**** BEGIN LOGGING AT Mon Mar 07 02:59:57 2011 Mar 07 04:07:43 woglinde: if you look at the logs for this channel, http://fastr.github.com/articles/node-waf-on-gumstix.html hacked around my 'import Scripting' issue. On new new issues now. Mar 07 05:04:49 okay, i'm hoping someone can help me piece things together Mar 07 05:05:45 how does OpenEmbedded, Angstrom, Poky, and Yocto all fit together? Please be kind, as I am a kernel developer and have been asked to look at some distributions. Thanks. Mar 07 05:06:58 sjhill, OE is not a distribution, its a framework Mar 07 05:07:15 Angstrom is a Distribution based on OE. Mar 07 05:08:47 Yocto is project to provides templates, tools and methods to enable a developer to create embedded distros Mar 07 05:09:06 it uses OE and other tools (like Poky) to do that. Mar 07 05:10:20 sjhill, I am an SA who dabbles in Kernel stuffs so I understand your concern about "being kind" Mar 07 05:11:18 ka6sox: thx, you're a ham too i see, n8fii is my call Mar 07 05:11:30 excellent Mar 07 05:11:37 so what is poky? Mar 07 05:12:16 http://www.yoctoproject.org/projects/poky Mar 07 05:12:30 thats a better explaination than I could give Mar 07 05:14:25 I've been using it for about 6yrs...starting with NAS boxes and now phone stuffs. Mar 07 05:15:40 so angstrom and poky just use OE to build a distribution for the specific platforms that they want to support, is that accurate? Mar 07 05:16:18 you define the recipies (or use a meta set), the target, and the toolchain. Mar 07 05:16:59 angstrom is a distribution that is built using OE and OE/Bitbake/Poky are parts of that framework to build it. Mar 07 05:17:44 recipies are probably closest to things in portage for Gentoo. Mar 07 05:18:23 since they aren't "packages" but rather the formula to build the packages. Mar 07 05:19:19 what this means for an embedded developer is that when the company decides to switch targets (ie migrate from MIPS to ARM) you only have to change your target. Mar 07 05:19:35 and rebuild Mar 07 05:21:27 sjhill, does this make sense? Mar 07 05:21:36 i'm trying Mar 07 05:21:52 i have tabs for OE, Angstrom, Pok, and Yocto all open and i'm trying to relate them Mar 07 05:22:25 Angstrom and Yocto are what i would classify as 'distributions' Mar 07 05:22:42 i can download a completed image and boot it on the desired target hardware Mar 07 05:23:23 OE has all of the recipes/directions to cross build hundreds of OSS packages Mar 07 05:23:26 OE is framework/recipies, Bitbake is tool for automating cross-compiles, Angstrom(aka angstrom-distribution) is just that..a distro. Mar 07 05:23:39 sjhill, correct Mar 07 05:24:07 poky and Openembedded-core are trying to standardize on a core of packages that every distro needs. Mar 07 05:24:44 Bitbake is constantly under development and debugging...it is the "engine" that creates the distros. Mar 07 05:25:34 bitbake is a package manager Mar 07 05:25:48 oe-core is the main repository Mar 07 05:26:04 yocto is the image builder / installer Mar 07 05:26:21 bitbake is the package manager in the same sense that emerge in Gentoo is Mar 07 05:26:26 yes Mar 07 05:26:47 meaning that it builds dependencies and stages them. Mar 07 05:26:57 as well as packaging up the final result. Mar 07 05:27:05 stages? Mar 07 05:27:35 if a lib is needed for DEPEND or RDEPENDS it builds and makes it available for building/running Mar 07 05:28:00 ahh yes, makes sure the build environment is ready for a particular packages Mar 07 05:28:02 yeah Mar 07 05:28:18 right... Mar 07 05:28:22 I was a bit confused, as to me stages imply merging to something other than a /tmp file system Mar 07 05:28:42 emerge can't do that on its own Mar 07 05:29:01 since we can't use x86 libs as build deps it has to create and "stage" them for use. Mar 07 05:29:11 for the "target" Mar 07 05:31:37 sjhill, any clearer? Mar 07 05:31:48 so is this more accurate? oe-core: core packages, bitbake: package manager, poky: image/distro builder, yocto: parent project, angstrom: poky-built distro Mar 07 05:31:53 jedahan: yes Mar 07 05:31:58 ka6sox: yes Mar 07 05:32:35 now if we only had a bot to remember all this. Mar 07 05:32:37 oh well. Mar 07 05:34:10 so angstrom and yocto have real bootable outputs, everything else are frameworks and tools to create them Mar 07 05:35:04 yes Mar 07 05:35:53 and narcissus will give you a bootable image (it builds images from packages that are already pre-built) Mar 07 05:37:29 which is how we got OE to work on the B&N Nook in record time. Mar 07 05:37:36 yeah....let's skip that one for now....my brain is full Mar 07 05:37:58 it is a drink from a firehose Mar 07 05:39:50 oh, narcissus is actually pretty nice...i like that Mar 07 05:41:03 sjhill, its one of the most instensive processes we support...I had to add RAID0 SSD to keep up with the demand. Mar 07 05:41:32 so to make things easier, angstrom and yocto should merge Mar 07 05:41:56 voila, one distribution Mar 07 05:42:14 sjhill, not really...different purposes. Mar 07 05:42:34 since OE is a framework it supports many different "configurations Mar 07 05:42:35 " Mar 07 05:42:44 minimal, x11, etc Mar 07 05:42:50 okay Mar 07 05:42:56 slugOS is another "distribution" Mar 07 05:44:32 just as gcc can be used for many different things(and supports other languages besides C) so does OE support other "distributions" Mar 07 05:44:45 that makes sense Mar 07 05:44:53 for example WebOS is built using OE. Mar 07 05:45:09 i guess i was just saying that i didn't see much differene between angstrom and yocto Mar 07 05:45:21 by i'm ignorant at this point Mar 07 05:45:25 s/by/but/ Mar 07 05:46:01 i do understand OE's purpose a lot better than 30 minutes ago Mar 07 05:46:25 I would say that Yocto seeks to make an "industrial strength" set of tools. Mar 07 05:47:21 and Poky is one of those Mar 07 05:48:13 http://www.yoctoproject.org/docs/yocto-quick-start/yocto-project-qs.html Mar 07 05:48:22 i think the diagram is mislabeled then Mar 07 05:48:34 'Poky Workflow' should be 'Yocto Workflow' Mar 07 05:52:21 correct? this is my last question, honest Mar 07 05:54:34 you would have to ask RP about that...I don't know. Mar 07 05:54:45 fair enough Mar 07 05:54:59 thx all of you, it helped tremendously Mar 07 05:55:03 like I said I'm a SysAdmin Mar 07 05:55:30 (who dabbles in kernel stuffs) Mar 07 05:57:37 73's Mar 07 06:12:41 oe-core does have a sato target that is recommended in the yocto docs. distributing that target via nightly builds would be the closest I'd expect to yocto having an official distro... Mar 07 06:17:35 I thought it was more tools oriented Mar 07 06:17:41 and a "core" of packages. Mar 07 07:20:13 Hi Mar 07 07:20:16 http://pastebin.com/FA4k3hhS Mar 07 07:20:24 bitbake failing Mar 07 07:20:59 following "http://www.openembedded.org/index.php/Getting_started" Mar 07 07:21:11 am i on the right path? Mar 07 07:27:19 relayhost = [smtpout.secureserver.net] is ignored Mar 07 07:30:26 oops Mar 07 07:30:34 sorry for that last line Mar 07 08:04:34 gm Mar 07 08:23:43 Hi, I'm trying to create a package wich provides openssl-conf, but when I'm building the final image, opkg fails because my-package conflicts with openssl-conf (When i use bitbake to create package-depends.dot, it is libcrypto which pulls openssl-conf) Am i missing something ? Mar 07 08:28:12 I tried PREFERRED_PROVIDER_openssl-conf without success Mar 07 08:37:57 which is a good distro for OE? Ubuntu? failing on debian and gentoo Mar 07 08:38:09 Unknown Event: Mar 07 08:38:10 Unknown Event: Mar 07 08:49:25 nataraj, angstrom Mar 07 09:32:44 letme get in straight Mar 07 09:33:09 Angstrom is a distro based on OE, if so any other distro on OE? Mar 07 09:44:48 nataraj, slugOS Mar 07 09:45:06 nataraj, WebOS Mar 07 09:46:35 webos really ? Mar 07 09:48:17 ho Mar 07 09:48:27 dj-death: yes, webos is OE based Mar 07 09:51:04 nataraj: there are about 30 distro's, see conf/distro Mar 07 09:51:09 * eFfeM_work uses minimal Mar 07 09:52:55 nataraj: and your build problem is probably because something is wrong in your config or in your tree (or you missed setting an env var) Mar 07 10:38:26 DISTRO_VERSION = "2010.7-test-20110307" Mar 07 10:39:43 is nano outdated? Mar 07 10:40:15 anyother package for testing? Mar 07 10:44:20 minimal Mar 07 10:44:32 like eFfeM_work said Mar 07 10:45:12 btw minimal is not a package it is a distro Mar 07 10:45:35 nataraj: but your problem is not nano, your problem is that you are missign some setting in your env vars or because your local.conf is not ok Mar 07 10:45:49 or because your files are not all thre Mar 07 10:47:12 nataraj: you might want to peek at contrib/testing/testscript.sh Mar 07 10:49:07 i guess you should be able to run that if you have all the prerequisites installed Mar 07 10:54:55 Cloning into openembedded.... Mar 07 10:55:08 strange Mar 07 10:58:37 eFfeM_work, testscript.sh started cloning Mar 07 10:58:50 that's good, it'll take some time Mar 07 10:59:36 testscript will create a complete tree for you, fetch bitbake, openembedded etc, it is a script aimed for regression testing on automated build systems Mar 07 10:59:53 it'll create the tree on the place you called the script Mar 07 11:02:41 i am on a normal user Mar 07 11:02:58 hope no need of root access Mar 07 11:05:47 not needed if all the packages are installed, there is not too much needed Mar 07 11:51:18 OE should be listed here: https://wiki.mozilla.org/Mobile/Platforms Mar 07 14:21:48 it would be nice to bump fennec Mar 07 14:51:29 Heh, I just updated to bitbake 1.12.0 and I'm getting some build failures Mar 07 14:52:07 http://pastebin.com/zg390uPk Mar 07 14:52:27 Why does it send the logfile to /dev/null? Mar 07 14:56:33 A bug Mar 07 15:01:01 Tartarus, is that directed to me? Mar 07 15:03:23 Does this happen to anyone else? Mar 07 15:04:37 B_Lizzard: for me everything is fine: BitBake Build Tool Core version 1.10.2, bitbake version 1.10.2 Mar 07 15:05:12 I said I updated to 1.12.0 Mar 07 15:05:27 1.10 worked OK for me Mar 07 15:07:03 ok Mar 07 15:07:39 B_Lizzard: you may even try to checkout bitbake master Mar 07 15:08:11 I'll check it out Mar 07 15:08:16 Heh Mar 07 15:08:18 B_Lizzard: btw about ben-nanonote, there is some work in progress Mar 07 15:08:27 kexec is broken on mips Mar 07 15:08:52 Ah, the kernel recipe doesn't use kexec Mar 07 15:08:54 B_Lizzard, yes, it's a bitbake bug Mar 07 15:08:55 well, there was a patch for kexec-tools and the kernel part Mar 07 15:09:05 the latter is not merged :/ Mar 07 15:09:13 Uh huh Mar 07 15:09:22 OK, thanks Tartarus Mar 07 15:09:29 Is it fixed or do you need help? Mar 07 15:09:50 I'm trying with kyak on #kexecboot Mar 07 15:10:08 in fact we have a semi/working linux-kexecboot Mar 07 15:10:33 btw, I don't own the NN Mar 07 15:10:42 B_Lizzard: fixed in master, see 90fc09ee050a24095a488d8ce98c7945bf282f6d in bitbake git for reference. just need to backport part of it to 1.12 Mar 07 15:11:10 kergoth, I'll just use git master, then, thanks a bunch ;) Mar 07 15:11:14 kergoth: is bitbake supposed to be working with openembedded-core? Mar 07 15:11:21 I wouldn't suggest going back to 1.10, I'd really recommend you try master for a while instead. The performance improvement with 1.12/master is such that there's no way in hell I'd go back to 1.10 Mar 07 15:11:30 kergoth: I mean bitbake/master Mar 07 15:11:32 JaMa|Off: it does work with it. Mar 07 15:11:40 B_Lizzard: np Mar 07 15:11:57 kergoth: linux-libc-headers do_install fails strangly, with bitbake version included in poky it works ok Mar 07 15:12:25 i built up to poky-image-minimal on friday or so without problems. Mar 07 15:12:29 did you set BBFETCH2=true? Mar 07 15:12:45 also, don't forget to continue to use the oe-core bitbake wrapper, or things *will* break eventually Mar 07 15:12:57 poky has a wrapper around the bitbake execution which runs the entire bitbake process under pseudo Mar 07 15:13:21 kergoth: http://paste.pocoo.org/show/349559/ and yes I have BBFETCH2 and I'll check oe-core wrapper Mar 07 15:13:27 here's what i use for my setup, currently (i don't use the poky env setup script): Mar 07 15:13:33 {~oeyocto} % cat setup.sh Mar 07 15:13:33 PATH=$PWD/oe-core/scripts:$PWD/bitbake/bin:$PATH Mar 07 15:13:34 export BUILDDIR=$PWD Mar 07 15:13:34 export BBFETCH2=true Mar 07 15:14:05 and $PWD/bitbake is checkout from master? Mar 07 15:14:35 as I'm installing bitbake to system (in gentoo with ebuild) so I don't set this to PATH.. Mar 07 15:14:52 yep Mar 07 15:15:15 that's fine, but still add oe-core/scripts/ to it Mar 07 15:16:05 ok, retrying now Mar 07 15:16:19 i'll update and try here too, since i haven't done it since last week Mar 07 15:17:46 I have this problem for longer.. so probably won't change Mar 07 15:20:22 kergoth: with wrapper it built tar-replacement-native first, but then failed the same.. I'll rebuild from scratch.. to see if using pseudo for whole build changes something Mar 07 15:21:10 very odd Mar 07 15:21:52 -c clean failed now too http://paste.pocoo.org/show/349568/ Mar 07 15:25:08 unrelated Mar 07 15:26:45 Ugh, bitbake from master is stuck at "Parsing recipes: 25%" for about 5 minutes now Mar 07 15:27:02 Previously it was stuck at rebuilding my cache Mar 07 15:28:27 So that's on a clean cache? Mar 07 15:29:10 Yeah Mar 07 15:29:42 It stuck on my previous cache, so I deleted the cache folder and it's stuck again Mar 07 15:30:03 what distro / python version? Mar 07 15:32:40 Python 2.6.6 on Ubuntu 10.10 Mar 07 15:32:51 heh Mar 07 15:33:04 That should work, but u10.10 has shitty disk performance Mar 07 15:34:06 how do i get bitbake to rebuild my toolchain? Mar 07 15:34:38 bitbake -c clean virtual/libc virtual/libc-for-gcc gcc gcc-cross gcc-cross-initial gcc-cross-intermediate Mar 07 15:34:45 bitbake virtual/libc Mar 07 15:36:12 I'll look into it Mar 07 15:36:19 Maybe I'm doing something wrong Mar 07 15:41:18 thanks Mar 07 15:41:45 03Koen Kooi  07org.openembedded.dev * r2ac51f0adb 10openembedded.git/contrib/angstrom/ (personal-feed.sh upload-packages.sh): Mar 07 15:41:45 angstrom feed scripts: make 2010.x the default and upload to the EU mirror by default Mar 07 15:41:45 Signed-off-by: Koen Kooi Mar 07 15:41:48 03Koen Kooi  07org.openembedded.dev * r208d503d74 10openembedded.git/recipes/bluez/bluez4_4.89.bb: bluez4: add 4.89 Mar 07 15:48:09 hello, i got a error related with glibc any suggestions https://gist.github.com/858663 Mar 07 15:49:22 check at this line | gconvert.c:55:2: error: #error GNU libiconv not in use but included iconv.h is from libiconv Mar 07 15:51:09 brolin: you built libiconv... Mar 07 15:51:09 i'm in a ubuntu 64bits 10.04 Mar 07 15:52:01 o... interesting Mar 07 15:52:21 sorry, no idea Mar 07 15:52:32 10.04 is newer then my OE builds Mar 07 15:52:43 i tried, but get that doesn't exists a recipe for libiconv Mar 07 15:53:11 this exactly ERROR: Nothing PROVIDES 'libiconv' Mar 07 15:53:14 no, it is rather about iconv.h from your system Mar 07 15:53:32 uhmm Mar 07 15:53:52 brolin: sudo apt-get install --reinstall libc6-dev Mar 07 15:54:00 brolin: that's glib, not glibc. you need to investigate where it is finding libiconv'c and stop that happening. Mar 07 15:56:30 or, perhaps conversely, if this is the -native build and you want it to use libiconv, figure out why the configure script didn't detect that. Mar 07 15:57:03 ok Mar 07 16:01:31 with the hrw suggestion the process pass the configure, now compiling Mar 07 16:02:37 error Mar 07 16:02:53 :'( Mar 07 16:04:40 the same thing, now follow suggestion of pb__ Mar 07 16:11:39 wow.. samba is fetching at 2-3Kb/sec Mar 07 16:17:32 bitbake -c clean world should delete/clean everything that bitbake built, right? Mar 07 16:17:43 ls Mar 07 16:18:34 gandhijee: no Mar 07 16:18:48 kergoth: what does that do then" Mar 07 16:18:49 ? Mar 07 16:18:53 clean runs against the recipe you ask it to run against only, not their dependencies, generally Mar 07 16:19:02 would you really expect it to clean *everything* up to ncurses when you bitbake -c clean ncurses? Mar 07 16:19:05 not likely :) Mar 07 16:19:27 i see Mar 07 16:19:54 so how do i get it to clean out my world and reindex my packages? i just moved my bitbake stuff from a 64bit machine to a 32bit machine Mar 07 16:19:59 and need to get my toolchain back up Mar 07 16:24:46 keep the packaged staging directory Mar 07 16:24:48 nuke the rest of tmp Mar 07 16:25:22 thats what i thought, just wasn't too sure Mar 07 16:26:24 thanks Tartarus Mar 07 16:37:00 hi kergoth Mar 07 16:37:25 hey pb__ Mar 07 16:57:03 morning Mar 07 17:03:59 fray: around ? Mar 07 17:15:42 morning khem Mar 07 17:16:04 khem here.. somewhat paying attention to IRC Mar 07 17:16:08 rebuild from scratch with scripts/bitbake finished (gcc-cross) do_install in linux-libc-headers still fail :/ Mar 07 17:16:09 (in meetings) Mar 07 17:17:00 fray: I have found that there is this rpm invocation where my do_rootfs is dying Mar 07 17:17:14 fray: its return code is 36 Mar 07 17:17:32 an RC !=0 simply tells you the number of files it tried to process.. :( Mar 07 17:17:59 fray: I ran it under strace Mar 07 17:18:15 and exit was called with value 36 Mar 07 17:18:30 if it was errno it could give some info Mar 07 17:21:06 ya, error messages are verbosly reported with "-vv".. they don't track rc codes.. (unfortunately but thats the reality of things) Mar 07 17:23:18 if you can figure out where in RPM it's dropping out and exiting, then I can investigate -- or if you can point me to a reproducer (so far I've not been able to find it) Mar 07 17:23:41 I put an echo before that command it printed it but echo after that it never reached Mar 07 17:23:56 so I am assuming that rpm invocation is problem Mar 07 17:24:00 brb Mar 07 17:24:19 the -vv output -- in generall -- should show you open of the db, processing of the files to install, installation of the files, close of the db Mar 07 17:24:20 you probably need ubuntu 11.04 Mar 07 17:24:35 from the logs from Friday, it showed open, process and close.. but no error.. Mar 07 17:25:02 when I've seen this in the past it's always been related to a failed chroot -- (which is handled by pseudo) Mar 07 17:31:40 So, why do we depend on target gettext in gettext.bbclass? Mar 07 17:32:50 tartarus: that does seem weird. the obvious answer would be "for libintl", but virtual/libintl seems to be in there as well. Mar 07 17:33:33 Tartarus: btw you'll be certainly happy to hear that I could not reproduce the lzma-native and coreutils-native races anymore Mar 07 17:33:41 k Mar 07 17:33:58 about compressors, some needs more love in the classes Mar 07 17:34:53 remember the lzma dep for kernel? should be the same for lzo, isn't? Mar 07 17:34:59 I do wonder why gettext.bbclass still exists, it seems like it would be better to fold it into autotools. Mar 07 17:35:17 ant_work, yeah Mar 07 17:35:24 ant_work, and please, lzo2 not lzo, heh :) Mar 07 17:35:30 and xz.. Mar 07 17:35:32 pb__, possibly due to toolchain unfun? Mar 07 17:37:10 ah, it looks like gettext was there first, and khem added libintl in 61f0980015f29d6050d6d44b8029b757978b51d6. Mar 07 17:37:21 so, maybe he intended to replace gettext with libintl but forgot to delete the former. :-} Mar 07 17:37:40 it seems unlikely that he meant to have both, since gettext PROVIDES libintl. Mar 07 17:49:51 gm Mar 07 17:51:16 can anybody confirm autoconf213-native breakage (yes this particular old version. it's needed by firefox, fennec) ? Mar 07 17:51:46 breakage in what way? Mar 07 17:52:31 some path problem. I want to make sure it's not a local thing Mar 07 17:52:33 | sed: can't read /home/luke/linux/openembedded/tmp/work/i686-linux/autoconf213-native-2.13-r13.0/autoconf-2.13/Makefile.am: No such file or directory Mar 07 17:53:48 03Tom Rini  07master * ra1a3a89566 10openembedded.git/recipes/docbook-sgml-dtd/docbook-sgml-dtd-native.inc: Mar 07 17:53:48 docbook-sgml-dtd-native.inc: Switch to subdir= in SRC_URI Mar 07 17:53:48 Signed-off-by: Tom Rini Mar 07 17:53:51 I will check if the directory exists Mar 07 17:53:57 03Tom Rini  07master * r26b9094010 10openembedded.git/recipes/freondemo/freondemo-media-files.bb: Mar 07 17:53:57 freondemo-media-files: Switch to subdir= in SRC_URI Mar 07 17:53:57 Signed-off-by: Tom Rini Mar 07 17:53:59 03Tom Rini  07master * r0d92284599 10openembedded.git/recipes/kxml2/kxml2_2.3.0.bb: Mar 07 17:53:59 kxml2: Switch to subdir= in SRC_URI Mar 07 17:53:59 Signed-off-by: Tom Rini Mar 07 17:54:00 03Tom Rini  07master * r15a758837c 10openembedded.git/recipes/netpipes/netpipes_4.2.bb: Mar 07 17:54:01 netpipes: Switch to subdir= in SRC_URI Mar 07 17:54:01 Signed-off-by: Tom Rini Mar 07 17:54:05 03Tom Rini  07master * r82040fcd3a 10openembedded.git/recipes/cnc/openredalert_r438.bb: Mar 07 17:54:05 openredalert: Switch to subdir= in SRC_URI Mar 07 17:54:05 Signed-off-by: Tom Rini Mar 07 17:54:07 03Tom Rini  07master * r1c47b7759b 10openembedded.git/recipes/microcom/microcom_1.02.bb: Mar 07 17:54:07 microcom: Switch to subdir= in SRC_URI Mar 07 17:54:07 Signed-off-by: Tom Rini Mar 07 17:54:10 03Tom Rini  07master * rff507097ac 10openembedded.git/recipes/qpf-fonts/ (qpf-arabic_1.0.bb qpf-unifont_1.0.bb qpf-unismall_1.0.0.bb): Mar 07 17:54:10 qpf-{arabic,unifont,unismall}: Switch to subdir= in SRC_URI Mar 07 17:54:10 Signed-off-by: Tom Rini Mar 07 17:54:11 03Tom Rini  07master * r2d64e45b7f 10openembedded.git/recipes/gtk-engines/elementary-icon-theme_2.3.bb: Mar 07 17:54:12 elementary-icon-theme: Switch to subdir= in SRC_URI Mar 07 17:54:12 Signed-off-by: Tom Rini Mar 07 17:54:18 03Tom Rini  07master * r286780bbd7 10openembedded.git/recipes/regina-rexx/regina-rexx_3.3.bb: Mar 07 17:54:18 regina-rexx: Switch to subdir= in SRC_URI Mar 07 17:54:18 Signed-off-by: Tom Rini Mar 07 17:54:22 03Tom Rini  07master * r07da5f9766 10openembedded.git/recipes/slugtool/slugtool.bb: Mar 07 17:54:22 slugtool: Switch to subdir= in SRC_URI Mar 07 17:54:22 Signed-off-by: Tom Rini Mar 07 17:54:25 * dcordes_ dodges -> barrel roll Mar 07 17:54:34 03Tom Rini  07master * ra0e396fff2 10openembedded.git/recipes/xml-commons/jaxp1.3_1.3.04.bb: Mar 07 17:54:34 jaxp1.3: Switch to subdir= in SRC_URI Mar 07 17:54:34 Signed-off-by: Tom Rini Mar 07 17:54:46 03Tom Rini  07master * r84812c64e7 10openembedded.git/recipes/ckermit/ckermit_211.bb: Mar 07 17:54:47 ckermit: Switch to subdir= in SRC_URI Mar 07 17:54:47 Signed-off-by: Tom Rini Mar 07 17:54:58 03Tom Rini  07master * r1f7fe5429c 10openembedded.git/recipes/ttf-fonts/ttf-arphic-uming_20080216.bb: Mar 07 17:54:58 ttf-arphic-uming: Switch to subdir= in SRC_URI Mar 07 17:54:58 Signed-off-by: Tom Rini Mar 07 17:55:09 03Tom Rini  07master * r70290793d5 10openembedded.git/recipes/zbedic/zbedic_1.2.bb: Mar 07 17:55:09 zbedic: Switch to subdir= in SRC_URI Mar 07 17:55:09 Signed-off-by: Tom Rini Mar 07 17:55:21 03Tom Rini  07master * rb51db30f34 10openembedded.git/recipes/zbedic/libbedic_1.1.bb: Mar 07 17:55:21 zbedic: Switch to subdir= in SRC_URI, add FILESPATHPKG Mar 07 17:55:21 Signed-off-by: Tom Rini Mar 07 17:55:22 03Tom Rini  07master * ra4576b2779 10openembedded.git/recipes/ttf-fonts/otf-junction_02.bb: Mar 07 17:55:22 otf-junction: Switch to subdir= in SRC_URI Mar 07 17:55:22 Signed-off-by: Tom Rini Mar 07 17:55:24 03Tom Rini  07master * rcc14abd3ba 10openembedded.git/recipes/ (3 files in 2 dirs): Mar 07 17:55:24 tzdata/tzcode: Switch to subdir= in SRC_URI Mar 07 17:55:24 Signed-off-by: Tom Rini Mar 07 17:55:26 03Tom Rini  07master * r66c7b423bb 10openembedded.git/recipes/mtd/mtd-utils-native.inc: Mar 07 17:55:26 mtd-utils-native.inc: Drop lzo-native Mar 07 17:55:26 Which lzo is needed is handled by the recipe itself. Mar 07 17:55:26 Signed-off-by: Tom Rini Mar 07 17:55:31 03Tom Rini  07master * r10ce4c0359 10openembedded.git/recipes/v4l2apps/v4l2apps_20020317.bb: Mar 07 17:55:31 v4l2apps: Switch to subdir= in SRC_URI Mar 07 17:55:31 Signed-off-by: Tom Rini Mar 07 17:55:33 03Tom Rini  07master * r69812cb4e7 10openembedded.git/recipes/zrally/zrally_0.90.bb: Mar 07 17:55:33 zrally: Switch to subdir= in SRC_URI Mar 07 17:55:33 Signed-off-by: Tom Rini Mar 07 17:55:34 03Tom Rini  07master * rfe23104045 10openembedded.git/recipes/woodstox/woodstox2_2.0.6.bb: Mar 07 17:55:34 woodstox2: Switch to subdir= in SRC_URI Mar 07 17:55:34 Signed-off-by: Tom Rini Mar 07 17:55:45 03Tom Rini  07master * ref1d9f0daf 10openembedded.git/recipes/zaurus-utils/zaurus-legacy-tar.bb: Mar 07 17:55:46 zaurus-legacy-tar: Switch to subdir= in SRC_URI Mar 07 17:55:46 Signed-off-by: Tom Rini Mar 07 17:57:13 pb__: so the workdir exists in expected path (see above) but it doesn't have Mafefile.am Mar 07 18:19:56 some autoconf tarballs <= 2.13 lack Makefile.am Mar 07 19:35:11 request for comments: http://ix.io/1wu Mar 07 19:51:21 03Koen Kooi  07org.openembedded.dev * r39a3744df0 10openembedded.git/recipes/gstreamer/ (4 files): Mar 07 19:51:21 gst-plugins: remove last traces of gnome-vfs Mar 07 19:51:21 Signed-off-by: Koen Kooi Mar 07 19:52:14 jedahan: I don't understand the role of poky in this text. is it some example ? Mar 07 19:52:32 hasn't that always been the confusion? Mar 07 19:53:22 Thought poky role was always clear. A project using OE to provide a stable baseline? Mar 07 19:54:11 the poky discription should say that then. Mar 07 19:54:24 03Tom Rini  07master * rc0aeb60414 10openembedded.git/recipes/util-linux-ng/util-linux-ng.inc: Mar 07 19:54:25 util-linux-ng: Drop lzo-native from DEPENDS Mar 07 19:54:25 After talking with Denys (who pushed the change before) we agree Mar 07 19:54:25 it's not needed now. Mar 07 19:54:25 Signed-off-by: Tom Rini Mar 07 19:56:29 I'm not sure I like calling bitbake a "package manager" as people not familiar with the Gentoo paradigm won't get that. Mar 07 19:58:17 jedahan: lots of instances of "bikeshedding", which is BSD-ism - http://en.wiktionary.org/wiki/bikeshedding Mar 07 19:58:20 :) Mar 07 20:22:08 denix, right I mentioned that I think a few times...make the names whatever is necessary for historical/legal reasons, but the key is splitting poky out of oe-core to make oe-core just another layer Mar 07 20:22:29 jedahan: I think, that what you propose is bikeshedding... Mar 07 20:22:32 hiho Mar 07 20:22:34 ka6sox, I think people are smart enough to learn Mar 07 20:22:37 hi woglinde Mar 07 20:23:03 ynezz, splitting poky from oe-core and making oe-core another layer is bikeshedding? Mar 07 20:23:24 i mean they seem to be two different projects Mar 07 20:24:21 jedahan: yep, they agreed on it Mar 07 20:24:38 jedahan, you say "package manager to a debian/ubuntu person and they think apt/adept/aptitude, same thing for fedora/centos? RPM things, Yum for other distros. Mar 07 20:25:03 ka6sox, bitbake is a package manager though...it does exactly what apt/adept/aptitude does... Mar 07 20:25:55 package manager, uh? Mar 07 20:26:24 then what does that make opkg? Mar 07 20:26:25 ynezz, so poky is a parent project to oe-core, or poky is a set of tools that bundle/depend on oe-core? Mar 07 20:26:25 you miss something, don't you? Mar 07 20:26:35 opkg is akin to dpkg Mar 07 20:26:59 Yeah it does seem like I am missing a lot, hence why I am trying to learn :) Mar 07 20:27:35 bitbake calculates dependencies from bb files, does it not? and then builds the necessary environment to create the packages you want? Mar 07 20:27:40 bitbake is not package manager, if you need some analogy, call it package builder, if you like, but it doesn't manage any package Mar 07 20:28:02 I don't know book definition, but for me it's task executor Mar 07 20:28:11 ahh so like bash? Mar 07 20:28:56 no more akin to wanna build and other automated package building systems. Mar 07 20:29:02 jedahan: http://bitbake.berlios.de/manual/ch01.html Mar 07 20:29:41 task executor is pretty vague, but i think we are dipping dangerously close to bikeshedding. what it is called matters little, as long as people understand what it does Mar 07 20:29:42 it can be used without yoctoepoky Mar 07 20:30:17 yeah! Mar 07 20:30:28 don't worry, you're not the only confused :) Mar 07 20:30:30 and bitbake can use oe-core without poky, right? Mar 07 20:30:55 it better :D Mar 07 20:30:58 i mean, if you split oe-core to not bundle poky, and made it in the same format as other meta- layers Mar 07 20:31:00 jedahan: there is some documentation that can be helpful if you new to bb and oe Mar 07 20:31:11 he's new to oe-core Mar 07 20:31:12 bb is an independent tool Mar 07 20:31:15 oe just uses it Mar 07 20:31:20 jedahan, bitbake builds packages etc.. apt/rpm/... don't Mar 07 20:31:24 oe is its own thing Mar 07 20:31:40 emerge builds and manages packages, hence its a package manager Mar 07 20:31:55 if bb is just like automake or make or redo then yeah, its a build system, not a package manager Mar 07 20:31:56 oe can be layered Mar 07 20:32:15 yes I get that Mar 07 20:32:25 what I dont get is why oe-core bundles poky Mar 07 20:32:33 so taking advantage of this oe-core is small set of recipes which will serve as basic building layer for rest of oe based projects Mar 07 20:32:48 like yocto, angstrom etc. Mar 07 20:32:50 yes that makes sense Mar 07 20:32:58 well wait, yocto is different than angstrom Mar 07 20:33:03 yes Mar 07 20:33:05 in that yocto aims to be like freedesktop.org Mar 07 20:33:09 angstrom is a distro Mar 07 20:33:14 layer Mar 07 20:33:15 lumping them together can confuse Mar 07 20:33:25 yocto is/will be based on oe-core and so will angstrom Mar 07 20:33:30 in coming future Mar 07 20:33:39 thats what I was trying to divine Mar 07 20:34:06 essentially there will be lot of layers Mar 07 20:34:08 I am trying to use as upstreamy-as possible build system, migrating away from angstrom Mar 07 20:34:12 perfect! thats how it should be Mar 07 20:34:20 and you can pick and choose to build your distro using what you want Mar 07 20:34:21 o.O? Mar 07 20:34:27 oe-core will be essential part Mar 07 20:34:35 jedahan: btw, why are you migrating from angstrom? :) Mar 07 20:34:53 because I prefer rolling release distro, and staying as close to oe-core as possible Mar 07 20:35:14 mainly, because I just started working for a company that has not upgraded from 2008.1 because it was too much of a pita Mar 07 20:35:17 hm I am migrating from angstroem by edit local.conf and add micro-uclibc as distro Mar 07 20:35:30 jedahan: oe-core will not be a distro Mar 07 20:35:37 i am not looking for a distro Mar 07 20:35:44 jedahan: its fine if you have your own distro based on oe-core Mar 07 20:35:49 khem, the targets available in oe-core are good enough for a distro Mar 07 20:36:00 but you have to do the same work that angstrom or yocto will be doing Mar 07 20:36:12 it will only have qemu targets Mar 07 20:36:31 we have to manage our own bsp for our board anyway Mar 07 20:36:39 as of now its evolving Mar 07 20:36:47 jedahan: thats correct Mar 07 20:36:55 jedahan: you need your bsp layer Mar 07 20:36:56 so no reason to rely on another layer Mar 07 20:37:03 however if you need extra packages Mar 07 20:37:15 we pull them from openembedded.git Mar 07 20:37:18 then you have to overlay other layers Mar 07 20:37:27 yes Mar 07 20:37:49 reduces developer cycles Mar 07 20:38:16 oe as of now will provide layers of different sort Mar 07 20:38:17 I guess I come from a distro where having many, many package repositories and building from scratch is the normal mode of working Mar 07 20:38:25 yes Mar 07 20:38:33 khem, how so ? I only see oe-core and openembedded Mar 07 20:38:43 and of course the intel bsp, of which I am basing our bsp Mar 07 20:39:04 there is a nicer layering feature of bb that oe architecture is trying to leavarage so you dont have to have a graveyard of packages Mar 07 20:39:21 jedahan: there is oe-meta Mar 07 20:39:34 ahh yes! i had some questions about that. you mean meta-oe , correct? Mar 07 20:39:37 and there will be more espeically BSP layers and probably arch layers Mar 07 20:39:43 yes Mar 07 20:39:45 meta-oe Mar 07 20:39:53 thats a beginning Mar 07 20:39:56 ok, so meta-oe looks slightly unmaintained Mar 07 20:39:57 yes Mar 07 20:40:03 I know Mar 07 20:40:08 it will not remain so though Mar 07 20:40:12 it's like 2 weeks old? :p Mar 07 20:40:15 ahh that is great news Mar 07 20:40:39 so, meta-oe will soon be what openembedded.git is? Mar 07 20:40:51 so may be different projects who want to add value on top of whats in oe-core can certainly have their own layer Mar 07 20:40:53 or will it be what oe-core.git is, minus poky? Mar 07 20:41:09 poky is another distro Mar 07 20:41:11 like angstrom Mar 07 20:41:14 so yes Mar 07 20:41:24 I thought poky was a set of build scripts that then call bitbake Mar 07 20:41:27 poky is now yocto Mar 07 20:41:30 and yocto was the distro Mar 07 20:41:33 ohhhhhhhhhhhh Mar 07 20:41:41 woooowwowow Mar 07 20:41:58 ok that was not clear, even though I had both the yocto and poky guides open Mar 07 20:42:04 and realized they were functionally the same Mar 07 20:42:09 well yocto is a project which started off from poky as base Mar 07 20:42:20 except yocto claims they want to be freedesktop.org Mar 07 20:42:26 as an umbrella organization Mar 07 20:42:38 with poky being one of thier projects Mar 07 20:42:46 so which is it? Mar 07 20:43:27 Is yocto just a rename of poky, or is it a suite of tools to develop on embedded devices, of which they use poky, oe-core, and other stuff? Mar 07 20:43:27 I will not speak on behalf of yocto as I may misrepresent what they do Mar 07 20:43:32 ok Mar 07 20:43:36 but from oe pov things are clear I suppose Mar 07 20:44:03 jedahan: yocto will use oe-core in future Mar 07 20:44:19 and oe-core will always bundle poky? Mar 07 20:44:28 jedahan: and so will oe distros like SHR micro minimal angstrom jlime kaeilos etc. Mar 07 20:44:47 will oe-core.git continue to bundle the poky build scripts? Mar 07 20:44:48 no oe core will not bundle poky Mar 07 20:45:16 build scripts etc. we are working through the oe-core designs ang goals Mar 07 20:45:21 everyday Mar 07 20:45:27 I do not have answer for you now Mar 07 20:45:45 so bitbake, on its own, will be able to build a very minimal image just by telling it to look at oe-core, which will essentially be a layer Mar 07 20:45:56 ahh Mar 07 20:45:58 ok thats fair Mar 07 20:46:03 it is nebulous Mar 07 20:46:34 jedahan: yes oe-core is like a layer Mar 07 20:46:35 nebulous is nice word :p Mar 07 20:46:59 I mean, for this situation Mar 07 20:47:06 it will have certain references but no distro policies as such Mar 07 20:47:17 ok, so what I 'proposed' in my request for comment, is what seems to be happening, though my wording may have been suboptimal Mar 07 20:47:33 that makes me feel better about the work I am doing right now Mar 07 20:47:56 glad to hear Mar 07 20:47:59 it will drag in gstreamer if bluez package seperating isnt made Mar 07 21:23:41 uh now its out Mar 07 21:23:44 nokia sold qt Mar 07 21:26:02 don't spread fud :) Mar 07 21:26:13 ? Mar 07 21:26:24 ynezz do you understand german? Mar 07 21:26:40 http://www.heise.de/newsticker/meldung/Nokia-trennt-sich-vom-Qt-Geschaeft-1203255.html Mar 07 21:26:42 nope Mar 07 21:26:53 ah its english too Mar 07 21:26:53 http://blog.qt.nokia.com/2011/03/07/nokia-and-digia-working-together/ Mar 07 21:28:06 They didn't actually sell Qt, just the licensing rights. Mar 07 21:28:16 yep, this is my understanding also Mar 07 22:39:52 Tartarus: ping Mar 07 22:57:59 03Chris Larson  07master * rbdd10e9b35 10bitbake.git/lib/bb/ui/goggle.py: Mar 07 22:57:59 goggle: exit quietly on ^C Mar 07 22:57:59 Signed-off-by: Chris Larson Mar 07 23:04:44 03Chris Larson  07master * r43dddb2f96 10bitbake.git/lib/bb/ui/crumbs/hobeventhandler.py: Mar 07 23:04:44 hob: adjust for master's eventHandler API Mar 07 23:04:44 Signed-off-by: Chris Larson Mar 07 23:23:23 kergoth: ping Mar 07 23:28:27 denix: pong Mar 07 23:42:12 kergoth: sorry. quick question on subdir= in SRC_URI - I see it places files in subdir in $WORKDIR, but what about $DL_DIR? Mar 07 23:44:07 hmm, think its for unpacking, not fetching, but would be best off checking the code to confirm. Mar 07 23:47:19 Yeah Mar 07 23:47:27 It's for fixing the unpack of stuff that doesn't normally unpack into a subdir Mar 07 23:47:29 yeah, I think that's right - I was checking bb/fetch code and base.bbclass Mar 07 23:47:38 ie foosrc.tar.gz just has Makefile rather tha foosrc/Makefile Mar 07 23:48:15 * kergoth nods Mar 07 23:48:23 does it matter? it goes to $WORKDIR, which is $BPN-$PV anyway... Mar 07 23:48:26 the fetching code should probably be a bit smarter about isolation Mar 07 23:48:30 rather than relying on a single flat namespace Mar 07 23:48:31 denix, well, that's it Mar 07 23:48:39 The second half of those patches is dropping S=WORKDIR Mar 07 23:48:48 So that the normal S where it's a subdir of WORKDIR happens Mar 07 23:48:54 s/happens/just works/ Mar 07 23:49:04 also consider cases with multiple tarballs Mar 07 23:49:09 hmm, I was hoping it also handles fetch isolation... :( Mar 07 23:49:18 where you need to unpack something into a subdir to use an embedded version rather than installed Mar 07 23:49:24 sadly no Mar 07 23:49:37 things like multiple versions of a recipe, or multiple recipes use media-files.tar.gz... Mar 07 23:50:15 i.e. when the filename is not unique - any plans for that? Mar 07 23:50:21 It'd be cool to do something with our knowledge of the sums to do better isolation Mar 07 23:50:37 ignore the name, we know what's what by the md5sum/sha1sum we define.. Mar 07 23:50:39 * kergoth shrugs Mar 07 23:52:02 yeah, as currently fetch fails with wrong checksums, when files collide in $DL_DIR... Mar 07 23:52:53 well, maybe in other life... :) Mar 07 23:53:08 well, fetch2 is under way, so maybe poke RP to add another requirement ;) Mar 07 23:58:35 the whole .md5 thing needs to DIAF Mar 08 00:04:53 kergoth, what would replace it? Mar 08 00:04:56 or nothing. Mar 08 00:05:11 Possibly nothing, yeah Mar 08 00:05:19 We need to checksum what we dl Mar 08 00:05:26 and then we have the good ones in the metadata Mar 08 00:05:34 So that's just being used for other purposes i guess, heh Mar 08 00:08:19 yeah, currently the checksumming is in do_fetch rather than the fetcher, but there's no reason we can't pass it into the fetcher and let it do the comparison where appropriate. currently the .md5 is only used as a stamp to indicate that the file was fetched by a fetcher through until completion, without interruption -- if we still need that, fine, but the name needs to reflect what it is, in that case Mar 08 00:08:36 it's not at all intuitive right now Mar 08 00:09:44 ah, you mean filename.md5 in $DL_DIR, not MD5 hashing in general... Mar 08 00:09:48 yep Mar 08 00:10:12 that was intended to be used fro some sort of hashing, but it never was, not really Mar 08 00:10:17 I think it now can even be empty, as it's used as a d/l stamp only... Mar 08 00:10:17 just a remnant Mar 08 00:10:21 yep Mar 08 00:11:14 *if* we need it, and i'm not sure if we do -- its possible we need a stamp to differentiate between "complete, but corrupt" vs "incomplete" -- but, its possible that case could be handled by attempting to resume, and if that fails, abort as corrupt Mar 08 00:11:15 * kergoth shrugs Mar 08 00:11:26 if we need it, make it .completed or something :) **** ENDING LOGGING AT Tue Mar 08 02:59:57 2011