**** BEGIN LOGGING AT Tue Sep 28 02:59:57 2010 Sep 28 03:08:01 no, i mean it should be possible to add, say, a curl fetcher, and the user should be able to specify whether they want wget or curl Sep 28 03:12:23 kergoth: I see Sep 28 03:21:53 kergoth, what's the advantage of supporting both curl and wget? Sep 28 03:22:52 that's not a particularly useful case, but it could simply be the availability -- wget can be built by oe, a pycurl based one would require the python dependency, for example Sep 28 03:23:48 yeah, but you just end up having to maintain more code. And if you're trying to diagnose a user's problem, its one more variable that needs to be determined Sep 28 03:25:54 what part of "that's not a particularly useful case" fails to sink in? Sep 28 03:25:55 if it was a different fetcher for, e.g. a BSD system (ftp instead of wget) that would make sense Sep 28 03:26:38 i'm sure you know better that i do... Sep 28 03:27:11 no, you're just being oddly dense Sep 28 03:27:53 could be Sep 28 03:27:55 :) Sep 28 03:40:35 hmm, no, I'm the one being dense. the main other case i was thinking of was custom content delivery mechanisms for companies, but it's just as easy to use a custom uri scheme and utilize it via PREMIRRORS rather than keeping the same schemes and prioritizing the fetcher Sep 28 03:40:37 heh Sep 28 05:49:39 khem, i've build console-image from your branch for beagleboard/minimal without problem Sep 28 05:49:52 however probably don't have time to test it before thursday evening Sep 28 07:09:09 morning Sep 28 07:10:28 good morning Sep 28 07:23:00 moin Sep 28 07:33:20 morning Sep 28 07:40:36 hi Sep 28 07:40:43 hi woglinde Sep 28 07:41:03 hm qemuarm kernel has no real usb support Sep 28 07:46:07 huh, wie nochmal? Sep 28 07:46:10 oops wrong window Sep 28 07:46:27 :) Sep 28 07:52:38 good morning Sep 28 07:53:03 hi ant Sep 28 07:53:06 hi Sep 28 08:01:48 the NOTE: /kernel/arch/arm/mach-at91/at91x40_time.c Sep 28 08:01:56 bug is really annoying Sep 28 08:06:19 it's a bug? Sep 28 08:07:16 sure Sep 28 08:07:26 I need 10 minutes to finish kernel-package Sep 28 08:07:46 because its scrolled down every file in image/kernel Sep 28 08:08:40 isn't it better to fix it in that 10 minutes? :) Sep 28 08:08:43 hi ynezz btw. Sep 28 08:08:47 hi Sep 28 08:09:08 one fix would be to disable sane stuff Sep 28 08:09:25 but only local Sep 28 08:09:30 one regexp to rule them all :) Sep 28 08:11:00 hm xorg image should include evtouch for qemu Sep 28 08:12:01 jo florian Sep 28 08:18:50 hrms kergoths new parser code dont works when you change a class Sep 28 08:19:25 woglinde: there was some discussion before, i think ericben peeked into it and added those files to kernel-dev as they might be needed for out-of-tree module building Sep 28 08:20:06 effem I try a rm -rf after stage_do_all Sep 28 08:20:13 (actually I thought out-of-tree modules would only require kernel headers and not src, but I am definitely not an expert in all kind of out-of-tree modules that might be out there) Sep 28 08:21:04 hope is sane class dont recognize kernel dir anymore Sep 28 08:21:17 for packaging Sep 28 08:21:34 gm effem btw. Sep 28 08:21:50 it's in package.class Sep 28 08:21:58 no Sep 28 08:22:07 its in kernel.bbclass Sep 28 08:22:25 kerneldir=kernel its copied to image Sep 28 08:22:36 you mean that additional sources, right? Sep 28 08:22:36 and not put into a package Sep 28 08:22:50 and I'm talking about that pile of NOTE messages Sep 28 08:22:56 yes Sep 28 08:23:19 sane class claims kernel/ is not put into a package Sep 28 08:23:25 and goes down all files Sep 28 08:23:42 no :) Sep 28 08:23:55 ynezz please Sep 28 08:24:01 believe me Sep 28 08:24:33 I'm just looking into it Sep 28 08:24:39 NOTE: the following files were installed but not shipped in any package: Sep 28 08:24:40 NOTE: /kernel/config-2.6.34 Sep 28 08:24:47 yes, package.class Sep 28 08:24:50 line 458 :) Sep 28 08:25:00 *sigh* Sep 28 08:25:01 for root, dirs, files in os.walk(dvar): Sep 28 08:25:16 thats the outcome Sep 28 08:25:19 dvar = PKGD = ${workdir}/package Sep 28 08:25:37 the reason is kernel.bbclass which copies the kernel-dir into image Sep 28 08:25:42 yes Sep 28 08:25:46 for do_stage_all Sep 28 08:26:01 but as eFfeM said, it might be necessary Sep 28 08:26:12 so when you delete the kerneldir after copy in do_stage_all Sep 28 08:26:16 all should be fine Sep 28 08:26:17 *sigh* Sep 28 08:26:35 i've no idea whether it is necessary, that is what ericben suggested Sep 28 08:26:36 I'll just filter the output and leave it as it's :) Sep 28 08:26:37 iirc Sep 28 08:26:49 ok, sorry Sep 28 08:26:55 is this the offending line? sysroot_stage_dir ${D}/kernel ${SYSROOT_DESTDIR}${STAGING_KERNEL_DIR} Sep 28 08:27:03 no Sep 28 08:27:26 thats putting the kernelsourceheaderneedsformodule the right place Sep 28 08:27:31 or rather makes the dir Sep 28 08:27:42 the cp in the next line put it then there Sep 28 08:28:01 after the cp I am trying a rm -rf Sep 28 08:28:06 to see if it helps Sep 28 08:28:14 so why cp at all? Sep 28 08:28:19 the line after the one I pasted only cp's .config ! Sep 28 08:28:20 if then do rm -fr :) Sep 28 08:28:28 cp -fpPR ${D}/kernel/.config ${SYSROOT_DESTDIR}${STAGING_KERNEL_DIR} Sep 28 08:28:37 btw this is sysroot_stage_all_append Sep 28 08:28:48 yes yes Sep 28 08:29:54 where does the kernel get copied to ${D}/kernel ? Sep 28 08:30:03 earlier Sep 28 08:30:25 but as ericben says its needed to compile out of kernel modules Sep 28 08:30:56 kernel_do_install Sep 28 08:31:10 after kerneldir=${D}/kernel/ Sep 28 08:31:12 03Steffen Sledz  07org.openembedded.dev * rcc233c9bee 10openembedded.git/recipes/linux/ (2 files in 2 dirs): Sep 28 08:31:12 ubifs-v2.6.24: avoid kernel error if ubifs superblock read fails Sep 28 08:31:12 original commit comment from ubifs-2.6.git: Sep 28 08:31:12 .get_sb is called on mounts with automatic fs detection too, so this Sep 28 08:31:12 function should print an error if it cannot read the superblock in Sep 28 08:31:13 debug mode only (new behaviour conforms the other fs types) Sep 28 08:31:14 Signed-off-by: Steffen Sledz Sep 28 08:31:37 hm currently they are not packaged so either someone refers directly to the kernel work dir, or no one is using it atm Sep 28 08:31:52 so I'm inclined to reduce build time and rm the copy Sep 28 08:32:07 NOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO Sep 28 08:32:10 *sigh* Sep 28 08:32:26 IT IS NEEDED FOR OUT OF KERNELSOURCETREE MODULE COMPILATION Sep 28 08:32:39 i know Sep 28 08:32:45 maybe make it optional somehow Sep 28 08:32:53 2 solutions Sep 28 08:32:57 make it a package Sep 28 08:33:09 somehow and make it package-stage aware Sep 28 08:33:18 or just delete it after the copying Sep 28 08:33:55 actually didn't make myself clear, for the systems I am building I don't need out of tree modules (actually making it optional would be a better phrase) Sep 28 08:34:05 I don't get something here, why the copy and then delete? Sep 28 08:34:11 why the copy at all Sep 28 08:34:37 ynezz: they are moved to staging Sep 28 08:34:41 i guess Sep 28 08:34:44 ah, so Sep 28 08:34:50 but no idea where this happens though Sep 28 08:35:00 sysroot_stage_dir Sep 28 08:36:14 ah, that's what I said at 10.28, but then you said no, and later "or rahter makes the dir" so that lead me to think that it did not copy to staging Sep 28 08:36:24 guess that caused the confusion Sep 28 08:36:29 sorry Sep 28 08:36:32 np Sep 28 08:36:52 but you could have looked up sysroot_stage_dir yourself Sep 28 08:36:54 *g* Sep 28 08:37:08 yeah, was still finding my way Sep 28 08:38:35 wouldn't it be more effiicient do do kerneldir=${SYSROOT_DESTDIR}${STAGING_KERNEL_DIR} (and maybe move the code from kernel_do_install to sysroot_stage_all_append) ? Sep 28 08:39:41 hm yeah that would be a solution too Sep 28 08:39:53 avoids the cp and rm Sep 28 08:40:00 could you raise it to the ml? Sep 28 08:40:35 yes, will try to write a msg have some time during the next test run Sep 28 08:41:56 hm the rm dont helps Sep 28 08:42:12 packaging and sysroot stuff is running parallel Sep 28 08:42:24 as I feared Sep 28 09:20:31 woglinde: been pondering about moving the code as suggested above, but I feel adding the files to kernel-dev is better as it will allow to build out-of-tree modules on target too (should that be desired) Sep 28 09:21:41 effem yes I tought about that too Sep 28 09:21:59 but that will blow up dev with 90 mb Sep 28 09:22:11 i think ericben already made that patch Sep 28 09:22:13 and you need to define a dir Sep 28 09:22:28 in /usr/src maybee Sep 28 09:23:23 hm yeah, +90m is not nice (then again embedded systems are not really intended to compile modules etc Sep 28 09:23:48 actually purists would not even want things like gcc on it Sep 28 09:23:59 let's talk about this with ericben when he pops in Sep 28 09:24:29 this = the dev pkg, not the purist approach Sep 28 09:31:41 * woglinde scratches head why no wacom xorg input driver was packaged Sep 28 09:32:49 so, what's that quick fix for that NOTE crap? Sep 28 09:34:43 ynezz as effem said Sep 28 09:35:03 put all after kerneldir= into the sysroot_foo Sep 28 09:35:13 ok, thanks Sep 28 09:35:24 just trying to get 2.6.34.7 working on ts72xx, means quite a few of clean&build cycles, so it's NOTE hell :) Sep 28 09:35:35 mom Sep 28 09:35:47 I will paste bin the patch Sep 28 09:35:54 cool Sep 28 09:36:04 other quick fix in kernel.bbclass wouldbe Sep 28 09:36:14 add kernel dir to files of kernel-dev package Sep 28 09:37:05 FILES_kernel-dev = "/boot/System.map* /boot/Module.symvers* /boot/config* /kernel/" Sep 28 09:37:35 oh, lovely Sep 28 09:38:05 thanks Sep 28 09:38:15 will tell you the result after the next cycle Sep 28 09:38:52 clean & build should be enough to trigger that change, right? Sep 28 09:39:33 yes Sep 28 09:41:11 btw. what happend to --interactive ? Sep 28 09:41:41 this was uber cool feature, I'm missing it a lot Sep 28 09:41:46 I mean bitbake -i Sep 28 09:41:53 hm you can use -c devshell now Sep 28 09:42:04 re ant Sep 28 09:43:44 if I understand devshell correctly, it's just good for patch/compile cycles Sep 28 09:44:32 what I would like to avoid at least, is parsing of the recipes every time I run bitbake Sep 28 09:44:51 with interactive it was easy, just parse;clean;build;clean;build Sep 28 09:45:36 to mimic the same workflow I do now bitbake -c clean virtual/kernel && bitbake virtual/kernel Sep 28 09:46:37 argsgsgs Sep 28 09:46:43 allowemptyinput Sep 28 09:46:43 ? Sep 28 09:46:59 now I have mouse and keyboard an qemuarm again Sep 28 09:47:08 ah :) Sep 28 09:47:22 ~lart xorgdevelopers for not makeing alloweymptyinput commandline option Sep 28 09:47:22 * ibot shoves a crumpet down xorgdevelopers's throat, happy now?! Huh? Want some JAM with that? for not makeing alloweymptyinput commandline option Sep 28 09:47:33 btw what do you use qemu for? Sep 28 09:47:40 or, what is your use case Sep 28 09:47:42 qemuarm image Sep 28 09:47:47 testing pyqt in the end Sep 28 09:47:52 ah Sep 28 09:47:59 pyqt must be fun :p Sep 28 09:48:09 programming or compiling? Sep 28 09:48:20 compiling Sep 28 09:48:26 hm no Sep 28 09:48:34 sip is doing its job Sep 28 09:49:05 so my son is finally awake Sep 28 09:49:08 till later Sep 28 09:49:15 are you building pyside bindings? Sep 28 09:49:26 no Sep 28 09:49:27 pyqt Sep 28 09:49:33 I dont trust pyside yet Sep 28 09:50:05 omg, it's nokia, you should trust them :) Sep 28 09:50:38 hm, the fix for NOTE don't work :) Sep 28 09:51:15 log http://pastebin.com/ZyJNEHc9 Sep 28 09:54:57 arrggsgsgsss Sep 28 09:55:10 mmmh... gcc 4.5 miscompiles eglibc's ld-2.12.1.so and other libs :( is it really stable enough to be the default compiler? Sep 28 09:56:20 woglinde: oh wait, it's my mistake maybe :) Sep 28 09:57:09 woglinde: just run menuconfig without the devshell, so there's even CONFIG_X86=y in my .config :) Sep 28 09:59:12 ~lart filebin Sep 28 09:59:12 * ibot stamps filebin on the forehead with the official Troll marker Sep 28 09:59:25 ARCH=foo Sep 28 09:59:34 otherwise the hostsone is taken Sep 28 09:59:35 hm, ibot is creative nowadays Sep 28 10:00:31 ynezz try this http://pastebin.com/1iZYc3gg Sep 28 10:01:52 jo zecke Sep 28 10:02:44 working gpe-terminal Sep 28 10:02:50 will try pyqt later Sep 28 10:05:39 woglinde: what did you sent me? :) http://pastebin.com/XRweNzgD Sep 28 10:06:14 but no problem, I'll do it manualy Sep 28 10:20:21 woglinde: one more http://pastebin.com/vkUTE8yf Sep 28 10:41:03 ynezz this one you should be able to fix your self Sep 28 10:41:41 cp -fpPR ${D}/kernel/.config ${SYSROOT_DESTDIR}${STAGING_KERNEL_DIR} shouldnt be there anymore Sep 28 10:47:06 it's saying missing .config, isn't it? Sep 28 10:48:09 sure Sep 28 10:48:18 because we dont copy it there anymore Sep 28 10:54:20 finally, Qt finished to build, pushing my changes Sep 28 10:57:10 *g* Sep 28 10:58:25 hrms Sep 28 10:58:29 new 4.6.3 Sep 28 10:59:24 now I need to fix gdb... then qt4-tools-sdk Sep 28 11:01:35 whats wrong with gdb? Sep 28 11:05:13 args Sep 28 11:05:21 damn now the iconv problem again Sep 28 11:06:05 we need to build it with expat support Sep 28 11:06:23 for remote gdb, the number of registers, their width, etc. is transfered in XML Sep 28 11:07:55 ah Sep 28 11:09:07 03Holger Hans Peter Freyther  07master * rf70e3e09fb 10openembedded.git/recipes/qt4/qt4.inc: qt4: Package the declarative debug packages into qt4-examples-dbg Sep 28 11:09:21 03Holger Hans Peter Freyther  07master * r5e42a46739 10openembedded.git/recipes/qt4/qt4-tools-native_4.7.0.bb: Sep 28 11:09:21 qt4-tools-native_4.7.0.bb: Use the mkspec files from qt-4.7.0 Sep 28 11:09:21 This needs to be streamlined that every Qt 4.x mkspec file Sep 28 11:09:21 is found in this directory. At least this is a start into Sep 28 11:09:22 the right direction. Sep 28 11:09:22 03Holger Hans Peter Freyther  07master * r249013a9b1 10openembedded.git/recipes/qt4/qt-4.7.0/g++.conf: Sep 28 11:09:22 qt-4.7.0: Use qmake feature to link with -rpath-link. Sep 28 11:09:23 This is instructing qmake to use the QMAKE_LFLAGS_RPATH for the Sep 28 11:09:23 lib directory in the build tree. It should be safe to use even Sep 28 11:09:24 for software using the resulting qmake. Sep 28 11:09:25 03Holger Hans Peter Freyther  07master * r66f479943a 10openembedded.git/recipes/qt4/qt-4.7.0.inc: Sep 28 11:09:26 qt-4.7.0.inc: Unpack the qte.sh to fix the qt4-embedded do_install Sep 28 11:09:26 Fix the qt4-embedded_4.7.0 do_install by unpacking the qte.sh like Sep 28 11:09:27 it is done in the other recipes. Sep 28 11:09:27 03Holger Hans Peter Freyther  07master * rf28f870243 10openembedded.git/recipes/qt4/ (10 files in 2 dirs): (log message trimmed) Sep 28 11:09:28 qt-4.7.0: Remove patches that cover up errors. Sep 28 11:09:28 * These patches lack a proper commit message, e.g. no explanation Sep 28 11:09:29 why they need to be applied Sep 28 11:09:40 woglinde: fixed, works, thanks! Sep 28 11:09:50 ynezz you are welcome Sep 28 11:09:52 *g* Sep 28 11:10:17 03Holger Hans Peter Freyther  07master * rcc2f6f81f9 10openembedded.git/recipes/qt4/qt-4.7.0/ (g++.conf linux.conf): Sep 28 11:10:17 qt-4.7.0: Update the spec files for Qt 4.7 Sep 28 11:10:17 The biggest change is the QMAKE_RPATH to QMAKE_LFLAGS_RPATH Sep 28 11:10:17 rename that broke linking for us. Sep 28 11:10:17 03Holger Hans Peter Freyther  07master * r722e6c8294 10openembedded.git/recipes/qt4/qt-4.7.0.inc: qt-4.7.0.inc: Pick the linux.conf/g++.conf from the right directory Sep 28 11:10:18 03Holger Hans Peter Freyther  07master * ra8436e70ed 10openembedded.git/: Merge branch 'holger/qt4-4.7' Sep 28 11:10:18 03Holger Hans Peter Freyther  07master * r41dedfdc3f 10openembedded.git/recipes/qt4/qt-4.7.0/ (g++.conf linux.conf): Sep 28 11:10:19 qt-4.7.0: Fork the mkspec used in previous Qt versions Sep 28 11:10:19 (1 lines omitted) Sep 28 11:10:44 zecke why is trolltech/nokia to dump to make a check if iconv is in a seperarte library Sep 28 11:10:49 dumb Sep 28 11:11:45 woglinde: hmm. just build with -no-iconv. :) Sep 28 11:13:16 * woglinde kicks zecke in the ass Sep 28 11:14:01 hm you removed the disable demos patches Sep 28 11:14:31 now the qt build times is 2 hours more again Sep 28 11:16:19 :) Sep 28 11:16:50 it is not two hours though Sep 28 11:17:02 okay 1 hour 55 minutes Sep 28 11:17:06 useless Sep 28 11:17:13 and another hour for examples Sep 28 11:17:16 cpuburn Sep 28 11:17:28 well, the whole build takes like 16 hours on my laptop Sep 28 11:17:45 you dont thing this is wrong? Sep 28 11:17:45 and there is little difference between 14 and 16 hours. Sep 28 11:17:59 I think it is wrong that Qt is so big Sep 28 11:18:14 but then you should compile with -no-webkit Sep 28 11:18:32 and build QtWebKit 2.0 standalone Sep 28 11:19:48 wtf is qtwebkit now dying with internal segfault Sep 28 11:22:00 hm Sep 28 11:22:09 must been a rare issuse Sep 28 11:22:12 now it works Sep 28 11:22:43 zecke is the va gcc change reflected in 4.7? Sep 28 11:23:38 woglinde: va == variable arguments? Sep 28 11:23:42 yes Sep 28 11:24:07 woglinde: it is an ABI change, so it should be transparent to Qt? Sep 28 11:24:27 ../../../../include/QtCore/../../src/corelib/tools/qstring.h:175:17: note: the mangling of 'va_list' has changed in GCC 4.4 Sep 28 11:29:28 oooh... __aeabi_uidiv is broken in gcc 4.5.2; it thinks that 5/3 = 3 Sep 28 11:32:55 ensc/w ieehks why they have the tests for? Sep 28 11:35:02 I am not sure why this is happening; here, everything coredumps during ld loading (there is calculated a hash which points outside the array due to bad % result) Sep 28 11:35:23 OE is using gcc 4.5 as default compiler, right? Sep 28 11:35:47 so this might be specific to a local build option Sep 28 11:36:48 I don't think it's default, it depends on what are you using Sep 28 11:37:12 it's using 4.4.3 here Sep 28 11:37:36 khem changed sane toolchain yesterday to 4.5 Sep 28 11:37:44 but angstroem might use 4.3 Sep 28 11:37:48 MACHINE="ts72xx" Sep 28 11:37:50 DISTRO="angstrom-2008.1" Sep 28 11:37:53 yep Sep 28 11:38:13 ah, did last rebase like day ago Sep 28 11:39:09 it might be due to the -Os Sep 28 11:39:36 non -Os when building gcc uses another branch for uidiv Sep 28 11:46:19 ensc|w: ouch Sep 28 11:46:29 woglinde: well, i think that is a -W flag for the compiler Sep 28 11:46:39 moin zecke Sep 28 11:47:00 fraxinas: hi, I am back in some time, did you manage to resolve the issue with khem? Sep 28 11:48:52 nah not really Sep 28 11:49:01 just started working on it again now Sep 28 11:49:43 hi fraxinas Sep 28 11:49:59 in principle... shouldn't the java line System.loadLibrary("aacskeys"); invoke a runtime import of that lib? Sep 28 11:50:03 hi linde Sep 28 11:50:10 i am going outo to by a desert. :) Sep 28 11:50:28 fraxinas: well, it should... but you try to link everything statically Sep 28 11:50:32 buy a desert? like the sahara? Sep 28 11:50:44 fraxinas: so all JNI symbols need to be there Sep 28 11:50:51 let me check the spelling. :) Sep 28 11:50:55 dessert? Sep 28 11:51:16 ah the jumminess that you stuff your tummy with? Sep 28 12:01:10 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45817 Sep 28 12:09:12 i figured screw it ill watch a movie Sep 28 13:31:11 03Holger Hans Peter Freyther  07master * rdee46c3eb9 10openembedded.git/recipes/qt4/ (qt4-tools-sdk.inc qt4-tools-sdk_4.7.0.bb): qt4-tools-sdk: Add a Qt 4.7.0 version, create a .inc file for reuse. Sep 28 13:44:14 03Martin Jansa  07master * r4ae02b5b27 10openembedded.git/recipes/shr/ (3 files in 2 dirs): Sep 28 13:44:14 initscripts-shr: drop ttySACx workarround and better ROOT_HOME detection Sep 28 13:44:14 * ttySACx are fixed in kernels 2.6.32.23 and 2.6.34.7 and newer, fixed Sep 28 13:44:14 by kernel patch from Radek Polak Sep 28 13:44:14 * ROOT_HOME problem found and fixed by Joachim Ott Sep 28 13:44:14 Signed-off-by: Martin Jansa Sep 28 14:02:44 ensc|w: I'm using gcc-4.5 on armv[4567] and all problems I've found were fixed by khem later..also the --enable-target-optspace problem Sep 28 14:03:02 JaMa: tried 4.5 with linaro patches? Sep 28 14:04:11 hrw: partially yes (the version from khem) Sep 28 14:04:29 hrw: now running rebuild from scratch on arvm[457] Sep 28 14:04:56 hrw: only problem sofar was tree-vectorize in mplayer Sep 28 14:05:41 JaMa: where is the --enable-target-optspace problem fixed? Sep 28 14:05:53 JaMa: can you extend that mplayer problem? Sep 28 14:06:21 I use recent git and see it Sep 28 14:06:40 hrw: check log from yesterday or day before Sep 28 14:07:16 ensc|w: it's quite long ago I don't remember if it was fixed in gcc or some workarround in OE.. Sep 28 14:07:49 kergoth_, ping Sep 28 14:07:57 hi kgilmer Sep 28 14:08:15 hrw: "./oe-2010-09-26:20:10 < JaMa> and I was able to build it with -fno-tree-vectorize instead -ftree-vectorize" , and arround Sep 28 14:09:01 reading Sep 28 14:09:31 jo hrw and jama Sep 28 14:10:01 and hi kgilmer Sep 28 14:10:15 JaMa: I use the gcc from e33feda488545dbec4a71c590fee8f6f3c2a7a0d and see the problem... Sep 28 14:11:02 hrw: a day later khem advised to use -mvectorize-with-neon-quad Sep 28 14:11:26 * ensc|w hates it when external git servers are unavailable and bitbake stops in the "Handling Bitbake files" phase Sep 28 14:12:20 nice read Sep 28 14:12:55 * hrw waits for dual A9 machine to run some tests Sep 28 14:13:22 hrw, will you and khem coordinate the linaro tool work? Sep 28 14:13:51 Crofton|work: Khem will maintain 4.5/linaro in OE Sep 28 14:13:59 ok Sep 28 14:14:10 this will be ok for you? Sep 28 14:14:27 yes, we had a discussion about it Sep 28 14:14:53 very good Sep 28 14:15:08 I am looking forward to seeing how much improvement they make for A8 Sep 28 14:15:22 depends on tests Sep 28 14:15:39 sure Sep 28 14:15:42 but we welcome any ideas/suggestions for tests Sep 28 14:15:53 my main concern is real apps Sep 28 14:16:03 as some of our devs work on automated tests Sep 28 14:16:06 particularly using NEON to vectorize Sep 28 14:16:11 ensc|w: did you find any solution for iwmmxt? Sep 28 14:16:31 Crofton|work: not all A8 have neon ;( Sep 28 14:16:42 well, the ones I care about do :) Sep 28 14:17:05 I know ;) Sep 28 14:17:40 anyway I am waiting for A9 Sep 28 14:17:49 zecke: no; only to avoid it :( Sep 28 14:18:20 howdy woglinde Sep 28 14:18:36 hi there hrw Sep 28 14:18:37 I will have to recheck it against recent kernels; perhaps it has been fixed there (last seen on some 2.6.34) Sep 28 14:19:04 I don't get to wait Sep 28 14:19:28 Crofton|work: ? Sep 28 14:19:47 http://pastebin.com/J18ZEsBe Sep 28 14:20:03 hrw, I have projects that need to work now, so waiting for A9 is not an option Sep 28 14:20:17 also, anyone have an idea how to find where my error really is? Sep 28 14:25:56 crfotn edit local.conf Sep 28 14:26:23 and try run again Sep 28 14:26:43 seems some bug in kergoth new bitbake parser code Sep 28 14:30:21 heh, yeah that is what I suspect Sep 28 14:31:22 ensc|w: no feedback from the devs? it used to work well Sep 28 14:35:42 "new bitbake parser code"? Sep 28 14:35:57 wb zecke :) Sep 28 14:36:04 playya_: pong Sep 28 14:36:08 hi Sep 28 14:36:41 the last parameter for ts_read: is it n finger in one sample? Sep 28 14:36:49 e.g. for multitouch Sep 28 14:36:50 zecke: about the lost iwmmxt state on ELF loading? There was no response, only other bugreports Sep 28 14:37:21 i try to update the palmpre module Sep 28 14:37:40 playya_: i understood it as number of requested events Sep 28 14:38:12 yes. that's what's the code is doing Sep 28 14:38:28 that's why I'm asking Sep 28 14:39:19 gm kergoth Sep 28 14:40:01 kergoth there seems to be an error with cache parser when you edit a .bbclass file Sep 28 14:47:14 hm newer sip dont uses qmake anymore Sep 28 14:48:57 Heinervdm, btw. there's a typo in ts.conf for palmpre in OE Sep 28 14:53:31 woglinde: not sure what you're referring to. Crofton|work's issue doesn't have anything to do with the cache stuff, his issue is with the conversion to logging -- parse failures in the configuration metadata aren't being reported, because they're seen before the UI is spawned Sep 28 14:53:51 ok :) Sep 28 14:54:09 suggestions welcome, my crystal ball is not working Sep 28 14:54:10 hm Sep 28 14:54:25 I had this error after editing a .bbclass Sep 28 14:54:48 Crofton|work: working on it now Sep 28 14:54:55 thanks Sep 28 14:54:58 woglinde: was it a bbclass that gets INHERITed? Sep 28 14:55:01 I will shower then :) Sep 28 14:55:03 INHERIT classes end up in the config metadata Sep 28 14:55:32 was kernel.bbclass Sep 28 14:55:49 but strangly after checkout and rediting wit worked Sep 28 14:55:56 well, dunno what to tell you. that error itself is meaningless, whats telling is the fact that you didn't actually see the cause of the parsing failure Sep 28 14:56:07 hm yes Sep 28 14:56:44 maybee I had an error there Sep 28 14:57:44 its certainly possible there's another bug, but i've never seen it, so can't do anything about it at this time without more information Sep 28 14:57:54 sure Sep 28 15:16:06 son of a bitch Sep 28 15:17:07 the parse error is an exception, which means the cooker isn't necessarily fully functional after it, which means we can't continue to spawn the ui, which means we can't flush the queued events to the ui -- will have to set up a temporary console handler to flush the queued logrecords to the console before exiting Sep 28 15:18:36 o.O Sep 28 15:20:02 ensc|w: in case of Qt, only some routines use mmx (e.g. drawing routines, color blending), so in general it should work as the kernel saves/restores the registers Sep 28 15:21:45 zecke: the kernel does not seem to restore the registers... Sep 28 15:22:24 ~lart sip Sep 28 15:22:25 * ibot plops sip into a giant vat of herring Sep 28 15:27:08 Crofton|work: okay, i have a fix Sep 28 15:27:43 Crofton|work: at least, i seem to :) Sep 28 15:28:40 03Chris Larson  07master * r73488aeb31 10bitbake.git/lib/bb/ (cooker.py event.py server/none.py server/xmlrpc.py): (log message trimmed) Sep 28 15:28:40 Queue up events before the UI is spawned Sep 28 15:28:40 - Queue up any events fired to the UI before the UI exists Sep 28 15:28:40 - At exit, check if UIs exist, and if not, flush the queue of LogRecords to Sep 28 15:28:40 the console directly. Sep 28 15:28:40 - When establishing a connection from the UI to the server, flush the queue of Sep 28 15:28:41 events to the queue in the server connection, so the UI will receive them Sep 28 15:28:55 ensc|w: ensc|w arch/arm/kernel/iwmmxt.S, was it build? was it called? Sep 28 15:29:02 Crofton|work: give that a shot, if you would Sep 28 15:29:23 thanks kergoth Sep 28 15:30:16 this would be somewhat less annoying if the ui was a separate script that spawned the server rather than a single binary that sucks it all in Sep 28 15:30:40 *g* Sep 28 15:30:41 dbus Sep 28 15:30:45 comes to my mind Sep 28 15:31:55 zecke: yes; stuff works in 99.999% of cases. The remaining 0.001% are the problem... Sep 28 15:33:22 ensc|w: did you write a test case with two processes storing constants in the register? Sep 28 15:34:21 zecke: yes; http://thread.gmane.org/gmane.linux.ports.arm.general/11795 Sep 28 15:35:00 I filled values 0-15 in wrX and after some time, they were 0 Sep 28 15:37:33 ensc|w: ah the last email Sep 28 15:39:15 zecke tell me if you get bored and need some excitement :) Sep 28 15:39:48 ensc|w: that udiv issue will only happen at -O0 Sep 28 15:40:00 ensc|w: or do u see it at Os too Sep 28 15:42:00 hi ph5 Sep 28 15:42:08 hi woglinde Sep 28 15:42:25 khem: -Os (--> -D__OPTIMIZE_FOR_SIZE__) when building libgcc (enabled by --target-optspace) Sep 28 15:42:49 ensc|w: hmmm Sep 28 15:42:51 khem: there is an alternative branch in libfunc1.asm for uidiv Sep 28 15:42:57 I see that Sep 28 15:43:12 and where does issue show up Sep 28 15:43:23 is it common enough problem Sep 28 15:43:50 at least in ld.so when it calls do_lookup_x() where foo[hash % elem_cnt] is accessed Sep 28 15:44:01 hmm Sep 28 15:44:22 but this affects every non-trivial division Sep 28 15:50:18 ensc|w: I wonder why it does not show up with gcc 4.4 Sep 28 15:50:26 code is same Sep 28 15:50:45 bah, typo Sep 28 15:51:16 kergoth where? Sep 28 15:53:06 kergoth hm there is another small glitch Sep 28 15:53:25 tasks summary is shown twice Sep 28 15:53:32 03Chris Larson  07master * rbed8e09971 10bitbake.git/lib/bb/ (cooker.py event.py): Sep 28 15:53:32 Fix syntax issue and don't include 'd' in ui_queue Sep 28 15:53:32 Signed-off-by: Chris Larson Sep 28 15:53:38 hmm Sep 28 15:55:24 khem: I see differences between 4.4.4 and 4.5 code (e.g. subcs vs subne) Sep 28 15:57:47 *sigh* package-index.bb fails Sep 28 15:58:16 not such good day today Sep 28 15:58:32 hm hm Sep 28 15:58:39 only fails with -f -b Sep 28 16:00:06 ensc|w: which one has subne ? Sep 28 16:02:07 khem: 4.4.4 has subcs, 4.5.2 has subne. https://www.cvg.de/people/ensc/a-4.4.4.s https://www.cvg.de/people/ensc/a-4.5.2.s Sep 28 16:10:48 hmm did you try with the patches I posted for linaro Sep 28 16:11:04 ensc|w: which that one I am getting cmpcs on gcc 4.5 Sep 28 16:11:13 ensc|w: may be worth trying Sep 28 16:14:27 I used the non-linaro gcc Sep 28 16:14:40 gcc on cmdline gives 'cs' too... strange... Sep 28 16:16:57 gm Sep 28 16:27:21 zecke: hi, do you have a sec for qt-4.7.0 question? Sep 28 16:27:43 JaMa: sure Sep 28 16:28:21 zecke: newer libx11-native I'm going to push later doesn't support --without-xcb, so they are built with xcb support Sep 28 16:28:32 zecke: then it fails in xlib test in qt4-tools-native Sep 28 16:28:55 zecke: this fixes it http://paste.pocoo.org/show/268589/, but it should be applied only for newer libx11 (1.3.6 and 1.4_RC+) Sep 28 16:29:32 JaMa: no, this looks wrong Sep 28 16:29:39 zecke: the error without: http://paste.pocoo.org/show/268590/ Sep 28 16:29:42 JaMa: the usage of xcb should be transparent to Qt Sep 28 16:30:05 but it adds -lX11 without needed deps Sep 28 16:30:33 which seems like same problem as "-lXau -lXdmcp" in that patch, isn't it? Sep 28 16:30:45 JaMa: no, gcc does not know how to find the libxcb.so.1. Sep 28 16:31:06 it needs -rpath-link, or khem's binutils fix. Sep 28 16:31:30 this is rebuild from scratch on today's HEAD Sep 28 16:31:53 the rule of thumb is: link to the libs you use, Qt does not use xcb_ symbols. so there is no need for it to link Sep 28 16:32:43 khem: building the lib1func.asm from 4.5.2 creates 'ne' both when building with gcc 4.4.4 and 4.5 (preprocessor translates the subhs to subne there) Sep 28 16:32:49 and what about -no-undefined? Sep 28 16:32:49 isnt that a .pc error Sep 28 16:33:40 JaMa: you will (strace) need to figure out where gcc is searching for libxcb.so.1, and maybe pass -Wl,rpath-link as part of the QMAKE_LIBDIR_X11. Sep 28 16:34:21 JaMa: in general we seem to have an issue with the toolchain layout (see my question on the ML), it is not searching for libs in the right folders Sep 28 16:34:26 hm pyqt needs some extra lib Sep 28 16:35:03 zecke: so whole fix.xlib-test.libs.patch should be removed if the core issue is fixed, right? Sep 28 16:36:17 (layout) well, this is a native recipe so nothing has changed there.. so I am already partially wrong. Sep 28 16:36:30 JaMa: yes, if we manage to pass -Wl,rpath-link down to the tests they should work Sep 28 16:41:00 woglinde: email on kernel NOTE issue send (as promised earlier today) Sep 28 16:41:39 JaMa: the other option is to pass -embedded on configure, and drop the x11 dependencies Sep 28 16:43:32 NOTE: package cairo-1.10.0-r0: task do_configure: Failed - checking whether cairo's image surface backend feature could be enabled... no (requires pixman-1 >= 0.18.4 http://cairographics.org/releases/) - seems to be a new issue (pulled last night) Sep 28 16:44:04 tharvey: what configuration? Sep 28 16:44:13 tharvey: newer pixman is available but with D_P -1 Sep 28 16:45:09 just trying to bitbake cario for MACHINE=x86 or MACHINE=overo Sep 28 16:45:19 tharvey: which distribution? Sep 28 16:45:22 JaMa, can you explain more? Sep 28 16:45:48 didn't have this issue when I last pulled a couple of weeks ago Sep 28 16:45:50 tharvey: you can get pixman new enought if you set PREFERRED_VERSION for it Sep 28 16:46:45 JaMa, ah... ok let me try that Sep 28 16:47:28 khem: my fault :( Sep 28 16:47:44 I applied fix for http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43999 which breaks non-thumb Sep 28 16:50:12 zecke: whatever sollution you think is better, I didn't look too much to qt (clearly - as my patches were only covering real problem :)) Sep 28 16:54:35 JaMa: i wonder what changed in regard to qmake/toolchain for this to happen Sep 28 16:55:00 JaMa: for the X11... what would be the right variable to find ${VAR}/libxcb.so.1? Sep 28 16:55:22 JaMa: it appears that somehow the -L is not evaluated for libraries to 'indirectly' search for Sep 28 16:56:00 zecke: nothing changed in qmake/toolchain afaik Sep 28 16:56:12 those xau xdmc were added long ago Sep 28 16:56:34 and xcb is added only to newest libx11-native (not yet pushed) Sep 28 16:56:56 or rather only newest libx11-native is built against xcb Sep 28 16:57:08 kergoth_, not good Sep 28 16:57:22 that patch I sent before covers this issue the same as it worked for xau/xdmcp Sep 28 16:57:48 Crofton|work, ? Sep 28 16:58:12 still uselss message Sep 28 16:58:19 I wiped tmp and am starting over Sep 28 16:58:24 then I will pastebin Sep 28 16:58:27 JaMa: I will go for -embedded, if dbus, etc. compiles with it Sep 28 16:58:52 JaMa: well, qmake received a lot of work this round (4.6 to 4.7) Sep 28 16:59:36 zecke: I meant that yesterday I was able to build qt4-tools-native with all your qt4 patches fine.. Sep 28 17:00:02 zecke: so the only change here was newer libx11-native Sep 28 17:01:48 Crofton|work, hmm, well, no idea then, offhand. would need further investigation. could always try 1.8.18 and 1.10.0 and see if the behavior differs, and if so, bisect Sep 28 17:02:31 lets see what happens after a full reparse Sep 28 17:03:49 Ok. I've been working on some new recipes. Is there a good place to put up the recipes and patches needed for comment? Sep 28 17:04:01 pastebin maybe? Sep 28 17:04:07 And to see if I'm not duplicating work? Sep 28 17:04:35 mailing list, patchwork, .. Sep 28 17:06:50 http://pastebin.com/a20XdT65 Sep 28 17:07:35 mmarker: if you are looking for advise on specific issues, probably pastebin and ask here, if you just want to share and get them reviewed and eventually in the repository send to the mailing list Sep 28 17:11:07 kergoth trying the 1.8 branch now Sep 28 17:13:50 ok Sep 28 17:13:58 found the actuall error with 1.8 Sep 28 17:18:51 JaMa thx PREFERRED_VERSION_pixman = "0.19.4" solved the issue - I assume angstrom should probably set that to avoid breakage Sep 28 17:25:11 tharvey: angstrom is using newer pixman already Sep 28 17:25:57 tharvey: DEFAULT_PREFERENCE_angstrom = "2" in pixman_0.19.4 so it should be used Sep 28 17:27:49 ah... I see. I have my own distro that includes angstrom-2008-preferred-versions.inc Sep 28 17:28:40 is DEFAULT_PREFERENCE_angstrom the way things are moving as opposed to a single/set of preferences in one file? Sep 28 17:29:29 tharvey: not really, those are rather complementary ways Sep 28 17:30:42 tharvey: in general you don't need your own distro. Sep 28 17:31:10 I'll need to split them out by package then post, it seems Sep 28 17:31:41 I realize that, but we've headed down a path of creating our own private OE distro for some projects. I would not do it that way again, its more trouble than its worth :) Sep 28 17:32:20 :) Sep 28 17:32:21 amen Sep 28 17:44:57 JaMa: with -embedded we would lose the qdbusviewer, but I don't think it is a loss nowadays Sep 28 17:58:53 hmm, our ast statement cache has no expiration policy, they're kept along indefinately Sep 28 17:58:56 that seems bad Sep 28 18:00:17 kergoth: in the parser? yes. :} Sep 28 18:00:23 yeah Sep 28 18:00:34 kergoth: was acceptable for a short-ran bitbake process Sep 28 18:00:36 * kergoth nods Sep 28 18:00:50 kergoth: I have never found a good LRU cache in python, it is a shame one needs to write one... Sep 28 18:01:18 JaMa: do you have a fast build machine and could test something for me? Sep 28 18:01:59 hmm, maybe something lie http://code.activestate.com/recipes/498245-lru-cache-decorator/ Sep 28 18:02:01 * kergoth reads Sep 28 18:08:57 zecke: sort of, but definetely can test Sep 28 18:11:17 JaMa: http://paste.lisp.org/display/114993 Sep 28 18:14:50 zecke: what about that fix.xlib-test.libs.patch ? remove with it? Sep 28 18:16:31 JaMa: well, I will need to fix that for Qt X11... but my workstation is still in the mail (from Taiwan) Sep 28 18:17:14 I don't understand, I'm talking about patch in qt4-tools-native.. Sep 28 18:18:44 JaMa: okay, yes, remove it. Sep 28 18:19:07 zecke: building it now Sep 28 18:26:01 zecke: NOTE: package qt4-tools-native-4.7.0-r7.1: task do_configure: Succeeded, waiting for qt4-x11-free build Sep 28 18:28:29 ensc|w: I am glad that the issue got sorted :) Sep 28 18:29:00 JaMa: did you try a build from scratch with gcc 4.5 from my branch Sep 28 18:29:06 khem: yes; I do not know how I could forget this patch :( Sep 28 18:32:17 JaMa: btw. the issue with mplayer is fixed too Sep 28 18:32:26 I have a patch here I can commit it to the branch Sep 28 18:33:24 kergoth: bitbake held well with the SCM export change with bitabake fetchall world Sep 28 18:33:50 nice Sep 28 18:34:22 will see about putting it in master Sep 28 18:34:41 cool Sep 28 18:35:17 btw khem how is your bitbake world doing? we still have build server issues so running on reduced capacity and I've been asked not to start any world builds... Sep 28 18:36:00 amend.inc does not look into FILESPATH anymore but only in the directory of the .bb file? Sep 28 18:36:01 khem: building atm Sep 28 18:36:11 ensc: no, that's not true Sep 28 18:36:29 it looks into the filespath locations but *not* within override specific dirs Sep 28 18:36:36 since you can use overrides inside the .inc instead Sep 28 18:36:43 FILESPATH is FILESPATHBASE + FILESPATHPKG + OVERRIDES Sep 28 18:36:49 now amend.bbclass searches the first two, not the last Sep 28 18:36:54 eFfeM_school: its progressing.. on back burner I look back into the logs and fix few things Sep 28 18:36:57 zecke: http://tinderbox.openembedded.net/packages/793224/ Sep 28 18:37:04 JaMa: cool Sep 28 18:37:04 khem: armv4 built fine Sep 28 18:37:07 which has a huge effect on bitbake memory usage Sep 28 18:37:12 khem, nice Sep 28 18:37:26 khem: now armv7 is building, then I'll try to remove workarround from mplayer and rebuild mplayer Sep 28 18:37:41 khem btw have build beagleboard console-image from your branch, it builds fine, but haven't had time to test it Sep 28 18:37:45 JaMa: OK, shoudl I commit the new patch to the branch Sep 28 18:37:52 I can sync the branch with master Sep 28 18:38:09 then you wont need any workarounds Sep 28 18:38:14 tonight i have to supervise a lab and tomorrow i have a meeting so guess it'll be thursday before i can test it (assuming you are still interested in test results by then) Sep 28 18:38:17 khem: does it change also other stuff than mplayer? Sep 28 18:38:38 khem: I'll build armv5 over night so I can test it with this patch too if you push it soon Sep 28 18:38:59 JaMa: ok, armv5t is not interesting for this change Sep 28 18:39:08 JaMa: its cortex-a8+ Sep 28 18:39:16 which will need this patch Sep 28 18:39:52 khem: armv7a is almost done here, is it possible to rebuild from scratch just one arch from tmpdir? Sep 28 18:40:16 khem: that way I would be able to test it without throwing whole armv4t build Sep 28 18:40:29 ugh Sep 28 18:40:36 I have one invasive way Sep 28 18:41:12 kergoth: this means, I have now to modify FILESPATHBASE instead of FILESPATH to modify search path of amend.inc? Sep 28 18:41:42 JaMa: hmm I think no Sep 28 18:42:05 ensc: you should always have been doing that. Sep 28 18:42:09 ensc: but yes Sep 28 18:42:58 JaMa: let me give you my script Sep 28 18:44:06 JaMa: http://pastebin.com/dHyCnQap Sep 28 18:44:43 JaMa: you need to adapt it a bit to your env Sep 28 18:46:46 khem: sure I'll check it later for armv7a-only rebuild Sep 28 18:47:27 JaMa: cool Sep 28 18:47:40 JaMa: are you cherry picking my fixes or using the branch Sep 28 18:48:04 cherry-picking Sep 28 18:48:15 JaMa: ok Sep 28 19:13:38 Hm, si there any way how to specify the name of the fetched file? To handle this crap for example SRC_URI = "http://www.ralinktech.com/download.php?t=U0wyRnpjMlYwY3k4eU1ERXdMekF6THpNeEwyUnZkMjVzYjJGa01UWXpPRGs1T0Rnek5pNTZhWEE5UFQxU1ZESTROekJmUm1seWJYZ" Sep 28 19:13:45 :p Sep 28 19:13:54 yikes Sep 28 19:15:06 it's using wget, so idealy something like the -O option Sep 28 19:15:48 maybe I'll just mirror it somewhere :) Sep 28 19:19:54 yeah, that should really be supported if its not already. check bitbake/lib/bb/fetch/wget.py :) Sep 28 19:20:04 something like a ';filename=/' modifier for SRC_URI would be nice... Sep 28 19:21:10 hmm, yeah, doesn't look like that exists yet -- see localpath() in bb.fetch.wget.Wget Sep 28 19:21:22 bb.fetch really bugs me Sep 28 19:25:22 JaMa: pushed the change needed for mplayer here Sep 28 19:25:24 http://git.openembedded.org/cgit.cgi/openembedded/commit/?h=khem/import-linaro-gcc-4.5&id=fa11285ca5b6fcdd9f91ac548ae275ea076201bd Sep 28 19:31:31 thx, trying mplayer with it Sep 28 19:32:39 khem: rebuilding gcc-cross should be enough, right? Sep 28 19:42:00 zecke: heh, problem with a smarter statement cache is what happens after BuildStarted -- subsequent cached accesses occur in child processes only, when the task processes reparse the recipes.. they'll never affect the access times/counts in the cache at that point :) Sep 28 20:12:45 JaMa: yes Sep 28 20:13:11 kergoth: I want to pass an env var QUILTRC to every invocation of quilt in OE Sep 28 20:13:22 kergoth: where to shove it in Sep 28 20:13:42 ah lib/oe/patch.py Sep 28 20:13:50 looks like the place Sep 28 20:25:57 kergoth: something like this would be ok? http://pastebin.com/VpfcQ3gL Sep 28 20:27:12 and then SRC_URI = "http://ynezz.ibawizard.net/openembedded/blah;filename=RT2870_Firmware_V22.zip" Sep 28 20:38:02 03Martin Jansa  07master * r6e7e356792 10openembedded.git/ (24 files in 5 dirs): Sep 28 20:38:02 xorg: add new recipes for versions from 2010-09-28 Sep 28 20:38:02 * remove 3 old recipes for versions not pinned by any preferred-xorg-versions Sep 28 20:38:02 * libx11 will be pushed soon Sep 28 20:38:02 Signed-off-by: Martin Jansa Sep 28 20:38:13 03Martin Jansa  07master * rbce1c1b96d 10openembedded.git/recipes/xcb/ (4 files): Sep 28 20:38:13 libpthread-stubs: add .inc file, INC_PR, native and nativesdk BBCLASSEXTEND Sep 28 20:38:13 * needed for libxcb-native Sep 28 20:38:13 Signed-off-by: Martin Jansa Sep 28 20:38:24 03Martin Jansa  07master * r83d7f52b5d 10openembedded.git/recipes/xcb/xcb-proto.inc: Sep 28 20:38:24 xcb-proto: add native and nativesdk BBCLASSEXTEND Sep 28 20:38:24 * needed for libxcb-native Sep 28 20:38:24 Signed-off-by: Martin Jansa Sep 28 20:38:36 03Martin Jansa  07master * r2fc04974d4 10openembedded.git/recipes/xcb/libxcb.inc: Sep 28 20:38:36 libxcb: add native and nativesdk BBCLASSEXTEND Sep 28 20:38:36 * needed for libx11-native-1.3.99.902 and newer where --witout-xcb is no longer available Sep 28 20:38:36 Signed-off-by: Martin Jansa Sep 28 20:45:04 03Khem Raj  07master * r02634fb85a 10openembedded.git/recipes/quilt/ (3 files in 2 dirs): Sep 28 20:45:04 quilt-native: Use the quilt.quiltrc from sysroot instead of /etc Sep 28 20:45:04 This fix is only applied on build systems using linux Sep 28 20:45:04 Signed-off-by: Khem Raj Sep 28 20:46:42 kergoth: tested, works, so if it's it enough to give you patch here, there's a patch http://pastebin.com/VVtUBLUV Sep 28 20:49:24 looks like data.expand() is common, and the basename of the url and of ud.path should be the same, so i'd expect you could move the data.expand() out of the if block -- just ud.localfile = data.expand(ud.basename, d) Sep 28 20:49:38 ok Sep 28 20:49:47 also, you should be using ud.localfile as the output-document Sep 28 20:49:50 not ud.basename Sep 28 20:49:58 other than that, looks reasonable Sep 28 20:53:21 kergoth: http://pastebin.com/8SvywSjW Sep 28 20:54:06 that looks better, indeed Sep 28 20:54:16 of course you wouldn't be able to use it until OE bumps to 1.12 or whatever is coming after 1.10 Sep 28 20:54:20 since itd go into master only, most likely Sep 28 20:54:30 and OE hasn't bumped to 1.10.0 has its minimum yet.. Sep 28 20:54:45 of course, your local only recipes will work fine, just can't push any using the feature Sep 28 20:55:03 I hope it's the only one :) Sep 28 20:55:23 thanks Sep 28 20:56:15 looks like the encoding got all out of whack in putting it in pastebin Sep 28 20:56:21 might be better off emailing to bitbake-dev Sep 28 20:56:24 hmm Sep 28 20:56:52 ok,no problem Sep 28 20:57:26 k, will throw it in whenever i get it Sep 28 20:58:55 sent Sep 28 21:05:40 khem: mplayer built ok (without any workaround in CFLAGS) thanks! Sep 28 21:07:20 mwester: hi Sep 28 21:15:15 hi , is Angstrom having WPA2 support ? Sep 28 21:15:37 wpa_supplicant works there just fine, so yes Sep 28 21:35:37 hi jin^eld Sep 28 21:35:40 re Sep 28 21:50:29 hi rkirti Sep 28 21:50:43 woglinde: hello Sep 28 21:50:58 woglinde: how are things ? Sep 28 21:51:17 sometimes difficult Sep 28 21:52:13 hmm.. Sep 28 21:52:38 * woglinde fights with pyqt Sep 28 21:52:52 sorry, /me is very slow today because its 3.30 am here and I am struggling with some qt setup :( Sep 28 21:53:43 hm Sep 28 21:53:48 whats the matter? Sep 28 21:57:01 woglinde: nothing - general qt/kde setup. just that I am not a kde person. setting up stuff for a hackfest starting tomorrow in univ. Sep 28 21:57:48 hey rkirti Sep 28 21:58:11 rkirti: welcome to the club, i would have to fight with this task too :) Sep 28 21:58:22 florian: hello :) Sep 28 21:58:57 * woglinde fights with .pro file now Sep 28 21:59:07 SUBDIRS = foo/moo Sep 28 21:59:25 foo/moo.file = my.pro seems not to work Sep 28 22:02:19 hi, all Sep 28 22:02:24 hi Sep 28 22:03:46 could anybody comment on this: http://pastebin.com/zyuEeV9M? Sep 28 22:03:48 thx Sep 28 22:05:11 woglinde: TEMPLATE = subdirs; SUBDIRS = foo moo; foo.file = my.pro ? Sep 28 22:05:22 ynezz no Sep 28 22:05:32 with path Sep 28 22:05:57 foo/moo is path with .pro file which is othernamed than moo Sep 28 22:06:10 push {r4, r5, r6, r7, r8, lr} Sep 28 22:06:12 hi Sep 28 22:06:27 that is from what kind of assembly? Sep 28 22:06:32 it doesn't seem plain arm Sep 28 22:06:37 SMID? Sep 28 22:08:41 woglinde: tried? SUBDIRS = foo/moo/my.pro Sep 28 22:10:15 woglinde: anyway, there's one great page about qmake, somewhere on qtcentre.org's wiki, can't find it now Sep 28 22:10:18 nite Sep 28 22:11:38 03Khem Raj  07master * ree4a70a8d3 10openembedded.git/recipes/quilt/ (quilt-0.48/use-cp-for-patch-backup.patch quilt.inc): Sep 28 22:11:38 quilt-0.48: Use cp instead of mv for backing up patches. Sep 28 22:11:38 * mv would not preserve symlinks but cp -a would. Sep 28 22:11:38 Signed-off-by: Khem Raj Sep 28 22:12:19 khem: -a is a gnu-ism -- can do it more portably than that Sep 28 22:14:45 anybody knows what is going on? Sep 28 22:16:59 IgorK: there is no recipe locale-base-* Sep 28 22:17:34 IgorK: it's created as part of *libc build Sep 28 22:18:28 cp -p ? Sep 28 22:19:55 JaMa: so how do I build it? because currently it is failing to build x11-gpe-image because of this.... Sep 28 22:21:31 IgorK: check what you have in GLIBC_GENERATE_LOCALES and rebuild *libc Sep 28 22:24:34 good nite Sep 28 22:24:44 Hi, a question about Xfbdev Sep 28 22:25:54 My keyboard doesn't work, so I tried specifying a -keybd option Sep 28 22:26:12 Specifically, -keybd evdev,,device=/dev/input/event1 Sep 28 22:27:02 It then goes on to say that Kbd option key (device) of value (/dev/input/event1) not assigned Sep 28 22:27:07 I have no idea why? Sep 28 22:27:09 Do you? Sep 28 22:27:10 :D Sep 28 22:28:04 this is with xserver-xorg-kdrive-1.4.2 Sep 28 22:30:38 Ugh Sep 28 22:30:47 It doesn't work with xkb enabled Sep 28 22:31:00 How am I supposed to have keyboard support if xkb is disabled? Sep 28 22:31:17 kergoth: bsd also has -a Sep 28 22:31:31 khem: osx doesn't. Sep 28 22:31:38 kergoth: hmmm Sep 28 22:32:02 er, hmm, nevermind, i could have sworn it didn't Sep 28 22:32:05 weird Sep 28 22:32:05 kergoth: do they all have a consistent way to copy symlinks Sep 28 22:32:33 kergoth: osx used bsd may be they have not stolen new stuff yet Sep 28 22:32:40 -a is just a convenience for a bunch of individual options Sep 28 22:33:03 yeah I used it because it preserved link Sep 28 22:33:09 both on freebsd and on lnx Sep 28 22:33:17 well, nevermind, seems to be a nonissue here Sep 28 22:33:24 can try on osx later Sep 28 22:33:27 at home Sep 28 22:33:31 and of course osx is far from working for a ton of other reasons anyway Sep 28 22:33:38 iirc OSX accepts shell aliases ;) Sep 28 22:34:09 that's not particularly useful. to translate arguments you'd have to use a shell function Sep 28 22:34:12 kergoth: right. Sep 28 22:34:27 I know someone was using OE on freebsd build machine Sep 28 22:34:34 nice Sep 28 22:34:40 I myself tried OSX and did not succeed Sep 28 22:34:59 so I keep freebsd ahead of osx in mind Sep 28 22:35:01 i'd really like to see it work on freebsd, osx, cygwin, and eventually msys, but that's a ways out :) Sep 28 22:35:11 ugh Sep 28 22:35:18 hehe Sep 28 22:35:20 osx would work Sep 28 22:35:24 with bunch of ports Sep 28 22:35:28 installed I guess Sep 28 22:35:41 * Jay7 will check OE on freebsd too :) Sep 28 22:35:47 firsly we need to stop programming for bash Sep 28 22:36:08 if possible Sep 28 22:36:16 i really need to find a project other than oe to code on, messingw ith the same project for work and personal time is getting a bit much Sep 28 22:36:36 well, months ago someone came out with bashisms in some recipes Sep 28 22:36:59 can't find the logs offhand Sep 28 22:37:32 kergoth: on the contrary I would like to work on OE to get paid for :) quite opposite Sep 28 22:37:43 ant_: did you solve that .so issue Sep 28 22:38:00 if .so exist in -dev it should be OK and QA will allow Sep 28 22:38:01 khem, yes but then I found a second one... Sep 28 22:38:10 this klibc recipe is like Rubik cube Sep 28 22:38:28 see...we package 'klcc' for use on target Sep 28 22:38:50 atm this has wrong prefix (the one for the cross-klcc) Sep 28 22:39:18 btw I'm not so sure one would native compile with klcc Sep 28 22:40:28 me neither Sep 28 22:40:32 so 1) I remove klcc before packaging 2) I split the prefix-patch Sep 28 22:40:45 klibc is not supposed to have a rich hosted env Sep 28 22:40:52 one would use glibc for that Sep 28 22:41:01 or uclibc to a lesser extent Sep 28 22:41:51 kergoth: do we still need that moblin stuff Sep 28 22:42:05 i hope not Sep 28 22:42:11 it broken Sep 28 22:42:47 and the omap stuff assumes to be the only armv7 arch Sep 28 22:43:24 I need to hunt who asks for glibc even when its overridden by eglibc Sep 28 22:43:32 khem: btw btw, there is a staging-patch removing the /usr prefix , i.e. we have /lib/klibc. I think this is ok for initranfs Sep 28 22:44:02 yeah /lib should be ok isnt it ? Sep 28 22:46:22 hm... I'm now thinking that we have a huge klibc-dev Sep 28 22:46:50 ant_: not an issue Sep 28 22:47:04 ideally klibc-dev should not be required to be deployed Sep 28 22:47:17 well, those are the kernel headers Sep 28 22:47:56 do u put those in image Sep 28 22:48:16 no, just toolchain Sep 28 22:50:41 FILES_klibc-dev: /lib/klibc.so /lib/klibc/lib/* /lib/klibc/include/* /bin/klcc Sep 28 22:51:00 FILES_klibc: /lib/klibc-*.so Sep 28 22:51:13 ^^ from the staging-package Sep 28 23:20:34 Can someone tell me how the hell I can get keyboard working under Xfbdev? Sep 28 23:20:41 -keybd doesn't help Sep 28 23:53:11 'nite Sep 28 23:54:36 03Andrea Adami  07org.openembedded.dev * rda451d4d55 10openembedded.git/recipes/kexecboot/ (400 files in 86 dirs): Sep 28 23:54:36 linux-kexecboot: move recipes in /linux Sep 28 23:54:36 * some old version might hopefully be pruned **** ENDING LOGGING AT Wed Sep 29 02:59:57 2010