**** BEGIN LOGGING AT Wed Feb 15 02:59:57 2012 Feb 15 03:46:33 hello everybody Feb 15 03:54:09 anybody here? Feb 15 04:11:13 i do face some problems while compiling a helloworld sample Feb 15 04:11:17 any help? Feb 15 04:11:55 thanks Feb 15 07:59:45 good morning Feb 15 08:19:29 is there any good reason for choosing systemd over sysvinit? or vica versa? We're developing a embedded linux application on the omap4460, and are quite new to embedded systems. Will there be any difference in boot time or anything? Feb 15 08:24:40 martiert: yes it could be faster and it has many advantages.. read the description online.. almost the same as for desktop applies to current embedded systems Feb 15 08:27:16 JaMa: ok, do I have to configure my kernel with some special parameters to make it work? Feb 15 08:27:29 yes as described Feb 15 08:27:48 ok, thanks Feb 15 08:28:39 you can also find koen's presentation from ELCE 2011 Feb 15 08:30:25 thanks, I'll have a look at that Feb 15 08:46:06 ~seen icanicant Feb 15 08:46:09 icanicant <~klawson@213.218.221.154> was last seen on IRC in channel #oe, 17h 17m 50s ago, saying: 'mckoan: can you tell you're the first person to follow the steps? ;)'. Feb 15 08:46:38 icanicant: http://pastebin.com/BSgkCBhY Feb 15 08:47:20 ~seen woglinde Feb 15 08:47:21 woglinde <~heinold@g229045192.adsl.alicedsl.de> was last seen on IRC in channel #oe, 5d 11h 10m 25s ago, saying: 'good nite'. Feb 15 09:36:05 morning all Feb 15 09:37:13 mornin Feb 15 10:29:28 morning Feb 15 10:30:02 hi otavio Feb 15 10:44:39 morning all Feb 15 10:46:20 hi pb Feb 15 10:52:51 hi pb_, otavio, woglinde Feb 15 10:53:02 hi bluelightning Feb 15 10:53:38 bye again Feb 15 10:53:41 till later Feb 15 10:57:53 I've broken out the migration section from the OE-Core page and extended it to create this: http://www.openembedded.org/wiki/Migrating_metadata_to_OE-Core Feb 15 10:57:59 anything missing? Feb 15 10:58:12 I'm not yet covering image recipe changes, that will need to be added Feb 15 10:59:40 looks pretty good Feb 15 11:03:08 bluelightning: maybe slice it thin and list all R* vars (RDEPENDS/RRECOMMENDS/RSUGGESTS/RPROVIDES/RCONFLICTS/RREPLACES) Feb 15 11:03:51 ant_work: sorry what do you mean by "slice it thin"? Feb 15 11:05:35 ah, it's an idiom/a way to say we (latin langs) have Feb 15 11:08:55 if smthg is too big to eat/understand, slice it thinner Feb 15 11:12:40 ant_work: ok, do you refer to the whole document or just the R* item? Feb 15 11:19:01 atm I'd say just the R* variables which have to do with packages Feb 15 11:23:38 moin folks Feb 15 11:26:17 mickey_office: moin moin Feb 15 11:27:47 hi mickey Feb 15 11:45:48 ant_work: ok, have reformatted it a little - any better? Feb 15 11:59:58 hey pb_, how's the house construction going? Are you finished by now? Feb 15 12:19:09 bluelightning: very nice, now the Main_Page of the wiki should be updated Feb 15 12:30:49 mickey_office: well, not entirely finished, but we're living in it now :-) Feb 15 12:31:29 pb_: finished houses are an illusion ;) Feb 15 12:31:41 heh Feb 15 12:31:43 yeah, it seems Feb 15 12:32:03 I should upload some pictures of our status so far, I guess Feb 15 12:32:07 our house has not been finished within 90 years ;) Feb 15 12:32:15 yes! Feb 15 12:32:17 hah Feb 15 13:08:40 Is there a neat way to add custom modifications to the kernel defconfig without duplicating the whole kernel into a custom layer? Is the "correct" way to have a small bbappend in a custom layer that patches the defconfig? Feb 15 13:12:32 sure Feb 15 13:12:45 are you using linux-yocto maybe? Feb 15 13:13:07 I wonder, can you use .cfg fragments even with linux.inc by virtue of still using kernel.bbclass? Feb 15 13:13:32 bluelightning: I've started tests in that direction Feb 15 13:13:56 ideally we provide the basic fragments Feb 15 13:14:27 atm seems only documented for linux-yocto Feb 15 13:15:13 but then you need the glue to compose the defconfig pieces together Feb 15 13:16:33 'our' linux.inc does little work in that regard Feb 15 13:20:48 ant_work: this is the linux-kirkwood but i'm generally interested for other platforms as well Feb 15 13:20:56 i have a few platforms on the go Feb 15 13:21:15 coming from oe-classic we kept the linux.inc structure Feb 15 13:21:46 but for machines supporting 3.x kernels it's maybe better to use linux-yocto Feb 15 13:22:50 i.e. you provide only the machine-specific lines of your defconfig and the rest is added in a modular way Feb 15 13:23:40 the kirkwood kernel recipe includes linux-dtb.inc and linux-tools.inc then pretty much just changes the SRC_URI and applies a few patches Feb 15 13:26:54 so i'm thinking: add a custom layer that has a simple "linux-kirkwood" recipe folder that bbappends the standard one, and just applies a patch to the defconfig file? Feb 15 13:30:30 I see the examples define KMACHINE and sometime KBRANCH...there are some vars to set Feb 15 13:31:12 icanicant: FYI Darren Hart gave a talk at ELC 2011 about the linux-yocto kernel: http://free-electrons.com/pub/video/2011/elc/elc-2011-hart-yocto-kernel-development-x450p.webm Feb 15 13:33:16 bluelightning: Thanks, will have a look. Feb 15 13:34:04 ant_work: perhaps i should move to a 3.x yocto kernel to keep in line with other folk Feb 15 13:34:13 yes Feb 15 13:34:28 I'd say look at the meta-intel examples Feb 15 13:34:40 (I was doing that) Feb 15 13:35:07 thanks Feb 15 13:36:13 and at the kernel-yocto.bbclass Feb 15 13:41:59 su root Feb 15 13:42:02 ops Feb 15 13:58:03 heh, bit early over there isn't it? Feb 15 13:59:19 hi, any idea on how to fix that: Feb 15 14:00:01 error: implicit declaration of function '__stpncpy_chk' Feb 15 14:00:24 it's part of string3.h Feb 15 14:00:30 which states in the comments: Feb 15 14:00:38 / XXX We have no corresponding builtin yet. Feb 15 14:00:50 s#/#//# Feb 15 14:02:58 in which lib could that be implemented Feb 15 14:03:10 it's an extern in libc or gcc if I remember well Feb 15 14:04:41 it's in glibc Feb 15 14:04:53 it's rather hard to say how to fix your problem without any context Feb 15 14:05:29 if you can provide an actual log, rather than just a single line from it, that would make it easier to suggest a course of action Feb 15 14:12:45 right it's in glibc Feb 15 14:12:48 in that file: Feb 15 14:12:56 ./eglibc-2_15/libc/debug/stpncpy_chk.c Feb 15 14:13:04 basically I'm trying to make a recipe for nss Feb 15 14:13:19 a mozilla security lib used in chrome for instance Feb 15 14:14:44 http://pastie.org/private/48zpizhmzwbjge0ebrbsog Feb 15 14:15:19 basically it includes strings.h Feb 15 14:15:26 and that includes string3.h Feb 15 14:15:39 which has '__stpncpy_chk' in extern Feb 15 14:16:17 ask Eric :) Feb 15 14:17:09 ok Feb 15 14:17:37 ah it's you gnutoo :) Feb 15 14:17:43 yes Feb 15 14:17:57 working for eukrea? :) Feb 15 14:18:06 yes, from my home tough Feb 15 14:19:38 it's new gcc goodies :) Feb 15 14:19:42 ok Feb 15 14:19:47 I've been bitten by it too Feb 15 14:19:56 I've gcc 4.6 with angstrom-bleeding Feb 15 14:20:06 they've enabled quite some checks by default Feb 15 14:20:16 ah ok Feb 15 14:20:17 I've fixed it by disabling those errors :) Feb 15 14:20:30 thoses checks you mean Feb 15 14:20:41 -Werror=implicit-function-declaration Feb 15 14:20:46 ah ok Feb 15 14:20:55 then I should do that? Feb 15 14:21:02 I know how to do that Feb 15 14:21:04 or you can patch eglibc for 4.6 version (fix that in code) Feb 15 14:21:09 some kernel do that Feb 15 14:21:11 ah ok ouch Feb 15 14:21:19 eh...missing prototypes Feb 15 14:21:24 yep Feb 15 14:21:32 just add that line in right place Feb 15 14:21:34 what's at line 124 of string3.h? Feb 15 14:21:43 that looks like the location of your real problem. Feb 15 14:21:45 ynezz, what line where? Feb 15 14:21:48 ok Feb 15 14:22:32 http://pastie.org/private/aa3zqsgulq1zcvfpf3ws3w Feb 15 14:22:55 extern char *__stpncpy_chk (char *__dest, __const char *__src, size_t __n,size_t __destlen) __THROW; Feb 15 14:23:03 line 124 is a comment Feb 15 14:23:14 and line 125 is what I just pasted Feb 15 14:23:36 Does it work if you change the // comment to be a /* .. */ comment? Feb 15 14:23:43 or just delete the comment altogether Feb 15 14:24:00 oh, 4.6 is that picky? :) Feb 15 14:24:18 seem so Feb 15 14:24:23 thanks a lot!!!! Feb 15 14:24:25 if you ask for "-ansi", yes Feb 15 14:24:31 ah so :) Feb 15 14:24:33 how do I fix upstream then Feb 15 14:24:51 well, it's a bug in glibc that it uses a c++ comment in a header that might be included from c89. Feb 15 14:25:12 and it's not the same header Feb 15 14:25:21 so, file a defect report against (e)glibc as a starting point. Feb 15 14:25:28 stdlib.h has it too Feb 15 14:25:39 ok Feb 15 14:25:51 in eglibc bug tracker? Feb 15 14:25:59 or in yocto one? Feb 15 14:26:31 and, maybe try to persuade mozilla to get with the 1990s and adopt a better language variant Feb 15 14:26:39 lol Feb 15 14:26:43 :D Feb 15 14:27:18 hm, but it's quite amazing to see -ansi -Wall in libc Feb 15 14:27:24 (at least for me) Feb 15 14:34:31 thanks a lot for the help Feb 15 14:34:53 but in which bugtracker should I report? Feb 15 14:35:00 eglibc, I guess Feb 15 14:35:06 ok thanks Feb 15 14:35:13 assuming the bug really is in upstream eglibc, not some crazed yocto patch Feb 15 14:35:24 I think it probably is but it'd be worth checking. Feb 15 14:45:40 ok Feb 15 15:32:25 so... if I am setting up OE on ubuntu, from what I have seen, Ubuntu 10.04 32-bit seems to be the best option currently Feb 15 15:32:28 am I correct on that? Feb 15 15:32:34 * cryptk has never set up OE before... Feb 15 15:33:08 64-bit should be fine, that's what I use. Feb 15 15:33:22 cryptk: yes, I had it 32b for some months Feb 15 15:33:47 the reason I was saying 32 bit was because of this line on the wiki... Feb 15 15:33:50 "You can also install Psyco Python JIT compiler to speed up BitBake. Psyco works on 32-bit x86 platforms only." Feb 15 15:33:52 then I updated and now I ignore 11. from bootloader list ;) Feb 15 15:34:03 you don't need/want psyco anyway Feb 15 15:34:20 ahh, is it one of those things where it makes the builds go faster but they are inconsistent? Feb 15 15:34:25 it doesn't make any noticeable difference to performance and I'm not sure it's even maintained anymore with newer python versions Feb 15 15:34:31 gotcha Feb 15 15:34:41 I think it did use to make a difference to performance with much older bitbakes, but not any more Feb 15 15:34:45 are there any problems doing OE on something newer like 11.10? Feb 15 15:34:58 I actually seem to be on 11.04 at the moment and that's fine. Feb 15 15:35:20 cool, I will give it a shot on 11.10 then, shouldn't be too different from your 11.04 setup Feb 15 15:35:22 my laptop is running debian sid and I haven't had any trouble there either. I expect 11.10 should also be ok Feb 15 15:35:34 yeah, should be good, thanks for the guidance! Feb 15 15:35:49 kinda funny that I know so little about OE and yet I run your servers ;) Feb 15 15:35:53 lol Feb 15 15:36:15 tell ya what... you help me out with OE and I will fix your servers whenever they break ;) Feb 15 15:36:24 I'm running 11.10, FWIW Feb 15 15:36:28 pb_: well, maybe that Unity is ok on laptops Feb 15 15:36:36 Kubuntu that is Feb 15 15:36:39 I'm on 12.04 Feb 15 15:36:46 * cryptk runs 11.10 on my home system but without unity Feb 15 15:36:49 I use Gnome Shell Feb 15 15:36:57 and systemd-image in angstrom bleeding builds just fine Feb 15 15:37:07 ant_work: no, unity was teh suck when I tried it on my laptop Feb 15 15:37:16 iirc atm only some Fedora give problems Feb 15 15:37:21 I use gnome 3 on my laptop and gnome 2 on my desktop right now Feb 15 15:37:24 is there any benefit to doing this on a desktop install with a GUI or is a server install fine? Feb 15 15:37:25 heh Feb 15 15:37:26 yes, unity is a such horrible mess Feb 15 15:37:41 as in, are there any good GUI tools that I would miss out on if I did this on something command line only? Feb 15 15:37:50 nope Feb 15 15:37:53 cool Feb 15 15:37:57 (I don't use any) Feb 15 15:37:57 I think there are some bitbake guis nowadays but I haven't tried them Feb 15 15:38:01 but there's hob Feb 15 15:38:06 (and, iirc, there is an ncurses one anyway) Feb 15 15:38:16 I've seen gtk :) Feb 15 15:38:34 video demo on ELCE at Intel booth Feb 15 15:38:38 * cryptk is what is hob? (pardon my ignorance) Feb 15 15:38:48 google? :) Feb 15 15:39:04 it's one of the bitbake guis Feb 15 15:39:14 http://www.linuxfordevices.com/c/a/News/Yocto-11-released/ Feb 15 15:39:20 hob can hog the server ;) Feb 15 15:39:21 tried google, I see lots of pull requests into the hob code, but nothing yet that shows that it is Feb 15 15:39:21 https://wiki.yoctoproject.org/wiki/BitBake/GUI/Hob Feb 15 15:39:31 ahh, I just found that linuxfordevices link actually Feb 15 15:40:09 so, is it worth it for me to go for yocto? or just vanilla OE? Feb 15 15:42:45 ask rather on #yocto :) Feb 15 15:42:52 can't help you with that, I use Angstrom, which is based on oe-core + some meta-meta-meta-layers Feb 15 15:43:16 I've never tried yocto... Feb 15 15:44:44 I guess that depends on what you plan on doing with it. Feb 15 15:45:44 well, eventually building the open source version of webOS ;) Feb 15 15:48:43 hm, you might not be the first with such idea Feb 15 15:49:13 I've seen some interest about webOS on #openmoko-cdevel Feb 15 15:49:51 it's SHR distro based on oe Feb 15 15:50:08 maybe that you should ask there to not duplicate some efforts? Feb 15 15:50:12 GNUtoo-desktop: ^ Feb 15 15:50:26 well, this will be for my own little side project ;) Feb 15 15:50:34 ok then :) Feb 15 15:50:36 much of the work that I am doing won't really apply Feb 15 15:58:03 hi Feb 15 15:58:27 cryptk, hi Feb 15 15:58:30 ynezz, thanks Feb 15 15:58:47 hola Feb 15 15:58:47 cryptk, just get commit access to meta-smartphone and do your own layer? Feb 15 15:59:17 + put the common stuff in openembedded-core or meta-openembedded Feb 15 15:59:27 that's the thing... my end goal has nothing to do with that really... it involves webOS, but getting it building with OE is just to learn about how to build it Feb 15 15:59:56 ok Feb 15 16:00:05 much of the work that I am planning on doing would cause a ton of un-necessary cruft for the vast vast majority of people Feb 15 16:00:18 but a lot of people will want to use your work done with oe+webos Feb 15 16:00:19 so I am weary of trying to upstream it Feb 15 16:00:24 true Feb 15 16:00:47 I guess it is more my webOS source hacks that most people won't want Feb 15 16:00:56 so just do it the right way within a layer(that works for you and for the people interested in it) Feb 15 16:01:10 I am new to all of this, so I don't even know what a layer is, lol Feb 15 16:01:12 maybe do a common layer Feb 15 16:01:17 and your hacks on top of it Feb 15 16:01:22 in another layer Feb 15 16:01:24 * cryptk needs to find an OE/bitbake for dummies guide somewhere Feb 15 16:01:39 basically recipes are splitted in layers nowadays Feb 15 16:01:54 splatted ? Feb 15 16:02:32 you have openembedded-core which contains the base, meta-openembedded that contains the shared stuff on top of the base Feb 15 16:02:37 and some extra layers Feb 15 16:03:14 http://www.openembedded.org/wiki/LayerIndex Feb 15 16:03:53 you'll need meta-palm at least Feb 15 16:04:00 + openembedded-core Feb 15 16:05:15 I'm gonna worry about getting OE actually setup first, lol Feb 15 16:06:18 cryptk, what's the machine you're building for? Feb 15 16:06:40 that's the trick, it's going to be building for an x86_64 platform Feb 15 16:06:52 ok Feb 15 16:06:55 I am mainly just messing around with the OE stuff to learn how webOS is built Feb 15 16:06:58 so no need of meta-palm Feb 15 16:07:12 the end goal is (for now) a secret ;) Feb 15 16:07:22 ok, maybe an x86 tablet Feb 15 16:07:24 but hopefully will be ready for SCaLE next year ;) Feb 15 16:07:28 ok Feb 15 16:07:42 cryptk, also note that not all the webos source code has been released yet Feb 15 16:07:42 depending on how long it takes HP to get all the sources out Feb 15 16:07:47 indeed Feb 15 16:07:51 yep, cryptk is in webos-internals ;) Feb 15 16:08:10 cryptk, also note that with oe can generate not only ipk but also deb and rpm Feb 15 16:08:20 yep, and I plan to leverage that Feb 15 16:08:24 ok Feb 15 16:15:54 hi otavio Feb 15 17:01:21 How do you use a .bbappend to change the kernel recipe's defconfig? If both files are named "defconfig" it just picks the original one from the kernel folder rather than overriding it. Feb 15 17:02:09 icanicant: are you prepending the path to FILESEXTRAPATHS? Feb 15 17:02:52 icanicant: you need to do that in your bbappend... i.e. FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:" if the defconfig is in a ${PN} named subdir Feb 15 17:03:42 bluelightning: ah, I set FILESEXTRAPATHS but i don't think i used _prepend. Feb 15 17:10:29 bluelightning: makes perfect sense, thanks. Feb 15 17:11:06 it is not openembedded related but maybe anyone knows it.. is there a kernel drive which creates an uart device using gpios? Feb 15 17:12:48 the .cfg kernel configuration files look useful. didn't seem to work for me though, do you have to use the yocto kernel for this? Feb 15 18:02:59 Does anyone know where the kernel config fragments for oe-core come from? Feb 15 18:03:11 The meta/cfg/kernel-cache/bsp/*/*scc files doesn't seem to be in the linux-yocto-3.2 tree. Feb 15 18:05:19 ah, they seem to be in the 'meta' branch Feb 15 18:05:20 http://git.yoctoproject.org/cgit/cgit.cgi/linux-yocto-3.2/tree/meta/cfg/kernel-cache/bsp?h=meta Feb 15 19:16:50 someone using ld gold on armv4t? Feb 15 21:06:18 evening all Feb 15 21:07:31 he pb Feb 15 21:33:57 hey pb_, woglinde_ Feb 15 21:34:31 hi woglinde_ Feb 16 01:30:50 am i right in thinking that module_autoload_ is not working with systemd at the moment? it adds to /etc/modutils but not /etc/modules-load.d which means you still need to use /etc/init.d/modutils.sh to auto-start. **** ENDING LOGGING AT Thu Feb 16 02:59:57 2012