**** BEGIN LOGGING AT Thu Mar 03 02:59:58 2011 Mar 03 07:24:43 03Martin Jansa  07master * r6536f01865 10openembedded.git/conf/ (5 files in 2 dirs): Mar 03 07:24:43 sane-toolchain-*: define DEBUG_FLAGS and add it to FULL_OPTIMIZATION to make -dbg packages more usefull Mar 03 07:24:43 Acked-by: Tom Rini Mar 03 07:24:43 Acked-by: Khem Raj Mar 03 07:24:43 Signed-off-by: Martin Jansa Mar 03 09:21:24 morning all Mar 03 09:22:36 am writing a comprehensive response to the top comment on http://lwn.net/Articles/430614/ Mar 03 09:25:24 seems man tried OE and was very frustrated because this is not what he wants.. Mar 03 09:26:48 heh.. nice collection of wrong opinions Mar 03 09:38:40 good morning Mar 03 09:39:05 Jay7: I hear this complaint about OE taking a long time and chewing up space every time an article comes out about OE Mar 03 09:39:45 that just different expectations Mar 03 09:39:57 people want something to build one package or image Mar 03 09:40:10 but OE is more Mar 03 09:40:15 given that we have an option to mitigate it, other build systems can't be much better with the same level of flexibility, and space is damn cheap I can't really understand it Mar 03 09:40:50 Jay7: but some claims are objectively correct Mar 03 09:40:52 so, we should have at least good docs Mar 03 09:42:08 Jay7: I wonder why they have never been examined by the TSC Mar 03 09:42:24 mckoan: ask TSC right now :) Mar 03 09:42:37 Jay7: LOL Mar 03 09:42:40 may be good idea to raise this in our ML Mar 03 09:43:00 Jay7: now it's worse than ever Mar 03 09:43:19 mckoan: imagine right after merging with Yocto Mar 03 09:43:37 old docs already obsoleted but no new docs :) Mar 03 09:43:37 Jay7: Unfortunately, I know Mar 03 09:43:56 at least we have now 'getting started with oe-core' Mar 03 09:44:43 Jay7: it is sufficient to see the traffic on IRC to figure out where we are going Mar 03 10:16:54 good morning Mar 03 10:17:10 ericben: hi Mar 03 10:17:40 hi mckoan Mar 03 10:18:09 lwn comments are funny. I saw the same kind of coments last week on a new on a french site annoucing buidlroot Mar 03 10:19:39 mickey|office: good morning Mar 03 10:20:12 what makes the difference is the "hmi" with make menuconfig which make people believe they have control over what they are doing :-) Mar 03 10:21:27 khem: patchwork still doesn't take my patches. http://lists.linuxtogo.org/pipermail/openembedded-devel/2011-March/030613.html is not present in patchwork Mar 03 10:22:17 ericben: the problem is that nobody tries to understand the feelings of a people trying OE for the first time Mar 03 10:22:51 ericben: I still remember it: frustration (more than with other tools) Mar 03 10:23:39 I don't think it's true that nobody tries to understand their feelings. It is true that nobody is doing much to improve their experience but this is at least partly because it is quite a hard problem to solve. Mar 03 10:23:40 ericben: and I know a lot of people who dropper OE after a few time because of these 'usual' problems Mar 03 10:24:56 pb_: IMHO nobody cares about that because all of us already know OE enough to use it proficiently (almost all of us) Mar 03 10:25:36 good morning pb_ Mar 03 10:27:11 mckoan: well, I guess it depends what you mean by "cares". I am a fairly proficient OE user and I still find it deeply annoying that you have to jump through so many hoops to get it to work on a new machine from scratch. But I haven't been able to think of any straightforward way of improving that. Mar 03 10:28:36 pb_: this sentence raises the actual problem, we all easily see the 'internal' point of view, but not who newbies see it from 'external' Mar 03 10:28:57 a (HUGE) black box Mar 03 10:30:29 I have always preached to provide a reduced set of configurations (BSP) stable and always working. Mar 03 10:30:53 For those approaches for the first time OE is frustrating to fail all the early build. Mar 03 10:31:09 ~hail translate.google.it Mar 03 10:31:10 * ibot bows down to translate.google.it and chants, "I'M NOT WORTHY!!" Mar 03 10:33:00 How do I get all "-dbg" packages installd into the rootfs image Mar 03 10:34:48 * mckoan goes to the soldering desk ;-) Mar 03 10:36:13 * hrw fights with cross gcj under ubuntu Mar 03 10:36:50 mckoan: sorry my line was cut "short circuit" on the phone line ;-) I'm now reading the archives Mar 03 10:39:44 03Eric Bénard  07org.openembedded.dev * r64503bd9bc 10openembedded.git/recipes/busybox/busybox_1.18.3.bb: Mar 03 10:39:44 busybox-1.18.3: add last fix Mar 03 10:39:44 Signed-off-by: Eric Bénard Mar 03 10:39:44 Acked-by: Khem Raj Mar 03 10:44:52 any idea what could be causing this in oe-core? ERROR: Error parsing /OE/shr-core/meta-shr/recipes-efl/efl/eeze_svn.bb: cannot concatenate 'str' and 'dict' objects Mar 03 10:45:15 SRC_URI = "${E_SVN}/trunk;module=${SRCNAME};proto=http;scmdata=keep" Mar 03 10:45:22 and E_SVN is added to shr.conf Mar 03 10:45:22 mckoan: adding a machine to other build systems is not always an easy task. check for doc for ltib, ptxdist or even buildroot ... at least when there is some doc detailing this. Mar 03 10:46:16 adding machine into OE can be done in 1 hour (+ build time) if you have working kernel tree Mar 03 10:46:25 hrw : true Mar 03 10:47:35 and people using buldroot and giving critic on OE should think twice as the goal of buildroot for future release is to provide feature already existing in oe : package generation, external toolchain support, sdk creation ... Mar 03 10:47:53 ericben: adding a machine is an andvanced feature Mar 03 10:48:16 ericben: buildroot supports external toolchains Mar 03 10:48:33 ericben: crosstool-ng support is integrated too Mar 03 10:48:36 hrw: +1 Mar 03 10:48:49 instead of opposing buildsystems, it would be great that build system work together, especially for cross compiling patches ... there was an attemps by ptx Mar 03 10:49:15 03Pau Espin Pedrol  07master * r569c6db89e 10openembedded.git/recipes/openmoko-3rdparty/emtooth2_svn.bb: Mar 03 10:49:15 emtooth2: bump SRCREV Mar 03 10:49:15 Signed-off-by: Pau Espin Pedrol Mar 03 10:49:15 Signed-off-by: Martin Jansa Mar 03 10:49:21 hrw: yes but the feature seems to be stabilizing seeing the work being done on it Mar 03 10:49:29 ericben: and for last two years crossdev ML is nearly empty Mar 03 10:49:42 I should have say "stabilizing and impriving external toolchain support" Mar 03 10:50:15 hrw: in fact I was talking about this tentative : http://www.send-patches.org/ Mar 03 10:50:54 ericben: which is same Mar 03 10:50:59 ericben: LOL that website has always been the motto of my old friend Schwebel Mar 03 10:51:03 hrw: OK ;-) Mar 03 10:52:15 the idea is good Mar 03 10:52:32 IMHO we are all too technical, we need a political vision too Mar 03 10:53:15 well ... then find political people to get the political vision and break everything :-D Mar 03 10:54:27 ericben: no, I didn't mean that, but a mis of technically-good and generally-useful could help Mar 03 10:54:34 s/mis/mix Mar 03 10:54:52 the root problem is that more and more people expect to build en embedded system without any knowledge only because they aregoing to use linux Mar 03 10:55:18 ericben: for 5/5 configuratons would be a must Mar 03 10:55:26 4/5 conf Mar 03 10:55:43 a sort of 'quick-start' Mar 03 10:56:25 if you see thinkg working you are eager to continue Mar 03 10:57:43 this is my personal way of evaluating a project that uses open-source time for the rhyme: First let's see if something works Mar 03 10:58:12 s/rhyme/first time Mar 03 10:58:33 mckoan: that exists : http://openembedded.org/index.php/Getting_started & http://www.angstrom-distribution.org/simplified-development-setup are good starting points Mar 03 11:00:01 ericben: ok, but usually you are evaluating it at home during your overtime, and you are using your laptop Mar 03 11:00:32 ericben: which would require at least all the night for the first build Mar 03 11:00:49 ericben: so the external toolchain is a must Mar 03 11:02:01 moreover that getting started is focused on beagleboard (favourite TI people board) Mar 03 11:02:26 I meant 4/5 generic configurations Mar 03 11:02:26 mckoan: maybe but that adds complexity. I think the added value of OE is building something from scratch without lot of dependencies. Mar 03 11:03:00 and a getting started focused on beagleboard is not a bad thing considering the popularity of this board and its small price. Mar 03 11:03:03 sstate support + external repo with sstate packages may help Mar 03 11:03:04 ericben: yes of course I love it and always wanted it (since I used PTXdist), but is not good for starting Mar 03 11:03:38 and also the amount of work done by Ti & Koen to have a good board support for this board Mar 03 11:04:08 mckoan: for starting narcissus + already build sdk is a good thing Mar 03 11:04:25 this allows people to concentrate on there application instead of loosing time building a base system Mar 03 11:04:58 * mckoan is moving to oscilloscope desk for bug hunting Mar 03 11:05:32 ok lunch time it's not every day that a semiconductor vendor invite us to restaurant ;-) Mar 03 11:06:35 no wonder, I've always thought, that you french only eat, drink wine and strike all the time... Mar 03 11:06:42 :) Mar 03 11:07:32 ynezz: btw, I'm going to add testbuilder.cond.d directory support :) Mar 03 11:07:47 I have about 5 build configs already.. starting hard to manage in one file Mar 03 11:07:47 hm, what's that? Mar 03 11:08:00 . testbuilder.conf.d/*.conf :) Mar 03 11:08:03 ah Mar 03 11:09:35 earlier, there was a checksum.ini file which consisted all the checksums of all packages, no the file is empty. Mar 03 11:10:03 yes, because the checksums moved to the corresponding recipes Mar 03 11:10:04 rsv: we moved it to recipes Mar 03 11:11:13 okay, so if i need a old package, i just need to change the checksum Mar 03 11:11:18 in the reciepe Mar 03 11:11:28 yep Mar 03 11:12:19 okay, i have not worked on oe for a long time. looks like i need to brush up Mar 03 11:13:34 hrw: then why is the checksum.ini exists Mar 03 11:13:46 i mean empty checksum.ini Mar 03 11:13:58 no one dropped it? Mar 03 11:15:13 someone remember who Junqian Gordon Xu is? Mar 03 11:20:12 hrw: xjqian ? Mar 03 11:21:02 haven't seen him around much lately Mar 03 11:22:09 can i send a patch to the OE mailing list to remove it? Mar 03 11:22:20 bluelightning: thx Mar 03 11:22:25 rsv: yes Mar 03 11:47:14 I'd like to get some advice how to solve an issue we're needing to deal with. We're using git submodule to "control" our layer and OE submodules and this way allowing us to have a reproducable environment without using a single. Mar 03 11:47:49 The problem is that we need to have an way to get the git tag of supermodule. Mar 03 12:06:01 otavio, your git foo is better than mine :) Mar 03 12:06:19 Crofton|work: not sure about it heh Mar 03 12:49:24 huhu Mar 03 13:43:24 03Martin Jansa  07master * r270f5d7144 10openembedded.git/recipes/ (7 files in 5 dirs): Mar 03 13:43:24 recipes: don't use weak assignment for LICENSE Mar 03 13:43:24 * while trying openembedded-core I've noticed that such LICENSE field is ignored anyways Mar 03 13:43:24 * it doesn't make much sense Mar 03 13:43:24 Signed-off-by: Martin Jansa Mar 03 13:43:30 03Martin Jansa  07master * r414823a3c0 10openembedded.git/recipes/ (5 files in 3 dirs): Mar 03 13:43:31 recipes: drop unused SHR_RELEASE field Mar 03 13:43:31 Signed-off-by: Martin Jansa Mar 03 13:43:34 03Martin Jansa  07master * r4dd61042a7 10openembedded.git/recipes/ (11 files in 4 dirs): Mar 03 13:43:34 recipes: s/LICENCE/LICENSE/g Mar 03 13:43:34 * while trying openembedded-core I've noticed that some recipes have LICENCE instead LICENSE which is checked in oe-core, just rename Mar 03 13:43:34 Signed-off-by: Martin Jansa Mar 03 13:50:52 hi otavio : what do you mean by "git tag of supermodule" ? Mar 03 13:51:48 ynezz: the most difficult here in Bordeaux is to find time to work between meals & wine tastery ;-) Mar 03 13:55:10 rofl Mar 03 14:26:17 03Chris Larson  07master * r0d732b3b54 10bitbake.git/lib/bb/server/process.py: (log message trimmed) Mar 03 14:26:17 server: use local fixed _bootstrap when appropriate Mar 03 14:26:17 When running on python versions 2.6.0 through 2.6.2, we use a local copy Mar 03 14:26:17 of the python 2.6.6 _bootstrap method of Process, to ensure that we have Mar 03 14:26:17 the fix for http://bugs.python.org/issue5313. This avoids the "hang" of Mar 03 14:26:17 the bitbake process at 0% progress during the parsing on older distros Mar 03 14:26:18 like Fedora 12. Mar 03 14:26:25 03Chris Larson  07master * r898f557cbd 10bitbake.git/bin/bitbake-layers: Mar 03 14:26:25 bitbake-layers: drop 2.6 from #!, per Joshua Lock Mar 03 14:26:25 Signed-off-by: Chris Larson Mar 03 14:26:46 ah, kergoth Mar 03 14:26:56 catch you :) Mar 03 14:27:13 heh :) Mar 03 14:27:36 I wish to say that trying to bitbake something with wrong MACHINE produces unreadable trace for user :) Mar 03 14:27:40 if anyone can test that hang fix on fedora 12 or another distro with the affected python version, I'd appreciate it. it worked for me, but.. Mar 03 14:27:58 something about INVALID machine Mar 03 14:28:16 Jay7: ah, yeah, i've seen that.. its due to our use of "${MACHINE}" in a python snippet in OVERRIDES, i think, if you're referring to the one with the python error about trying to find a matching ' or somesuch Mar 03 14:28:21 i think.. Mar 03 14:28:24 * kergoth_ tests Mar 03 14:28:43 right, yeah, EOL while scanning string literal, ugly indeed Mar 03 14:28:59 I've just made mistype in my MACHINE and got it :) Mar 03 14:29:20 we could avoid it by having a fallback Mar 03 14:29:32 MACHINE = "INVALID" in bitbake.conf or unknown or somesuch Mar 03 14:29:39 or we could change how we add machine to overrides / filespath Mar 03 14:29:43 hmmm Mar 03 14:31:22 hmm, bitbake-layers shows more deprecation warnings than bitbake, i forgot the warnings stuff i do there isn't in a common place Mar 03 14:31:26 * kergoth_ wonders what to doa bout that Mar 03 14:37:55 hmm Mar 03 14:38:25 Jay7: if you set, e.g. DISTRO = "unknown" MACHINE = "unknown" in bitbake.conf, it gets past this issue, and then appears to hang when it hits the unknown endianness Mar 03 14:38:27 * kergoth_ rolls eyes Mar 03 14:38:51 bletch Mar 03 14:38:54 an yes, I have seem that about unknown endianiess :) Mar 03 14:38:57 that stuff is so full of suck Mar 03 14:39:32 indeed Mar 03 14:45:37 of course, really, target arch being INVALID should be caught at ConfigParsed time, before the recipes are parsed.. Mar 03 14:45:37 hmm, sanity.bbclass operates at BuildStarted time, not ConfigParsed Mar 03 14:45:37 i could see needing to wait till then, but most of it could likely be done at config parsed Mar 03 14:45:37 s/needing/some things needing/ Mar 03 15:00:38 03Koen Kooi  07org.openembedded.dev * rc8bbbb90cc 10openembedded.git/recipes/sdr/soft66_git.bb: Mar 03 15:00:38 soft66: add git version Mar 03 15:00:38 "Library and tools for Soft66ADD and related SDR radio receivers" Mar 03 15:26:51 Hrm. OE keeps grinding to a halt during boot just after starting portmap. Mar 03 15:31:33 OE does? Mar 03 15:31:43 or an image created with OE :) Mar 03 15:51:41 morning all Mar 03 15:53:10 are there any tutorials for setting up a new oe platform? want to add a stream for my old Transmeta Crusoe based internet appliance Mar 03 15:59:21 Soopaman, I think so Mar 03 15:59:35 look in the manual and in the wiki Mar 03 15:59:40 for adding a new "machine" Mar 03 15:59:45 look for the machine keyword Mar 03 16:04:43 k Mar 03 16:21:27 Soopaman, did you find what you were looking for? Mar 03 16:25:37 kind of but not really Mar 03 16:26:24 hoping to reduce the amount of guess-build iterations Mar 03 16:28:00 Well, you're just treating it as another x86 right? Mar 03 16:28:02 basically you need to add : Mar 03 16:28:08 Look at qemux86 as an example machine.conf Mar 03 16:28:13 *a machine config in conf/machine/ Mar 03 16:28:16 *a kernel Mar 03 16:28:19 Then find openembedded/recipes -name qemux86 -type d Mar 03 16:28:20 that's the minimum Mar 03 16:28:39 then add other configs like xorg.conf or /etc/network/interfaces etc... Mar 03 16:28:42 (Along with a git grep _qemux86 recipes/*/*.{inc,bb}) Mar 03 16:32:29 Soopaman: which device exactly you have? Mar 03 16:33:11 Soopaman: progear may fit your device as it was for my webpad based on Crusoe Mar 03 16:33:47 hrw, I have the Gateway Connected Touchpad Mar 03 16:33:58 lemme see if I can dig up a link Mar 03 16:34:39 Tartarus: I was thinking about that approach, but wasn't exactly sure if it matters that the final build will have to be written to a cf card Mar 03 16:34:45 I see Mar 03 16:35:26 Soopaman: try progear - should give you something working Mar 03 16:35:42 http://www.zdnet.com/news/gateway-connected-touch-pad/299005 Mar 03 16:35:49 k Mar 03 16:37:17 96MB of ram... Mar 03 16:37:30 progear has 128 Mar 03 16:38:02 Soopaman: you will probably have to adapt X11 config a bit but should work Mar 03 16:38:22 Soopaman: I will be interested in informations does backlight control works Mar 03 16:39:35 I have my old x11 configs Mar 03 16:40:13 and the backlight scripts, so hopefully i'll be able to work it back in if it works Mar 03 16:40:23 I get the following errors when compiling qt4-embedded with latest org.openembedded.dev: http://pastebin.com/9H2nYtb5. Mar 03 16:46:27 anr78: qt4e with armv7a ? good luck :-D Mar 03 16:51:35 mckoan: works fine here why should that create trouble ? Mar 03 16:52:44 works fine for me as well (when it builds) Mar 03 16:53:03 so I also wonder why :) Mar 03 16:53:18 ericben: anr78: good to know, thx Mar 03 16:55:09 have a nice rest of the day Mar 03 16:59:59 ericben: do you use 4.7.1? Mar 03 17:00:32 anr78: yes Mar 03 17:01:29 anr78: with angstrom 2010 Mar 03 17:02:06 ericben: hm. I still use angstrom 2008, but have been looking for reasons to switch :) Mar 03 17:03:22 still, 4.7.1 has compiled for me on angstrom 2008 until todays fetch Mar 03 17:09:11 anr78: I'll launch a build from scratch Mar 03 17:09:18 denix, any typos in the gdbserver patch from Ben, before I git am it? :) Mar 03 17:19:06 What is the support for CMake and scons like in OE. Is it on par with autotools? Which build system would you choose for a project that will eventually be built in OE? Mar 03 17:20:47 cmake in general is horrible.. Mar 03 17:21:10 I don't know how well it's supported in oe, but if I was doing a new project -- it would be "regular" make based.. if it's at all complicated -- autoconf based Mar 03 17:21:10 and scons? Mar 03 17:21:23 ok Mar 03 17:21:30 I don't know anything about scons Mar 03 17:21:55 fray, can you edit the wiki now? Mar 03 17:22:04 autotools has one advantage, a heck of a lot of people know how to work with it.. cmake, not a lot of people (in comparison) know how to -- or refuse to.. Mar 03 17:22:17 ka6sox ya I do, but I had a question -- how do I add a new topic to the dev section list?\ Mar 03 17:22:18 I guess a majority of projects are still autotools based.. but I have seen a few now moving to cmake... Apache QPid Mar 03 17:22:25 guyvdb_: I personally hate cmake very much, it's number one on my build systems shitlist :) I did have to create some recipes for scons based projects and it did work Mar 03 17:22:33 fray, which topic oe-core? Mar 03 17:22:36 I really think it's rate.. cmake is really a terrible make system.. Mar 03 17:22:49 ya.. I havn't created it yet.. but thats what I want "OpenEmbedded-Core" Mar 03 17:23:03 'er.. it's rare (cmake that is)... Mar 03 17:23:19 aem, cmake I mean, not scons, although I remember we had a scons based project too long time ago Mar 03 17:23:41 So with being able to be cross compiled with OE a priority... i should stick to autotools Mar 03 17:23:43 03Michael 'Mickey' Lauer  07org.openembedded.dev * re238a7a949 10openembedded.git/recipes/meego-cellular/libcmtspeechdata_git.bb: libcmtspeechdata: new recipe; library for cellular speech data path on n900 Mar 03 17:23:55 03Michael 'Mickey' Lauer  07org.openembedded.dev * r8f4d4e7b5a 10openembedded.git/: Merge branch 'org.openembedded.dev' of git@git.openembedded.net:openembedded into org.openembedded.dev Mar 03 17:24:03 or no helper environment Mar 03 17:24:03 fray, does it need to be on the front page? or can it be under Developers:More? Mar 03 17:24:23 developers:more Mar 03 17:24:31 guyvdb_: as I said, "inherit cmake" did work for me, but I'd go with autotools Mar 03 17:24:51 fray, okay will do in a bit..meeting Mar 03 17:24:54 in the latter case I would at least know what to do if anything does not work out, with cmake I'd be lost Mar 03 17:25:05 thanks.. let me know.. I have an errand in a few minutes, but I should be back in an hour or so Mar 03 17:25:23 Jin^eLD ya, I echo that as well Mar 03 17:25:23 I have a recipe that looks like it is working for me for zeromq.. not much too it... but if there is interest how would i contribute it? Mar 03 17:45:32 ericben: great Mar 03 17:50:12 gm Mar 03 18:24:38 ericben: is your build done yet? gotta run in 5, and I'll check back tomorrow if it's not Mar 03 18:32:59 03Denys Dmytriyenko  07master * ra47c14c83c 10openembedded.git/recipes/meta/external-toolchain-csl.bb: (log message trimmed) Mar 03 18:33:00 external-toolchain-csl: handle packaging of gdbserver based on PREFERRED_PROVIDER Mar 03 18:33:00 Allow packaging independent copy of specific version of gdbserver with Mar 03 18:33:00 external-toolchain-csl by setting PREFERRED_PROVIDER. E.g. for GPLv2 Mar 03 18:33:00 gdbserver, add these lines to your distro/local.conf: Mar 03 18:33:00 PREFERRED_PROVIDER_gdbserver = "gdbserver" Mar 03 18:33:01 PREFERRED_VERSION_gdbserver = "6.6" Mar 03 18:40:03 03Denys Dmytriyenko  07master * rdf59f77a2e 10openembedded.git/conf/distro/include/toolchain-external.inc: Mar 03 18:40:03 toolchain-external.conf: set the default provider for gdbserver Mar 03 18:40:03 Signed-off-by: Denys Dmytriyenko Mar 03 18:42:30 hi , does someone know openmika jvm ? Mar 03 18:53:26 hrw: around? Mar 03 18:53:52 a little confused what to send the DISTRO value of my local.conf Mar 03 18:55:10 trying to make sure I understand what you, Tartarus, GNUtoo|laptop and the website suggest for creating a new machine Mar 03 18:55:41 Soopaman, hi again Mar 03 18:56:10 try to look at similar machine than yours Mar 03 18:56:14 in conf/machine Mar 03 18:56:19 and try to make a kernel recipe Mar 03 18:56:41 basically create conf/machine/foo.conf Mar 03 18:56:53 in the org.openembedded.dev git Mar 03 18:57:00 that you checked out Mar 03 18:57:19 then try to create a kernel recipe for your machine Mar 03 18:57:33 and well...that's the minimum required Mar 03 19:01:56 Soopaman: what does DISTRO have to do with anything? it's completely orthagonal vs MACHINE, they're two entirely different things Mar 03 19:06:28 GNUtoo|laptop: that's where I'm at, just trying to make sure where the DISTRO suggestion from http://openembedded.org/index.php/Getting_started fits in Mar 03 19:06:34 hey kergoth, long time Mar 03 19:07:23 Soopaman, what's your machine already? Mar 03 19:07:47 i made a gctp.conf from hrw's suggestion of progear.conf Mar 03 19:09:52 getting started is for setting up your first build, it's the basics, you might want to try doing that from beginning to end before you get anywhere near creating a machine .conf of your own.. Mar 03 19:10:15 :) Mar 03 19:13:09 it's a transmetta Mar 03 19:13:12 but what is it? Mar 03 19:13:15 a laptop? Mar 03 19:13:18 a pda? Mar 03 19:13:46 ha sorry I didn't realize that he didn't have his build setup Mar 03 19:14:21 Soopaman: make sure you can build for an existing distro and machine before you go any further. i'd suggest qemux86 or qemuarm for the machine Mar 03 19:16:41 GNUtoo|laptop: it's an "internet appliance" (webpad w/o battery), using the transmeta crusoe tm5500 chip (if my memory serves me) Mar 03 19:16:52 ok Mar 03 19:18:52 fray: around ? Mar 03 19:19:04 fray: I can reproduce the rpmbuild coredump Mar 03 19:19:39 khem.. there is a patch in poky-contrib qhe/zypper can you apply that and rebuild RPM -- and retry? (the patch is actually to libpcre) Mar 03 19:19:55 fray: ok pointer ? Mar 03 19:20:12 git://git.pokylinux.org/poky-contrib Mar 03 19:20:28 :) beat me to it Mar 03 19:20:34 (it's the top patch in that tree) Mar 03 19:21:00 I thought the url was yoctolinux Mar 03 19:21:17 pokylinux.org and yoctoproject.org are the same.. Mar 03 19:21:20 there is no yoctolinux.org Mar 03 19:21:36 (well there is, but it's to a holding site -- no content and not part of the yoctoproject) Mar 03 19:22:32 http://git.yoctoproject.org/cgit/cgit.cgi/poky/ Mar 03 19:22:37 is what I am at Mar 03 19:22:50 poky-contrib, not poky Mar 03 19:23:08 http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=qhe/zypper&id=e16bcffe0a3ce527cdd91aa01c21fc7305a2ab99 Mar 03 19:23:12 is one ? Mar 03 19:23:29 ya, thats it.. Mar 03 19:23:46 when yuo apply it.. be sure to rebuild libpcre before rpm.. (native and target) Mar 03 19:23:50 GNUtoo|laptop: Soopaman's machine generally is sort of 'try does it works, recycle if not after 2 days of hacking' Mar 03 19:24:05 why would target be needed ? Mar 03 19:24:15 its dying in rpmbuild on host Mar 03 19:24:32 once you get it working, assuming you want to be able to use rpm on the target.. :P Mar 03 19:24:38 thats patch has patch=1 btw which is deprecated Mar 03 19:24:44 ? Mar 03 19:24:58 then we should fix that as well Mar 03 19:26:00 and patch needs origin information Mar 03 19:26:12 basically something similar to what I did for the rpm patch I have sent Mar 03 19:26:15 thats fine -- this is a test thing Mar 03 19:26:17 did you look at the patch ? Mar 03 19:26:21 is it ok? Mar 03 19:26:32 which patch? Mar 03 19:26:57 (BTW I am in the process of writing up the commit [and patch msg] guidelines) Mar 03 19:27:04 I hope to have them done today adn to the oe-tsc list Mar 03 19:27:30 fray: there is one for rpm I posted to oe-core Mar 03 19:27:55 fixes the build problems I was seeing on ubuntu Mar 03 19:27:57 found it, I'd overlooked it.. it's fine Mar 03 19:28:04 ack it then Mar 03 19:28:30 just did Mar 03 19:28:34 thx Mar 03 19:30:17 I've forwarded it on to rpm-devel@rpm5.org Mar 03 19:30:35 cool Mar 03 19:30:54 ok so libpcre-native is rebuilding lets see Mar 03 19:31:08 ya, based on your symptoms this is likely the issue Mar 03 19:31:19 conflict between pcre and libc regex functions.. Mar 03 19:31:33 hmm likely the issue Mar 03 19:32:03 the back trace you sent me was very similar to the one we'd been tracking on the target side.. the fix resolves the target side issue we had Mar 03 19:32:27 I am using eglibc libc-2.13 on build host Mar 03 19:33:55 ah forgot to clean rpm-native Mar 03 19:34:14 real fix will likely bump the PR which will trigger rpm-native to rebuild... Mar 03 19:36:21 one day bitbake will be able to figure out reverse deps :) Mar 03 19:38:21 well, theoretically the checksumming enhancements take into account versiong and other attributes of depends -- this should trigger rebuilds.. but no PR update (of libpcre) means no rebuild (of RPM).. Mar 03 19:38:52 well PR for libpcre got incremented Mar 03 19:39:12 hmm, that should have triggered rpm to rebuild then.. :( Mar 03 19:39:28 we'll have to bring it up to RP and see if it's implemented the way I thougth it was Mar 03 19:39:47 and are you both using the same bitbake? i'm sure tehre are still some siggen commits that haven't been pulled to master yet Mar 03 19:39:49 ok sounds like it will fix the problem Mar 03 19:40:00 I am always on master Mar 03 19:40:06 ahh that could be it.. Mar 03 19:40:09 I dont use others Mar 03 19:40:27 i think there's a few siggen/stamp commits in particular, which is likely this behavior Mar 03 19:40:32 along with the server changes Mar 03 19:40:36 and thats hould be it for differences Mar 03 19:40:45 maybe a few fetch2 fixes left, not sure Mar 03 19:41:37 fray: seems that patch was what I needed please propose it for oe-core Mar 03 19:41:57 I'm waiting for the final version, then I'll make sure it gets sent to oe-core Mar 03 19:42:08 sure thing Mar 03 19:42:28 so it seems finally I will have a qemu image that I can boot (hopefully) Mar 03 19:44:48 Hi. I'm making a class and I want to call do_fetch and do_unpack from base.bbclass when SRC_URI is defined on the recipes, instead of the ones defined on the class I'm making. How can I do that? Explicitly calling base_do_fetch and base_do_unpack doesn't work. Mar 03 19:50:00 I'm looking at a diff of Poky's bitbake and the upstream bitbake now.. Mar 03 19:50:22 one thing we added was Summary's to each recipe... it doesn't look like that made it upstream Mar 03 19:52:06 fray: one more thing for oe-core patch from yocto which have bugid mentioned should say yocto bugid blah Mar 03 19:52:17 or may be have a link to bugzilla Mar 03 19:52:38 I reckon that might become invalid in future if bugzilla changes Mar 03 19:52:45 its location Mar 03 19:58:29 fray: here is another error its running into http://pastey.net/146728 Mar 03 20:00:21 error: LOOP: well isn't.. there is no error msg there Mar 03 20:00:36 (it's actually a warning of circular dependencies) Mar 03 20:00:54 if rootfs creation failed look above that Mar 03 20:02:25 * fray used to have a patch that removed those messages -- but I'll probably generate a new one that changes them into visible warnings Mar 03 20:02:48 at some point we should likely try to figure out a way to resolve the circular dependencies Mar 03 20:02:51 re Mar 03 20:02:58 fray: http://paste.ubuntu.com/575142/ is log_do_rootfs Mar 03 20:03:07 I also thought about not exporting my classe's functions when SRC_URI is not empty, so base.bbclass' ones would be used, but I could not make such a condition in the class toplevel environment (like `if (not empty SRC_URI) EXPORT_FUNCTIONS do_fetch do_unpack ...') Mar 03 20:04:08 boag Mar 03 20:04:08 the documents are slightly confusing, and i cant see any decent examples of BSPs Mar 03 20:04:10 mario-goulart: I think you have to override the defaults Mar 03 20:04:10 between lines 336 and 337 there should be actual error messages.. Mar 03 20:04:17 I am not sure what to put in layout.conf and what in machine/bug20.conf ... Mar 03 20:04:25 free beer Mar 03 20:04:38 khem: but how to do that conditionally? Mar 03 20:04:42 since we don't have any, my suggestion is to enable debugging on that line.. Mar 03 20:05:22 damn fair parties Mar 03 20:05:37 woglinde: too drunk ? Mar 03 20:05:44 layer.conf seems to just let poky find the layer, why does it have bbfiles defined? Mar 03 20:05:57 but now I know where my taxes go Mar 03 20:05:59 khem on line 308 of classes/package_rpm.bbclass Mar 03 20:06:06 add "-vv" before the --root ... Mar 03 20:06:08 because shouldnt machine/bsp.conf have all the bb file stuff in Mar 03 20:06:18 re run your build.. that should give us a lot more clue as to the failure Mar 03 20:06:19 khem I stopped right Mar 03 20:08:05 fray: that verbose thing should be make available as knob may be ? Mar 03 20:10:12 it really slows things down -- and really shouldn't be needed.. Mar 03 20:10:43 ...but at the top is a RPM="rpm"... it was done this way so we could change it to RPM="rpm -vv" easily.... Mar 03 20:11:15 khem, I'm working with the RPM maintainer -- he doesn't think thats the right fix.. Mar 03 20:11:24 but we're not sure what the correct one is.. Mar 03 20:11:26 fray: http://paste.ubuntu.com/575148/ Mar 03 20:12:28 fray: ok whatever is the right fix is fine since rpmio depended on these missing libs and mtree was using rpmio I thought it was right thing to do Mar 03 20:12:42 the beecrypt/syck should bec oming in via the linkage of librpmmisc Mar 03 20:12:46 thats as much as we know right now Mar 03 20:12:57 all of the librpm* should be including librpmmisc already Mar 03 20:13:02 I suspect this is an "as-needed" issue Mar 03 20:13:22 but if librpmio needs them then it should bring them along I would think would be the right way Mar 03 20:13:37 it does -- when you are not on Ubuntu 11.04-alpha Mar 03 20:13:39 yeah as-needed could be Mar 03 20:14:24 fray: that last paste is the link with verbose enabled Mar 03 20:14:38 ya.. there is no obvious error.. it just "stopped" Mar 03 20:14:55 let me try a build locally and see what I can get.. which target did you try to build? Mar 03 20:14:56 rpm has been PITA for me :) Mar 03 20:15:04 qemuarm Mar 03 20:15:22 I think it knows you are on ubuntu and doesn't want to behave properly.. ;) Mar 03 20:16:14 ARG! I still can't build because it says linux-libc-headers do_install fials Mar 03 20:16:23 is anyone else seeing that behavior with oe-core? Mar 03 20:16:33 fray: whats the error Mar 03 20:16:44 ERROR: Function 'do_install' failed (see /home/mhatle/git/oe-core/build-oe-core/tmp/work/i586-poky-linux/linux-libc-headers-2.6.37.2-r0/temp/log.do_install.20306 for further information) Mar 03 20:16:48 thats the whole error message Mar 03 20:16:54 hmmm Mar 03 20:16:54 another of these mystery failures Mar 03 20:17:06 and is there anything in /home/mhatle/git/oe-core/build-oe-core/tmp/work/i586-poky-linux/linux-libc-headers-2.6.37.2-r0/temp/log.do_install.20306 Mar 03 20:17:29 that -is- the only line in that file Mar 03 20:17:39 hmm Mar 03 20:17:52 but this doesn't fail on my poky system.. so something screwy is going on Mar 03 20:18:06 * fray trys switching out bitbake Mar 03 20:18:13 which bitbake are you using ? Mar 03 20:18:18 thats with master Mar 03 20:18:25 I have master too Mar 03 20:18:50 but I see that you are trying qemux86 prolly Mar 03 20:18:58 yes.. it's what I had Mar 03 20:19:21 try bitbake -DDDD Mar 03 20:19:28 it might give some hints Mar 03 20:19:42 I'll try that next.. it looks like it might be working with teh poky bitbake Mar 03 20:19:50 (but I'm not yet sure) Mar 03 20:26:57 yes it worked for me with Poky's bitbake.. Mar 03 20:27:08 so there is something in the Poky bitbake patches that is causing failures for me.. Mar 03 20:27:13 anyway, back to trying to reproduce the issue Mar 03 20:27:28 (might turn out it's fakeroot/pseudo related.. if so the poky bitbake might be worth trying) Mar 03 20:28:01 if you're using bitbake master, make sure you're still using the oe-core/scripts/bitbake wrapper.. Mar 03 20:28:36 yes, I was/am Mar 03 20:28:51 i was referring more to khem, but good :) Mar 03 20:28:58 I assume Khem is as well, otherwise I'm not sure he would have gotten that far into the build Mar 03 20:28:58 * kergoth_ forgot about it at first :) Mar 03 20:29:24 no I am not using wrapper Mar 03 20:29:30 well I did Mar 03 20:29:38 the wrapper makes sure the whole build runs under pseudo Mar 03 20:29:48 --ahh-- that would explain it! Mar 03 20:30:07 when RPM tries to chroot into the new filesystem it's failing and I suspect asserting -- without displaying an error message.. Mar 03 20:30:10 I was doing it and then I was running into all sort of issues Mar 03 20:30:13 * fray looks at the RPm code Mar 03 20:30:13 lets retry using it Mar 03 20:30:19 khem: but if mario-goulart provides a do_fetch on his class, the base's one won't be used. The problem is that it needs to be done conditionally. Mar 03 20:30:34 hi , does someone know openmika jvm ? Mar 03 20:30:41 khem: in case we have VAR=1 it uses the class do_fetch, otherwise the base's one Mar 03 20:30:58 otavio: then reimplement base ones to have two versions and use other if first one is not there Mar 03 20:31:01 hi , does someone know openmika jvm ? Mar 03 20:31:01 hi , does someone know openmika jvm ? Mar 03 20:31:02 hi , does someone know openmika jvm ? Mar 03 20:31:02 hi , does someone know openmika jvm ? Mar 03 20:31:03 hi , does someone know openmika jvm ? Mar 03 20:31:09 hmm Mar 03 20:31:18 khem: duplicate the code? Mar 03 20:31:22 gosh! Mar 03 20:31:24 JDuke128: don't do that again Mar 03 20:31:40 otavio: is there some reason you can't just call the base one from yours? what are you trying to do? Mar 03 20:31:56 kergoth_: mario-goulart tried and it didn't work. Mar 03 20:31:57 if (chroot(rootDir) == -1) { Mar 03 20:31:57 rpmlog(RPMLOG_ERR, _("Unable to change root directory: %m\n")); Mar 03 20:31:57 rc = -1; Mar 03 20:31:57 goto exit; Mar 03 20:32:07 kergoth_: base_do_fetch failed. mario-goulart am I wrong? Mar 03 20:32:11 it's possible the error message never makes it out.. but I'm not sure why that would be Mar 03 20:32:23 otavio: so make it work. what's the failure? Mar 03 20:32:51 args Mar 03 20:32:55 mario-goulart: can you paste.bin the error? Mar 03 20:32:57 jduek here too Mar 03 20:33:02 kergoth_, otavio: it failed. If that's supposed to work, maybe I made some syntax mistake Mar 03 20:33:25 mario-goulart: oe was *designed* to facilitate overriding tasks and calling the base ones from them Mar 03 20:33:33 so yes, if that doesn't work, something is wrong. Mar 03 20:34:13 (BTW we're trying to figure out a way to determine if pseudo is running or not in a sanity check -- but we need to make sure to only do this on the stage2 build, not stage1.. so far I don't think we've added the check because of that -- in yocto) Mar 03 20:35:45 hmmm that script can only be run if I use poky-init-env Mar 03 20:35:50 since it needs BUILDDIR Mar 03 20:36:02 set BUILDDIR. Mar 03 20:36:04 :P Mar 03 20:36:04 yes, the environment needs to be setup Mar 03 20:36:20 you can choose not to use the poky-init-env -- but you still need the environment configured Mar 03 20:36:43 kergoth_: I see. I'm getting a traceback now. Mar 03 20:37:06 (and for anyone who really doesn't want to use a POSIX shell, I have patches that enache (t)csh support in poky-init-env... but they're pretty nasty) Mar 03 20:37:17 * kergoth_ shudders Mar 03 20:37:19 ...and no I don't use (t)csh, but I have some customers who do) Mar 03 20:37:31 * kergoth_ sticks to zsh (bash if not available) Mar 03 20:37:45 kergoth_: http://paste.lisp.org/display/120180 Mar 03 20:38:00 so we are going to promote pople to use this env Mar 03 20:38:03 poky-init-env works on zsh, bash, ash (dash), and pdksh Mar 03 20:38:12 wow, what the hell. i just used a python ast visitor and hit a recursionerror Mar 03 20:38:15 since I see pseudo needing few env settings Mar 03 20:38:24 the values are needed -- it's up to the end distribution to use an initscript or not.. Mar 03 20:38:25 What I did was making my own_do_fetch() simply call base_do_fetch Mar 03 20:38:31 mario-goulart: base_do_fetch is python, not shell. Mar 03 20:38:35 you can't run python from shell Mar 03 20:38:51 Hmmm! Mar 03 20:39:01 PSEUDO needs to be setup in the ld_preload and ld-Library-path.. the problem is that you can't do that from within bitbake because there is no way to say "I don't have pseudo -- so don't do this.. ok now I have pseudo .. set this and re-exec ourselves" Mar 03 20:39:11 thus the wrapper script Mar 03 20:39:19 so to do what you want, you'd have to define a python do_fetch (though you could have it exec_func a shell funciton to do some of what you want to do, for convenience) Mar 03 20:39:23 kergoth_: so I should make my class' do_fetch in python? Mar 03 20:39:36 if you want to run base_do_fetch, your do_fetch will have to be python, yes Mar 03 20:39:52 kergoth_: ah, ok. Thanks a lot. Mar 03 20:40:02 np Mar 03 20:40:07 (pseudo is loaded in memory at all times, so there is 'fakeroot' like ability for python and shell tasks.. without forcing the python tasks for fork/exec (which would have been required with libfakeroot.. the overhead of the fork/exec vs just fork is very measurable in build performance) Mar 03 20:40:28 we resisted the wrapper as well -- until we found it cut 20+ minutes off a build.. :P Mar 03 20:40:46 fray: I guess we need to document the environment settings Mar 03 20:40:46 fray, how long was the build :) Mar 03 20:41:11 although it would take an excessively long build for me not to notice 20 minutes ... Mar 03 20:41:33 I think a poky-image-sato build was up around 200 minutes on a 4 core 2.9GHz Mar 03 20:41:41 so 20 minutes was roughly 10% Mar 03 20:41:51 oouch Mar 03 20:42:16 we're currently at around 110 minutes for the same build, on the same machine due to other performance improvements Mar 03 20:43:29 this is to fix a bug that AFAIK exists in existing oe.. if you run tasks with python -- there is currently no way to do the fakeroot behaviors.. which can cause incorrect permissions in the resulting target packages Mar 03 20:43:46 our first fix to that was simply allow python functions (non-shell) for fork/exec.. Mar 03 20:43:55 that added a lot of unexpected time to the build.. Mar 03 20:44:20 add pseudo -- and suddenly things took longer because it handles the build interception in a more complete (IMHO) fashion that fakeroot does.. Mar 03 20:45:19 (currently with the wrapper there is a variable PSEUDO_DISABLED=1 that is set in the wrapper.. pseudo just stays in memory, watches for fork/exec calls -- when it sees one, quickly checks the environment.. if that goes away or is set to 0, it then activates... otherwise stays dormant) Mar 03 20:46:55 fray: if I need to use BUILDDIR outside the control of these scripts then I do source ./poky-init-build-env BUIDLDIR=/mydir Mar 03 20:47:11 source ./poky-init-build-env Mar 03 20:48:55 and MACHINE and DISTRO Mar 03 20:49:09 machine and distro are set in the conf/local.conf Mar 03 20:52:58 hmm it copies conf into builddir Mar 03 20:53:14 yes.. there is a sample conf, which gets copied as conf/local.conf Mar 03 20:53:20 I usally am used to have it along with sources Mar 03 20:53:21 then you can add/remove or edit the entries in there Mar 03 20:53:36 thats how you would disable package_rpm and defualt onto to ipk for instance Mar 03 20:53:38 so I can delete the builddit Mar 03 20:53:40 builddir Mar 03 20:54:05 (note rpm is still needed for dependency scanning, even in deb and ipk processing -- but it just doesnt' build RPM packages) Mar 03 20:54:18 what? Mar 03 20:54:25 yeah rpm is must it seems Mar 03 20:54:50 I didnt see rpm building Mar 03 20:55:11 fray: is there some reason to use only rpm Mar 03 20:55:21 there were a number of (run-time) dependencies that didn't appear in the original work.. I added code that used "rpmdeps" program to go in and capture provide and requires information per-file.. Mar 03 20:55:28 this solved some nagging dep issues for us.. Mar 03 20:55:28 I think if someone does not want to use rpm he/she could see it potential let off Mar 03 20:55:48 rpm for the dep scanning was simply the easiest way to do it -- and most reliable I found.. Mar 03 20:55:55 but rpm for the packaging is strictly optional.. Mar 03 20:56:08 I doubt a lot of people doing OE work reallyw ant RPM as the default packaging format Mar 03 20:56:28 (company I work for and our customers do need rpm as the default packaging format for one reason or another) Mar 03 20:56:57 (specifically larger systems seem to want rpm or similar, while smaller systems seem to want something like opkg or no packaging method) Mar 03 20:57:14 fray: thats fine however what I was thinking is if its possible to use say opkg instead to do dep scanning and make it pluggable and if user has no choice then use rpm Mar 03 20:57:39 opkg doesn't do dep scanning.. it relys on the information from OE Mar 03 20:58:19 for details, look in "package.bbclass" search for "rpmdeps" Mar 03 20:59:19 trying to remember the missing details.. I believe it was versioned symbols, specific soname requirements, and something else.. there were three things that were missing.. Mar 03 20:59:26 (Sorry I'm blanking now on exactly what they were..) Mar 03 20:59:48 rpmdeps just uses the same code RPM does to generate dependencies, without using the rest of the rpm packaging system to either package or install.. Mar 03 21:00:20 (the other side effect is we now have -per-file- dependency information available.. I hope to use this in the future to allow for post package generation "repackaging" behavior through an external tool) Mar 03 21:00:45 I think its a good thing no doubt Mar 03 21:01:16 * khem needs food Mar 03 21:01:19 specifically if say package "foo" generates with three binaries "/bin/A" "/bin/B" and "/bin/C".. the user can look at it and understand what each A, B, C have for runtime dependencies and then repackage it and say remove 'B' without having to re-write the recipe and such Mar 03 21:01:38 in otherwords, the customization a lot of people do already Mar 03 21:02:36 gotta run.. sounds like my car is done at the body shop.. bbiab Mar 03 21:51:13 kergoth: I am trying to setup a local git repo which should track an upstream projects Mar 03 21:51:29 kergoth: have you done that before Mar 03 21:51:35 and any experiences to share ? Mar 03 21:56:02 khem: git clone ; git pull; Mar 03 21:56:56 robtaylor: I am looking for shareing this repo and make it writable to local folks Mar 03 21:57:03 at the same time Mar 03 21:57:36 khem: ah, righty, then git clone --bare, somewhere accessible by http Mar 03 21:58:44 khem: do you want yur local toeam to be abel to push to it? Mar 03 21:58:54 yes Mar 03 21:59:10 atleast they should clone/pull it Mar 03 21:59:14 easily Mar 03 21:59:29 and I should be able to take their changes and at the same time changes from upstream Mar 03 22:00:58 khem: are there any port restrictions between the mirror machine and your developers? Mar 03 22:01:04 nope Mar 03 22:02:40 khem: ok so :http://book.git-scm.com/4_setting_up_a_public_repository.html Mar 03 22:03:20 khem: thoughm you probably want to use git clone --mirror , which will mirror all the branches in the public repo Mar 03 22:05:04 khem: then you'll want a cron job running GIT_DIR=/path/to/git/repo git pull --all Mar 03 22:05:47 khem: your team would them be able to create their own branches there for sharing their work, and you can upstream your work in the ususal way Mar 03 22:06:16 khem: though I have one other suggestion that might be worthwhile - just use github? Mar 03 22:06:28 depends how secret the work is, i guess :) Mar 03 22:07:21 robtaylor: well its internal so no github for now Mar 03 22:07:53 robtaylor: git clone --bare --mirror is what i need Mar 03 22:07:55 khem: you can have private github repos Mar 03 22:08:09 khem: no need for --bare, --mirror implies bare :) Mar 03 22:08:17 k Mar 03 22:08:28 and for getting changes from upstream Mar 03 22:08:39 I can just track upstream as remote repo right ? Mar 03 22:08:41 khem: it is possiblyt to set up an internal gitorius - which is quite nice for code review, Mar 03 22:08:47 khem: spot on Mar 03 22:08:59 gerrit Mar 03 22:09:10 khem: ah, that's better :) Mar 03 22:09:49 khem: the main issue with gitis there's so many differnt options on setting up the workflow Mar 03 22:10:05 yes Mar 03 22:10:28 robtaylor: so that page does not explain the part how to setup git access on server Mar 03 22:10:41 khem: tbh, one main q here - is the mirror just for local access speed? Mar 03 22:11:00 robtaylor: speed + local patch history Mar 03 22:11:44 khem: tbh, i'd probably use two repos, just to save mess Mar 03 22:12:01 khem: one as the mirror - for your team to use as their remote/upstream Mar 03 22:12:11 robtaylor: what needs to be done on git server to enable git:// Mar 03 22:12:14 khem: read only, of course Mar 03 22:12:22 khem: and one for dev Mar 03 22:12:33 robtaylor: it will mostly be pull model Mar 03 22:12:39 is there an easy way to test if bitbake has been installed correctly? Mar 03 22:12:47 so devs will clone in and send patches wont push directly Mar 03 22:12:50 khem: well, i'd need more info to talk you through that. what distro? Mar 03 22:12:57 ubuntu Mar 03 22:13:14 do I need git-daemon Mar 03 22:13:29 khem: yep Mar 03 22:13:48 khem: there's a bit of trickyness with permissions etc Mar 03 22:13:57 I have apache running Mar 03 22:14:12 so probably I can hook in cgit too Mar 03 22:14:33 khem: the easiest thing is to make a single git user, and have all your devs put their public keys in /home/git/.ssh/authorized_keys Mar 03 22:14:58 khem: the issue is,I can't really point to a simple how to-guide Mar 03 22:15:32 khem: a bit of googling will give you some useful inputs though. Mar 03 22:15:57 khem: might be gitosis is a good option Mar 03 22:15:58 robtaylor: yeah probably I need gitolite or somesuch Mar 03 22:16:57 khem: sounds like you've got a good handle on it now ;) good luck! Mar 03 22:17:08 thx Mar 03 22:18:42 where does the r100 in uImage-2.6.36-r100-overo.bin come from? Mar 03 22:18:53 is there any way I can track down the cause of an "error: printing the evironment of the function" error? googling and searching the wiki haven't given much help Mar 03 22:18:57 nelim: Its PR Mar 03 22:19:06 probably look in conf/ Mar 03 22:19:11 for omap machines Mar 03 22:19:26 Soopaman1: some python error prolly Mar 03 22:20:19 all I can find is MACHINE_KERNEL_PR = "" Mar 03 22:21:26 look in machine/include/omap3.inc Mar 03 22:22:06 hey khem Mar 03 22:22:09 MACHINE_KERNEL_PR = "r97" Mar 03 22:22:20 Tartarus: howdy Mar 03 22:22:26 If I said I wanted to put --hash=both for -cross/-native stuff, would you kill me? :) Mar 03 22:22:39 RHEL5 defaults to --hash=gnu which doesn't run on RHEL4 Mar 03 22:22:55 EOL it Mar 03 22:23:02 2012 Mar 03 22:23:06 hmm Mar 03 22:23:06 is EOL for RHEL4 Mar 03 22:23:18 hmmm Mar 03 22:23:20 So gcc-cross-sdk for example won't work on RHEL4 w/o that kind of change Mar 03 22:23:22 khem: using python 2.6.6 on ubuntu 10.10 which seems to be kosher in the req page Mar 03 22:23:24 its getting to be like cygwin Mar 03 22:23:30 I'll let you think for a little, heh Mar 03 22:23:38 If we need to carry it forward locally that's fine too Mar 03 22:23:45 Tartarus: there is some penalty Mar 03 22:24:06 and since we already suffer with long build times I would ask you to measure build times Mar 03 22:24:25 with both as well as gnu Mar 03 22:24:51 Does bitbake default to fetching the "master" branch if the branch name given in the recipe doesn't exist? Mar 03 22:25:07 B_Lizzard: I would think so Mar 03 22:25:29 Tartarus: in theory it should not be that bad but measuring is probably right thing Mar 03 22:25:33 Does git or bitbake emit a log? Mar 03 22:25:35 :/ Mar 03 22:25:52 B_Lizzard: there should be log.do_ Mar 03 22:25:56 in your build tree Mar 03 22:26:00 OK Mar 03 22:26:03 khem, you mean =both vs system default? Mar 03 22:26:04 if you have enabled QA_LOG Mar 03 22:26:06 Thanks. Mar 03 22:26:17 Tartarus: =both and =gnu Mar 03 22:26:41 I think it's some issue in the git repo because it stopped working after a rebase or something like that Mar 03 22:26:42 hello, I heard abaout that OE repository will be split in different repositories. Where can I read about It and what will happen with openembedded repo ? Mar 03 22:26:43 Sucky Mar 03 22:27:11 nschle85: there is some email from RP in detail about it Mar 03 22:27:13 on oe ml Mar 03 22:27:29 nschle85: and more details on openembedded-core ml Mar 03 22:27:32 oh, that's good news Mar 03 22:31:47 03Joshua Lock  07master * rc4328e5c5e 10bitbake.git/lib/bb/cooker.py: (log message trimmed) Mar 03 22:31:47 bitbake/cooker: add generateTargetsTree method Mar 03 22:31:47 The generateTargetsTree() command needs to return a model which includes more Mar 03 22:31:47 metadata than the one generated by generateDepTree(). Mar 03 22:31:47 This patch adds a new method generateTargetsTreeData() to the cooker, based Mar 03 22:31:47 on generateDepData(), and switches generateTargetsTree() to use it. Mar 03 22:31:48 (From Poky rev: dcfc5ae7b1f5925587b0675e1cba6c60f098267c) Mar 03 22:31:51 03Joshua Lock  07master * red9178cc40 10bitbake.git/lib/bb/ (command.py cooker.py event.py): (log message trimmed) Mar 03 22:31:51 implement command to get all possible targets and their dependencies Mar 03 22:31:51 Add a new command generateTargetsTree() which returns a dependency tree of Mar 03 22:31:51 possible targets (tasks and recipes) as well as their dependency information. Mar 03 22:31:51 Optional parameter 'klass' also ensures any recipes which inherit the Mar 03 22:31:51 specified class path (i.e. 'classes/image.bbclass') are included in the model Mar 03 22:31:51 (From Poky rev: 1b3eb0c35f504e8f652303a4b238034ecc5c5d02) Mar 03 22:31:53 03Joshua Lock  07master * rb8cfc85ddc 10bitbake.git/lib/bb/cooker.py: (log message trimmed) Mar 03 22:31:53 bitbake/cooker: reduce code duplication Mar 03 22:32:04 introduce crumbs.TaskListModel a gtk.ListStore subclass Mar 03 22:32:04 Provide a gtk.ListStore subclass which includes a function, Mar 03 22:32:11 populate(), which takes as input the data emitted by Mar 03 22:32:11 bb.event.TargetsTreeGenerated and fills the ListStore model Mar 03 22:32:11 appropriately. Mar 03 22:32:11 Furthermore convenience functions are provided by which the caller can Mar 03 22:32:11 03Joshua Lock  07master * r4a2fb1b550 10bitbake.git/lib/bb/event.py: Mar 03 22:32:11 bitbake/event: fix some whitespace issues Mar 03 22:32:11 (From Poky rev: b14cda62d075d1213fc4769aa6b3622a491b57d5) Mar 03 22:32:12 Signed-off-by: Joshua Lock Mar 03 22:32:13 Signed-off-by: Richard Purdie Mar 03 22:32:22 03Joshua Lock  07master * r97c405f0d7 10bitbake.git/lib/bb/ui/crumbs/progress.py: (log message trimmed) Mar 03 22:32:23 bitbake/progress: add method to pulse the progress bar Mar 03 22:32:23 When we're running a long operation with indeterminate duration it's useful Mar 03 22:32:23 to use the gtk.ProgressBar's pulse method to show that something is happening Mar 03 22:32:23 but we don't know how long it will take. Mar 03 22:32:23 (From Poky rev: fb62c54e13e875dd81e0b5220c54a7753b4d5fa2) Mar 03 22:32:24 Signed-off-by: Joshua Lock Mar 03 22:32:35 03Joshua Lock  07master * ref1ba91719 10bitbake.git/lib/bb/cache.py: Mar 03 22:32:35 bitbake/cache: store a list of inherited files in the cache Mar 03 22:32:35 (From Poky rev: 920c402342bd490cd94b365c3e151de735dec0d6) Mar 03 22:32:35 Signed-off-by: Joshua Lock Mar 03 22:32:35 Signed-off-by: Richard Purdie Mar 03 22:32:47 03Joshua Lock  07master * rf491f1fd60 10bitbake.git/lib/bb/cooker.py: (log message trimmed) Mar 03 22:32:47 bitbake/cooker: don't drop possible_world ref count Mar 03 22:32:47 We need this if we want to run the buildWorldTargetList function more than Mar 03 22:32:47 once, for example in a UI where we can change the MACHINE and DISTRO as much Mar 03 22:32:47 as we like before triggering a build. Mar 03 22:32:47 (From Poky rev: c2814caa5d80d3a1097c83b82b7a3f8eeb8da548) Mar 03 22:32:48 Signed-off-by: Joshua Lock Mar 03 22:32:50 03Joshua Lock  07master * rc2ee2b71ec 10bitbake.git/lib/bb/ui/ (crumbs/hobeventhandler.py hob.py): Mar 03 22:32:50 Add new UI hob, a prototype Gtk+ GUI for creating images Mar 03 22:32:50 Hob is a first stab at implementing an interactive GUI for BitBake. Mar 03 22:32:50 (From Poky rev: 6dbceb0be9a1b8d7d5124b4fbd74f18609bc6146) Mar 03 22:32:50 Signed-off-by: Joshua Lock Mar 03 22:32:51 Signed-off-by: Richard Purdie Mar 03 22:32:53 03Joshua Lock  07master * rc72ad4fdeb 10bitbake.git/lib/bb/ui/crumbs/__init__.py: Mar 03 22:32:53 bitbake/crumbs: update documentation header Mar 03 22:32:53 (From Poky rev: 7f8aa691c52547ee5f1272a7931ca9cccd1a120a) Mar 03 22:33:11 03Joshua Lock  07master * r42cfa8e261 10bitbake.git/lib/bb/ui/crumbs/progress.py: Mar 03 22:33:11 bitbake/progress: make progress dialog modal for parent window Mar 03 22:33:11 (From Poky rev: f258cedfe8432d61eebd21a239381e9510be7109) Mar 03 22:33:11 Signed-off-by: Joshua Lock Mar 03 22:33:11 Signed-off-by: Richard Purdie Mar 03 23:01:11 they keep pushing out the ELF for RHEL4.. I wouldn't be surprised if someone pays RH enought to push out the EOL past 2012 Mar 03 23:05:57 heh Mar 03 23:05:59 So I'm not crazy Mar 03 23:06:06 I thought it was already gone, but had to check today Mar 03 23:06:20 original EOL was end of 2010 Mar 03 23:06:38 rumor (I heard) was that some big company paid them a lot of money in a 2 year extension.. Mar 03 23:06:58 I know that's always been possible, but I assumed it would be kept private Mar 03 23:07:35 the rumor was it was a large enough support deal that a stipulation was that they continue public support as well.. (for this companies customers).. and no I have no idea who that company is Mar 03 23:10:38 * ReaperOfSouls wonders how much that extra two years cost. Mar 03 23:10:51 I'm guessing someone's soul (at RH) Mar 03 23:12:17 Well it prolly crushed someone's soul, not sure about cost them. Mar 03 23:12:51 heh Mar 03 23:14:02 goodness, now i'm starting to remember how frustrating bitbake is Mar 03 23:14:29 if not bitbake, just getting it to "work" Mar 03 23:27:41 03Andrea Adami  07org.openembedded.dev * rf54d43fc6d 10openembedded.git/recipes/kexecboot/ (kexecboot-klibc_git.bb kexecboot_git.bb): Mar 03 23:27:42 kexecboot: bump to SRCREV 1464e897e416f7458e93fb30148e87e60509a667 Mar 03 23:27:42 * adds new icons and fonts, logs, text-ui, multiple kernels per partition Mar 03 23:27:42 * bump PR Mar 03 23:27:42 Signed-off-by: Andrea Adami Mar 03 23:27:53 03Andrea Adami  07org.openembedded.dev * re8a78acb1e 10openembedded.git/recipes/kexecboot/kexecboot-cfg_0.1.bb: Mar 03 23:27:53 kexecboot_cfg: update boot.cfg example to sync with kexecboot. Mar 03 23:27:53 * clean some cmdline cruft Mar 03 23:27:53 * bump PR Mar 03 23:27:54 Signed-off-by: Andrea Adami Mar 03 23:27:54 03Andrea Adami  07org.openembedded.dev * r60ede7b913 10openembedded.git/recipes/kexecboot/kexecboot.inc: Mar 03 23:27:55 kexecboot.inc: update configure options. Mar 03 23:27:55 * --enable-evdev-rate default is disabled so readd it to Zaurus machines Mar 03 23:27:56 Signed-off-by: Andrea Adami Mar 03 23:31:17 03Andrea Adami  07org.openembedded.dev * r5da6fe894b 10openembedded.git/recipes/kexecboot/kexecboot.inc: Mar 03 23:31:17 kexecboot.inc: fix typo. Mar 03 23:31:17 Signed-off-by: Andrea Adami Mar 03 23:38:38 03Andrea Adami  07org.openembedded.dev * r688d359b9c 10openembedded.git/recipes/linux/linux-kexecboot.inc: Mar 03 23:38:38 linux-kexecboot.inc: bump PR to follow kexecboot changes. Mar 03 23:38:38 Signed-off-by: Andrea Adami Mar 04 00:16:35 are there any circumstances under which a recipe (say xserver-common) would be built even though it contains DEFAULT_PREFERENCE = "-1" ? Mar 04 00:18:22 neilm: yes if it appears in PREFERRED_VERSION_blah Mar 04 00:19:56 which could appear in any recipe that depends on blah? Mar 04 00:41:13 neilm, no. It could appear in distro conf files or files that they include. Search for blah in openembedded/conf/ Mar 04 00:43:29 ahhh.. distro/include/angstrom-2008-preferred-versions.inc:PREFERRED_VERSION_xserver-common = "1.33" Mar 04 00:43:32 thank you Mar 04 02:29:12 God I love the OE community Mar 04 02:29:43 someone has just sent in a patch that looks like it fixes my system-tools-backends build issue Mar 04 02:29:52 that I have been too dumb to figure out Mar 04 02:43:24 how fast does patchwork update? **** ENDING LOGGING AT Fri Mar 04 02:59:59 2011