**** BEGIN LOGGING AT Tue Jul 12 02:59:58 2016 Jul 12 03:00:04 how about a ternary operator construct in the depends? Jul 12 03:00:43 heeen: ordering shouldn't cause a problem here Jul 12 03:00:58 heeen: you mean ${@ ... } ? you could alternatively do that yes Jul 12 03:01:09 bluelightning: actually would a DEPENDS_append_imx6="fooo" work? Jul 12 03:01:40 that only works if it's do_configure -> do_populate_sysroot that you want in terms of a task dependency Jul 12 03:01:46 (that's what DEPENDS translates into) Jul 12 03:02:04 so it wouldn't work for this specific example I don't think Jul 12 03:02:05 not sure I get it Jul 12 03:02:11 why? Jul 12 03:02:34 because it's do_shared_workdir that you want to depend upon, right? not do_populate_sysroot Jul 12 03:03:02 my more abstract goal is I just need the path to that header and that it exists Jul 12 03:03:23 which is part of the kernel I think but for some reason not installed in my sysroot anymore Jul 12 03:03:47 right - and you're on the right track it sounds like Jul 12 03:03:56 I need the path during configure for the CFLAGS and it should exist during compile obviously Jul 12 03:04:04 STAGING_KERNEL_DIR is where you would find that file based on the path you've given Jul 12 03:04:07 so DEPENDS would not work? Jul 12 03:04:29 no, because as I said, that will give you a dependency on do_populate_sysroot, which is not the task you need for this Jul 12 03:04:37 hm Jul 12 03:05:23 no big deal though because you have two alternatives either of which should work Jul 12 03:06:53 bluelightning: what is the correct way to make it work for all imx6 variants Jul 12 03:07:15 some MACHINE beginswith imx6 hack? Jul 12 03:08:31 is there a hierarchy that checks for these _machinename vars existing for imx, imx6, imx6{dl,q}, imx6dlsabreauto etc? Jul 12 03:10:37 well I'm not too familiar with the imx6* BSPs but those are presumably represented by some other variable which makes its way into OVERRIDES... so you just need to find out what that variable is (bitbake -e | less and search for ^OVERRIDES and you should see it) and check that in the code Jul 12 03:17:00 bluelightning: having an issue on a build config that is a little older: Jul 12 03:17:42 ERROR: Task do_configure in myrecipe.bb depends upon non-existent task do_shared_workdir in meta-fsl-bsp/meta-fsl-bsp-release/imx/meta-fsl-arm/recipes-kernel/linux/linux-imx_3.14.28.b Jul 12 03:22:55 heeen: what OE-Core branch are you building with there? Jul 12 03:25:40 bluelightning: how do I find out? Jul 12 03:25:56 last tag of meta-openembedded? Jul 12 03:26:34 it would be the tag of openembedded-core (or poky, if you are using that), rather than meta-openembedded Jul 12 03:26:39 though they ought to match up Jul 12 03:27:24 poky Jul 12 03:28:01 tag: yocto-1.7.2, tag: dizzy-12.0.2 Jul 12 03:31:21 bluelightning: the newer one where it works is 1.8 Jul 12 03:32:53 right, the kernel shared workdir didn't happen until 1.8 Jul 12 03:33:14 dizzy isn't officially supported anymore FYI Jul 12 03:34:59 looks like they are moving to 1.8, just some teams lagging behind Jul 12 03:35:47 bluelightning: is there a way to make it work until I can uncomment the 1.8 way with the share_workdir depends? any task that comes close? Jul 12 03:36:54 I'm not sure, I don't know what would have happened to those files before the shared workdir rework Jul 12 03:37:48 STAGING_KERNEL_DIR expands to BUILD/sysroots/imx6dlsabreauto/usr/src/kernel Jul 12 03:38:03 and that works, just the depends does not Jul 12 03:38:41 might not work on a clean temp dir? but those files are required for libc, so it should happen usually, right? Jul 12 03:38:58 that task doesn't exist in dizzy, so it doesn't matter what the state of TMPDIR is Jul 12 03:39:51 for dizzy and earlier I would assume do_populate_sysroot would be what wrote out files to that path Jul 12 03:40:07 http://stackoverflow.com/questions/34793697/how-to-write-a-bitbake-driver-recipe-which-requires-kernel-source-header-files Jul 12 03:40:22 from that post, it looks like kernel headers are required to build libc Jul 12 03:40:36 so they should be put there on clean builds Jul 12 03:40:53 unless you use sstate? Jul 12 03:42:56 from linux-libc-headers, but that is intended for use by the C library only Jul 12 03:43:52 sstate or not shouldn't make a difference - if you have the appropriate dependency set up, the headers will be there Jul 12 03:44:23 it's when dependencies are missing that you'll see things missing when sstate is involved Jul 12 03:45:26 but what is the correct dependency on dizzy Jul 12 03:46:14 I have a working one for 1.8, but dizzy does not have it. Jul 12 03:47:22 I mentioned that already - I assume it would be do_populate_sysroot Jul 12 13:04:08 Crofton: any chance we can get the dates for OEDEM and Yocto dev day soon ? Jul 12 13:04:18 We'd like to book our flights :) Jul 12 13:10:37 99% certain dev day is Monday and OEDAM Friday Jul 12 16:27:43 Hi there Jul 12 16:28:30 I'm trying to produce a static binary of ostree for ARM Jul 12 16:28:41 for ease of deployment Jul 12 16:29:38 I'm wondering if openembedded/bitbake is appropriate for producing single application binaries Jul 12 16:29:54 rather than a complete system image Jul 12 16:31:07 It was easy enough to write the new rcepies Jul 12 16:31:13 it can do that Jul 12 16:31:19 just turn on static linking in the build Jul 12 16:31:23 and you'll get a package Jul 12 16:31:38 I'm not sure how to do this Jul 12 16:31:53 what bit? Jul 12 16:32:05 just passing -static in LDFLAGS will do a static build Jul 12 16:32:48 Just in the application recepie? Jul 12 16:33:43 I'll try that Jul 12 16:34:28 Thanks Jul 12 16:37:11 My concern was that I don't seem to be getting a glib-2.0-staticdev package Jul 12 16:38:04 which I assume is necessary for statically linking binaries that depend on glib Jul 12 16:41:09 yeah you'l want to enable static packages Jul 12 16:42:01 the poky distro pulls in no-static-libs, which disables static libs Jul 12 16:42:15 this is where i point out that poky is an example and you should make your own :) Jul 12 16:42:35 I don't think I'm using poky Jul 12 16:42:36 or unset DISTRO and use the defaults instead of poky Jul 12 16:42:59 I've got DISTRO = "nodistro" Jul 12 16:43:38 I'm building with TCLIBC = "musl" too Jul 12 16:44:18 WillManley: OE is transparent for this, you have to add static linking options to your package Jul 12 16:44:28 component build system Jul 12 16:44:48 I followed the instructions on http://www.openembedded.org/wiki/Getting_started Jul 12 16:44:51 bitbake just delegated the build jobs to the package's build system itself Jul 12 16:45:31 Right Jul 12 16:45:32 WillManley: as rburton suggested add -static to LDFLAGS and see what breaks Jul 12 16:46:04 Yep, will do. Currently rebuilding after upgrading to kragoth from jethro Jul 12 16:46:07 you can also inject it via recipe using something like LDFLAGS += "-static" Jul 12 16:47:10 It's good to know that I should persevere, I was becoming unsure whether this was a use-case supported by openembedded Jul 12 16:47:26 Thanks Jul 12 16:50:05 np Jul 12 16:51:01 I need to club a libatomic into a recipe for aarch64, any suggestions? Jul 12 17:18:41 Crofton|work: libatomic-ops is on OE-core Jul 12 17:18:44 already Jul 12 17:18:59 Can you use/patch that ? Jul 12 17:20:16 I am trying Jul 12 17:20:26 the symbol is in libatomic.a Jul 12 17:20:35 so I'd liek to insert that into the lnk line Jul 12 17:20:49 -latomic Jul 12 17:20:51 but I can only get it a the beginning of the list of -l's Jul 12 17:21:13 I guess you need to modify your package's make system for sequence issues Jul 12 17:21:19 yeah Jul 12 17:21:23 recipes are limited in that Jul 12 17:21:25 that is what I am fighting Jul 12 17:21:47 does is use gmake/autohell or cmake Jul 12 17:21:50 cmake Jul 12 17:21:53 or somehting more esoteric Jul 12 17:21:55 ok Jul 12 17:22:16 sounds like you confirm what I have learned so far, need to go into cmake Jul 12 17:24:32 find_package(libatomic REQUIRED) Jul 12 17:25:11 include_directories(${libatomic__INCLUDE_DIRS}) Jul 12 17:25:22 set(LIBS ${LIBS} ${libatomic_LIBRARIES}) Jul 12 17:25:31 something like this will help ya Jul 12 17:26:14 yeah< i was just hoping for something a little easier Jul 12 19:18:58 I hate cmake Jul 12 19:19:16 yes Jul 12 19:43:16 not as bad as qmake, so i hear Jul 12 19:43:28 true Jul 12 20:26:29 horros Jul 12 23:17:39 infact all build systems are horror, but its a necessary evil **** ENDING LOGGING AT Wed Jul 13 02:59:58 2016