**** BEGIN LOGGING AT Thu Dec 17 02:59:58 2009 Dec 17 05:05:16 jesus, we *really* need USE Dec 17 05:06:45 that would be awsome Dec 17 05:07:04 we do have some commonality with gentoo :) Dec 17 07:43:38 good morning Dec 17 07:44:01 gm :) Dec 17 07:50:55 Hi, does anybody compile scim under oe? Dec 17 07:51:52 i got error while compiling scim 1.4.9 with scim bb file download from oe, anybody can help me? Dec 17 08:02:20 leslie: never tried, but pastebin error would help Dec 17 08:12:54 mckoan: thanks man, I have pasted the log.do_compile here: http://pastebin.com/d26ab0a5 Dec 17 08:14:10 i just download scim_1.4.9.bb files and the other dependent files, and do bitbake scim, with no modification, and it got error there Dec 17 08:28:25 leslie: sounds like a missing include Dec 17 08:44:41 morning Dec 17 08:46:22 leslie: hmm, that looks like some C++ that used to be valid but no longer is type error Dec 17 09:02:38 Hey, I'm currently fixing a dependency on a kernel module package by replacing it with RRECOMMENDS. However, someone asked me if I should not use the _${PN} for for RRECOMMENDS. What would be the reason for wanting this? Dec 17 09:09:38 I checked the code, it seems missing some architecture dependency, i'm asking scim developers for what this error means Dec 17 09:18:31 I got another problem, I found the built libraries are so big than old one, when I update my oe tree Dec 17 09:19:13 like same libpng version, the old oe built it for 900k, Dec 17 09:19:44 but the new download oe built it to 4M Dec 17 09:20:10 I think I have missed some settings in oe to strip the libraries? Dec 17 09:20:13 odd sounds like debug symbols arent getting stripped, this is the size inside the ipk file you are talking of? Dec 17 09:20:45 it's the file size in image folder Dec 17 09:20:57 like in qt4-x11-free, I got 900M!! Dec 17 09:21:20 leslie: your looking at the wrong thing Dec 17 09:21:28 leslie: they are stipped before packaging Dec 17 09:21:49 leslie: -ggdb3 is now in the co Dec 17 09:21:52 mpile flags Dec 17 09:21:58 XorA: so it changed since the newest oe? Dec 17 09:22:13 leslie: it changes about 6 months ago, not sure what you mean by newest Dec 17 09:22:23 XorA: oh, I see Dec 17 09:23:23 XorA: so where can i find the stripped files? Dec 17 09:23:58 leslie: inside the .ipk files Dec 17 09:24:11 leslie: possibly in the package* directories Dec 17 09:24:21 Im not sure at which point they get stripped Dec 17 09:32:13 RP: hi Dec 17 09:42:21 This hasnt anything todo with oe but I'll try anyway. I need a way to verify that the upgrades people loads into our application (using an usb stick) comes from us and have not been modified. I need some kind of signing.. Any ideas about that? Dec 17 09:43:06 jovox_: gpg Dec 17 09:43:30 Oki, I will look into that Dec 17 09:43:36 XorA: do you know a way to force oe put stripped files in the image folder? anybody knows? Dec 17 09:43:42 jovox_: sign with you key and have the software verify it Dec 17 09:44:13 jovox_: you can use libgpgme to use gpg from inside a program Dec 17 09:44:48 leslie: no Dec 17 09:44:56 Ok, thanks. I will go thought the docs Dec 17 09:45:30 jovox_: why not look in packages-split? Dec 17 09:45:43 what's that? Dec 17 09:46:06 jovox_: where the packages are built from, from here looks like they are stripped by this stage Dec 17 09:47:13 Are you talking about the packages bb generates? Dec 17 09:49:57 yes Dec 17 09:51:23 jovox_: it seems only in .ipk, the libraries is stripped Dec 17 09:55:44 But this is not only ipk files. It could be any kind of file not created using bb Dec 17 09:57:07 jovox_: sorry, wrong person :-) Dec 17 09:57:18 * XorA is a little hung over Dec 17 10:03:36 heh ok :) Dec 17 10:03:51 gpg is just what I was looking for b.t.w Dec 17 10:06:39 dont feed me anymore rum :-( Dec 17 10:14:19 XorA: been on the RUM again ;) Dec 17 10:15:20 lrg: it was sooooo cheap Dec 17 10:15:25 2.50 for a double Dec 17 10:15:52 anybody is experienced using an Angstrom build without gpm? I am trying a x11 system with minimal Xorg xinit but I'm facing to some problems and I'd like to understand where to dig Dec 17 10:16:11 XorA: hehe, that is cheap Dec 17 10:16:18 mckoan: we use gpm? Dec 17 10:16:28 1. /usr/lib/X11/xinit/xinitrc seems ignored Dec 17 10:16:40 lrg: and the barmaid was bloody hot even before I started drinking :_) Dec 17 10:16:43 XorA: I mean gpm-dm (gpe) Dec 17 10:17:09 mckoan: openmoko-devel-image does Dec 17 10:17:32 2. launching xinit cause an error (EE) FBDEV(0): FBIOBLANK: Invalid argument Dec 17 10:18:03 3. /etc/X11/xorg.conf seems ignored Dec 17 10:18:12 I'm puzzled Dec 17 10:18:30 XorA: I'm taking a look at openmoko thx Dec 17 10:19:26 hi all Dec 17 10:19:37 anyone know who's running the embedded room at FOSDEM? Dec 17 10:19:46 mckoan: Xorg stopped using the config file for most stuff Dec 17 10:23:14 XorA: in you opinion, would be possible to build openmoko-image with a different ARM machine, just to test it? or it fails by sure? Dec 17 10:24:32 mickey|office: good morning! Dec 17 10:24:52 * mckoan couldn't resist and is doing bitbake openmoko-image Dec 17 10:25:11 good morning pb Dec 17 10:25:11 good morning pb Dec 17 10:25:11 good morning pb_. Dec 17 10:25:13 oops Dec 17 10:25:17 heh Dec 17 10:25:30 this xchat here is somewhat strange Dec 17 10:25:40 the first couple of lines do not properly get displayed Dec 17 10:25:57 might be OS X Dec 17 10:25:58 and quite a pleasant morning it is here, too. the weather forecast was predicting 200mm of snow overnight, but in fact it is warm and sunny this morning. Dec 17 10:26:05 crazy met office, heh Dec 17 10:26:14 mickey|office: doh, that must be annoying Dec 17 10:26:26 It was snowing here ~30mins ago Dec 17 10:26:36 interestingly only after I wake up the iMac from sleep Dec 17 10:26:41 couple of minutes later it just works Dec 17 10:26:44 pb_: this morning I found snow :-) Dec 17 10:26:44 *shrug* Dec 17 10:26:55 sunny here as well, but slightly below 0 degree C Dec 17 10:27:13 the time where i wish i had enough money to spend half of the year on the canarian islands Dec 17 10:27:15 oh well... Dec 17 10:27:36 heh, indeed Dec 17 10:27:39 hehe Dec 17 10:27:42 taiwan is cold too Dec 17 10:27:44 hey z! Dec 17 10:28:02 car still broken this morning as well, stupid audi. Dec 17 10:28:02 where are you these days? Dec 17 10:28:06 bitbake openmoko-image is a mess Dec 17 10:28:11 hi zecke_ Dec 17 10:28:26 mickey|office: I have a cold, and I'm freezing cold taiwan Dec 17 10:28:31 pb_: bummer, why's that taking so long? Dec 17 10:28:44 ERROR: '['/home/koan/devel/openembedded/recipes/tasks/task-openmoko-base.bb']' RDEPENDS/RRECOMMENDS or otherwise requires the runtime entity 'matchbox-panel-2-applets' but it wasn't found in any PACKAGE or RPROVIDES variables Dec 17 10:28:51 mickey|office: houses here don't have the concept of isolation here... inside temp == outside temp Dec 17 10:29:05 ugh, right, i remember Dec 17 10:29:24 mickey|office: dunno, they seem to be struggling to find out what is wrong. last time I spoke to the garage (yesterday) they said that they thought they had identified the fault as being the alternator (!) but that Audi were double-checking it. Dec 17 10:29:40 hopefully I'll have it back this afternoon Dec 17 10:30:02 right Dec 17 10:30:05 how old is that vehicle? Dec 17 10:30:20 about 9 years I think Dec 17 10:30:33 hmm, yeah, that's about the timeframe problems like these show up Dec 17 10:30:51 yeah. oh well, luckily it is still under warranty. Dec 17 10:31:36 XorA: is openmoko ok in dev or needs any patches? Dec 17 10:38:41 I have a trouble with busybox, because I need to execute complex strings using: find, cp, ls, grep, pipes. How may change the "OE build conf" to stop to use busybox and use the normal version of cp, find, ls and grep? Dec 17 10:39:22 roliveira: install the coreutils package(s)? Dec 17 10:40:44 zecke_: Yes, Have I to remove the busybox from IMAGE_INSTALL and write "coreutils"? Dec 17 10:41:29 roliveira: you can start with IMAGE_INSTALL += "coreutils" Dec 17 10:41:41 roliveira: and then see if busybox is still used for anything? Dec 17 10:42:22 zecke_: I will try, thanks Dec 17 11:03:47 zecke_: I'm interested and I would be grateful if you ciould inform about the final result, thx :-) Dec 17 11:35:19 I generated a minimal system whick asks me a password at boot, does exist a default root parrsord in OE generated images? Dec 17 11:37:02 mckoan: should be blank Dec 17 11:38:33 XorA: it isn't thx :-( Dec 17 11:39:20 this line should remove root password? Dec 17 11:39:23 ROOTFS_POSTPROCESS_COMMAND += '${@base_conditional("DISTRO_TYPE", "release", "zap_root_password; ", "",d)}' Dec 17 11:39:44 mckoan: that will make it not possible to log in if your DISTRO_TYPE is release Dec 17 11:39:52 mckoan: make your distro type debug Dec 17 11:40:36 XorA: I'm using standard .dev angstron Dec 17 11:41:40 mckoan: then the password is certainly blank Dec 17 12:41:45 I belive the zip recipe is broken. It points out an incorrect url: http://ftp.info-zip.org/pub/infozip/src/zip232.tar.gz I think it should be ftp:// instead of http:// Dec 17 12:49:16 jovox_: actually looks like this ftp://ftp.info-zip.org/pub/infozip/src/zip232.zip Dec 17 12:53:48 mckoan, not in stable/2009 Dec 17 12:54:04 if it hasnt recently being updated Dec 17 12:54:55 SRC_URI = "http://ftp.info-zip.org/pub/infozip/src/zip${@bb.data.getVar('PV',d,1).replace('.', '')}.tar.gz" Dec 17 13:15:34 XorA|gone: the problem was missing task-base Dec 17 13:15:54 jovox_: I meant that you're right Dec 17 13:29:26 morning Dec 17 13:46:13 morning hrw Dec 17 13:48:26 bye Dec 17 14:00:37 mckoan, ah ok Dec 17 14:02:09 How do I get the absolute path to ${KERNEL_CC} ? Dec 17 14:02:33 Hasse__: Hi hasse Dec 17 14:03:32 sgh: God dag Hr. :) Dec 17 14:03:51 Hasse__: crazy :D Dec 17 14:03:57 ... oh and that had NOTHING to do with God... sorry! Dec 17 14:04:11 03Sebastian Spaeth  07org.openembedded.dev * r727b276ed3 10openembedded.git/conf/distro/include/sane-srcrevs.inc: Dec 17 14:04:11 sane-srcrevs.inc: Bump shr apps to latest Dec 17 14:04:11 Signed-off-by: Sebastian Spaeth Dec 17 14:05:09 sgh: I'll move back to UMC channel - waiting for a job from the Jeddy ;) Dec 17 14:05:23 Hasse__: :( Dec 17 14:05:33 Hasse__: ;( Dec 17 14:05:39 :P Dec 17 14:07:32 sgh: Booting your own image from VirtualBox is not easy for me ... Dec 17 14:08:48 How do I get the location of the target-compiler ? Dec 17 14:16:29 I'm trying to compile a kernel module. Apparently it compile allright. The problem is that it complains about undeclared marcos and such. There are no "missing header"-messages. The message is here http://pastebin.ca/1718199 . It compiles fine on my local computer. Dec 17 14:22:48 sgh: you should check gcc and linux versions on OE and local computer Dec 17 14:23:24 mckoan: do you have a hint on what I might look for ? Dec 17 14:25:01 'TASK_INTERRUPTIBLE' undeclared looks like a different linux includes (kernel version) which one are you using on both systems? Dec 17 14:25:37 mckoan: 2.6.32 on both systems. Dec 17 14:26:38 mckoan: hmm ...seems like sched.h does not get included under oe.... Dec 17 14:28:34 mckoan: thanks for yuo help .... I will patch you sources since I don't know why the header is not included Dec 17 14:31:49 morning Dec 17 14:31:59 hi kergoth Dec 17 14:32:32 hey mckoan Dec 17 14:38:04 otavio: ping Dec 17 14:41:33 hello Dec 17 14:43:20 do you happen to recall why xcb was disabled in libx11, by chance? I can't find an actual reason for it, just the bug, which says to remove it. (again, we really need USE.. :) Dec 17 14:45:14 kergoth, DISTRO FEATURES Dec 17 14:48:59 i worry about distro features being overloaded, due to the namespace being so flat.. guess we'll see how it goes Dec 17 14:50:49 In my distro, I want to use bash instead busybox. Is possible to build the distro with sh linking to bash? I put "bash" in the IMAGE_INSTALL, but the "sh" link continued pointing to busybox. Anone may help me, please? Dec 17 14:51:52 03Chris Larson  07org.openembedded.dev * r7b49682d1b 10openembedded.git/recipes/lzma/lzma-native_4.65.bb: Dec 17 14:51:52 lzma-native: dep on zlib-native, not zlib Dec 17 14:51:52 Signed-off-by: Chris Larson Dec 17 14:52:34 hi; i have a problem with the latest angstrom. I'm trying to build a kernel module (for the igep0020 board) and when I load the module, i get the following error: Version magic '2.6.28.10 mod_unload modversions ARMv7 ' should be '2.6.28.10 mod_unload ARMv7 ' - I checked in the corresponding .config and module versioning seems to be activated. Any idea? Dec 17 14:52:47 (i tried asking in #angstrom but noone knows, maybe someone has an idea here..) Dec 17 14:54:16 roliveira: iirc there was danger in screwing with /bin/sh via packaging scripts. i really don't see the point in switching it anyway, if oyu want bash in a script, use #!/bin/bash, if you want bash for your user, change the password file / use usermod.. but if you *really* want to, you could make the change in a rootfs postprocess script. Dec 17 14:56:04 otavio: in addition to the xcb question, i was wondering about the lzma upgrade. are any of the patches/changes for 4.65 worth trying to pull back to 4.17? I'm syncing mvl6 with oe, but don't want to pull in newer versions of things just yet (and we have to keep the old ones indefinitely) Dec 17 15:19:35 kergoth: about the xcb change I don't remember any reason for it but I'm intending in January to update our base here and wish to use xcb to everything Dec 17 15:19:45 okay Dec 17 15:19:54 kergoth: about lzma I don't know about anything of t Dec 17 15:19:57 we had a need for it for some crazy graphics driver Dec 17 15:20:09 kergoth: those patches Dec 17 15:20:17 ah. where'd they come from? Dec 17 15:20:24 kergoth: but would be nice to check if there's something interesting Dec 17 15:20:28 kergoth: debian IIRC Dec 17 15:20:32 okay, thanks Dec 17 15:21:00 found a great way to stay in sync with OE without cherry picking hundreds, if not thousands of commits, finally Dec 17 15:21:40 kergoth: how? Dec 17 15:22:59 amend-recipes.inc allows you to put an amend.inc file in a recipe's FILESPATH to add on changes to its metadata. switched *all* of our changes to it, then i rm -rf'd my working copy in my git repo for our collection, re-deployed all the collection's recipes from OE using newcollection.inc (deploys a recipe and all the files it needs given current config into another directory), then compared the old and new with git status after resurrecti Dec 17 15:22:59 amend.inc files and patches we added Dec 17 15:23:09 moving the changes outside the recipes themselves avoids most issues Dec 17 15:23:18 so sync via copy, rather than cherry-pick Dec 17 15:42:03 re Dec 17 15:42:15 hey hrw Dec 17 15:45:03 kergoth: lzma for a graphics driver? Dec 17 15:45:47 no, xcb Dec 17 15:45:52 w/ libx11 support for it enabled Dec 17 15:46:08 compressed framebuffer would be pretty eleet Dec 17 15:46:18 pb_: haha Dec 17 15:46:48 pb_: fbpng? Dec 17 15:58:38 03Chris Larson  07org.openembedded.dev * r2ac4f6cdee 10openembedded.git/recipes/portmap/portmap_6.0.bb: Dec 17 15:58:41 portmap 6.0: kill -O2 and replace -fpie with -fPIC. Dec 17 15:58:43 4fe22eed50ce7c45da30b0efc6106b90281ff5f5 changed the CFLAGS bits to match the Dec 17 15:58:45 makefiles, but bitbake.conf is what controls optimizations--recipes shouldn't Dec 17 15:58:47 be forcing -O2, and -fpie can cause problems on some archs, and should be a Dec 17 15:58:49 distro policy decision regardless. Dec 17 15:58:51 Signed-off-by: Chris Larson Dec 17 15:59:31 have any of you guys played with firmware linux? Dec 17 15:59:40 i think we should think about doing what it does wrt build machine binaries Dec 17 16:02:09 you mean that iso image which checks does your x86 HW comply with ACPI? Dec 17 16:02:51 no, that was http://linuxfirmwarekit.org thing Dec 17 16:05:00 with firmware linux, its all shell scripts, but the concept is cute. you run ./build.sh , it builds some build tools, crosscompiler, minimal target fs w/ development tools, a filesystem image, then a ./run-from-build.sh drops you inside of a qemu in that fs, with distcc automatically set up pointing at a spawned distccd on the host side using the crosscompiler, so you can just do native development and not worry about cross st Dec 17 16:05:13 but.. the cool thing I'm talking about is, it overrides PATH to not include the host paths at all Dec 17 16:05:13 a Dec 17 16:05:13 n Dec 17 16:05:16 Man, I am so not a fan of the qemu method Dec 17 16:05:27 instead creating a symlink tree with *only* the build machine binaries it knows it can rely on Dec 17 16:05:41 so you don't sneak in inadvertant deps on build environment tools Dec 17 16:06:43 CosmicPenguin: s/qemu/kvm/ and you get it a bit faster Dec 17 16:06:50 and total separation Dec 17 16:08:58 what i think would be cool for oe is, just a helper script + MACHINE to drop you into the qemu image you built.. but add in the distcc automatic stuff that fwl has, and also add in automatic setup of a minimal http server to set up an ipkg feed pointing at tmp/deploy/ipk Dec 17 16:09:09 so you could just contrib/qemu/qemu.sh, and you're in an env where you can opkg install foo Dec 17 16:10:08 VMware's LabManager Dec 17 16:12:25 kergoth: kind of Poky run-qemu stuff on steroids? Dec 17 16:14:59 yeah Dec 17 16:16:02 i like the qemu stuff for the app developer use case, as they tend to not want to have to deal with tools like bitbake for their own code, until they add it to the automated build process, imo Dec 17 16:17:53 I can understand that Dec 17 16:19:45 * kergoth is looking over all the alternative embedded Linux build tools to see what we can pull over without revamping the works Dec 17 16:23:58 * likewise thinks that's a good idea Dec 17 16:25:07 I have had a discussion with Rob about FWL. His major statement is: big iron hardware will outgrow embedded hardware, thus we can safely take the qemu approach without caring about anything cross*. Dec 17 16:25:43 So in its essence, FWL is quite different from OE. However, there might be low hanging fruit elsewhere Dec 17 16:25:55 * hrw -> off Dec 17 16:28:23 likewise: aye, they're really used for two completely different things. OE builds and maintains distributions, but FWL is *explicitly* removing the distro piece, his presentation talks about his annoyance with the fact that buildroot become a distro. Dec 17 16:29:19 kergoth: what's in a name? FWL can be seen as a single distro itself. Dec 17 16:29:42 A distro being a set of packages, versions and their interdepencies like configuration files etc. Dec 17 16:29:44 it can, but the difference is the focus. fwl isn't gaining new "recipes" :) Dec 17 16:29:51 it's more static, single purpose Dec 17 16:29:57 true Dec 17 16:30:11 and the lack of a target package management system Dec 17 16:30:19 whereas currently we pretty much require it Dec 17 16:30:25 but yes, its a distro of sorts, just not like us :) Dec 17 16:35:39 03Chris Larson  07org.openembedded.dev * rbce2054085 10openembedded.git/recipes/glib-2.0/ (3 files): Dec 17 16:35:39 glib-2.0-native: kill unnecessary FILESPATH overrides. Dec 17 16:35:40 Signed-off-by: Chris Larson Dec 17 16:36:05 whew, syncing mvl6 with oe is taking a while... hopefully it'll be less painful going forward Dec 17 16:48:15 yikes, oe's python needs tk now? we need USE like yesterday Dec 17 16:50:02 it always needed Dec 17 16:50:08 Poky removed that dep Dec 17 16:50:59 Call that "improved DISTRO_FEATURES" :) Dec 17 16:51:03 ugh Dec 17 16:51:49 well, that oedem discussion seemed to imply we should be leveraging DISTRO_FEATURES more, maybe we should just start adding features for these optional deps asap Dec 17 16:52:14 indeed Dec 17 16:52:18 yeah, I think that is the idea Dec 17 16:53:01 having a DISTRO_FEATURE for tk would be fine. Dec 17 16:59:42 03Chris Larson  07org.openembedded.dev * ra7c752d20f 10openembedded.git/recipes/python/ (python_2.6.1.bb python_2.6.2.bb): Dec 17 16:59:50 python: only dep on tk if tk is in DISTRO_FEATURES. Dec 17 16:59:57 Signed-off-by: Chris Larson Dec 17 17:00:03 there, all better Dec 17 17:01:31 kergoth: so angstrom users do not get python-thinker now? Dec 17 17:05:20 hrw: here, fixed Dec 17 17:05:42 (CIA is slow) Dec 17 17:05:54 thx Dec 17 17:05:59 03Chris Larson  07org.openembedded.dev * rf25fb0eb70 10openembedded.git/conf/distro/include/angstrom.inc: Dec 17 17:06:09 angstrom: add tk to DISTRO_FEATURES for compatibility (python-tkinter module) Dec 17 17:06:16 Signed-off-by: Chris Larson Dec 17 17:06:44 hrw: thanks for the reminder Dec 17 17:06:49 np Dec 17 17:09:50 which public GIT server would you recommend for keeping some LGPL sources? Dec 17 17:10:25 I've been using github, but I'm sure there are new ones around that i haven't tried yet :) Dec 17 17:27:40 I have a trouble with udev and fstab and the partitions mounted. The "mount" command show this three root partitions: http://pastebin.com/m6bb638ec . The "df -h" show only /dev/root and /dev/hda1, and i only need one rootfs partition, how can i solve that? Dec 17 17:50:16 roliveira: your mount command shows only two root partitions (and not three, the 3rd one is not on /), guess you get two because you have two entries in /dev/fstab. Dec 17 17:50:43 roliveira: but perhaps better ask your Q in the cannel for your distro (e.g. #angstrom), this is a developer channel Dec 17 17:53:23 eFfeM1: Thanks, I will try there Dec 17 18:13:14 http://marcin.juszkiewicz.com.pl/2009/12/17/released-sources-of-my-protracker-module-player/ Dec 17 18:14:16 see you tomorrow Dec 17 18:31:31 mickey|dinner: any idea why ee14422b23949fa59555d86be8f12386dc5ae827 removed python-bzip2? Dec 17 18:35:25 mickey|dinner: yum rdeps on it, but python no longer packages it :) Dec 17 18:41:11 hi k. Dec 17 18:41:26 ah i see Dec 17 18:41:29 packaging changes Dec 17 18:41:36 bzip2 has moved into python-compression Dec 17 18:41:45 we need to change yum to rdepend on that instead Dec 17 18:45:35 ah, okay, thanks Dec 17 18:47:44 03Chris Larson  07org.openembedded.dev * r502ca1bf89 10openembedded.git/recipes/yum/yum_3.2.18.bb: Dec 17 18:47:44 yum: kill python-bzip2 dep, as python-compression includes it now, per mickeyl Dec 17 18:47:44 Signed-off-by: Chris Larson Dec 17 18:55:00 RP, ping, did you by any chance see my mail on staging and libpcre from a few days ago ? still stuck with it Dec 17 19:26:39 anyone know how to get ibot to log some new channels? Dec 17 19:30:16 Crofton send an email to riker, he wants you to be op in the channel (so not everyone can add a bot), turned out to be very easy, did it for hawkboard Dec 17 19:30:47 hmm, will need to figure out how to get ops then Dec 17 19:36:29 for what channel? is there a channel owner ? otherwise you *might* be able to become ops if there is a split near you (not sure about that) Dec 17 19:39:12 #gnuradio Dec 17 20:02:47 re Dec 17 20:17:26 ~curse opkg Dec 17 20:17:27 May you be reincarnated as a Windows XP administrator, opkg ! Dec 17 20:23:08 Crofton|work: I was wondering if we could teach bitbake if gcc-cross is built sucessfully dont bother to build gcc-cross-intermediate and gcc-cross-initial Dec 17 20:24:04 IOW there could be another property to recipes which make them overridable Dec 17 20:25:29 khem: hmm, gcc-cross could PROVIDES them, but i don't know if bitbake would be smart enough to avoid a self dependency there. if so, that would solve it Dec 17 20:33:11 kergoth: when building from scratch they should be provided by the respective recipes Dec 17 20:33:18 i know Dec 17 20:33:51 I hope bitbake wont decide oh let me pick gcc-cross for all of them Dec 17 20:34:04 so may be something like weak provide Dec 17 20:34:09 is what is needed Dec 17 20:35:44 well, I'm thinking gcc-cross depends on gcc-cross-intermediate, and it also PROVIDES it, and bitbake is smart enough to know that if you try to build gcc-cross, it can't depend on itself, so fall back to the other Dec 17 20:35:50 so yeah, weak provide of sorts would do Dec 17 20:36:57 hmmm Dec 17 20:37:00 if I understand your problem right, I think the correct answer is to teach bitbake to distinguish between dependencies which affect the binary output from a given package (which are useful to rebuild if they have changed version) and those that don't (which aren't). Dec 17 20:37:19 most of the native-* tools would fall into the latter category, and so would the initial bootstrap compilers. Dec 17 20:37:47 that's a good point Dec 17 20:38:41 pb__ yes Dec 17 20:40:07 pb__: then staging is another problem. e.g. gcc-cross and gcc-cross-intermediate install common files Dec 17 20:40:23 yah, but that problem will go away with fully packaged stagin Dec 17 20:40:24 g Dec 17 20:40:42 it's just an artifact of the old shared staging mechanism, I don't think there is much point trying to solve it in the old world. Dec 17 20:40:56 hmmm Dec 17 20:42:08 right now if I remove gcc-cross it also cramps out gcc-cross-intermediate though it appears to be installed but its broken by removing gcc-cross Dec 17 20:42:39 how would this problem get solved by packaged staging ? Dec 17 20:44:07 unless there is a way to obsolete package by another packages Dec 17 20:44:49 Could implement CONFLICTS/REPLACES, and making pstage obey it as Conflicts/Replaces in the packages, but i dont' know if that's the best way to go Dec 17 20:45:20 yeah that would work Dec 17 20:45:25 * khem thinks of rpm Dec 17 20:49:15 if I have gcc-cross staged and glibc staged properly and now I update metadata and get changes in gcc recipes and glibc recipes if I bitbake my image then it should only rebuild gcc-cross and glibc and not the intermediate packages Dec 17 20:50:21 that's what pb was talking about in the distinguishing deps which affect binary output of a pcakge vs those that dont Dec 17 20:50:56 correct Dec 17 20:51:18 kergoth: may be we should teach bitbake to differentiate such recipes Dec 17 20:51:34 that's what pb suggested, yes Dec 17 20:51:45 cool Dec 17 20:51:49 lzma hates me Dec 17 20:52:06 before these changes, this patch applies, after it doesn't. SRC_URI and S are identical Dec 17 20:52:09 what the hell? Dec 17 20:52:55 a diff comparing them doesn't seem to show anything that could possibly affect patching Dec 17 20:52:58 hmm Dec 17 20:53:09 ah, i see Dec 17 20:53:55 lzma SRC_URI is interesting Dec 17 20:54:21 silly do_unpack_append Dec 17 20:56:22 updating to the latest recipes for everything in mvl6 that came from oe is time consuming Dec 17 21:11:57 I'd really like bitbake to not bother rebuilding a task even if its dependent tasks change. i really, really don't give a shit about fetching if I've already compiled the thing.. /me tries to remember why we did it that way Dec 17 21:18:21 hi, I've updated the udev 141 recipe...I bet it's considered as core and should be submited for review Dec 17 21:18:40 it's for adding a machine specific udev rule Dec 17 21:18:55 if in doubt send it to the list Dec 17 21:18:59 ok Dec 17 21:19:03 thanks Dec 17 21:25:27 doh, power cut there Dec 17 21:26:06 pb__: howdy. long time. (jacques here) Dec 17 21:30:32 g'day linuxjacques Dec 17 21:36:16 jo Dec 17 22:33:13 raster raster Dec 17 22:33:24 woglinde: wogz! Dec 17 23:06:03 i think i finally have a documented, functional process for syncing OE changes into the MVL6 collections. will have to see if they'll let me post them, might be useful for others maintaining collections which are subsets of OE Dec 17 23:08:02 awesom Dec 17 23:08:19 i may be insane, though Dec 17 23:08:38 true, but how's that related? ;) Dec 17 23:08:42 heh :) Dec 17 23:10:08 two separate processes, one for config files & classes, one for recipes. the former uses a squashed merge, but only for those parts of the tree, and the latter involves doing a version lockdown, wiping your working copy of the collection, then re-deploying the entire collection based on a previously stored list of recipes from bitbake -s coupled with newcollection.inc, then using git status to compare (one can leverage amend.inc files to a Dec 17 23:10:08 stepping on oe's toes here) Dec 17 23:14:01 sounds convoluted at first glance, but might actually work Dec 17 23:15:35 you can't just merge everything, cause then you get a *shitload* of new recipes all over the place that you don't likely want (we want to explicitly add new versions), plus you get way more "removed in HEAD but modified in remote" conflicts to deal with Dec 17 23:15:49 and cherry picking just doesn't scale, if you want to track OE closely Dec 17 23:18:21 right Dec 17 23:18:39 i spent mon-thurs trying to pull back recipe & config & class changes, got most of the config files & classes done, and had 400 more cherry picks to do in the recipes with that workflow... with this, yesterday and today and I have 15 config files & classes left to merge, and this collection is done Dec 17 23:18:50 fun fun :) Dec 17 23:19:25 hehe Dec 17 23:20:43 course, all I've done is confirmed that the patches still apply after the oe changes are pulled in, still have to actually *build* everything and make sure nothing exploded horribly.. but even so, 2 days isn't bad, especially since some of our local changes were still directly in the recipes rather than in amend.inc files Dec 17 23:22:48 *nod* definitely a good start. i think that'd be interesting for lots of parties Dec 17 23:23:19 kergoth: I use to spend more than a day in merging nptl uclibc branch :) Dec 17 23:23:26 yikes Dec 17 23:24:04 so 2 days is not bad I say Dec 17 23:33:57 I'm getting the following error when compiling mono-mcs-intermediate: http://pastebin.com/m4c0c56c5 before I did into it I wanted to make sure it was not some known error. Dec 17 23:34:57 03Denis 'Gnutoo' Carikli  07org.openembedded.dev * rd4c00c4ffd 10openembedded.git/recipes/xorg-xserver/xserver-xorg-conf/bug/xorg.conf: Dec 17 23:35:02 xserver-xorg-conf: improved buglabs's bug device's xorg.conf Dec 17 23:35:04 *each screen has now its respective tslib input device Dec 17 23:35:06 *added usb keyboard support Dec 17 23:36:56 I'll go bye Dec 17 23:37:05 (need sleeping) Dec 17 23:37:22 #/msg nickserv identify blaat Dec 17 23:38:05 likewise, I saw that : "#/msg nickserv identify blaat" Dec 17 23:38:16 so change your password as soon as possible Dec 17 23:43:41 #/msg nickserv set passwd blaat boe Dec 17 23:44:08 likewise: we still saw your password change to boe :-) Dec 17 23:44:41 I know, I'm f*cking up stuff all day :-| Dec 17 23:44:42 switch to the freenode server window when you do things like that so that an error won't show your password since its not a channel. Dec 17 23:45:10 thanks Dec 17 23:45:19 likewise: ahh, I do that quite a bit. I can't count the number of times I have broadcasted my passwords by fat fingers. Dec 17 23:46:11 I am a bit pissed off by installing Ubuntu 9.10 where GRUB2 doesn't recognize my disk anymore. Took me three hours to get back to a working setup. Dec 17 23:46:48 I'm now happily bringing over all my applications, settings etc... and even passwd's :-) Dec 17 23:47:07 and happily showing the passwords to us, too? :D Dec 17 23:47:28 yes, sure. Dec 17 23:47:37 sharing makes me happy :-) Dec 17 23:47:57 Guess I'm surprised that grub would fail with the disk - that's not supposed to happen. Dec 17 23:48:15 Do you have some sort of weird disk device? Dec 17 23:48:23 even the grub2 rescue shell 'ls' command didn't show the disk Dec 17 23:48:37 Samsung F1 on a Dell e520 Dec 17 23:48:48 I'm always amazed when my dead stupid PC managed to restart and GRUB doesn't fail Dec 17 23:50:32 Going to reboot chatzilla and see if I copied its prefs ok... cu Dec 17 23:51:32 ~seen ant Dec 17 23:51:36 ant was last seen on IRC in channel #oe, 289d 23h 44m 1s ago, saying: 'nite all'. Dec 17 23:54:41 is ibot's clock ok? i'm pretty sure i've seen ant here more recently than taht Dec 18 00:06:04 ~seen ant_ Dec 18 00:06:07 ant_ was last seen on IRC in channel #oe, 327d 55m 22s ago, saying: 'yes but then it sneaks out on framebuffer :-/'. Dec 18 00:06:12 hmm. Dec 18 00:06:29 Tartarus: hey long time Dec 18 00:07:39 grg: I have a local patch to enable -mplt in mips toolchain. I will post it for review soon I have added that as DISTRO_FEATURE and tested it on minimal Dec 18 00:08:22 mwester: hello Dec 18 00:08:48 khem, ok. I've not been keeping my oe tree up to date.... Dec 18 00:09:09 grg: I remember once you wanted -mplt for gcc Dec 18 00:09:28 khem, yes i've been using it successfully for some time now Dec 18 00:09:56 grg: cool do you have any measurement on perf Dec 18 00:10:02 probably time i did a git pull in my oe tree Dec 18 00:10:20 khem, nah. no performance numbers Dec 18 00:10:23 I think it cant be used on uclibc but glibc/eglibc should be ok Dec 18 00:10:29 grg: OK Dec 18 00:10:50 I will try to do some numbers but I have qemu lets see Dec 18 00:11:01 i did keep a rootfs built without mplt for testing, but haven't found the time for it Dec 18 00:11:40 intuition tells me that application start up time is improved with mplt Dec 18 00:12:32 khem, are you targetting a mips board, or just playing within qemu? Dec 18 00:13:08 I have no hardware so just qemu for now :) Dec 18 00:13:22 I only have arm hardware Dec 18 00:15:04 has anyone considered an alpha port? i recently got given one and am keen to play Dec 18 00:15:37 grg: no alpha here even debian stopped supporting it IIRC Dec 18 00:15:52 but why not Dec 18 00:15:56 yeah. there's only gentoo left. and i don't like gentoo Dec 18 00:16:10 then port OE over :) Dec 18 00:18:06 i see an arm-darwin site file there... wtf? Dec 18 00:18:54 heh Dec 18 00:42:36 * khem 's beagle is talking now Dec 18 00:44:02 cat /dev/urandom > /dev/dsp will make it bark! Dec 18 00:44:14 :) Dec 18 00:44:29 that's my default audio test :") Dec 18 00:45:06 hmm... that looks like a smiley with a hitler moustache! Dec 18 00:45:11 I did not have right cables so far, I was making the combinations myself **** ENDING LOGGING AT Fri Dec 18 02:59:56 2009