**** BEGIN LOGGING AT Fri Aug 14 02:59:58 2015 Aug 14 06:11:49 does journalctl store everything from stdout stderr on tmp ? where can i change the configuration of journalctl ? Aug 14 06:26:24 hi guys Aug 14 06:32:15 i've got a problem with a recipe. I'd like to build nmea lib. everything looks ok when i build the recipe, but when i want to add this dependancy into an image, i get an error, saying that nothing provides "libnmea.so" from -dev pkg Aug 14 06:36:30 drou: mind posting the error message when u tried to bitbake the image? Aug 14 06:42:25 here the error log: http://paste2.org/MeIYkHEg and here is the recipe: http://paste2.org/xW63tdbC (recipes-devtools/nmea/nmea_0.5.3.bb) Aug 14 06:42:30 error occurs when bitbake is doing the do_rootfs Aug 14 06:46:10 rpm -qlp ./tmp/deploy/rpm/core2_32/nmea-dev-0.5.3-r0.core2_32.rpm (contains no files) Aug 14 06:46:29 drou: perhaps I would recommend http://pastebin.com next time....somehow the links you posted contained some NSFW materials that wouldn't be pleasant when I'm on work Aug 14 06:46:55 sorry for that, didn't know :( Aug 14 06:48:02 a quick try would be to clear the tmp dir and try again Aug 14 06:48:38 you mean, clear everything ? Aug 14 06:48:52 u can retain the sstate-dir Aug 14 06:49:08 just clear the tmp directory inside your build dir Aug 14 06:49:56 i'll try but i guess the build will take a while then Aug 14 06:50:15 hopefully not too long since your builds are already cached Aug 14 06:57:37 is there any way to avoid those -dev pkgs ? Aug 14 06:57:51 drou: not really sure Aug 14 06:58:03 tmp trick didnt work? Aug 14 06:58:56 i don't know atm, it's building. but i was just asking Aug 14 07:08:42 no, it's not working :( Aug 14 07:11:45 you can take a look at your packages-split directory Aug 14 07:12:23 here is what i have in the rpm generated : http://pastebin.com/eyYAYU1h Aug 14 07:12:30 where is the packages-split dir located ?! Aug 14 07:13:56 tmp/work//nmea//packages-split Aug 14 07:17:20 there is no such directory Aug 14 07:17:39 under ~/yocto/build/tmp/work/core2-32-poky-linux/nmea/0.5.3-r0 i only have 'pseudo' & 'temp' directories Aug 14 07:21:46 hmm...that's weird Aug 14 07:23:44 one strange thing is that when i tried to build the image, the last build failed and when i start it again, bitbake is doing do_package_setscene it hasn't do in the previous build Aug 14 07:25:38 now i got the packages-split directory Aug 14 07:26:47 i have 3 subdirectories, nmea, nmea-dbg, nmea-dev. the two first are not empty, the last one is... Aug 14 08:04:23 i guess there is something wrong with my recipe :( Aug 14 08:07:11 i also have another error: ERROR: QA Issue: bni-debug rdepends on bni-debug-dev [dev-deps] Aug 14 08:07:44 drou: I'm afraid the libnmea.so is not really installed Aug 14 08:07:57 so like u said u will need to check your recpie Aug 14 08:08:00 *recipe Aug 14 08:09:29 yes, but i can't understand why the lib is not added into the -dev package Aug 14 08:09:31 http://pastebin.com/nviPDhPP Aug 14 08:12:20 I noticed that both FILES_${PN} and FILES_${PN}-dev contain ${libdir}/*.so* Aug 14 08:12:30 not sure if that's the roote Aug 14 08:12:34 *root Aug 14 08:12:57 perhaps bluelightning: can help? Aug 14 08:13:30 this stuff can be tricky at times unfortunately Aug 14 08:13:36 (morning all) Aug 14 08:13:49 let me examine the recipe a bit more closely Aug 14 08:14:31 bluelightning: it certainly can be :-( Aug 14 08:14:34 hi bluelightning Aug 14 08:15:10 yes parrot1, the install error tells that nothing provides "/usr/lib/libnmea.so" so I added into the -dev package Aug 14 08:15:17 otherwise, i don't really need this Aug 14 08:16:55 drou: so.. this is probably the result an RPM per-file dependency that for some reason isn't being provided by the package Aug 14 08:17:13 I'm a little hazy on how those get set for shared libraries Aug 14 08:19:44 i understand, so how can i get rid of this ? is there any way to avoid generating those dev-pkgs ? Aug 14 08:21:40 hello guys. Aug 14 08:23:03 I want to know. if my linux can't mount my mmc device (SDCARD) .. is it a kernel problem or an electronic one? Aug 14 08:25:10 drou: I don't think the -dev packaging is an issue with the nmea recipe Aug 14 08:27:41 bluelightning: you think this is something related to the image recipe? Aug 14 08:28:08 drou: no, it's either about the packaging or about how the library is being compiled Aug 14 08:28:31 (and when I say packaging, I mean what rpmbuild is doing to actually create the package file) Aug 14 08:28:51 drou: with bni-debug I suspect a binary (probably a non-symlink .so) is going into the -dev package when it really ought not to be in this case Aug 14 08:29:59 drou: to explain, our default FILES values are set up assuming that when you build a library, it uses standard library versioning such that the main libxyz.so file is a symlink to the library rather than the library itself, and the library itself has a version number in the name Aug 14 08:30:16 the .so symlink is for development only, hence it goes into the -dev package Aug 14 08:31:37 bluelightning: actually, this library has a makefile i don't use, because it builds a static library, maybe the problem is the way of building nmea as a shared library? Aug 14 08:35:41 it was working as a static library, but i wanted to get rid of a warning message that said one of my library has a relocation in .text Aug 14 08:38:24 drou: well, just building it as a shared library isn't going to be the cause, we build just about everything as a shared library Aug 14 08:39:19 yes, i guess so Aug 14 08:46:51 is there a command to build all packages? I want to create a package feed. Aug 14 08:47:05 tasslehoff: bitbake world perhaps? Aug 14 08:47:43 bluelightning: sounds promising :) Aug 14 08:47:46 yo, verified on gentoo x86 and amd64 (with full multilib ABI_X86="64 32" build) Aug 14 08:48:06 tasslehoff: that's probably a lot more than you want though... you could either use a bunch of EXCLUDE_FROM_WORLD_pn-xyz = "1" settings to reduce that, or create a "meta" recipe which depends on all the others you want - so basically blacklist vs. whitelist Aug 14 08:48:14 no more libpseudo fail Aug 14 08:48:49 * nerdboy building ancient kernel fsl-imx28 image on laptop Aug 14 08:49:31 and still warning me i'm "unsanitary" or some ting... Aug 14 08:49:39 bluelightning: Ok. I'll try the build and see how bad it is :) Aug 14 08:49:47 *or something even Aug 14 08:50:33 Host distribution "Gentoo-2.2" <= works just fine Aug 14 08:50:55 always did on 32-bit... Aug 14 08:52:00 drou: can you pastebin the libnmea file from pkgdata/runtime in the libnmea workdir? Aug 14 08:52:06 morning all Aug 14 08:56:59 bluelightning: http://pastebin.com/QjymcWjE Aug 14 09:07:49 drou: interesting, so the lib mentioned by name in the error is in FILERPROVIDESFLIST Aug 14 09:11:59 bluelightning: what does that mean? Aug 14 09:13:42 can you do something like "python do_configure_prepend() { ..."? Aug 14 09:14:30 oh, and i needed to make a stupid perl patch to build that old fsl kernel Aug 14 09:14:46 drou: it means that I don't understand how you'd see that error, since what it's complaining about *is* provided Aug 14 09:15:30 topik: I'm afraid not, however you can do something like do_configure[prefuncs] += "myfunc" and then define myfunc as a python function Aug 14 09:15:45 i guess only fsl cares about that, although i can't figure why it's not building 4.1 instead... Aug 14 09:16:44 probably only otavio knows for sure... Aug 14 09:17:20 bluelightning: so do_configure_append is basically a shorthand for adding a python wrapper around a shell function and adding that to do_configure[myfunc]? Aug 14 09:17:52 bluelightning: it's complaining about bni-debug-dev. -dev can't find /usr/lib/libnmea.so because nmea-dev doesn't provide it Aug 14 09:23:29 LetoThe2nd: no, not at all, do_configure_append just says stick this on the end of do_configure without regard for its content - which means it must be written in whatever language the original function was written in Aug 14 09:24:07 drou: what's in bni-debug-dev? Aug 14 09:24:27 here is the pkgdata/runtime/nmea-dev http://pastebin.com/eYX0ZKNL Aug 14 09:24:43 hmm, ok thanks Aug 14 09:24:45 bluelightning: ah ok. *mentalnote* Aug 14 09:24:48 i don't know, because this recipe is not explicitely written. Aug 14 09:25:02 LetoThe2nd: prefuncs/postfuncs on the other hand can be whatever you want them to be since they're executed through separate explicit function calls Aug 14 09:25:16 drou: er, what do you mean by that? Aug 14 09:25:41 bluelightning: so the prefuncs come before _prepend, while postfuncs come after _append? Aug 14 09:26:53 LetoThe2nd: effectively yes - we execute the prefuncs (if any), then the entire task function with all overrides/prepends/appends applied, and then all the postfuncs (if any) Aug 14 09:27:10 bluelightning: i see, thanks for the explanation! Aug 14 09:27:37 prefuncs and postfuncs aren't often needed, though we do use them quite a bit internally in OE-Core classes and within bitbake Aug 14 09:28:23 guessed something like that, yes. Aug 14 09:28:57 sounds very much like half-internal facility that *can* be used in a recipe, but isn't primarily intended to. Aug 14 09:29:19 bluelightning: i defined a bni-debug.bb but no bni-debug-dev.bb. the error given when doing the image is on bni-debug-dev (??) Aug 14 09:29:25 bluelightning: here is the recipe: http://pastebin.com/56StaXD6 Aug 14 09:29:58 error was: error: Can't install bni-debug-dev-1.0-r0@core2_32: no package provides /home/cedric/yocto/build/tmp/sysroots/vecow-jdebug-bni-desk-chief/usr/lib/libnmea.so Aug 14 09:34:38 drou: oh... well... that is definitely wrong Aug 14 09:35:21 drou: you definitely shouldn't have that sysroot path mentioned at this stage, that points to something wrong in how a binary within that package has been linked Aug 14 09:36:12 drou: you wouldn't have a bni-debug-dev.bb, bni-debug-dev is a _package_ defined by the bni-debug _recipe_ (different namespaces) Aug 14 09:40:59 yes but i can't figure out what part of the recipe defines this -dev package Aug 14 09:44:30 drou: it's defined in meta/conf/bitbake.conf, it provides default values for PACKAGES and FILES Aug 14 09:44:39 among a number of other things Aug 14 09:45:05 drou: if you need to examine everything that's being set in the recipe context, I'd recommend bitbake -e recipename | less Aug 14 09:45:22 it provides full history of where each value is set also Aug 14 09:46:47 yes, -e option is very useful Aug 14 09:47:00 # set /home/cedric/yocto/build/../meta/conf/bitbake.conf:264 # "${PN}-dbg ${PN}-staticdev ${PN}-dev ${PN}-doc ${PN}-locale ${PACKAGE_BEFORE_PN} ${PN}" Aug 14 09:48:53 but obviously i cannot edit this Aug 14 09:57:33 if my linux can't mount my mmc device (SDCARD) .. is it a kernel problem or an electronic one? Aug 14 09:59:11 drou: you can set your own value easily in the recipe though Aug 14 09:59:26 which will override what's set there Aug 14 10:12:17 Hello guys. Aug 14 10:12:26 I have a tiny issue. Aug 14 10:12:40 binaries built with my sdk use the wrong dynamic loader. does anyone know how the loader is chosen? could be I'm misusing the sdk... Aug 14 10:13:12 hi agherzan - what's the issue? Aug 14 10:13:52 I want to add a file to SRC_URI which has an odd name. This is due to the way systemd encodes strings while using mount services. So i need the seq "\2d" to be in the file name. Well, this confuses yocto badly with errors. Aug 14 10:14:12 bluelightning: was typing :) thanks for fast reply Aug 14 10:15:06 Did anyone before used mount services with yocto so it would be able to add services like this? Aug 14 10:15:30 agherzan: does the source filename need to be that or could it be named correctly at do_install time? Aug 14 10:22:09 bluelightning: If i do it at install time i loose the auto detection of systemd class and installment. But i can do it manually I think... Aug 14 10:22:21 You propose to rename it at install time? Aug 14 10:23:10 Obviously, as a workaround. Aug 14 10:32:38 agherzan: what is the actual error? I'm not sure where this is going wrong yet... Aug 14 10:34:40 bluelightning: https://bugzilla.yoctoproject.org/show_bug.cgi?id=8161 Aug 14 10:34:41 Bug 8161: normal, Undecided, ---, ross.burton, NEW , Can't use "\" characters in SRC_URI entries - not even escaped Aug 14 10:36:05 agherzan: ok thanks Aug 14 10:36:16 I think you may have to work around this for now Aug 14 10:36:44 I know. I submitted the bug just for traceability . Aug 14 10:36:56 I'll take a look on it if I have time. Aug 14 10:37:34 ok, great... even if not, it's definitely worth having the bug so we have it tracked as you say Aug 14 10:38:19 bluelightning: thanks mate. Aug 14 10:39:12 tasslehoff: not sure; I know the SDK itself has its own dynamic loader for the host side, but that shouldn't be affecting anything you build with it Aug 14 10:39:40 bluelightning: it is my own fault in some way. a simple Makefile project did the right thing after sourcing the env setup Aug 14 10:40:07 my attempt at bringing the env into our SCons setup fails. some envvar that affects the linker selection. Aug 14 11:22:23 Hi guys. Yesterday I ask about g_ether driver for rndis-usb device (actually my smartphone as hotspot over usb). And now driver is on board, but modprobe says "could not insert 'g_ether': No such device" Aug 14 11:23:22 usb-devices command says, that "Driver=(none)", and I think this is the problem Aug 14 11:23:35 any suggestions? :) Aug 14 11:24:21 pohmelie: I think there might be some device list that the driver recognises, but there is possibly a parameter that lets you influence that? I don't recall the details, but there should be some documentation for g_ether out there on the web somewhere - maybe part of the kernel docs Aug 14 11:25:41 bluelightning: hm, nice idea, thanks ;) Aug 14 11:34:19 bluelightning: I think -mfloat-abi=hard should be in LDFLAGS as well. Aug 14 11:34:56 tasslehoff: isn't that a CFLAGS type option? Aug 14 11:35:45 bluelightning: if I don't include it in my linker command line it chooses the wrong dynamic loader Aug 14 11:36:23 tasslehoff: ah ok... Aug 14 11:37:56 bluelightning: unless I'm missing/misunderstanding something, I think that is a real bug. Aug 14 11:41:30 that I'll have to resume next week. have a nice weekend. Aug 14 12:24:07 I want to know what happens if I set SRC_URI="file://myprog.c" and myprog.c file is located at ../file/mydir/myprog.c ??? Aug 14 12:38:21 i'm buidling the image with multilib support (config as wiki) and a hello world app with the lib32- prefix. Should i build the image as lib32- prefixed? Aug 14 12:40:14 if i try to deploy the deb pkg to the image in the target pkg manager complains about package architecture (i386) does not match system (amd64). so im missing something. Aug 14 13:00:11 sm0ketst: I'm not sure if deb + multilib is very well tested in our system Aug 14 13:01:43 sm0ketst: the lib32- image may not be that interesting.. typically you build just the libs you need to support whatever it is that requires them, rather than having the entire contents of the image also mirrorored in the multilib (which is what lib32-imagename will do) Aug 14 13:02:46 bluelightning: so is it better to use ipk? i think i was doing the monkey... defining the ml support in the bsp/machine config. BTOH the image complains about lib32-glibc not provided. so I'm indeed testing to put the multilib:lib32 in the local.conf. seems to build and the mannifest is including the lib32-glibc... im trying... Aug 14 13:03:34 bluelightning: i saw the lib32- google regarding to multilib support... but no way in my case Aug 14 13:05:35 what tools do you guys use to flash a UBI image in NAND from u-boot ? Aug 14 13:10:37 bluelightning: works. i had to move the ml32 lines to the local.conf instead of the bsp layer/machine. then add to the corresponding helloworld.bb the corresponding lines. seems to work. Aug 14 13:13:02 sm0ketst: yeah the machine conf is not the right place for this kind of thing - local.conf will work for testing, otherwise it really belongs in a custom distro conf Aug 14 13:13:17 zecke, hi ! -- when you flash a ubi/ubifs image in NAND, do you flash it from u-boot using 'ubi write' or use flash it form linux using ubiformat ? -- If I don't use ubiformat, i'm getting a bunch of ECC error when I boot a linux from a NAND Aug 14 13:15:39 bluelightning: I'm still confused about bsp/local.conf... my head sometimes is like a rock Aug 14 13:18:02 bluelightning: http://pastebin.com/MYh8grUh Aug 14 13:19:49 sm0ketst: I don't think you need that BBCLASSEXTEND to the multilib Aug 14 13:26:44 bluelightning: which line? native? both? Aug 14 13:27:21 sm0ketst: BBCLASSEXTEND += " multilib:lib32" Aug 14 13:28:15 bluelightning: going to try, thanks Aug 14 13:28:30 sm0ketst: also you shouldn't need the MLPREFIX line Aug 14 13:28:50 sm0ketst: this is all I need to add to my local.conf to enable multilib here last time I needed to do it: http://pastebin.com/3yuY8JZL Aug 14 13:36:28 bluelightning: ok, so then nothing is required in the bb recipe for the application/lib since it is enabled in the config if I understood well. Going to check it. Thanks again! Aug 14 13:36:42 sm0ketst: no worries Aug 14 13:36:58 sm0ketst: and yes you understood perfectly :) Aug 14 13:40:49 Good Morning all, I am wondering if anyone else is having issues with qtwebengine linking from the meta-qt5 layer? I keep getting a link failure related to icu. From what I can tell qtwebengine seems to be looking for icu 4.6 but poky has it at 52.1. Aug 14 13:42:26 bluelightning: indded folks/oe/yocto ur doing good work. have a nice weekend all Aug 14 14:17:15 damnit autobuilder Aug 14 14:20:04 For log: to use usb-eth device (RNDIS) (your smartphone, for example), which is not your board and just is a way to share internet, you need build kernel with "CONFIG_USB_NET_RNDIS_HOST=y" and all prereqs (probably specific modules for kernel, to activate your usb AMD5536UDC as example for geode, but probably you don't need this). Aug 14 14:21:25 For log: next step after start system and plug-in your "internet" you should see your usb0 in "ifconfig" command result. Then you need start dhcp via "udhcpc -i usb0". Aug 14 14:21:47 For log: after that you can ping hosts and open them in browser Aug 14 16:10:23 NOTE: Tasks Summary: Attempted 6530 tasks of which 3467 didn't need to be rerun and all succeeded. <= yippee-doo Aug 14 16:11:36 just the one kernel fail and a few QA warnings... Aug 14 16:31:39 hi, is opkg still developed these days? It seems that opkg list-installed foo would return 0 in both cases whether or not foo is installed Aug 14 16:32:03 it is a bit difficult this way to differentiate between the two outputs without actually parsing the string. Aug 14 16:33:57 lpapp: I believe there is a new maintainer yes Aug 14 16:37:17 cool Aug 14 16:37:52 I think opkg list-installed | grep foo will be my friend this time. I prefer exit code over parsing output :) Aug 14 16:37:55 (especially in C) Aug 14 16:39:07 I guess for now you'd use grep -q and you'd have your exit code, but I know what you mean Aug 14 16:40:49 the maintainer is stepping back a bit now iirc Aug 14 16:41:08 lpapp: if you're going to do that, you might want grep -x, which matches a whole line, not substring Aug 14 16:41:20 * kergoth yawns Aug 14 16:41:58 hmm, yes, but that means doing output parsing based on expectation :) Aug 14 16:43:00 but I can hear you because grep foo can also fall apart. Aug 14 16:43:14 the best would be to address this in opkg, I think. Aug 14 16:43:21 either that or use anchors (^/$) Aug 14 16:43:22 indeed Aug 14 16:47:27 thanks to all Aug 14 16:57:16 rburton: the one after Paul as well? Aug 14 16:57:59 ah maybe i didn't notice someone take up after paul Aug 14 16:58:21 * kergoth loses track of opkg maintenance status Aug 14 16:58:53 ah yes, alejandro del castillo has been adding tests and fixed a bug Aug 14 16:59:41 i should tell him that we don't need three ways of doing opkg compare-versions, where the only sensible candidate (using the same logic) makes random comparisons and returns inverted responses Aug 14 17:00:14 i seem to recall tableflipping and giving up when i root-caused that problem in my test suite Aug 14 20:04:27 * nerdboy glad he saved state-cache Aug 14 20:34:02 otavio: yo Aug 14 20:34:15 * nerdboy having a thought... Aug 14 20:34:26 and please don't call the fire dept Aug 14 20:35:05 as behan runs from the bldg... Aug 14 20:35:18 i said it wasn't an emergency Aug 14 20:39:26 nerdboy: hi :) what's up Aug 14 20:39:58 who would care about broken imx28 kernel build? Aug 14 20:40:57 and i guess i'm wondering why linux-fslc isn't default provider for imx28-evk Aug 14 20:41:24 the fail in the olde kernel is a perl thing Aug 14 20:41:42 what was once a deprecation warning is now a hard error Aug 14 20:41:52 *in perl 5.22 Aug 14 20:42:07 plus it appears not to use perl-native Aug 14 20:42:08 meh Aug 14 20:42:21 easy fix is change provider Aug 14 20:43:58 we'll push a couple of imx28 products onto latest yocto/fsl Aug 14 20:44:07 *be pushing even Aug 14 20:44:43 so we'll probably have to fix stuff, maybe even make a new machine Aug 14 20:45:47 details are still forthcoming, so not sure how much hacking is required yet Aug 14 20:46:14 i'm rebuilding fsl-image-machine-test with new kernel now Aug 14 20:46:30 it completed last night with kernel patch Aug 14 20:46:38 *the old one Aug 14 20:48:37 otavio: this is all plain follow-the-instructions build with fsl-community-bsp btw Aug 14 20:49:40 nerdboy: but this failure is in master? Aug 14 20:50:15 and gentoo-x86/amd64-multilib-strict are now certified platforms as far as i'm concerned Aug 14 20:50:38 otavio: is that what "fido" is? Aug 14 20:50:57 meta-yocto = "(nobranch):08f9fbbc97909698b56200aa9be5fe50b99a44b2" Aug 14 20:50:57 nerdboy: fido is stable; 1.8 release Aug 14 20:51:16 whatever that commit is i guess... Aug 14 20:51:55 * nerdboy poking debian people Aug 14 20:53:03 otavio: after this test build, and when i get more hardware data, i will make my own layer setup (no repo manifest junk) and do a master build of everything Aug 14 20:53:40 *everything imx28 that looks close anyway... Aug 14 20:53:53 nerdboy: fido uses linux-imx; master uses linux-fslc Aug 14 20:54:16 okay, well, stable is broken then Aug 14 20:54:32 you guys get more like gentoo every day Aug 14 20:54:37 congrats Aug 14 20:55:26 * nerdboy been on full ~arch for a long time now... Aug 14 20:55:45 who needs "stable" anyway Aug 14 20:56:50 still, i suppose it would be nice if the default Just Worked when new people try following instructions Aug 14 20:59:32 here's the kernel patch for "fido" if you want it... Aug 14 20:59:46 http://www.gentoogeek.org/files/0001_fix_kernel_timeconst_perl_array_error.patch Aug 14 21:57:17 nerdboy: send the error report Aug 14 21:57:33 nerdboy: but as far as I know it is working; it was working last time I tried it Aug 14 21:57:48 nerdboy: but we can fix it, if you send the fix Aug 14 23:23:33 otavio: it's an error in perl 5.22 Aug 14 23:24:04 i assume that's newer than build machine but it should not be using build host anyway Aug 14 23:24:27 should depend on perl-native i would think... **** ENDING LOGGING AT Sat Aug 15 02:59:58 2015