**** BEGIN LOGGING AT Wed Sep 06 03:00:01 2017 Sep 06 07:38:34 Hi, I have a recipe "webinterface" which creates 3 packages, webinterface, webinterface-www, webinterface-extra Sep 06 07:38:56 Is is possible to change local recipe variable from local.conf? Sep 06 07:39:07 in my image I would like to add webinterface, and webinterface-extra but not webinterface-www Sep 06 07:41:42 catch22, yes : with the variable name, and did you ever tried to fix a package version ? SRCREV_pn-glibc = "b8079dd0d360648e4e8de48656c5c38972621072" Sep 06 07:42:30 Yeah I seen examples like this, but can't seen anything in the docu that talks about it as a general mecanhism Sep 06 07:42:44 Does this work for all local variables? Sep 06 07:44:28 catch22, I have variables in my custom recipes to handle global URL and branches (to change branch or tag for a build for every recipe) Sep 06 07:45:13 aurele: yes this would be handy Sep 06 07:45:39 aurele: I'll go play, thanks for the hint Sep 06 07:47:02 catch22, I never tried to play with "_pn-package_name" with a custom variable Sep 06 07:48:24 catch22, I'm just a user, maybe there is more documentations and hint in bitbake or openembedded documentation but I never needed to read those ;) Sep 06 09:36:18 aurele: appreciate the help Sep 06 10:29:23 are there any way standard way of checking if an override is set? Sep 06 10:31:11 s/any way standard/any stadnard way/ Sep 06 10:44:42 who should I assign/CC issues wrt git.yoctoproject.org (the git service itself)? Sep 06 10:45:46 to my surprise bug 12040 seems to be reproducable Sep 06 10:45:47 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=12040 normal, Undecided, ---, jeffrey.osier-mixon, NEW , git server git.yoctoproject.org hung up unexpectedly after 1.00GiB Sep 06 10:56:11 jku: reassign to halstead Sep 06 10:57:20 thanks Sep 06 11:12:58 RP: hey, I think rburton was just talking with you about bug #139 and the on-target GCC not inheriting the same defaults/tune/etc as the cross & SDK compilers Sep 06 11:12:59 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=139 enhancement, Low, Future, raj.khem, NEW , Set default gcc instruction generation to TARGET_ARCH on target Sep 06 11:13:42 RP: it's causing a bit of grief with the freedesktop.org flatpak SDK, which (for better or worse) uses the target GCC Sep 06 11:14:05 we're specific to GCC 6.2 at the moment - is there a way we could bodge this with a .bbappend or something? :) Sep 06 11:17:01 ramcq: I'm just looking at the gcc code to try and figure that out Sep 06 11:17:20 Heyho, I have managed to use pkgconfig in my layer. However, Im wondering whats the best workflow for my application developer as he doesnt gets in touch with recipes. We ware using CMake. It is recommended to use different CMakes and project structures for application developer and yocto integration? Sep 06 11:17:22 ramcq: I think setting TARGET_CPU_DEFAULT and mabe TARGET_FPU in the gcc recipe would likely help Sep 06 11:17:55 ramcq: its slightly complicated by the fact that the gcc sources are shared... Sep 06 11:34:23 RP: TARGET_CPU_DEFAULT seems not to be used by the gcc recipes - might be TARGET_ARCH? where are the tunes expanded into their different bits? Sep 06 11:37:58 ramcq: its used in the gcc sources Sep 06 11:38:13 ramcq: just trying to figure out a magic incantation Sep 06 11:39:20 RP: ah, ok - cool. thanks! Sep 06 11:39:22 beer++ Sep 06 11:50:35 ramcq: still working on it sorry, gcc is convoluted :/ Sep 06 11:50:42 no worries Sep 06 11:51:11 this is a good thing to catch and fix as it means we kinda have to rebuild all of the armhf binaries we ever built for flatpak so far Sep 06 12:05:36 ramcq: care to try an experiment? Sep 06 12:07:10 ramcq: I think you may need to patch gcc/config.gcc in the sources to comment out the "target_cpu_default=" line, then add export target_cpu_default = "arm946e-s" to gcc-target.inc Sep 06 12:07:27 RP: I can, but I have to forewarn that, I've never built this, and I dont have access to the box that Alex builds them on, and he's off sick today it seems Sep 06 12:07:42 ramcq: replace arm946e-s with a suitable armv7 machine you want to default to Sep 06 12:08:25 ramcq: as far as I can tell, the above tweaks gave me a gcc that would have the right target settings for the qemuarm machine I targetted Sep 06 12:08:42 so it might be just as quick / quicker for you to check out freedesktop-sdk-base and built it on one of the 128-core magical mystery boxes you no doubt have handy...? :) Sep 06 12:08:59 ramcq: not sure if you might have to do something with the armhf piece too, depends if it was neon or hf causing you issues Sep 06 12:09:24 ramcq: I'm nested X deep in other issues :( Sep 06 12:17:28 ramcq: the other option may be --with-arch= and/or --with-cpu options to gcc configure. I'm not 100% sure what that would do with the output... Sep 06 12:17:53 ramcq: couple of things to try anyway... Sep 06 12:27:46 khem: have you ever tried cross-localedef-native/2.25-r0 on host system with glibc-2.26? Sep 06 12:29:47 khem: it's choking on it's own locale_t https://pastebin.com/udCzFA9L but more difficult to resolve here when it mixes host's headers from glibc-2.26 with glibc-2.25 code Sep 06 12:32:15 I'll probably use LOCALE_GENERATION_WITH_CROSS-LOCALEDEF to bypass this, but it there is a plan to support Pyro on more recent hosts than this should be resolved as well Sep 06 12:32:38 toghether with backport of icu and qemu-native fix from master Sep 06 14:22:02 Does anybody know what package includes the modinfo utility? Sep 06 14:25:12 ansierch: try kmod? Sep 06 14:25:40 aratiu: kmod was it, thanks much! Sep 06 14:25:49 $ oe-pkgdata-util find-path \*/modinfo* Sep 06 14:25:50 ... Sep 06 14:25:53 kmod: /sbin/modinfo.kmod Sep 06 15:03:21 RP: hey, so I'm getting access to an ARM box to try the build, because this bit of the SDK at least has only ever been tested native->native Sep 06 15:03:38 (maybe it would work as cross? maybe not? I don't want to find out now) Sep 06 15:04:07 RP: is it possible to influence that variable with a bbappend or is that "too late"? yocto is incorporated as a submodule so patching it is, er... tricky Sep 06 15:25:04 Hi all. I've added gtest as a dependency to my recipe in order to get gtest-native in my SDK. However looking at my host.manifest it's not listed in there. It's been aeons since i've last worked with bitbake so I'd be glad if someone had a hint for me. Sep 06 15:39:37 ramcq: yes, just set the export in the bbappend Sep 06 15:39:47 RP: ok, cool Sep 06 16:00:27 RP: I've noticed that you have just pushed few changes to morty to fix build on glibc-2.26, at least the qemu-native one is missing in pyro, should I send pull-request or is it already staged somewhere? Sep 06 16:02:58 JaMa: I was just about to look at that, needed to get morty running as a build and that was the last issue Sep 06 16:04:08 RP: cross-localedef-native works for you on that host? Sep 06 16:04:48 JaMa: it seemed to? Sep 06 16:07:21 RP: ok, pull request for pyro sent Sep 06 16:07:53 JaMa: ah, I just fixed it! Sep 06 16:08:05 for pyro as well? Sep 06 16:08:10 JaMa: yes Sep 06 16:14:15 JaMa: I've merged the other patches but there is a difference in that I was only tweaking them to work with native glibc newer versions Sep 06 16:15:13 same here Sep 06 16:15:56 JaMa: interesting. I guess there is native usage in meta-oe which core doesn't trigger :/ Sep 06 16:17:21 there is still that cross-localedef-native issue, I'm surprised you're not seeing it, but maybe you don't have binary locale generation enabled? Sep 06 16:17:39 but just building cross-localedef-native explicitly should reproduce it Sep 06 16:20:28 JaMa: we should be using binary locales... Sep 06 16:20:36 JaMa: I'll keep an eye open for it... Sep 06 16:20:54 armpit: I've fired a 2.2.2 build, I figured out fixes for qemu and python-numpy Sep 06 16:21:13 RP: what host distro is this? Sep 06 16:21:16 armpit: I've also ensured the patches are in pyro (and some from JaMa) Sep 06 16:21:23 JaMa: opensuse tumbleweed Sep 06 16:23:51 RP, k, thanks Sep 06 16:26:44 * armpit time to rebase Sep 06 16:47:27 hi Sep 06 16:47:31 anyone online? Sep 06 16:50:09 hi Sep 06 16:51:47 I am trying to get wxpython working. It compiles successfuly after days of work, but I have to run 'python setup.py install' so it's seen by python as well. Is there a command or something that I can run or should i just run python setup.py install? Sep 06 17:30:20 ioanb7: always check the layer index: https://github.com/bmwcarit/meta-ros/blob/master/recipes-extended/wxwidgets/wxpython_2.9.5.0.bb Sep 06 17:30:56 that's broken Sep 06 17:30:59 i fixed it Sep 06 17:31:07 i fixed wxwidgets Sep 06 17:31:11 and now im trying to fix wxpython Sep 06 17:31:13 im one step away Sep 06 17:31:26 you need to install, yes. the distutils class will do that for you Sep 06 17:31:33 how Sep 06 17:31:38 because it doesn't do that currently Sep 06 17:32:34 should i just do ${STAGING_BINDIR}/python install ....wxpythonfolder.py ? Sep 06 17:32:37 is that gonna work? Sep 06 17:34:57 https://hastebin.com/unolebibin.bash Sep 06 17:35:01 this is my code so far Sep 06 17:47:02 i dont actually know where i would have to call python setup.py Sep 06 17:47:49 the library compiles the C++ bit of it in about 4-5 minutes, though. so it does something Sep 06 17:47:56 the recipe * Sep 06 18:00:49 Does anyone here have experience managing system power states? Specifically handling loss of power or low-power conditions? Sep 06 18:01:56 I have an application where the system will regularly have power cut with no warning. Super caps will provide about 4 seconds of power after the main goes dark. Sep 06 18:02:45 I think I could just have a process register for SIGPWR and then just call shutdown with -h now. Sep 06 18:03:08 Am I in the weeds on that? Sep 06 18:04:06 xthunderheartx: https://designs.apertis.org/latest/robustness.html might have some stuff? Sep 06 18:04:19 it's about system reliabiliity in an automotive IVI system Sep 06 18:06:26 Thx ramcq. Very apropos as that's essentially our problem space. Data collection on HDX vehicles. Sep 06 18:09:09 can somebody give me an example of python library using distutils automatically running setup.py ? :) Sep 06 18:18:16 ioanb7: just use the distutils or setuptools classes (depending on whether the pacakge uses distutils or setuptools) Sep 06 18:42:54 rburton, RP: so, er, can you .bbappend an .inc? Sep 06 18:43:05 or should I .bbappend the real .bb that the .inc goes into Sep 06 18:57:38 ramcq: the latter Sep 06 19:01:28 rburton: so RP told me to patch a config.gcc file I don't have :O Sep 06 19:48:07 if a recipe has a "_a.b.c.bb" version and a "_git.bb" version and the "_git" version is 'DEFAULT_PREFERENCE = "-1"', how do I request the "_git.bb" version anyway? 'PREFERRED_VERSION_$pn = "git"'? Sep 06 19:49:00 you'll need to fiddle the version to be what the git recipe sets PV to, but yes Sep 06 19:49:06 you can use % as a wildcard Sep 06 19:49:21 ie PREFERRED_VERSION_glib-2.0 = "2.51.0+git%" Sep 06 19:49:29 (for the git version of glib that i have locally) Sep 06 19:50:20 rburton: perfect, thanks! :-) Sep 06 19:56:30 ramcq: gcc/config.gcc ? Sep 06 19:58:44 I have a recipe for a libA and a recipe for BinaryB that uses libA. However, I get undefined reference error when linking the binary. Im using CMake and pkgconfig. The include files are in libA-dev package. Is this working or do I have to put the include files in libA package? Sep 06 20:04:41 FabKna: if the issue is at link time, include files aren't the issue Sep 06 20:05:12 FabKna: you may be missing a -l or the -l options may be in the wrong order perhaps Sep 06 20:05:38 bluelightning: Ok, I will look in my target_link_libraries than. Sep 06 20:12:25 bluelightning: Dump Question: I only have to link the binary not libA if libA depends on libC? So in the end I link binaryB with libA and libC? Sep 06 20:13:13 FabKna: I think both have to be linked but I'm not too sure to be honest Sep 06 20:13:43 this is where pkg-config can be helpful since it gives you the linker arguments needed for a library Sep 06 20:14:51 Yeah I use the libA_LIBRARIES variable but the pkgconfig does not help with the order? Sep 06 20:45:21 bluelightning: It was indeed wrong ordering. FYI: linking libraries in CMake is not needed but possible and sometimes recommended as explained here: http://floooh.github.io/2016/01/12/cmake-dependency-juggling.html Sep 06 21:42:12 Hello - with qtwebkit embedded, and qtbase-fonts no longer being a thing, how can I determine what fonts qtwebkit is looking for? I've installed ttf-dejavu-sans, and running fc-match on any CSS rule on my web page returns it. Sep 06 21:42:43 (This broke switching from Krogoth to Morty) Sep 06 21:46:34 Hey all - Ive got a python-xxx.bb recipe I've added based off another (simple) recipe. There's nothing clever beyond SRC_URI= S= and inherit setuptools. However it appears to run setuptools locally on the dev machine to install into the rootfs which then fails as it inherits smbus which isn't on my dev machine. I'm not sure what the "proper" way to do this is? Can I disable checks like this (I cant see a way) or should I just bodge the recipe to copy files to inte Sep 06 21:46:34 ded locations and leave it at that? Sep 06 23:07:30 yocto linux ! Sep 06 23:07:31 woooo Sep 06 23:07:52 does anyone know how to set the proper mteric for new routes in yocto ? Sep 06 23:08:07 all or my interfaces get a metric of 0 Sep 06 23:08:23 I'd like to make these higher because I'm dealing with routing problems Sep 07 00:10:15 theorem: are you using systemd to manage the i/f Sep 07 01:10:57 khem: I was managing it with ifconfig directly and the route command Sep 07 01:11:09 are there tools in systemd that allow me to change the route metrics ? Sep 07 01:51:41 theorem: then you should be able to change the /etc/network/interfaces file Sep 07 01:52:29 you should try the changes on device and then tranfer the changes to metadata may be stick in a new interfaces file via a bbappend Sep 07 01:52:57 hm Sep 07 01:53:03 so, what if those files don't exist ? Sep 07 01:53:39 Linux FZEDC713D00BKJ501 3.10.17-poky-edison+ #1 SMP PREEMPT Fri Jun 19 12:06:40 CEST 2015 i686 GNU/Linux Sep 07 01:53:57 check ./meta/recipes-core/init-ifupdown/ Sep 07 01:54:17 where is the relative ./meta ? Sep 07 01:54:26 thats in yocto metadata Sep 07 01:54:34 path ? Sep 07 01:54:35 so how is your interface brought up ? Sep 07 01:54:43 it in poky dir Sep 07 01:54:57 er, can I get a full path ? Sep 07 01:54:58 is it using ifup/ifdown or something else Sep 07 01:55:15 /sbin/ifup Sep 07 01:56:00 ok then it mus be using init-ifupdown Sep 07 01:56:19 so add the interfaces files and define your n/w if params Sep 07 01:56:41 I have nothing by that name on my filesystem Sep 07 01:57:41 this seems to be close to busybox Sep 07 01:57:48 how does busybox get interfaces online ? Sep 07 02:00:26 it looks like wifi is brought online with wpa_supplicant_wlan0_event.service Sep 07 02:00:32 the systemd service Sep 07 02:01:40 there's also systemd-networkd.service Sep 07 02:30:44 Anybody knows whether it is possible to add something to do_image_myimage[prefuncs]? If yes, how? Sep 07 02:31:15 It seems not to work when I just add Sep 07 02:31:35 do_image_myimage[prefuncs] =+ "myfunc" Sep 07 02:31:53 Presumably because do_image_myimage gets generated dynamically in image.bbclass... Sep 07 02:31:59 theorem: that means you are using systemd Sep 07 02:39:23 sagner: master fixed that recently Sep 07 02:39:28 yes Sep 07 02:39:33 sagner: long standing issue where it set rather than appended/prepended Sep 07 02:39:38 so, it took me a while to figure that out Sep 07 02:39:49 but the wireless itnerface is not enumerated through systemd Sep 07 02:40:16 there's no "wireless" section in /etc/systemd/network as far as I can tell. Sep 07 02:41:55 kergoth: hm I see, it uses now appendVarFlag... Can I work around that easily? Sep 07 02:42:03 generally you need wpa-supplicant to set up wireless interfaces Sep 07 02:42:25 sagner: if it uses appendVarFlag, then it won't override what's already there, and yours should be fine Sep 07 02:43:08 kergoth: working on morty, so appendVarFlag is not there yet.. Just noticed appendVarFlag in image.bbclass in master... Sep 07 02:44:05 no good way to work around it short of overridding the class, afaik, but could be missing something Sep 07 02:46:45 kergoth: I was thinking to just call the function in my image shell function, but afaik there is no way to call python from shell. Is there a way to create a image format in python? Sep 07 02:47:24 IMAGE_CMD_... **** ENDING LOGGING AT Thu Sep 07 03:00:00 2017