**** BEGIN LOGGING AT Sun May 30 02:59:56 2010 May 30 07:15:57 03Khem Raj  07org.openembedded.dev * red878a05bc 10openembedded.git/recipes/linux/ (2 files in 2 dirs): May 30 07:15:57 linux_2.6.34.bb: Fix the boot issue due to naked functions when compiled with gcc 4.5 May 30 07:15:57 Signed-off-by: Khem Raj May 30 07:21:20 khem: ^ this means that -fno-ipa-ira is not needed anymore for kernel? May 30 07:36:13 khem: and there is probably extra "patches" in SRC_URI "file://patches/use-noclone-..." May 30 07:46:47 03Khem Raj  07org.openembedded.dev * r57ae3a86f9 10openembedded.git/recipes/linux/linux_2.6.34.bb: May 30 07:46:47 linux_2.6.34.bb: Correct the patch path for use-noclone-attribute-for-naked.patch May 30 07:46:47 Signed-off-by: Khem Raj May 30 07:46:54 JaMa: hey May 30 07:47:04 yes just figured out the paste error :) May 30 07:47:19 JaMa: can you try kernel with new patch on your box May 30 07:47:20 * JaMa trying that patch atm ;) May 30 07:47:26 and let me know if it fixes ok May 30 07:47:30 I will wait May 30 07:47:34 how long will it take you May 30 07:47:38 I have to go to bed soon May 30 07:47:46 tomorrow is my daughter's bday May 30 07:47:56 and -fno-ipa-ira in kernel's Makefile is not needed, right? May 30 07:48:05 yep May 30 07:48:13 you need nothing just this patch May 30 07:48:24 it's built already.. so just need a bit juice to my battery to boot spitz (2-5 mins) May 30 07:49:39 ok *fingers crossed* May 30 07:54:36 grr few more mins.. cannot get IP from dhcp and I don't have sd card reader here .. May 30 07:58:53 reboot.. May 30 08:01:15 heh ok np May 30 08:01:19 I am still here May 30 08:01:23 seems fine sofar May 30 08:01:30 I wont sleep unless I hear it booted May 30 08:02:06 booting debian rootfs, but it booted fine now May 30 08:02:43 cool May 30 08:02:55 enjoy May 30 08:03:14 now you can recompile whole rfs with 4.5 and see what happens May 30 08:03:16 thanks a lot man! you're great! May 30 08:03:38 I have it prepared.. just need to cp the kernel there from internal hd to sd :) May 30 08:03:53 you mean rfs with 4.5 ? May 30 08:04:00 yes May 30 08:04:20 and with -fno-ipa-ira in FULL_OPTIMIZATION May 30 08:05:15 oh you did not need that for root file system but hey who knows we only booted a kernel thus far :) May 30 08:06:49 rebooting to 4.5 rootfs May 30 08:07:58 lots of segmentation faults May 30 08:09:21 seems like init itself died too May 30 08:11:44 hmm now that could be another beast May 30 08:12:03 I guess busybox May 30 08:12:16 btw. I am building rfs with 4.5 so I will know more tomorrow May 30 08:12:32 great, thanks again and good night May 30 08:17:22 good night May 30 08:17:41 * khem -> sleep() May 30 10:52:56 how much disk space should be required for building console-image ? May 30 10:54:24 about 20G if you're building all at once and rm_work is called only in the end May 30 10:55:00 is there a way to call rm_work earlier? May 30 10:56:57 well, or in other words, is the workdir still necessary after a certain package was staged, packaged and everything? May 30 10:59:00 I usually call multiple targets before building final image; ie bitbake -k virual/kernel; bitbake -k console-image; will make tmpdir much smaller (because of calling rm_work after kernel) May 30 10:59:21 even better with git-native; gcc-cross; virtual/kernel May 30 10:59:44 git-native in Angstrom is huge (around 13G) May 30 11:00:33 JaMa, i think your 20GB estimate is low; on my 26GB disk it got full before half of console image May 30 11:00:47 on my 40gb disk i'm currently at 23GB May 30 11:01:13 mhm, so it's worth using git-native May 30 11:03:17 thanks for the workaround, JaMa May 30 11:08:42 CMoH-notebook: ah sorry it was estimeate only for tmpdir not whole (as I have downloads,tmpdir and the rest on separate partitions) May 30 11:10:41 JaMa, my current goal is to determine a spec for our build machines. being used with portage cleaning up after each package is built, I found myself surprised to see that bitbake and rm_work doesn't do that May 30 11:10:52 and only at the end May 30 11:12:52 so that's why I'm asking if this implementation of rm_work is the best-so-far implementation? or is there a build process requirement for workdirs to be available until the last package is staged? May 30 11:14:09 but for the time being splitting the build in multiple targets should work May 30 11:15:44 thanks again, JaMa May 30 13:00:56 03Martin Jansa  07org.openembedded.dev * r165079ba2c 10openembedded.git/recipes/kexecboot/ (2 files in 2 dirs): May 30 13:00:56 linux-kexecboot-2.6.34: add patch for building with gcc-4.5.0 (from linux_2.6.34) May 30 13:00:56 Signed-off-by: Martin Jansa May 30 13:48:52 khem: replaced gcc-4.5.0 built busybox with the same built with gcc-4.4.4 and rootfs is usable again (opkg install coreutils doesn't end with lots of segmentatin faults and now it booted fine). Whole rootfs was built with -fno-ipa-ira. I'll try another build from scratch without it. May 30 14:02:04 * JaMa rebuilding from scratch again May 30 14:20:24 khem: and broken busybox's sed seems known bug http://bugs.archlinux.org/task/19403 May 30 14:21:01 JaMa: do you use archlinux ? May 30 14:25:51 djszapi: no May 30 14:26:07 khem: and proper fix is in gcc svn http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43987 May 30 14:45:31 Build of qemuarm is very funny .... it is building multiple kernels and I remove each not needed recipe to stop this behavior. The zImage image that has been built is not also booting! May 30 14:48:07 khem: I've updated gcc-4.5.0.inc to use newer snapshot from 4.5 branch, is it worth pushing if everythings builds and works fine or do you prefer to wait for 4.5.1 release? May 30 14:53:26 JaMa: How can I stop oe from building multiple kernels for qemuarm? May 30 14:54:36 JaMa: The kernel image for 2.6.33 got built but it does'nt boot! Any clues? May 30 14:57:56 is PREFERRED_PROVIDER set for virtual/kernel? May 30 14:58:37 iirc khem updated qemu machines to use linux-2.6.34 May 30 15:07:24 shazkhan: it was updated ~22 hours ago.. so check you have latest oe.dev May 30 15:08:30 JaMa: I'll update it if it should improve the conditions ... but I have used this commit to build for gta02. Updating .... May 30 15:22:30 JaMa: What do you think about http://www.angstrom-distribution.org/narcissus/? May 30 15:51:02 JaMa: cool. May 30 15:51:53 JaMa: locally I am building 4.5 from branch only:) May 30 15:52:13 JaMa: for gcc in future I only want to have one recipe for one major release May 30 15:52:24 so I have renamed the recipes too May 30 15:58:58 hi Khem, JaMa May 30 15:59:05 03Andrea Adami  07org.openembedded.dev * rd963abc321 10openembedded.git/recipes/linux/linux-rp.inc: May 30 15:59:05 linux-rp.inc: Rename getline in unifdef.c to parseline to avoid conflict. May 30 15:59:05 * sync with linux.inc May 30 15:59:05 * see https://patchwork.kernel.org/patch/11166/ May 30 15:59:05 * committed in 2.6.29 (commit d15bd1067b1fcb2b7250d22bc0c7c7fea0b759f7) May 30 15:59:15 03Andrea Adami  07org.openembedded.dev * r4a8053b1e5 10openembedded.git/recipes/klibc/ (klibc_1.5.15+1.5.16.bb klibc_1.5.15+1.5.16.inc): May 30 15:59:15 klibc: bump to 1.5.15+1.5.16 May 30 15:59:15 * tested on arm with 2.6.26 and 2.6.33, 2.6.34 kernels May 30 15:59:19 JaMa: pls test once more ^^ May 30 16:00:08 khem: I had to make a patch depending on kernel version, probably makefile hack would be more elegant May 30 16:00:35 at least we moved from 1.5.15 meadows ;) May 30 16:01:15 JaMa: there are some strangeness in klibc-utils... May 30 16:01:55 insmod, dmesg, sync and ls are not installed ? May 30 16:02:13 ok, those come from OE patches, not upstream May 30 17:34:35 ant__: I'm using only 1.5.18 version, is it worth trying that old klibc? May 30 17:38:49 khem: I've moved gcc-4.5.0.inc to that snapshot with PR bump (too lazy to update PREFERRED_VERSIONS because of that change) so I guess this is not acceptable http://gitorious.org/~jama/angstrom/jama-shr-experimental/commit/fba694e6011c7d14760d2b39f21019dfb214b333 and I should at least git mv it with better PV May 30 19:11:12 JaMa: to use 1.5.17 I have to revert the socket patch :/ May 30 19:11:48 and apply a signal-cleanup for arm May 30 19:12:44 JaMa: /include/sys/socket.h:146: error: redefinition of typedef 'sa_family_t' -> this was mainlined in 1.5.17 May 30 19:13:07 I dunno how sane is reverting back to old behavior to make it 2.6.26 cokpatible... May 30 19:18:14 JaMa: the good news is 1.5.18 fails in the same manner... May 30 19:18:44 but works for you with newer kernel right? May 30 19:18:55 only testing on 2.6.26 May 30 19:19:59 it is the socket.h patch May 30 19:20:11 I'll have to apply it in negative May 30 19:20:48 which will probably break support of newer kernels May 30 19:20:51 JaMa: at that point we could remove 1.5.17 May 30 19:21:10 JaMa: I'll make conditional as in previous patch May 30 19:22:10 how can this last patch work properly? May 30 19:22:18 actually we could remove 1.5.15 and 1.5.17 May 30 19:22:35 +# upstream in 2.6.33 May 30 19:22:36 +SRC_URI += '${@base_version_less_or_equal("PV", "2.6.32", "", "file://socket.h.patch", d)}' May 30 19:22:50 you're testing klibc version ($PV), aren't you? May 30 19:22:59 I built 1.5.17 here May 30 19:23:18 do you think we should add a var inbetween? May 30 19:23:43 build it with kernel > 2.6.32 first May 30 19:23:58 JaMa: omg..I was using the same func we have in kexecboot.inc... May 30 19:24:27 I'll correct this now May 30 19:24:28 this change means that you removed socket.h.patch from klibc < 2.6.32 May 30 19:24:42 it is only for >=2.6.33 May 30 19:25:00 but KERNEL >=2.6.33 not klibc.. May 30 19:25:12 sorry... May 30 19:26:26 * ant__ hates working on 2.6.26... May 30 19:26:51 that function in linux-kexecboot.inc is OK May 30 19:27:04 I'll just rename it May 30 19:27:22 because it checks PV of linux-kexecboot recipe thus testing for > 2.6.31 makes sense May 30 19:27:32 I don't see how renaming will help May 30 19:27:47 not only that May 30 19:28:01 renaming avoids naming clash if you include linux-kexecboot May 30 19:30:20 * JaMa still doesn't see how to check version of other recipe inside klibc May 30 19:30:30 but sure, we need KERNEL_MAJOR_VER May 30 19:30:41 I'm checking which KERNEL_* May 30 19:32:48 hmm still don't understand, sorry May 30 19:33:22 just checked bitbake -e -b klibc_1.5.18.bb and closest variable is MACHINE_KERNEL_VERSION="2.6" which won't help May 30 19:34:21 and better is PREFERRED_VERSION_linux="2.6.34", but it's not safe to expect it to be defined and it can be ie PREFERRED_VERSION_linux-rp instead May 30 19:34:40 KERNEL_VERSION = "${@get_kernelversion('${B}')}" May 30 19:34:49 in klibc recipe? May 30 19:34:54 kernel.bbclass May 30 19:35:07 I'm grepping in the classes May 30 19:36:42 yes.. but if you look what this function does.. it's not possible to use inside klibc recipe May 30 19:37:30 ./kernel.bbclass: kernelver = bb.data.getVar('KERNEL_VERSION', d, 1) May 30 19:37:37 and even if it does work (and klibc has /include/linux/utsrelease.h included) then it's checking something else than what kernel version is built in sysroots May 30 19:38:07 JaMa: the we'll have similar issues May 30 19:38:49 where? I'm a bit sleepy, so maybe I'm overlooking something May 30 19:39:22 look only this May 30 19:39:23 http://fr.pastebin.ca/1874446 May 30 19:39:36 we'll have to work around other things May 30 19:39:53 ant then, again, that socket patch upstream... May 30 19:40:03 they know compat is broken :/ May 30 19:40:28 perhaps we can stop at 1.5.15+1.5.16 for older kernels May 30 19:40:30 are we talking about checking in recipe or in code? May 30 19:40:45 eh..in code is easy sometimes... May 30 19:40:50 sometimes not May 30 19:41:23 we are swimming against the stream... May 30 19:41:28 don't think about code May 30 19:41:29 in code it's easy as shown in that pastebin May 30 19:41:31 it's easy May 30 19:41:47 make -DSOMEDEFINE :) May 30 19:42:04 but you were talking about bbclasses and recipes and 4a8053b1e522a31ad7e008c9dac723998bd62491 is just wrong May 30 19:42:58 let me see if I can quickly fix it May 30 19:43:46 and that's the hard thing in recipe, but you can update socket.h.patch with check inside May 30 19:44:05 and then include it in SRC_URI in call cases May 30 19:44:15 finally, do you think we should insist and support 2.6.26 in klibc 1.5.18? May 30 19:44:38 seems not so far, really... May 30 19:45:14 as you know some Z models are still 2.6.2x :/ May 30 19:47:08 no idea how hard would be to build klibc_1.5.18 on 2.6.26 kernel headers, but I expect it to be very difficult (but you track klibc development and requirements more than I do.. ) May 30 19:49:05 what is main problem with building new klibc on old headers? May 30 19:58:30 atm two patchs : the signal patch I linked above May 30 19:59:16 and a second patch reverting socket.h patch of 1.5.16 wich got mainlined May 30 20:04:56 do we have some methods to 'glue' versions of software? May 30 20:05:32 e.g. 'build this initrafms with that kernel and that klibc' May 30 20:08:05 well, setting preferred version in distro May 30 20:11:57 this will glue all distro to one versions 'chain' May 30 20:12:12 I mean rather DEPEND's May 30 20:13:02 e.g. DEPEND += 'linux-kexecboot-2.6.26 klibc-1.5.16' May 30 20:27:15 The toolchain built with OE can be used from anywhere? I mean if I copy it to my disk where I store my deployments and toolchains ... May 30 20:28:45 Still I will need a script to handle the environment. Right? May 30 20:29:36 JaMa ... is this possible? May 30 20:30:39 How do openmoko guys provide a toolchain similar to this one? May 30 20:33:00 check this http://www.angstrom-distribution.org/simplified-development-setup May 30 20:33:15 and this http://www.angstrom-distribution.org/toolchains/ May 30 20:33:26 may be you will find answer there :) May 30 20:37:22 Jay7: I just went through http://www.angstrom-distribution.org/narcissus/ May 30 20:37:36 I checking your tips May 30 20:43:37 jay7: I found two recipes for toolchain .... one was met and the other was external ... whats the difference? May 30 20:43:57 I don't know :) May 30 20:44:03 k May 30 20:45:07 may be external toolchain is toolchain build somewhere else May 30 20:45:21 not in current build env/session May 30 20:45:43 at least you can ask google or wikipedia :) May 30 20:46:00 Maybe that is something that I require ... bb in progress May 30 20:46:38 oe dudes are much cool than google and wikis May 30 20:47:59 * ant__ is forced to write spaghetti-code for that patch :/ May 30 20:51:02 Jay7: The external toolchain seems to be the one that is to be used imported for OE builds May 30 20:52:09 shazkhan: looks like May 30 20:52:37 is that a comment or a question? May 30 20:52:51 it is external relative of OE May 30 20:53:06 *relative to May 30 20:53:12 that was comment :) May 30 20:53:18 :) May 30 21:29:20 anybody working with clutter? trying to get a "stable" 1.2.8 recipe out. May 30 21:37:28 03Michael 'Mickey' Lauer  07org.openembedded.dev * r7b117a45ac 10openembedded.git/recipes/vala-dbus-binding-tool/ (2 files): vala-dbus-binding-tool: remove obsolete version 0.1.5 and bump git version to 0.3.1 May 30 21:37:38 03Michael 'Mickey' Lauer  07org.openembedded.dev * r11be21a228 10openembedded.git/recipes/vala/vala_0.8.1.2.bb: vala: add 0.8.1.2 May 30 21:38:41 thanks mickeyl for adding 0.8.1.2 May 30 21:39:12 HopsNBarley, sorry I don't know clutter,I even never used it but it would be nice to have it in openembedded May 30 21:39:33 :-) May 30 21:43:55 03Andrea Adami  07org.openembedded.dev * rd37b0c814f 10openembedded.git/recipes/klibc/ (klibc-1.5.15+1.5.16/socket.h.patch klibc_1.5.15+1.5.16.inc): klibc: rework socket.h.patch for v. 1.5.15+1.5.16. Fix ugly c&p bug. May 30 21:44:07 JaMa: still ugly but better now May 30 21:49:07 ant__: yeah, thanks May 30 21:49:15 ant__: +inherit linux-kernel-base is needed? May 30 21:49:20 now thinking to rename it 1.5.16 May 30 21:49:22 ant__: or left-over from tests? May 30 21:49:30 sh@t May 30 21:49:51 ok, I'll rename and correct this second conceive&push error May 30 21:52:00 I'm looking here.. May 30 21:52:11 unsure we have all patches for 1.5.16 May 30 21:52:15 can't remember May 30 21:52:24 checking now... May 30 21:57:03 JaMa: I think we are at utils: add simple ls May 30 21:57:21 last commit of 2009 May 30 22:00:26 03Andrea Adami  07org.openembedded.dev * r9040854409 10openembedded.git/recipes/klibc/klibc_1.5.15+1.5.16.inc: klibc: remove forgotten cruft in klibc_1.5.15+1.5.16.inc May 30 22:07:36 HopsNBarley: hi, long time no see Andrew May 30 22:08:07 hey leon! yeah, been doing other (misguided :-) things for a while - back to working in embedded. May 30 22:08:19 HopsNBarley: all cool :-) May 30 22:09:06 HopsNBarley: welcome back in OE. While you were away, we made a big mess out of it :-) May 30 22:11:40 lol, I just read about Lunch 2.0, never heard of it before. May 30 22:25:01 hm.. May 30 22:25:09 NOTE: The checksums for .. did not match. May 30 22:25:20 Expected and Got May 30 22:25:24 are imho swapped... May 30 23:08:03 Jay7: klibc_1.5.16 compiled vs. 2.6.26 May 30 23:08:08 step by step... May 30 23:17:10 03Andrea Adami  07org.openembedded.dev * r3f8149a4f0 10openembedded.git/recipes/klibc/ (8 files in 2 dirs): klibc: initial commit of ver. 1.5.16. Tested on arm. May 30 23:28:35 03Andrea Adami  07org.openembedded.dev * rc6a662cd60 10openembedded.git/recipes/klibc/ (4 files in 2 dirs): May 30 23:28:35 klibc: add some patches to 1.5.17 and make it build vs. old kernels. May 30 23:28:35 * increase default preference, tested on arm May 30 23:40:19 03Andrea Adami  07org.openembedded.dev * r921f2ae278 10openembedded.git/recipes/klibc/ (4 files in 2 dirs): May 30 23:40:19 klibc: add some patches to 1.5.18 and make it build vs. old kernels. May 30 23:40:19 * increase default preference, tested on arm May 30 23:40:37 Jay7: here we are May 31 00:00:28 Jay7: kexec/kexecboot still working compiled against klibc_1.5.18 :D May 31 00:00:34 good night May 31 02:46:09 I'm just getting started with Embedded Linux. I have a rugged PC device, 1GHz fanless brick, boots from Compact Flash. Have a 512MB CF card booting DSL Linux. What Linux should I put on this? I want a good terminal, VT100, and ability to run Java would be nice. Any sugestions? May 31 02:53:11 uchobby, sounds like you want to play. Try everything you can get your hands on. May 31 02:53:28 I do want to play grg May 31 02:54:25 I could use DSL (dam small linux) but I want to install a terminal app and apt-get does not seem to be there. May 31 02:55:46 i suggest you look at the docs for DSL. I would be surprised if it didn't have some kind of package management May 31 02:56:28 btw, this channel is not really for embedded linux in general. It is for OpenEmbedded, a framework for building embedded linux distributions. May 31 02:57:19 There are several OE based distros you could try on your system. Check the OE site. **** ENDING LOGGING AT Mon May 31 02:59:56 2010