**** BEGIN LOGGING AT Fri Jan 25 02:59:59 2013 Jan 25 09:15:35 morning all Jan 25 09:37:00 morning bluelightning Jan 25 09:38:36 good morning Jan 25 09:40:04 hi lyang0 Jan 25 09:40:50 hello Jan 25 09:42:48 it seems there is no variable is for "sysroots" Jan 25 09:43:18 lyang0: there is Jan 25 09:43:27 which Jan 25 09:43:31 I can't find it Jan 25 09:43:41 just see it's export to $PATH Jan 25 09:44:14 STAGING_DIR Jan 25 09:44:42 you asked about adding the sysroot bin path to PATH the other day Jan 25 09:44:55 note that that may not be a good idea Jan 25 09:45:33 you can extract it from the "pseudodone" file if you need to, or alternatively you can do bitbake -e | grep ^STAGING_BINDIR_NATIVE= in a script Jan 25 09:46:02 however, that's not necessarily going to give you an environment capable of running all tools in the native sysroot Jan 25 09:46:30 depending on what you're trying to do, you may want to use devshell or a proper SDK Jan 25 09:51:07 hi all, i'm trying to add some perl modules to our custom layer, i used a template coming from existing perl modules, but when trying to compile module, i get Errno.pm (native) complaining about different architectures. Jan 25 09:51:43 I will check thanks Jan 25 09:52:07 something similar to https://lists.yoctoproject.org/pipermail/poky/2011-January/002371.html Jan 25 09:53:43 thanks , that's what I need Jan 25 10:01:19 pr0gg3d: I'm no perl expert but it would seem to me something on the host is being read during compilation in order to determine the architecture Jan 25 10:01:34 pr0gg3d: first port of call would be the configure/compile logs Jan 25 10:02:37 bluelightning: yes, Errno.pm module in perl-native has that check on startup, and in "straight" perl, there's a patch to remove that check. Jan 25 10:03:03 bluelightning: i'm trying to figure out, why it's patched as it. Jan 25 10:03:30 pr0gg3d: git history might be helpful there.... git blame then git log Jan 25 10:13:26 sometimes I like git log than git blame, git blame only can find the latest modification, git log -GXXXXX. can find the line contains XXXXX ever modified by someone in the history Jan 25 10:13:53 bluelightning: i will check that. thanks for now Jan 25 10:15:40 lyang0: my version of git reports -G is not a valid option Jan 25 10:16:12 you can specify a revision with git blame and it will look at that version of the file and before Jan 25 11:40:29 bluelighting I checked git blame append revision , it can't give a full history , it can git the info before revision Jan 25 11:41:10 lyang001@lyang0-9010:~/kvm_32/layers/oe-core$ git log -GFULL_OPTIMIZATION --pretty=format:%H:%an meta/conf/bitbake.conf Jan 25 11:41:10 2864ff6a4b3c3f9b3bbb6d2597243cc5d3715939:Richard Purdie Jan 25 11:41:10 5b9877aa55e28ba06410592350e7e5106ebdf6cb:Chris Larson Jan 25 11:41:10 b22831fd63164c4db9c0b72934d7d734a6585251:Richard Purdie Jan 25 11:41:10 9cb7113790d716a4c5cf7d511535ba87fdecd1ac:Khem Raj Jan 25 11:41:10 2096c315df2d7d0a8fbf9df1be73f56b834c3340:Richard Purdie Jan 25 11:41:10 b2f192faabe412adce79534e22efe9fb69ee40e2:Richard P Jan 25 11:42:09 this give the commit list which ever change the line contains FULL_OPTIMIZATION in file bitbake,conf Jan 25 11:42:21 lyang001@lyang0-9010:~/WORK/leaning-work/git$ git version Jan 25 11:42:22 git version 1.7.10.4 Jan 25 11:47:33 lyang0: ah right, I was checking on my centos 6 machine which has 1.7.1 Jan 25 11:48:41 git log -S is similar though, only difference is -G supports a regex whereas -S is just a plain string Jan 25 12:00:26 bluelighting run above cmd, -S get less commit, it seems not get the full history Jan 25 12:02:18 hmm, you're right... Jan 25 12:03:44 looks like a bug to me Jan 25 12:26:03 bluelightning ,where is a common place to put tar.bz2, and there is no http or ftp location for tar.bz2 , it just my local tar.bz2 Jan 25 12:58:57 is anyone here a cmake expert? Jan 25 13:03:38 I have seen it Jan 25 13:04:04 depending on the question, I may be able to impersonate one Jan 25 13:05:58 heh Jan 25 13:06:16 libmusicbrainz does this thing where it does a native build of a tool it needs first, and then builds the library (using that tool) Jan 25 13:06:22 but, if it ends up re-building, it fails Jan 25 13:06:54 from what i can tell, because when it does the native build it thinks that its cross-compiling (via the toolchain file, i think), and that blows up Jan 25 13:08:15 CMakeFiles/ is all generated, right? Jan 25 13:08:27 so i could just wipe that away in configure Jan 25 13:33:14 lyang0_: probably in a files directory next to the recipe as with other files Jan 25 13:33:20 other local files that is Jan 25 13:33:45 I see someone put it in downloads dir which defined in layer.conf, I think that's bad Jan 25 13:34:18 downloads dir often put the http or ftp tar.bz2 locally Jan 25 13:35:58 lyang0_: if it is a small file put it in /files as suggested by bluelightning, otherwise create a public github repo nowadays Jan 25 13:36:37 yes Jan 25 13:36:41 you don't want bloated metadata, don't you ? Jan 25 13:36:49 I mean tar.bz2 here Jan 25 13:37:09 what does bloated mean Jan 25 13:37:27 a tar.bz2 without a remote address Jan 25 13:37:29 we do that only for some kernel/psplash logo Jan 25 13:37:37 bloated = big sized Jan 25 13:37:53 inlflated Jan 25 13:38:18 *inflated I mean Jan 25 13:38:45 I can imagine some bootloader blob as well Jan 25 13:41:24 is there an example? Jan 25 13:43:16 ant_work do you mean a git bare clone repos Jan 25 13:44:20 rburton: cmake is a mystery to me Jan 25 13:44:35 its being weird Jan 25 13:44:42 and making rebuilds a nightmare in libmusicbrainz Jan 25 13:44:52 rburton: I noticed that fail here too :/ Jan 25 13:45:05 (which is making me want to move it to meta-multimedia more and more, its only in oe-core for a gstreamer plugin that nobody uses) Jan 25 13:46:53 ant_work, the tar.bz2 only locally without a github or ftp http. Jan 25 13:47:39 I remind only the files has a remote address can be put in download dir of the top dir Jan 25 13:47:48 as said you can use the local fetcher (file:// ) or you can create a github repository Jan 25 13:47:55 see http://cgit.openembedded.org/openembedded-core/log/meta?qt=grep&q=github Jan 25 13:49:54 yes here, I mean where is right place for local file Jan 25 13:50:48 in meta-foo/bar/files/xyz.tar.bz2 of bar recipe Jan 25 13:51:07 local file=local tar.bz2, someone put it in downloads dir on the top dir eg: met-foo/downloads Jan 25 13:51:14 I think it's wrong place Jan 25 13:51:57 if so, file:// will can't fetch it rightly Jan 25 13:52:45 file://xyz.tar.bz2 Jan 25 13:53:42 yes, then it will fetch it from meta-foo/bar/files/, but not met-foo/downloads Jan 25 13:53:55 correct Jan 25 13:54:57 is there a reason why populate_sdk brings in qemu? Jan 25 13:55:48 is there somewhere user definable where you can set the 'sdk tools' Jan 25 13:56:13 or should I fall back to using a mutant meta-toolchain image? Jan 25 15:17:33 <_Lucretia__> hi, what is the module parameter for in a hg:// SRC_URI? Jan 25 15:19:37 I don't understand Jan 25 15:19:57 <_Lucretia__> a mate is trying to get an url working, it's failing for somereason Jan 25 15:20:37 'module' is really pretty meaningless outside of svn, afaik Jan 25 15:21:11 <_Lucretia__> well, apparently it's demanding it Jan 25 15:21:28 <_Lucretia__> ERROR: Function failed: URL: 'hg://hg.assembla.com/openada.8;protocol=https' is missing the required parameter 'module' Jan 25 15:21:56 <_Lucretia__> I've said to try: SRC_URI = "hg://hg.assembla.com;protocol=httpsmodule=openada.8" Jan 25 15:22:02 <_Lucretia__> I've said to try: SRC_URI = "hg://hg.assembla.com;protocol=https;module=openada.8" Jan 25 15:23:07 _Lucretia__: cmd = "%s clone %s %s://%s/%s %s" % (basecmd, " ".join(options), proto, hgroot, ud.module, ud.module Jan 25 15:23:34 <_Lucretia__> yeah I saw that already Jan 25 15:24:17 ugh, stupid. so it's just being interpreted as the last element of the uri Jan 25 15:27:54 <_Lucretia__> yup Jan 25 15:29:20 * kergoth shakes head Jan 25 15:35:51 i sent a series of patches for ssh/sftp support for the bitbake fetch2 library early this week. i haven't got any feedback (for understandable reason, it was somewhat lengthy). should i open a ticket? Jan 25 15:37:12 zibri: fwiw I haven't noticed them and I don't see them on patchwork, but I'm interested Jan 25 15:37:33 I like the idea of the uri class -- but then i tried to get something similar merged in the past and RP shot it down. but maybe the fact that yours doesn't use the existing urlparse uri class and namedtuples will change the response Jan 25 15:37:33 jama: it was on the bitbake-devel list Jan 25 15:37:37 ah found it http://patchwork.openembedded.org/patch/43027/ Jan 25 15:38:00 (didn't know about patchwork :-), thanks!) Jan 25 15:38:11 using the secsh uri stuff is obviously the Right Thing, imo, so i back that 100%, but richard makes the calls on merges for bitbake Jan 25 15:40:04 * imolit lurks Jan 25 15:40:39 zibri: I read through it and didn't find anything I strongly disliked. Need to spend more time reading it Jan 25 15:41:36 rp: thanks. i'll wait then :). no hurry Jan 25 15:42:22 someone seeing hangs when trying to interrupt bitbake with Ctrl+C? Jan 25 15:42:22 I have a native recipe which during do_install is installing some executables, and then copying some libs (via cp) into ${libdir} Jan 25 15:42:22 I see the executables are put into the sstate populate-sysroot.tgz file but the libs I copied in are not, so when other develoeprs try to reference the libs they get errors. What do I have to do to get the libs to in that file? Jan 25 15:42:49 I think it's also related to PRServ like #3742 Jan 25 15:43:30 JimNH: to ${D}${libdir}? Jan 25 15:48:03 JaMa: I was having issues with bitbake hanging at parsing recipes the other day, seems resolved now though Jan 25 15:48:24 couldn't ^C out either Jan 25 15:48:35 JaMa hmm I was doing ${libdir} Jan 25 15:48:47 let me try ${D}${libdir}.... Jan 25 15:48:51 ${libdir} is /usr/lib/. unless you're running as root, that won't do much Jan 25 15:50:17 my ${libdir} is an absolute path pointing into my build tree at tmp/sysroots/x86_64-linux/usr/lib Jan 25 15:50:53 that's only true for native/cross recipes. Jan 25 15:51:05 and even then won't get packaged as you've discovered Jan 25 15:51:17 right this is a native build where I need to export some libs into the native sysroot Jan 25 15:51:24 ${D} is the root of the destination filesystem, everything is found relative to there Jan 25 15:51:30 then you need to install to ${D} like every other recipe does :) Jan 25 15:51:34 ah Jan 25 15:52:09 files don't get installed into sysroots directly, they get installed to the location where they get packaged, and the files get pulled into the sysroot and binary packages both from there Jan 25 15:56:11 JimNH: use openembedded-core/scripts/sstate-sysroot-cruft.sh to find where else did you break this rule Jan 25 15:56:49 ooh, good idea, i need to play with that more Jan 25 15:57:19 :) tu JaMa and kergoth; I realize I was being too clever in lookin for a symbol in the bitbake -e output; Jan 25 15:57:44 oh -e is a very valuable tool, most people *underuse* it :) Jan 25 15:57:50 but in this case less helpful Jan 25 15:58:36 hey guys, just wanted to say thanks, I got my project ported over to meta-systemd on top of yocto danny release and everything is booting up properly (mostly), thanks to a couple of you that provided pointers. Jan 25 16:00:11 mebrown: good news! :) Jan 25 16:00:43 a hint for people that have problems with this: boot to init=/bin/sh and then strace /lib/systemd/systemd --system Jan 25 16:01:40 One question: are there any docs for packaging systemd services? Jan 25 16:03:10 mebrown: if there is nothing in meta-systemd, I wouldn't have thought so Jan 25 16:03:43 mebrown: however, it's likely we will add something to the documentation for 1.4 now that the systemd support is going into the core Jan 25 16:03:54 I'm just reading systemd.bbclass, but there are not really any just overview docs Jan 25 16:05:54 mebrown: agreed. basically you just set SYSTEMD_PACKAGES and SYSTEMD_SERVICE_* Jan 25 16:06:00 have a look at the bbappends in meta-systemd Jan 25 16:06:16 thanks, that's what I was using Jan 25 16:06:29 the lack of documentation is why i didn't merge systemd.bbclass into oe-core last week Jan 25 16:06:37 and it still makes me uneasy :/ Jan 25 16:06:38 reading code, it looks like it is ok to have multiple services in SYSTEMD_SERVICE? Jan 25 16:07:50 afaik, yes Jan 25 16:09:36 and, stupid question that I'll find out the answer to in an hour anyway when I finish this build: in the *.service file, if i have [Install] section that has WantedBy (iirc), that will automatically be linked? Jan 25 16:44:41 rburton: integrated eth on my Asus motherboard got renamed to p4p1.. :) Jan 25 16:49:04 heh and usbnet from openmoko gta02: Jan 25 16:49:06 ID_OUI_FROM_DATABASE=OPENMOKO, INC. Jan 25 16:49:06 ID_NET_NAME_PATH=enp0s19f0u3 Jan 25 16:52:19 is there a way to prioritize IRQs in the RT-kernel? can't find a recipe for rtctl in yocto/OE Jan 25 16:53:15 JaMa: ouch Jan 25 17:13:04 is udev missing 70-persistent-net.rules? can't find it anywhere and /lib64/udev/write_net_rules is broken as well Jan 25 17:13:39 schramae: atm you can override 80-net-name-slot.rules putting a copy in /etc/udev/rules.d and copying there the old 70-persistent-net.rules Jan 25 17:14:20 http://www.freedesktop.org/wiki/Software/systemd/ Jan 25 17:14:20 PredictableNetworkInterfaceNames Jan 25 17:15:21 ant_work: i have no 80-net-name-slot.rules as well. should this be in meta/recipes-core/udev/? Jan 25 17:15:27 udev 197, isn't? Jan 25 17:15:38 I've been hit updating entoo Jan 25 17:15:43 *Gentoo Jan 25 17:16:45 i have 164 from danny release tarball, that explains it. thanks Jan 25 17:18:06 ok, atm you're immune but be prepared :) Jan 25 17:18:24 thought youwere testing latest stuff Jan 25 17:19:52 i was using the git version before, but switched back to the tarball because of bug #3634 and to keep things simpler Jan 25 17:19:53 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=3634 critical, Medium+, 1.3.1, ross.burton, IN PROGRESS IMPLEMENTATION , Udev broken in Danny branch (0140519ba15bfb27ccbfb3d41c7e678a1118fc5c) Jan 25 17:20:28 wasn't that fixed? Jan 25 17:22:28 rburton: have you seen http://patchwork.openembedded.org/patch/41691/ for danny? Jan 25 17:22:47 rburton: yes, but i thought it would be easier and more reproducible to use the tarball instead of a git version with manually pulled in commits when finally releasing the source for end customers. i might have to rethink this... Jan 25 17:24:49 JaMa: btw I'll send one morel patch for fixing v4bx in klibc.bbclass, this times for the external binaries built by klcc Jan 25 17:25:11 the other patch is for the klibc-utils, built by gcc Jan 25 17:25:53 btw, I see it is still in your queue Jan 25 17:27:49 JaMa: yes. my danny-pending mail folder is rather full. i plan to empty it on monday. Jan 25 17:29:41 zibri: did you do any performance testing of encode/decode w/ the class vs without? Jan 25 17:29:50 rburton: ok, thanks Jan 25 17:30:09 ant_work: feel free.. I just wanted to have bootable armv4t images with vanila danny Jan 25 17:30:47 atm we produce valid kernels (> 2.6.31) and binaries for armv4 Jan 25 17:32:57 (in master) Jan 25 17:35:20 khem: ping Jan 25 17:35:41 oh, nice: ARM: compressed/head.S: work around new binutils warning Jan 25 17:35:54 have youseen the: Jan 25 17:35:56 arch/arm/boot/compressed/head.S: Assembler messages: Jan 25 17:35:57 arch/arm/boot/compressed/head.S:565: Warning: (null) Jan 25 17:36:01 ... Jan 25 17:36:44 khem: eglibc-mtrace and eglibc-scripts provide the same mtrace binary, eglibc-mtrace also has 'RDEPENDS_${PN} = "perl"' eglibc-scripts only bash Jan 25 17:37:36 khem: s/binary/script/ Jan 25 17:40:41 kergoth: nope... what would be a good way to test it? i did try this just now, but it's probably a bit too naive: http://codepad.org/SYsVsckY Jan 25 17:41:54 naive, but it does at least show a lack of an obvious problem. i usually use the timeit module myself, usually with repeat() - http://docs.python.org/2/library/timeit.html Jan 25 17:45:49 does anyone saw this problem running an image on qemuarm? Jan 25 17:45:51 sh: -list: command not found Jan 25 18:02:56 rburton: wipe-sysroot why not remove BUILDDIR completely? Jan 25 18:38:14 kergoth: there does seem to be a performance hit when using timeit... http://codepad.org/Vzu7Pqqw am I doing anything wrong? otherwise i'll investigate it Jan 25 18:42:30 hmm, I'd suggest profiling them to investigate further Jan 25 18:45:56 kergoth: thanks, will do Jan 25 18:48:09 we should really try to reduce the bitbake module interdependencies, so we can do better at testing and profiling them in isolation Jan 25 18:48:26 ah well, we're making progress, slowly but surely Jan 25 18:56:18 JaMa: takes longer :) Jan 25 19:15:25 rburton: to delete or recreate? Jan 25 19:15:54 rburton: I mean, what else is there to keep? (unless you keep stuff like buildhistory and precious deploy images) Jan 25 19:17:42 FWIW in my setups I always change PERSISTENT_DIR outside TMPDIR, because having something "persistent" in directory starting with tmp doesn't sound right Jan 25 19:47:23 hi all ... i'm hoping to build an app for gumstix using yocto (with _Lucretia__) Jan 25 19:50:10 it looks like we may need blas and lapack libs but, so far, i haven't been able to find much info on these libs re yocto Jan 25 19:50:26 would anyone have any suggestions or pointers ? Jan 25 20:06:10 need some pointers on how to debug an issue... when I build my image it is rerunning all the copy_image and write_package_ipk tasks for my 'localsource' packages. Jan 25 20:55:02 What is the tools which we can use with bitbake to check dependencies? I always forget it Jan 25 20:56:45 -g Jan 25 21:04:42 but there're a gui app for it Jan 25 21:09:14 someone asked that on ML Jan 25 21:10:16 otavio: http://lists.linuxtogo.org/pipermail/bitbake-devel/2013-January/004082.html Jan 25 21:10:23 I usually just grep them Jan 25 21:11:06 JaMa: I found the -u depexp in bitbake Jan 25 21:11:15 JaMa: bitbake -g -u depexp Jan 25 21:14:59 does linux-yocto in oe-core have device tree support Jan 25 21:15:21 gah Jan 25 21:15:26 linux-yocto.inc Jan 25 21:18:10 it should work. we use device tree everyday @ Wind River based on it. Jan 25 21:19:36 Crofton|work, I take it you want to avoid linux-yocto.inc ? We could always break the include out into two parts. Jan 25 21:21:34 not, I am trying to figure out why something didn't get made, so looking at the code Jan 25 21:21:39 and went to the class first Jan 25 21:22:09 ok. yell in my direction if something is foobar'd. but I'll be afk shortly. Friday! :) Jan 25 21:22:15 yeah Jan 25 21:22:17 same here Jan 25 21:22:27 need to get ready to hike to the bar in the snow Jan 25 21:22:59 once upon a time, some guys from Toronto where trying to impress me with their winters Jan 25 21:23:09 :P Jan 25 21:23:18 they were a bit surprised it got cold and snowed here Jan 25 21:23:27 and they are on the north side of a lake Jan 25 21:23:36 so not quite as coold as you Jan 25 21:26:25 freeezing here. Jan 25 21:27:09 define freezing? Jan 25 21:28:27 -7 here (and I converted to units you would understand) Jan 25 21:28:32 shoudl get over 0 tomorrow Jan 25 21:28:34 -20c Jan 25 21:29:10 my jacket gets brittle in about a minute outside :P Jan 25 21:30:55 lovely Jan 25 21:36:10 RP: ping Jan 25 21:53:04 anyone know what the heck the default root password is for meta-ivi? Jan 25 21:53:14 they patch one into shadow via a base-passwd bbappend Jan 25 21:53:25 Hi all, I have a question about packages with init scripts (e.g., netbase). I don't want my custom distro to try to start networking on boot. Is there a good way to remove the /etc/rc?.d/[SK]40networking scripts? Jan 25 21:53:45 look at the recipe, it defines variables that the update-rc.d bbclass obeys Jan 25 21:53:53 any variable can be overrirden with an override Jan 25 21:53:57 yeah, I tried that Jan 25 21:54:00 FOO_pn-somerecipe = "bar" in a .conf Jan 25 21:54:13 ah, I didn't try that. Jan 25 21:54:45 I created a netbase_5.0.bbappend in my layer and did: INITSCRIPT_PARAMS = "" Jan 25 21:54:49 recipes can and do override global config metadata Jan 25 21:54:58 that's a perfectly reasonable way to do it too Jan 25 21:55:02 but that didn't seem to do what I wanted Jan 25 21:55:05 but use bitbake -e to make sure it did it. Jan 25 21:55:23 i'm not sure thats the right variable to adjust, offhand, but use of an append in general is a reasonable approach Jan 25 21:56:40 yeah, it seemed like it, but I rebuild my image and the files were still there. Maybe they're there because the last build left them there? Jan 25 21:56:56 is there some way to clean things out? Jan 25 21:57:50 images are generally wiped each time Jan 25 21:57:58 but you can use -c clean like you do with any other recipe Jan 25 21:58:56 I'll give that a shot -- thanks Jan 25 21:59:11 you mean the rootfs is cleaned each time? **** ENDING LOGGING AT Sat Jan 26 02:59:59 2013