**** BEGIN LOGGING AT Mon Jul 25 02:59:57 2011 Jul 25 06:16:55 Hi all Jul 25 06:43:01 dorileo: try to find missing sources on http://sources.openembedded.org/ Jul 25 07:44:17 morning Jul 25 07:58:30 morning all Jul 25 08:09:39 How do I find the same point within the git repository that narcissus.angstrom-distribution.org is using to build am image for me? My /etc/angstrom-version is branch: 2011.03-maintenance, rev 09bd785 but If I check out 2011.03-maintenance it is using different versions of gcc., clib, stdc++lib and qt Jul 25 08:33:42 gm Jul 25 08:55:28 hi all Jul 25 08:55:51 I'm trying to build a kernel image for a machine (colibri toradex tegra2), using overlays (oe-classic) Jul 25 08:56:01 interesting config files are http://pastebin.com/fVPL0yaV Jul 25 08:56:18 the problem is that bitbake virtual/kernel try to do linux-nios Jul 25 08:56:28 instead of linux-colibri ... :( Jul 25 08:56:31 what's wrong ? Jul 25 08:58:20 brb Jul 25 10:36:39 How can I find a point in the git repository that narcissus is using to build an image for me? Jul 25 10:43:21 guyvdb_: you get the revision from the /etc/angstrom-version file Jul 25 10:57:34 hi guys Jul 25 10:57:53 I'm creating a custom git recipe..but I don't know what to set on SRCREV Jul 25 10:58:04 which value must I use there? Jul 25 10:58:37 latest revision known to work and tested by you Jul 25 10:59:42 this recipe is for bluez git repository Jul 25 10:59:52 I want to have always the latest revision Jul 25 10:59:58 is this possible? Jul 25 11:01:04 yes, put SRCREV_pn-bluez = "${AUTOREV}" to your local.conf Jul 25 11:01:23 but for others put some "sane" revision in recipe Jul 25 11:11:20 ok thank you! Jul 25 11:28:08 HeinervdmOff: I have looked at /etc/angstrom-version and it contains branch: 2011.03-maintenance, rev 09bd78 but if I checkout 2011.03-maintenance it has different versions of clib, stdc++lib, gcc and qt Jul 25 11:29:02 JaMa, I queried in the internet using its md5 and found it, sources neither mirrors.oe.org have it Jul 25 11:30:10 guyvdb_: did you checkout right revision or just 2011.03-maintenance HEAD? Jul 25 11:30:32 I am at head of 2011.03-maintenance Jul 25 11:30:48 then checkout rev 09bd78 Jul 25 11:31:49 hmm how do i do that? I did a "git checkout -b local-branch-name origin/2011.03-maintenance" originally Jul 25 11:34:47 09bd78 is giving me an unknown path or revision Jul 25 11:39:19 JaMa any idea how I get to 09bd78 instead of head? I am getting invalid revision so I must be doing something wrong Jul 25 11:42:23 git checkout 09bd78 Jul 25 11:43:18 JaMa it looks like i do not have the commit I am trying a git remote update Jul 25 11:44:09 hmm still do not get it... " error: pathspec '09bd78' did not match any file(s) known to git." Jul 25 11:46:00 pb_: ping Jul 25 11:46:35 ant_work: hi Jul 25 11:46:51 hello, not yet on holiday? Jul 25 11:47:13 heh, no Jul 25 11:47:18 no holidays here for the foreseeable future Jul 25 11:47:47 taking advantage of the fresh air we have here on south, I was busy enough with my pets Jul 25 11:48:18 now, I have the n- question for you: where is defined 'syscall' ? in unistd.h? Jul 25 11:48:37 either there or sys/syscall.h, I don't recall exactly Jul 25 11:48:49 I'm trying to remove 'magic' patches from klibc/kexec-tools-klibc Jul 25 11:49:01 if I dothat I lack definition of syscall Jul 25 11:49:06 yeah,it's in Jul 25 11:50:08 hm... as always I do it the hard way: klibc-static, --nostdinc, Jul 25 11:50:17 of course, klibc might do something different, or possibly not have syscall() at all Jul 25 11:50:44 ther is a pactch for klibc which I don't understand, well I appreciate the effects :) Jul 25 11:51:39 but this patches are years old and I tried wit a vanilla klibc..seems still lacking smthg Jul 25 11:52:10 what I dislike is patching klibc for the sle purpose of compiling kexec-tools against it Jul 25 11:52:13 *sole Jul 25 11:55:02 pb_: similar issue was solved for uclibc lik ethat: Jul 25 11:55:04 http://lists.busybox.net/pipermail/uclibc/2001-June/022475.html Jul 25 12:02:33 is the revision in /etc/angstrom-version just a partial git commit id? Jul 25 12:03:27 and would that commit id be in the branch also listed in /etc/angstrom-version (i assume so) Jul 25 12:11:08 Ok I am a little stumpped. Could my build from narcissus have incomplete information in /etc/angstrom-version ?? The revision hash seems really short and I cannot find a commit with such a hash. What am I doing wrong? Jul 25 12:12:00 JaMa : About the bluez4 git recipe I'm doing...I'm getting this error : AttributeError: 'FetchData' object has no attribute 'branches' Jul 25 12:12:44 what do you think? Jul 25 12:13:41 guyvdb_: with narcissus you're probably best off emailing the angstrom-devel mailing list Jul 25 12:14:00 thx bluelightnig Jul 25 12:14:47 FWIW I would really like it if Narcissus would clearly report the revision it used also Jul 25 12:16:29 yea.. i am struggling with this for a long time now... but i will post to the angstrom list Jul 25 12:20:57 mrAlmond: I guess incompatible bitbake Jul 25 12:21:24 mrAlmond: I had this in few revisions for bitbake master.. so first try to upgrade to bitbake/master/HEAD Jul 25 12:21:32 I'm using yocto 1.0 Jul 25 12:22:17 From my understanding this problem seems related to the AUTOREV option Jul 25 12:23:09 mrAlmond: and was fixed later in bitbake Jul 25 12:23:25 mrAlmond: if you are using 1.0 I would suggest upgrading to 1.0.1 Jul 25 12:23:51 thank you for the suggestion but unfortunately now I can't do an upgrade Jul 25 12:24:10 how can I obtain the latest SRCREV version from a git repository? Jul 25 12:24:17 just to do a test Jul 25 12:24:38 I mean the latest SRCREV string Jul 25 12:24:48 mrAlmond: http://git.openembedded.org/cgit.cgi/bitbake/commit/?id=eb799b44adab4701825f3dda5d982256781d7f78 Jul 25 12:25:30 mrAlmond: just checkout that repository or use gitweb to see latest hash and use that Jul 25 12:26:34 tnx! Jul 25 12:27:43 mrAlmond: btw there is bluez4_4.95.bb in meta-openembedded.. Jul 25 12:28:13 JaMa : yes I know but the git has some fixes that are not available in 4.95 Jul 25 12:28:24 this is what the bluez guys tell me Jul 25 12:28:37 my problem is with a crash of bluetoothd Jul 25 12:28:43 JaMa: I don't know if there's a good reason not to upgrade the version in oe-core... probably worth asking sgw Jul 25 12:29:47 mrAlmond: ok.. I was just surprised that you want bleeding edge bluez_git in "old" yocto 1.0 Jul 25 12:31:16 JaMa : at the moment I'm using yocto only to cross compile single packages as I'm not ablt to generate a full rootfs for a cramfs read-only fs Jul 25 12:32:29 I think that at the moment the lack of read-only filesystems support is something that should be improved as embedded devices most of the times have ro filesystems Jul 25 12:37:16 bluelightning: agreed, first I've added it to meta-oe to get more testing.. but I guess it's there long enough already :) Jul 25 12:43:39 As I'm a git noob...how can I find the hash of the latest version using gitweb...I'm not able to find it : http://git.kernel.org/?p=bluetooth/bluez.git;a=summary Jul 25 12:44:28 mrAlmond: click on last patch and copy&paste 53e6e763d252882921c0b7f17c56baa132f3b311 Jul 25 12:45:02 JaMa : Ok so it's the "commit" value :-) Jul 25 12:45:24 yes Jul 25 13:25:11 ant_work: that uclibc link you sent is specific to sysinfo(), not syscall() generically Jul 25 13:26:09 pb_: yes, I dug and found more bad news...http://www.zytor.com/pipermail/klibc/2003-January/000021.html Jul 25 13:27:04 seems having been replaced Jul 25 13:27:50 I'm now in touch with the klibc maintainer wrt kexec syscall Jul 25 13:29:08 are there known issues with iconv and external codesourcery toolchain? Jul 25 13:29:11 this one is listed in asm/unistd.h but not in klibc SYSCALLS.def Jul 25 13:30:16 ah Jul 25 13:30:43 http://git.kernel.org/?p=utils/kernel/kexec/kexec-tools.git;a=blob;f=kexec/kexec-syscall.h Jul 25 13:31:21 is failing on those syscall( Jul 25 13:31:34 yeah, seems like klibc suckage Jul 25 13:31:53 but the patch we have does dirty things I see,,, Jul 25 13:31:54 if that klibc list message is to believed, you could use _syscall4() rather than syscall() there. Jul 25 13:32:26 I wonder they ever tried to compile kexec-tools Jul 25 13:32:28 inside some suitable #if KRAZY_KLIBC kind of guard of course Jul 25 13:32:51 ok, I see Jul 25 13:32:53 it does seem a little bit outlandish that the klibc h4x0rs would not have tried to build kexec, but still Jul 25 13:33:05 that's subject of my mail :) Jul 25 13:34:11 or you could patch klibc to implement syscall(), it isn't all that hard. Jul 25 13:34:22 you could borrow the implementation from eglibc I guess Jul 25 13:34:27 the patch we have now just adds kexec_load Jul 25 13:34:34 as direct call Jul 25 13:34:54 yeah, or that, but then you'd still end up with kexec-tools having to be different for klibc vs other libcs Jul 25 13:34:56 but then in kexec code ugly things happen.... Jul 25 13:35:02 since (afaik) no other libc implements kexec_load() Jul 25 13:35:08 hm.. Jul 25 13:35:15 ah Jul 25 13:35:34 how is it it builds fine with uclibc/eglibc? Jul 25 13:35:54 they do have syscall() Jul 25 13:35:59 iirc you even removed dependency on kernel Jul 25 13:36:15 which is the patch that you're talking about? Jul 25 13:36:21 for kexec-tools Jul 25 13:36:22 let me have a look at what it actually does Jul 25 13:36:25 right Jul 25 13:36:30 http://cgit.openembedded.org/cgit.cgi/openembedded/tree/recipes/kexec-tools/kexec-tools-klibc-2.0.2/syscall.patch?h=org.openembedded.dev Jul 25 13:37:17 as far as I understand __NR_kexec_load is defined in asm/unistd.h Jul 25 13:37:30 different for each ARCH Jul 25 13:38:37 right Jul 25 13:38:41 wow, that patch looks like madness Jul 25 13:38:59 it 'happens' to work on armv5te Jul 25 13:40:00 I don't see how that can possibly work for either eglibc or uclibc, it seems to just remove kexec_load() altogether Jul 25 13:41:04 happily allthat stuff is klibc-specific Jul 25 13:41:10 :) Jul 25 13:41:30 ah, heh Jul 25 13:41:32 I guess that when thesing did commit that klibc was lacking some Jul 25 13:41:38 defs Jul 25 13:41:53 or maybe kexec was not supported by all archs Jul 25 13:42:55 return __reboot( was Jul 25 13:44:03 pb_: please shed some light Jul 25 13:44:06 :] Jul 25 13:45:00 actually all other binary are compiling just fine against klibc, is just kexec-tools needing special care Jul 25 14:16:15 ant_work: well, as far as I can tell, the problem is just that klibc lacks syscall(). Judging from that patch I am guessing that it does implement kexec_load() itself although I haven't checked. Jul 25 14:16:25 (and __reboot()) Jul 25 14:21:55 morning kergoth Jul 25 14:25:56 pb_: yes, the patch adds kexec_load() prototype Jul 25 14:26:20 er, does it? Jul 25 14:26:26 the copy I am looking at doesn't seem to Jul 25 14:26:38 this is the one you posted the url to earlier, right? kexec-tools-klibc-2.0.2/syscall.patch Jul 25 14:26:52 +long kexec_load( Jul 25 14:27:29 ah, the klibc side I mean Jul 25 14:27:42 klibc_kexecsyscall.patch Jul 25 14:29:29 pb_: I put all there http://paste.debian.net/124075/ Jul 25 14:34:35 ~seen khem Jul 25 14:34:37 khem is currently on #oe (6d 20h 8m 24s) #uclibc (6d 20h 8m 24s). Has said a total of 193 messages. Is idling for 2d 7h 59m 46s, last said: 'I dont think this info is there '. Jul 25 14:34:45 I'm pretty confident once we get syscall() defined we could use syscall(__NR_kexec_load,... ) and discard the kexec-tools-klibc/syscall patch Jul 25 14:36:12 ant_work: oh right, yeah Jul 25 14:36:19 yes, indeed Jul 25 14:36:32 or, as I mentioned earlier, I think you could probably use _syscall4(__NR_kexec_load, ...) now Jul 25 14:36:48 which I guess is probably what klibc_kexecsyscall.patch is doing internally Jul 25 14:37:03 I think those macro have been removed from klibc, I'll have to check Jul 25 14:37:31 oh right Jul 25 14:37:57 maybe copy the syscall() code from http://www.zytor.com/pipermail/klibc/2003-January/000021.html Jul 25 14:38:10 heh..10 yrs old code :/ Jul 25 14:38:20 almost.. Jul 25 14:40:31 pb_: you gave me some new ideas, thx Jul 25 15:14:46 Is there a variable which gives the current kernel version, I need to install a kernel module to ${D} Jul 25 15:14:48 ? Jul 25 15:15:51 KERNEL_VERSION? Jul 25 15:17:14 Do I need to inherit from something to pick that up? module, kernel class perhaps? Jul 25 15:17:40 module-base Jul 25 15:17:56 Ah ha, cool. Thanks pb_ Jul 25 15:22:55 Hmmm, shame. Something inherited from module-base is interfering with the build env of my package. Jul 25 15:23:14 Ah, suck. Well, you could just read kernel-abiversion directly I guess. Jul 25 15:23:38 I think I'll have to. Jul 25 15:24:06 I think it's the 'export OS' that's getting me. Jul 25 15:25:54 pb_: We haven't stopped working with OE, btw. Things have gone quiet while we try and justify to management the move to an OE based system. Now we wait for the machinery to turn and spit out a decision... Jul 25 15:28:03 heh, righto Jul 25 15:41:54 Is the use of ${STAGING_KERNEL_DIR}/kernel-abiversion deprecated? Jul 25 15:42:20 I think it used to be created in kernel.bbclass, but it isn't anymore. Jul 25 15:44:04 hm, really? Jul 25 15:44:10 it's still in my copy of kernel.bbclass Jul 25 15:44:31 (as of 4f03d922ce32f602a5a407cf422d82a0d036e4a0) Jul 25 15:47:49 I am potentially out-of-date WRT oe-core Jul 25 15:47:52 pb_: I'll do syscall home-works tonite. I hope to get you interested enough to build yourself a very small klibc image one day... Jul 25 15:47:56 bbl, bye Jul 25 15:49:12 Latest log I have for kernel.bbclass is 8c6bf9f6775fc5aaa8 Jul 25 16:16:55 celston: oh, right. it was deleted at one point and then reinstated in 6978774595f892e6a819ee64c932d475c67a6a0f. Jul 25 16:17:03 you must be in the "hole" where it didn't exist Jul 25 16:19:23 pb_: A re-pull is on the 'todo' list :) Jul 25 16:20:18 righto Jul 25 16:20:25 I guess you could just cherry pick that one commit if you wanted. Jul 25 16:41:17 hi all Jul 25 16:41:30 I'm trying to write a recipe Jul 25 16:41:57 for build on a new machine (toradex colibri tegra 2) Jul 25 16:42:01 I wrote this Jul 25 16:42:02 http://pastebin.com/fVPL0yaV Jul 25 16:42:14 but when bitbaking virtual/kernel Jul 25 16:42:21 it tries to compile linux-nios Jul 25 16:42:32 (and don't work) Jul 25 16:42:37 what's wrong ? Jul 25 17:09:48 * sgw sez: fyi www.yoctoproject.org is going readonly as we transition it to a new server Jul 26 01:24:48 hello, am a beginner on embedded system, so i dont know why many people didn't recommend me to use arduino ?? can tell me why Jul 26 01:48:39 arduino is only so usefull. but can be a intro to the subject Jul 26 01:58:43 so do you recommend me to use in the first case ??? Jul 26 01:59:21 am very good on linux and i know bases on C :) Jul 26 02:02:46 well its not a bad idea Jul 26 02:02:56 but ideally what to you want to learn? Jul 26 02:03:14 s/to/do Jul 26 02:08:38 i wanna to start courses of embedded system (especially Linux embedded), so i wanna to learn more about it. i know that i must know very well C and ASM. Jul 26 02:09:12 pick a beagleboard Jul 26 02:09:18 and start Jul 26 02:10:00 ok, but have you any good docs Jul 26 02:10:02 ? Jul 26 02:13:01 ichigo90: http://beagleboard.org/ Jul 26 02:13:03 start there Jul 26 02:13:43 ok thanks Jul 26 02:14:25 last question please, what language must i use to program on it??? Jul 26 02:16:48 should i have courses on electronic ??? because i did never learn it :s am a special technician on system and computers communication :D Jul 26 02:36:40 How about pandaboard???? is it better?? **** ENDING LOGGING AT Tue Jul 26 02:59:56 2011