**** BEGIN LOGGING AT Thu Feb 23 02:59:58 2012 Feb 23 03:28:35 arr Feb 23 07:45:04 Hello I am trying to compile qt4.7.1 for arm9 this is my recipe http://dpaste.com/707129/ I am getting error http://dpaste.com/707131/ please tell me where I am going wrong.. I am totally new to this Feb 23 08:45:48 * Jay7 have found this in bookmarks: http://openembedded.uservoice.com/forums/6963-general Feb 23 09:04:19 good morning Feb 23 09:16:52 hi mckoan, all Feb 23 09:21:54 hi bluelightning, all Feb 23 09:22:37 good morning Feb 23 09:25:53 i checked out org.openembedded.dev in the bitbake i found following problems Feb 23 09:25:55 ecipes/konqueror/konqueror-embedded_20070316.bb (# --disable-scrollbars \) which is in the middle of a multiline expression. Feb 23 09:26:03 org.openembedded.dev/recipes/kdepimpi/kdepimpi-base.inc (# pwmanager/libcrypt/cipher \) which is in the middle of a multiline expression. Feb 23 09:29:44 Ananth_: use bitbake 1.12 Feb 23 09:31:51 i have now is __version__ = "1.15.1" in that one is this problem fixed ? Feb 23 09:33:22 bluelightning: i have git version for bitbake, what head need to checkout Feb 23 09:34:14 Ananth_: the problem is in the metadata (which is OE-Classic), it has not been updated to work with bitbake newer than 1.12.x Feb 23 09:34:31 Ananth_: if you're starting something new I'd recommend starting with OE-Core instead Feb 23 09:34:41 Ananth_: http://www.openembedded.org/wiki/OpenEmbedded-Core Feb 23 09:35:06 otherwise you can continue to use OE-Classic but you will need to use bitbake 1.12.x Feb 23 09:40:12 Thanks Feb 23 10:10:08 anyone tried using the pvr driver for the omap4 instead of the omapfb driver? Any tips on how to do that? Feb 23 10:21:11 what is the best practise when updateing a package via opkg and it brings a new conffile .. right now opkg mentions that and doesnt copy it over but places a .opkg backup file ? Feb 23 10:21:46 rob_w: it depends on your use case, if you do not want the user to be able to modify the conf file, then you should not list those files als CONFFILEs in your recipe Feb 23 10:21:58 if its defined as conf file, then opkg will only update it if there were no changes Feb 23 10:22:08 if the user did any changes - you'll get the backup scenario Feb 23 10:22:20 just like with .rpmnew and so on on fedora for example Feb 23 10:26:38 right , thx Feb 23 12:17:03 Hi. I'm new to openembedded, so this is probably a dumb question: what to do if I encounter this kind of error? ERROR: Function 'Fetch failed: Unable to fetch URL http://www.angstrom-distribution.org/unstable/sources/git_git.savannah.gnu.org.config.git_e35217687ee5f39b428119fe31c7e954f6de64f0.tar.gz from any source.' failed Feb 23 12:18:50 roland: the anstrom-distribution.org site seems to be down Feb 23 12:19:18 hence, 'Fectch failed' means that it could not resolve the site to download the required sources :) Feb 23 12:19:51 jackmitch: all mirrors are down too? Feb 23 12:21:13 what's the usual work procedure in this kind of situation? Feb 23 12:21:16 I'm not sure what the mirror situation is with angstrom, it may just be a case of waiting Feb 23 12:21:42 i'm no expert myself so unfortuantly cannot be of much more help :( Feb 23 13:06:22 hello Feb 23 13:07:04 after a long time i'm back to openembedded. But i have some questions. Feb 23 13:12:16 First: which branche is the current best for beginnig to create a new distri not based on Ångström? currently i'm using master Feb 23 13:15:00 master has lots of problems with broken download links. I have added lots of mirrors to finish a minimal distri image. But now i would like to switch to a new 3.x Kernel. The wiki show "change kernel TBD" Feb 23 13:16:44 PREFERRED_VERSION_linux = "3.0" at local.conf din't work Feb 23 13:17:55 KERNEL = "kernel30" and MACHINE_KERNEL_VERSION = "3.0" at conf/distro/minimal.conf din't work Feb 23 13:18:29 you're trying oe-classic right? look at new layers instead for new work Feb 23 13:19:33 DEFAULT_PREFERENCE = "1" and DEFAULT_PREFERENCE_qemux86 = "1" at recipes/linux/linux_3.0.bb seem to work. The 3.0 was compiled but the header din't fit Feb 23 13:20:55 i used http://www.openembedded.org/wiki/Getting_started to start. Is there a new howto for the new "layers"? Feb 23 13:22:52 decembersoul: the closest we have to an explanation of the new structure is http://www.openembedded.org/wiki/OpenEmbedded-Core Feb 23 13:25:09 ok. Now i have to read ... Thanks. Feb 23 13:59:58 decembersoul: you may also have a look at this table version using oe-classic http://www.kaeilos.com/download Feb 23 14:00:06 s/table/stable Feb 23 14:07:03 thanks. I will try oe-core. If it din't become clear i will try kaeilos. Feb 23 16:14:03 hi khem Feb 23 16:24:39 did anyone use gold + armv4t? Feb 23 16:34:09 using oe-core, under what circumstances might I get an error "ERROR: Function failed: Fetcher failure for URL: 'file://speedup.patch'. Unable to fetch URL from any source." ? There is a directory named "files" in the same director as my recipe, and it contains a speedup.patch file Feb 23 16:37:35 awozniak: are you sure it is in FILESPATH or whatever the var name si? Feb 23 16:37:59 are you maybe adding this in your .bbappend and you forgot FILESEXTRAPATHS_prepend or something? just a wild guess :) Feb 23 16:38:26 I'm not familiar with either of those. Looking at other working recipes, I don't see those variables. Feb 23 16:38:36 are you on classic or core? Feb 23 16:38:56 oe-core Feb 23 16:39:18 and is it your own recipe "from scratch" or are you using a .bbappend to modify an existing recipe in some other layer? Feb 23 16:39:46 bbappend Feb 23 16:39:55 ok thats what I was getting at :) Feb 23 16:40:08 then you need FILESEXTRAPATHS_prepend := "${THISDIR}/files:" Feb 23 16:40:12 on top of your .bbappend Feb 23 16:40:23 kthx Feb 23 16:41:50 still no joy Feb 23 16:42:27 well where did you put your speedup.patch? Feb 23 16:42:39 under files/speedup.patch as seen from the dir where the .bbappend is? Feb 23 16:42:44 yep Feb 23 16:42:48 there are *many* examples of bbappends using FILESEXTRAPATHS Feb 23 16:42:51 just use grep Feb 23 16:45:04 my bad there, I used "FILES_EXTRAPATHS" not FILESEXTRAPATHS Feb 23 16:45:09 thanks for the help Feb 23 17:12:38 bluelightning: Thanks for sharing the elc-videos url! Feb 23 17:13:03 kenws: no problem... thank the people who gave presentations :) Feb 23 17:13:12 : ) Feb 23 17:28:09 GNUtoo|laptop: yes Feb 23 17:39:13 khem, because on SHR we got thumb interworking issues Feb 23 17:39:26 like it switches to thumb with LDR Feb 23 17:40:23 GNUtoo|laptop: oh hmmm I see Feb 23 17:41:20 GNUtoo|laptop: yes it does not have the relevant fix for armv4 Feb 23 17:41:26 let me see something Feb 23 17:41:34 ok thanks a lot!!!! Feb 23 17:42:48 what is your machine underlying arch ? Feb 23 17:42:53 armv4 or armv4t Feb 23 17:43:01 om-gta02: armv4t Feb 23 17:44:00 bluelightning, http://pastebin.com/VAHM6wnB Feb 23 17:44:01 should work alright for armv4t I guess Feb 23 17:45:00 Crofton|work: hi, Feb 23 17:45:03 Crofton|work: hmm... didn't see that one myself Feb 23 17:45:09 hmmm lookin at your fix theere is no error Feb 23 17:45:10 => 0x402ae6ac: ldr pc, [r12, #2404]! ; 0x964 Feb 23 17:46:00 then that points to: Feb 23 17:46:10 0x7cf64 : push {r3, lr} Feb 23 17:47:50 GNUtoo|laptop: ok Feb 23 17:47:59 I didn't think so Feb 23 17:48:01 then on stepi it illegal instructions Feb 23 17:48:11 doing this in the backgroundd and I wanted to let you know what I see Feb 23 17:48:21 need to make sure I didn't suck in qmake from somewhere else Feb 23 17:48:29 I can try to reproduce and get better infos if you want Feb 23 17:48:44 GNUtoo|laptop: let me look up something Feb 23 17:48:50 * khem pulls out the red book Feb 23 17:49:24 and I was told that it was still in arm modem when executing thumb instructions Feb 23 17:49:29 hence the illegal instruction Feb 23 17:49:34 *arm mode Feb 23 17:50:16 that certainly won't work on armv4t Feb 23 17:50:28 you need an interworking sequence, you can't just jump from thumb to arm like that. Feb 23 17:50:40 indeed that's what I was told Feb 23 17:51:01 if gcc is generating that code under -march=armv4t then it is a bug. Feb 23 17:51:02 but in the recent arm book they say you can jump with LDR but I was told that it's not the case Feb 23 17:51:26 I'm not sure what "recent arm book" you're talking about, but if it was written in the last decade it is probably not referring to armv4t. Feb 23 17:51:40 let me find it Feb 23 17:51:50 Crofton|work: going to run a build again just to check Feb 23 17:51:57 that code sequence will work fine on armv5t and newer, it's just v4t where it will break. Feb 23 17:51:58 ok Feb 23 17:52:09 that was a clean build with angstrom Feb 23 17:52:31 Crofton|work: chances are it's not related to angstrom Feb 23 17:52:44 yeah, I just believe in full disclosure :) Feb 23 17:52:49 but that's a useful bit of info in any case, thanks :) Feb 23 17:53:30 GNUtoo|laptop: yes pb_ is right. LDR pc .. does not do mode switch arches older than armv5t Feb 23 17:53:53 ok, must work on getting stable support in place for existing products Feb 23 17:54:32 GNUtoo|laptop: I wonder if gcc is configured wrongly Feb 23 17:54:57 GNUtoo|laptop: or options passed by the package in question are wrong Feb 23 17:55:09 I think the options are ok Feb 23 17:55:09 GNUtoo|laptop: which package is that btw. have you identified that Feb 23 17:55:15 bash Feb 23 17:55:24 ok Feb 23 17:56:02 I wonder if its some linker veneer Feb 23 17:56:08 so you're compiling with -march=armv4t, right? Feb 23 17:56:23 whereabouts in the code does the offending sequence occur? is it actually in bash's sources? Feb 23 17:56:23 yes, JaMa|Off just switched to gold and the issue appeared right after Feb 23 17:56:37 but maybe the toolchain was bumped too, I've no idea Feb 23 17:57:23 arm-oe-linux-gnueabi-gcc -march=armv4t -mthumb -mthumb-interwork -mtune=arm920t --sysroot=/home/gnutoo/embedded/oe/oe-core/shr/oetmps/build/sysroots/om-gta02 -c -DHAVE_CONFIG_H -DSHELL -I. -I.. -I.. -I../include -I../lib -I. -O2 -pipe -g -feliminate-unused-debug-types break.c Feb 23 17:57:30 example of how it's compiled ^^^^ Feb 23 17:57:47 I think it's in bash's malloc.c Feb 23 17:58:29 I think its definitely the linker induced veneer code Feb 23 17:58:51 can you disassemble the code around that ldr Feb 23 17:59:15 yes Feb 23 17:59:18 what version of gold do you have? Feb 23 17:59:20 let me gdb to it Feb 23 17:59:25 I just eyeballed the code for 2.21 and it does look broken. Feb 23 17:59:30 how to know the gold version Feb 23 17:59:32 pb_: whatever comes with 2.22 Feb 23 17:59:40 "gold --version"? Feb 23 17:59:43 ok Feb 23 18:00:10 brb Feb 23 18:00:11 well, anyway, have a look at the logic it uses to decide when to set_may_use_blx(). Feb 23 18:00:23 2.21 looks wrong, I don't have any newer version on hand to look at right now Feb 23 18:01:05 pb_: I dont think gold is so well excercised on arches less than armv5t I suppose Feb 23 18:01:15 isn't it arm*gold? Feb 23 18:01:24 because gold points to /usr/bin/gold in the devshell Feb 23 18:01:35 well, yes, whatever gold you are using Feb 23 18:01:46 armv4t-oe-linux-ld probably Feb 23 18:01:51 ok Feb 23 18:02:19 2.22 Feb 23 18:02:51 http://paste.debian.net/157320/ Feb 23 18:04:52 ok, let me have a quick look at cvsweb Feb 23 18:06:10 looks like it was probably fixed in november or thereabouts Feb 23 18:06:16 so I would guess 2.22 is probably still broken Feb 23 18:06:27 ok Feb 23 18:06:36 could try just fixing up that one line yourself, or get a newer version Feb 23 18:06:51 ok Feb 23 18:06:52 phone ringing Feb 23 18:06:55 will be back soon Feb 23 18:08:45 back Feb 23 18:08:53 hi woglinde Feb 23 18:09:21 hi gnutoo Feb 23 18:11:19 JaMa|Off, ^^^ Feb 23 18:45:08 hey, I get this out of a fresh checkout as of this afternoon: http://pastebin.com/94ybFQLd Feb 23 18:45:21 MACHINE=omap4430-panda ./oebb.sh bitbake console-image Feb 23 18:45:27 anyideas? Feb 23 18:51:57 panto are you sure that this the latest from git? Feb 23 18:53:06 woglinde, pretty sure Feb 23 18:55:34 * woglinde wonders because someone else should have been stumbled over it Feb 23 18:55:42 hm maybee a problem with meta-oe Feb 23 18:55:45 not oe-core Feb 23 18:59:13 trying it now here too Feb 23 19:05:02 no problem here Feb 23 19:05:05 hm Feb 23 19:34:42 panto, woglinde maybe previous workdir were re-used, I often have to clean the toochain because of that Feb 23 19:43:53 he said clean Feb 23 19:44:01 but I made it clean here too Feb 23 19:44:03 and it worked Feb 23 19:45:39 diff --git a/conf/local.conf b/conf/local.conf Feb 23 19:45:39 index 1c8dc76..51f17b9 100644 Feb 23 19:45:39 --- a/conf/local.conf Feb 23 19:45:39 +++ b/conf/local.conf Feb 23 19:45:39 @@ -13,8 +13,10 @@ IMAGE_FSTYPES += "tar.bz2" Feb 23 19:45:40 # Make use of SMP: Feb 23 19:45:42 # PARALLEL_MAKE specifies how many concurrent compiler threads are spawned per bitbake process Feb 23 19:45:44 # BB_NUMBER_THREADS specifies how many concurrent bitbake tasks will be run Feb 23 19:45:46 -PARALLEL_MAKE = "-j2" Feb 23 19:45:48 -BB_NUMBER_THREADS = "2" Feb 23 19:45:50 +# PARALLEL_MAKE = "-j2" Feb 23 19:45:52 +# BB_NUMBER_THREADS = "2" Feb 23 19:45:54 +PARALLEL_MAKE = "" Feb 23 19:45:56 +BB_NUMBER_THREADS = "1" Feb 23 19:45:58 Feb 23 19:46:00 DISTRO = "angstrom-2010.x" Feb 23 19:46:02 Feb 23 19:46:06 this fixes it as well Feb 23 19:47:08 ehm, that could have been two lines, rather than 15 Feb 23 19:47:10 no paste here Feb 23 19:47:18 ~pastebin Feb 23 19:47:19 [~pastebin] A "pastebin" is a web-based service where you should paste anything over 3 lines so you don't flood the channel. Here are links to a few : http://www.pastebin.com , http://pastebin.ca , http://channels.debian.net/paste , http://paste.lisp.org , http://bin.cakephp.org/ , http://asterisk.pastey.net/ , or install pastebinit with yum or aptitude. Feb 23 20:09:00 woglinde, yeah I know, this was faster Feb 23 20:22:07 looks like disabling parallel make and threads fixes it Feb 23 20:28:45 panto strange Feb 23 20:28:55 that should not happen Feb 23 20:29:02 it is what it is Feb 23 20:29:06 and your are not using this much threds Feb 23 20:29:21 I know Feb 23 20:31:19 khem, i need a way to build without threads. So, in DISTRO_FEATURES, do you prefer me to use "pthread" or "pthreads" ? Feb 23 20:32:59 blindvt: use pthreads Feb 23 20:33:07 k Feb 23 20:33:20 are you getting far with that btw. Feb 23 20:34:08 khem, btw.. IIRC i have some commentary typo fix in one of the repos in the one aldot/fixes (or fixups, dunno) branches. What should i do to get that applied upstream? Feb 23 20:34:39 khem, didn't start yet. So if you beat me to it, i'd be glad since i don't have much time ATM Feb 23 20:35:11 khem, should be pretty straight forward. No magic involved, as you know. Feb 23 20:36:43 khem, the most involved step is to fixup/add dependencies but for me as micro-interested that's not too involved either Feb 23 20:36:51 (obviously) Feb 23 20:37:59 woglinde: it happens here also http://fpaste.org/8pvT Feb 23 20:39:25 I've seen similar problems before, but was unable to track it down... Feb 23 20:39:49 usually reruning the bitbake "fixed it" for me Feb 23 20:42:08 looks like something fishy here http://fpaste.org/ODeU Feb 23 20:43:04 older bitbake? Feb 23 20:43:16 1.15.1 Feb 23 20:43:19 is that old? Feb 23 20:43:48 no its the actual Feb 23 20:43:58 I dunno, it shouldn't matter which bitbake version Feb 23 20:44:07 it might Feb 23 20:44:16 woglinde, khem, http://git.openembedded.org/openembedded-core-contrib/log/?h=aldot/fixes Feb 23 20:44:27 typo fix in base.bbclass Feb 23 20:45:07 commentary typo fix in base.bbclass, even Feb 23 20:45:07 blindvt send it to the oe-core ml Feb 23 20:59:48 GNUtoo|laptop: can you try this patch http://sakrah.dontexist.org/files/0001-binutils-gold-Correct-interworking-checks.patch Feb 23 21:00:10 GNUtoo|laptop: this should hopefully fix the gold issue on armv4t you had Feb 23 21:00:18 wow thanks a lot Feb 23 21:01:14 GNUtoo|laptop: please rebuild bash after applying this patch. It should already take care of rebuilding binutils since I have pr bump in there Feb 23 21:01:23 and report back if the issue is gone Feb 23 21:01:24 ok Feb 23 21:01:48 GNUtoo|laptop: I'll start rebuild as soon as you confirm this fix and push it to oe-core-contrib/shr Feb 23 21:01:55 JaMa|Off, ok Feb 23 21:02:08 * JaMa|Off started sdk build to keep buildhost busy for now :) Feb 23 21:02:47 ok I'll stop building angstrom for my bug and build bash for shr Feb 23 21:03:42 JaMa|Off: I still owe you the gcc rebuild fix. I have not looked into it Feb 23 21:04:51 khem: reboud or headerfix? :) Feb 23 21:04:58 s/reboud/rebuild/g Feb 23 21:05:17 What is the current bitbake task that replaces "mrproper"? That is, what do i "bitbake -c" to remove downloaded source? Feb 23 21:05:27 reboud Feb 23 21:05:33 teamnoir: cleanall Feb 23 21:05:35 khem: but not with more sstate fixes it seems to be a bit better (haven't seen it in last rebuild from scratch) Feb 23 21:05:39 ty Feb 23 21:05:43 GNUtoo|laptop: that patch may not apply directly Feb 23 21:05:45 (which imo is misleading, as we have tasks like 'fetchall' which do something en tirely different) Feb 23 21:05:46 let me refix it Feb 23 21:06:13 I concur, "cleanall" is weird by comparison to fetchall. Feb 23 21:06:16 khem: maybe it was caused by this http://git.shr-project.org/git/?p=meta-smartphone.git;a=commit;h=c6be068de2f716c9fdfc54970fd9908b4f2cd0e6 Feb 23 21:06:25 khem: so feel free to ignore rebuild issue for now Feb 23 21:07:08 khem, it applies Feb 23 21:07:53 GNUtoo|laptop: oh it does hmm ok Feb 23 21:08:01 may be then I messed up my local build heh Feb 23 21:08:23 ok Feb 23 21:08:37 thanks a lot, I'm building now Feb 23 21:08:47 np thank me if it works :) Feb 23 21:09:04 JaMa|Off: so you seem to be one happy man than without any issues :) Feb 23 21:09:09 Can I get bitbake to tell me where, specifically, it is fetching from? Feb 23 21:10:42 khem: sure I'm man without issues :) just ladies doesn't know it yet (at least not all of them :)) Feb 23 21:12:12 hmmm heh Feb 23 21:12:29 dont fall prey thats all I would suggest :) Feb 23 21:12:44 GNUtoo|laptop: I wonder what version of binutils do u have there Feb 23 21:12:55 teamnoir: I assume you mean for a specific app? Feb 23 21:12:58 since the patch is already applied to binutils 2.22 in oe-core Feb 23 21:13:10 wmat: yes, for a specific component. Feb 23 21:13:24 khem, I'll look but I think it's 2.22 Feb 23 21:13:34 yup should be 2.22 Feb 23 21:13:39 http://paste.debian.net/157320/ Feb 23 21:14:12 khem, the oe patch applies cleanly Feb 23 21:14:20 but not the gold patch Feb 23 21:14:42 GNUtoo|laptop: the patch I gave you is failing right ? Feb 23 21:15:21 yes Feb 23 21:15:28 at do_patch Feb 23 21:15:32 ok. thats what I expected Feb 23 21:15:33 should I fix locally? Feb 23 21:15:45 no, I just checked that patch is already in Feb 23 21:15:52 let me look more Feb 23 21:23:50 teamnoir: grep the recipe file for SRC_URI Feb 23 21:27:12 GNUtoo|laptop: I guess I will have to build for armv4t myself here and see whats going wrong in bash Feb 23 21:27:30 GNUtoo|laptop: so you have ARM_INSTRUCTION_SET = thumb" Feb 23 21:27:35 by default right ? Feb 23 21:28:12 khem: are you sure about this patch being applied in oe-core/master? I don't see it there Feb 23 21:28:33 JaMa|Off: its applied to original sources of gold in 2.22 Feb 23 21:28:47 ah I see Feb 23 21:28:59 so I guess there is a real bug here. Feb 23 21:29:05 I want to reproduce it Feb 23 21:29:34 does shr use thumb by default ? Feb 23 21:32:00 hmmm I git pulled bitbake wrong idea probably Feb 23 21:32:08 lets see what breaks Feb 23 21:33:37 JaMa|Off: what errors were you getting with gcc-cross-initial Feb 23 21:33:47 I guess I am seeing same probably Feb 23 21:33:56 as: line 87: /home/kraj/work/angstrom/sources/openembedded-core/build/tmp-uclibc/sysroots/x86_64-l Feb 23 21:33:59 inux/usr/bin/i586-oe-linux-uclibc/i586-oe-linux-uclibc-as: No such file or directory Feb 23 21:34:01 something like that ? Feb 23 21:38:06 hmmm so in gcc-cross-initial.inc we say DEPENDS = "virtual/${TARGET_PREFIX}binutils ... Feb 23 21:38:22 and then few line below we say INHIBIT_DEFAULT_DEPS = "1" Feb 23 21:39:01 now if I have a configured gcc-cross-initial and my binutils-cross wants to rebuild I have a problem since the dep is not respected ? Feb 23 21:40:23 no my problem was in do_patch Feb 23 21:40:25 it didn't apply Feb 23 21:40:36 GNUtoo|laptop: thats unrelated to your problem Feb 23 21:40:59 ok Feb 23 21:43:14 khem: no work-shared/gcc-4.6.2+svnr181430-r22/gcc-4_6-branch/gcc/calls.c:1089:9: error: 'STACK_CHECK_MAX_VAR_SIZE' undeclared (first use in this function) Feb 23 21:43:25 khem: that's why I was thinking about headerfixes.. Feb 23 21:43:29 DEPENDS="virtual/i586-oe-linux-uclibc-binutils Feb 23 21:43:45 so thats not issue Feb 23 21:44:02 any gettext guru here? I wonder if it's correct patch http://paste.ubuntu.com/854578 "gettext infrastructure mismatch" type of error.... Feb 23 21:44:44 JaMa|Off: that define is in defaults.h Feb 23 21:44:55 which we munge in gcc's do_configute Feb 23 21:45:24 JaMa|Off: so whast the scenario when this happens Feb 23 21:47:01 khem: today every build from scratch was just failing like this Feb 23 21:47:13 hmmm Feb 23 21:47:24 thats the failure in gcc-cross-initial ? Feb 23 21:47:25 khem: most builds I did with qemux86-64 Feb 23 21:48:12 khem: yes.. see cooker.. 16:34:23 < JaMa|Off> khem: cooker log http://paste.pocoo.org/show/555585/ and referenced do_compile.log http://paste.pocoo.org/show/555586/ Feb 23 21:48:54 so just clean dir and "bitbake perl" Feb 23 21:49:14 and this was happening with distroless oe-core Feb 23 21:49:44 khem: and yes shr uses thumb by default Feb 23 21:51:48 JaMa|Off: back to gcc-cross-initial issue its not using sstate to populate from ? Feb 23 21:52:05 so its a clean clean build from scratch not just tmpdir deletion Feb 23 21:53:45 no sstate and yes clean Feb 23 21:53:56 you can see that from cooker log :) Feb 23 21:54:41 I am seeing it but confirming is better Feb 23 22:10:24 what could be the root cause, that I'm getting this error *** error: gettext infrastructure mismatch: using a Makefile.in.in from gettext version 0.17 but the autoconf macros are from gettext version 0.18 Feb 23 22:10:28 just sometimes? Feb 23 22:11:13 just doing "bitbake -c cleanall elfutils-native && bitbake elfutils-native" Feb 23 22:11:29 sometimes it builds fine, sometimes it ends with that gettext error Feb 23 22:11:49 reproducible builds :) Feb 23 22:34:39 esbenh: ping Feb 23 22:48:43 ynezz: not the best solution, oe should gettextize it.. but if there is autopoint called before then gettextize isn't run and such error is shown Feb 23 22:49:36 ynezz: see autotools.bbclass or 45d7250d64bf7d7c2d29d2dbf61bb6c30b683a80 0d9457c6773ad9d9de7576b5b7ed5610b2d2b8d5 716d12924e03d7ad54517ccbcd8c5439b8f1c348 in meta-oe Feb 23 22:49:45 it looks like it's impossible to build angstrom on oe-core now Feb 23 22:49:56 with more than one bb thread Feb 23 22:50:13 10 errors and still counting, now git... Feb 23 22:50:32 there were some changes in bitbake recently? Feb 23 22:51:26 I'm building incrementally for past 2 months, say once at least every week, but don't remember it that broken Feb 23 22:52:19 bitbake shouldn't affect that at all. and doesn't, here Feb 23 22:52:37 the only bitbake issues could be in potential parse hangs (fixed) and error handling (mostly fixed) Feb 23 22:55:11 ok, what could cause so many errors due to increased (4) number of bb threads? Feb 23 22:55:35 it's just a bad luck today? :) Feb 23 22:56:36 I did few rebuilds from scratch today with 2 threads fine Feb 23 22:57:50 ok, trying with 2 also Feb 23 22:59:34 with 4 BB_NUMBER_THREADS and -j 4 it feels like always waiting for IO (and eats lot more memory) so for my quad core system I always use just 2 BB_NUMBER_THREADS Feb 23 23:00:49 I use 4 BB threads on my quad core, though actually the dependency tree seems to mean that I seldom get 4 tasks executing in parallel for any length of time. Feb 23 23:01:22 I use 1.5*cores for BB_NUMBER_THREADS and 2*cores for PARALLEL_MAKE, usually, but will occasionally adjust the multipliers Feb 23 23:01:48 It is a bit sad that the build process still seems to be fundamentally a rather serial one. I guess there are just these unavoidable pinch points that everything has to wait for. Feb 23 23:02:38 I'm sure there's still work to be done on that. The more we can trim down the dependency digraph, the fewer of those points we should have. e.g. i'm sure we have a ton of stuff depending on -natives that might or might not actually be needed for a given build machine.. Feb 23 23:02:41 * kergoth shrugs Feb 23 23:02:48 heh what type of disks do you have? my TMPDIR is on 3 SATA2 RAID0 and still everytime I check htop it's doing just IO operations Feb 23 23:03:28 I usually just do a multi-build of a variety of combinations and see how they each perform in practice before I settle for a given machine Feb 23 23:04:06 well, thank you all for the input Feb 23 23:04:42 and also git status from buildhistory takes a lot of time after build.. sort of expected with 751M buildhistory/ Feb 23 23:05:03 hmm, wonder if we can make that more efficient Feb 23 23:05:40 kergoth: git status in my shr-chroot is even slower then this :) Feb 23 23:05:56 but that's a lot of binaries.. so again sort of expected Feb 23 23:06:38 hehe, yeah Feb 23 23:38:10 WARNING: /home/kraj/work/angstrom/sources/meta-ti/recipes-misc/payload/bonescript.bb: it is recommended to set SYSTEMD_PACKAGES as -systemd Feb 23 23:38:13 ERROR: Command execution failed: Exited with 1 | ETA: 00:01:29 Feb 23 23:38:16 Summary: There was 1 WARNING message shown. Feb 23 23:38:18 Summary: There was 1 ERROR message shown, returning a non-zero exit code. Feb 23 23:38:25 updated bitbake and others Feb 23 23:38:41 now what bitbake does is that it does not tell me what the error is Feb 23 23:51:34 did you update it about hour ago? Feb 23 23:51:42 no about 3 Feb 23 23:51:59 I do have that WARNING (from different recipe) but error messages should be fixed now Feb 23 23:52:44 khem: then update again, you need this one: http://git.openembedded.org/bitbake/commit/?id=356c42ed00541dea2c382c57f768a4f431d22db0 Feb 23 23:53:25 JaMa|Off: yes updare helped Feb 23 23:53:30 but I still have an error Feb 23 23:53:56 http://pastebin.com/84uXYqza Feb 23 23:55:15 someone needs to fix meta-opie may be Feb 23 23:56:35 yes, koen droped it from setup-scripts and added meta-initramfs Feb 23 23:57:48 yeah I guess best Feb 23 23:58:28 ok so now I am back on track Feb 23 23:59:27 * JaMa|Off going to bed build is running and it won't finish before I have to wake up for daywork :) Feb 24 00:00:15 gn Feb 24 00:11:19 hello. are there any tools to create a mirror of packages locally for bitbake? Feb 24 00:12:14 i'd like to eliminate surprises, like that of lsof i'm getting: http://pastebin.com/mVZZWLqg Feb 24 00:13:24 CMoH-notebook: usually with oe-core we have sane default mirrors Feb 24 00:14:14 you can always INHERIT = "own-mirror" in local.conf and then provide SOURCE_MIRROR_URL Feb 24 00:15:09 khem, ty. so do i need to copy the locally cached download folder to SOURCE_MIRROR_URL? Feb 24 00:15:33 yes Feb 24 00:15:59 ah, btw... khem... after that 530 error the do_fetch just hangs, and to me it seems the mirrors never get tried Feb 24 00:16:18 hmmm Feb 24 00:16:22 well, at least within 5 minutes :) Feb 24 00:18:02 i've also had the experience that your master went too much ahead our versions, so having the packages cached locally might be worth the effort Feb 24 00:19:44 thanks for your help, khem Feb 24 00:25:14 khem: do you include preferred-opie-versions ? Feb 24 00:26:56 i.e. require conf/distro/include/preferred-opie-versions-1.2.5.inc in your local.conf? Feb 24 00:27:06 ant__: no Feb 24 00:27:18 something angstrom should do Feb 24 00:27:33 ${OPIE_GIT_PV} unexpanded, is undeclared Feb 24 00:30:34 I know Feb 24 00:30:48 something wrong with ltg.org Feb 24 00:31:59 CMoH-notebook: you can also use BB_FETCH_PREMIRRORONLY option Feb 24 00:32:07 which means it wont search outside of your premirror Feb 24 00:32:30 ah, even better - thanks Feb 24 00:32:34 i'll have to set this up Feb 24 00:32:58 and BB_NO_NETWORK would asssume you have no network Feb 24 00:35:35 i see - although that would prevent having a common mirror for a team Feb 24 00:36:37 it is for remote network access Feb 24 00:36:44 local network should still work Feb 24 00:37:39 aha! the variable name seemed more restrictive Feb 24 00:37:43 thanks for all the info Feb 24 00:40:25 one other question: how can i force bitbake to rebuild a package? now it seems that the -c clean command does not operate so thouroughly as in oe-classic Feb 24 00:40:56 bitbake -f -c compile Feb 24 00:42:10 i've read that, but it only seems to force a recompile; what about the do_package, etc ? Feb 24 00:43:18 Khem, how would one share say an Angstrom project with other teams that are distributed around the globe? Host the project on some git server? I don't know what this would look like or if there are any best practices out there. Feb 24 00:45:39 bitbake -c clean package and then bitbake package again? Feb 24 00:47:55 hutchman: you can use sstate Feb 24 00:48:12 hutchman, it seems that for some recipes bitbake 'optimizes' away the actual operation Feb 24 00:48:27 I heard you guys talking about that at ELC but I'm not familiar with it. Is that a OE thing or a Yocto thing? Feb 24 00:48:54 yocto is not different in terms of OE features Feb 24 00:49:02 dont be confused Feb 24 00:49:17 I missed Beth's talk and that is probably the one I should have been in. Feb 24 00:49:27 shared state is a OE-Core feature. Feb 24 00:49:56 Ah, OK. So I need to search and read up on it. Feb 24 00:50:31 hutchman: yes. sstate lets you store prebuilt packages and had sophesticated signature mechanism Feb 24 00:50:50 and can smartly invalidate a pakcage buiild Feb 24 00:51:08 you can put shared state on a common server that all your devs then can use Feb 24 00:51:14 and they dont have to rebuild everything Feb 24 00:51:45 for sharing project metadata I guess you could just use a git server that all see Feb 24 00:51:49 and they can clone from it Feb 24 00:52:22 i'm also interested in details about this - i've only found items on mailing lists from people that already know what it's all about, but less documentation on how to set it up Feb 24 00:52:38 Problem is I have to share things with other groups in India and Italy and our corporate firewall always gets in the way ... we may have to do it through vpn or something. Feb 24 00:52:43 like source mirrors Feb 24 00:52:48 you just set SSTATE_MIRROR Feb 24 00:52:52 is there a documentation on how it operates, how to set it up for the server/client side? Feb 24 00:53:02 yes Feb 24 00:53:18 I think either on yocto wiki or poky manual Feb 24 00:53:24 describes it IIRC Feb 24 00:54:15 or you can pay me and arm and leg and I can train your folks :) Feb 24 00:54:47 Already paid arm and leg to go to ELC :) Feb 24 00:55:15 heh Feb 24 00:55:39 hehe, if arms and legs you're missing :P Feb 24 00:55:53 * khem hops on to road Feb 24 00:55:55 I'm typing with one hand and one foot. Feb 24 00:59:54 sstate stuff https://lists.yoctoproject.org/pipermail/yocto/2011-April/003504.html Feb 24 01:00:31 damn... i wanted to meet khem :) Feb 24 01:04:37 Go to ELC Europe next and you can :) Feb 24 01:07:34 :) Feb 24 01:07:59 well, thanks for the freebies :) keep in touch Feb 24 01:08:07 nn Feb 24 01:11:24 .... didn't see anything in the poky manual about sstate. Feb 24 01:11:50 hutchman: the next release version of the docs cover it in some detail, FWIW Feb 24 01:13:46 Are you talking about the April Yocto release or is there a OE core release coming up? Feb 24 01:17:14 hutchman: the April Yocto release (since you refer to the poky manual) Feb 24 01:17:43 Cool. Thanks, I'll be looking for it. Feb 24 01:23:29 * bluelightning -> bed Feb 24 01:23:30 night all Feb 24 01:24:39 same here ... tired of waiting for build to finish so I'm going to vege on TV a while. Guess I need 16 cores. Feb 24 02:26:49 Does anyone know what might be wrong if "systemctl" command responds "Failed to get D-Bus connection: No connection to service manager."? **** ENDING LOGGING AT Fri Feb 24 02:59:58 2012