**** BEGIN LOGGING AT Thu Aug 24 03:00:01 2017 Aug 24 07:54:49 If I might ask Aug 24 07:55:22 I do have one meta layer supporting one board Aug 24 07:55:55 What is the recommended way to support another board (with the same SoC, and 90% of packages)? Aug 24 07:56:56 Shall I define in /config/conf/local.conf new MACHINE ? Aug 24 07:57:06 or is there any other (better) way Aug 24 08:00:11 The only tweaks would be with kernel and u-boot recipes Aug 24 08:00:49 (to adjust the patch for the remote server repo) Aug 24 08:06:44 grrr... first look (grep) then ask ;-) Aug 24 08:06:48 PREFERRED_PROVIDER_virtual/kernel ?= "linux-yocto-custom" Aug 24 10:30:40 heyho Aug 24 10:31:53 i'm trying to build code for the zedboard (zynq7020), currently, with a custom receipe, it fails to compile the code, complaining with: arm-poky-linux-gnueabi-g++: error: unrecognized argument in option '-mtune=generic Aug 24 10:32:10 looking at the compiler call, -mtune was not set at all Aug 24 10:33:13 what would be the right way to set the flag? putting it into the receipe is imho wrong, since the receipe should not be tied to this platform Aug 24 11:18:40 crazy_imp: nrossi is the meta-xilinx expert Aug 24 11:19:02 nrossi: He normally seems to wake up when I mention is nick.. :P Aug 24 11:19:10 s/is/his/ Aug 24 11:19:14 * nrossi rawwwrrr Aug 24 11:21:18 crazy_imp: is the makefile/etc for you recipe setting '-mtune=generic' if so, it's probably doing it wrong. Those flags should come from the tune config (e.g. http://git.openembedded.org/openembedded-core/tree/meta/conf/machine/include/tune-cortexa9.inc#n6) Aug 24 11:24:20 crazy_imp: also for arm, '-mtune=generic' is invalid, you see to specific the with a "-" on the end, see https://gcc.gnu.org/onlinedocs/gcc/ARM-Options.html Aug 24 11:25:14 nrossi: i did not set it, nor was it in the compiler parameters. see https://pastebin.com/SuRMekSt Aug 24 11:27:35 crazy_imp: what version of OE/yocto you using? and is the feature-extraction recipe private or public? Aug 24 11:28:25 private. poky DISTRO_VERSION = "2.2.2", morty branches Aug 24 11:35:02 crazy_imp: you sure the error is for the exact same command and not that of a different thread/job? Aug 24 11:36:42 yes, was only rebuilding my receipe, all other stuff was already build at that point Aug 24 11:37:28 crazy_imp: i mean within your recipe... e.g. if your building with PARALLELL_MAKE = "-j" Aug 24 11:40:18 nrossi: yes, i'm sure, patched my receipe to add the correct do_compile and it blew up afterwards, so there were no other receipes involved at this point, deps were already build Aug 24 11:42:38 in meta-oe/recipes-support/openldap/openldap_2.4.45.bb there is LDFLAGS += "-pthread" which fails the build of openldap-native (simple openldap recipe works fine) because pthread is not searched for at link time, however if I replace it with LDFLAGS_append = " -pthread" then it works... looking at the run.do_compile of openldap-native, LDFLAGS += "-ptrhead" does not append to LDFLAGS, only LDFLAGS_append + " -pthread" does Aug 24 11:43:02 can someone explain to me the difference between LDFLAGS += vs LDFLAGS_append = Aug 24 11:43:04 ? Aug 24 11:43:34 crazy_imp: so the only command running is the g++ one in your log? theres no makefile or other g++ commands being run by your recipes do_compile? Aug 24 11:44:03 I tried searching the manuals, but didn't find anything concluding, maybe use BUILD/TRAGET_LDFLAGS in the openldap recipe? http://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#var-LDFLAGS Aug 24 11:47:28 aratiu: https://www.yoctoproject.org/docs/1.6/bitbake-user-manual/bitbake-user-manual.html#appending-and-prepending 3.1.6 and 3.1.8 (specifically the last note) Aug 24 11:47:36 nrossi: it calls some other makefiles, but all those files do, is to generate some more soucre files out of binary data (which runs fine and succeeds). next up it to build my code, that's what was in the pastebin. Aug 24 11:48:23 also tried it again now, the buildlog says, the output is from the receipes do_compile, so it's not from another source Aug 24 11:50:28 nrossi: thanks a lot, the relevant info in that manual I think is that _append happens after parsing, so in my case during prasing something overwrites the LDFLAGS variable for the native recipe Aug 24 11:52:11 crazy_imp: tbh its a weird error, since your not passing an -mtune option. It might be a bug in gcc. Add "-mtune=cortex-a9" to CFLAGS/your command and see if it resolves it. Aug 24 11:52:57 aratiu: If this is a -native recipe, it might be that something setting LDFLAGS_class-native or similar. Your can see the applied values and how its being set with bitbake -e Aug 24 11:55:00 crazy_imp: if forcing mtune doesn't help, you can try and track down where the arg is introduced by using -verbose on g++, it will show gcc sub processes Aug 24 12:01:58 interestingly, the error comes from the lto-wrapper, so maybe disabling -flto will do the trick Aug 24 12:07:55 nrossi: I think I found the root cause: native.bbclass sets $LDFLAGS to "${BUILD_LDFLAGS}" after the openldap-native recipe appends to it, here's the relevant log from bitbake -e: https://pastebin.com/raw/4Y9mpGFY Aug 24 12:14:22 nope, that's not the issue :| Aug 24 12:16:44 hello guys.. I have one question. I have a custom application that must use "libmosquitto1" which is rprovided by mosquitto Aug 24 12:17:41 So in my custom app recipes I've added RDPENDS_{PV}="libmosquitto1" Aug 24 12:18:44 the problem is that when I try to compile I got the following error: ld: cannot find -lmosquitto1 Aug 24 12:18:56 what can I do ? any suggestions ? Aug 24 12:19:56 fberg: you still need "DEPENDS = mosquitto" in the recipe so it'll be available for building Aug 24 12:20:17 also, there's a typo Aug 24 12:20:21 fberg: RDEPENDS_${PN} not PV Aug 24 12:20:22 it should be RDEPENDNS Aug 24 12:20:34 not RDPENDS (missen E) Aug 24 12:21:05 that adds a runtime dependency for a specific package, you need a build dependency, use DEPENDS as smurray suggested Aug 24 12:21:08 h Aug 24 12:21:14 hi everyone Aug 24 12:22:59 adding mosquitto to DEPENDS= doesn't solve the issue. I've got the very same error. I Have also corrected PV to PN, reguarding RDEPNDS it was a typo ! Aug 24 12:29:36 nrossi: https://pastebin.com/HFaMKNnt added the -mtune option and tried to build it by hand, still the same... Aug 24 12:33:55 fberg: which mosquitto recipe? Aug 24 12:34:50 http://git.yoctoproject.org/cgit/cgit.cgi/meta-intel-iot-middleware/tree/recipes-connectivity/mosquitto/mosquitto_1.4.10.bb?h=master Aug 24 12:37:06 it's very strange Aug 24 12:38:47 fberg where does "-lmosquitto1" come from? Aug 24 12:39:02 I would have expected -lmosquitto Aug 24 12:39:13 when I try to build my application that needs the mosquitto library Aug 24 12:39:55 I've a cmake directive that add it when compiling Aug 24 12:40:00 right but are you sure it's correct? does it come from pkg-config or something? Aug 24 12:41:12 I thoughtn as usual.. so i've dumped the lib prefix and added the -l to the library Aug 24 12:44:12 try "-lmosquitto" that looks to match what the recipe packages Aug 24 12:45:28 also if you're linking to it, you don't need any RDEPENDS_${PN} at all, DEPENDS is enough Aug 24 12:46:39 ok Aug 24 12:48:08 crazy_imp: mmmm i'm out of ideas, try removing all your custom optimization args and see if that resolves it, then slowly add them back one by one Aug 24 12:48:34 ty jku ! your hint fixed the problem Aug 24 12:48:53 really appreciated ! Aug 24 13:09:12 I need to set a setgid bit on a directory, where multiple packages install files, I know I can add a ROOTFS_POSTPROCESS_COMMAND, but does anyone knows a better way? Aug 24 13:10:12 and obviously I need to set a specific group which is not root... Aug 24 13:13:04 Whats the best way to search for pre-included recipes for packages? Specifically, I need my project to include the following: Flask, Flask-HTTPAuth, flask-restful, flask-sqlalchemy, flask-migrate, flask-marshmallow, marshmallow-enum, marshmallow-sqlalchemy, flask-script, flask-cors, gunicorn Aug 24 13:13:23 How can I find the ones that I can just include? Aug 24 13:15:19 tcpdump, you want to know where to find the recipies for flask and other things? Aug 24 13:17:05 nrossi: should i expect to see -mtune in env if i dump it inside the do_compile function? Aug 24 13:17:58 tcpdump, usually i'm using : http://layers.openembedded.org/layerindex/branch/master/recipes/ but maybe it is not what you are looking for ;) Aug 24 13:18:34 tcpdump: oe-pkgdata-util is useful for digging info on packages that have been do_package'd already... but really layerindex is easier for finding recipes if you can guess the name Aug 24 13:19:17 crazy_imp: nope (at least in oe-core), for arm -mcpu is preferred as it is more specific Aug 24 13:19:58 another piece of hope gone... Aug 24 13:20:14 aurele: I think that may be what I want. Thanks! Aug 24 13:21:25 after doing a devtool modify/build/finish cycle, if i do not push my local git changes up (because i don't own the parent repo), will devtool do the right thing if i do a fresh devtool modify/build/finish and the append and patch files are left in the layer from the initial modify/build/finish? Aug 24 13:23:02 jku: Yea, it seems like that should work out pretty good. Aug 24 13:23:12 It looks like that's a pretty extensive repo. Aug 24 13:23:27 i.e., i'm adding more changes to the src files in the second finish Aug 24 13:27:03 tcpdump : layerindex is v. useful as long as you remember that layers have varying levels of maintainership (in other words review before using new layers) Aug 24 13:27:19 jku: makes sense. Thanks Aug 24 13:27:20 ! Aug 24 13:28:24 So, if they're in here all I need to do is add them to my .bb file? ex: IMAGE_INSTALL_append = " gunicorn"? Aug 24 13:32:14 tcpdump: assuming you've "git cloned" the layer, checked out the correct branch, and added the layer dir to BBLAYERS in conf/bblayers.conf (and done the same thing to any layers that the new layer depends on) Aug 24 13:42:29 how can I deploy a *-native package to my SDK? I noticed that TOOLCHAIN_HOST_TASK does not work... Aug 24 13:46:30 if i do a "make menuconfig", make a change, save changes, and exit, what is changed? is a new file created? Aug 24 13:46:35 CoLa|work: you can't do that. you can add nativesdk packages to it, however. Aug 24 13:46:37 i'm seeing .config, .config.new; some are symlinks. this is under yocto. so i'm not sure who's doing what how/when. Aug 24 13:46:42 and i don't see any .config in my linux src when doing a yocto "devtool modify -x kernel-recipe src/linux" Aug 24 13:46:52 yates: .config in the build directory of the recipe, that's all. build dir != src dir Aug 24 13:48:11 CoLa|work, kergoth maybe with "TOOLCHAIN_TARGET_TASK_append" ? I added a package to my toolchain with this one Aug 24 13:48:20 kergoth: I see that the wayland package is deploying wayland-scanner to the sdk's x86_64 sysroot, so there must be a way to do the same with qtwayland-scanner, or not? Aug 24 13:48:26 kergoth: but isn't the Makefile that runs "make menuconfig" a makefile supplied by the linux kernel? why would it know anything about a recipe? Aug 24 13:48:45 CoLa|work: agian, add nativesdk packages rahter than native recipes to TOOLCHAIN_HOST_TASK Aug 24 13:49:33 nativesdk just adds the ARM binaries that I cannot use when cross-compiling from the host machine Aug 24 13:49:36 yates: oh, i see, i misread, thought you ran bitbake -c menuconfig, which is the better idea. make menuconfig will almost guaranteed be wrong, unless you're passing in the right env varss.. are you in devshell? Aug 24 13:49:50 CoLa|work: no, nativesdk is host binaries targeted for sdkmachine Aug 24 13:50:01 excatly the same as native, just for sdkmachine rather than the current build machine Aug 24 13:50:08 the two are generally compatible, however Aug 24 13:50:14 OK, then I might have some configuration problem, thanks Aug 24 13:50:53 note that nativesdk packages are prefixed rather than suffixed like -native. i.e. nativesdk-foo, not foo-native Aug 24 13:51:05 CoLa|work: laso note the BBCLASSEXTEND you need for nativesdk (see the wayland recipe for example) Aug 24 13:51:15 kergoth: if by "devshell" do you mean have i run ". oe-init-build-env build-dir", then yes, i think so Aug 24 13:51:46 i'm in an xterm that's running bash (i think) after running the oe-init-build-env Aug 24 13:52:39 very likely i'm totally hosing up the intended process Aug 24 13:53:26 your best bet is to just use bitbake -c menuconfig, it'll run things in the correct environment and pass the correct vars along. you can even do a bitbake -c diffconfig afterward to generate a kernel configuration fragment of the changes you made Aug 24 14:03:44 if i am in a devtool modify cycle with source in src/linux, and i modify the kernel config in src/linux, won't devtool finish generate the appropriate patch for the config file? Aug 24 14:04:07 the config file isn't part of the source tree at all, at least once configure runs Aug 24 14:04:37 as i said earlier, it ends up in the build directory, which isnt' the same as the source directory at all in this case Aug 24 14:04:49 so nope, won't work with the samew orkflow as source modifications Aug 24 14:05:52 now, once you have a new config file, if you're using a defconfig that lives in the source tree at all, which not all bsps do, then you could copy the new one into the correct path in the source tree and thereby patch it, but patching a defconfig is a pretty messy way to handle updating a kenrel configuration, you're better off using an out of tree defconfig, or using kernel config fragments if your recipe supports it Aug 24 14:06:27 of course, linux-yocto handles this differently, using fragments in the meta repository Aug 24 14:09:40 kergoth: how would you go about it if you had srctree changes to implement (e.g., in arch/arm/boot/dts/somefile.dts), and also needed to update the kernel config file? Aug 24 14:10:23 would you run devtool modify/build/finish first, then do the bitbake -c menuconfig? Aug 24 14:11:58 ..then of course finally bitbake image-recipe.. Aug 24 14:12:30 or am i still confused? sorry if i am not hearing you. some of what you suggested (e.g., defconfig) went over my head. Aug 24 14:25:01 Do we have some script to find unused .patch and .inc files (or even .bbclasses) in the metadata? Aug 24 14:25:06 by "defconfig" do you mean a default ".config" file? Aug 24 14:25:31 I think I've seen something like that on ML or somewhere, before I write something to cleanup after meta-oe removals Aug 24 14:28:48 RP: I have sent a v3 for db here https://patchwork.openembedded.org/patch/143058/ few days ago give it a try when you can Aug 24 14:30:00 khem: I know, I've had to prioritise other things, sorry. Will include it next along with the exploding python patches Aug 24 14:33:50 RP: thats why I said when you can :) Aug 24 14:35:03 RP: I am trying to build core image sato with clang and currently only 5 packages are forcibly built with clang and db is one of them. This will fix it Aug 24 14:35:28 others are pulseaudio, busybox, tcp-wrappers, elfutils, pixman Aug 24 14:36:10 once I get this then we will have pure clang based sato image Aug 24 14:39:25 https://www.openembedded.org/wiki/OEDEM_2017 Aug 24 15:12:27 anyone know of a way to nuke all connections on nmcli? Aug 24 16:47:58 Does anybody happen to know how to remove a configure option? I'd like to configure gdb with the "--with-python" flag but the "--without-python" flag is appended by default. Aug 24 17:31:56 samyin: the easiest would be to include it in your local.conf, i.e PACKAGECONFIG_gdb += " python" Aug 24 17:32:19 _pn-gdb Aug 24 17:32:29 and even then you probably want _append_pn-gdb, not _pn-gdb += Aug 24 17:32:40 otherwise you're removing every other config and replacing it with python Aug 24 17:34:38 kergoth: correct Aug 24 18:00:00 lsandov: Thanks much, that did the trick! Aug 24 18:59:09 smartin: great. take a look at bbappends, that would be the best place to include your statement. In the latter case, just remove the pn-gdb part, because you are already in the recipe context Aug 24 19:21:55 what is the correct way to fix "No GNU_HASH" QA issues? Aug 24 19:23:05 crazy_imp: fix the component to respect LDFLAGS from environment Aug 24 19:34:31 thanks JaMa :) Aug 24 20:12:22 Hello - has anyone built a Krogoth-based system on Debian Stretch(9)? I'm getting module 'bb' import errors trying to run bitbake, even though the python version hasn't changed. Aug 24 20:15:12 Well, only python revision change. 2.7.9->2.7.13 Aug 24 20:30:39 When building my sdk, the HOST python installation is incomplete (it's missing the json module). How do I fix this? Aug 24 20:31:13 I have an incomplete python3.5 installation only in the HOST side of my sdk sysroot. Aug 24 20:34:07 Circuitsoft: log? Aug 24 20:34:34 Circuitsoft: I know that master branch works as expected Aug 24 20:34:45 Circuitsoft: which of course, is python3 Aug 24 20:36:22 Master and Pyro work, but they deprecate packages that I need (xf86-video-nouveau) Aug 24 20:36:32 Krogoth is the challenge. Aug 24 20:37:10 https://clbin.com/AEGSd Aug 24 20:39:20 Circuitsoft: no idea. it would be good if you file a bug Aug 24 21:28:28 Is there a way to make part of recipe conditional execution if MACHINE = XXX and the other part when MACHINE = YYY. Aug 24 21:28:28 I can add suffix to variables and bash functions, but this is a lot of writing. Aug 24 21:28:28 I could also make large if and embrace with it the block of code for one machine and another Aug 24 21:28:37 What is the preferred way ? Aug 24 21:59:38 lukma: I think the preferred way would be to add usffix to variables in bash Aug 24 22:00:44 so group the parts for XXX and YYY and extend them with suffixes? **** ENDING LOGGING AT Fri Aug 25 03:00:02 2017