**** BEGIN LOGGING AT Wed Mar 28 02:59:58 2012 Mar 28 04:35:45 is there a recommended test framework that integrates into the oe build chain nicely? Mar 28 04:37:29 thinking about trying to get http://code.google.com/p/kyua/wiki/ATF compiled. I've never had great luck trying to autoconf a cross compile. There doesnt seem to be a pkg-config in the oe toolchain I have Mar 28 05:54:33 <_tasslehoff_> I get "ImportError: No module named _md5" when compiling python-numpy in oe-classic dev. Was this openssl and python building in the wrong order? I tried cleaning all-things-python and rebuilding, but it doesn't help. Mar 28 06:01:39 yes rebuild python-native too Mar 28 06:01:49 after openssl Mar 28 06:12:01 <_tasslehoff_> JaMa: thanks (again) :) Mar 28 07:26:59 What is a "anonymous" function? How can I access these function-types? Where can I read more? Mar 28 07:28:25 I don't understand this concept Mar 28 07:40:56 How can I build more than one image with one bitbake-command call? I use the classic-oe ... Mar 28 07:43:04 bitbake image-foo image-bar Mar 28 07:43:19 ou - this is simple Mar 28 07:50:57 good morning Mar 28 08:23:17 mornin Mar 28 08:27:19 hi Jin^eLD, all Mar 28 08:31:06 mickeyl: ping Mar 28 09:19:43 hi sebaz Mar 28 09:23:05 oops wrong window :) Mar 28 09:25:32 what's the way to go if I want to modify the S variable of a package globalle in an inherited custom .bbclass? I figure something like bb.data.setVar(src, S, d) but I am not quite sure where exactly to do it, something via addtask? at what position would I add this task? Mar 28 09:31:06 1. analyzing network problem during update, architecture meeting (or should that go to point 3?) Mar 28 09:31:09 oops wrong window agai Mar 28 09:40:33 hmm actually the setVar thing does not work out for me, or will it only be visible within the task function that is doing it? Mar 28 10:13:24 <_tasslehoff_> what must I throw into my sdk-recipe to get lib-opengl32? Mar 28 10:40:17 Jin^eLD: you can do it within an anonymous python function, that should work Mar 28 10:43:32 bluelightning: thx will try, so far I tried in python blah_do_something() but that did not set it everywhere Mar 28 10:48:49 what is the heck with the anonymouse function? When starts this function? Mar 28 10:50:36 bluelightning: thanks, that did work! Mar 28 10:50:51 silviof: it gets run during parsing Mar 28 10:51:05 Jin^eLD: cool :) Mar 28 10:51:27 silviof: if you "git grep anonymous" you can find some examples Mar 28 10:52:11 bluelightning: thx Mar 28 10:53:58 bluelightning: is a function declared with python () { ... } a anonymous-function too (without the __anonymous-flag) ? Mar 28 10:54:21 silviof: I'm not entirely sure Mar 28 10:55:11 bluelightning: thx again :-) Mar 28 11:30:19 yes, it is Mar 28 11:30:46 iirc, the "__anonymous" decoration was mostly to allow you to declare anonymous shell functions without the syntax being really obscure Mar 28 11:31:26 for python, I think it's basically the same as if you didn't give a name at all Mar 28 11:31:50 pb_: ah ok - with anonymous i can write a shell script? Mar 28 11:32:12 like __anonymous () { #shell..syntax here } ? Mar 28 11:32:17 yes. __anonymous() { # shell here } Mar 28 11:32:21 exactly Mar 28 11:32:23 :-) Mar 28 11:32:49 pb_: thx - now i understand this mechanism Mar 28 11:33:16 that said, anonymous shell is not particularly useful and I don't think there are any examples of it in oe-core at the moment. Mar 28 11:49:43 <_tasslehoff_> I've added qt4-embedded-gles to my image, and to get a fitting sdk I added libqt-embeddedopengl4-dev, but when I try to use it I get a message about libopengl32 missing. Mar 28 13:21:43 <_tasslehoff_> I'm on a Dell laptop with a quad-core i7 (Q740). Any gut feelings on good choices for BB_NUMBER_THREADS and PARALLEL_MAKE? Mar 28 13:24:06 how much ram do you have? Mar 28 13:24:18 <_tasslehoff_> JaMa: 4GB Mar 28 13:24:27 * JaMa finds high thread numbers always too hungry for IO Mar 28 13:25:05 <_tasslehoff_> I have j4/4 now, and builds seem so slow Mar 28 13:25:12 now building in ramfs makes more cores usable, but before even on raid0 I was using less threads then others Mar 28 13:25:46 I had j4/2 Mar 28 13:25:47 <_tasslehoff_> looking for a new pc with ssd and more ram, but will have to make the best with what I have a while longer Mar 28 13:25:50 on quad core Mar 28 13:27:36 <_tasslehoff_> I'll try that and see if things get better. Mar 28 13:33:07 _tasslehoff_: I use "watch -n 0.5 ps xf" to get a good look of how many processes the build is running Mar 28 13:35:01 <_tasslehoff_> Zagor: thanks Mar 28 13:35:33 <_tasslehoff_> it's been compiling qt4-embedded-gles for 2 hours now, and seems to be running 4 make processes most of the time Mar 28 13:38:31 I would run as many threads as you have cores/hyperthreads and only lower if you run out of memory Mar 28 13:39:18 yes but 4 BB_NUMBER_THREADS can create 4 times more gcc processes then cores if all happens to do do_compile at the same time Mar 28 13:40:38 true. but in my experience that doesn't happen terribly often. in fact I'm a little puzzled why bitbake is so poor at using many cores. Mar 28 13:41:25 Usually it seems to be because the dependency graph forces a lot of things to be serialized. Mar 28 13:41:57 Very I/O intensive as well. Mar 28 13:42:50 mwester: seek intensive, yes. but that should only mean things don't finish fast. not that they don't start. Mar 28 13:43:47 * _tasslehoff_ will try -j4/2, but has to go now. thanks Mar 28 13:43:48 pb_: yes. I have a sneaking suspicion there is a bug hiding in the build scheduler/queue. but I have nothing more to go on than a feeling. Mar 28 13:44:05 Yeah, I have had that feeling sometimes as well. Mar 28 13:44:17 I've never been quite bothered enough to try to debug it though. Mar 28 13:44:39 but, but... it's *suboptimal*! ;-) Mar 28 13:45:11 I'm hoping to get some time to peek into it in a not too distant future Mar 28 13:45:52 with new footer patch it's nice to see that one BB thread is e.g. doing do_compile (more CPU hungry) and other do_package (more IO hungry) Mar 28 13:46:37 JaMa: what patch is that? Mar 28 13:47:36 hmm patchwork is showing wrong patch, but this one http://patchwork.openembedded.org/patch/24023/ Mar 28 13:48:19 http://patchwork.openembedded.org/patch/23475/ Mar 28 13:49:30 and also package/populate-sysroot etc are using just single thread so do_compile with -j8 and some single threaded IO hungry process is good combination Mar 28 13:51:26 my use case is simple: I have a powerful machine and just want bitbake to use all of it. I'd prefer brute force over "clever" scheduling. Mar 28 13:51:47 but I acknowledge mine is not the only use case Mar 28 13:52:32 yeah, that's more or less my situation too. My machine is not all that powerful but it is hard to get bitbake to use very much of it. Mar 28 14:12:23 if load can be considered as good metric then here it's using my machine well load average: 4,86, 4,94, 4,19 Mar 28 14:12:30 just image build running Mar 28 14:15:38 yeah, I think load is a fairly good metric Mar 28 14:16:01 iirc, I seldom see a load of much more than 10-12 on a 16-core machine, which is obviously not so good Mar 28 14:17:57 and your PARALLEL_MAKE/BB_NUMBER_THREADS combination? Mar 28 14:18:08 this is with -j2/2 on 8-core Mar 28 14:18:12 this is with -j8/2 on 8-core Mar 28 14:18:13 sorry Mar 28 14:18:33 16/16, I think Mar 28 14:45:04 hi, journald uses 100% CPU here with xfce-nm-image, I've setup mdbus(python version) for debugging and it seems that /org/freedesktop/systemd1 is there but in another hand I've that: Mar 28 14:45:27 Failed to get D-Bus connection: No connection to service manager. #when running systemctl Mar 28 14:50:06 when I run that another journald spawns: Mar 28 14:50:19 ./mdbus -s org.freedesktop.systemd1 /org/freedesktop/systemd1/unit/systemd_2djournald_2eservice org.freedesktop.systemd1.Unit.Stop "" Mar 28 15:24:15 where do i set the ARCH var ? should i put it in the machine config or in the special machine kernel recipe? Mar 28 15:35:52 VtS, I think that if your machine config include the correct inc files you don't need that Mar 28 15:36:20 so if you're not writing a machine config you very probably messed up your setup Mar 28 15:46:47 hmm, how would one use variables like PN in a bbclass when creating a postinst ipk hook? Mar 28 15:48:12 the hook itself is shell, so I'd somehow need t expand PN when the hook is applied/written into the package or how does it work? Mar 28 15:51:00 but I guess I see a possible solution in update-rcd.bbclass Mar 28 16:04:57 denisATeukrea i've created a machine file for the cubox Mar 28 16:05:06 and included the armv7a Mar 28 16:05:18 but the arch isn't set there Mar 28 16:06:11 or belongs this error to something other then the arch var because the compiler is searching in x86 dir Mar 28 16:06:20 sed: can't read /storage/raid6/Dev/cubox/cubox-oe/build/tmp-eglibc/work/cubox-oe-linux-gnueabi/linux-cubox-2.6.32.9-cubox+r0+gita02fe1e396e39102689c9faf2ec8579112f034ee-r0/git/arch/x86/include/asm/unistd_32.h: No such file or directory Mar 28 16:07:35 hmmm Mar 28 16:08:25 try bitbake -e Mar 28 16:08:33 and look at the expression commented Mar 28 16:08:42 like #ARCH = ${... Mar 28 16:08:50 and grep for that Mar 28 16:10:11 tune_arch host_arch package_arch arch = arm Mar 28 16:10:19 build_arch = x86_64 Mar 28 16:10:28 sdk_arch is also x86_64 Mar 28 16:11:42 all that is set in bitbake.conf Mar 28 16:12:00 but I wonder where ARCH is set Mar 28 16:12:28 since I cannot stop my compilation you will have to grep yourself Mar 28 16:12:56 with bitbake -e i get ARCH="arm" Mar 28 16:13:09 redirect the output Mar 28 16:13:23 do bitbake -e your_package > package.e Mar 28 16:13:26 vim package.e Mar 28 16:13:37 and look for the line before ARCH="arm" Mar 28 16:13:43 or just use grep -B Mar 28 16:13:46 ok Mar 28 16:16:15 HOST_VENDOR="-oe" Mar 28 16:16:15 # ARCH=${@map_kernel_arch(d.getVar('TARGET_ARCH', True), d)} Mar 28 16:16:15 export ARCH="arm" Mar 28 16:16:15 # PREFERRED_PROVIDER_gdk-pixbuf=gdk-pixbuf Mar 28 16:16:40 then grep for that line in the metadata: Mar 28 16:16:45 ARCH=${@map_kernel_arch(d.getVar('TARGET_ARCH', True), d)} Mar 28 16:16:55 or look for TARGET_ARCH Mar 28 16:17:04 let me look in the current machines Mar 28 16:18:18 TARGET_ARCH = "${TUNE_ARCH}" #bitbake.conf Mar 28 16:19:02 machine/include/arm/arch-arm.inc:TUNE_ARCH = "${@bb.utils.contains("TUNE_FEATURES", "bigendian", "armeb", "arm", d)}" Mar 28 16:19:22 so that must be included somehow Mar 28 16:20:22 this is my machine config https://git.teamcoders.org/lmc/meta-cubox/blobs/master/conf/machine/cubox.conf Mar 28 16:20:27 ok let me look Mar 28 16:21:05 hmmm Mar 28 16:21:18 first don't include arch-armv7a.inc directly Mar 28 16:21:27 use rather a higher level CPU include Mar 28 16:21:56 like cortex a8 Mar 28 16:22:18 then test if the TARGET_ARCH is still necessary Mar 28 16:22:31 cubox is a marvell 510 Mar 28 16:23:55 i've tried the tune_armv7 but there i get a build error with gcc-initial Mar 28 16:24:00 ok Mar 28 16:24:08 which is? Mar 28 16:24:31 mom ... i've to restart a build with that Mar 28 16:24:48 ok let me explain something before Mar 28 16:24:55 the marvell 510 is a SOC Mar 28 16:25:06 that contains a CPU + some other stuff Mar 28 16:25:15 the cpu can be for instance cortex A8 Mar 28 16:25:26 so find out what CPU that SOC has Mar 28 16:26:41 ARM PJ4 Mar 28 16:29:49 mhm ... its past this point where it filed with gcc-cross-inital ... is it because the tmp is not clean? Mar 28 16:30:56 I don't understand Mar 28 16:31:01 could you give more infos? Mar 28 16:32:30 i've tried to compile with the tune_armv7 days ago with a clean oe tmp dir Mar 28 16:32:42 Does anyone know of a way to determine how much space can be safely used on tmpfs? Mar 28 16:32:42 and always failed at gcc-cross-initial Mar 28 16:33:03 so someone here told me that i should include the armv7a and then it worked Mar 28 16:33:17 VtS, ok Mar 28 16:33:33 now i've changed back to armv7 and now there is no gcc-inital error Mar 28 16:35:11 checking for suffix of object files... configure: error: in `/storage/raid6/Dev/cubox/cubox-oe/build/tmp-eglibc/work/armv7-vfp-oe-linux-gnueabi/gcc-cross-initial-4.6.3+svnr184847-r24/gcc-4_6-branch/build.x86_64-linux.arm-oe-linux-gnueabi/arm-oe-linux-gnueabi/libgcc': Mar 28 16:35:11 | configure: error: cannot compute suffix of object files: cannot compile Mar 28 16:35:40 no i've got it ... ^^ Mar 28 16:36:10 <-- short afk Mar 28 16:37:13 look in config.log Mar 28 16:37:24 but for which file to include I've no idea Mar 28 16:37:32 because your marvell CPU looks strange Mar 28 16:37:52 it doesn't seem to be a classic IP core from arm Mar 28 16:38:06 s/IP/licensed Mar 28 16:38:37 they say it support armv6 and armv7 Mar 28 16:38:52 but it says nothing about armv4t and armv5te Mar 28 16:39:57 anyway all that is in CPU datasheet Mar 28 16:40:04 s/CPU/SOC Mar 28 16:40:15 http://www.marvell.com/application-processors/armada-500/assets/Armada-510-Functional-Spec.pdf Mar 28 16:44:23 so maybe you need to create one for the "sheeva" CPU Mar 28 18:13:27 denix: =+ and += are not the same Mar 28 18:14:29 gm Mar 28 18:15:57 Tartarus: yes, I know that well :) Mar 28 18:16:54 Tartarus: so what was that about? Mar 28 18:17:07 We haven't come full circle, wrt IMAGE_FSTYPES Mar 28 18:18:48 how can i determine in which order the task of a package is executed? Mar 28 18:19:04 so, you now want to prepend, instead of append? Mar 28 18:19:45 It's also about when it's applied, iirc Mar 28 18:25:01 really? I thought only overrides are delayed Mar 28 18:25:20 and += as well as =+ are applied right away... Mar 28 19:02:33 hi, I've a strange error: Mar 28 19:02:35 ERROR: QA Issue: non -dev/-dbg/-nativesdk package contains symlink .so: chromium path '/work/armv7a-angstrom-linux-gnueabi/chromium-19.0.1049.3-r0/packages-split/chromium/usr/bin/chrome/libffmpegsumo.so' Mar 28 19:02:49 I guess I shouldn't simlink manually the libraries right? Mar 28 19:03:46 I'll move the shared library instead... Mar 28 19:10:05 Anyone got an idea what to add to an sdk in oe-classic to get libopengl32? After adding the "libqtopengl" ipk ld fails because of -lopengl32. Mar 28 19:11:11 and in general, how do I figure out what packages I need to add to an sdk to get what I need? is it usually -dev? Mar 28 20:04:31 I have a system that uses two physically identical boards, but with different peripherals such that they need different kernels (and bootloaders). Is there a good way to set up a kernel recipe such that the output image names don't collide, or do I need two MACHINE values? Mar 28 21:16:23 Tartarus: as I thought, =+ wasn't something special - http://thread.gmane.org/gmane.comp.handhelds.openembedded.core/17731/focus=18040 Mar 28 21:16:43 Then we're almost back to where we started :) Mar 28 21:16:49 :) Mar 28 21:17:42 denix: whats the issue ? Mar 28 21:18:45 khem: no issues, trying to decide between =, ?=, += and =+ for IMAGE_FSTYPES :) :) Mar 28 21:19:55 where Mar 28 21:20:34 do you want to honor the defaults ? then go with .= or =. Mar 28 21:20:57 thats more like append/prepend logic Mar 28 21:23:21 .= oe =. won't be safe - no space added, += or =+ would be better Mar 28 21:23:38 but read the thread for the back story :) Mar 28 21:29:38 ok Mar 28 21:29:57 so you want to add to and not override what distro wanted Mar 28 21:30:27 in other words machine dictated FSTYPES and anything user defines should complement Mar 29 02:30:56 anyone know what would cause bitbake to hang after the "NOTE: Preparing runqueue" step? Mar 29 02:36:07 looks like it's trying to download shared sstate cache files Mar 29 02:36:11 sigh **** ENDING LOGGING AT Thu Mar 29 02:59:58 2012