**** BEGIN LOGGING AT Wed Aug 31 02:59:58 2016 Aug 31 03:24:49 Hi there, I need to set-up a continuous integration server and release server for a yocto project. Is there a standard way of doing this? Aug 31 03:25:25 Is the the autobuilder project widely used? Aug 31 07:35:49 if I set GLIBC_GENERATE_LOCALES = "en_US.utf-8" in local.conf i get this error: glibc-locale_2.22.bb, do_package) failed with exit code '1' Aug 31 07:40:15 What I want is to have default locale set to "en_US.utf-8" on image, but I haven't found a way yet - is possible ? Aug 31 07:41:08 I've tried setting GLIBC_GENERATE_LOCALES = "en_US.UTF-8" and set IMAGE_LINGUAS ?= "en-us" Aug 31 07:41:20 but nothing worked so far Aug 31 07:42:15 I'm using fsl image from variscite which is based on poky jethro Aug 31 08:14:32 Hi! I am trying to have a variable with overrides depending on the platform, like this: Aug 31 08:14:36 UBOOT_CONFIG_LOADADDR = "test_2" Aug 31 08:14:43 UBOOT_CONFIG_LOADADDR_platform-one = "0x12000000" Aug 31 08:14:48 UBOOT_CONFIG_LOADADDR_platform-two = "0x18000000" Aug 31 08:14:53 That works great for usual recipes (the ones which get crosscompiled) but it is not working for a native recipe Aug 31 08:14:58 how could I make that work for the native ? Aug 31 08:19:06 hi, Is there a way to determine the origin of all the packages present into my image. As example, libqtguie4 come from qt4-embedded? Aug 31 08:35:36 TuTizz: https://community.nxp.com/docs/DOC-94953 perhaps it can help Aug 31 08:36:22 TuTizz: you can start by Showing image's packages Aug 31 09:09:36 Hey guys, I've got a question. I'm building a core-image-minimal. Now I changed my SERIAL_CONSOLE from ttyS2 to ttyS1 in my machine config. I do a bitbake core-image-minimal again and this change is not picked up. Aug 31 09:10:03 How can I find the recipe that is responsible for building the inittab so I can force it to rebuild? Aug 31 09:12:26 isn't it a parameter for u-boot ? Aug 31 09:13:16 I thought it also influcend the serial port selection for the getty in the inittab Aug 31 09:14:20 SERIAL_CONSOLE specifies parameters for serial debug console Aug 31 09:14:49 and it's passed to the kernel by u-boot Aug 31 09:15:48 Any clue what causes the inittab to pick ttyS2 then? And how I can change this to ttyS1? Aug 31 09:17:26 i think sysvinit-inittab recipe is responsible for generating inittab Aug 31 09:17:28 meta/recipes-core/sysvinit/sysvinit-inittab_2.88dsf.bb: tmp="${SERIAL_CONSOLES}" Aug 31 09:18:22 Ahh thnx Aug 31 09:20:36 @jwessel what is the easiest and fast way of find this? Aug 31 09:20:54 grep -r SERIAL_CONSOLES meta Aug 31 09:20:56 @jwessel is it just experience or do you have some thriks Aug 31 09:21:09 how did you know it was SERIAL_CONSOLES Aug 31 09:21:19 instead of SERIAL_CONSOLE Aug 31 09:21:35 I meant SERIAL_CONSOLE Aug 31 09:21:46 meta/conf/bitbake.conf:SERIAL_CONSOLE ??= "" Aug 31 09:21:46 meta/conf/bitbake.conf:SERIAL_CONSOLES ??= "${@d.getVar('SERIAL_CONSOLE', True).replace(' ', ';')}" Aug 31 09:22:08 You can see the definition of serial consoles comes from the setup of bitbake.conf Aug 31 09:22:40 I know from experience the parse order of how the different conf files work. It is also documented in the bitbake manual I believe. Aug 31 09:26:41 I normaly try to debug that with -e option of bitbake Aug 31 09:26:50 I'll look it up, thnx for the help Aug 31 09:27:42 bitbake -e will tell you want you need for a single recipe. In this case you needed to find the recipe. Aug 31 09:29:06 I figured the SERIAL_CONSOLE was in a recipe or a bbclass file. Aug 31 09:29:41 Having done some work with serial ports before, the grub bits are found elsewhere as an example. Aug 31 09:29:48 meta/classes/grub-efi.bbclass:GRUB_SERIAL ?= "console=ttyS0,115200" Aug 31 09:30:10 I realize of course you are not using grub for your system. Aug 31 09:48:11 Why is MACHINEOVERRIDES="" for native recipes? Aug 31 09:48:28 I need to change the way a host tool is compiled depending on which machine it is going to be used with Aug 31 09:48:35 I know it is a bit of a corner case Aug 31 09:48:56 how could I somehow change the recipe depending on the machine of the current yocto project? Aug 31 09:49:07 (the recipe extends native and sdknative) Aug 31 10:31:25 Kakounet, thanks for your answer, this is exactly what I m doing (this is how I get libqtguie4) but now I m facing with a list of different packages and I don't know how to make corresponding them with the original source (qt_embedded4.8.5 for my example) Aug 31 10:47:50 @jwessel thnx for the help. Indeed I'm not using a grub based system. It's an embedded target Aug 31 11:40:06 Hi! I wonder if you can help. I have trying for a few weeks to solve an error with recipe "fatal error: new: No such file or directory." The source has "#include " Aug 31 11:55:16 Is there a C++ standard library DEPENDS that I need to include in a recipe for C++ project? Aug 31 11:58:04 How can i tell when the _append magic is in play, and when its not? E.g. two consecutive DISTRO_FEATURES_BACKFILL_CONSIDERED="" statements will the latter override the first, or does it behave in a append sort of way? Aug 31 12:43:17 hi Aug 31 12:43:38 I am seeing a build failure with util-linux-native Aug 31 12:43:39 ../util-linux-2.28.1/schedutils/chrt.c:88:17: error: ‘__NR_sched_setattr’ undeclared (first use in this function) Aug 31 12:43:53 Updating the build machine fixed the issue Aug 31 12:44:21 I've seen this on 3 different machines and it seems like a problem with the native kernel headers Aug 31 12:44:53 is there a way to fix this through yocto or is updating the build machine the only way around? Aug 31 13:08:11 has anyone else seen this error? Aug 31 13:43:13 i know this is wacky, but due to cmake limitations. I want to build a project using bitbake to build a component for the host. Then later use that output in a second recipe to rebuild for the target. Aug 31 13:43:48 is it possible to tell bitbake to build a component for the host or is it strictly for target builds? Aug 31 14:00:39 so what ive found is that there is a bitbake keyword BBCLASSEXTENDED="native" and a inherit native. It seems like to build a cmake project for native host, you do "inherit native cmake" Aug 31 14:04:17 yes, this seems to be on the right path. the build dir has moved from the cross compiler build dirs to the native dirs. Aug 31 14:10:26 Heey guys, I've been playing around with adding my own tasks in a bbappend. I see something really strange. addtask foo after do_unpack doesn't work addtask foo before do_compile works. addtask foo after do_unpack before do_compile works fine as well. Why? Aug 31 14:10:56 Can I only use after when a before is specified? Aug 31 14:15:50 flitjes: i'm unfamiliar with add task Aug 31 14:16:27 but i'm thinking if its some extra processing done in configure before normal configure, could you not just modify the configur step? Aug 31 14:16:43 flitjes: i am a noob, so keep that in mind. Aug 31 14:16:53 @johns_ ping Aug 31 14:17:37 flitjes: tasks will only get executed if you request their execution or something depends on them Aug 31 14:18:04 davis: Ok we are both noobs then :D Aug 31 14:18:06 flitjes: "addtask foo after do_unpack" means bitbake -c foo $recipe will work, and $recipe:do_unpack will have run Aug 31 14:18:52 "addtask foo before do_compile" means bitbake -c foo $recipe will work, but $recipe:foo may be run as the first task (it doesn't even have to unpack or fetch the sources for that) Aug 31 14:20:21 neverpanic: I exeucte the linux-ti-staging recipe, bitbake linux-ti-staging in my bbappends file I got the addtask foo after do_unpack. I execute bitbake linux-ti-staging -f and the foo is not executed. Aug 31 14:21:26 neverpanic: i'm trying to build a native version of this cmake recipe. I've noticed i have a bc in my native build dir, where I can do file bc and the executable is definitely for the host. However, when I look at the only recipe in my tree for bc, ./sources/openembedded-core/meta/recipes-extended/bc/bc_1.06.bb I don't see an inherit native in there. I thought you used the inherit native to specify a native Aug 31 14:21:27 neverpanic: If I do the same thing but with add task before do_compile it works fine. So I was wondering if the build process can't figure out where to place the foo command in the task order if it only contains an after do_unpack. It can litteraly be queued everywhere. Aug 31 14:21:32 build. Aug 31 14:22:04 it uses the BBCLASSEXTEND="native" syntax Aug 31 14:22:10 is that the alternative? Aug 31 14:27:04 fischerm: why would it be executed? bitbake $recipe is the same as bitbake -c build $recipe, and do_build does not depend on do_foo, unless you say "addtask foo before do_build" Aug 31 14:28:58 so you "addtask foo" without any dependencies just adds a possible new task, it does cause it to be executed at all. You've created a new leaf in the tree of tasks, but only the tasks required by the do_build leaf will be run by default Aug 31 14:30:58 s/fischerm/flitjes/g, sorry Aug 31 14:31:06 Also, see http://www.yoctoproject.org/docs/latest/bitbake-user-manual/bitbake-user-manual.html#dependencies-internal-to-the-bb-file where this is documented. Aug 31 14:32:15 davis: http://www.yoctoproject.org/docs/latest/mega-manual/mega-manual.html#var-BBCLASSEXTEND Aug 31 14:34:07 thanks, i'll try to add nativesdk as well. Aug 31 14:34:34 Why would you need that if you only need a version for your host? Aug 31 14:35:15 good question Aug 31 14:35:33 so here is the deal, this project i'm trying to get to work is in cmake Aug 31 14:36:43 the project has code which is built by make using the host compiler, later this code is run by the host to generate new files. these new files are then used for the next stage of make to build the final output. Aug 31 14:37:36 it seems that cmake can be configured to build for cross development, but it has a limitation where it can not be used in conjunction with bitbake to accomplish this. Aug 31 14:38:01 so, i hope to get bitbake to build what i've got working stand alone to simply build Aug 31 14:38:50 then use these executables to be imported into cmake recipe for the second build so that it uses prebuilt .cxx/.h files in the target build. Aug 31 14:39:26 So why would you need a version built for the SDK architecture then? Aug 31 14:39:52 im sorry i'm not being clear, let me be more detailed. Aug 31 14:40:13 stand alone, outside of bitbake I have managed to build this project. Aug 31 14:40:28 $cmake . generates a slew of makefiles. Aug 31 14:41:34 $ make then builds some .exe's. These .exes run on the host and build some additonal .cpp and .h files. Then make uses the .cpp files in src dirs and the .cpp files in generated dirs to build a set of final .exes. Aug 31 14:41:49 I understood what you're trying to do, I'm wondering what you hope to achieve by using BBCLASSEXTEND = "nativesdk" if there's no SDK involved anywhere in your process? Aug 31 14:42:16 i'm simply trying to get bitbake to build a project which builds standalone. Aug 31 14:43:06 and i've noticed that once I got cmake . to run in bitbake as part of its configure step, I ran into an error with do_make as part of its make step. Aug 31 14:44:01 the first thing it does is it builds some .exe's and then tries to run these .exes to make .cpp files and since the .exes are built for the target, they fail to run, and make stops and then bitbake fails in do_build. Aug 31 14:45:25 so in an attempt to get bitbake to build the same code, i've tried to get it to build as BBCLASSEXTENED="native", I've tried inherit cmake native, and it still fails. I thought nativesdk might help. Aug 31 14:47:50 that link you pointed me to, had a BBMAKEEXTENDED = "native and nativesdk" line. Aug 31 14:47:51 It won't. Aug 31 14:48:29 Yes, but that link also starts with an explanation what the different values mean Aug 31 14:49:17 from my limited understanding of bitbake and cmake, i thought nativesdk directive would say build this component for the native host using the native sdk. Aug 31 14:50:57 I have a lot of confusion, but looking at the build tree and examing output of what is there prior to my tampering, there are some native apps and some target apps. Aug 31 14:51:33 bc is an example of a native app, i can use file to determine its built for the host. I can even cd to its build dir and I can ./bc and it runs. Aug 31 14:51:56 so I thought bc was a good example of a native recipe to use Aug 31 14:54:44 what gets me is that bc shows up in this output dir for its build Aug 31 14:55:00 build/tmp-glibc/work/x86_64-linux/bc-native/1.06-r3 Aug 31 14:56:00 but my "native build" shows up in the target work dir Aug 31 14:56:28 build/tmp-glibc/work/corei7-64-oe-linux/pcmx-native/1.0.0+gitAUTOINC+d3_d3v16_pcmx-r0 Aug 31 14:57:54 then the only thing I can get make to build at this point is a single library, which appears to be built for the target (ie. its built with the cross compiler and not the native compiler) Aug 31 14:58:09 strings libcppreflection.a | grep arch Aug 31 14:58:25 shows GNU C++ 5.3.0 -m64 -march=corei7 -mtune=corei7 ... Aug 31 14:58:40 that looks to me like its not using my native host gcc. Aug 31 15:00:12 i'm not sure if ar or objdump or nm can tell which compiler built a .a file or what it architecture is. I think strings can detemrine that. Aug 31 15:00:36 on my standalone, build strings on that same .a has a different output. Aug 31 15:01:15 so it looks to me like bitbake using native is not built native, Aug 31 15:03:51 khem: around? Aug 31 15:04:03 khem: We've a couple of musl failures with the latest kernel headers :( Aug 31 15:48:34 davis: if your build ends up in tmp/work/$targetarch, it's not a native build, even though the recipe name may contain -native. Aug 31 15:50:03 you get a native build using 'inherit native' (or BBCLASSEXTEND += "native", which for all usual purposes is like "inherit native"). You seem to have a recipe that ends with -native, but doesn't inherit native. Aug 31 16:17:48 <_william_> hi all, i am trying to generate a Yocto Krogoth for an freescale imx6 target with opengl enabled. It goes until the end, GL demos are available from /opt/fsl-gpu-sdk, but all demos to start. It looks like something is missing... Do i have to enable another distro freature than "opengl" ? Aug 31 16:51:44 khem: I have fixes now FWIW Aug 31 16:55:06 oh hey, someone got a reproducer for the dumb-init thing. Aug 31 17:10:22 Is there any way to get systemd to output the stdout/stderr of spawned processes to its log? I'm trying to debug why a service is failing, but it seems I have pretty much nothing to go on... Aug 31 17:11:01 Cwiiis: Isn't that the default? It should go into the journal. Aug 31 17:11:01 Running what it should be execing manually works fine - it kind of looks like the substitution is broken, but it's not giving me enough log information to know what it's doing or how it's failing... Aug 31 17:11:59 stwcx: it's quite hard to tell, but it doesn't look like it is - though maybe it's just not getting to the executing the process part and perhaps that's my problem... Aug 31 17:14:47 journalctl -b -u Aug 31 17:15:10 That should give you all of the journal output from the current boot for a particular systemd unit. Aug 31 17:16:08 I think that 'systemctl status ' yields the same content though. Aug 31 17:16:26 stwcx: that was my problem, the binary directory is wrong. Kicking myself here. Aug 31 17:16:43 stwcx: thanks for your help :) Aug 31 17:19:26 Cwiiis: journalctl seems to indeed give more information that just the systemctl status. Aug 31 17:20:05 stwcx: I'll make a note of that - systemctl could definitely be more verbose, I'd have thought not being able to find the binary would output an obvious error message... Aug 31 17:20:40 By default all of stdout / stderr goes into the journal. I'm not certain how systemd decides what is in 'status' vs what is not. Maybe stderr goes to 'status'? Aug 31 17:21:02 Ah. 'systemctl status' is very useful for that part of it. ;) Aug 31 18:41:50 neverpanic: sorry, i missed your reply. Aug 31 18:42:59 I've got the inherit and extended syntax right, its doing a native build now. i can verify it with the strings mechanism i was showing earlier. ie. the .a file built standalone and with bitbake both match. Aug 31 19:47:12 RP: I think the svn fetcher was broken by the chdir changes. I have a downloads/Tremor/ svn dir while downloads/svn/svn.xiph.org/trunk is empty, and the mirror tarball construction fails due to the missing directory Aug 31 19:47:29 has anyone else tested BB_GENERATE_MIRROR_TARBALLS="1" for tremor recently? Aug 31 19:51:04 RP: found it, runfetchcmd(svnfetchcmd, d, ud.pkgdir) -> passes ud.pkgdir as the 'quiet' argument, not the 'workdir' argument, fetches the svn module into the completely wrong path Aug 31 19:53:39 sounds like we're missing a test Aug 31 20:10:33 indeed, if it's not noticed that a svn module showed up at the root of DL_DIR :) Aug 31 20:36:26 kergoth: patch to fix it would be most welcome Aug 31 20:36:35 kergoth: svn isn't used so heavily now I guess :/ Aug 31 20:42:57 join Aug 31 20:51:17 bitbake makes the deb package but doesn't install them in the image. Know where to look? Aug 31 20:53:09 ok, so i have got a very similar standalone cmake build and a native bitbake cmake build working. Aug 31 20:53:50 i have the next problem where I when I try to invoke the step for the generators, it differs between standalone and bitbake Aug 31 20:54:26 in standalone, it invokes the generator with a fully qualified path of an exe in the bin dir where it has built a binary. Aug 31 20:55:01 bobt: installing items in the image is controlled through IMAGE_INSTALL - just building a package isn't enough Aug 31 20:55:05 in bitbake, make cd's to the dir as fully qualified, but executes just the exe by name. not fully qualified or ./ Aug 31 20:55:27 bobt: see here if you haven't already: http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#usingpoky-extend-customimage Aug 31 20:55:45 is there some kind of switch to cmake or setting that modifies that? I dont see it showing up in the cmakecache.txt Aug 31 20:56:32 Thank you - thank you !! Aug 31 21:04:18 here is a more succient description of the situation Aug 31 21:04:21 i have a standalone build which works with cmake, if I use the same code in bitbake it fails. The failure is when it tries to run the generator code. in Standalone, it does cd /home/..../bin && /home/.../bin/foo_generator . In bitbake where it fails, it does cd /somebitbakedir/.../bin && foo_generator I do not have ./ in may path, so it can not find the generator. I've compared the Aug 31 21:04:27 two cmakecache.txt files on the standalone and bitbake builds, and nothing pops up for specifying the path differently. Any ideas? Aug 31 21:05:02 the bin dir it cd's to does exist. Aug 31 21:06:44 I have php in CORE_IMAGE_EXTRA_INSTALL += should I just be using - IMAGE_EXTRA_INSTALL += Aug 31 21:08:44 there's no such thing as IMAGE_EXTRA_INSTALL. Aug 31 21:13:00 Thank you kergoth - gee that might make a difference. Thanks again. Aug 31 21:47:15 Any idea why no scripts are ran during first boot with systemd ? Aug 31 21:52:04 make sure you have correct system time Sep 01 00:53:51 RP: ok thanks, fighting fevers this week, so afk most of time :( **** ENDING LOGGING AT Thu Sep 01 02:59:58 2016