**** BEGIN LOGGING AT Wed Aug 07 02:59:58 2013 Aug 07 04:39:41 hii Aug 07 08:01:05 morning all Aug 07 08:06:43 morning all, hi bluelightning Aug 07 08:07:00 morning silvio_ Aug 07 08:07:14 hi, i copied a kernel-recipe, changed the name and replaced the git source with my own local source. building it with "bitbake linux-faarm" builds the kernel. now when i try to build my image (it's for a colibri board) i get "NOTE: multiple providers are available for virtual/kernel", also no uImage gets built. what am i doing wrong? Aug 07 08:07:21 hi bluelightning Aug 07 08:07:27 hi woglinde Aug 07 08:07:31 also, good morning @ all ;) Aug 07 08:07:38 hi woglinde Aug 07 08:08:04 thesignal: hi - basically you need to set PREFERRED_PROVIDER_virtual/kernel, probably in your machine configuration Aug 07 08:09:16 bluelightning: the configuration should be in something like /build/conf/machine, right? Aug 07 08:09:59 thesignal: usually meta-(something)/conf/machine/machinename.conf Aug 07 08:12:05 i got this already. what i also recognized, my kernel-module* ipk packages get deployed in a subdirectory called "morgue" Aug 07 08:13:52 and because of that, when i build my image, i get an error, cause the package "kernel-modules" can't be found, copying the packages with cp ../morgue/* . fixes this issue Aug 07 08:13:59 but this seems a bit strange to me Aug 07 08:14:09 er... that's not normal Aug 07 08:15:11 I wonder how that is happening Aug 07 08:16:52 thesignal: ok, got it.. the morgue directory is created for old packages that have been displaced by newer versions Aug 07 08:18:57 bluelightning: this makes sense. and how do i fix this? just replace the original version with a higher one? Aug 07 08:19:21 thesignal: I'm still not sure why it would be happening if there were no newer version Aug 07 08:21:08 thesignal: though it may have something to do with two kernels being built at the same time Aug 07 08:21:31 thesignal: you should set PREFERRED_PROVIDER_virtual/kernel to stop that from happening Aug 07 08:22:33 bluelightning: it is set, at least in /meta-my/conf/machine/faarm-t20.conf Aug 07 08:23:00 bluelightning: or is this also needed somewhere else? Aug 07 08:23:03 thesignal: are you still getting the multiple provider warning? Aug 07 08:25:29 bluelightning: just a moment, i trying to reproduce it again Aug 07 08:30:22 bluelightning: okay, i added a "COMPATIBLE_MACHINE" to my recipe yesterday, i didn't get "NOTE" message any more. Aug 07 08:30:56 thesignal: ah ok right, I'm assuming the default COMPATIBLE_MACHINE did not include your machine then Aug 07 08:32:04 bluelightning: but i have another problem now. i ran bitbake -c clean linux-faarm (my kernel recipe) and the package was built, without building the kernel Aug 07 08:32:51 bluelightning: and in the ipk/ directory, there is something with a different git sha1 sum Aug 07 08:37:25 thesignal: er, I don't quite understand - did you do something else after -c clean? Aug 07 08:39:04 bluelightning: no nothing, just -c clean the image and the kernel, i don't even know where it gets that packages from (kernel-modules) Aug 07 08:39:56 thesignal: maybe those are the other kernel you built previously - if you know what that is you should be able to do -c clean on that as well Aug 07 08:55:20 bluelightning: i found my mistake. there were backupfiles from every recipe with a "~" at the end. with -v -c clean i found out. Aug 07 08:55:27 bluelightning: thanks for your help Aug 07 08:55:35 thesignal: ok, np Aug 07 09:16:22 can someone tell me what the "?=" operator does/means? Aug 07 09:17:39 thesignal: http://docs.openembedded.org/bitbake/html/ch02.html#id398548 Aug 07 09:22:04 Zagor: thx Aug 07 09:47:24 what expands the TMPDIR from tmp to tmp-eglibc ? Aug 07 10:10:50 suihkulokki: see TCLIBCAPPEND in meta/conf/distro/defaultsetup.conf Aug 07 10:14:19 hm so finally I wrote khem an email about merge_config.sh for uclibc Aug 07 13:53:53 JaMa: it does kind of look like libav and libpostproc are interdependent Aug 07 15:49:55 is there a collection of recipes somewhere? i'm looking for a recent octave recipe Aug 07 15:50:22 thesignal: http://layers.openembedded.org Aug 07 15:50:41 thanks Zagor Aug 07 15:58:47 thesignal: I think you'll probably need to migrate the one from OE-Core Aug 07 15:58:52 er I mean OE-Classic Aug 07 16:00:16 * kergoth yawns Aug 07 16:00:52 thesignal: http://cgit.openembedded.org/openembedded/tree/recipes/octave Aug 07 16:01:00 thesignal: http://www.openembedded.org/wiki/Migrating_metadata_to_OE-Core Aug 07 17:31:45 moin Aug 07 17:39:03 hi mr_science Aug 07 18:17:38 so if kergoth (or someone) can point me at the right stuff i can cleanup the nginx-1.4.0 recipe... Aug 07 18:18:11 i'm thinking there has to be a better way to set file/dir permissions Aug 07 18:19:01 but i found exactly zero usage examples for the fs_perms.txt thing... Aug 07 18:34:15 mr_science: honestly i'm not sure on perms, i've just chmod'd in do_install where necessary. it's under pseudo, so that works Aug 07 18:40:06 that's what i'm doing now... Aug 07 18:40:18 just seems like i'm missing something Aug 07 18:40:38 kergoth: have you looked at package.bbclass lately? Aug 07 18:42:22 check out fixup_perms in meta/classes/package.bbclass Aug 07 18:42:46 seems like that should be The Right Way Aug 07 18:50:58 I always kind of liked the idea of using some sort of install manifest. generate it when you create the recipe, taking into account where the files came from, and drop do_install Aug 07 18:51:01 heh Aug 07 18:51:03 i still need some time to dif a little deeper and try it out Aug 07 18:51:06 but no, i haven't looked at that, it sounds interesting though Aug 07 18:51:08 * kergoth nods Aug 07 18:51:19 *dig even Aug 07 19:18:16 okay, so i just added a really basic vixie-cron and the logrotate package to our classic ro-rootfs build Aug 07 19:18:32 so cron is now barking at me Aug 07 19:19:31 /USR/SBIN/CRON: can't open log file log: Read-only file system /USR/SBIN/CRON: can't write to log file and so on Aug 07 19:20:27 the funny part is a) /var/log is on a writable partition, and b) it doesn't actually say what file/path it can't write to Aug 07 19:22:18 anoybody know off the top of their head what "log file" cron is trying to open? Aug 07 21:00:18 it was the log directory perms Aug 07 21:00:39 * mr_science is still not sure why root's crontab has this problem **** ENDING LOGGING AT Thu Aug 08 02:59:59 2013