**** BEGIN LOGGING AT Thu Aug 30 03:00:01 2012 Aug 30 06:46:36 hi Aug 30 06:47:35 when I build an image I get an extra tarball containing the kernel modules configured to be build in the defconfig of the kernel, but they are not included in the rootfs image Aug 30 06:47:50 is there someway to configure them to be included in the image? Aug 30 06:49:11 hico hm? Aug 30 06:49:19 normaly its in rootfs /boot Aug 30 06:49:34 but maybee that changed lately Aug 30 06:49:43 not the kernel image, but kernel modules which should be in /lib/modules/... Aug 30 06:50:30 there's an extra tar ball in the deploy directory which contains them but I'd like to have them directly included in the image Aug 30 07:01:33 ok i see there seems to be a meta package called kernel-modules Aug 30 07:01:50 i guess i must include this Aug 30 07:04:25 yes not all kernel modules are installed by default Aug 30 07:04:36 add them to the image depend Aug 30 07:04:41 in your image file Aug 30 08:14:41 hi bluelightning Aug 30 08:39:23 hi silvio_, all Aug 30 08:46:35 I try to compile and install qwt with my own recipes, but looking the install log I read these strange linees, it is normal or I had wrong recipes Aug 30 08:47:10 install -m 755 -p "../lib/libqwt.so.6.0.1" "/home/oe/output/work/armv7a-vfp-angstrom-linux-gnueabi/qwt-6.0.1-r0/image/home/oe/output/sysroots/i686-linux/usr/lib/libqwt.so.6.0.1" Aug 30 09:07:03 khem: ping Aug 30 09:32:13 hello Aug 30 09:32:19 hi Aug 30 09:34:16 got a patch for git://github.com/openembedded/oe-core.git about cronie Aug 30 09:34:50 in sources/openembedded-core/meta/recipes-extended/cronie/cronie/crontab which does not contain the run-parts stuff Aug 30 09:34:58 (or is it that way on purpose ?!) Aug 30 09:35:35 shall I post that to the patchwork mailing list ? Aug 30 09:35:51 please post the patch to the OE-Core list yes Aug 30 09:36:11 ok, got to find the RTFM again in my bookmarks :) Aug 30 09:37:10 silvio_ - it is normal, that's the package's tree before it get actually packaged (and copied in the rootfs archive) Aug 30 10:42:42 morning all Aug 30 10:44:12 hi pb_ Aug 30 10:47:00 hi bluelightning Aug 30 10:48:27 pb_: hi Aug 30 10:49:27 hi jama Aug 30 10:49:42 have you seen my request for git hooks? Aug 30 10:49:48 um, no Aug 30 10:49:51 where did you make that request? Aug 30 10:49:57 here :) Aug 30 10:50:00 ah, heh Aug 30 10:50:05 no, I didn't see that Aug 30 10:50:18 I whould like to use same git hooks OE is using to sync with github Aug 30 10:50:22 can you share them? Aug 30 10:50:26 oh, yeah, sure Aug 30 10:50:31 IIRC you wrote them Aug 30 10:50:39 though, we don't actually have a git hook, it's just a cron job that runs every 5 minutes Aug 30 10:50:54 ah I see Aug 30 10:51:01 we did consider making a hook but it wasn't clear that this would be an improvement. Aug 30 10:51:21 because of possible delay when github is not responding or something? Aug 30 10:51:41 yeah, and because of possible issues to do with multiple people committing simultaneously/in quick succession Aug 30 10:52:02 I have found some hooks for syncing but all looked a bit overcomplicated for such job Aug 30 10:52:06 maybe that's why Aug 30 10:52:15 but cron job should work for me too Aug 30 10:56:22 ok Aug 30 10:56:27 I can fish out the cron script we use if you like Aug 30 10:56:42 yes, please Aug 30 10:57:45 seems to be as simple as: Aug 30 10:57:46 for i in *.git; do Aug 30 10:57:47 (cd $i && git fetch origin && git remote prune origin && git push github ) || true Aug 30 10:57:47 done Aug 30 10:57:48 heh Aug 30 10:58:24 I just run that every 5 minutes out of cron, protected with flock. Aug 30 10:58:32 great, thanks :) Aug 30 11:05:47 hi pb Aug 30 11:11:22 hi woglinde Aug 30 12:28:57 I made a recipe to build qwt from source, the binary was right compiled for arm, but are put in a strange dir, someone could help me to find error o rwhere what to it?: Aug 30 12:29:10 the dir is: Aug 30 12:30:00 home/oe/output/work/armv7a-vfp-angstrom-linux-gnueabi/qwt-6.0.1-r0/image/home/oe/output/sysroots/i686-linux/usr/lib/libqwt.so.6.0.1 Aug 30 12:30:06 but shoul be: Aug 30 12:30:18 home/oe/output/work/armv7a-vfp-angstrom-linux-gnueabi/qwt-6.0.1-r0/image/usr/lib/libqwt.so.6.0.1 Aug 30 12:30:22 any suggest? Aug 30 12:49:32 silvio_: that is the expected path within the image directory Aug 30 12:50:05 silvio_: if you have a look at the package/ subdirectory (and packages-split) you can see that the final target path there is the expected path for the target Aug 30 12:52:17 hm dont we already have a recipe for qwt? Aug 30 12:54:52 yes but is old Aug 30 12:55:03 woglinde: yes but is old Aug 30 12:55:17 than send patches Aug 30 12:56:51 woglinde: if I could I would already done ;) Aug 30 12:57:13 woglinde: i mean to write good patches Aug 30 13:00:32 even when they are not so good Aug 30 13:00:41 you will get feedback and getting better Aug 30 14:51:59 re Aug 30 14:52:31 what's the best way to change the default hostname of the image I'm generating ? Aug 30 14:54:09 where can I find the defintion of a 'virtual package'? I see refs in some recipes but no description of what it is... Aug 30 14:56:48 'virtual' usually just means multiple recipes exist providing that functionality, and you have the ability to cohose which you prefer Aug 30 14:57:11 e.g. virtual/libc can be eglibc or uclibc Aug 30 14:57:54 ""By default empty packages are ignored. Occasionally you may wish to actually create empty packages, Aug 30 14:57:54 typically done when you want a virtual package which will install other packages via dependencies Aug 30 14:57:54 without actually installing anything itself. Aug 30 14:58:27 JimNH: if you mean virtual/* as kergoth is referring to, that's a recipe-level thing rather than package Aug 30 14:59:38 yeah, ahrd to tell if he's talking about tasks or virtual/'s Aug 30 14:59:55 stupid 'package' word ambiguity :) Aug 30 15:00:07 :) Aug 30 15:01:02 for added flavour there's also the variable naming convention for runtime package selections (VIRTUAL_RUNTIME_*) Aug 30 15:01:28 that *is* at the package level Aug 30 15:02:04 ah ok tks; I was dealing with the recipies and see them used but didnt see a description of exactly why or how Aug 30 15:02:52 ultimately virtual/* is also just a naming convention - one recipe can put anything in PROVIDES and another can add that thing to DEPENDS and then a relationship is created Aug 30 15:03:07 (a build-time relationship) Aug 30 15:03:20 I just read the section about /var files, and actually I do need to store volatile stuff in /var for my needs, but the documentation does not really says how to declare those as volatile Aug 30 15:05:19 Just did a clean install of bitbake / oe on Ubuntu 12.04; after resolving some package issues, I tried bitbake nano.... and got an error configuring quilt: Aug 30 15:05:29 "You appear to have a `cp' that does not support hard links" Aug 30 15:06:29 I followed the directions on Aug 30 15:07:14 "NOTE: package quilt-native-0.48-r7.1: task do_configure: Failed" Aug 30 15:08:02 Anyone know how to fix this? Aug 30 15:08:22 robtow: are you starting some new effort or do you need to be using the old OpenEmbedded? Aug 30 15:08:56 bluelightning - started a totally new effort; this is intended as a calesthenic. Aug 30 15:09:12 Deliberately following the getting started directions. Aug 30 15:09:23 robtow: ok, in which case I would recommend starting with OE-Core rather than those instructions Aug 30 15:09:35 there is a red notice at the top of that page, maybe the wording needs to be changed Aug 30 15:09:37 http://www.openembedded.org/index.php/OpenEmbedded-Core Aug 30 15:09:37 AH, got it. Aug 30 15:09:42 Thanks. Aug 30 15:10:22 Yes, I think the "getting started" page should have the new directions, and the old should be pushed back to a secondary page. Aug 30 15:12:19 that's what we'd like to do, unfortunately few seem to be able to find time to work on restructuring the wiki Aug 30 15:13:13 there's loads of outdated info that needs sorting through :( Aug 30 15:22:32 bluelightning - I understand. I recently found that TI is using bitbake to support a SOC family, and a wireless networking company is trying to use it - and running into configuration difficulties. Aug 30 15:53:40 first patch sent \o/ Aug 30 15:56:56 bb Aug 30 16:05:35 second patch sent \o/ Aug 30 16:05:51 Tried setting up the new oe; failed on initial build.... Aug 30 16:06:24 | ERROR: Function failed: sysroot_stage_all (see /mnt/hgfs/rob/Linux-extras/oe/oe-core/build/tmp-eglibc/work/x86_64-linux/quilt-native-0.60-r0/temp/log.do_populate_sysroot.15224 for further information) Aug 30 16:07:29 robtow: any more details than that? Aug 30 16:10:17 yes..... Aug 30 16:10:20 DEBUG: Executing python function sstate_task_prefunc Aug 30 16:10:21 DEBUG: Python function sstate_task_prefunc finished Aug 30 16:10:21 DEBUG: Executing python function do_populate_sysroot Aug 30 16:10:21 DEBUG: Executing shell function sysroot_stage_all Aug 30 16:10:21 tar: ./quilt/compat/awk: Cannot utime: Operation not permitted Aug 30 16:10:21 tar: Exiting with failure status due to previous errors Aug 30 16:10:21 DEBUG: Python function do_populate_sysroot finished Aug 30 16:10:22 ERROR: Function failed: sysroot_stage_all (see /mnt/hgfs/rob/Linux-extras/oe/oe-core/build/tmp-eglibc/work/x86_64-linux/quilt-native-0.60-r0/temp/log.do_populate_sysroot.15224 for further information) Aug 30 16:10:39 is it the "cannot utime"? Aug 30 16:11:50 bluelightning some kind of awk error? Aug 30 16:14:09 Hmmmm....... Aug 30 16:14:14 robtow@Ubuntu-64bit-12:/mnt/hgfs/rob/Linux-extras/oe/oe-core/build$ utime Aug 30 16:14:14 The program 'utime' is currently not installed. You can install it by typing: Aug 30 16:14:14 sudo apt-get install atfs Aug 30 16:15:32 Still fails.... Aug 30 16:15:42 robtow: that's not the solution Aug 30 16:16:09 Anbluelightning - any idea? Aug 30 16:16:25 I'm puzzled as to why you would see that error and I don't, I'm also running ubuntu 12.04(.1) here Aug 30 16:16:36 Indeed Aug 30 16:18:25 ERROR: Function failed: sysroot_stage_all (see /mnt/hgfs/rob/Linux-extras/oe/oe-core/build/tmp-eglibc/work/x86_64-linux/quilt-native-0.60-r0/temp/log.do_populate_sysroot.16224 for further information) Aug 30 16:18:25 ERROR: Logfile of failure stored in: /mnt/hgfs/rob/Linux-extras/oe/oe-core/build/tmp-eglibc/work/x86_64-linux/quilt-native-0.60-r0/temp/log.do_populate_sysroot.16224 Aug 30 16:18:25 Log data follows: Aug 30 16:18:25 | DEBUG: Executing python function sstate_task_prefunc Aug 30 16:18:25 | DEBUG: Python function sstate_task_prefunc finished Aug 30 16:18:25 | DEBUG: Executing python function do_populate_sysroot Aug 30 16:18:26 | DEBUG: Executing shell function sysroot_stage_all Aug 30 16:18:26 | tar: ./quilt/compat/awk: Cannot utime: Operation not permitted Aug 30 16:18:27 | tar: Exiting with failure status due to previous errors Aug 30 16:18:27 | DEBUG: Python function do_populate_sysroot finished Aug 30 16:18:28 | ERROR: Function failed: sysroot_stage_all (see /mnt/hgfs/rob/Linux-extras/oe/oe-core/build/tmp-eglibc/work/x86_64-linux/quilt-native-0.60-r0/temp/log.do_populate_sysroot.16224 for further information) Aug 30 16:18:41 could you please use pastebin for long messages? thanks Aug 30 16:18:46 sorry Aug 30 16:18:47 sure Aug 30 16:19:33 http://pastebin.com/wAa9TnKV Aug 30 16:20:20 what do you get if you run: ls -ld /mnt/hgfs/rob/Linux-extras/oe/oe-core/build/tmp-eglibc/work/x86_64-linux/quilt-native-0.60-r0/sysroot-destdir /mnt/hgfs/rob/Linux-extras/oe/oe-core/build/tmp-eglibc/sysroots/x86_64-linux Aug 30 16:20:42 this is some kind of odd file ownership issue judging from google results Aug 30 16:21:02 ls: cannot access /mnt/hgfs/rob/Linux-extras/oe/oe-core/build/tmp-eglibc/sysroots/x86_64-linux: No such file or directory Aug 30 16:21:22 what about the other? Aug 30 16:21:39 Ah. I bet it's because I am using an OS X dir mounted through the shared file system under VMWare. Aug 30 16:21:47 oh, well that's bound to be it Aug 30 16:21:52 drat. Aug 30 16:22:01 you should use a real local filesystem Aug 30 16:22:23 I guss I'll have to mount an external USB drive to do this then, don't have enough space in the native Ubuntu virtual machine. Aug 30 16:22:31 Thanks for the help. Aug 30 16:22:39 robtow: no problem Aug 30 16:22:45 :-) Aug 30 16:40:24 ahah, I had the exact same issue as robtow :) Aug 30 16:40:38 (but I was doing it on full OSX) Aug 30 16:42:02 apparently there is some semantic difference in how permissions work between the two filesystems Aug 30 17:01:24 bluelightning - well afaict the problem is not really the permissions or the utime stuff Aug 30 17:01:38 it's because HFS+ is per default case-insensitive Aug 30 17:02:42 (though HFS+ also has an extensive use of xattr) Aug 30 17:12:13 damn Aug 30 17:12:26 my patches were rejected from the list Aug 30 17:12:33 (though I sent them *after* subscribing) Aug 30 17:12:36 with the same mail Aug 30 17:12:52 (though, I did use my mailer and not git to send the mails) Aug 30 17:13:05 and damn^2 Aug 30 17:13:23 I configured a package to use systemd, and created its service file, but it's not enabled per default Aug 30 17:17:16 guyzmo: right, I think not having case-sensitivity might be a problem Aug 30 17:17:57 and another thing is that I need to modify /etc/lighttpd.conf for another package that serves stuff through lighttpd, I would do it using sed in postinstall script of the package... how would I set that in the .bb file ? (or is there a best way to do it ?) Aug 30 17:18:26 for that kind of customisation, I would recommend a bbappend in your own customisation layer on top Aug 30 17:18:42 (unless you are fixing something that's genuinely broken, in which case send a patch :) Aug 30 17:19:18 well, actually this situation may be slightly different Aug 30 17:19:33 so you tell me I should do some lighttpd.bbappend to configure /etc/lighttpd.conf ? Aug 30 17:20:12 it depends Aug 30 17:20:44 if the other package may be optionally installed and definitely depends on lighttpd and no other web server, then it could be OK to do the sed in postinstall as you suggest Aug 30 17:22:52 well for the time being, that's the case. ultimately, if I want to add the support of other webservers, I can always create new specialized packages Aug 30 17:23:09 but how do I tell bitbake what shall go in postinstall ? Aug 30 17:24:04 pkg_postinst_packagename() { } Aug 30 17:24:19 where packagename should be ${PN} if it's the main package Aug 30 17:24:33 git grep for pkg_postinst for examples Aug 30 17:24:48 ok thanks Aug 30 17:25:37 and last thing, how do I force the enabling of my systemd enabled package at install time ? Aug 30 17:27:17 you mean, how do you get it installed into the image? Aug 30 17:29:41 I mean, how do I get it to start at first boot of the image Aug 30 17:30:11 I already defined : Aug 30 17:30:11 SYSTEMD_PACKAGES = "${PN}-systemd" Aug 30 17:30:11 SYSTEMD_SERVICE = "pollux_gateway.service" Aug 30 17:30:14 and it's installed Aug 30 17:31:32 (I can always do that in postinst, but I think that's not the way :-p) Aug 30 17:38:55 oh, right - sorry, know very little about systemd usage, I've been sticking with sysvinit as far as OE is concerned Aug 30 17:39:10 ok :-/ Aug 30 17:39:26 well, I'll do it the dirty way then Aug 30 17:39:32 let's postinst systemctl enable Aug 30 17:41:54 bbl Aug 31 02:10:28 hi. i noticed oe-core doesn't have a native package for libsdl, but instead specifies it as a host dependency. is there any reason against providing a libsdl-native? Aug 31 02:12:27 i have a SLED host, and its SDL relies on libxcb. but oe-core also has libxcb. you could imagine the version incompatibility **** ENDING LOGGING AT Fri Aug 31 02:59:59 2012