**** BEGIN LOGGING AT Wed Feb 09 02:59:57 2011 Feb 09 07:19:21 hello all i build gdb for arm using oe .. Feb 09 07:19:49 but when i run the gdbserver binary it says as illegal instruction Feb 09 07:19:50 # /bin/gdbserver :1234 --attach 552 Feb 09 07:19:50 Illegal instruction Feb 09 07:20:02 any ideas .. where might be the things gone wrong Feb 09 07:36:28 whats the difference between /dev/mtd0 and /dev/mtdblock0 ? where would i use mtd_debug write against to ? Feb 09 07:42:04 good morning Feb 09 07:42:30 vlrk: add gdb-cross recipe and use it Feb 09 07:50:01 ok mckoan ; i will try that Feb 09 07:55:32 could someone look at this and find it good to apply -- very much needed for certain omap devices Feb 09 07:55:35 http://git.infradead.org/mtd-utils.git/commitdiff/772f5e08faa9d2b5677aa6e01fce06860186a49f?hp=a8214921bdc7d78b75d133a37ff9f8d4ec919650 Feb 09 08:44:56 mckoan:i did bitbake gdb-cross gdbclient Feb 09 08:51:54 even now i get same error any mis match with the processor differences Feb 09 08:52:58 i mean previously i used oe to get build for arm it was showing me armv5te Feb 09 08:53:09 it is now armv7a Feb 09 08:53:29 how can i go back to armv5te .. with oe .. Feb 09 08:54:48 hm.. our fbreader have no recipe for opie Feb 09 08:55:44 and we have 6 fbreader versions.. Feb 09 08:56:45 w/o fbreader.inc Feb 09 08:56:47 cool Feb 09 08:59:43 :) Feb 09 09:01:24 Jay7: hello Feb 09 09:01:30 ant_work: hi Feb 09 09:01:46 can you test with 6 bb threads? Feb 09 09:02:03 curl-native fails *always* with -j5 and 64 threads Feb 09 09:02:11 arr s/64/4/ Feb 09 09:02:12 ant_work: I've built opie-image this night w/o any problems Feb 09 09:02:47 on my buildhost (Gento) no issues with 2 threads and always breakage with 4 Feb 09 09:03:05 what distro/image? Feb 09 09:03:09 and machine :) Feb 09 09:03:14 minimal - console-image Feb 09 09:03:30 machine is c7x0 but iirc spitz had same issue Feb 09 09:04:02 ok, I'll run set of builds from 2 to 6 threads Feb 09 09:04:16 good morning Feb 09 09:05:16 I'll retry to build from within Ubuntu with the same metadata Feb 09 09:19:51 ant_work: builds started Feb 09 09:20:26 I'll go now, look to tinderbox for Jay7 builds (not Jay7-tb) Feb 09 09:20:42 ok, cu later Feb 09 09:20:47 thx Feb 09 09:21:35 should take about 2-2.5 hrs per build, 5 builds.. about 10 hrs :) Feb 09 09:22:56 he, the curl-native filure happens after 20-30 mins, around task 1490 on my Gentoo's buildhost with 4 threads Feb 09 09:24:03 seems collecting BB_THREADS_NUMBER with oestats is good idea as well Feb 09 09:24:50 and btw BB_SCHEDULER may affect it Feb 09 09:24:59 because of different task ordering Feb 09 09:25:06 I have completion now Feb 09 09:25:17 if it's only Gentoo, perhaps is my self-compiled kernel (scheduler, etc) Feb 09 09:26:30 i.e. some thread 'stalling', but that would be strange.... Feb 09 09:35:15 03Stefan Schmidt  07master * re88ef734d9 10openembedded.git/recipes/iw/iw_0.9.20.bb: iw: Its compatible with libnl2 as well use it. Feb 09 09:35:25 03Stefan Schmidt  07master * r495be66248 10openembedded.git/recipes/images/bug-x11-image.bb: bug-x11-image: Add X11 buglabs image Feb 09 09:35:27 03Stefan Schmidt  07master * ra7e3f3276a 10openembedded.git/recipes/wpa-supplicant/ (3 files in 2 dirs): Feb 09 09:35:27 wpa-supplicant-0.7: Enable the nl80211 driver Feb 09 09:35:27 Also enable support for libnl2 and depend on it. Feb 09 09:35:29 03Stefan Schmidt  07master * rb7e1882fe3 10openembedded.git/classes/jni-library.bbclass: Feb 09 09:35:29 jni-library.bbclass: Override JAVAH for our jni recipes. Feb 09 09:35:29 We need the javah from icedtea6 here. Feb 09 09:35:30 03Stefan Schmidt  07master * r343b243cf2 10openembedded.git/recipes/buglabs/dauber-feed-configs.bb: buglabs/dauber-feeds: Feed config for internal repos Feb 09 09:35:31 03Stefan Schmidt  07master * r0e9b38a840 10openembedded.git/recipes/buglabs/repo-feed-configs.bb: buglabs/repo-feeds: Feed configs for external use Feb 09 09:43:07 03Koen Kooi  07org.openembedded.dev * r82b466cec9 10openembedded.git/recipes/iw/iw_0.9.20.bb: Feb 09 09:43:08 iw: bump PR for nl2 change Feb 09 09:43:08 Signed-off-by: Koen Kooi Feb 09 09:44:11 hi all...what is quilt in recipes... Feb 09 09:45:59 10 hrs was very optimistic Feb 09 09:47:05 arg, forget the PR bump Feb 09 09:47:25 vicky_: patchstack management tool Feb 09 09:47:42 vicky_: takes care of adding our patches to the source Feb 09 09:48:07 Jay7: I only see one build running, started 2011-02-09 09:20:11. Is it with 2 threads? Feb 09 09:48:15 yes Feb 09 09:48:24 788 task of 6484 Feb 09 09:48:34 yes, it's slower Feb 09 09:48:40 much slower :) Feb 09 09:48:59 next time start with 6 and go down to 2 :) Feb 09 09:49:02 -j6 is helping sometimes, but overall progress is slow Feb 09 09:49:24 good idea :) Feb 09 09:50:01 I'll try with 6 threads too, even if on quad core -j5 and 4 bb threads seems ok Feb 09 09:50:48 (apart those races...) Feb 09 09:53:51 when i am using "bitbake nano" ... it taking long time... Feb 09 09:54:58 hang up at "package quilt-native-0.48-r7.1: task do_fetch: Started " Feb 09 10:06:51 vicky_: no erros around this? Feb 09 10:07:03 vicky_: are you able to write to the set DL_DIR? Feb 09 10:07:27 vicky_: maybe it fails to save the fetched file due to permission problems Feb 09 10:12:58 or it is fetching slowly Feb 09 10:16:49 or that Feb 09 10:30:21 hmm, I have a weird problem with a library package... I have a library, when compiled it's called libname.so.1.0.0, the ipk that gets created out of it is called libname-1.0.0_${PV}, now the library got updated to a new interface and is now called libname.so.1.1.0, OE makes a libname-1.1.0_${PV} ipk and opkg treats it like two absolutely different packages Feb 09 10:30:40 so updating to a newer version does not work anymore Feb 09 10:30:51 is this something that is expected or am I doing something wrong? Feb 09 11:25:10 03Koen Kooi  07org.openembedded.dev * rdc77eb8254 10openembedded.git/recipes/networkmanager/networkmanager_git.bb: Feb 09 11:25:10 networkmanager git: update to 0.8.3 Feb 09 11:25:10 Signed-off-by: Koen Kooi Feb 09 12:46:04 jo Feb 09 12:46:09 hi woglinde Feb 09 12:46:18 git-server ha some hickups? Feb 09 12:46:33 ah Feb 09 13:20:36 03Stefan Schmidt  07master * r235a67532c 10openembedded.git/recipes/buglabs/net.contentobjects.jnotify_0.91.bb: buglabs/jnotify: Add recipe for java notify usage Feb 09 13:20:48 03Stefan Schmidt  07master * ra25c44e70f 10openembedded.git/recipes/buglabs/ (bug2v4l2_0.01.bb bug2v4l2_git.bb): Feb 09 13:20:48 bug2v4l2: Fix recipe to atcually work. Feb 09 13:20:48 The first commit of this one was not intended. Sorry for that. The recipe is now Feb 09 13:20:48 cleaned up and working. Feb 09 13:40:25 jo mickeyl Feb 09 13:40:59 hi wogt Feb 09 13:41:03 woglinde, that is Feb 09 13:46:11 hi woglinde Feb 09 13:46:14 hi mickeyl Feb 09 13:46:20 morning steq Feb 09 13:46:25 err, what's up with my typing today Feb 09 13:46:29 heh Feb 09 14:06:01 * kergoth yawns Feb 09 14:06:11 gm kergoth Feb 09 14:10:12 Have question regarding reading u-boot environment from Linux Feb 09 14:10:24 As I understand u-boot env has partition Feb 09 14:10:39 Can it be access from Linux (should be possible) and read the data from it? Feb 09 14:10:52 What kind of structure it would have? Feb 09 14:11:43 ? Feb 09 14:11:54 u-boot runs before linux Feb 09 14:12:11 and by the way its like a linux kernel but stripped of most stuff Feb 09 14:16:07 davidlt: http://git.denx.de/?p=u-boot.git;a=tree;f=tools/env;h=756f60c3de12e895859fcf382ef1a5716db78e80;hb=HEAD Feb 09 14:19:11 ericben: thanks Feb 09 14:27:45 Crofton|work: get out of bed :-D Feb 09 14:29:13 I am Feb 09 14:29:19 gotta run some errands Feb 09 14:29:30 Need to find an SSD for my "Logic Analyzer" Feb 09 14:30:24 crofton what? Feb 09 14:30:52 hi xora btw. Feb 09 14:35:43 hey woglinde Feb 09 14:42:01 Is there any way of generating a detailed list of the packages installed into a image. preferrably with version number. Feb 09 14:43:08 INHERIT += testlab and inspect the testlab output would be a start Feb 09 14:43:13 hm rootfs log dont has it? Feb 09 14:45:42 I'll look into testlab. Feb 09 14:46:37 Isnt this a usual question? "Which packages, version numbers, licenses used do I have in my newly built image?" Feb 09 14:50:56 SH_ -> http://narcissus.angstrom-distribution.org/ Feb 09 14:54:35 Narcissus looks nice, but I want to generate a list from my OE builds Feb 09 14:55:05 its build from oe Feb 09 14:55:07 *g* Feb 09 14:55:21 but that was for your question with the licenses Feb 09 15:07:00 SH_: testlab Feb 09 15:07:40 woglinde: hey, about licenses... Feb 09 15:07:50 I thought once to commit beatrix http://people.dsv.su.se/~fk/beatrix_download.html Feb 09 15:07:58 the license is scary... Feb 09 15:08:23 well, it seems the author recently sold the software to a company Feb 09 15:09:01 question is: is one allowed to push a patch in OE metadata against the original sources? Feb 09 15:12:19 I am not a lawyer Feb 09 15:17:03 03Michael 'Mickey' Lauer  07org.openembedded.dev * r64dc157ca8 10openembedded.git/ (5 files in 2 dirs): fsogsmd: bump; we rely on libgisi / gisicomm now instead of libisi Feb 09 15:18:54 ant_work: the license used when the author released the code can chenge but is not retroactive Feb 09 15:19:24 03Marco Cavallini  07master * rdd15beecb5 10openembedded.git/recipes/fbv/fbv-1.0b/fbv-1.0b-BGR666.patch: fbv-1.0b-BGR666.patch: patch for support BGR666 @ 24bpp Feb 09 15:19:34 03Marco Cavallini  07master * rba8d1f493d 10openembedded.git/recipes/fbv/fbv_1.0b.bb: Feb 09 15:19:34 fbv_1.0b.bb: added support for new fbv-1.0b-BGR666.patch Feb 09 15:19:34 * bump PR Feb 09 15:19:34 * support for BGR666 @ 24bpp Feb 09 15:19:39 OU MAY NOT BUNDLE, RENAME, SELL, REDISTRIBUTE, REPACKAGE OR IN ANY OTHER WAY ALTER WHOLE OR PART OF THE SOFTWARE AND ITS DISTRIBUTION WITHOUT PRIOR, EXPRESS AND WRITTEN PERMISSION FROM THE AUTHOR. Feb 09 15:19:54 its pretty clear Feb 09 15:26:15 mckoan: so, if he sold a 'new' version to keyb, th eold one stay unchanged? Feb 09 15:27:44 see the argumentations (for distributiong a patch) here http://rg42.org/wiki/beatrix Feb 09 15:28:11 anyway, here in OE is not allowed I'd think Feb 09 15:30:08 ant_work: I'd avoid its inclusion in OE, stuff with unclear license are crap Feb 09 15:30:26 ant_work: unclear = not free :-D Feb 09 15:30:42 sure :p Feb 09 15:38:21 does anyone here use u-boot to read the kernel directly from ubifs volumes already ? Feb 09 15:51:59 Can someone take a look in a pull-request done in OE ml? Feb 09 15:52:08 * otavio his pull request Feb 09 15:52:50 * otavio is unable to push those himself for now and wants to bring those into upstream to avoid people having duplicated work Feb 09 15:55:56 otavio, in patchwork yes? Feb 09 15:56:05 Oh, er, pull request, heh, maybe Feb 09 15:56:09 link? Feb 09 15:56:15 Tartarus: ml Feb 09 15:56:22 Tartarus: I can past it here if you prefer Feb 09 15:56:57 Can you post all of the patches for review? Feb 09 16:14:13 Tartarus: I can but dunny if it is required. Feb 09 16:14:21 Tartarus: most of them are trivial Feb 09 16:16:18 Please help. Do not compile curl_7.21.2 distro = "angstrom-2008.1" mashine = "at91sam9263ek" error = "configure: error: one or more libs available at link-time are not available run-time. Libs used at link-time: -lldap -lrt -lz" how to fix it? Feb 09 16:17:27 03Koen Kooi  07org.openembedded.dev * r7b9e2fed3d 10openembedded.git/recipes/networkmanager/networkmanager_git.bb: Feb 09 16:17:27 networkmanager git: up DP for angstrom Feb 09 16:17:27 Signed-off-by: Koen Kooi Feb 09 16:19:35 otavio, can you rebase on current master? Feb 09 16:20:49 Tartarus: yes .. or if you wish you can do that. I did it 5min before posting the p-r. Feb 09 16:20:55 Tartarus: so it won't conflict Feb 09 16:21:24 So you're OK w/ me rebase -i'ing it? Feb 09 16:21:26 Then pushing Feb 09 16:21:44 Tartarus: sure. Just: git rebase origin/master Feb 09 16:21:49 Tartarus: it should be fine Feb 09 16:22:07 k Feb 09 16:22:47 CIA-2: what is DP? Feb 09 16:23:02 spirin: D_P? default-preferrence Feb 09 16:23:47 03Otavio Salvador  07org.openembedded.dev * r92c386850d 10openembedded.git/recipes/slang/slang_2.2.2.bb: Feb 09 16:23:47 slang: fix 2.2.2 URL Feb 09 16:23:47 Signed-off-by: Otavio Salvador Feb 09 16:23:51 03Otavio Salvador  07org.openembedded.dev * r0978356c5b 10openembedded.git/classes/gitpkgv.bbclass: Feb 09 16:23:51 gitpkgv.bbclass: return None if repository does not exist Feb 09 16:23:51 Signed-off-by: Otavio Salvador Feb 09 16:23:52 03Otavio Salvador  07org.openembedded.dev * r71a1e1a6e6 10openembedded.git/recipes/newt/ (3 files in 2 dirs): Feb 09 16:23:53 libnewt: include latest fixes and allow python removal Feb 09 16:23:53 Signed-off-by: Otavio Salvador Feb 09 16:23:55 03Otavio Salvador  07org.openembedded.dev * rd0b8fbdbd2 10openembedded.git/recipes/cmake/cmake-sdk_2.8.2.bb: Feb 09 16:23:56 cmake-sdk: add 2.8.2 recipe Feb 09 16:23:56 Signed-off-by: Otavio Salvador Feb 09 16:24:06 03Otavio Salvador  07org.openembedded.dev * rb762c15327 10openembedded.git/recipes/tzcode/tzcode-native_2011a.bb: Feb 09 16:24:06 tzcode-native: update to 2011b Feb 09 16:24:06 Signed-off-by: Otavio Salvador Feb 09 16:31:30 Tartarus: about the damned race (?) with curl-native and bb_number_threads = 4 Feb 09 16:31:43 well, Jay7 is building with 2,3,4,5,6 threads Feb 09 16:32:10 hm never hit this racing? Feb 09 16:32:27 what is the bitbake target to get when starting from scratch? Feb 09 16:32:33 ant_work, since you can reliably hit it, make up a script for git bisect run Feb 09 16:33:21 sure, I'll do a desperate test building from within Ubuntu (OE is on a separate partition) Feb 09 16:33:37 I start to doubt about Gentoo's kernel / scheduling Feb 09 16:34:06 (I compiled it, my bad :P) Feb 09 16:35:50 CIA-2: in openembedded/recipes/networkmanager/networkmanager_git.bb DEFAULT_PREFERENCE = "-1" change to DEFAULT_PREFERENCE = "1"? Feb 09 16:44:44 CIA-2: in openembedded/recipes/networkmanager/networkmanaget_git.bb DEFAULT_PREFERENCE = "-1" change to DEFAULT_PREFERENCE = "1" ? Feb 09 16:45:23 woglinde: the issue is do_configure http://www.scribd.com/doc/48366410/aa or (alternatively) do_compile http://www.scribd.com/doc/48372004/bb Feb 09 16:45:34 as you prefer :) Feb 09 16:45:46 strange, it ìs related to sockets Feb 09 16:52:58 Tartarus: thanks :-) Feb 09 16:53:25 Added --enable-shared to ruby recipe and got this after bitbake ruby : ERROR: QA Issue with ruby: No GNU_HASH in the elf binary Feb 09 16:53:51 is this still the fix for that? TARGET_CC_ARCH += "${LDFLAGS}" Feb 09 16:54:15 atl19711 yes Feb 09 16:54:23 or fix the makefile Feb 09 16:54:32 to accept LDFLAGS from enviroment Feb 09 16:55:24 woglinde: thanks, good enough, hard to trust postings that are 2 years old Feb 09 17:09:22 CIA-2: in openembedded/recipes/networkmanager/networkmanaget_git.bb DEFAULT_PREFERENCE = "-1" change to DEFAULT_PREFERENCE = "1" ? Feb 09 17:10:48 does anyone have any ideas for a better name for oe.types.value()? Feb 09 17:10:54 it sucks, i hate it Feb 09 17:10:56 (name-wise) Feb 09 17:14:23 hmm, maybe it should just be elsewhere. its part of variable typing, yes, but conceptually it's part of the metadata. oe.value() or oe.data.value() might be better Feb 09 17:19:12 g'day kergoth Feb 09 17:19:38 hey pb_ Feb 09 17:19:42 hi pb Feb 09 17:20:25 hi woglinde Feb 09 17:21:13 anybody had success targeting mingw with oe? Feb 09 17:21:25 I was wondering whether it would help with a windows build problem that I have. Feb 09 17:21:38 i tried, it's definitely more work than cygwin is Feb 09 17:21:58 ah. how far did you get? Feb 09 17:22:20 not very :) Feb 09 17:22:23 opkg-native dies Feb 09 17:22:46 oh, funky. I wonder why opkg-native cares. Feb 09 17:23:12 also, pretty sure bitbake doesn't work at all Feb 09 17:23:51 there's multiple issues there, the lack of windows fcntl for file locking, you can replace that with a cross-platform file lock module Feb 09 17:24:08 ah, I wonder if we are talking about the same thing Feb 09 17:24:09 then there were, iirc, issues with task execution, due to the forks Feb 09 17:24:18 oh, right Feb 09 17:24:19 what I want to do is build for a mingw target, but on a linux host Feb 09 17:24:21 *targeting* Feb 09 17:24:23 gotcha Feb 09 17:24:27 okay, never tried that :) Feb 09 17:24:33 don't see any reason why it wouldn't work though Feb 09 17:24:41 opkg might still care, the target one Feb 09 17:24:53 yeah, luckily I don't need/want opkg on target :-} Feb 09 17:24:57 don't think it does anything particularly bad though, i asked on the list Feb 09 17:25:00 ah, better yet Feb 09 17:25:29 * kergoth ponders Feb 09 17:25:32 all I really want to do is build gstreamermm, but doing that by hand on native windows seems to be something of a mission. Feb 09 17:25:49 uhaa Feb 09 17:25:49 seems like that would be quite useful, the msys distro is pretty weak Feb 09 17:25:58 why you need gstreamer on windows? Feb 09 17:26:03 hell, targeting cygwin would be nice too, setup.exe can DIAF as far as i'm concerned Feb 09 17:26:12 there are helpful prebuilt windows binaries of gstreamer itself, but less helpfully they don't include any .pc files so no way to compile against them without manual hackery Feb 09 17:26:12 isn't vlc enough? Feb 09 17:26:21 hm Feb 09 17:26:21 ah Feb 09 17:26:36 woglinde: vlc doesn't help much with running my gstreamer-based applications Feb 09 17:26:59 is gstreamer on windows using all the windows video accel frameworks? Feb 09 17:27:00 unrelated: if anyone here hasn't discovered 'ack' as an alternative to grep, i strongly encourage checking it out, it's great (betterthangrep.com, iirc) Feb 09 17:27:22 woglinde: dunno, I only care about audio Feb 09 17:27:30 hm okay Feb 09 17:27:31 (mostly) Feb 09 17:27:44 kergoth: sounds hoopy, I'll take a look Feb 09 17:28:02 evening Feb 09 17:28:17 hi jay7 Feb 09 17:28:30 * woglinde needs to fix pulsecrap for uclibc again Feb 09 17:28:55 has some quite nice features. save a bit of typing. e.g. it defaults to recursive, defaults to searching the current directory, and only shows results in file types it knows about, ignoring scm dirs, etc Feb 09 17:29:01 great for finding things in source trees Feb 09 17:29:05 "Top 10 reasons to use ack instead of grep: (2) ack is pure Perl" Feb 09 17:29:14 * kergoth tries to ignore the fact that its perl Feb 09 17:29:19 not sure I am entirely convinced by that particular one. :-} Feb 09 17:29:21 ant_work: all builds was fine Feb 09 17:29:21 hehe Feb 09 17:29:29 6-threaded is in progress Feb 09 17:29:56 but yeah, if it can ignore scm dirs and editor backups then that alone is a win Feb 09 17:30:26 yeah Feb 09 17:31:02 "9. Color highlighting of search results. " Feb 09 17:31:05 you can also do export ACKRC=.ackrc to get it to use the one in the current dir in preference to ~/.ackrc, which means you can create one in your project and make it ignore, say tmp/ or build*/ dirs Feb 09 17:31:08 hm, grep seems to be doing that for me nowadays Feb 09 17:31:09 heh, grep does color, has for a while Feb 09 17:31:10 yeah Feb 09 17:31:34 it groups results though, which is nice. file path header, then the lines. if stdout isn't a tty, it outputs grep format, for scripting Feb 09 17:31:42 The requested URL /svn/tags/latest/ack was not found on this server. Feb 09 17:31:43 doh Feb 09 17:31:49 hah Feb 09 17:37:59 * pb_ finds a working download link Feb 09 17:45:00 where is Xorg pulling the info that it uses tslib.so for a touchscreen ? Feb 09 17:49:32 echo '--type-set=bitbake=.bb,.bbappend,.bbclass,.inc,.conf' >> ~/.ackrc is helpful Feb 09 17:49:36 hmmm Feb 09 17:49:50 * kergoth wonders whether to add a expected_type assertion to the typing support Feb 09 17:50:28 thinking itd be good for catching the situation where the metadata's type of a variable changes, but not all the classes were adjusted for it, so it'd immediately fail on that assert rather than waiting and failing in the actual usage of the type Feb 09 17:58:02 does https://github.com/kergoth/openembedded/commit/d269078 seem like an improvement to people? Feb 09 17:58:09 (not that anyone but me is likely using this stuff yet..) Feb 09 18:19:11 kergoth: oe.types -> oe.typing Feb 09 18:19:23 I would say make it oe.maketype Feb 09 18:19:26 or somesuch Feb 09 18:19:41 typing is continuous tense :) Feb 09 18:20:39 why is sysvinit-inittab separated out of sysvinit? Feb 09 18:20:48 for alternative inittabs? Feb 09 18:21:07 oe.maketype sounds good Feb 09 18:21:21 I want upstart anyway Feb 09 18:21:26 next 2 tasks Feb 09 18:21:31 actual upstart Feb 09 18:21:39 kergoth: oe.data.value Feb 09 18:21:44 and fixing pulse for uclibc Feb 09 18:21:55 may be since it returns properly types values it should show up in name Feb 09 18:21:56 hi khem Feb 09 18:22:08 woglinde: hey Feb 09 18:22:44 woglinde: how are things Feb 09 18:26:49 hi khem, ever tried qemuarm+uclibc+qt4? Feb 09 18:29:28 khem: oe.data.typed_value() or so? Feb 09 18:29:46 | /var/hudson/workspace/sb_openembedded_autobuilder/BBPKGS/qt4e-demo-image/DISTRO/minimal-uclibc/MACHINE/qemuarm/label/sb-rhel5/tmp/sysroots/i686-linux/usr/armv5te/lib/gcc/arm-oe-linux-uclibceabi/4.5.3/../../../../arm-oe-linux-uclibceabi/bin/ld: /var/hudson/workspace/sb_openembedded_autobuilder/BBPKGS/qt4e-demo-image/DISTRO/minimal-uclibc/MACHINE/qemuarm/label/sb-rhel5/tmp/sysroots/armv5te-oe-linux-uclibceabi/usr/lib//uclibc_nonshared.a(atexit.os)(.text+0x12): u Feb 09 18:29:59 i was thinking we could subclass DataSmart in oe.data eventually, possibly even automatically return typed values from __getitem__ Feb 09 18:31:03 oe.maketype.create() seems kind of redundant, but it probably doesn't matter too much, better to be explicit than implicit Feb 09 18:31:34 did anyone have an opinion on the type assertions thing? Feb 09 18:32:08 e.g. oe.data.typed_value('SOMEVAR', d, 'list') -- if the metadata changes the type of the variable and thsi code doesn't change to match, it'd immediate error out there Feb 09 18:32:17 not sure if that would be good or not Feb 09 18:33:13 Tartarus you have to use uclibc-git Feb 09 18:33:36 all other will not work Feb 09 18:33:46 woglinde, yes, this is what sane toolchain is pulling in Feb 09 18:33:47 because latest release has not log2 function Feb 09 18:33:54 hgm Feb 09 18:33:59 stupid angstroem Feb 09 18:34:08 minimal Feb 09 18:34:29 094d82d3b3e8f8c8460a802b7a9548f3ae4fda46 good enough? Feb 09 18:34:41 hmm Feb 09 18:39:12 kergoth: ok seems better Feb 09 18:39:24 what does? :) Feb 09 18:39:33 03Chris Larson  07master * rf41d2fdfa6 10openembedded.git/lib/oe/types.py: Feb 09 18:39:34 oe.types: for invalid type, list valid ones Feb 09 18:39:34 Signed-off-by: Chris Larson Feb 09 18:40:44 Tartarus: whats the error in the end Feb 09 18:40:55 Tartarus: I suppose it only happens on qemuarm Feb 09 18:41:04 | /var/hudson/workspace/sb_openembedded_autobuilder/BBPKGS/qt4e-demo-image/DISTRO/minimal-uclibc/MACHINE/qemuarm/label/sb-rhel5/tmp/sysroots/i686-linux/usr/armv5te/lib/gcc/arm-oe-linux-uclibceabi/4.5.3/../../../../arm-oe-linux-uclibceabi/bin/ld: final link failed: Nonrepresentable section on output Feb 09 18:41:15 and yes, only arm Feb 09 18:44:36 hm Feb 09 18:44:45 might be some armv6/armv7 foo Feb 09 18:44:49 in qt Feb 09 18:57:01 kergoth: oe.data.typed_value() seems better yes Feb 09 18:58:09 Tartarus: couple of things to try Feb 09 18:58:19 Tartarus: I know if you compiled all in ARM mode it will work Feb 09 18:58:26 so thats last resort Feb 09 18:58:40 problem is thumb/arm mixing Feb 09 18:59:00 Tartarus: one thing to try is downgrade to 2.20.1 binutils Feb 09 18:59:18 and rebuild this image this way we can rule out ld problem Feb 09 18:59:43 Tartarus: second thing to look at is how uclibc is compiled w.r.t. atexit.os Feb 09 19:00:05 Tartarus: I have a qemuarm build lying around I can do this part quickly Feb 09 19:03:54 at4.inc has ARM_MODE="arm" or whatever it is Feb 09 19:03:57 qt4.inc Feb 09 19:04:10 I'll try dropping down to 2.20.1 next Feb 09 19:04:31 (doing a qemumips64 build now, that's the other failure I see which is a postgresql linking issue I think I see how to fix..) Feb 09 19:12:23 03Chris Larson  07master * r2fbd56c0ef 10openembedded.git/ (8 files in 3 dirs): (log message trimmed) Feb 09 19:12:23 More sane naming for the variable typing code Feb 09 19:12:23 oe.types.value -> oe.data.typed_value Feb 09 19:12:23 This name has been bugging me. This function is the primary interface to Feb 09 19:12:23 the module for OE metadata, as it takes a variable name and datastore and Feb 09 19:12:23 returns an object of the correct type. While this function is part of the Feb 09 19:12:24 variable typing implementation, in reality it's more about giving you a Feb 09 19:12:28 03Chris Larson  07master * r8d22023a06 10openembedded.git/lib/oe/test_types.py: Feb 09 19:12:28 oe.test_types: improve naming/organization Feb 09 19:12:28 Signed-off-by: Chris Larson Feb 09 19:13:03 if anyone hasn't used this typing stuff and wants to see a real example: http://git.openembedded.org/cgit.cgi/openembedded/tree/recipes/editline/editline_20100424-3.0.bb#n18 Feb 09 19:26:35 Tartarus: with mips64 mostly it should be missing a autoconf cached var in site/ Feb 09 19:27:13 mostly, yes :) Feb 09 19:30:50 hm, I have a problem with dpkg packages Feb 09 19:30:51 tmp/work/armv4-oe-linux-gnueabi/dpkg-1.14.29-r1/image/usr/lib/perl/5.10.1/Dpkg.pm: Perl5 module source text Feb 09 19:31:04 this is not included in any (dpkg, dpkg-dev, dpkg-dbg) Feb 09 19:31:15 yet I can't find any reason in recipes/dpkg Feb 09 19:31:32 is this behaviour inherited? Feb 09 19:33:20 filip: so explain the problem Feb 09 19:36:37 khem: update-alternatives-dpkg needs Dpkg.pm which is in dpkg, but is not being installed Feb 09 19:36:55 khem: and I see no FILES_ lines in dpkg recipe, so I don't know why it gets discarded Feb 09 19:38:40 filip: if its needed then it should be added to one of the packages Feb 09 19:39:25 khem: yeah, but where are these defined? Feb 09 19:39:47 filip.zyzniewski@jlime:~/openembedded/recipes/dpkg$ grep -r dbg . Feb 09 19:39:47 filip.zyzniewski@jlime:~/openembedded/recipes/dpkg$ Feb 09 19:39:58 khem: this confuses me Feb 09 19:40:17 filip: recipes aren't the only source of metadata.. Feb 09 19:40:26 filip: you can put all /usr/share/perl5/Dpkg under a new package too Feb 09 19:40:34 filip: read bitbake.conf Feb 09 19:40:35 say dpkg-perl or something Feb 09 19:41:33 and yes if its installed but not packaged then you should get warnings from bitbake about it Feb 09 19:42:19 kergoth: aaah, this helped a lot, I see it now :) Feb 09 19:42:36 khem: it hasn't printed any Feb 09 19:43:17 hmm Feb 09 19:44:36 ok, I'll define FILES_${PN}-perl Feb 09 19:49:12 now this is a sick shebang: #!/home/builds/jornada7xx/tmp/sysroots/x86_64-linux/usr/bin/perl -- Feb 09 19:50:17 welcome to the evil of perl.. ;) I recommend python style #!/usr/bin/env perl :) Feb 09 19:50:42 can I add a directory recursively in FILES_.. ? Feb 09 19:51:24 you can either use globs or directories, yes Feb 09 19:51:25 just listing it is enough it seems... Feb 09 19:51:26 it'll traverse dirs Feb 09 19:51:28 filip: use wildcards but if you add dir it should add stuff below it too unless someone else included it Feb 09 20:26:02 Tartarus: I've added my future builds to wiki, check format please Feb 09 20:26:22 not sure that I should group all-in-one-row Feb 09 20:26:59 Jay7, works for me for now Feb 09 20:27:06 Have to see once we've got failures to deal with :) Feb 09 20:28:06 Tartarus: when first tests are planned? Feb 09 20:28:48 tomorrow instead of testing-next Feb 09 20:29:11 I think one line for al passes, so long as it's really right, is fine and lines for failures Feb 09 20:29:21 (right means all comos in the set work) Feb 09 20:29:37 *combos Feb 09 20:31:29 well.. Feb 09 20:31:49 I just scared about filling that table again :) Feb 09 20:32:24 heh Feb 09 20:35:59 I'll try to add proposed fields to oestats-client Feb 09 20:38:07 * kergoth fights with sqlite3 Feb 09 20:38:41 kergoth: still locking problems? Feb 09 20:39:44 just trying to verify that its fixed :) Feb 09 20:39:51 since it was random.. Feb 09 20:45:42 03Graeme Gregory  07org.openembedded.dev * r7794a0c735 10openembedded.git/recipes/linux/ (5 files in 3 dirs): Feb 09 20:45:43 linux-omap_2.6.38.bb: add 2.6.38-rc4 version Feb 09 20:45:43 I have only enabled this kernel for the omapzoom2/3 machines that I Feb 09 20:45:43 take care of and it has been boot tested on the zoom3. Feb 09 20:45:43 This recipe has PARALLEL_MAKE disabled as multiple jobs is showing Feb 09 20:45:43 a dependency issue in the kernel which I hit 100% of time in OE. Feb 09 20:45:44 Signed-off-by: Graeme Gregory Feb 09 20:45:46 03Graeme Gregory  07org.openembedded.dev * r8b2bf26a68 10openembedded.git/conf/machine/ (omapzoom2.conf omapzoom36x.conf): Feb 09 20:45:46 omapzoom2/3.conf : switch to linux-omap kernel Feb 09 20:45:46 Signed-off-by: Graeme Gregory Feb 09 20:55:19 kergoth: can I help you anyhow? Feb 09 20:56:07 I can run build with 12 threads ;) Feb 09 20:56:39 nah, i think i've got it, just a matter of doing -p N times with and without the change Feb 09 21:12:14 khem, do you know how to force 'ld' into elf64-tradbigmips for output when it wants to go elf64-ntradbigmips? Feb 09 21:13:33 hm, I have a problem with ttys Feb 09 21:13:35 stat64("/dev/tty4", 0xbebd6b88) = -1 ENOENT (No such file or directory) Feb 09 21:14:00 crw-rw-rw- 1 root root 4, 4 Feb 9 21:08 dev/tty4 Feb 09 21:14:30 this kind of low level inconsistency freaks me out Feb 09 21:18:47 I mean, what can sneak in between stat() and the filesystem? Feb 09 21:19:35 filip: could the leading / matter? Feb 09 21:20:45 hollisb: I don't think so, my $PWD is the root partition (mounted on a PC) Feb 09 21:21:36 are you using an initramfs (or similar) on the target? Feb 09 21:21:53 aaah, ramfs on /dev Feb 09 21:21:56 sorry for this Feb 09 21:21:57 or is something (udev?) mounting a new /dev on the target? Feb 09 21:21:58 yeah Feb 09 21:27:47 filip: check *device-table.txt Feb 09 21:27:51 somewhere in recipes :) Feb 09 21:28:20 ah, no Feb 09 21:28:33 files/*.txt Feb 09 21:28:40 in root OE dir Feb 09 21:29:13 Jay7: huh, there's hardly anything in device_table-jlime.txt Feb 09 21:29:35 Jay7: how is it deployed into the tmpfs? Feb 09 21:29:59 03Koen Kooi  07org.openembedded.dev * rd02fea18c2 10openembedded.git/recipes/linux/linux-omap_2.6.31.bb: Feb 09 21:29:59 linux-omap 2.6.31: remove DP for zoom2 Feb 09 21:29:59 Signed-off-by: Koen Kooi Feb 09 21:30:01 not sure, but somewhere at do_rootfs I think Feb 09 21:30:34 hm, S05devices should create tty* Feb 09 21:30:44 if no udev ;P Feb 09 21:31:11 filip: extract initramfs image with cpio and look there Feb 09 21:31:42 Jay7: no initramfs for j7xx :) Feb 09 21:31:43 anyway, according to device_table you should have tty0-9 Feb 09 21:31:57 ah.. Feb 09 21:32:17 then add ls /dev/* to rc :) Feb 09 21:32:26 I think I know what happened Feb 09 21:32:37 my device cache got spoiled on first unsuccessful boot Feb 09 21:32:41 somehow Feb 09 21:36:47 hm, makedevs displays usage on boot Feb 09 21:36:50 this can't be right :P Feb 09 21:39:30 ok, there is a bug. I have: Feb 09 21:39:32 filip.zyzniewski@jlime:~/openembedded$ grep makedevs recipes/initscripts/initscripts-1.0/devices /sbin/makedevs -r / -D /etc/device_table Feb 09 21:39:39 yet makedevs won't accept -r Feb 09 21:40:12 hrm Feb 09 21:40:22 my makedevs is symlinked to busybox Feb 09 21:40:46 kergoth: any suggestions maybe ;)? Feb 09 21:41:05 or maybe there is nothing to fix but I have built an image in a wrong way Feb 09 21:41:19 another alternative for makedevs or something Feb 09 21:43:51 would a patch removing '-r /' be a good idea? Feb 09 21:46:48 busybox'es makedevs should be called: makedevs -d /etc/device_table / Feb 09 22:10:05 ERROR: Logfile of failure stored in: /dev/null Feb 09 22:10:54 NOTE: package lzma-native-4.65-r5.1: task SRC_DISTRIBUTECOMMAND: Failed Feb 09 22:10:55 ERROR: Function 'SRC_DISTRIBUTECOMMAND' failed (see /dev/null for further information) Feb 09 22:11:02 build with 12 bb threads ;) Feb 09 22:17:23 can new people make wiki logins? Feb 09 22:17:28 or is that restricted? Feb 09 22:24:58 Tartarus: ask ka6sox or cryptk|offline Feb 09 22:25:08 or just try :) Feb 09 22:25:22 Well, I don't see the link to just try Feb 09 22:25:30 ah, hm.. Feb 09 22:25:35 (got an email from someoen that wants to help but can't see where to register for the wiki) Feb 09 22:25:46 btw, yes, there was only login link Feb 09 22:55:39 So, here's a thing Feb 09 22:55:50 for postgresql, it uses ld, raw, for linking at times Feb 09 22:56:00 And was doing on mips64, default to mips32 Feb 09 22:56:07 binutils config issue? Feb 09 22:56:18 Otherwise we need TARGET_LD_ARCH Feb 09 22:56:24 (which I've patched in and seems to be working) Feb 09 22:58:10 using ld for linking is wrong in most cases Feb 09 23:00:44 I can see if 8.4.x or so uses CCLD Feb 09 23:02:16 the only reason to use ld directly is if you need a non standard binary, are using a linker script, etc. Which should not be the case for a userspace application. Feb 09 23:10:04 So, yeah, wow Feb 09 23:10:11 postgresql 8.1 is EOL Feb 09 23:10:16 and 8.2 goes EOL this year Feb 09 23:10:25 8.4 is 2014 at least Feb 09 23:10:47 So maybe it's time to grab current 8.4.x (we have 8.4.4, 8.4.7 is latest) and drop the DP=-1 from it Feb 09 23:11:08 huh Feb 09 23:37:00 are there docs in-progress on using an external toolchain? the ones in git still have a TODO: on that section Feb 09 23:45:54 03Filip Zyzniewski  07org.openembedded.dev * r8a01a9750f 10openembedded.git/recipes/base-files/base-files/jornada7xx/fstab: Feb 09 23:45:54 base-files: update fstab for jornada7xx Feb 09 23:45:54 We use the scsi subsystem for PCMCIA ATA access, so device files are Feb 09 23:45:54 /dev/sda*, not /dev/hda*. Feb 09 23:45:54 See recipes/linux/linux-jlime-jornada7xx-2.6.34+git/defconf_jlime:702. Feb 09 23:45:54 Signed-off-by: Filip Zyzniewski Feb 09 23:45:55 Signed-off-by: Tom Rini Feb 09 23:47:20 03chase maupin  07org.openembedded.dev * r51ed875e94 10openembedded.git/recipes/cups/ (2 files in 2 dirs): (log message trimmed) Feb 09 23:47:21 cups: fix desktop file QA issues Feb 09 23:47:21 * This patch fixes the QA issues with version 1.2.12 of cups. Feb 09 23:47:21 * The patch being applied was taken from Feb 09 23:47:21 http://bugs.gentoo.org/show_bug.cgi?id=236706 Feb 09 23:47:21 * Bump the PR Feb 09 23:47:21 Signed-off-by: Chase Maupin **** ENDING LOGGING AT Thu Feb 10 02:59:57 2011