**** BEGIN LOGGING AT Sat Feb 13 02:59:58 2016 Feb 13 07:07:47 Hey, WHen I set a cmake flag in my recipe, is it possible for me to confirm if it worked? Or in other words, I just want to see the cmake output. Feb 13 08:12:07 Hi. Is there a command to list QA warnings? Feb 13 08:18:45 cat tmp-eglibc/qa.log Feb 13 08:19:09 rtr_: but not all QA warnings are listed there Feb 13 09:36:06 @JaMa Thanks Feb 13 09:52:48 Hello, I am trying to compile opencv with cuda support. So what I did is to create my own layer with higher priority, copy opencv recipe from meta-oe and added -DWITH_CUDA=ON to EXTRA_OECMAKE. But the opencv that gets compiled doesnot have cuda support Feb 13 09:53:14 I am asked to set the WITH_CUDA=ON cmake flag. Feb 13 10:10:27 I tried EXTRA_OECMAKE = "-DWITH_CUDA=ON" and got this error: CMake Error: The source directory "/run/media/fury/storage/yocto/poky/build/tmp/work/cortexa15hf-vfp-neon-poky-linux-gnueabi/cudacv/3.0+gitAUTOINC+424c2bddb3-r0/build/-D WITH_CUDA=ON" does not exist. while do_configure Feb 13 10:39:15 issue solved Feb 13 10:53:45 How do I rebuild all the dependencies along with the main recipe? Feb 13 10:54:51 why would you want that? Feb 13 10:55:45 you can delete tmpdir and sstate-cache Feb 13 10:59:16 @JaMa One of the recipes is failing when compiled individually but passing when other recipe which depends on it is compiled Feb 13 10:59:34 Just want to recompile everything in that line Feb 13 11:08:48 maybe it's trying to use some file from WORKDIR of the dependencies, so just fix the recipe Feb 13 11:14:56 @JaMa will removing sstate-cache remove everything? Feb 13 11:15:12 All the recipes get rebuilt from scratch? Feb 13 11:16:29 hello, anyone have any idea whta DEBUG: Removing manifest: in log.do_compile mean? Will it be a problem? Feb 13 11:18:33 rtr_: yes Feb 13 12:26:51 Hello, this is my first recipe. It is a simple one which installs a binary from a .deb file. Can anyone please take a look and identify the mistakes? Feb 13 12:26:52 https://paste.ubuntu.com/15035420/ Feb 13 12:43:28 why is the package not being created? Feb 13 13:10:35 If I am writing a recipe to install binaries from a .deb file, do I have to write the do fetch? Feb 13 13:36:48 I am trying to write a simple recipe that installs a few files to /usr from a .deb file. This is the recipe I wrote: https://paste.ubuntu.com/15035766/ and this is the error I get: https://paste.ubuntu.com/15035771/ Will somebody please point out what is wrong? Feb 13 14:12:12 hey, package is there in build/temp/deploy/deb, so why is it not getting installed? Feb 13 14:20:49 yay! I found a clue!!!' Feb 13 14:21:47 So -dev and -dbg packages are boing installed but not the actual package itself. And I can't install the -dev package because it lists the original package as a dependancy Feb 13 14:45:41 I want to package whatever is in my ${WORKDIR}/output but doing a FILES_${PN}-dev = "${WORKDIR}/output" didnot work. Feb 13 14:48:59 where is everybody? Feb 13 14:52:02 how do you install the package? Is it install in a image recipe or in local.conf? Feb 13 14:52:45 there is a big difference of the package being build and the package being isntalled Feb 13 14:53:10 aravinPV: Feb 13 14:53:35 local.conf Feb 13 14:54:51 The package being build does not have the files I want to be installed. I have opened them and checked, fl0v0 Feb 13 14:55:50 you need to add the files in the recipe Feb 13 14:56:55 FILES_${PN} += " " Feb 13 14:57:05 have you done that? Feb 13 14:57:16 This is the part where I am confused. Feb 13 14:57:40 I have tried many variations of the above said thing, but none worked Feb 13 14:58:33 there are many recipes who do this, maybe just 'grep -r FILES_' in your source folder Feb 13 14:58:33 my ${WORKDIR}/output has "/usr/..." so how should FILES_${PN} look Feb 13 14:59:09 you dont use hard coded pathnames Feb 13 14:59:53 see here: http://docs.openembedded.ru/directories_installation.html Feb 13 14:59:56 Why is FILES_${PN}-dev = "/" should do the job, right? Feb 13 15:01:52 fl0v0: Ok, I am tryinh FILES_${PN}-dev = "${prefix}" Feb 13 15:01:53 are we talking of the recipe above? https://paste.ubuntu.com/15035766/ dont use those pathes Feb 13 15:02:14 normaly you do not install anything in your root Feb 13 15:02:21 what are you trying to do? Feb 13 15:02:28 In desperation I tried those. Feb 13 15:02:41 I will post my updated recipe Feb 13 15:03:33 https://paste.ubuntu.com/15036462/ Feb 13 15:04:08 I have some binaries from nvidia in the .deb format. Feb 13 15:04:15 I want to install them Feb 13 15:05:39 ok idnt really now how one handles precompiled packages Feb 13 15:05:55 maybe there is a class for this Feb 13 15:06:35 SImply tell me how to do this. Whatever I want to package is present in ${WORKDIR}/output. SO how will my FILE variable look like? Feb 13 15:08:04 fl0v0: Feb 13 15:09:16 is there a source distribution of the software you want to install? this would it make easer for me Feb 13 15:10:04 if not Feb 13 15:10:28 the simples way to do it right now would be set the package management of your build to deb Feb 13 15:10:41 than copy the file tou your build image and try to install it Feb 13 15:11:01 no they are proprietary binaries whose source I don't have Feb 13 15:11:19 kk i dont have experience of how to ship blobs Feb 13 15:12:07 I am kinda stuck. I have tried what you said. It works like a charm. But this thing is a depandancy for something else that want to build. Feb 13 15:13:55 and the irony is that I am stuck with something so simple! Feb 13 15:14:27 look if you can find an installer for deb packages Feb 13 15:15:02 what do you mean by installer? Feb 13 15:15:02 i dont know if tsometjhing like that exists Feb 13 15:15:19 on yocto? Feb 13 15:15:31 then you could use 'inherit deb'in your recipe Feb 13 15:16:03 bin_package exists Feb 13 15:16:05 http://yocto.yoctoproject.narkive.com/yoITgHc1/installing-deb-file-in-build Feb 13 15:17:20 yeah that sounds good Feb 13 15:17:53 look at classes/bin_package.bbclass thats a good advice Feb 13 15:18:47 ok you need to extract the files in the deb Feb 13 15:18:57 Hi i cant remember the name of the network manager made by intel (for embedded use) for yocto :D Feb 13 15:19:03 can somebody help refresh my mind Feb 13 15:19:17 was it connman ? Feb 13 15:19:27 yeah I think so Feb 13 15:19:45 https://01.org/connman Feb 13 15:19:48 its this one Feb 13 15:20:03 is it not done by base.class? I see them in ${WORKDIR}/output Feb 13 15:20:22 it is already extracted to ${WORKDIR}/output Feb 13 15:20:29 ah ok Feb 13 15:20:39 so maybe set S to ${WORKDIR}/output Feb 13 15:20:48 S = "${WORKDIR}/output" Feb 13 15:21:05 then also inherit from bin_package3 Feb 13 15:21:09 -3 Feb 13 15:21:54 then base class should extract your package and bin_package installs it Feb 13 15:22:03 but Feb 13 15:22:04 moment Feb 13 15:22:09 it uses tar Feb 13 15:22:52 so you need a tar archive with the binaries Feb 13 15:23:52 I added the S parameter and something happened Feb 13 15:24:06 ok :) Feb 13 15:24:16 do_package_qa failed Feb 13 15:24:32 can you pastebin your current recipe? Feb 13 15:25:33 wait a sec Feb 13 15:25:50 https://paste.ubuntu.com/15036633/ Feb 13 15:27:08 https://paste.ubuntu.com/15036639/ -> looks like this contains something Feb 13 15:27:57 ok Feb 13 15:29:21 hmm now you dont need FILES_${PN} = "/*" in the recipe anymore Feb 13 15:29:32 but thats not the problem Feb 13 15:29:49 i think Feb 13 15:30:28 hm the recipes looks good to me Feb 13 15:31:56 :-( Feb 13 15:33:45 I silenced the "already stripped" QA errror by adding INSANE_SKIP_${PN} = "already-stripped" to my recipe Feb 13 15:35:30 does it work? Feb 13 15:36:05 !!YES!! Feb 13 15:36:15 ncie Feb 13 15:36:34 well it builds :) Feb 13 15:36:34 thanks a lot! Feb 13 15:36:45 so the problem was that S was not set Feb 13 15:36:53 ok Feb 13 15:37:03 for bin_package to work on Feb 13 18:27:50 CUDA is a build dependency for OpenCV when it is compiled with CUDA support. CUDA binaries are available from nvidia as .deb files. So what I did was to create a recipe that unpacks this deb and install the binaries. Then I added this as a build dependency(DEPENDS) for opencv. Also I set the WITH_CUDA=ON flag for opencv's cmake. But still cmake output (from do_configure logs) indicate that CUDA support has not been turned on. This Feb 13 18:28:13 has failed to locate the binaries. What possibly when wrong? How to fix it? Feb 13 18:36:28 anyone else seeing perl build failures? 5.3.0/ld: cannot find -lssp_nonshared Feb 13 18:39:17 Also where is the staging area where I can inspect? Feb 13 19:02:52 JaMa: i did, but in another context... Feb 13 19:03:44 nerdboy: ty, I have a fix already, it was something in my config Feb 13 19:04:22 we have bbclass preventing staging static libraries unless explicitly whitelisted (to prevent licensing issues) Feb 13 19:04:26 are you even building with ssp? Feb 13 19:04:45 and new perl seems to be first recipe which require libssp_nonshared.a Feb 13 19:05:45 perl needs a diaper... Feb 13 20:59:40 I am getting this weird issue Feb 13 20:59:48 I have this warning: Feb 13 20:59:50 cuda-npp-6-5: found library in wrong location: /usr/local/cuda-6.5/targets/armv7-linux-gnueabihf/lib/libnppi.so.6.5 Feb 13 21:00:37 cuda-npp-6-5 is a build dep for opencv, but it is not there in the sysroot! Feb 13 21:00:48 even after there is this warning. Feb 13 21:01:28 in short, there is no /usr/local/cuda-6.5/targets/armv7-linux-gnueabihf/lib/libnppi.so.6.5 in the sysroot Feb 13 21:01:36 I am puzzled Feb 13 21:17:11 But they appear in the correct position in th final image, **** ENDING LOGGING AT Sun Feb 14 02:59:58 2016