does yocto strip sed ptest files and even fail because of it? O.o May 03 10:02:16 and more important: How do I stop it from doing that?! O.o May 03 10:06:02 LeifSo: its not stripping, it's running file on an executable to see if its an ELF and should be stripped. file then crashes, because you've got a buggy file binary. May 03 10:06:18 (see if your distro has fixed it, if not file a bug, and downgrade file) May 03 10:09:03 Anyways I got PREFERRED_VERSION when set to "git" works for me May 03 10:16:11 rburton: Ok my image is the dey-image-graphical for the IMX6 boards. I assumed the Ethernet drivers are installed by default since one can set the ip address before the build May 03 10:19:44 That was bit stupid question I asked here. Really felt sorry because git will be taken by default even if stable versions are there. And hence no need of having PREFERRED_VERSION for the same :( May 03 10:19:48 my bad May 03 10:20:22 I even missed the DEFAULT_PREFERENCE = "-1" in the git recipe which lead to all my confusion ... May 03 10:23:04 rburton: Although I can not find a package containing the Ehternet card drivers though? May 03 10:23:51 cart_man: kernel drivers? no idea, never used imx6 so have no idea if they install as modules, or built-in to the kernel. why not just run ifconfig on the target and see? May 03 10:26:03 rburton: I did.. at first it received one of those bogus random IPs EVEN though there are legit settings inside /etc/network/interface file. So I forced it going ifconfig eth0 xxx.xx.xx.xxx netmask 255.255.255.0 but still nothing May 03 10:26:20 Just the Orange/Amber light staying ON and nothing from the green light May 03 10:27:26 cart_man: you want to talk to someone like otavio who knows the HW May 03 10:34:21 Does OE contain any functionality to record uptime, # of reboots that kind of thing? May 03 10:40:18 rburton: Ok on it.. thanks allot May 03 10:58:16 rburton: thanks :) Seems that the most recent File version is broken. May 03 10:59:00 yes May 03 11:00:46 rburton: since that is likely a distribution agnostic issue: Is it already reported upstream? May 03 11:03:29 LeifSo: no idea May 03 11:33:12 as for now one can only build python-pygobject with python2, im having trouble making a recipe for building it with python3, anyone has some experience ? May 03 11:50:29 trying to make my own sdk installer for cross compilation, but my cmake hello world project fails to compile. Even configure, actually. May 03 11:50:54 error: cmTC_06a92 uses VFP register arguments May 03 11:50:54 CMakeFiles/cmTC_06a92.dir/testCCompiler.c.o does not May 03 11:51:49 I can prepare a better log, but is there any "usual" problem that causes that sort of error? May 03 11:52:33 (to be clear, that is cmake verifying and failing to compile a simple test program to verify the toolchain works) May 03 12:04:25 Here's a log: http://pastebin.com/PyB6MBvJ May 03 12:51:36 neverpanic, thanks for the info about where to send pseudo patches May 03 13:09:11 kergoth: around? May 03 13:10:30 kergoth: basically wondering how to make meta/lib/oe/maketype.py work under py3 :/ May 03 13:26:13 kergoth: no matter, figured it out May 03 13:38:48 Hi! Is there a reason UNINATIVE_URL is set using a regular assignment instead of a conditional assignment in yocto-uninative.inc? I would like to be able to override it in local.conf. May 03 13:46:25 all it does is assign 3 variables — a URL and the checksums for the downloaded file, just set those variables directly in a conf file along with making sure you INHERIT += "uninative" May 03 13:46:46 semi_: you can still assign it directly from local.conf as far as I understand it? May 03 13:50:08 RP: It seems to me that I am getting the value from yocto-uninative.inc if I check it with bitbake -e. May 03 13:51:27 If I change the assignment to conditional in yocto-uninative.inc, I get the value from local.conf. May 03 15:01:18 YPTM: Joshua is on the call May 03 15:01:22 YPTM: Ready-Access Number: 8007302996/9139049836 Access Code: 2705751 May 03 15:01:29 YPTM: Stephen Joined May 03 15:01:46 YPTM: sona joined, hi * May 03 15:02:53 YPTM: Richard joined May 03 15:07:44 YPTM: I've joined May 03 15:07:46 YPTM armin is on May 03 15:08:12 lol.. you say that every week.. ;) May 03 15:08:40 guy with the beard and long red hair.. ;) I occasionally do some Yocto Project work as well.. May 03 15:11:46 YPTM: Jussi here, also on the call May 03 15:25:43 folks - how do I file a bug? May 03 15:27:25 never mind, found it May 03 15:28:14 mortderire: just to confirm what you found, it is https://bugzilla.yoctoproject.org May 03 15:28:29 billr: thanks got it May 03 15:40:38 Hmm, the krogoth release says 4.1 is ltsi, but master shows it's 4.1.18 while the ltsi tree upstream is 4.1.17. does master no longer use ltsi, or was the linux-yocto tree updated from 4.1.17 to 4.1.18 for some reason? May 03 15:44:55 versioning schemes as random number generators ..... May 03 15:45:23 looks like 4.1.22 is coming, but 4.1.17 is released, from a quick look at http://git.linuxfoundation.org/?p=ltsi-kernel.git;a=summary, so i'm still confused May 03 15:47:32 rburton: https://bugzilla.yoctoproject.org/show_bug.cgi?id=9578 May 03 15:47:33 Bug 9578: normal, Undecided, ---, paul.eggleton, NEW , INHIBIT_PACKAGE_STRIP breaks -dbg May 03 15:47:44 rburton: we discussed this the other day. May 03 15:48:26 rburton: wanted to make sure it wasn't forgotten. May 03 15:59:09 Hmm, I don't think its appropriate for devtool to use /tmp/devtool*/ to extract sources. /tmp/ is tmpfs on some systems May 03 16:00:46 hello May 03 16:01:02 I was wondering if there is a publicly visible CI for the jethro branch? May 03 16:01:19 I'm getting an issue and want to make sure it's not something in my environment May 03 16:03:46 fredcadete: looks like the focus is primarily master, see http://autobuilder.yoctoproject.org/ May 03 16:04:05 kergoth: thanks, I had seen that and that was also my feeling May 03 16:04:08 fredcadete: http://errors.yoctoproject.org/ may b eworth checking to see if anyone else has hit the issue you're hitting May 03 16:20:12 kergoth: If I cam a cross a bug, and have a fix for it ... do I need bugzilla in order to submit a patch to fix it? May 03 16:21:03 mortderire: no just submit the patch to the lists May 03 16:21:32 rburton: l May 03 16:21:35 rburton: k May 03 16:29:36 Hmm, is meta-freescale rebased at the moment? May 03 16:41:17 anybody using busybox with systemd? May 03 16:43:16 denix: not on real h/w, but I've been working on a qemu instance the last couple of days with systemd and busybox. May 03 16:44:14 denix: mel has defaulted to systemd for our images for multiple releases now May 03 16:45:45 kergoth: busybox (and other recipes) seems to lack bunch of systemd service files and only provide initscripts, but those won't get installed w/o sysvinit DISTRO_FEATURE though. May 03 16:46:36 kergoth: so, it seems you can't do pure systemd with busybox - you also need to enable sysvinit just for the initscripts? May 03 16:47:05 I doubt one missing initscript will suddenly cause your image to explode, particularly if it's syslog, since systemd has journald May 03 16:47:06 * kergoth shrugs May 03 16:47:58 kergoth: actually, syslog is the only one that has service file :) there are bunch of others, like dhcp, http, etc. May 03 16:48:28 of course those can be provided by real apps and not busybox, but what's the point of using busybox then? :) May 03 17:02:41 I am using busybox with systemd May 03 17:02:55 only, no services from busybox, just utils May 03 17:17:48 it seems that on jethro tip, perf fails to compile for linux-yocto 3.19; it compiled ok at tag yocto-2.0 May 03 17:18:12 I'll move to linux-yocto 4.1 and will not be able to investigate further May 03 17:18:34 is there some place where I should flag this issue for the next user? May 03 17:19:37 fredcadete: bugzilla? May 03 17:20:12 denix: even if it's on an maintenance kernel on a maintenance branch? May 03 17:21:03 my own approach is WONTFIX and drop linux-yocto 3.19, I don't mean to presume the project will do otherwise :) May 03 17:21:13 bugzilla isn't master only May 03 17:21:26 and stable branches which are still in maintenance mode, supported, will stillg et bugfixes May 03 17:21:33 don't know if jethro is included in that, offhand May 03 17:21:38 it's all resource-driven anyway May 03 17:23:13 k, thanks May 03 17:23:16 submitting May 03 17:23:38 indeed May 03 17:34:55 hi kergoth! May 03 17:35:03 hey May 03 17:48:13 hi mranostay! May 03 17:49:15 georgem: oh you are here.. can i have one safe place? May 03 17:49:36 just look for the sticker in the window May 03 17:50:00 mranostay: https://upload.wikimedia.org/wikipedia/en/thumb/b/bd/Safeplacelogo.svg/1024px-Safeplacelogo.svg.png May 03 19:19:54 RP: I was mulling over the possibility of other file formats for bitbake, but it'd be problematic since the file format and the metadata are so tightly bound, two aspects of the same thing. it's not just a matter of a new parser, we'd need a clean api to wrap both aspects. maybe someday.. May 03 19:19:55 * kergoth adds to his someday/maybe list May 03 19:20:05 oh, i didn't realize bmaptool can use compressed images directly, that's handy May 03 19:20:27 so .wic.bz2 + .wic.bmap can be used directly.. nice, was wondering about that May 03 19:24:23 i like the write of ./tmp/deploy/tools/bmaptool, the forthcoming qemu bits should write a runqemu script there too, to be less dependent upon the environment May 03 19:25:01 okay, bmaptool requiring file: for arguemtns which are clearly local paths is a bit silly May 03 19:30:05 kergoth: interestingly that really annoyed me May 03 19:30:11 (install to deploy) May 03 19:31:37 we explicitly tell people not to poke into sysroots directly.. May 03 19:31:46 so where else would you write a convenience script? May 03 19:31:50 :) May 03 19:45:27 * kergoth mulls over poisoning the default path prefix for cross/native to make sure reloc issues are caught immediately, not only when used from sstate in a different location May 03 20:28:12 hmm, am I missing something when sharing sstate and appending "PATH" to the SSTATE_MIRROR? it ends up dropping the 2-char hierarchy and symlinks files in the root of sstate-cache? is that right? May 03 20:28:51 denix: by default it downloads to SSTATE_DIR as the destination, even though it came from a different location May 03 20:29:13 denix: i recommend ;downloadfilename=PATH May 03 20:30:22 kergoth: a-ha, thanks! May 03 20:30:44 should probably change that default at some point, but this'll do for now May 03 20:30:44 np May 03 22:15:41 kergoth: FWIW top 3 commits of http://git.yoctoproject.org/cgit.cgi/poky-contrib/log/?h=rpurdie/wip has enough to make "bitbake bash" work under python3, admittedly a bit hacked together May 03 22:16:00 kergoth: gives an idea of the sized hammers we need May 03 22:16:11 nice, will check it out May 03 22:16:52 kergoth: some bits can be merged straight in as 2.7 compatible and we just regressed, some of the character conversion pieces will need to be better thought out May 03 22:17:24 * kergoth nods May 03 22:17:50 kergoth: do you think we just have a flag day and switch? trying to support 2 and 3 seems like it would be problematic... May 03 22:19:07 a *lot* of modules seem able to support both with modules like six, 2to3 or 3to2, etc. I'm not sure if that'd be viable for us, though, given that we have so much python outside bitbake itself, and bitbake isn't just a python package May 03 22:19:11 If someone knows... I'm looking for the quick module in meta-qt5... May 03 22:19:47 On krogoth branch. May 03 22:20:37 search for 'quick' on the layer index May 03 22:20:38 scroll down May 03 22:20:39 I tried: qtdeclarative qtquick1 qtquickcontrols qtquickcontrols2 May 03 22:20:40 click on http://layers.openembedded.org/layerindex/recipe/4774/ May 03 22:20:41 done May 03 22:20:57 tried where? in what context? doing what? May 03 22:21:02 you've not given any useful information about what you're trying to do May 03 22:21:26 kergoth: its the python in the metadata which would worry me the most with that May 03 22:21:31 Ok ok. May 03 22:21:44 RP: agreed May 03 22:23:17 I just have my Qt application inside with a recipe inside Qt. And I am trying to find the proper DEPENDS to get my application compiling. And there is now way the module "quick" can be found. May 03 22:23:22 RP: it does seem possible to use six to create code which works with either, there are few actual *syntax* problems, i just don't know if we care enough to bother with that. we'll have to modify code to use that anyway. my main concern would be conversions of external layers May 03 22:23:47 *inside yocto May 03 22:23:51 present: my guess is you didn't inherit the needed qt bbclasses in your recipe May 03 22:24:31 kergoth, can be true! I just have "inherit qmake5" for now. May 03 22:25:37 Still not that sure since the QML module can be found at least. May 03 22:25:53 "QT += qml quick" from the .pro file May 03 22:27:36 Though I was kind of sure this inherit was sufficient... May 03 22:29:41 I'll try this one then: require recipes-qt/qt5/qt5.inc May 03 22:31:16 Nope same error :/ May 03 22:31:18 Project ERROR: Unknown module(s) in QT: quick May 03 22:34:10 kergoth: I don't doubt its possible, as you say its just a question of effort. We'll have to change the code regardless so should we not just move to 3? I guess we need to have this discussion on the architecture list May 03 22:36:03 I just removed quick and the recipe ended properly. Added again the module quick... and complained. May 03 22:36:56 I have to find which recipe from meta-qt5 provides the "quick" module. May 03 22:37:42 kergoth, thanks for trying to help by the way! :) May 03 22:46:16 There are renaming in Qt that are hard to follow... May 03 22:49:57 RP: yeah, I'm inclined to say just go to 3, bite the bullet and get it over with, as you've suggested May 03 23:00:11 g'night **** ENDING LOGGING AT Wed May 04 02:59:58 2016