**** BEGIN LOGGING AT Tue Sep 14 02:59:57 2010 Sep 14 04:43:34 why do so many projects insist on shipping tarballs with 0 byte README files? Sep 14 04:52:35 grg, because autotools complains Sep 14 04:53:19 playya_, Mystery solved. Thanks. But that doesn't mean they can't put some content in it. Sep 14 04:53:51 coders want to code, not to write READMEs Sep 14 06:02:04 ka6sox4: finally obtained a ping from work: http://www.pastie.org/1157527 Sep 14 06:02:12 btw & gm everyone Sep 14 06:26:50 Hi all Sep 14 06:27:04 any llvm build expert over here Sep 14 06:46:42 03Brijesh Singh  07org.openembedded.dev * r5513bd1f9a 10openembedded.git/recipes/ti/ (ti-codec-engine.inc ti-codec-engine_2.26.00.08.bb): (log message trimmed) Sep 14 06:46:42 ti-codec-engine_2.26.00.08: add new codec engine version. Sep 14 06:46:42 ti-codec-engine_2.26.00.08: add new codec engine version. Sep 14 06:46:42 * Add CE 2.26 GA release version. Sep 14 06:46:42 * .make target is no longer supported on 2.26 hence commenting out from .inc. Sep 14 06:46:43 Signed-off-by: Brijesh Singh Sep 14 06:46:43 Signed-off-by: Denys Dmytriyenko Sep 14 06:47:34 03Roger Monk  07org.openembedded.dev * r0ee1e44969 10openembedded.git/recipes/ti/ (2 files in 2 dirs): Sep 14 06:47:34 ti-codec-engine: Add initial recipe for latest ce 2.26.00.08 Sep 14 06:47:34 Signed-off-by: Roger Monk Sep 14 06:47:34 Signed-off-by: Koen Kooi Sep 14 07:20:20 good morning Sep 14 07:23:26 morning Sep 14 08:08:52 does anyone know where RPATH variable is being set in LLVM packahe Sep 14 08:47:35 hello? Sep 14 09:46:50 aja, packi, ganz vergessen Sep 14 09:46:53 oops wrong window :) Sep 14 10:18:04 gm Sep 14 10:18:14 hi likewise Sep 14 10:18:34 hi Jin^eLD Sep 14 10:44:38 I am trying to add install -d $D${sysconfdir}/${PN} into do_install() function, but the recipe fails with Sep 14 10:44:40 + install -d /etc/mypackage Sep 14 10:44:47 install: cannot change permissions of `/etc/mypackage': Operation not permitted Sep 14 10:44:55 any idea why? Sep 14 10:45:44 turanmo: try ${D} maybe Sep 14 10:46:25 i.e. ${D}${sysconfdir}/${PN} Sep 14 10:46:27 oh how dumb, oh how dumb of me Sep 14 10:46:30 Hi again Sep 14 10:46:38 Jin^eLD: that's a million Sep 14 10:46:40 :) Sep 14 10:46:43 :) Sep 14 10:46:48 s/that's/thanks Sep 14 10:46:48 need help on cmake in llvm package Sep 14 10:46:51 well, just 4 pair of eyes ;> Sep 14 10:47:00 yeah Sep 14 10:47:51 in llvm package when do configure task is execute an rpath variable is set Sep 14 10:48:15 need to change that value but could not be able to figure it it out how to do that Sep 14 11:04:10 can anybody give me a hint Sep 14 11:14:52 noor, no idea on llvm, but there are a few options Sep 14 11:15:03 patch configure in a _prepend and change the path Sep 14 11:15:15 bye Sep 14 11:15:30 nm Sep 14 11:18:26 what is the tag i use to copy files from the "files" directory ? i think i could try ${FILE_DIRNAME}/files. Or is it that all contents of "files" dir, will be available in ${S} ? Sep 14 11:18:56 so i just use ${S} ? Sep 14 11:20:57 turanmo: stuff that you install additionally via adding your own file:// entries to SRC_URI land in ${WORKDIR}, all sources from the main package land in ${S} Sep 14 11:21:46 Jin^eLD: ah, excellent. so i need to use ${WORKDIR} Sep 14 11:21:59 yes, i will specify it via SRC_URI Sep 14 11:22:56 * turanmo glances page 55. oe manual Sep 14 11:56:49 hi, does anyone remember if SRC_URI = "git:///path/gitrepo/" is supposed to work or is it broken ? Sep 14 11:58:42 here it seems git clone the repo but instead of checking out master branch, it tries to checkout an unknow object "1" Sep 14 11:59:16 ericben: I think you need the SRCREV or something, let me look it up Sep 14 11:59:41 yes, otherwise it will check out rev 1 and that does not exist Sep 14 11:59:46 Jin^eLD: oh yes you are right Sep 14 11:59:53 thanks Sep 14 12:00:51 that explains why it was working on my previous recipe, I had : SRCREV = "${AUTOREV}" Sep 14 12:00:54 :) Sep 14 12:25:31 ericben: hmm, what did AUTOREV do? Always pick the latest? Sep 14 12:30:25 likewise: hey Sep 14 12:30:53 I was fooled by UTC... my build started at 23.08 UTC ... Sep 14 12:37:16 ant_work: and it failed? or worked? Sep 14 12:37:40 builds were fine, is logged in tinderbox Sep 14 12:38:08 but console-image was done before midnight..so angstrom-version Sep 14 12:38:29 iirc issue arises crossing midnight Sep 14 12:39:11 I remember I've seen huge time-shiftings...several hours Sep 14 12:39:20 in the oestats logs Sep 14 12:40:04 but was long ago... Sep 14 12:43:14 likewise: yes AUTOREV always picks that HEAD I think Sep 14 12:44:23 could someone tell me, how i may give instructions to the user after the installation in do_install() is complete? Sep 14 12:44:25 perhaps i must add something in pkg_postinst()? what function can i call? if i use "oenote" i get stuff like this on target bootup Sep 14 12:44:30 /usr/lib/opkg/info/mypackage.postinst: line 1: oenote: not found Sep 14 12:44:32 /usr/lib/opkg/info/mypackage.postinst: line 2: oenote: not found Sep 14 12:44:33 /usr/lib/opkg/info/mypackage.postinst: line 3: oenote: not found Sep 14 12:44:55 turanmo: postinst is shell Sep 14 12:45:02 turanmo: I think oenote is only for usage inside OE but not on the target Sep 14 12:45:16 right Sep 14 12:45:25 anything else i can use? Sep 14 12:45:27 you only have basic sh stuff there and no "special" oe things Sep 14 12:45:28 echo ? Sep 14 12:45:44 ericben: :) Sep 14 12:45:55 cat file-with-instructions.txt Sep 14 12:46:36 * turanmo was thinking something sophisticated. for example gentoo has "elog" Sep 14 12:47:07 thanks, i will try echo or cat, for the time being Sep 14 12:50:43 shouldn't qemu depend on native recipes? Sep 14 13:10:42 ah, I bet BBCLASSEXTEND adds -native to DEPENDS? Sep 14 13:11:33 yes Sep 14 13:38:14 hi all Sep 14 13:38:47 I am getting errors in building target nkf Sep 14 13:39:31 anyone else faced this problem? Sep 14 13:42:01 i don't think that is a recipe that is build very often Sep 14 13:45:18 ya, that's right Sep 14 13:47:01 can't test it now, but what is the problem ? Sep 14 13:47:06 do you have a tinderbox log ? Sep 14 13:47:46 http://wiki.openembedded.net/index.php/How_do_I_send_automatic_success_and_failure_reports Sep 14 13:53:12 OK, now getting mmap: permission denied Sep 14 13:53:20 looks like when trying to run qemu-arm Sep 14 13:54:22 Crofton: on ubuntu? I remember someone on ubuntu was having a problem that sounded similar Sep 14 13:54:40 and he had to echo some value somewhere, doh, if I only could remember Sep 14 13:56:26 Fedora 13 Sep 14 13:59:59 oh hmm, I'm on fedora 13 Sep 14 14:00:23 and i did build an image for ARM with binary locale generation enabled (I think qemu is used for that?) Sep 14 14:03:47 eFfeM_work: i just tried it with oestat, http://tinderbox.openembedded.org/builds/86629/ Sep 14 14:05:38 Jin^eLD, yes Sep 14 14:05:39 weird Sep 14 14:05:42 hmmm Sep 14 14:06:21 Crofton: i686 Sep 14 14:06:32 I'm on x86_64 btw Sep 14 14:06:33 selinnux is denying it Sep 14 14:06:45 dunno then :/ Sep 14 14:06:46 aah, selinux is the first thing I turn off after installing ;) Sep 14 14:06:53 heh Sep 14 14:07:01 heh Sep 14 14:07:08 makes a lot of problems like these go away ;) Sep 14 14:07:11 if the NSA thought it was a good idea, I leave it on :) Sep 14 14:07:32 there are plenty of good ideas out there, but then there are the working and the not yet quite working ones ;) Sep 14 14:07:35 and someone has to be able to answer selinux + OE questions Sep 14 14:09:02 fahad: thre is no info on that page Sep 14 14:09:15 i just posted it on pastebin Sep 14 14:09:16 http://pastebin.com/9SA0tS1T Sep 14 14:09:53 the gnu hash stuff is simple, has to do with ldflags, don't know details but google can tell you Sep 14 14:10:46 see eg what this patch does Sep 14 14:10:47 http://www.mail-archive.com/openembedded-devel@lists.openembedded.org/msg00625.html Sep 14 14:12:00 thanks, i am trying this now Sep 14 14:18:51 later, have fun Sep 14 14:30:22 03Koen Kooi  07org.openembedded.dev * r3b1b9336b4 10openembedded.git/recipes/opencv/libdecodeqr_git.bb: libdecodeqr: add 0.9.4+git Sep 14 14:34:13 03Koen Kooi  07org.openembedded.dev * rd374d1c754 10openembedded.git/conf/distro/include/angstrom-2008-preferred-versions.inc: angstrom 2008: switch to cairo 1.10 Sep 14 15:18:08 Hi, what's the best way of overriding "autoconf" in a recipe? Sep 14 15:18:31 autoconf, which runs during do_configure, breaks my configure script Sep 14 15:18:40 I'd like to have it not run. Sep 14 15:22:57 do_configure () { Sep 14 15:22:59 oe_runconf Sep 14 15:23:00 } Sep 14 15:24:51 Thank you, kergoth. Sep 14 15:25:15 np Sep 14 15:25:38 there are, of course, complications resulting from not running those. if it uses libtool, you'll almost certainly fail to build, for example Sep 14 15:25:51 the better bet is to fix it to work with current autoconf Sep 14 15:26:12 but avoiding the run is a good temporary fix Sep 14 15:28:15 hi, has nayone here tries to build gs (ghostscript) ? It fails here because of tools compiled for the target but used on the host. I've prepared a fix for this but I would like to be sure others also have the problem. Sep 14 15:28:39 sorry, I put the right letters at the right place : has anyone here tried to build ... Sep 14 15:38:39 It's a nethack recipe which already has a couple of hacks in, and hasn't been accepted yet into OE after a year so it should be fine :) Sep 14 15:38:58 hehe Sep 14 16:01:54 03Klaus Kurzmann  07org.openembedded.dev * r6f04cfcf9b 10openembedded.git/recipes/xserver-common/ (3 files in 2 dirs): Sep 14 16:01:54 xserver-common: patch to add a section for nokia900 to xserver-common Sep 14 16:01:54 Signed-off-by: Klaus Kurzmann Sep 14 16:01:54 03Klaus Kurzmann  07org.openembedded.dev * ra54c780914 10openembedded.git/recipes/watchdog/ (4 files in 2 dirs): Sep 14 16:01:54 watchdog_5.6.bb: add config for the nokia900 machine Sep 14 16:01:54 Signed-off-by: Klaus Kurzmann Sep 14 16:16:12 ibot, message for zecke: your freyther.de mail is bouncing Sep 14 16:28:52 ooh Sep 14 16:28:54 http://lateral.netmanagers.com.ar/weblog/posts/BB917.html Sep 14 16:28:55 i like that Sep 14 16:29:24 using an iterator is very clean Sep 14 16:45:03 hmm, seems I found a bug in opkg, pkg_parse_line function will miss certain fields Sep 14 16:46:22 bug in opkg? i guess there is a first time for everything. Sep 14 16:46:30 still, bit of a shocker Sep 14 16:46:43 are you being sarcasting or are you serious? :) Sep 14 16:46:50 sarcastic that is Sep 14 16:47:23 I really have no idea about how good or bad opkg is, was never digging deeper there Sep 14 16:50:59 it's fairly bad Sep 14 16:51:42 hmm, considered that is's a kind of important and widely used software... it's not really nice then Sep 14 17:01:14 hi,what's the better way to reduce the memory usage while cross-compiling Sep 14 17:01:20 I've only 4GB on the build system Sep 14 17:01:40 I've the following settings: Sep 14 17:02:02 PARALLEL_MAKE = "-j 9" Sep 14 17:02:04 BB_NUMBER_THREADS = "4" Sep 14 17:02:08 maybe that's too much? Sep 14 17:06:05 maybe Sep 14 17:06:16 where's it crapping out? Sep 14 17:06:31 it's a new laptop Sep 14 17:06:44 I'll try kernel compilations Sep 14 17:06:46 without oe Sep 14 17:06:59 and I bet it won't block Sep 14 17:07:13 oops Sep 14 17:07:28 I'll restart from scratch Sep 14 17:07:39 I've a new laptop, Sep 14 17:07:57 when cross-compiling stuff it uses a lot of ram Sep 14 17:08:08 more than what i have Sep 14 17:08:13 and it start using the swap Sep 14 17:08:17 and the laptop freeze Sep 14 17:08:25 maybe my swap has an issue? Sep 14 17:08:37 but anyway if I reduce the ram used.... Sep 14 17:08:40 GNUtoo: I use 4/4 settings, have 4GB too, but even that seems too high, because working while OE is building something big in the background is tough Sep 14 17:08:55 everything lags terribly Sep 14 17:08:59 ok Sep 14 17:12:38 oh man Sep 14 17:13:01 two days solving a problem i had already solved some months ago. Sep 14 17:13:38 ups, wrong window Sep 14 17:20:41 python-native question: actually python-native is built using UCS2 build (unicode size), this create atrouble when trying to run sdk's gdb on a distribution which has python UCS 4 (example ubuntu 9.10 or 10.04) as I get this error : Unexpected gdb stderr: "/usr/local/angstrom/arm/bin/arm-angstrom-linux-gnueabi-gdb: symbol lookup error: /usr/local/angstrom/arm/bin/arm-angstrom-linux-gnueabi-gdb: undefined symbol: PyUnicodeUCS2_FromEncodedOb Sep 14 17:21:01 hmmm no money for ram Sep 14 17:21:04 is it a bad thing to switch python-native to ucs4 by adding --enable-unicode=ucs4 to configure ? Sep 14 17:21:06 ram isn't cheap Sep 14 17:21:08 hi ericben Sep 14 17:21:12 hi GNUtoo Sep 14 17:21:23 lowering -g and BBTHREADS is cheapper ;-) Sep 14 17:21:35 indeed but I want speed too Sep 14 17:21:39 exact reasons of the problem is detailed here : http://docs.python.org/faq/extending#when-importing-module-x-why-do-i-get-undefined-symbol-pyunicodeucs2 Sep 14 17:22:07 any idea from python experts of potential problems to switch to ucs4 for python-native ? Sep 14 17:22:53 hmmm about python Sep 14 17:23:29 I've an idea: compcache? Sep 14 17:23:44 what is compcache ? Sep 14 17:24:10 basically compressed swap in ram Sep 14 17:24:22 else I could get a faster swap Sep 14 17:24:28 because my swap is encrypted Sep 14 17:24:31 compression will have a CPU cost Sep 14 17:24:35 ok Sep 14 17:24:45 why do you encrypt it if you want speed ? Sep 14 17:25:30 I now want speed Sep 14 17:25:58 but anyway using a raid-like swap won Sep 14 17:26:05 using BB=2 and PARALLEL=4 will be not bad with laptop hard drives Sep 14 17:26:09 't help I bet Sep 14 17:26:15 ok Sep 14 17:26:20 It'try that Sep 14 17:27:11 anyone for python-native ucs2 vs ucs4 ? Sep 14 17:27:31 for python I read but I was not aware of the problem....but... Sep 14 17:27:42 I know mickey|away maintained python stuff Sep 14 17:27:42 Python before 3.0 doesn't handle unicode very well at all Sep 14 17:27:43 but... Sep 14 17:27:47 he's on hollidays Sep 14 17:28:47 nar: it's enabled in ubuntu using 2.6 python Sep 14 17:29:23 Yeah. It's always enabled, it just doesn't handle unicode very well. You probably won't run into any problems with ucs4 that you wouldn't run into with ucs2 Sep 14 17:30:09 ok, that would solve my compatibility problem (trying to run sdk's gdb from qtcreator which seems to use python things to talk to gdb) Sep 14 17:30:27 What version of gdb? Sep 14 17:30:55 GNU gdb (GDB) 7.1 Sep 14 17:32:44 ahh, ok, so with python stuff built into gdb Sep 14 17:32:48 ok I'm trying to add the configure option to check this solves my problem and f that works I'll send a patch for ack Sep 14 17:33:07 I'd verify what version python your gdb is built against -- you might not need to rebuild Sep 14 17:33:59 nar: I've checked it : servers's python has ucs4, workstation python has ucs4, python-native build by OE has ucs2 Sep 14 17:34:07 ahh indeed Sep 14 17:34:10 you're on the right path then Sep 14 17:34:43 server has 2.6.4, oe builds 2.6.5, workstation has 2.6.5 Sep 14 17:34:48 rgr Sep 14 17:35:06 will let you know the result later, dinner time Sep 14 17:38:15 ok indeed for me too Sep 14 18:20:52 kergoth: ping Sep 14 18:38:51 gm Sep 14 18:50:36 gm all Sep 14 18:50:43 gm khem Sep 14 18:50:48 likewise: hey Sep 14 18:53:54 is it better to use git tags in SRCREV or by adding "tag=" to SRC_URI? both work, but from bitbake fetcher perspective, which one is better? Sep 14 18:54:28 Is it different from the fetcher pov? Sep 14 18:55:27 not sure Sep 14 18:56:35 It shouldn't be much different, if any really, iirc Sep 14 18:56:55 if "tag=" used in SRC_URI, you need to remove SRCREV. and if you forget and there's different SRCREV, I don't know which one will be prefered - the tag in SRC_URI or in SRCREV. Sep 14 18:56:55 I kinda think tag= is just translated to checkout this, rather than what's passed in as the rev Sep 14 18:57:37 won't it checkout by rev anyway? Sep 14 18:57:42 yeah, my quick read of the code is it uses the tag Sep 14 18:57:47 unless tag is master Sep 14 19:00:06 does it translate internally to "tag=SRCREV" anyway? Sep 14 19:03:50 rev is put into the tag variable, yes Sep 14 19:03:53 if i read this right :) Sep 14 19:04:37 yeah, that was my understanding too, thanks Sep 14 19:18:19 03Simon Busch  07org.openembedded.dev * r60716e8657 10openembedded.git/recipes/netbase/ (netbase/palmpre/interfaces netbase_4.21.bb): Sep 14 19:18:20 netbase: add interface file for palmpre machine and bump PR Sep 14 19:18:20 Signed-off-by: Simon Busch Sep 14 19:18:22 03Simon Busch  07org.openembedded.dev * rdaa103a9c6 10openembedded.git/recipes/linux/ (linux-palmpre/defconfig linux-palmpre_git.bb): Sep 14 19:18:23 linux-palmpre: enable uinput support in defconfig Sep 14 19:18:23 Signed-off-by: Simon Busch Sep 14 19:18:23 03Simon Busch  07org.openembedded.dev * r6464424bfd 10openembedded.git/recipes/tslib/tslib_git.bb: Sep 14 19:18:23 tslib_git.bb: bump SRCREV Sep 14 19:18:23 Newer version of tslib contains better support for the cy8mrln touchscreen used on the Sep 14 19:18:24 palmpre machine. Sep 14 19:18:24 Signed-off-by: Simon Busch Sep 14 19:22:39 hi, can someone push this: http://thread.gmane.org/gmane.comp.handhelds.openembedded/36911 don't have my ssh key here and want to start a build with this fix to see if thigns now build fine Sep 14 19:26:10 patchwork url? Sep 14 19:29:12 eFfeM_school: will do Sep 14 19:30:14 re Sep 14 19:37:08 khem, is MIPS64 ld still giving you fits? Sep 14 19:37:54 nar: configuring python-native with ucs4 solved the gdb problem on my workstation Sep 14 19:41:57 ka6sox-work: yes Sep 14 19:42:09 ka6sox-work: it works well on native ubuntu64 Sep 14 19:42:24 ka6sox-work: I will dig more Sep 14 19:43:12 likewise: thanks, don't have a git tree here at school Sep 14 19:44:00 need to go now before I'm kicked out, catch you in an hour Sep 14 19:49:04 looks like the patchqueue has been growing lately - is that just due to the summer holiday ? Sep 14 19:49:21 And people not clearing out their stuff? Sep 14 19:52:13 I have several patches that need ack beforoe being pushed : Sep 14 19:52:23 http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024052.html Sep 14 19:52:53 http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024167.html or http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024166.html Sep 14 19:53:00 http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024192.html Sep 14 19:53:07 http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024231.html Sep 14 19:53:14 http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024233.html Sep 14 19:53:46 sorry if some of them seems basic, until I get more experience I prefer not breaking things even if that works for me (c) :) Sep 14 19:54:01 http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024192.html has my ack Sep 14 19:54:25 http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024231.html too Sep 14 19:54:50 http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024233.html too Sep 14 19:59:46 thanks khem Sep 14 20:00:51 03Eric BENARD  07org.openembedded.dev * r203eb8601e 10openembedded.git/recipes/python/python-native_2.6.5.bb: (log message trimmed) Sep 14 20:00:51 python-native: configure unicode to ucs4 Sep 14 20:00:51 * while trying to run gdb from an SDK created by OE, I got the following Sep 14 20:00:51 errors: /usr/local/angstrom/arm/bin/arm-angstrom-linux-gnueabi-gdb: undefined symbol: Sep 14 20:00:51 PyUnicodeUCS2_FromEncodedObject Sep 14 20:00:52 * the reason is that the workstation (ubuntu 10.04 or 9.01)'s Python is compiled using ucs4 Sep 14 20:00:53 and the python-native used to create arm-angstrom-linux-gnueabi-gdb is using ucs2, the Sep 14 20:00:56 03Eric BENARD  07org.openembedded.dev * rae447cd948 10openembedded.git/recipes/busybox/files/mdev.conf: Sep 14 20:00:56 busybox/mdev.conf: add net/tun Sep 14 20:00:56 having tun in /dev/net/tun allows openvpn to run without any hack Sep 14 20:00:56 on the target Sep 14 20:00:56 Signed-off-by: Eric Bénard Sep 14 20:00:57 Acked-by: Khem Raj Sep 14 20:00:57 03Eric BENARD  07org.openembedded.dev * re6094c0b00 10openembedded.git/recipes/tasks/task-qte-toolchain-target.bb: Sep 14 20:00:58 task-qte-toolchain-target: add libqt-embeddedmultimedia4-dev Sep 14 20:00:58 Signed-off-by: Eric Bénard Sep 14 20:00:59 Acked-by: Khem Raj Sep 14 20:05:59 there should be ack button in patchwork too Sep 14 20:09:24 who are the admins of patchwork ? Sep 14 20:09:44 Crofton is one I guess Sep 14 20:10:02 it would be great if the committers could close the patches they commit when the submiter has'nt done it Sep 14 20:10:14 yes thats what I do Sep 14 20:10:18 generally Sep 14 20:10:27 I pick patches from patchwork Sep 14 20:10:40 it gets all authors signoffs right Sep 14 20:10:44 I dont have to worry Sep 14 20:10:48 and picking it close the patch ? Sep 14 20:10:58 no Sep 14 20:11:02 I have to do it manually Sep 14 20:11:13 even if you're not the initial author ? Sep 14 20:11:18 yes Sep 14 20:11:35 ah ok, it's the fact that you added your signoff which gives you the right ? Sep 14 20:11:52 here are those from me that need ack too: Sep 14 20:11:55 http://patchwork.openembedded.org/patch/2543/ Sep 14 20:11:59 no I think I have special rights Sep 14 20:12:03 :) Sep 14 20:12:07 http://patchwork.openembedded.org/patch/2544/ Sep 14 20:12:09 ah OK :) Sep 14 20:12:33 http://patchwork.openembedded.org/patch/2548/ Sep 14 20:13:20 http://patchwork.openembedded.org/patch/2543/ PR bump is strange Sep 14 20:13:44 /usr/libexec/git-core should use OE vars Sep 14 20:14:18 I had asked here about the policy for PR bumps, but got no answer besides "increasing should be good" :} Sep 14 20:14:20 http://patchwork.openembedded.org/patch/2548/ Sep 14 20:14:27 same you have to use OE vars Sep 14 20:14:35 /usr/lib/python*/site-packages/gtk-2.0/rsvg.* Sep 14 20:14:38 is not ok Sep 14 20:14:57 there is PR bump policy on wiki Sep 14 20:15:22 http://wiki.openembedded.net/index.php/When_To_Bump_PR Sep 14 20:15:27 ok ths Sep 14 20:15:30 thx Sep 14 20:16:23 what is the problem with the rsvg files ? I wrote that to avoid hardcoding the python version Sep 14 20:18:10 yann: /usr/lib is ${libdir} Sep 14 20:18:55 khem: the wiki does tell when, but not about the amount - I was not comfortable to do +1 before knowing whether anyone would commit something else in the meantime - that was meant to make it obvious what the base revision was Sep 14 20:19:01 Running task 51500 of 80400 Sep 14 20:19:10 ERROR: [Errno 28] No space left on device Sep 14 20:19:12 :( Sep 14 20:19:18 and it seems there is ${PYTHON_DIR} in classes/pythoin-dir.bbclass Sep 14 20:19:42 ericben: yeah I was bitten wit that catch 22 Sep 14 20:20:00 khem: here bitbake world is stuck at ~50k and doesn't manage to progress anymore Sep 14 20:20:15 yann: PRs are usually integers Sep 14 20:20:18 but now I know the new build machine is stable :) Sep 14 20:20:26 r6 and then r7 Sep 14 20:30:58 khem: ok, will reroll all of that Sep 14 20:34:36 khem: http://www.amazon.com/Western-Digital-Intellipower-Desktop-WD20EARS/dp/B002ZCXK0I/ref=sr_1_sc_1?ie=UTF8&qid=1284496215&sr=8-1-spell :-) Sep 14 20:34:57 "Amazon Has Certified That This Packaging Is Frustration-Free" Sep 14 20:35:08 WTF? :-) Sep 14 20:35:53 likewise, lol what if you're frustrated while messing with the firmware of the hdd? Sep 14 20:36:12 don't know! :-) BTW, I hope they deliver a top cover with that drive. ;-) Sep 14 20:36:15 khem: any comments about my newly-sent http://patchwork.openembedded.org/patch/2952/ ? Sep 14 20:36:26 like for instance if you reverse enginner the firmware, you surely get a lot frustrated Sep 14 20:36:34 lol Sep 14 20:37:43 likewise: thanks :) this is a laptop but I need a serious hardware upgrade Sep 14 20:38:12 khem: http://www.amazon.com/Minnesota-Vikings-Pro-Grip-Hammer/dp/B000HX1NBK/ref=sr_1_9?s=electronics&ie=UTF8&qid=1284496436&sr=1-9 Sep 14 20:38:42 high quality stainless steel head, fiberglass shaft. That should get the spindle going. Sep 14 20:39:41 likewise: the second item I have 3 of them in different sizes and I use them based on frustration level :) Sep 14 20:52:03 ericben, do you have an account on pw Sep 14 20:52:16 the account system for it sucks by the way :) Sep 14 20:53:50 Crofton: yes I do Sep 14 20:54:10 username is ebenard I think Sep 14 20:54:27 khem: out of curiosity - the only difference in recipes/git between master and shr/testing2010 is a PR bump (because of libssl) in master - I suppose my patches should be rerolled for the master PR, but what should the PR be when porting back to shr ? Sep 14 20:56:29 what does #oe use to log its irc transcripts? anyone know? Sep 14 20:56:38 home :-) Sep 14 20:56:51 jconnolly: rikers ibot Sep 14 20:57:07 see rikers.org Sep 14 20:57:24 eFfeM: excellent I'll look into it. thanks eFfeM. I want to do the same for #buglabs... so many times I grep through my own irc logs to find a simple fix for something... would be nice to make it available to others Sep 14 20:57:45 ok Sep 14 20:57:52 I'l make sure it has admin rights Sep 14 20:58:43 ok thanks Sep 14 21:01:47 ericben, can you check the username again Sep 14 21:02:07 jconnolly: send tim riker an email if you are channel op, he'll add the bot for you, i did the same for hawkboard Sep 14 21:02:29 eFfeM: great thanks Sep 14 21:04:08 yw Sep 14 21:05:12 03Graham Gower  07org.openembedded.dev * rc4bdd84957 10openembedded.git/recipes/foonas-init/ (8 files in 2 dirs): (log message trimmed) Sep 14 21:05:12 foonas-init: remove broken and unused recipe. Sep 14 21:05:12 This recipe has references to non existant files in its SRC_URI, I could not Sep 14 21:05:12 find them in the git logs either. Sep 14 21:05:12 A grep -r foonas-init shows no in tree users. Sep 14 21:05:13 Signed-off-by: Graham Gower Sep 14 21:05:14 Verified with Tim Ellis and Oyvind Repvik that this can be deleted Sep 14 21:05:21 03Frans Meulenbroeks  07org.openembedded.dev * reb9f07b76e 10openembedded.git/recipes/fftw/ (fftw_3.2.2.bb fftwf_3.2.2.bb fftwl_3.2.2.bb): Sep 14 21:05:22 fftw: added --enable-threads to the 3.2.2 recipes Sep 14 21:05:22 Signed-off-by: Frans Meulenbroeks Sep 14 21:06:03 Crofton: ebenard, email eric eukrea.com Sep 14 21:06:06 yann: whatever you have it should be larger than that Sep 14 21:06:11 grrr Sep 14 21:09:01 I need to delete all the spam user names Sep 14 21:09:26 user admin on pw is getting to be a nightmare Sep 14 21:09:40 ericben, you should be able to change states now Sep 14 21:09:52 Crofton: thanks Sep 14 21:44:39 Tartarus: for soe form of RedHat : is Fedora ok ? Sep 14 21:48:49 hi dd Sep 14 21:49:23 ericben, sure Sep 14 21:49:35 this should create some trouble there ? Sep 14 21:49:36 I'd really love to see an older thing like centos5 also tried Sep 14 21:49:47 Well, diff python versions compiled possibly differently :) Sep 14 21:50:28 what is the link between host python versions and that change ? Sep 14 21:50:46 maybe I misread, rereading Sep 14 21:51:07 * the reason is that the workstation (ubuntu 10.04 or 9.01)'s Python is compiled using ucs4 Sep 14 21:51:07 and the python-native used to create arm-angstrom-linux-gnueabi-gdb is using ucs2, the Sep 14 21:51:07 problem is detailed here : Sep 14 21:51:08 http://docs.python.org/faq/extending#when-importing-module-x-why-do-i-get-undefined-symbol-pyunicodeucs2 Sep 14 21:51:24 So how is the host python built on redhat, etc? Sep 14 21:51:27 ok I understand, to test I should not compile test but test the SDK on other hosts Sep 14 21:51:40 Yes, make sure the result works elsewhere Sep 14 21:51:45 as well as ever anyhow Sep 14 21:51:54 ucs Sep 14 21:52:21 What I'm saying is move and retry your testcase on redhat :) Sep 14 21:52:45 well that may be compicated as to trigger the problem I have to use QTCreator which interfaces with gdb through python so I hope QTcreator will work fine Sep 14 21:53:26 ok I checked the SRPM of Redhat 9 (python 2.3 and they have : --enable-unicode=ucs4 Sep 14 21:54:19 the FAQ says : RedHat Linux 7.x, in particular, provided a “python2” binary that is compiled with 4-byte Unicode. so in fact RedHat was in advance and it should run on redhat from 7.x ! Sep 14 21:58:05 but I'll try a VM with RedHat or Fedora asap Sep 14 21:58:45 k Sep 14 22:00:50 btw, for the sdk, I have library name problems depending on the host Sep 14 22:01:29 for this one I had to make symlinks for libtinfo and libmpfr Sep 14 22:01:49 is there a way to prevent this ? Sep 14 22:02:08 foro libtinfo: this is because the host puts tinfo into ncurses Sep 14 22:02:42 for libmpfr, this is because the binary was linked to .so.4 but the host only had .so.1 's symlink Sep 14 22:02:49 Well, the SDK binaries are not cross host compatible Sep 14 22:02:52 We don't build them to be Sep 14 22:04:18 OK Sep 14 22:04:35 Tartarus: to what extend are the SDK binaries portable across hosts or relocatable within one host? Not at all I assume? Sep 14 22:05:03 not at all portable, should be fully relocatable Sep 14 22:05:12 assuming you don't use the .rc file anyhow Sep 14 22:05:20 (without fixing it up for where you put it) Sep 14 22:05:39 here we are using them on several debian & ubuntu versions without major problem, just some libraries links to fix sometimes Sep 14 22:06:17 You could call that not a major problem, if you want :) Sep 14 22:06:27 but this was until I learn this is not the right thing to do:-) Sep 14 22:06:41 yes of courses this is a major problem the 1st time :) Sep 14 22:06:52 Switching to using a static version of mpfr/gmp isn't too hard Sep 14 22:06:59 Haven't had to tackle the ncurses issue yet Sep 14 22:07:32 foro ncurses this is triggered using gdb Sep 14 22:08:30 using qte sdk 2 lib symlinks and a symlink for mkspecs, QTCreator is happy with OE's sdk and cross compiles QT projects fine Sep 14 22:09:09 I plan to work on qte sdk to generate one fully compatible with what QTCreator expects Sep 14 22:11:44 You work with Qt a lot? Sep 14 22:12:08 likewise: I don't but customers ask for this so I discovered QTCreator yesterday ! Sep 14 22:13:19 ericben: I learned it last year at the Qt training days. Didn't touch Qt afterwards, too little time. Sep 14 22:13:27 it interfaces well on the SDK and can use gdb + gdbserver to get one clic remote debug + has support for git Sep 14 22:14:49 btw Cliff Brake has done review of QTCreator on his website : http://bec-systems.com/site/704/qt-creator-for-cc-development Sep 14 22:16:38 showing it's not only QT centric Sep 14 22:24:57 hello folks! I try to build my own image but when I add "xserver-xorg" to IMAGE_INSTALL I get a "satisfy_dependencies_for" error message which misses the libxau6 and libxdmcp6 packages; oe builds libxau and libxdmcp packages but xserver-xorg needs the *6 packages - what is the problem here? Sep 14 22:40:11 gn all Sep 14 22:40:48 bye Sep 14 22:56:18 * Tartarus gets bit, again, by angstrom not using sane-toolchain.inc Sep 14 22:56:59 Excuse me if this is a stupid question, but why do certain packages end up with different package names? Sep 14 22:57:26 debian.bbclass, often Sep 14 22:57:35 eglibc ends up as libc6, zlib ends up as libz1, libgcc ends up as libgcc1 Sep 14 22:57:41 Is this intentional? Sep 14 22:57:45 Ah Sep 14 22:57:56 This a new addition for compatability? Sep 14 22:58:02 I haven't used OE in 6 months Sep 14 22:58:12 Naw its been around for ages. Sep 14 22:59:55 Ah Sep 14 23:00:01 Maybe it was a distro thing Sep 14 23:00:06 yes Sep 14 23:00:36 I also see that DEPENDS no longer has dependancies with their versions Sep 14 23:00:51 libc6 ( <= 2.11) or something like that. Sep 14 23:02:43 Distro-related again? Sep 14 23:11:01 metadata hasn't ever supported that Sep 14 23:11:04 it just didn't puke on it Sep 14 23:11:08 ipks still support it Sep 14 23:29:02 Tartarus: is nandfs implementation available for lnx do you know Sep 14 23:29:21 Tartarus: or ubifs has all goodies Sep 14 23:30:47 http://git.iksaif.net/?p=nandfs.git;a=summary Sep 14 23:30:57 seems to be 19 months old Sep 14 23:31:40 ubifs works fine here on several arm targets (atmel, freescale, TI) Sep 14 23:31:47 on NAND & NOR flashs Sep 14 23:32:14 ericben: what are adv on ubifs when it comes to nand Sep 14 23:32:22 on say yaffs2 Sep 14 23:32:45 khem: vs yaffs2 : I don't know Sep 14 23:32:52 or in general Sep 14 23:32:58 vs jffs2 : mount time very short Sep 14 23:33:01 whats good about ubifs Sep 14 23:33:05 scalable ? Sep 14 23:33:23 good for large flashs size Sep 14 23:33:39 right Sep 14 23:34:14 + UBI which seems to allow complicated things for people who need it Sep 14 23:34:24 * Tartarus checks nandfs Sep 14 23:34:48 Hadn't heard of nandfs until now Sep 14 23:34:53 khem: http://www.embedded-linux.co.uk/downloads/ESC-5.4-flash_filesystems-slides.pdf Sep 14 23:34:56 But yes, UBI allows for complicated things Sep 14 23:34:59 same here for nandfs Sep 14 23:35:07 And abstracts away the wear leveling Sep 14 23:35:41 iirc now Sep 14 23:35:54 You can do say 2 partitions of stuff on one chip in a ubi volume Sep 14 23:35:59 complicated means what here Sep 14 23:36:00 and get wear leveling over the whole volume Sep 14 23:36:11 ah i see Sep 14 23:36:34 UBIFS supports write caching Sep 14 23:36:37 thats cool Sep 14 23:36:39 UBI allow for flash what LVM allow for disks Sep 14 23:36:53 dynamic volumes Sep 14 23:36:58 resizing Sep 14 23:37:01 That part of LVM, yes :) Sep 14 23:37:25 UBI is part of LVM? Sep 14 23:37:27 no Sep 14 23:37:32 ah ok Sep 14 23:37:48 just the concepts of LVM you said are part of UBI concepts too Sep 14 23:37:52 yes Sep 14 23:37:52 Not all LVM concepts ;) Sep 14 23:38:23 possible, I don't really know LVM Sep 14 23:38:23 I think somewhere there's benchmarks of ubi/ubifs vs yaffs2 vs some other stuff Sep 14 23:38:28 grg: hello Sep 14 23:38:28 And there's wins and losses Sep 14 23:38:35 logfs can also be interesting to throw into the mix Sep 14 23:38:38 mornin' Sep 14 23:39:05 http://tree.celinuxforum.org/CelfPubWiki/ELCEurope2008Presentations?action=AttachFile&do=get&target=flash-filesystems.pdf Sep 14 23:39:51 ubi is a good thing for u-boot Sep 14 23:39:53 and ubifs is mainlined Sep 14 23:43:12 nice Sep 14 23:43:20 so ubifs seems to be future Sep 14 23:43:48 the present, Nokia has it on the N900 :) Sep 14 23:44:55 and android works on ubifs but not on jffs2 Sep 14 23:47:55 mainlined android does ubi now? Sep 14 23:48:04 Or just has been done (we did that as a one off back in 1.5 days) Sep 14 23:48:25 we have i.MX51 board running froyo on ubifs here Sep 14 23:48:55 ericben: what products does your company make Sep 14 23:48:55 yeah, but did you guys add that (or whoever did the i.mx51 port) or is that all in google's codebase? Sep 14 23:49:27 I think this is a local hack Sep 14 23:50:36 khem: we are designing boards, bringing them up and porting the OS on them Sep 14 23:51:07 meego uses btrfs Sep 14 23:51:29 Tartarus: I have to check this is an intern who worked on this so I don't have all the details yet Sep 14 23:52:01 k Sep 14 23:52:23 It shouldn't be a big deal to do so, yeah, it's just adding the tools and a little makefile magic Sep 14 23:52:32 google officially uses yaffs2 still i bet, heh Sep 14 23:53:20 Tartarus: yes the official FS seems to be still yaffs2 Sep 14 23:53:37 how exactly are the ipk files generated during an OE build? Sep 14 23:54:09 I see that the feed file has fields like "HomePage" but opkg knows nothing about it Sep 14 23:55:47 ah, found it I think, its the package_ipk bbclass Sep 14 23:56:12 hm sort of Sep 14 23:56:48 meego = moblin + maemo Sep 14 23:57:07 may be they should use poky/oe :) Sep 14 23:57:11 = Intel + Nokia :) Sep 14 23:57:46 can they stop the droids Sep 14 23:57:54 hope they make progress on battery life once they put a iCore in a smart phone Sep 14 23:58:56 because actually, when I want ot be sure to be able to phone for one day, I have to take a basic nokia as a rescue when android has eaten all the energy of my Acer Liqui :-) Sep 14 23:59:00 unless they have an arm vendor as partner in meego effort it might be difficult if they go solo on atom Sep 14 23:59:29 this would be a strange strategy for Intel to come back to ARM Sep 14 23:59:39 not so bad for nokia Sep 14 23:59:53 and if you have enough money you can have your say Sep 15 00:00:14 I wonder if nokia is going to go intel only in future Sep 15 00:00:16 yes but nokia is (was ?) partner with TI and now they chose Intel Sep 15 00:00:32 nokia is losing anyway Sep 15 00:00:50 they need to have a smartphone seriously Sep 15 00:01:10 and cortex-a8/a9 is the easiest path Sep 15 00:01:49 hm thats odd, package_ipk.bbclass writes "Homepage" but in the Packages file it becomes "HomePage", so far could not figure out who is doing this Sep 15 00:01:57 they've just changed their manager so that may change in the future Sep 15 00:02:22 the new one comes from Microsoft ... Sep 15 00:03:12 hehe Sep 15 00:03:24 windoze mobile 7 Sep 15 00:03:32 they can keep playing this cat and mouse game Sep 15 00:03:49 one needs to invest in one path and remain on it to be successful Sep 15 00:04:05 android good or bad did that and they are relatively successful Sep 15 00:04:34 they have invested so much in opensource these past years to only have 2 or 3 products using Linux & maemo Sep 15 00:05:54 but if Microsoft manages to put Widnows on Nokia phones this would mean that all the software strategy of nokia (Maemo + Symbian) has failed Sep 15 00:06:14 heh Sep 15 00:07:39 nokia has such a large market share if they rolled for themselves it will be like 50% of world market in mobile phones :) Sep 15 00:08:48 one good thing they seems to have done was to puchase QT. After spending so much time to try to get a good user interface from scratch Sep 15 00:08:59 * khem nods Sep 15 00:11:08 and their new models still have Symbian not MeeGo Sep 15 00:11:26 meego is like 6mos old Sep 15 00:11:41 linaro has also hinted to use meego Sep 15 00:11:46 one other project I don't realy understand is linaro Sep 15 00:11:54 I don''t understand where they want to go Sep 15 00:14:17 prolly netbooks Sep 15 00:14:38 it has ubuntu instead of maemo+moblin Sep 15 00:17:05 we will see in the future. I fonly this can bring better mainline support for i.MX51 that will be a good point :-) Sep 15 00:22:11 builds launched, good night Sep 15 00:48:59 Jin^eLD, chances are the "HomePage" field is written out by ipkg-utils/opkg-utils. The "Homepage" spelling would therefore still be found within the .ipk control file. Sep 15 00:51:36 where has linaro hinted to use meego? I'd like to read that Sep 15 00:52:04 eFfeM: thanks again for the tip. Tim Riker hooked us up. Sep 15 01:35:00 grg: thanks for the hint, I'll have another look Sep 15 01:35:36 grg: ah, you're the opkg guy :) Sep 15 01:36:10 Jin^eLD, and you're the guy who just sent some patches :) Sep 15 01:36:16 right, right :) Sep 15 01:36:27 i like patches Sep 15 01:36:40 I'll take the chance and bug you with a couple of questions on opkg if that's ok?:> Sep 15 01:36:48 yeah, sure Sep 15 01:37:45 how usable is the library/API? I want to write a small app that would do something similar like opkg list_installed and opkg list, but output more data (like opkg info), and do that in a JSON format Sep 15 01:38:03 is that possible by using the library as a real library, or would I have to actually hack into libopkg? Sep 15 01:38:20 I saw that most things are done internally in libopkg, including printing and so on Sep 15 01:38:50 its minimally usable. Needs some work to fix up the error checking, because it currently sucks Sep 15 01:39:14 i have a gtk interface to opkg lying around somewhere... i'll dig it up and put it in a pastebin Sep 15 01:39:38 so, I would not be able to get a package list and iterate over the pkg structs outside of libopkg? Sep 15 01:40:48 yeah, you can do that Sep 15 01:40:53 ok, cool Sep 15 01:41:14 since you like patches, you may have a look at this one as well: http://gitorious.digitalstrom.org/dss-oe/dss-oe/blobs/master/targets/dss11/recipes/opkg/opkg/separate_libdirs.patch Sep 15 01:41:26 what I needed to achieve is - complete separation between two feeds/opkg configurations Sep 15 01:41:56 one part is handled nicely by OPKG_CONF_DIR Sep 15 01:42:11 but then /usr/lib/opkg is still shared and opkg list would how packages from all feeds Sep 15 01:42:20 so I added OPKG_LIB_DIR Sep 15 01:42:50 if you think that's useful for upstream you can either take it or tell me if you dislike some parts or want it reworked Sep 15 01:43:14 i understand where you're coming from... its really outside opkg's design Sep 15 01:43:27 i noticed ;) Sep 15 01:43:39 it works however Sep 15 01:44:22 basically we have the regular system feeds and then some feeds for plugins, which are kind of "special" and the user should be able to list/install them via a web based UI Sep 15 01:44:55 so it made sence to separate the system stuff from that other special stuff Sep 15 01:45:09 have you tried using the --dest flag? Sep 15 01:45:31 well.. I wanted the packages to get installed relative to / Sep 15 01:45:39 right ok Sep 15 01:46:13 because various parts of the contents go to different system directories due to various reasons Sep 15 01:46:40 I was first hoping the conf option would be my salvation, but ended up hacking the code Sep 15 01:47:38 well the patch is not at all intrusive, and i can see the benefit, so you want to clean it up a bit i'll put it in Sep 15 01:48:22 ugghh.. the indentation is rather nasty in that file Sep 15 01:48:37 indentation throughout the whole code seems a little messy Sep 15 01:48:55 very Sep 15 01:49:06 I'd run it through some tool or something Sep 15 01:49:17 i have fixed it here and there, but there is still a lot of tab/space misuse. Sep 15 01:49:31 what would you want me to clean up in the patch? indents + remove patch comments? Sep 15 01:49:48 i briefly played with indent -i8 -kr, but it made other things really ugly Sep 15 01:50:06 yeah, those are the changes i'd like Sep 15 01:50:37 ok, I'll do that and send it to the ml later today (after sleep time ;) Sep 15 01:50:44 no probs Sep 15 01:50:54 there's another one: http://gitorious.digitalstrom.org/dss-oe/dss-oe/blobs/master/targets/dss11/recipes/opkg/opkg/print_homepage.patch Sep 15 01:51:08 basically it enables parsing of the homepage field that OE is adding Sep 15 01:51:14 and also adds it to the info output Sep 15 01:51:28 not sure if you'd be interested in that though Sep 15 01:52:42 ah crap, I think I forgot two debug printouts there :) Sep 15 01:54:27 anyway, I'm off for now, thanks for the hints Sep 15 01:54:31 nite everyone Sep 15 01:54:35 bye Sep 15 02:12:09 Jin|away: fyi, i uploaded some opkg interfacing code. http://code.google.com/p/opkg/source/browse/#svn/branches/gtkopkg Sep 15 02:23:36 debian has the notion of "Essential" packages, ones that the package manager will refuse to remove. Opkg supports this too. Is there a reason why this was never used in OE, or is it just that no one implemented it? Sep 15 02:29:21 probably no one implemented Sep 15 02:29:39 There's a lot of things that package managers could be supporting that oe metadata isn't today apparently **** ENDING LOGGING AT Wed Sep 15 02:59:57 2010