**** BEGIN LOGGING AT Tue Mar 06 02:59:58 2012 Mar 06 06:08:49 hi Mar 06 06:09:27 any idea how to specify SRC_URI in bb version 1.10.2 Mar 06 06:09:43 for source in local system Mar 06 06:10:23 iin 1.8 i use to specify SRC_URI = "file://../../../build/sources/git-${PV}.tar.bz2" in this way Mar 06 06:10:38 but its not working in 1.10.2 Mar 06 07:01:13 any idea about this issue Mar 06 09:24:08 good morning Mar 06 11:34:18 had anyone success with compiling openjdk-6 , i get allways segfaults with qemu in do_compilesteptwo Mar 06 11:38:35 <_tasslehoff_> in oe-core fsck is ran automatically on partitions at boot (when needed), but not on oe-classic. is this something I must enable / turn on someplace? Mar 06 11:40:39 _tasslehoff_ man fstab. its the fitht and sixth field , that tunes this behavior, i assume oe-core and oe-classic are using difftenyt templates Mar 06 11:49:00 <_tasslehoff_> pwgen: actually I have the same fstab in both cases. only difference is that I use relatime on oe-core and noatime in oe-classic Mar 06 11:49:36 _tasslehoff_ : systemd or clissic sysvinit ? Mar 06 11:52:44 usually the /etc/init.d/checkroot.sh script is responsible doing a check on boottime Mar 06 11:52:57 <_tasslehoff_> pwgen: classic sysvinit. systemd on the oe-core that works Mar 06 11:55:06 there should be a symlink in /etc/rcS.d that points to /etc/init.d/checkroot.sh. this "should" make this checkon boot time . Mar 06 11:57:07 <_tasslehoff_> pwgen: that script is ran, I've seen some of the prints from it. will have to debug it to find out what fails. thanks. Mar 06 11:57:22 np[ Mar 06 12:26:21 <_tasslehoff_> seems to be some differences between the checkroot scripts in core and classic, and the fact that I have ENABLE_ROOTFS_FSCK=no in /etc/default/rcS likely doesn't help Mar 06 12:28:56 *FG* i think you are right "ENABLE_ROOTFS_FSCK=n" sounds like something that disable the ROOTFS_CHECK ... Mar 06 12:49:34 hi Mar 06 12:50:16 i'm working with i2c from kernel space, using linux/i2c-dev-user.h Mar 06 12:51:04 i include the file and work fine with it when compiling native on my beagleboard, but i don't know how to do stuff on my pc, since i cannot find this header Mar 06 12:51:28 i'm using Linux rrosa-X220 2.6.38-13-generic #56-Ubuntu SMP Mar 06 12:52:17 rlros install required kernel headers. or the i2c dev packages ( this should install the required header files. Mar 06 12:53:16 pwgen, i have i2c-tools and libi2c-dev, can't find any other i2c related packages Mar 06 12:55:44 rlrosa, lm-sensors may help Mar 06 12:58:53 do we have any support for karo-elektronics.de boards , like tx25 tx28 ? Mar 06 13:19:06 rob_w: oe classic should have support for the tx25. it is likely to be out of date, but replacing the board patch should be pretty easy. Mar 06 13:39:15 Is someone working at xinput-callibrator improvement? Mar 06 13:43:53 rob_w: Do you know by any chance what's the price of those modules (and the base board, Starter-Kit V)? Mar 06 14:26:39 ynezz, no idea about pricing .. Mar 06 14:27:52 thanks florian i just found it that some evm board support existis . those should be fine to start with .. thanks btw for the embedded world suggestions .. as you see i take them and we will try to based on those boards our "lowend" products Mar 06 14:46:00 morning all Mar 06 14:56:53 hi, all! Mar 06 14:57:11 I'm in zillion problems with python in oe-core again Mar 06 14:58:17 when building _curses modules it for some reason includes curses.h from -native curses instead one in arm sysroot, which fails quite fast with inability to include curses-32.h Mar 06 14:58:51 when i just put proper curses.h into python tree includes, it compiles properly. Mar 06 14:59:34 also, weird problems with hashlib, and attempts to look for libcrypto.so in native directory Mar 06 15:00:04 so python is basically unusable for me now Mar 06 15:00:50 I use amd64 as build host, and everyone else seems to be using 32-bit build host so no problems are seen Mar 06 15:01:04 and nobody uses python for that matter. Mar 06 15:07:00 well, I use a 64-bit build host, but I don't think I use python on the target. Mar 06 15:07:17 (I do build it, because it's a silly build dependency of something or other, but I don't think it ever gets installed.) Mar 06 15:15:39 so to confirm that target python is in non-working unmaintained state for a long time Mar 06 15:19:22 and libcrypto.so is never installed by openssl package, which seems to be reason for broken hashlib module in python Mar 06 15:20:23 with breakage of hashlib python will not work with even with that little use it had before Mar 06 15:21:32 so it means basically there's no python support with OE and I ask what id does withi oe-core, which should be rock-solid by design? probably its place is in meta-oe for its quality. Mar 06 15:23:34 sorry, trying to fix all this breakage now, and a bit frustrated, thus typing errors. Mar 06 15:24:13 I mean I expected that oe-core is tested, but it seems to be tested for very specific use cases. Mar 06 15:24:23 bad luck this time Mar 06 15:27:36 saying target python is 'broken' because a couple of its modules are is not accurate Mar 06 15:27:53 but it probably is accurate to say it doesn't get the attention it should. maybe you're volunteering to do so? ;) Mar 06 15:31:14 kergoth: well, it is hard to volunteer these days, as I find the root of problem I will submit patches, but I will do just that. Mar 06 15:32:23 kergoth: technically python is not usable for anything at the moment. Probably will work for some hello-worlds. Mar 06 15:32:59 'anything'? really? you tried every non-trivial project? Mar 06 15:33:21 well, yes, so why I try to fix it Mar 06 15:34:06 that's so stupid and illogical I don't even know where to begin. Mar 06 15:34:07 I can't say there's no way it will work for something, but this something is hard to find Mar 06 15:34:13 but I'm sure your contributions will be welcome Mar 06 15:36:01 well, I'm not that optimistic. There are so many ways for patch to be rejected and I will have time for only one attempt. Mar 06 15:37:10 is there a recipe for subprocess (python) in recipes? Thank you Mar 06 15:37:37 what is the difference between python and python-native? Mar 06 15:39:22 Spider-Pork: python-native is for host Mar 06 15:39:41 Spider-Pork: it is part of stock python (subprocess) Mar 06 15:39:53 stock python? Mar 06 15:39:57 Spider-Pork: probably try to install python-modules package Mar 06 15:40:05 Spider-Pork: yes Mar 06 15:40:27 python recipe will build it for you Mar 06 15:40:49 python recipe is present but if i do: import subprocess it can't find it Mar 06 15:41:07 add MACHDEP=linux2 to EXTRA_OEMAKE in python.inc in case if won't Mar 06 15:41:13 but i not have python-modules installed Mar 06 15:41:26 Spider-Pork: try to install python-modules package and in particular python-misc Mar 06 15:41:28 what? Mar 06 15:41:36 ok i'll try it Mar 06 15:42:46 (thank you slapin) Mar 06 15:45:37 and again .a file in /lib. where this world is going to? Mar 06 15:46:39 slapin: i'm a bit newbie. What are you talking about? Mar 06 15:47:07 Spider-Pork: bad openssl packaging, that's why hashlib is broken on oe-core python. Mar 06 15:47:22 ah good... Mar 06 15:48:12 i have no .a file in my /lib Mar 06 15:49:27 and i haven't no python-modules.bb Mar 06 15:49:47 so i don't know howto generate python-modules Mar 06 15:53:24 Spider-Pork: you should understand recipes vs packages Mar 06 15:53:39 Spider-Pork: single recipe can generate multiple packages Mar 06 15:53:44 python is the case Mar 06 15:53:57 you build python and it generates lots of packages Mar 06 15:54:20 Spider-Pork: sho look in your tmp*/deploy/ipk Mar 06 15:54:27 ah ok Mar 06 15:54:38 thank you slapin! I'll study it better :) Mar 06 15:54:43 also you might have your device connected to internet Mar 06 15:55:14 and do opkg update; opkg upgrade if you have packaging installed and some luck Mar 06 15:55:40 or don't do upgrade and directly do opkg install python python-modules Mar 06 15:55:55 thay might work for you Mar 06 15:56:34 ok thank you slapin :) your help is like the rain in the desert Mar 06 15:56:35 or you can copy python-*ipk files to your device and do opkg install python-*ipk on it directly Mar 06 15:57:22 yep i've done so for install packages from host build machine Mar 06 15:57:44 bitbake yadda than scp ipk and opkg install Mar 06 15:58:06 Spider-Pork: yeah Mar 06 15:59:08 I've found the packages, cool :) Mar 06 15:59:19 a big thank to you slapin :) Mar 06 16:01:14 Spider-Pork: my pleasure Mar 06 16:02:22 if you need some good places where to eat in Italy i'll be glad to help you :) Mar 06 16:02:36 anyway I found why hashlib was broken, that's openssl packaging of devel stuff in /lib which is again problem. _curses module failure is weird, so that's for later. Mar 06 16:02:50 just had to revert some patches. Mar 06 16:05:25 God I hate computers at times Mar 06 16:05:51 opkg_install_pkg: Package python-modules md5sum mismatch. Either the opkg or the package index are corrupt. Try 'opkg update'. :/ Mar 06 16:08:48 Spider-Pork: just remove package lists and try again Mar 06 16:09:03 if you install from local package Mar 06 16:10:35 slapin: you mean *.conf in /etc/opkg? Mar 06 16:12:20 oh found it Mar 06 16:17:52 Spider-Pork: something inside /var/lib/opkg Mar 06 16:18:14 yep Mar 06 16:18:29 was present python. I've removed it Mar 06 16:18:45 Spider-Pork: /var/lib/opkg/lists/ Mar 06 16:19:39 list is not present Mar 06 16:19:56 I've removed python file and now i can install without problems Mar 06 16:22:13 Spider-Pork: ok then Mar 06 16:51:17 is it straightforward to get a glibc with hardware floating point for a cortex-A8 to build? (sadly not for a beagleboard, but for an AM3517_EVM) Mar 06 17:35:20 ../../gdk-pixbuf/gdk-pixbuf.h:31:18: fatal error: glib.h: No such file or directory Mar 06 17:35:31 gdk-pixbuf is unbuildble too Mar 06 17:35:49 even that I don't need it at all. Mar 06 17:36:26 it seems to ignore GLIB_CFLAGS Mar 06 17:36:38 so can't find header Mar 06 17:37:05 most of gtk-pixbuf is ok, only contrib directory is failing totally. Mar 06 17:41:41 I hate building in a vm Mar 06 18:03:40 Crofton: what is your host OS? Mar 06 18:04:30 ubuntu running VBoz with Fedora 16 Mar 06 18:06:35 Crofton: you may use lxc e.g. instead of vbox Mar 06 18:06:49 or openvz Mar 06 19:43:05 anyone an idea how I can get the arm compiler for my machine use Mar 06 19:43:06   Tag_ABI_align8_preserved: Yes, except leaf SP Mar 06 19:59:04 eFfeM: just use EABI and it should do it why do you ask ? Mar 06 19:59:06 8-byte stack Mar 06 19:59:08 alignment is required and preserved Mar 06 19:59:25 its a EABI requirement Mar 06 20:00:31 if you are having pure asm files do something like this Mar 06 20:00:43 #if defined(__ARM_EABI__) Mar 06 20:00:52 khem, I have a complete system that is build with buildroot, don't have sources, and I want to run those apps from my oe kernel Mar 06 20:00:55 .eabi_attribute 25, 1 Mar 06 20:00:58 #endif Mar 06 20:01:20 or be able to make a toolchain that will allow me to produce code for this box Mar 06 20:01:45 OK you should not need to do _anything_ then Mar 06 20:01:52 khem: http://pastebin.ca/2124867 Mar 06 20:02:05 first file is from the box, 2nd is what I made, this is armv6 hw Mar 06 20:02:24 that pastebin doesnt open Mar 06 20:02:30 Does it have some malware :) Mar 06 20:02:38 ah ok, will make a new one, Mar 06 20:02:46 no malware, expired i guess Mar 06 20:04:22 khem: http://pastebin.ca/2124889 Mar 06 20:04:36 first is what I got from the device, 2nd is the helloworld I made myself Mar 06 20:05:20 if I run the helloworld on the box I get "Killed", then compiled an oe kernel for the board runs nicely with init=/helloworld but refuses to work with the existing busybox Mar 06 20:07:40 OK whats your own toolchain versions Mar 06 20:09:52 i don't really know the original toolchain, maybe code sorcery actually only thing I know is that it was made with buildroot Mar 06 20:10:05 what about your ? Mar 06 20:10:15 mine, oe build gcc Mar 06 20:10:35 oe classic, tried both armv5 and armv6; hw is armv6 Mar 06 20:11:07 gcc is 4.5 ? Mar 06 20:11:25 yes Mar 06 20:11:44 anyway, i can generate my own rootfs, wanted to do this in smaller steps so decided first to try their binaries Mar 06 20:11:52 figured that they should be compatible but they are not Mar 06 20:14:05 so not really a problem, just did not understand why on this armv6 I could not run their v5 code Mar 06 20:14:33 don't worry about it, I know how to progress, and mixing stuff from different origins is probably not a good plan anyway Mar 06 20:18:50 what are your cmdline options to compile helloworld Mar 06 20:19:16 its valid thing thats the whole purpose of having EABI Mar 06 20:19:53 let me check, I used the std recipe Mar 06 20:20:50 khem this is what I used Mar 06 20:20:51 log.do_compile.7696:+ arm-oe-linux-uclibceabi-gcc -march=armv5te -mtune=xscale -mthumb-interwork -mthumb --sysroot=/home/frans/oe/tmp/sysroots/armv5te-oe-linux-uclibceabi -fexpensive-optimizations -fomit-frame-pointer -frename-registers -Os -pipe -g -c -o helloworld.o helloworld.c Mar 06 20:21:06 log.do_compile.7696:+ arm-oe-linux-uclibceabi-gcc -march=armv5te -mtune=xscale -mthumb-interwork -mthumb --sysroot=/home/frans/oe/tmp/sysroots/armv5te-oe-linux-uclibceabi -Wl,-O1 -Wl,--hash-style=gnu -o helloworld helloworld.o -static Mar 06 20:21:27 ok thats sane Mar 06 20:21:32 not sure what was used for the other file Mar 06 20:21:51 so what happens if you remove -Os and use say -O2 Mar 06 20:21:54 and not really an idea how I can find out; the only thing I have is an obj file Mar 06 20:22:15 hm ok, that might help Mar 06 20:22:49 will try that in a short while Mar 06 20:23:00 its an abi mandate so either you have it or not Mar 06 20:23:15 but since in current world of OE we do have it Mar 06 20:23:24 I want to understand why its not there in oe.classic Mar 06 20:23:29 when you are using gcc 4.5 Mar 06 20:23:54 what worried me in the pastebin I gave before that their code tells about alignment except leave SP and the OE binary does not Mar 06 20:24:32 holy crap... now I remember, why it takes fucking forever to bitbake an image – ⅔ of the time is wasted downloading all that java-crap in the gcc-branch >_< Mar 06 20:24:53 dFence: there are ways to avoid it Mar 06 20:25:01 tar files are snapshotted Mar 06 20:25:09 khem: hello, since some days ca-certificates download is broken, i sent a patch to angstrom mailing list, but in oe mailing list a patch should also be available do you know why this issue is not fixed ? Mar 06 20:25:37 khem: but they still include libjava and all that testsuite-crap, right? Mar 06 20:26:07 dFence: yes it does Mar 06 20:26:13 speak of the devil – I start bitching, gcc finished checking out ;) Mar 06 20:26:20 dFence: and we have option to turn on java in gcc if one wants Mar 06 20:26:27 and same for other lang frontends Mar 06 20:26:36 khem: ah... makes sense ;) Mar 06 20:26:41 its just that commonly used are c/c++ Mar 06 20:27:05 nschle85: it is with meta-oe ?? Mar 06 20:27:20 khem: yes Mar 06 20:27:29 nschle85: hmm Mar 06 20:27:32 khem: I didn't know there were recipes based on java Mar 06 20:27:39 send a ping if you are waiting for feedback Mar 06 20:28:18 dFence: there are, people might use gcj we just dont want to let that possiblity out Mar 06 20:29:31 eFfeM: can you paste the .s file from OE compiled helloworld Mar 06 20:29:33 khem: is there a way to only selectively checkout a project like gcc? would definitely speed out the building process imho Mar 06 20:29:43 no Mar 06 20:29:51 its a one time thing though Mar 06 20:29:54 khem, ok will do, one sec Mar 06 20:30:01 unless you keep deleting your dl_dir Mar 06 20:30:35 khem: I definitely have to do some heaving cleaning – OE got fat since the last time I used it ;) Mar 06 20:31:05 khem moved back to the armv6 compiler as this is v6 hw and I want to use that one (and it already shows to run under my kernel) Mar 06 20:31:06 dFence: there is work done in removing stale downloads from dl_dir Mar 06 20:31:35 khem: stale as in non-necessary anymore or one-time use only? Mar 06 20:31:41 eFfeM: doesnt matter which architecture it is as long as its arm Mar 06 20:31:59 dFence: it looks at access times Mar 06 20:32:02 khem, ok will do Mar 06 20:32:06 cool Mar 06 20:32:16 OE-maid ;) Mar 06 20:34:29 eFfeM: I think if OE gcc should generate the attribute Mar 06 20:34:36 if its not doing that its a bug Mar 06 20:35:22 eFfeM: wait a second you are using thumb ?? Mar 06 20:35:29 khem, yes Mar 06 20:35:35 hmmm Mar 06 20:36:05 that could mean different things then Mar 06 20:36:17 Does your box provide for interworking ? Mar 06 20:36:23 if not those apps wont run Mar 06 20:36:59 once I look at .s file I can tell more whats going on Mar 06 20:38:47 khem: here its is: http://patches.openembedded.org/patch/22283/ Mar 06 20:39:13 khem: the .s file: http://paste.ubuntu.com/872038/ Mar 06 20:40:41 nschle85: send a gentle ping Mar 06 20:40:45 or acked by Mar 06 20:40:50 if you have tested it Mar 06 20:41:00 this helps catch maintainers eye Mar 06 20:41:11 khem: ping to whom ? Mar 06 20:41:34 nschle85: reply to patch email Mar 06 20:41:41 khem not sure if there is interworking support, how can I find out ? Mar 06 20:41:52 or should I be better off disabling thumb ? Mar 06 20:42:42 eFfeM: .eabi_attribute 24, 1 Mar 06 20:42:42 .eabi_attribute 25, 1 Mar 06 20:42:47 so you have it Mar 06 20:43:07 ah ok, Mar 06 20:43:08 25 means preserve the 8 byte stack align Mar 06 20:43:18 and 24? Mar 06 20:43:45 This code may require 8-byte alignment from caller Mar 06 20:44:02 because the existing code has: Tag_ABI_align8_preserved: Yes, except leaf SP Mar 06 20:44:41 that is 25 Mar 06 20:46:15 ah ok Mar 06 20:47:58 can do another test now, originally it failed on init=/bin/sh, but now have a workaround for it Mar 06 20:49:48 khem, seems the old busybox does run, but does not do a thing if I run one of its applets it just returns Mar 06 20:50:53 eFfeM: thats your buildroot ? Mar 06 20:51:04 the busybox is from buildroot Mar 06 20:51:18 and whats not from buildroot Mar 06 20:51:56 the helloworld is made with oe Mar 06 20:52:29 and to complicate things in order to bring up things I used a root fs from debian (with armv4 executables) Mar 06 20:52:47 using that rootfs I can run my helloworld but not their busybox Mar 06 20:53:26 ok Mar 06 20:53:43 so buildroot busybox only run with busybox rfs and kernel Mar 06 20:54:14 yes Mar 06 20:54:47 OE runs ok with debian ? Mar 06 20:54:53 and that rfs will not run oe made images, and the oe kenrel with debian rfs will run oe make executables Mar 06 20:54:53 yes Mar 06 20:55:29 is it some illegal instr error or what do u see as failure Mar 06 20:56:55 oe image on buildroot system gives Killed in the shell, buildroot busybox on oe/debian system gives nothing, just a new prompt Mar 06 21:01:57 eFfeM: I think cause is something else Mar 06 21:02:05 ok Mar 06 21:02:15 you can try to build in arm mode Mar 06 21:02:18 and see if that helps Mar 06 21:03:25 actually it is not too much of a problem as I do not really need the old stuff, basically I'm moving to my own setup, but figured that this was a nice way to ttest my kernel, but apparently it was is not, and I do not really need the interoperability any more, this was mostly for testing Mar 06 21:03:31 but thanks a lot for your help Mar 06 21:10:14 eFfeM: ok. Mar 06 21:10:24 sometimes things get abused. Mar 06 21:10:31 I would blame buildroot here Mar 06 21:10:39 I learned a lot and made progress, so this is a good evening for me Mar 06 21:10:56 yeah buildroot or the guys who made the system goofed Mar 06 21:11:42 khem, please remind me that if we ever meet in person, that I owe you a beer :-) Mar 06 21:12:18 heh, thanks but I can only accept it on that mobile beer bar in NL Mar 06 21:12:32 * khem does not drink since 10 years Mar 06 21:13:01 well say 6 to be totally abstaining Mar 06 21:25:17 i think the mobile bar is forbidden; anyway a coke or so is also fine Mar 06 21:25:40 the mobile bar was against regulations, too wide or so (but actually probably caused too many problems) Mar 06 21:27:16 DAFUQ?! compilation of gettext fails with error "Cannot create temporary file in /tmp/: No space left on device" – are 4.2G not enough!? Mar 06 21:27:57 khem btw noticed that my kernel says upon booting: [ 0.000000] CPU: ARMv6-compatible processor [410fb024] revision 4 (ARMv7), cr=00c5387f Mar 06 21:27:57 why would it say v6 and v7 in one line ? Mar 06 21:28:52 eFfeM: it says v6-compatible (which v7 should be imho) Mar 06 21:29:12 this is supposed to be a v6 chip Mar 06 21:29:18 i thought Mar 06 21:29:31 i'd say your chip has a little identity crisis Mar 06 21:29:43 dFence: you can google for rm_work to delete intermediate build artifacts Mar 06 21:29:44 happens to the best of us Mar 06 21:30:38 hollisb: as INHERIT in local.conf? Mar 06 21:31:07 dFence: :-) Mar 06 21:31:54 I would suggest to buy a bigger HD instead Mar 06 21:32:17 eFfeM: what is on this board for a processor do you know Mar 06 21:32:56 cavium cns3420 Mar 06 21:33:34 the interesting thing is that the board is in CA, and I am in NL Mar 06 21:34:14 oh heh Mar 06 21:34:25 are you working for cavium btw. Mar 06 21:34:35 that board is armv6 AFAIK Mar 06 21:39:25 khem, I'm not working for cavium, I'm hacking a nas that has that chipset Mar 06 21:41:29 ok Mar 06 21:42:54 the chip is indeed listed as armv6; cavium does not really have much chip info for public download Mar 06 21:46:50 somehow status bits are confusing kernel them Mar 06 21:58:07 probably Mar 06 21:58:30 anyway, bedtime here, bye all Mar 06 23:58:33 which svn version is currently used by OE for opkg-utils? (seems the releases have changed since they moved to google code) **** ENDING LOGGING AT Wed Mar 07 02:59:58 2012