**** BEGIN LOGGING AT Thu May 28 02:59:58 2015 May 28 06:05:20 <[w00t]> Hi, I've been having trouble with booting the yocto-built linux kernel using extlinux. I've tried both the standard (no change in config) and altering the config (both with and without initrd) but it always complains that it is unable to find a valid root file system. Booting with my custom made kernel is however possible. Has anyone else encounter May 28 06:05:20 <[w00t]> ed similar issues? May 28 06:13:25 [w00t], does your custom include a different fstypes? May 28 06:15:38 <[w00t]> redengin: I've tried including as many fs types as possible in both the custom and for the yocto config May 28 06:16:46 have you tried a diff between the yocto .config and your custom .config? May 28 06:18:32 <[w00t]> redengin: I'm about to :) I'm guessing that there will be quite a big difference as the gentoo one features a lot less modules May 28 06:21:03 [w00t], sorry, I don't have any experience with extlinux to know what it may specifically need May 28 06:22:35 what fstype is your root? May 28 06:22:57 <[w00t]> ext3 May 28 06:23:16 <[w00t]> And it's included in the kernel May 28 06:24:29 [w00t], what device is it on? May 28 06:24:38 <[w00t]> It seems like either the module isn't loaded properly or it is somehow unable to find the root... at first I got a similar issue when the initrd failed to locate /sbin/init due to it being symlinked to /sbin/init.sysvinit May 28 06:24:56 <[w00t]> redengin: It's a virtual machine May 28 06:25:05 <[w00t]> redengin: vmware May 28 06:25:20 [w00t], it'll still need to be mapped to some sort of device scsi, ide, sata, etc. May 28 06:26:11 <[w00t]> sata May 28 06:27:34 and its only the kernel that changes between your test cases? May 28 06:29:27 <[w00t]> redengin: yes May 28 06:36:44 <[w00t]> I'll try to build it as similar to my custom one as possible and se if it works. Probably just missed some small feature.. May 28 06:46:49 while you're building, do a diff between the two configs... thats probably going to show the critical difference May 28 06:54:01 <[w00t]> redengin: I did.. It differs a lot May 28 06:54:40 can you pastebin it, there aren't a lot of things that would influence initrd or filesystems May 28 06:58:00 <[w00t]> redengin: Unfortunately I'm not allowed to May 28 06:59:24 <[w00t]> redengin: But I've got different architectures in the config, and also the yocto sets CONFIG_EMBEDDED May 28 07:01:01 <[w00t]> redengin: Pentium M vs i686 May 28 07:02:10 hi guys May 28 07:11:57 hi May 28 08:13:03 is there a way to create system-wide environment variables in yocto? May 28 08:15:24 jaeckel: can you explain what you mean by that? May 28 08:20:03 brb May 28 08:25:47 I currently have a problem when using gnupg from init scripts because it doesn't correctly resolve ~ to /root but to /, so I wanted to set GNUPGHOME to point to the correct folder May 28 10:32:09 Is anyone here familiar with the difference between inheriting distutils and setuptools when building python packages? May 28 10:33:26 I'm building this recipe: http://pastebin.com/aSfcLFDd (it builds the SOAPpy package) May 28 10:35:11 which has dependencies... if I use setuptools it will put the SOAPpy folder directly under the site-packages, like this: http://pastebin.com/KS5YnNBz May 28 10:45:49 However, when using distutils it will extract all dependencies but put in incorrectly under site-packages, like this: http://pastebin.com/sdRguwMq May 28 10:47:25 I could just move all the site-packages subdirs to the dir above but I feel like I'm not using the tool correctly and should perhaps pass a flag or something May 28 11:43:10 good morning May 28 11:43:39 is Bernhard Reutner-Fischer available in this channel? May 28 11:44:35 khem: ping May 28 11:45:31 khem: what exactly did you mean by "may be you can insert a custom task after useradd_sysroot ?"? May 28 12:10:58 bluelightning: perhaps you know what a custom task would be? Sorry for the highlight. May 28 12:17:39 do_sometask() { } May 28 12:17:58 addtask sometask after useradd_sysroot May 28 12:27:16 abelloni: literally like that in the recipe? May 28 12:34:13 lpapp: sure May 28 12:40:53 abelloni: actually it looks even simpler, it seems that the useradd task runs before do_install, so I can probably do my own thing in there? May 28 12:41:13 Please refer to: https://github.com/openembedded/oe-core/blob/master/meta-skeleton/recipes-skeleton/useradd/useradd-example.bb#L60 May 28 12:41:21 "# The new users and groups are created before the do_install # step, so you are now free to make use of them:" May 28 12:43:00 although it is unfortunate that things like ${datadir} is not documented. May 28 12:43:47 I'm not sure you understand what a community project is May 28 12:44:50 given that I have been doing that for well over a decade, it is unlikely. :) May 28 12:45:10 (not that my comment does not stand still) May 28 12:45:20 (community project or not) May 28 12:47:31 I mean it has been at least one year that I see you complaining about things not working properly or not being documented May 28 12:48:20 yeah, not very long time. I am relatively new to the project. May 28 12:48:37 and by complaining you mean providing feedback, right? May 28 12:48:47 (most of which were fixed by Scott, thankfully) May 28 12:51:03 hmm, it seems to be documented actually, just not searchable by the direct variable names: http://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#var-CONFFILES May 28 12:51:22 (I guess I did not know what else to look for other than the variable names) May 28 12:54:15 I wonder if there is a better place to put this information at, in the documentation as this knowledge is not CONFFILES specific. May 28 12:54:30 I was personally looking for it under the bitbake section. May 28 13:14:10 hello guys I meet the following do_rootfs warning: libstdc++-staticdev-4.9.1-r0.0@ppce500mc is already installed May 28 13:15:01 indeed the package appears in the log file defined twice May 28 13:15:14 in the "NOTE: Installing the following packages:" May 28 13:15:29 but I cannot seem to understand why May 28 13:15:37 could you please help me? May 28 13:15:54 which recipe does it come from? May 28 13:16:00 two different or the same? May 28 13:17:12 to be honest I do not know May 28 13:17:20 I am now searcing for them :D May 28 13:17:44 I think that is the first thing to figure out May 28 13:17:59 but if you do not mean to install it on your own, then PRIVATE_LIBS will not be necessary. May 28 13:19:00 it seems it is brought by two different packagegroups May 28 13:19:12 that shouldn't be a problem May 28 13:19:27 that is what I am thinking also May 28 13:19:56 I think there's a bug here somewhere (possibly in smart) and we are only seeing it now because we are exposing warning messages printed during do_rootfs May 28 13:20:00 but I believe bitbake know to reunite two lists of packages May 28 13:20:39 smart could impact the do_rootfs task? May 28 13:20:58 noob question because i do not know much about it :D May 28 13:22:42 also my image does not contain smart in its packages May 28 13:25:09 AlexVaduva: when you have RPM packaging selected, we use smart to do the actual image assembly during do_rootfs - ultimately we just send a list of packages to smart and it figures out all of the dependencies and determines the final list of packages that need to be installed May 28 13:25:24 aah ok May 28 13:25:37 so it must be a smart bug somewhere May 28 13:25:53 well, the question is, are we sending the same package twice to smart, or is this coming about due to the internal behaviour of smart itself May 28 13:26:02 so this is why it finds duplicate packages May 28 13:26:12 if we are sending the same package twice, that would probably be a bug on our side May 28 13:26:25 well the packages May 28 13:26:27 another option is we could whitelist this warning May 28 13:26:32 because there are a few May 28 13:26:45 hmm, there is no bitbake variable for the home holder. May 28 13:26:46 seem to apear in more then one packagegroup May 28 13:26:49 folder, sorry. May 28 13:26:51 though I don't know that we already have whitelisting functionality for the do_rootfs warning code May 28 13:27:03 something like ${homedir} May 28 13:27:22 588:NOTE: warning: libstdc++-staticdev-4.9.1-r0.0@ppce500mc is already installed May 28 13:27:22 589:warning: libexpat-dev-2.1.0-r0.0@ppce500mc is already installed May 28 13:27:22 590:warning: libc6-staticdev-2.20-r0.0@ppce500mc is already installed May 28 13:27:22 591:warning: libstdc++-dev-4.9.1-r0.0@ppce500mc is already installed May 28 13:27:22 592:warning: libusb-1.0-dbg-1.0.19-r0.0@ppce500mc is already installed May 28 13:27:23 593:warning: gettext-dev-0.18.3.2-r0.0@ppce500mc is already installed May 28 13:27:25 594:warning: libgomp-dev-4.9.1-r0.0@ppce500mc is already installed May 28 13:27:27 595:warning: libc6-dev-2.20-r0.0@ppce500mc is already installed May 28 13:27:29 596:warning: libc6-dbg-2.20-r0.0@ppce500mc is already installed May 28 13:27:31 597:warning: libgcc-s-dev-4.9.1-r0.0@ppce500mc is already installed May 28 13:27:33 these are the warnings May 28 13:27:48 right, that doesn't really mean much without context though May 28 13:28:21 lpapp: if you mean the home directory for a specific user, that wouldn't work May 28 13:28:31 bluelightning: for the generic /home location. May 28 13:28:51 I guess noone has cared to move that up to now... May 28 13:28:54 it just feels a bit out of place to type /home/foo/.profile beside the other variables. May 28 13:29:05 there is only ROOT_HOME that specifies the home directory for the root user May 28 13:29:09 I am still pondering if it is a valid thing to do, but I think it is. May 28 13:29:26 since it is common in server setup that services are run under a particular user, which is intentionally not root. May 28 13:29:42 sure, /home does not have much meaning in desktop content for packages, but on servers, it may. May 28 13:29:53 server is a bit specific. May 28 13:29:54 yes but for service users you shouldn't really be using .profile for anything May 28 13:30:07 so what should I use then for extending the PATH? May 28 13:30:09 .profile AFAIK is for interactive shells May 28 13:30:13 or I should not extend the PATH in the first place? May 28 13:30:31 either don't, or do it in the script that starts the service May 28 13:30:33 I think it is OK to put suid binaries into /sbin. May 28 13:31:03 bluelightning: that is fine as long as the devs remember to run that services through that script and not manually. May 28 13:31:12 so it is interactive May 28 13:31:19 in which case .profile would be the way to do it then May 28 13:31:25 bluelightning: the only problem with that approach is that you have to take care in the code when doing exec, etc. May 28 13:31:38 with .profile, that is no issue in the code. May 28 13:32:06 well, the other problem is obviously that you cannot run the services easily without using the init service May 28 13:32:08 did you consider auto-extending PATH in the global /etc/profile based on some other criteria (group membership) as I suggested the other day? May 28 13:32:11 one more thing to replicate otherwise. May 28 13:32:29 yes, I did consider and deemed complex :) May 28 13:32:33 or unusual. May 28 13:32:35 at least for me. May 28 13:32:37 complex? it's trivial May 28 13:32:54 complex in the sense that it makes the system non-standardy May 28 13:33:00 profiles were meant for this, I think. May 28 13:34:28 therefore I was looking for ${homedir/foo/.profile. /home/foo feels out of place when everything starts with variables in the recipe for the installation instructions. May 28 13:34:34 it is just stylistic, sure, but still :) May 28 13:34:48 well, not necessarily stylistic, but yeah. May 28 13:34:50 it's up to you, but that is my advice May 28 13:35:22 bluelightning: etc has security issues, too. May 28 13:35:34 -> /home/nic/.profile is not visible by others May 28 13:36:32 therefore, it is a bit unusual request from desktop point of view, but it makes sense to me from server point of view. May 28 13:37:49 1) I would like to make it distro policy, not hard coded in the services 2) I would like to constrain its visibility to that particular user 3) I would like to have consistent variable usage for that installation - These are my requirements. May 28 13:38:27 4) Ideally, I would like to make it so that the devs can avoid the initscripts, so that they can just use NFS and run the binary from there without adjustments. May 28 13:38:39 it seems to me these are only fulfilled by /home/foo/.profile May 28 13:38:59 unless you revoke the read access for /etc/profile May 28 13:40:53 (but that would have other problems, like users would not be able to read while this information relates to them, too!) May 28 13:41:17 so that is why I think I wanted /home/foo/.profile over other solutions. It is not an easy decision as /home/foo/.profile looks painful from package point of view. May 28 13:41:21 I may have missed something. May 28 13:44:10 I considered the busybox change that Bernhard created after my suggestion, but that also has issues :) May 28 13:44:33 1) You pollute your filesystem with a skeleton file 2) Busybox specific solution, not buildsystem specific. May 28 13:44:58 3) It is not yet in busybox, so it would mean some additional maintenance to me. May 28 13:45:17 (it is also questionable whether it gets at all, at this point) May 28 13:45:21 in* May 28 14:40:54 seems the fetcher change to not use sums for hg broke vim downloads... May 28 14:41:37 reverting commit 70c1d222c08ba5bdca416c6f1c2bcb0f15c93126 fixes it May 28 14:42:05 you need to cleanall in order to force a fetch, else you won't see it. May 28 14:42:28 ERROR: Function failed: Fetcher failure: Fetch command failed with exit code 255, output: May 28 14:42:28 abort: repository /home/paul/poky/build/downloads/hg/vim.googlecode.com/hg/vim not found! May 28 14:49:31 morning May 28 14:56:51 hmm, interesting QA issue: gcc-cross-canadian-arm: found library in wrong location May 28 15:02:27 bluelightning: it seems that you have done some fixes for that claiming that it was still over zealous May 28 15:03:34 lpapp: I believe it is yes, you may choose to simply disable it May 28 15:04:18 bluelightning: so how can I pin it down whether or not it is a false positive? May 28 15:44:52 hi guys May 28 15:51:34 Exception: ExpansionError: Failure expanding variable SRCPV, expression was ${@bb.fetch2.get_srcrev(d)} which triggered exception FetchError: Fetcher failure: Fetch command failed with exit code 128, output: fatal: '/home/xxx/DriverSrc' does not appear to be a git repository fatal: Could not read from remote repository. Please make sure you have the correct access rights and the repository exists. May 28 15:53:56 any ideas? May 28 15:54:58 sounds like you made your recipe point to a nonexistent local repositroy May 28 15:55:05 which would be exactlyw hat it says May 28 16:01:56 bitbake core-image-weston gives me following error. Please check Exception: ExpansionError: Failure expanding variable SRCPV, expression was ${@bb.fetch2.get_srcrev(d)} which triggered exception FetchError: Fetcher failure: Fetch command failed with exit code 128, output: fatal: '/home/xxx/DriverSrc' does not appear to be a git repository fatal: Could not read from remote repository. Please make sure you have the correct access rig May 28 16:02:13 yes, you already said that. in multiple channels May 28 16:04:25 kergoth:my low connection May 28 16:53:46 Hello! I just tried baking an image (something very common), however I just got ~33 error messages like this one: "ERROR: QA Issue: Package version for package kernel went backwards which would break....... [version-going-backwards]". I googled and it appears the maintainers of some layer/recipe decided to change stuff around. How can I repair this in my system? Do I need to clean everything and start from scratch, or will this go away th May 28 18:18:20 hi all. What is the best way to use the linux-yocto-rt? dependency in the image and preferred in the machine.conf only? May 28 18:20:28 if i use the yocto-bsp to create mybsp, i can select only the 3.10 for the linux-yocto-rt kernel, and it also provides 3.10 bbappend with .scc files but not for the 3.14, no choice. May 28 18:21:56 but i can see the 3.15 for the rt in the dizzy branch, and if i remove the preferred version from 3.10 i can build the 3.14. In any case i can run only the standard but not the rt May 28 18:46:33 How can I update my local.conf to include pgrep and pkill? Thanks! May 28 18:48:37 I added IMAGE_INSTALL_append = " procps" but it didn't give me pgrep or pkill May 28 19:00:16 halstead: I'm trying to upload an opkg release to downloads.yoctoproject.org but getting a time out from ssh. It's been ages since I last did this, anything changed? May 28 19:11:18 paulbarker, Yes. We moved it to a faster network but ssh is now blocked by the firewall. You'll need to bounce via another server. I'll send the config via e-mail. May 28 19:18:03 halstead: Cheers May 28 21:03:22 hi May 28 21:58:42 kergoth: there is a problem with this update_data removal :(. d.keys() doesn't list the override expanded version of keys. Turns out making it do so is expensive too :( May 29 01:36:13 knock knock May 29 01:37:16 Trying to follow http://www.yoctoproject.org/docs/1.1/yocto-project-qs/yocto-project-qs.html but I get an error from runqemu - "version `GLIBC_2.14' not found". All bug reports I can find are from 2012 :( May 29 01:37:31 This is on kubuntu 14.04LTS May 29 01:49:50 Hmmm - busy channel **** ENDING LOGGING AT Fri May 29 02:59:59 2015