**** BEGIN LOGGING AT Thu Oct 06 02:59:58 2016 Oct 06 08:08:15 when generating an image, does bitbake(?) populate the root_fs using the previously generated RPMs (or whatever is used as package format)? Oct 06 08:12:08 could somebody help me with this error: http://dpaste.com/0SKAR6N ? Oct 06 08:13:12 0x4: you're missing the layer that r(untime)provides the 'x11' package Oct 06 08:14:23 something like 'grep -R RPROVIDES*x11* *' within the directory that contains all your meta-* folders (layers) should help you determine the layer that provides it. Oct 06 08:14:39 then put that layer into your bblayers.conf and retry :) Oct 06 08:15:25 unless it was you who added the x11 dependency to the arago image recipe. Then you might have picked the wrong dependency name :) Oct 06 08:18:33 T_UNIX: yup, the RPMs are simply installed into the image Oct 06 08:18:53 Ulfalizer: thank you :) Oct 06 08:19:00 np Oct 06 08:20:25 re. the (?), much of the behavior of the standard build task is implemented in bbclasses that are inherited globally rather than being something built into bitbake itself Oct 06 08:20:43 bitbake + standard metadata like that is what the manual calls the "openembedded build system" Oct 06 08:21:33 http://www.yoctoproject.org/docs/2.2/ref-manual/ref-manual.html#usingpoky-debugging-viewing-variable-values has a note on how to see what classes are inherited by default, near the end Oct 06 08:21:47 the more you know :) Oct 06 08:22:21 the 2.2 version of the manuals has a bunch of new stuff :) Oct 06 08:22:50 unfortunately I'm stuck with dylan for now -.-' Oct 06 08:23:14 most of it also applies to older versions Oct 06 08:35:08 *standard build tasks Oct 06 09:01:43 T_UNIX: grep shows me nothing about x11 :-( Oct 06 09:32:22 what is the correct approach: append IMAGE_INSTALL += " blah" to the layer image file, or to the build/conf/local.conf? Oct 06 09:35:30 Ox4: sorry, network was down. So then it seems to me that either the grep expression is wrong :D or you're missing the corresponding layer ;-) Oct 06 09:51:36 hey all Oct 06 09:51:58 I have some software that I download via git, but can't be downloaded anonymously Oct 06 09:52:10 moreover my SSH key is protecte by a password. Oct 06 09:52:33 I use ssh-agent to keep my key unlocked, but somehow bitbake seems to loos that info and can't download the git sources Oct 06 09:53:01 (doing a git fetch in the same shell as bitbake correctly uses the ssh agent and downloads whatever needs to be downloaded) Oct 06 09:53:11 is there a workaround for that ? Oct 06 10:07:10 I can't just add BBCLASSEXTEND="native" to get a native package from a recipe, can I? Oct 06 10:23:49 No real function for mknod: /home/jrosen/projets/nexter-lafp/lafp-os/lafp-toradex-wayland-head/tmp/sysroots/x86_64-linux/usr/bin/../lib/pseudo/lib64/libpseudo.so: undefined symbol: mknod Oct 06 10:23:49 ^^^has anybody encountered that ? (i'm using jethro) Oct 06 11:02:28 yeah, fixed in master and discussed on yocto@ recently Oct 06 11:02:57 thread is https://lists.yoctoproject.org/pipermail/yocto/2016-October/032238.html Oct 06 11:03:33 Fix is http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=e090775f7e3b4af2345e835bbebcd5d4a580798b Oct 06 11:13:19 joshuagl: ok, thx a lot Oct 06 11:14:20 boucman_work: if you backport that to jethro it's probably worth submitting to the list Oct 06 11:14:46 yeah, good point Oct 06 11:16:20 If a recipe generates host tools in addition to the target files, how should that be split out in a recipe? Oct 06 11:17:31 BBCLASSEXTEND="native" really only adds the ability to build the whole recipe once more for the -native target, doesn't it? Oct 06 11:20:03 sveinse: yes, but you can make a recipe that generates two packages, the native and the non-natvie package, and deal with the details at pkg-split time Oct 06 11:20:10 I think that would work... Oct 06 11:20:54 boucman_work: just by naming convention? e.g. canfestival-objdictgen-native Oct 06 11:21:04 i'm not sure... Oct 06 11:24:34 not sure either. Just tested it, and it generates canfestival-objdictgen-native_*_armel.deb, which is not correct. Oct 06 11:25:54 Using BBCLASSEXTEND="native" does work, thou. But it checks out and rebuilds the whole shebang once more for native. Seems superfluous where there's only a small part of the entire recipe needed for native Oct 06 13:09:03 In yocto devshell neovim is not working. where can i find more info on why is not working ? Oct 06 13:38:55 could somebody take a look at this error: http://dpaste.com/0A5STM3 ? Oct 06 14:00:05 Hi. Does anybody know how IMGDEPLOYDIR gets its value? Or is it something I need to set manually? I am trying to create an image including meta-raspberrypi but the SD image creation seems to fail as it is trying to write the image to my root directory i.e. IMGDEPLOYDIR is empty in the beginning of the path. Oct 06 14:37:35 muppe: try to use "grep" to understand how things are being set Oct 06 14:43:37 Does yocto have a mechanism for sharing the same SRC_URI across multiple recipes? E.g. our project repo, which contains the code for multiple recipes/outputs. I'd rather avoid having to check out the repo many times Oct 06 14:44:21 easiest way, if there are multiple recipes int he same directory sharing.. Oct 06 14:44:38 wait I missread.. Oct 06 14:44:56 If you have identical sources on multiple recipes, it will only be downloaded once -- but will be extracted multiple times Oct 06 14:45:23 there are ways to even share the extracted sources -- but it takes a lot of care to ensure you don't break things -- so it's generally not recommended.. Oct 06 14:45:42 gcc and the kernel are the two cases where we have shared source use... but it can be done elsewhere, but again care is needed.. Oct 06 14:46:17 easiest way is to follow the gcc model -- have a 'source' package that does nothing but fetches, and patches the sources.. then the other recipes depend on that and use it from the shared installation location.. avoiding any local patching or other changes Oct 06 15:03:31 So, I found why nothing PROVIDES lib32-my-app; it's because my LICENSE_FLAGS_WHITELIST="commercial_my-app" not "commercial_lib32-my-app" Oct 06 15:09:31 hi guys Oct 06 15:09:49 I am trying to overcome a "fatal error: sys/types.h: No such file or directory" Oct 06 15:10:33 I am using a custom do_compile() Oct 06 15:10:46 and am basically running a script with some arguments Oct 06 15:11:03 one or the arguments being the toolchain prefix to which the script adds "gcc" Oct 06 15:11:17 so it basically gets the cross gcc and invokes it on a .c file Oct 06 15:11:38 problem is that it does not pick-ul the sysroot, hence the error about not finding sys/types.h Oct 06 15:11:53 how would I go about adding the sysroot when not using oe_runmake? Oct 06 15:16:56 So, my issue seems to be an issue with license name expansion with multilib. Oct 06 15:36:36 gartin: may be your application overrides CC and CXX and family of make variables in its Makefiles ? Oct 06 16:02:02 guys, I don't see symlink to my latest image build. How can I generate it automatically? Oct 06 16:08:36 khem, yeah, that was it Oct 06 16:08:41 thanks for the pointer Oct 06 16:32:47 Is there an alternative to ${PN} that doesn't get ${MLPREFIX} added to it automatically? Oct 06 16:32:59 BPN Oct 06 16:33:37 Perfect, thank you. Oct 06 16:33:59 I guess my INITRD_IMAGE = line needed BPN-initramfs rather than PN-initramfs Oct 06 17:40:00 moto-timo: ping Oct 06 17:40:09 denix: pong Oct 06 17:40:26 moto-timo: is there are recipe for pybind11? Oct 06 17:40:35 a recipe Oct 06 17:41:07 I don't see one in layer index. Oct 06 17:41:45 yeah, I already checked that... thought maybe there was a different name or something Oct 06 17:42:42 denix: sounds useful :) Oct 06 17:42:53 denix: new to me, so I'm not much help Oct 06 17:43:41 moto-timo: ok, thanks :) I'll look into it Oct 06 17:44:08 denix: let me know if you have problems Oct 06 17:44:49 moto-timo: sure, thanks! Oct 06 18:15:14 hi guys, having an issue where qtquickcontrols-qmlplugins is not available. Anyone know how to turn it on? I'm using jethro, meta-qt5, qt version is 5.5 Oct 06 18:15:35 read something in an old forum post about PACKAGECONFIG_append_pn-qtbase = " accessibility" would that do it? Oct 06 18:20:51 I've tried including qtdeclarative-qmlplugins and qtquickcontrols-qmlplugins in my image but I get this error: "cannot satisfy the following dependencies: qtquickcontrols-qmlplugins" Oct 06 19:24:05 How can I build 32 and 64-bit kernels against my 32-bit-only userspace? Oct 06 19:24:16 I'm using multilib, with lib32-my-image. Oct 06 19:24:41 bitbake lib32-linux-yocto doesn't seem to have built anything. Oct 06 19:28:43 Circuitsoft: 32bit vs 64bit kernel is selected based on the CONFIG_64BIT Kconfig value in the kernel .config Oct 06 19:30:04 right, and multilib explicitly does not apply to the kernel Oct 06 19:31:26 correct Oct 06 19:31:42 it will be the BSP selecting the kconfig via kernel config frags Oct 06 19:34:19 not sure if there is a good or documented way to build multiple kernels in a single build Oct 06 19:34:57 I would almost suggest you create a second build and use a customer layer with kernel scc and cfg frag to set disable CONFIG_64BIT Oct 06 19:35:13 use the second build to build the kernel and modules... Oct 06 19:37:07 RP you around, I have a question for you based on a finding I have with fray asking me to look at some bitbake performance stuff Oct 06 19:54:34 marka: I am, kind of Oct 06 19:54:49 no worries, if you are busy I can get you another time Oct 06 19:55:22 Are the items in TARGET_HOST_TASK package names or recipie names? (trying to get mkfs.ubifs to be included, and mtd-tools puts it in a seperate package) Oct 06 19:55:29 RP if you want to look as a distraction I want to ask about fn_from_tid() vs taskfn_fromtid() Oct 06 19:55:32 er Oct 06 19:55:35 from my take these are the same function Oct 06 19:55:36 TOOLCHAIN_HOST_TASK Oct 06 19:55:52 taskfn_fromtid() is just more complex Oct 06 19:56:53 fishey1: package names, in the nativesdk- namespace Oct 06 19:58:17 marka: they do look very similar don't they Oct 06 19:58:34 they do Oct 06 19:58:38 unless I am missing something Oct 06 19:58:44 which is entirely possible Oct 06 19:58:56 taskfn_fromtid() lacks "taste" as Linus would say Oct 06 19:59:04 since it has a secial case Oct 06 19:59:27 special even Oct 06 20:00:23 marka: also note that in nearly every usage, it calls split_tid before it Oct 06 20:00:37 RP yes, this is why I was asking Oct 06 20:00:44 as this is an optimization I am looking at Oct 06 20:01:03 marka: are the profiles showing slowdown in these functions? Oct 06 20:01:35 it is the dependency building recursive loop which I am focused on Oct 06 20:01:43 and they use the split and build funcs Oct 06 20:02:28 in the case I am using I was able to go from ~26,000,000 to ~23,000,000 function calls by cutting out the split in an inner loop Oct 06 20:02:53 marka: right, and that is worth something with python Oct 06 20:02:57 it is Oct 06 20:03:20 marka: I suspect this is an artefact of how I converted the code from the older IDs Oct 06 20:03:28 that was my guess Oct 06 20:03:45 I was just wanting to avoid wading in over my head Oct 06 20:04:01 marka: I think you're right and this function can be dropped Oct 06 20:04:18 it will be a nasty patch Oct 06 20:04:31 marka: why? Oct 06 20:04:36 since the variable names which use these for assignment are fn and task Oct 06 20:04:43 sorry taskfn Oct 06 20:04:58 I suppose it won't be that bad Oct 06 20:05:38 would be nice to have a variable name for "multiconfig:mc:fn" vs "fn" Oct 06 20:06:14 marka: right, its confusing Oct 06 20:06:21 marka: mcfn? Oct 06 20:06:44 marka: I can have a look at cleaning this up a bit if it helps Oct 06 20:07:05 ya, since I suspect you spend far more time in this code than I do it might be safer for you to make the change Oct 06 20:07:20 unless you want to be stuck with a stupid var name, which I am bound to give it Oct 06 20:07:34 task_prefix Oct 06 20:07:39 marka: let me finish off what I was doing and I'll have a look Oct 06 20:07:46 for sure, no rush Oct 06 20:07:52 marka: I might go a bit further than you would ;-) Oct 06 20:08:01 I will continue with my performance analysis Oct 06 20:08:12 this won't get in my way, so no rush Oct 06 20:08:32 agreed, if I was making the change I might be 'shy' Oct 06 20:08:46 and it is a good opportunity for cleanup while the rename is obvious Oct 06 20:09:34 * marka ducks back in to the shadows **** ENDING LOGGING AT Fri Oct 07 02:59:58 2016