**** BEGIN LOGGING AT Thu Jan 19 02:59:57 2012 Jan 19 07:56:29 good morning Jan 19 10:07:14 morning folks Jan 19 10:07:22 mornin bluelightning Jan 19 10:08:04 hi Jan 19 11:03:35 hi mckoan Jan 19 12:53:28 hello Jan 19 12:53:52 i am trying to get atk bitbaked but fail horrible with this error "ExpansionError: Failure expanding variable SUMMARY, Jan 19 12:54:38 expression was ${SUMMARY}" , Jan 19 12:54:53 which to me translates that there is something wrong with a DESCRIPTION = "An accessibility toolkit for GNOME." Jan 19 12:55:31 because there is no SUMMARY Variable.... but it aint, for that is perfectly fine :( Jan 19 13:18:27 here is the pastebin if someone wants to see a "screenshot" of such a rarely ocuring incident. :) http://pastebin.com/0FqecMeC Jan 19 13:30:58 ipaqlor: never noticed such failure for missing $SUMMARY Jan 19 13:38:06 ant_work: yeah, me neither, and google not at all :) Jan 19 13:39:21 also, doing bitbake -DDD -v...and what not, does not realls show more. Jan 19 13:39:54 I had just read this thread http://lists.linuxtogo.org/pipermail/openembedded-core/2012-January/016558.html Jan 19 13:43:04 ant_work: well, no idea. i allready tried using SUMMARY="wtf", DESCRIPTION="nop"... pasting the atk.inc into the bb. well... alls the same :) Jan 19 13:43:09 btw why do you build with -b? Jan 19 13:43:36 ant_work: because besides that task and its dependencies everything has been allready built Jan 19 13:44:02 ant_work: the eroor is independet on how the atk recipe is triggered Jan 19 13:45:25 ant_work: or clearer. i can produce the error with every build target available that inherits the atk package (gtk stuff) Jan 19 13:48:23 is there some bash -x equivalent to bitbake/python ?.... something that lets you look while VARIABLES get handled (or mishandled) Jan 19 13:48:37 Hiya, is anyone working on recipes for LightDM? Started but having a bit of fun getting it all working so thought it was worth checking that another layer does not have a working recipe ;). Jan 19 13:49:13 have you inspected the files in $WORKDIR? The run_do_package_ipk has porobably the solution Jan 19 13:50:41 ant_work: sorry, cant follow you. the what in the what? :) Jan 19 13:50:42 see line #79 of your pastebin Jan 19 13:51:04 hehe, ok Jan 19 13:51:37 ant_work: yes thos elogfiles, are for people without scrollbackbuffers :) nothing more in there then a snipplet of the full stdout dump i put on pastebin Jan 19 13:51:57 check the run. file Jan 19 13:52:02 is lower layer priority higher or lower? Jan 19 13:52:58 Crofton|work: happily it seems we only have *one* atk recipe Jan 19 13:53:10 heh Jan 19 13:53:22 but there is a .bbappend ;) Jan 19 13:53:27 in meta-oe Jan 19 13:53:42 well, damn, i should clean up Jan 19 13:53:43 mom Jan 19 13:54:13 a few hundred do_bla...files in atk$WORK :) Jan 19 13:55:36 ant_work: ok, now i can see again. what do.bla file do you suggest? Jan 19 13:56:02 well, the corresponding task Jan 19 13:56:25 is embedded in the name Jan 19 13:58:14 ant_work: nevermind, i now have run.do_package_ipk.24358, and run.do_package_write_ipk.24358, and run.read_subpackage_metadata.24358, and run.sstate_task_prefunc.24358 ...greping vor priority on them didnt give back any hit. Jan 19 14:00:33 ipaqlor: grep for SUMMARY Jan 19 14:01:38 summary = localdata.getVar('SUMMARY', True) or localdata.getVar('DESCRIPTION', True) or "." Jan 19 14:01:47 :) Jan 19 14:03:47 ant_work: i got both ;) so that shoukd be fine :) except when /home/gl/IPAQ/angstrom/setup-scripts/sources/bitbake/lib/bb/data_smart.py hits me :) Jan 19 14:06:15 hello GNUtoo seems like you are the other human on earth who got :: ExpansionError: Failure expanding variable SUMMARY, expression was ${SUMMARY}" . let me greet you :) Jan 19 14:06:15 ...REGARDING TO GOOGLENETS) Jan 19 14:06:15 sorry caps Jan 19 14:07:49 GNUtoo: care to share your secret? i assume you managed to get over it or you for shure wouldnt be hanging around here for 2 months waiting for a solution ;) Jan 19 14:08:22 ipaqlor: psss.. I tell you a secret now. Don't mix that many layers for an ipaq ! Jan 19 14:08:58 ant_work: woow, thanks. Jan 19 14:09:21 ant_work: so its a feature :) Jan 19 14:09:35 to remind one of a much to layerd livestyle Jan 19 14:11:11 this layerstuff Jan 19 14:11:22 isnt that what its all about? Jan 19 14:11:42 i thout i understand that stuff some 5 years agfo. but i must have missed the whole pint. Jan 19 14:11:54 I, for me, don't use any script to manage the layers Jan 19 14:12:17 I know you are following som einstructions Jan 19 14:12:55 ant_work: i am a good robot Jan 19 14:13:05 anyway, I suppose the images I'm building lately do not contain atk... Jan 19 14:13:37 I'll try later on and tell you. I just use oe-core + meta-handheld + meta-openembedded (reduced) Jan 19 14:14:13 betetr said, only some recipes of meta-openembedded Jan 19 14:14:17 ant_work: lucky you :) but seriously thats no solution for me. Jan 19 14:14:49 have you built a normal console-image? Jan 19 14:15:09 ant_work: whats a normal console-image? Jan 19 14:15:36 which image are you building? Jan 19 14:15:41 console-image is shell only Jan 19 14:16:10 MACHINE=h3600 bitbake -ak console-image Jan 19 14:16:52 bluelightning, koen is thinking I can get qmake on the target with the qt-tools package? Jan 19 14:16:56 ipaqlor, hi Jan 19 14:17:02 hi GNUtoo :) Jan 19 14:17:22 ipaqlor, can you describe more your problem , I cleansstated some stuff rebuilt and it worked Jan 19 14:17:35 some stuff beeing the recipe that caused that Jan 19 14:17:48 ant_work: well, mei console-image somehow depends on atk Jan 19 14:18:00 Crofton|work: the one in OE-Core does not include it; I believe it gets explicit deleted since it's for the host arch Jan 19 14:18:01 ipaqlor, where's the pastebin where I pastebined? Jan 19 14:18:12 Crofton|work: although when I fix it that's the package it will end up in Jan 19 14:18:28 GNUtoo: its a irclog archive i found via google http://ibot.rikers.org/%23openmoko-cdevel/20111030.html.gz Jan 19 14:18:38 ipaqlor: you are working on h3600? Jan 19 14:18:55 GNUtoo: here is my paste http://pastebin.com/0FqecMeC Jan 19 14:19:17 that is what i thoguht to :) Jan 19 14:19:27 do you have a timeline for fixing it? Jan 19 14:19:46 ipaqlor, ahhh it was loong ago Jan 19 14:19:51 bluelightning: working, well, id like to. but it seems like it is a dead target Jan 19 14:20:08 I tought you were talking about the problem I had yesterday Jan 19 14:20:22 ipaqlor, bitbake -e will help you debug it Jan 19 14:20:29 ipaqlor: a few people have talked about fixing that machine, some even made reasonable progress but somehow it never gets finished :/ Jan 19 14:20:54 and all those buildsystems, which most arent, but just "layers" done make me less confused... hh..easy..oe, poky, angstroem, yocto......all put together in a pot and steered :) Jan 19 14:20:56 ipaqlor: I know I would really appreciate it if it did get fixed... if you need any help (besides kernel coding) I'd be happy to assist Jan 19 14:21:32 GNUtoo: ill try that Jan 19 14:21:53 Crofton|work: I have quite a few things on my plate atm but perhaps I can schedule some time next week to look into it Jan 19 14:22:08 ok Jan 19 14:22:12 Crofton|work: surely in the mean time the qmake2 recipe can be beaten into shape? it did work not that long ago after all Jan 19 14:22:24 I am still trying to get something built that is close to what I made from oe-clssic Jan 19 14:22:45 bluelightning: well i havent even an idea where to start :) hh.org is kind of dead but some git clones seem to exist. angstrom, poky or oe "native" no idea.... Jan 19 14:22:48 so that timeframe works for me Jan 19 14:23:38 this will be a great thing btw, much less confusion Jan 19 14:23:57 it like, once, there was hh, and all was easy. and now every ebookvendor or mp3playerdealer seems to make his layers..uon layers :) Jan 19 14:24:20 i really need to read about thjese layers. they make my head crack Jan 19 14:24:41 http://www.openembedded.org/wiki/LayerIndex Jan 19 14:25:07 http://www.openembedded.org/wiki/OpenEmbedded-Core Jan 19 14:25:17 my favorite, is irex :) ion on poky, on oe :) Jan 19 14:25:26 ipaqlor: a copy of the old hh.org wiki can be accessed via archive.org FYI Jan 19 14:25:38 ipaqlor: also, this was the last attempt: http://elinux.org/Flameman/ipaq Jan 19 14:25:48 bluelightning: its not that googd to browse Jan 19 14:26:14 ipaqlor: no, but it's pretty much all there is Jan 19 14:27:38 oh...flameman Jan 19 14:27:39 bluelightning: git clone... http://pilppa.com/gitweb/?p=familiar-h63xx-build.git , website+sources until 0.8.4 : http://ftp.gwdg.de/linux/handhelds Jan 19 14:28:35 bluelightning: someone hasnt paid their dns fees it seems Jan 19 14:28:44 bluelightning: no elinux.org over here Jan 19 14:29:09 hh.org will be back on Dec 1'st! Jan 19 14:29:12 woah, just found this: http://mad.is-a-geek.org/h3600/index.html Jan 19 14:29:22 Crofton: hehe l) Jan 19 14:29:25 Crofton|work: they just keep moving the date ;) Jan 19 14:29:30 I know :) Jan 19 14:29:30 whoever it is anyway Jan 19 14:31:02 bluelightning: sorry, my fault :) *lookin at elinux* Jan 19 14:32:23 http://web.archive.org/web/20071003095830/http://www.handhelds.org/moin/moin.cgi/HandheldsDevices_2fh3600 Jan 19 14:33:12 gah, the number of hours I spent working on that wiki, and now we can only look at it via archive.org :.( Jan 19 14:34:18 bluelightning: hey it will be active again from December 2011 ;-) Jan 19 14:34:27 bluelightning: i stopped playing around with my ipaq in 2003. seems there has been development going on until 2k6. and now its laying down to die for 5 years Jan 19 14:34:49 florian: yeah, I'm heading out to my delorean now ;) Jan 19 14:35:00 *g* Jan 19 14:35:25 ipaqlor: well, there are still a few people out there who care :) Jan 19 14:35:27 hi, all. How can I switch on cursor in X ? Jan 19 14:35:49 really bad, found asilverslider+512mb kingston flashcard on the secondhandmarket for 5 euro, now flashing makes really fun :) (before i only had a serial cradle) ... Jan 19 14:36:27 ipaqlor: yeah, you can pick up old ipaqs for peanuts now Jan 19 14:37:36 bluelightning: shure. with 16 or 32mb of ram, no android target whatsoever Jan 19 14:38:14 ipaqlor: yeah, you have to have reasonable expectations though... there are devices from just a few years ago that can't handle the current version of android Jan 19 14:38:57 bluelightning: 2.6.29 sounds better than 2.6.21 wrt udev & compatibility Jan 19 14:39:00 bluelightning: hehe, i have any expections about android :) it kids toys, and bloated in my opionion :) Jan 19 14:39:16 no idea how both will cope with Angstrom requirements, though Jan 19 14:39:28 ant_work: yes, it does look very promising Jan 19 14:39:39 ipaqlor: have you actually built and booted an image? Jan 19 14:39:49 ipaqlor: I guess I just find it too hard to let go of old tech :) Jan 19 14:40:21 those 16-32MB devices are good candidates for an opie revamp Jan 19 14:40:22 ant_work: no, it fails. but i got a kernel, libc and some stuff compiled allready :) Jan 19 14:40:47 ant_work: i totally miised the e17@familiar, id like to try that too :) Jan 19 14:41:04 I see. I fear you'll be disappointed :/ Jan 19 14:41:27 ant_work: well, i dont :) Jan 19 14:42:09 ther have been so many new things since hh. like kexec boot, how great is that? that would be great having on the h3600 Jan 19 14:42:11 I mean, once sorted out this atk issue, you'll end up with a system probably booting but failing early in userspace Jan 19 14:42:37 (I hope to be wrong) Jan 19 14:42:38 ant_work: i have time :) Jan 19 14:43:07 ant_work: and also started to locally mirror some hh stuff before the allmighty internets totally forgets about hh Jan 19 14:43:19 heh Jan 19 14:44:14 ant_work: a big thing, i know now. before the hh incident i wouldnt ever have thought that a project like that. more or less the roots of modern androidphonees simply gets forgotten and vanishes Jan 19 14:44:39 within 5 years! :) Jan 19 14:44:57 i allways thought the internets memory is limitless :) Jan 19 15:01:44 mhm, ok, so for a clean start. i should only take oe core and put a h3600 layer on top and fill that with recipies ? Jan 19 15:02:12 the h3600 is in meta-handheld layer Jan 19 15:02:20 pls use that Jan 19 15:02:49 ant_work: ok, i still dont get it. i just read the small OpenEmbedded-Core entry Jan 19 15:03:59 i jst use angstrom with target5 h3600 and that "too much" layers then. so...mhm, in my opionon i should start by removing all layer except core. so that this doesnt happen again in the near future. Jan 19 15:06:05 ipaqlor: you surely need oe-core and a BSP layer on top of it (meta-handheld) Jan 19 15:06:24 still confused, maybe reading the oe pdf book helps to clean things up in my brain. but honestly, it seems oe by itselof i a bit messy :) Jan 19 15:06:35 then you can extend with more layers Jan 19 15:07:05 we are trying to clean out meta-handheld and make it another 'official' layer Jan 19 15:07:29 ant_work: is there o way to look at all this dependencies and layers and override mayhem that is acceptable to human eyes and brains? Jan 19 15:08:23 ant_work: aso you said, i have to many layers, and the whole house will be put a part in the near futurew anyhows, so why build on a system that is allready said to die? Jan 19 15:08:40 heh, it's too bad opkg-utils cvs seems to be gone forever due to hh (among others). it really bugs me when project history is lost. there are a number of oss projects whose old releases just don't seem to exist anywhere online :| Jan 19 15:08:59 if only the core oe is the thing thats stable, then i shall not use the other layers but buiold em myself, aint i? Jan 19 15:09:30 ipaqlor: in oe-core there is support for qemu only Jan 19 15:09:57 in short, you do need a BSP layer Jan 19 15:10:03 ant_work: yes, that where the h3600 recipes with sa1100 and co come into play Jan 19 15:10:35 also, whats bsp :? Jan 19 15:10:38 (ant thinks at sa1100 patches reverted by Russel today ) Jan 19 15:10:48 bigg mess ! Jan 19 15:11:27 ipaqlor: does meta-handheld not support your device? Jan 19 15:11:28 http://www.openembedded.org/wiki/LayerIndex Jan 19 15:11:34 Board Support Package Jan 19 15:11:45 http://cgit.openembedded.org/meta-handheld/tree/conf/machine Jan 19 15:11:56 3 letter acronyms do suck more and more, every day, with every new meaqning :) Jan 19 15:12:07 bsp isn't new, it's very old, starting to lose meaning, actually Jan 19 15:12:20 but the point is valid :) Jan 19 15:12:39 kergoth: http://www.acronymfinder.com/BSP.html Jan 19 15:13:26 ipaqlor: nonissue. determine meaning based on context. this is an embedded channel, therefore bsp == board support package. this will always be the case :) Jan 19 15:14:36 hehe :) true, but BSP is only mentioned twice in the oe wiki :) Jan 19 15:14:37 hi ant and kergoth Jan 19 15:14:43 so its not the usual lingo i assume :) Jan 19 15:15:04 FYI, meta-handheld supports h3600 in spirit Jan 19 15:15:16 but actual working support is another story entirely Jan 19 15:15:18 spirit is good Jan 19 15:15:37 but it has a definitive place to live, is the key. contribute to that rather than duplicating hte effort Jan 19 15:15:46 I have not had as much time as I would have liked to resurrect the older ipaq machines Jan 19 15:15:57 kergoth: yes indeed Jan 19 15:16:01 I welcome all patches :) Jan 19 15:16:36 :) Jan 19 15:17:25 well, from my perspective, the build works just fine and i should build my gpw and opie images in the nearer future. only the buildsystem does not allow me too :) Jan 19 15:18:43 i havent even come to a point yet where i could say h3600 is broken :) Jan 19 15:19:14 but bitbake is :) Jan 19 15:19:29 all is broken Jan 19 15:19:33 in some cases Jan 19 15:20:23 mhm, but bitbake -e is nice Jan 19 15:21:36 bitbake -g too Jan 19 15:21:55 didnt yet took a look at the graphical stuff for it Jan 19 15:23:54 I've not been able to build gpe Jan 19 15:24:03 haven't tried opie Jan 19 15:24:30 woglinde: -g doenst change anything obviuos to me. Jan 19 15:24:48 -g prints the dep graphs for all tasks Jan 19 15:24:50 muriani: could you conquer atk? Jan 19 15:25:09 woglinde: ok, havent looked to close. where should they be found? Jan 19 15:26:11 ls *.dot Jan 19 15:26:35 bluelightning is maintainer for opie Jan 19 15:27:43 i have bitbake 1.15.0 over here. nothing "graphical" happens with -g or -ge Jan 19 15:27:52 I have no cursor in X :-( Jan 19 15:28:27 I said I didnt looked at the graphic stuff for displaying the dep graphs Jan 19 15:28:34 I didnt say bitbake has it build in Jan 19 15:28:58 reisei: dont know if its for much hepl, but have you tried swcurser? Jan 19 15:29:18 wogline: ic Jan 19 15:30:34 reisei: using matchbox? Jan 19 15:30:44 ipaqlor: Yes, i've tried swcursor "true" in X in section Device and it don't work Jan 19 15:30:58 florian: what's the matchbox? Jan 19 15:31:14 reisei you mabyee called Xorg with -nocursor? Jan 19 15:31:16 reisei: a window manager Jan 19 15:31:18 check ps Jan 19 15:31:25 or check it with ps Jan 19 15:31:30 hi florian Jan 19 15:31:37 hi woglinde_ Jan 19 15:31:40 florian: no, I don't use it Jan 19 15:31:54 ok Jan 19 15:32:09 then you might have a transparent cursor theme installed Jan 19 15:32:12 woglinde_: in xinit there is no cursor, how can I check the arguments of Xorg? Jan 19 15:32:33 florian: how can I check it? Jan 19 15:33:36 reisei ps Jan 19 15:33:42 look vor Xorg Jan 19 15:33:44 or X Jan 19 15:33:56 opkg help Jan 19 15:34:20 you can find out the option your self for getting installed packages Jan 19 15:34:27 look for transparent Jan 19 15:36:02 or -nocursor Jan 19 15:36:31 * JaMa should read backlog firxt Jan 19 15:37:12 jama no problem Jan 19 15:37:17 has anyone worked on Wt in OE lately? Jan 19 15:37:26 xumbi what? Jan 19 15:37:57 http://www.webtoolkit.eu/wt, recipes/wt Jan 19 15:38:31 just curious, I'm trying to get the latest version working Jan 19 15:38:52 but it seems to be very picky about the version of boost used, and how it's compiled Jan 19 15:39:04 hm ah I remember dark Jan 19 15:39:25 I have no idea and nobody seen working on it for oe-core and meta-oe Jan 19 15:39:40 oops.... I didn't install libcursor :) Jan 19 15:39:49 elo Jan 19 15:41:54 it didn't help :) Jan 19 15:49:59 250 hello cs_nbp ;) Jan 19 15:51:26 hii Jan 19 16:03:21 kergoth: do you find any reason for bitbake 1.15 to break like that? http://pastebin.com/0FqecMeC Jan 19 16:04:06 ipaqlor: if any, kergoth will tell you why it fails Jan 19 16:06:10 check the recipe and package_ipk, see SUMMARY. the output is clearly unpleasant, but i highly doubt the failure h as anything to do with bitbake itself. it's telling you SUMMARY is referencing itself, so investigate that first Jan 19 16:06:42 btw in the oe-core recipe SUMMARY is unset Jan 19 16:07:28 which is why i also suggested looking at what package_ipk is doing. see also bitbake -e foo | grep SUMMARY Jan 19 16:08:18 iirc I had once a bad checkout. fschk did solve Jan 19 16:08:26 some strange parsing errors Jan 19 16:09:02 what is happening is that SUMMARY_${PN} is supposed to reference SUMMARY Jan 19 16:09:03 ant_work: sorry, was lost in the VARIABLES :) Jan 19 16:09:25 but since a default SUMMARY isn't set, SUMMARY == SUMMARY_${PN} after the translation stuff happens Jan 19 16:09:28 i can provide a bitbake -e dump if it helps Jan 19 16:09:53 (Note a default SUMMARY should be set on the bitbake.conf -- perhaps it fell out? Jan 19 16:10:22 SUMMARY ?= "${PN} version ${PV}-${PR}" Jan 19 16:10:27 that should be the default summary Jan 19 16:10:53 SUMMARY_${PN}-dbg ?= "${SUMMARY} - Debugging files" Jan 19 16:10:53 ah Jan 19 16:10:57 thats the line that caused the failure.. Jan 19 16:11:05 it's not self referencing -- or well shouldn't be Jan 19 16:11:05 fray: ill try that. Jan 19 16:11:42 (when I revamped the summary/description code a while back I ran into this.. it was fixed, unless something recently broke things... I doubt it was bitbake, but maybe?) Jan 19 16:11:58 fray: if you look at the pastebin, you see its retrieving SUMMARY, not SUMMARY_. at a guess, i'd say its adding pkg to overrides, applying them, then exploding due to the original SUMMARY having been overwritten Jan 19 16:12:12 but i didn't think package_ipk did that, and it does seem odd it'd blow up out of the blue Jan 19 16:12:13 so i should put SUMMARY_${PN}-dbg ?= "${SUMMARY} - Debugging files" into bitbake.conf? Jan 19 16:12:19 kergoth yes, that is what I was trying to say, but forgetting the name "overrides" Jan 19 16:12:22 :) Jan 19 16:12:30 ipaqlor it should already be there Jan 19 16:12:51 my meta/conf/bitbake.conf has a comment called "# Package info" and below that you will see the default values Jan 19 16:16:12 fray: ok, no change here Jan 19 16:16:36 ya, that I don't know what is wrong.. it's been working for me fine for at least a year.. Jan 19 16:18:04 maybe thats a clou, the atk package spliting fails at first because there is no atk debug directory created. i allmost forgot that, but i just created it manually and then hit the summary summar thingy Jan 19 16:18:34 one of these days we need to sit down and just go through our classes and clean them up. we have a lot of remnant/legacy/crap python floatting around Jan 19 16:18:42 dunno.. there won't be a debug directory if the debug creation stuff is disabled Jan 19 16:18:52 (or there is no debug info to pull out) Jan 19 16:22:57 SUMMARY="atk version 2.2.0-r1" , SUMMARY_atk-staticdev="atk version 2.2.0-r1 - Development files (Static Libraries)" , SUMMARY_atk-dbg="atk version 2.2.0-r1 - Debugging files" ... all set Jan 19 16:23:45 maybe the () :) Jan 19 16:40:16 should i use bb.note or something else to print the variablevalues for debugging? Jan 19 16:41:17 the problem seems to apear in do_package_ipk, so i thought i just smear it with printfs Jan 19 16:42:31 ipaqlor: bb.note() if it's python code, bbnote if it's shell script Jan 19 16:43:21 bluelightning: ok, thnkas Jan 19 16:58:55 mhm, ok, im feeling more and more stupidified, i cant find the function do_package_ipk :) Jan 19 16:59:37 so i grep over ...mhm lets say 6gb of osurcecode and binaries... may take an hour Jan 19 17:00:06 ipaqlor: "git grep" is your friend Jan 19 17:00:27 im not comfortable with versioning systems Jan 19 17:00:29 never used any Jan 19 17:00:31 no clue Jan 19 17:00:42 thats aother world. i dont want to go there. Jan 19 17:00:58 the othere stuff wastes my time enough. Jan 19 17:02:30 git grep is just like grep, except it limits itself to the version controlled files... really helps if you have TMPDIR underneath where your recipe files are Jan 19 17:04:07 bluelightning: thanks. but i dont even know in which project to search for. Jan 19 17:05:25 bluelightning: also my grep just scrolled out into nirvana leaving me with thousands of nice lines telling me that i have entered a recursive directory loop Jan 19 17:05:36 *gnar* Jan 19 17:05:42 ipaqlor: I think the file you want to look at is meta/classes/package_ipk.bbclass Jan 19 17:07:47 hi, when using layers is it possible in a layer to have a bbappend which would only be taken in account for one machine ? Jan 19 17:08:28 ericben: not really, but you can pretty much make all of its contents only apply for the machine via overrides Jan 19 17:08:33 actually the only solution I found was to enable/disable the layer in bblayers.conf which is quite annoying when developing for several machines Jan 19 17:08:59 bluelightning: thanks. but here i stop for today. since.... the princess is anothere castle == the codepart that does the SUMMARY thingy. Jan 19 17:09:24 hi bluelightning : thanks that would work this way Jan 19 17:09:53 highly modularised Jan 19 17:10:16 im blind, stupid and frusttrated. cu tomorrow Jan 19 17:10:26 ipaqlor: cya tomorrow Jan 19 17:16:35 hi Jan 19 17:17:14 hi eFfeM Jan 19 17:41:36 any ideas on fixing this? Jan 19 17:41:37 Starting system message bus: Unknown username "gdm" in message bus configuration file Jan 19 19:47:20 re Jan 19 19:47:35 hey Jan 19 19:48:10 any idea why muriani and me still get * extract_archive: Cannot create symlink from ./var/run to 'volatile/run': File exists. Jan 19 19:48:25 dont know if we already solved that-.- Jan 19 19:48:35 anyways its not in my list of known problems :D Jan 19 19:49:39 are you using oe-core or oe classic? Jan 19 19:49:50 oe-core and meta-oe Jan 19 19:49:57 this message is usually a result of one package having /var/run as a directory, and other having it as a symlink.. Jan 19 19:50:36 Within oe-core, there is a control file that is supposed to sync the usage of these directories, meta/files/fs-perms.txt Jan 19 19:50:50 ${localstatedir}/run link volatile/run Jan 19 19:51:12 that says that whenever the package post processing sees ${localstatedir}/run (as a directory) it is changed into a link to volatile/run Jan 19 19:51:36 ok Jan 19 19:52:16 so I'd suggest you check the packages quickly that they have the link and not directory.. if you can find a bad one, youll have to figure out why the fs-perms.txt didn't run Jan 19 19:52:24 (or does meta-oe replace it? and is missing something?) Jan 19 19:52:58 so somewhere there is a recipe that mkdirs a directory which it shouldnt? Jan 19 19:53:12 I think so Jan 19 19:53:22 gonna try to find that Jan 19 20:09:27 is that mkdir ok or souldn't it be there (udev/init) http://pastebin.com/FK5eEmD5 Jan 19 20:10:31 I think that is fine.. Jan 19 20:11:17 the only other reference to a /run I find looks like this mkdir -p ${D}/run Jan 19 20:11:48 that might be it.. but even so the creation of the directory should be fixed by the packaging system -- unless the error is coming from an initscript or post install script or something Jan 19 20:12:55 I usually get it during do_rootfs Jan 19 20:13:02 y Jan 19 20:13:10 cant extract archive Jan 19 20:13:17 for various images Jan 19 20:13:30 then it's the package itself.. Jan 19 20:13:44 I'd simply start by querying all of the packages and figure out which ones have the bad directory in them Jan 19 20:13:55 (I suspect it's a very small set) Jan 19 20:13:59 I would think that, if florian hadn't said that he doesn't get that error Jan 19 20:14:15 k ill look Jan 19 20:14:50 I don't get the error either -- but I'm not suing meta-oe Jan 19 20:15:16 actually florian is a good example because he built exactly the same thing were failing at Jan 19 20:15:46 I dunno, I'm failing at a different image. Jan 19 20:15:49 meta-netbookpre with jlime-lxde-image Jan 19 20:15:54 :p Jan 19 20:16:09 I fail at xfce-nm-image Jan 19 20:16:19 if you would build jlime-lxde-image you would fail at the same point Jan 19 20:16:24 lemme try Jan 19 20:16:25 ;) Jan 19 20:16:29 if you change the packaging type to RPM from opkg, you may get a better error message, I'm not sure.. Jan 19 20:16:33 I usually use RPM for th packaging Jan 19 20:16:59 hmm that would take a lot of time, too :) Jan 19 20:17:00 but this is an error we need to recreate and fix... the fs-perms -should- have done that Jan 19 20:17:12 I'll take a look at my configuration Jan 19 20:17:15 no.. just swtich your config to rpm instead of opkg and build.. Jan 19 20:17:17 i will try to get the cuplprit Jan 19 20:17:26 the only time is the repackage, since it doesn't rebuild everything.. Jan 19 20:17:31 ah k Jan 19 20:17:44 ill try that for a sec Jan 19 20:19:13 so i change this PREFERRED_PROVIDER_virtual/update-alternatives = "opkg" to "rpm" Jan 19 20:19:18 no Jan 19 20:19:22 that's bblayers.conf, should be harmless: http://pastebin.com/vq1HX6VR Jan 19 20:19:43 conf/local.conf: PACKAGE_CLASSES ?= "package_ipk" to "package_rpm" Jan 19 20:19:46 thats it Jan 19 20:20:00 ok and ok Jan 19 20:20:06 *gets brain overload Jan 19 20:20:07 eeks rpm Jan 19 20:20:11 you can even do PACKAGE_CLASSES ?= "package_rpm package_ipk" and it'll keep building both.. (first is teh default rootf package type) Jan 19 20:20:34 florian RPM is cpio archive + meta data.. opkg is ar archive that holds two tar archives.. one w/ meta data, one w/ files Jan 19 20:21:12 fray: even cpio is evil... and ipk is much better tested Jan 19 20:21:36 cpio can handle devices and fifos and other special files.. -posix- tar can not.. (gnu tar can for the most part) Jan 19 20:22:01 at least for oe-core, ipk, rpm and deb are equally tested.. if not they are supposed to be Jan 19 20:22:28 this particular issue should be the same for all three.. and I know what the error messages come from and mean in RPM, but I don't know how to reconcile them in opkg Jan 19 20:22:40 well it will take a while, 5300 tasks left Jan 19 20:22:54 btw my layers look very similar Jan 19 20:22:55 my guess is it'll take less then 10 minutes Jan 19 20:22:58 ^^ Jan 19 20:23:14 (glibc-locales are by far the longest part.. they take about 5 minutes on my system.. everything else is quick) Jan 19 20:23:21 of course I'm assuming you are doing parallel builds Jan 19 20:23:38 fray: hum well... sounds like some hot iron :) Jan 19 20:23:52 you mean j4? yes Jan 19 20:24:02 and yes i got 4 cpus on th build machine Jan 19 20:24:06 florian a full oe-core build takes me about an hour and a half or so.. Jan 19 20:24:18 rebuilding just the packages (everything is already built) takes minutes Jan 19 20:24:31 rebuilding the packages, is quick since you aren't compiling or installing anything Jan 19 20:24:33 hi, is there a chromium recipe in new OE? or oly in classic OE? Jan 19 20:24:52 pespin I thought I saw someone submit it for the meta-oe layer.. Jan 19 20:25:02 ok I'll look thanks :) Jan 19 20:25:05 so it's not part of oe-core, but if it's in meta-oe is part of the new stuff Jan 19 20:25:38 cs_nbp: run more compliers than cpus Jan 19 20:25:49 that's local.conf without comments: http://pastebin.com/SWQ0vEAg Jan 19 20:25:50 florian: rly? Jan 19 20:26:38 florian, what I've found is if I have a lot of ram.. BB_NUMBER_THREADS should be the same or higher then parallel_make.. Jan 19 20:26:46 lot of ram => 8 GB Jan 19 20:26:52 cs_nbp: yes... Jan 19 20:27:08 fray: ... which is quite cheap by now Jan 19 20:27:18 florian: how many times more should it be? Jan 19 20:27:27 I usually run (on an 8 core machine w/ 10 GB ram) 16 or 20 for the parallel builds.. and -j 16 for the parallel make arg Jan 19 20:27:31 if you're not buying ddr1 ram Jan 19 20:27:57 ah so -j8 should pose no probs Jan 19 20:27:59 my 4 core, 6 GB ram box.. I run 6 / 8 or 8 / 8 dependning on how much I am using the machine Jan 19 20:28:27 oh the machine stands far away in a chilled room Jan 19 20:29:04 (I have access to a 16 core -- 48 GB ram machine.. that one I usually do 48/48... it's really nice, but unfortunately runs an ancient RHEL) Jan 19 20:30:19 do we have recipe for SIP client in core based oe? Jan 19 20:30:32 not AFAIK Jan 19 20:30:35 u win :p highest i can go is 8 cpus and 16 gb Jan 19 20:30:37 ouch Jan 19 20:30:41 I don't remember seeing one in meta-oe either.. Jan 19 20:30:45 ah salut Jan 19 20:30:48 but we'd welcome one.. :) Jan 19 20:30:55 core based means not oe-classic Jan 19 20:30:59 yes Jan 19 20:31:02 the whole stuff Jan 19 20:31:05 in any possible layer Jan 19 20:31:11 but I think there is nothing Jan 19 20:31:19 A lot of folks start w/ oe-classic packages and forward port them into meta-oe (used w/ oe-core) Jan 19 20:31:41 yes but if you have some patches what to do? Jan 19 20:31:50 cs_nbp I wouldn't consider teh 16 core machine to be "mine" since it's a shared machine for everyone in my office.. ;) Jan 19 20:32:28 we (OE Technical Steering Committee) want people to migrate software that is actively being used into meta-oe, or a "new layer" if they're willing to maintain the code.. Jan 19 20:32:30 it's impossible to port forward patches without a patch header Jan 19 20:32:39 so at the end we put in our own layers Jan 19 20:32:45 instead of sharing the recipes Jan 19 20:33:02 Why? Can't you add the patch header(s)? Jan 19 20:33:16 because I've no background infos on the patch Jan 19 20:33:28 git log or git blame.. figure out who originally submitted it Jan 19 20:33:29 sometimes I don't even know why the patch is there Jan 19 20:33:32 ok Jan 19 20:34:11 some are even the packages -> recipes renames Jan 19 20:34:43 fray, I can't find any chromium related stuff in meta-oe Jan 19 20:35:50 We used to have a patch for xtscal reading the calibration data from eeprom Jan 19 20:40:26 1300 to go Jan 19 20:57:23 oh well Jan 19 21:01:12 we need a working tool for touchscreen calibration Jan 19 21:10:51 * florian tries to build another image Jan 19 21:14:51 ok i get failed dependencies error now Jan 19 21:16:51 it's rpm! Of course there's failed dependencies :D Jan 19 21:18:23 BTW dependency scanning is something RPM is -really- good at.. Jan 19 21:18:25 whast the bad dep? Jan 19 21:18:32 It's most likely a real bug Jan 19 21:18:36 nawt for vim Jan 19 21:18:43 Koen was working on a fix for that Jan 19 21:18:44 libgl1 Jan 19 21:18:51 for webkit Jan 19 21:18:53 dunno about libgl1... Jan 19 21:18:57 webkit is in meta-oe? Jan 19 21:19:13 err didt look :D Jan 19 21:19:30 I suspect it is.. we've been really careful in Yocto to track down these issues when they're found.. Jan 19 21:19:45 ah and /bin/csh is needed by vim Jan 19 21:19:52 the vim one is there are a couple of things that have incorrect dependencies --or are examples that are executable.. Jan 19 21:20:07 yup.. Jan 19 21:20:16 i lost the libgl1 error by bitbaking libgl and libsdl by hand Jan 19 21:20:32 that means there is a RDEPENDS issues.. Jan 19 21:20:35 now only nawt and csh Jan 19 21:20:43 in vim recipe? Jan 19 21:22:02 if you fixed a problem by building libgl and libsdl by hand.. ther eis a runtime bug somewhere.. Jan 19 21:22:28 the vim recipe, there are two issues.. one was the csh dep, and the other is an incorrect python dependency.. (might only hit you at runtime though) Jan 19 21:22:49 the csh one is simple to fix, since it's an example type script.. in do_install_append simply do chmod -x ${D} Jan 19 21:22:54 it will no longer generate dependencies Jan 19 21:23:05 ok sec Jan 19 21:23:44 I'm surprised it's not fixed though.. Koen and i talked about it a couple of days ago, and he's normally really quick at checking in simple fixes like that Jan 19 21:24:28 the nawt one I don't know.. I've not used it at all.. Jan 19 21:24:40 sry but where can I find this do_install_append? Jan 19 21:25:04 (BTW something I'm planning on working on in a while is to get opkg using the extended dependency information that RPM is using to help better find these issues and avoid runtime problems) Jan 19 21:25:18 modify the vim recipe and add a do_install_append (or modify one that exists) Jan 19 21:25:25 aah Jan 19 21:32:02 should that go before or after the "install -m 06400 ...etc" line in vom recipe? Jan 19 21:32:05 :) Jan 19 21:32:39 (0644 typo) Jan 19 21:33:17 ah ok after ;) Jan 19 22:32:37 is someone named Phil Blundell here? Jan 19 22:33:00 pb_ is sometimes here Jan 19 22:33:16 ah ok Jan 19 22:33:33 I just found something from something called Oedem (developers conference maybe?) from 2009 that says "Why is opkg so bad?" hehehe Jan 19 22:33:49 (attributed to him) Jan 19 22:34:01 hm its over 2 years old Jan 19 22:34:06 yeah it is. Jan 19 22:34:09 *I* don't think opkg is that bad. Jan 19 22:34:16 hm Jan 19 22:34:23 I just think it's poorly documented as crap, Jan 19 22:34:27 cannt remember but I was at oedem Jan 19 22:34:30 and I really would like to see better flexibility Jan 19 22:34:30 2009 Jan 19 22:34:33 good nite Jan 19 22:34:37 bye Jan 19 22:34:48 Particularly in the areas of connecting to private repositories... Jan 19 22:34:53 ok i'll be back later too. Jan 19 22:37:04 btw there's an GLIB_GENERATE_LOCALES variable, is there one for eglib too or is it just the same? Jan 19 22:47:13 cs_nbp should be the same variable.. Jan 19 22:49:03 when a machine conf file has PREFERRED_PROVIDER_virtual/kernel = linux how does one go about finding where that maps to a particular git source tree? Jan 19 23:13:06 spacecolonyone: the kernel recipes can have compatibility information Jan 19 23:15:59 florian: btw in the lxde console, I get shell version as prompt but in the config (in etc) username@hostname$/# is specified Jan 19 23:16:15 does that mean it cant read the config and falls back on version prompt? Jan 19 23:16:28 I guess so Jan 19 23:16:46 * florian patiently waits for his new image to upload Jan 19 23:17:40 what image^ Jan 19 23:17:47 hi I'm trying to build angstrom on lpc3250 (arm) but I got /usr/lib/gcc/x86_64-pc-linux-gnu/4.5.3/../../../../x86_64-pc-linux-gnu/bin/ld: cannot find -lgmpxx Jan 19 23:18:36 libgmpxx is clearly missing from that flder, I'm also using oe from git and bb 1.10.2 Jan 19 23:19:44 is there a reason this library is missing? Jan 19 23:20:28 cs_nbp: one with display calibration stuff Jan 19 23:21:05 florian: for nbp? Jan 19 23:21:19 sure Jan 19 23:21:21 swell Jan 19 23:21:40 well with u I don't know u seem to have many gadgets at hands reach Jan 19 23:32:09 good thing theres no syslog yet -.- Jan 20 00:44:41 florian I'm not sure I follow Jan 20 00:46:35 spacecolonyone: I'm not sure he dint disconnect ;) Jan 20 00:56:36 yea, I just noticed that. :( Jan 20 01:23:09 spacecolonyone: I thinks he wanted to tell you to look directly in the recipe for your kernel Jan 20 01:27:11 hi folks Jan 20 01:27:16 cs_nbp, that is what I'm trying to figure out. I'm not clear how to figure out wich recipe "linux" corresponds to. there are so many variants! Jan 20 01:27:47 I'm off guys, thanks for the help. I'll check in tomorrow Jan 20 01:28:53 gn8 Jan 20 01:29:54 spacecolonyone: PREFERRED_VERSION_linux? Jan 20 01:33:11 cyaZ Jan 20 01:40:40 anyone using oe-core + meta-handheld ? i setup ala https://wiki.yoctoproject.org/wiki/OpenEmbedded-Core and it is supposed to work. but its missing classes/klibc.bbclass which is in openembedded... Jan 20 01:41:27 i know i could just copy everything together....but id like to do it right and thought i could do a compilerun overnight..... Jan 20 01:42:14 so how is it done? anyone a clou? the manuals i found where alltogether outdated as far as i understand. Jan 20 01:43:15 do i get openembedded and put oe-core somwhere and add the path to bblayers or the other way around. is there any real system behind all this or is it random and my problem is that i think there is some way it should be done? Jan 20 01:45:25 well, no nightly compile then. Jan 20 01:45:41 ipaqlor: do you need meta-oe? Jan 20 01:45:52 denix: i have no idea Jan 20 01:46:40 denix: i was told i need oe-core and a "BSP" *eg* like meta-handheld ...and off i fly Jan 20 01:48:26 like you said, you need klibc.bbclass, which lives in meta-oe: http://cgit.openembedded.org/meta-openembedded/tree/meta-oe/classes Jan 20 01:54:58 denix: thanks. Jan 20 02:04:26 well, yet another time building the crosscompile toolchain. i am still not confinced about the features of such a "buildsystem" untill now it looks more like many people reordering the same set of thing over and over again Jan 20 02:05:19 blue socks to blue socks green pees to green pees Jan 20 02:05:32 good night alltogether. **** ENDING LOGGING AT Fri Jan 20 02:59:57 2012