**** BEGIN LOGGING AT Fri Jan 18 02:59:59 2013 Jan 18 06:53:37 hi, does anybody know how to force bitbake to regenerate the accidentally deleted build/tmp/deploy/licenses folder and its content ? Jan 18 08:27:20 good morning Jan 18 08:38:10 mckoan: hello, solved your boot issues? Jan 18 08:39:53 ant_work: hi, yes thank you. was udev version too new for kernel 2.6.30 as you suggested ;-) Jan 18 08:40:20 nice Jan 18 09:02:33 hello all! Jan 18 09:05:01 i there any good documentation on bitbake inetrnals ? Jan 18 09:05:52 internal code Jan 18 09:06:23 yes but is hard to undestand what code do what Jan 18 09:08:57 try the development manual at: https://www.yoctoproject.org/documentation/current Jan 18 09:09:07 in a bbclass file i want to add some extra dependency based on what is actually installed on the image IMAGE_INSTALL, so i want to reconstruct the complete dependecy chain and then compute extra packages to install Jan 18 09:09:22 any hint on that ? Jan 18 09:12:00 see the 5.2.3. section in that manual Jan 18 09:12:09 seems to address your needs Jan 18 09:14:52 ok thanks Jan 18 09:18:57 hmmm this is not what i need, i want to rebuild a complete list of packages that will be installed on the image, and based on that list add some extra packages Jan 18 09:29:46 b737800: maby add a layer? Jan 18 09:36:30 could you elaborate please ? Jan 18 09:43:12 morning all Jan 18 09:45:53 b737800: look at the complementary package code in image.bbclass Jan 18 09:46:29 b737800: note that it already handles installing all dbg/dev etc. packages for each package selected to be installed Jan 18 09:49:51 i'v tested the feature using poky danny but this semms to work partially. i'v added a package on IMAGE_INSTALL added IMAGE_FEATURES+="dev-pkgs" and checked if the dev version of the package is isntalled on the rootfs. and it is not installed Jan 18 10:01:23 bluelightning: hi, seems thre have been some cleanings wrt modules auto-loading Jan 18 10:03:11 I'm still puzzled by the interaction btw modutils.sh and systemd, I'll have to test Jan 18 14:01:28 ant_work: I'm refreshing yaffs2 today. that will fix up your build failure. Jan 18 14:03:56 great, thx Jan 18 14:05:52 I'll send you the dubious specified-non-hardware Jan 18 14:06:19 triggered by included fragment Jan 18 14:08:41 (features/mac80211/hostapd.scc, mac80211.scc) Jan 18 14:09:51 zeddii: btw 3.8-rc4 is out Jan 18 14:10:11 release is near Jan 18 14:14:35 yep. I merged -rc4 last night and tested, didn't push to the servers yet. Jan 18 14:14:47 I'll do final yaffs2 testing and then push the two at the same time. Jan 18 14:15:08 zeddii, I am thinking of mixing config fragments with a defconfig Jan 18 14:15:12 should that work? Jan 18 14:15:22 ie, I use the defconfig from the evil vendor tree Jan 18 14:15:36 and use frags to change it inot something that works Jan 18 14:16:05 reading through the kernel tools, I think this might work, but sh tends to make my eyes bleed after a while Jan 18 14:16:20 zeddii: some yaffs2 stuff is also appearing in specified-non-hardware Jan 18 14:16:29 Crofton|work, should definitely work, the approach is fine. if it doesn't it's a bug. Jan 18 14:16:36 greta Jan 18 14:17:04 basically, rather than editing the defconfig, I want it to be clear how to cahnge the config into something usable Jan 18 14:17:14 Crofton|work, re: the code, I'm addressing the horror of that in 1.5 ;) Jan 18 14:17:15 Crofton|work: you can rename it, evil.cfg would also work Jan 18 14:17:17 I think the default is setup for a minimal busybox based rootfs Jan 18 14:18:05 Crofton|work, definitely. and then you'll start generating some config blocks you like, and have an explicit/visual history of what you've changed. Jan 18 14:18:16 yep Jan 18 14:18:27 now we just need common config blocks Jan 18 14:20:33 Crofton|work: ideally we should have MACHINE_FEATURES -> KERNEL_FEATURES translation so you just defin it in the machine.conf and the kernel adapt itself Jan 18 14:21:14 ofc not in depth, just for the most exposed stuff Jan 18 14:21:39 yes, but we also have some baseline stuff Jan 18 14:22:16 ant_work. We've got a few entries in the bugzilla for the translation, and I've got one more to put in for 1.4 for yanking in common blocks. I just have to get to it next week. Jan 18 14:22:17 CONFIG_UEVENT_HELPER_PATH="/sbin/hotplug" needs wiping I think :) Jan 18 14:22:49 zeddii: I'll be happy to test and help to improve Jan 18 14:22:57 I'm all annoyed at this patch also: http://git.xilinx.com/?p=linux-xlnx.git;a=commitdiff;h=9b9b5842791e4a3462e39be5651c51da7c67a512 Jan 18 14:23:22 if that is in, the kernel build fails without giving a useful error message Jan 18 14:23:34 basically, it breaks makefile parsing with make 3.82 Jan 18 14:23:43 Crofton|work, nasty. Jan 18 14:23:48 yeah Jan 18 14:24:06 it would be nice to get a useful message from oe in this scase Jan 18 14:24:11 not just make oldconfig fails Jan 18 14:24:40 ant_work: perfect. I'll ping you when it is tangible. I do actually have some changes done as of yesterday that should allow you to throw out your fragment that unsets options. Jan 18 14:25:03 Crofton|work, funny. I was fighting with oldconfig last night, but for a less exciting reason .. but it was a bit of a pain to dig into :) Jan 18 14:26:09 heh Jan 18 14:26:22 I didn't figure it out until I tried a build outside OE Jan 18 14:26:47 sounds similar to what I ended up doing :) Jan 18 14:27:07 sounds like we need better messages then! Jan 18 14:28:39 I managed to whack a space where I needed tab. wasn't obvious at all. Jan 18 14:47:37 new Testopia HOWTO wiki is up and running! https://wiki.yoctoproject.org/wiki/Testopia_HOWTO Jan 18 14:49:41 hi. is there a way to specify the partition scheme of the syslinux hddimg install target? i scanned the classes/recipes, but could not find sth. appropriate Jan 18 15:07:23 zeddii, also, do you intend to move the multi-dtb.inc file into oe-core? Jan 18 15:11:21 Crofton|work, which part in particular, I'm probably not aware of the gap. Some multi-dtb supported merged over this past summer, right ? Jan 18 15:11:41 not sure, i am reading the meta-zynq layer on yp.org Jan 18 15:11:54 and there is a multi-dt.inc file there Jan 18 15:12:05 for multiple dt's oper machine Jan 18 15:13:18 right. there was some work from the FSL guys on this as well. if I could get sustained comms on that zynq topic, I'd have moved on it more. but I'll make a note here and see if there's a gap. Jan 18 15:17:04 I will need to build multiple dt's to support people using different kernel args depending on boot device Jan 18 15:17:14 which is what multi-dtb was doing Jan 18 15:17:48 * zeddii nods Jan 18 15:29:53 Crofton|work: zeddii: we build multiple dtbs Jan 18 15:29:59 i think everything is in oe-core proper Jan 18 15:30:03 ok Jan 18 15:30:14 I'll look there, before copying the old multi-dtb file Jan 18 15:45:55 How do I see what options the KERNEL_FEATURES variable has? Jan 18 16:33:39 i've got major garbage on the terminal with a do a menuconfig through bitbake (on danny) Jan 18 16:33:40 Doing a build that needs libedit, my coworkers are running into the error below. We already have libedit-20110802-3.0.tar.gz in a common donwloads area (and its readable); How can we tell bitbake to stop going to the web? ERROR: Function failed: Fetcher failure for URL: 'http://www.thrysoee.dk/editline/libedit-20110802-3.0.tar.gz'. Unable to fetch URL from any source. Jan 18 16:33:44 anyone else see something similiar? Jan 18 16:34:38 it occurs w/ screen and xtern Jan 18 16:36:27 JimNH2: is there a .done file for the download as well? Jan 18 16:37:02 yes, there is a .done Jan 18 16:37:55 but I note the .done is from 2am this morning; do .done files get updated periodically? Jan 18 16:38:11 (2am is about the time of a nightly build for us) Jan 18 16:42:39 JimNH2: I don't know. Maybe -D will help to trace it down? Jan 18 16:43:03 ok tu Jan 18 17:23:44 trying to do a build of yocto/meta-fsl-arm core-image-sato (danny/danny branches) the the vivante gpu installs its GL headers which break the build of xserver-xorg, any idea how to solve this in a clean way? Could I provide egl/gles1/gles2/etc but not the -dev packages so that things build properly? Jan 18 17:25:08 markos_: turn off the glx packageconfig in xserver-xorg Jan 18 17:25:36 rburton, that's a clean solution, thanks :) Jan 18 17:25:57 alternatively, if your vivante gpu headers are actually good for glx, change the dep in xserver-xorg to virtual/libgl and submit a patch :) Jan 18 17:26:38 it might be best to make glx a machine-specific override tbh Jan 18 17:27:49 re PACKAGECONFIG: meta/recipes-graphics/xorg-xserver/xserver-xorg.inc: -> PACKAGECONFIG ??= "udev ${@base_contains('DISTRO_FEATURES', 'opengl', 'glx', '', d)}" just remove the glx entry, right? Jan 18 17:28:40 you can override it in your local.conf, PACKAGECONFIG-pn_xserver-xorg = "udev" Jan 18 17:28:44 something like that anyway! Jan 18 17:28:49 the vivante headers redefine many things and miss out others, too much breaks with those at least from what I've tried Jan 18 17:28:49 they esp miss MESA* GL defines Jan 18 17:29:20 even better! Jan 18 17:29:33 (sorry if these are dumb questions, I'm relatively new to yocto) Jan 18 17:30:00 markos_: not dumb at all Jan 18 17:31:03 ok trying with that Jan 18 17:42:44 so anyone know why meta-systemd installs the keymaps service when systemd-vconsole-setup already does loadkeys? Jan 18 17:45:08 rburton, seems to do the trick, thanks :) Jan 18 17:47:23 actually, it was rather PACKAGECONFIG_pn-xserver-xorg (reversed _ & -) but the idea was that mattered :) Jan 18 17:48:19 close :) Jan 18 17:48:28 i had a feeling i might have got those bits wrong Jan 18 18:50:32 fray: there ? Jan 18 18:50:50 fray: smart or rpm is puking on me while doing meta-toolchain Computing transaction...error: Can't install libc6-2.17-r0.0@x86_64: no package provides /bin/sh Jan 18 18:51:06 Do you have an idea Jan 18 18:51:17 poky-contrib mhatle/au-fix Jan 18 18:51:21 top two patches Jan 18 18:51:52 'er.. sorry.. "ua-dep" Jan 18 18:52:28 ok thanks Jan 18 18:52:31 let me try Jan 18 18:52:36 this was not seen during testing because all of the SDK testing occured with the populate_sdk task, not meta-toolchain* Jan 18 18:52:45 hmm Jan 18 18:52:48 ok Jan 18 18:53:02 poky-contrib is so big Jan 18 18:53:06 the underlying "problem" is that RPM enforces dependencies that deb and ipk don't.. Jan 18 18:53:12 you can fetch just the one branch Jan 18 18:53:13 can we delete the dead trees Jan 18 18:53:34 thats what I usually do.. fetch the one branch down.. and then it'll only pull the two commits Jan 18 18:53:49 fray: yeah that explains why I dont see it with OE-Core standalone Jan 18 18:54:08 ya, the issue is specific to meta-toolchain processing Jan 18 18:54:43 the software in the packages requires /bin/sh, /usr/bin/env and a few other things.. but there is no DEPENDS that enforces the build.. OE is relying on them being built as part of the normal system process (in an image) Jan 18 18:54:46 how do you pull just one branch I wonder Jan 18 18:54:54 it works fine for an image, but meta-toolchain is so much smaller of a set.. Jan 18 18:54:55 you still have to fetch the whole git metadata Jan 18 18:55:02 git fetch : Jan 18 18:55:12 (verify that.. thats from memory) Jan 18 18:56:16 ya.. I'm pretty sure thats right Jan 18 18:56:38 that just changes what you checkout Jan 18 18:56:44 clone is still fat Jan 18 18:56:44 I finished this right before I left yesterday.. Jan 18 18:57:04 git fetch pulls from a remote, and it'll only pull those items not the whole tree like a clone or git remote update Jan 18 18:57:21 and if you already have an oe-core repository.. then it'll only pull the top two commits, as you alreayd have all of the rest Jan 18 18:58:28 if I can get verification from someone else it's working right.. I'll submit it upstream Jan 18 19:01:30 I am testing it right now Jan 18 19:01:43 if it works for me. Add my Tested-by Jan 18 19:02:49 ok Jan 18 19:03:08 (I still need to add the YP bug reference numbers to the commit msg as well.. but I think the patches are solid.. Jan 18 19:03:31 the top commit we'll see what RP thinks.. it resolves the dependency issue, but I'm not sure he'll think it's reasonable.. Jan 18 19:03:38 (that only fails in meta-toolchain-gmae BTW) Jan 18 19:05:45 I am trying both applied Jan 18 19:19:23 guys, does anyone here knows how to configure smartrpm ? Jan 18 19:20:22 fray: my SDK is generated successfully so Add my Tested-by if that matters Jan 18 19:25:01 msm: any update on the oprofile/libpfm patches? Jan 18 19:29:20 sgw1: Jan 18 19:29:25 sgw1: hi ;) Jan 18 19:29:57 sgw1: i'm testing these Jan 18 19:29:59 http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/log/?h=mattsm/master-next Jan 18 19:30:07 but i don't know if the libpfm stuff is really needed right now Jan 18 19:30:19 msm: great, I will pull to alocal build and test also Jan 18 19:30:23 the oprofile patch alone fixes the build Jan 18 19:30:26 you just need that one Jan 18 19:30:35 http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=mattsm/master-next&id=3b3a21ad76ee911a2a8f4a729e25a4666ca9ea09 Jan 18 19:30:59 ftonello smartpm you mean? Jan 18 19:31:11 khem thanks for verifying.. I'll try to get the work done and submitted today Jan 18 19:31:22 sgw1: i will warn you i've been unable to actually test oprofile Jan 18 19:35:02 msm: Ok thanks, starting a qemuppc build now and heading out for a swim, back later Jan 18 21:19:59 Is Yocto going to have an event in the ELC in San Francisco? Jan 18 21:24:06 khem: your kernel CC / modules patch needs to be rebased again :/ Jan 18 22:01:52 I've been googling around and not finding any good instructions for integrating the meta-systemd layer from meta-oe into an existing build. I just got my build updated to 1.3 and I'm trying to get systemd integrated. Jan 18 22:02:07 anybody know how this is supposed to work? Jan 18 22:03:06 i'd say look at angstrom Jan 18 22:03:09 it's a systemd distro Jan 18 22:03:16 so presumably they already integrate it Jan 18 22:03:50 I was hoping to just add meta-oe/meta-systemd as a layer in my bblayers.conf, but it looks like it's dragging in practically everything from meta-oe. Jan 18 22:04:37 i'm not at all surprised that meta-systemd would require meta-oe Jan 18 22:05:09 quick question, what do I need to tweak to get my Yocto sugarbay BSP to build x86 instead of x64? Jan 18 22:05:12 yes and README says so Jan 18 22:05:18 it is requiring meta-oe/meta-oe, but then I get errors for missing bb files for bbappend's included in meta-systemd that I'm not using. Jan 18 22:05:33 mebrown: like what? Jan 18 22:06:09 http://pastebin.com/g56573KC Jan 18 22:06:47 do I need to BBMASK that? (just browsing the README) I haven't used BBMASK before... Jan 18 22:07:01 are you using master? Jan 18 22:07:09 danny Jan 18 22:07:17 then you need to use BBMASK Jan 18 22:07:59 ok, just finished reading docs on BBMASK. i will try it. Jan 18 22:25:55 msm: your patch built ok in my env, trying to do a quick qemu test, but unsure of the options to pass Jan 18 22:26:11 fray: meta-toolchain built OK this time with your patch Jan 18 22:27:29 if meta-toolchain-gmae is where the ua failed.. Jan 18 22:34:58 the BBMASK doc in the ref guide could really use some elaboration. Took a couple google searches to get the syntax right. Jan 18 22:36:48 fray, seem to remember that Jan 18 22:38:31 it's just a regular expression Jan 18 22:38:36 that's all Jan 18 22:38:47 matched against the recipe full absolute path Jan 18 22:41:33 kergoth, when you have several paths to exclude, the syntax is non-obvious Jan 18 22:44:09 mebrown: again, it's just a regular expression Jan 18 22:44:12 google for how regexes work Jan 18 22:44:28 the syntax is standard as is used in regular expressions in any software Jan 18 22:44:50 I understand, but interactions between +=, etc with the bitbake vars is subtle. += adds a space, I think. Jan 18 22:44:56 you have to use .= Jan 18 22:45:07 and "|..." Jan 18 22:45:44 not saying it's difficult, just not obvious to a neophyte Jan 18 22:46:09 http://www.mail-archive.com/yocto@yoctoproject.org/msg10573.html Jan 18 22:46:12 was very helpful Jan 18 22:47:03 += works the same way everywhere in the metadata, there's nothing new there either Jan 18 22:47:17 i appreciate that it needs to be better documented, but it's still not complex Jan 18 22:47:46 you just said exactly the same thing as I just did, just in a different way. :) Jan 18 22:48:28 you feel it's non-obvious, i don't. regexes work this way, the metadata works this way. combining them is trivial Jan 18 22:48:55 and seeing the results of your actions is quite easy with bitbake -e Jan 18 22:49:10 and bitbake -DDD, of course, depending on what exactly you want to see Jan 18 22:49:38 msm: oprofile worked with your change, so please post when you are ready with it, I would like to get it in a MUT build over the weekend for M3, thanks Jan 18 22:49:53 * kergoth wanders off Jan 18 22:51:39 fray: did you see bug 3753 regarding rpm2cpio? Jan 18 22:51:39 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=3753 normal, Undecided, ---, bogdan.a.marinescu, NEW , rpm2cpio fails to unpack emgd 1.16 binary rpms Jan 18 22:52:31 if it's what I'm thinking of it's likely a new version of an old bug Jan 18 22:53:04 Hmm.. nope, that's a new and interestign one.. Jan 18 22:53:28 my guess is that whoeveer build the rpm did it in a Fedora specific manner.. so xz or some other alternative archive format was used.. Jan 18 22:55:17 My kernel recipe has a patch, which adds a custom kernel module. When baked, the module is built, but doesn't end up in the modules__.tgz file in the deploy directory. Jan 18 22:55:23 added comments Jan 18 22:55:29 Any pointers? Jan 18 22:56:38 chrisw957: do_deploy in kernel.bbclass Jan 18 23:01:18 sgw - what is the bugzilla entry for those SDK meta-toolchain bugs/ Jan 18 23:03:37 found it Jan 18 23:08:19 fray found the rpm2cpio binary works on F18 Jan 18 23:09:27 fray: found the script and it seems to work also, will create a patch! Jan 18 23:10:14 otavio, there is a Developer Day scheduled Jan 18 23:10:20 and a Bof has been accepted Jan 18 23:10:30 they likely added "yet another arbitrary compression type(tm)"... Jan 18 23:10:33 Jefro should have better details Jan 18 23:14:00 fray: looks like lzma! Jan 18 23:14:03 sgw1 -- the meta-toolchain fix has been sent.. we'll see what RP says about the ua update stuff.. Jan 18 23:14:21 in the past there has been no magic to identify an lzma archive.. maybe that finally changeD? Jan 18 23:20:23 strange it's actually xz which we support, I wonder if it's a 'file' issue Jan 18 23:20:58 Crofton|work: :) Jan 18 23:22:26 it's not file.. lzma archives historically have had no magic identifiers.. so thereis no way to know you are using them Jan 18 23:22:56 fray: I just checked the header and it is xz not lzma, but lzma was added in the rpm4 version Jan 18 23:23:14 ok.. Jan 19 00:38:33 fray: bumped into ppc64 sdk issue now Jan 19 00:38:38 fray: error: Can't install libstdc++6-4.7.2-r14.0@ppc64e5500: no package provides libgcc_s.so.1(GCC_3.4.4)(64bit) Jan 19 00:39:45 should come from the libgcc package.. Jan 19 00:40:05 figure out what provides the file and see what the package provides are... or grep through the pkgdata, looking for file provides Jan 19 00:44:27 yes this sdk contains libgcc for 32bit as well as 64bit ppc Jan 19 00:44:45 ../build/tmp-eglibc/deploy/rpm/lib32_ppce5500/libgcc1-4.7.2-r14.0.lib32_ppce5500.rpm Jan 19 00:44:57 ../build/tmp-eglibc/deploy/rpm/ppc64e5500/libgcc1-4.7.2-r14.0.ppc64e5500.rpm Jan 19 00:45:40 I wonder if its ignoring the lib32_ppce5500 and ppc64e5500 part Jan 19 00:50:57 fray: http://pastebin.com/AHBb08yv Jan 19 00:51:15 fray: those are the provides of libgcc1-4.7.2-r14.0.ppc64e5500.rpm Jan 19 00:51:55 I wonder why it could not figure the rpm out from its deploy area Jan 19 01:25:50 khem -- back Jan 19 01:26:21 it all looks like it's there.. Jan 19 01:30:11 sgw1: im unsure too Jan 19 01:30:21 never used oprofile so far **** ENDING LOGGING AT Sat Jan 19 02:59:59 2013