**** BEGIN LOGGING AT Thu Nov 29 02:59:59 2018 Nov 29 07:55:01 Hello, i try to move my project from fido to thud. currebtly i have a problem to build busybox. the error message is: rpc/rpc.h - No such file. the problem was already mentioned here: https://lists.yoctoproject.org/pipermail/yocto/2018-May/041163.html but i don't find the solution if there is already one. What can i do? Nov 29 08:09:27 How can I add gcc compiler on my target? Nov 29 08:09:41 And also tools like make Nov 29 08:17:17 learningc: IMAGE_INSTALL_append = " packagegroup-core-buildessential" Nov 29 10:06:00 Hi. Yesterday I asked a question about how to know that a certain configuration fragment that is added in a custom layer is actually included in the final build, without having to flash Yocto on the actual device. aeh29 came with some good hints but then my build server went offline so I had to postpone this to today. To be a bit more specific, I am using Digi Embedded Yocto 2.4, and I have created a custom layer with a configuration fr Nov 29 10:06:15 contains "CONFIG_TUN=y". When I build, I can see in the project directory that my fragment file was picked up, but I don't know if the kernel was built using this option. Does anyone know how I can check that without flashing the OS? Nov 29 10:07:31 With "project directory" I mean: /home/dev/workspace/ccimx6ulsbc/tmp/work/ccimx6ulsbc-dey-linux-gnueabi/linux-dey/4.9-r0/tun_interface.cfg Nov 29 10:31:33 Hey guys, does anyone happen to know why trace ends up in the image even if I have no libaudit and trace states on target at execution "trace command not available: missing audit-libs devel package at build time."? Nov 29 10:32:10 Whats the purpose of having trace (3.3MB) on the image if it does not work? Nov 29 10:37:00 tristanram: look at the package dependencies to find out how it gets pulled in, and maybe also oe-pkgdata-util to see where it comes from Nov 29 10:37:06 tristanram: sounds like a bug, yes. Nov 29 10:46:27 LetoThe2nd: No results for "grep package-name recipe-depends.dot". Could it be that trace is built because my Kernel config has "CONFIG_AUDIT=y"? Nov 29 10:47:13 tristanram: i doubt it, at least on first glance. have you checked which package/recipe provides that trace binary? Nov 29 10:47:25 -> oe-pkgdata-util Nov 29 10:51:16 LetoThe2nd: It comes from perf: oe-pkgdata-util list-pkg-files -p perf ... /usr/bin/trace ... Nov 29 10:54:30 tristanram: and nothing pulls in perf? package-depends.dot being completely silent? Nov 29 10:55:22 LetoThe2nd: I am pulling in perf myself. But I would only like to have perf and not trace if possible Nov 29 10:55:50 LetoThe2nd: There is no point to me in having a non working trace Nov 29 10:56:06 tristanram: well then, look at the perf recipe if it is built unconditionally or what causes it, fix the recipe, then submit a patch :) Nov 29 10:56:57 IMHO this is a bug in the recipe Nov 29 11:11:15 LetoThe2nd: Trying to look into https://github.com/torvalds/linux/blob/v4.12/tools/perf/Makefile.config and https://github.com/torvalds/linux/blob/v4.12/tools/perf/Build on how NO_LIBAUDIT and CONFIG_AUDIT are working together. The latter has changed been changed to "CONFIG_TRACE" in newer Kernels. So the issue in the recipe might affect me because I am still on 4.12. Nov 29 11:20:45 hello! Nov 29 11:32:17 where to put my MACHINE_EXTRA_RDEPENDS? Nov 29 11:32:22 image recipe ? Nov 29 11:32:29 distro recipe? (i guess not) Nov 29 11:35:47 ak77: i'd say your machine config Nov 29 11:36:48 derRichard: ok. how do i then extend existing machine config ? (case of rpi, it's meta layer defines all the machines) Nov 29 11:37:35 i define always my own machine type and inherit the existing one Nov 29 11:38:21 i depends on how you want to work with yocto. i try to avoid settings in my local.conf as much as possible Nov 29 11:38:30 others stuff almost all settings there.... Nov 29 11:39:10 derRichard: yes, i try to do the same Nov 29 11:45:26 derRichard: I did so, but now it's missing BSP or defconfig. Nov 29 11:54:02 RP: not sure what to do with 'xxx binconfig: only mangle files'. intention was to stop attempting to rewrite absolute symlinks (wx has wx-config as an absolute symlink so doesn't do the right thing) but what if the link is relative... Nov 29 11:56:02 rburton: tricky, I keep hoping we don't have many of these things. Error if there is a symlink in there? Nov 29 11:57:03 rburton: it would probably be reasonable only to have files in BINCONFIG_GLOB Nov 29 12:00:10 rburton: good news is I've figured out the ccache failure and the stoptasks one Nov 29 12:00:15 RP: yay Nov 29 12:00:17 master emits lots of warnings from missing dependencies from ptest to locale packages. Is this a known issue? Nov 29 12:00:39 <[Sno]> I currently try to add a "run test" task for perl modules to prove before creating an image whether all/reasonable tests succeed Nov 29 12:00:48 e.g. WARNING: libxml2-2.9.8-r0 do_package_qa: QA Issue: libxml2-ptest rdepends on locale-base-en-us, but it isn't a build dependency? [build-deps] Nov 29 12:01:09 <[Sno]> is there any chance to add the interpreter into ${STAGING_DIR_TARGET}/usr/bin? Nov 29 12:01:13 mcfrisk: are you using musl? Nov 29 12:01:52 [Sno]: how do you suggest running a target binary at build time? Nov 29 12:02:13 <[Sno]> rburton: kind-of - similar to pgo run in python compile stage Nov 29 12:02:38 [Sno]: so i've a proof of concept for that for python. qemu-user is *really* limited Nov 29 12:03:16 [Sno]: see "XXX DOESNT WORK try package check" in poky-contrib:ross/py Nov 29 12:03:41 qemu-user can't fork/exec which may be a problem Nov 29 12:03:56 <[Sno]> rburton: I tried by copying target perl into ${STAGING_DIR_TARGET}/usr/bin and then "cloned" the qemu-wrapper to run "perl -e ... t/*.t" Nov 29 12:03:58 if that's not a problem then what i hacked up in that commit could be generalised Nov 29 12:04:22 <[Sno]> rburton: will check ;) Nov 29 12:04:37 for a module, yes, getting the binary it the tricky bit Nov 29 12:05:08 rburton: not that I know of, just building core-image-minimal with USE_NLS = "no" and GLIBC_GENERATE_LOCALES="en_US" Nov 29 12:05:35 <[Sno]> rburton: BTW - any decision how to deal with the python3 related issues like http://errors.yoctoproject.org/Errors/Details/200715/ ? Nov 29 12:05:44 mcfrisk: in that case its telling you that the tests won't work because youre missing the locales... Nov 29 12:06:01 [Sno]: properly and upstream would be best Nov 29 12:06:11 but upstream and cross is an uphill struggle Nov 29 12:06:22 though my float detection change made it in Nov 29 12:06:29 <[Sno]> I must reject upstream :P Nov 29 12:08:00 <[Sno]> And properly is maybe impossible for me - as I wrote, I could appy a similar hack to sysconfig as Perl 5.x is using to load the "right" Config_heavy.pl Nov 29 12:08:31 how to inherit from existing machine ? Nov 29 12:08:32 <[Sno]> when this is acceptable - I give it a shot Nov 29 12:09:19 RP: the cpio patch i posted replaces the ones i had in mut last night Nov 29 12:18:03 rburton: cool, I was expecting that :) Nov 29 12:18:13 rburton: just posted my various fixes for this morning... Nov 29 12:19:22 <[Sno]> rburton: to make it clear - I could hack something like https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/recipes-devtools/perl/perl/perl-configpm-switch.patch into Python-3.7.1/Lib/distutils/sysconfig.py Nov 29 12:21:48 [Sno]: haven't looked deeply enough, but i think we're at one with needing hacks there currently. poking and working with upstream is a long-term effort that needs someone willing to get really dirty with the build Nov 29 12:23:11 <[Sno]> rburton: that should probably do someone who is familiar with Python community etc. Nov 29 12:23:20 aye Nov 29 12:30:24 various fixes, theme of the day :) Nov 29 12:31:23 mcfrisk: i guess the tests should use recommends? Nov 29 12:32:34 rburton: yea, that would be nice Nov 29 12:32:45 if they really don't need more locales Nov 29 12:34:54 the tests will fail Nov 29 12:36:26 ok, then they need to depend. I can live with the warning. Was just wondering if it's by design and that seems to be the case. Nov 29 12:43:52 rburton: down to one major problem now I think, the bitbake server restart one :/ Nov 29 12:44:19 rburton: Is there a "rebuild most things" patch I can put into -next to run another round of selftest testing? Nov 29 12:45:20 oh i'm sure i have one somewhere ;) Nov 29 12:45:49 rburton: I've put a few questionable ones in -next already Nov 29 12:45:59 well there's the python manifest patches from aehs in mut Nov 29 12:46:17 touches python-native so surely will cause a fairly heavy rebuild Nov 29 12:46:54 rburton: I'm not sure it will actually rebuild Nov 29 12:48:52 Hi! I am building a yocto image for raspberrypi3. I am trying to get ufw to work. If I am trying to create some rules, I get an error message saying: FATAL: Module ip6_tables not found in directory /lib/modules/4.14.68 Nov 29 12:49:00 I have tried adding IMAGE_INSTALL_append = " kernel-modules". Then it is working. But I guess it is a better idea to just include the needed modules? Nov 29 12:49:09 If I am adding CONFIG_IP6_NF_FILTER=y, CONFIG_IPV6_SEG6_HMAC=y and CONFIG_IPV6_SEG6_LWTUNNEL=y, it is also working, but then the whole system is very unstable, and I get kernel panic after a few minutes. Nov 29 12:49:17 What should I do to get it working? Nov 29 12:49:30 rburton: I guess we'll see. Local test says it might be enough Nov 29 13:24:22 hi there! Nov 29 13:29:48 I moved from morty to sumo and I now have troubles to built images Nov 29 13:30:29 There is an error in the new recipe "make-mod-scripts" Nov 29 13:30:30 https://pastebin.com/86pB1M22 Nov 29 13:31:52 The kernel sources are 4.1.15: https://github.com/fmntf/linux_kernel/tree/4.1.15_2.x-udoo Nov 29 13:32:48 The error says: "scripts/genksyms/lex.lex.c_shipped:567:10: fatal error: parse.tab.h: No such file or directory" Nov 29 14:14:01 when using multilib yocto, what do i have to set to have the 32bits version of a library installed in my sdk? Nov 29 14:21:59 <[Sno]> rburton: see fwd mail - works in general Nov 29 14:23:23 <[Sno]> rburton: unfortunately there is a manual step ('cp tmp/work/cortexa8hf-neon-poky-linux-gnueabi/perl/5.28.0-r0/package/usr/bin/* tmp/work/cortexa8hf-neon-poky-linux-gnueabi/text-csv-xs-perl/1.37-r0/recipe-sysroot/usr/bin/') which needs somehow eliminated Nov 29 14:23:32 if you -c cleansstate for a recipe that DEPENDS on recipe recipeB, will recipeB be rebuilt? Nov 29 14:23:43 ...when bitbaking recipeA Nov 29 14:24:35 no Nov 29 14:25:51 you mean in binconfig.bbclass? Nov 29 14:25:57 [Sno]: you mean in binconfig.bbclass? Nov 29 14:26:35 so override it in your recipe with a new binconfig_XYZ_preprocess() Nov 29 14:27:21 <[Sno]> yates: it's worth a try - thank you very much Nov 29 14:28:21 [Sno]: i've been squirreling with binconfig stuff for my wxwidgets library build for days now... :( Nov 29 14:28:37 s/squirreling/fidgeting/ Nov 29 14:29:02 <[Sno]> good luck ;) Nov 29 14:29:11 i need it - thank you. Nov 29 14:53:25 I can't seem to be able to push to ssh://git@push.yoctoproject.org/poky-contrib anymore "Permission denied (publickey)". Is it just me? Nov 29 15:37:00 Hi, can anyone help me with a current problem? I am trying to migrate a project from yocto-morty to yocto-pyro. I have two recipes, recipe A produces a static library and puts it into the $STAGING_LIBDIR, recipe B uses this library for built-in compilation with reference to $STAGING_LIBDIR (so that the lib and header from recipe A don't need to go to the target sysroot, but can be used in different recipes). That worked perfectly Nov 29 15:37:27 specific sysroots and I don't know a way to access staging directory from recipe A in my recipe B Nov 29 15:38:49 all recipes install files to ${D} in do_install Nov 29 15:38:58 they should not be putting an ything in staging directly, even in morty Nov 29 15:39:22 and there's no need for recipe B to do anything, libraries and headers are in the default linker search path already Nov 29 15:39:32 linker/compiler, that is Nov 29 15:40:44 hi, thanks for the answer - I want to avoid that the library and headers from recipe A are put into the target system, I just want to share them with recipe B Nov 29 15:41:43 the correct process is to use do_install. Nov 29 15:41:56 if you don't want to package those files for target use, then set inherit nopackage or set PACKAGES = "" Nov 29 15:43:59 okay - ahm so recipe A installed the output to "install -m 0644 libgsm-utils.a ${STAGING_LIBDIR}/", where exactly should that go now? Nov 29 15:44:43 ${D}${libdir} is where it should always have gone, including in morty Nov 29 15:45:01 ah allright, I'll try that, thank you! Nov 29 15:45:27 no problem Nov 29 15:53:04 is libz named something else in the oe layers? Nov 29 15:53:26 | /opt/fslc-x11/2.4.1/sysroots/x86_64-fslcsdk-linux/usr/libexec/arm-fslc-linux-gnueabi/gcc/arm-fslc-linux-gnueabi/7.2.0/real-ld: warning: libz.so.1, needed by /Storage/production-hardware-revision-A-1.0/sources/poky/build-hw-test-image/tmp/sysroots/imx6ul-var-dart/usr/lib/libwx_baseu-3.0-arm-fslc-linux-gnueabi.so, not found (try using -rpath or -rpath-link) Nov 29 15:53:53 zlib? Nov 29 15:53:55 flib? Nov 29 15:55:57 https://paste.fedoraproject.org/paste/PByWgPtxko0VbCE476iYEQ Nov 29 15:57:47 sounds like zlib isn't in DEPENDS? Nov 29 15:58:25 that's what i'm trying now. but the linker was looking for libz, not zlib... Nov 29 15:58:49 and the linker looks for libc, but you don't depends on libc,l the project is glibc Nov 29 15:59:03 not naming the proejct the name of the lib is extremely common Nov 29 15:59:15 it's glib, not libglib, gtk+ not libgtk, etc Nov 29 16:05:51 OK, so I have a kernel module that I do not want installed on a bootstrap image Nov 29 16:05:57 but want it on all others Nov 29 16:06:15 and I have kernel-modules in machine definition, so all module send up in image Nov 29 16:06:30 How do I remove one module in one specific image? Nov 29 16:06:45 don't include kernel-modules in the machuine definition if you don't want it installed in all images Nov 29 16:07:20 kergoth, not the answer I am looking for :) Nov 29 16:07:22 it gets pulled in via a packagegroup, so any image with that packagegroup installed will end up with the kernel-modules package, you cant' use _remove to get rid of it at that point Nov 29 16:07:40 ah Nov 29 16:07:41 bother Nov 29 16:07:47 it's not directly installed via image_install or whatever Nov 29 16:07:51 it's part of packagegroup-base's machine package Nov 29 16:08:03 ok Nov 29 16:08:05 you could check and see if it's rrecommended or rdepended. if the former, you could remove it in the image Nov 29 16:08:12 if the latter, not doable without excluding packagegroup-base Nov 29 16:08:13 Need to deal then :) Nov 29 16:08:21 hmm Nov 29 16:08:41 comes in via MACHINE_EXTRA_RRECOMMENDS Nov 29 16:08:59 iirc paul put something in for removing packages, if it's rrecommended that won't result in removing the packagegroups Nov 29 16:09:01 = "kernel-modules Nov 29 16:09:06 PACKAGE_REMOVE Or something? or bad_recommends? Nov 29 16:09:07 * kergoth shrugs Nov 29 16:09:12 yeah Nov 29 16:09:13 ok Nov 29 16:17:04 kergoth: yeah, i guest "z" would be a pretty terse library name... Nov 29 16:17:09 s/guest/guess/ Nov 29 16:17:39 -lz Nov 29 16:18:14 there's no real consistency in library vs project naming Nov 29 16:21:36 ok someone please help me out here. i'm getting this error when building client-6300-devel recipe, which DEPENDS on wxwidgets, and the wxwidgets recipe DEPENDS on zlib (among others). https://paste.fedoraproject.org/paste/blyjn4sZaZWc68LYu~bt9g Nov 29 16:26:11 @kergoth thank you very much, that worked like a charm and solved my problem (and got rid of an ugly workaround) Nov 29 16:26:29 does yocto only pull in a DEPENDS library if something in the recipe directly depends on the library? it seems other libraries in the wxwidgets recipe DEPENDS (e.g., libglu) are getting pulled in. Nov 29 16:27:08 pschneider: np Nov 29 16:27:49 so now I am stumbling into the next problem - I don't know if yocto related, I probably miss some kind of configuration or a package dependency - I try to run a python script found on github to test bluetooth and when I execute it i get that error: "ImportError: cannot import name GObject, introspection typelib not found" Nov 29 16:28:38 but I set: DISTRO_FEATURES += "gobject-introspection-data" Nov 29 16:29:01 probably need to import the appropriate classes for gobject introspection. i don't recall what they are offhand Nov 29 16:29:56 but i think zlib is a secondary dependency Nov 29 16:29:57 yeah but I can't figure out which ones - tried googling the error Nov 29 16:30:01 looks like you'll need to inherit the gobject-introspection class, and possibly others depending on the project. gnomebase, gsettings, gtk-icon-cache, gtk-doc, etc. grep oe-core and go based on the particular recipe Nov 29 16:30:05 grep is your friend Nov 29 16:30:11 i just did a grep inherit.\*gobject Nov 29 16:30:26 yeah well it's just a *.py file I downloaded from github Nov 29 16:30:59 yeah, but there are 36^N possibilities for a grep search string of length N... Nov 29 16:31:08 at least.. Nov 29 16:31:43 yates: not sure offhand how indirect deps are handled in recipe specific sysroot construction Nov 29 16:31:54 I don't have a recipe I only try to execute the python script here Nov 29 16:32:49 if you're trying to run something on target, you'll need to ensure its dependent packages are installed, i.e. via adding to the image's image_install Nov 29 16:33:16 kergoth: was that intended for me? Nov 29 16:33:22 no Nov 29 16:33:26 yeah and I have problems figuring out which dependencies that *.py script has, that seems to be my root problem Nov 29 16:34:30 pschneider: i'd su ggest using recipetool to create a recipe for it. even if you don't need one, i t'll use the pythondeps script to scan it for python moduel/package deps and map them to the packages that've been built in your build dir Nov 29 16:34:45 so you can examine the resulting RDEPENDS_${PN} Nov 29 16:35:08 it's not full proof, since it naïvely scans the ast of the py thon script, but it's a starting point Nov 29 16:35:33 okay that is again very helpfull, I usually don't use the tool, can you point me in the right direction where to find documentation on how to do that? Nov 29 16:36:12 recipetool create --help Nov 29 16:36:28 it's probably int he yoctoproject.org docs too of course Nov 29 16:36:41 but it's rpetty straightforward. recipetool create Nov 29 16:38:01 thanks a whole lot! I'll do that! Nov 29 16:39:36 foolproof Nov 29 16:40:18 i see this in one of the lines from my linker error: libgio-2.0.so.0: undefined reference to `deflateSetHeader@ZLIB_1.2.2' Nov 29 16:40:29 does that mean it's looking for version 1.1.1 of zlib? Nov 29 16:40:39 sorry, 1.2.2 Nov 29 16:40:58 my sysroot has 1.2.8, not 1.2.2... Nov 29 16:43:26 it likely still has 1.2.2 versioned symbols, however, unless they do it much differently than glibc does Nov 29 16:44:02 right. for backwards compatibility. Nov 29 16:44:25 exactly Nov 29 16:46:50 does the build system search /lib as well as /usr/lib? Nov 29 16:47:05 libz.so.1 is in the former but not the latter Nov 29 16:47:48 and i noticed a lot of symlinks to /lib, via ../../lib Nov 29 16:49:06 is there an option to ld that says "spit out all your library search paths" Nov 29 16:49:18 s/spit out/emit to the console" Nov 29 16:49:35 i've looked and don't see it. Nov 29 16:50:31 or maybe i should just "echo $"? Nov 29 17:13:55 kanavin: I'm impressed you fixed a bug assigned in triage before the triage call finished :) Nov 29 17:24:44 DOH!!!! Nov 29 17:25:28 that was it. "ln -s ../../lib/libz.so.1" in the /usr/lib of sysroot fixed the link. Nov 29 17:25:49 is this a bug? Nov 29 17:26:09 maybe i should just put it in my do_install? Nov 29 17:26:13 sounds like a bug in whatever is building, not libz Nov 29 17:26:17 * rburton has to go again Nov 29 17:26:39 option 1: zlib is broken and nothing that uses it builds Nov 29 17:26:45 (that's a lot of software) Nov 29 17:26:55 option 2: zlib is fine and you're building something which is doing it wrong Nov 29 17:27:30 rburton: i'm simply putting zlib in the DEPENDS. it's yocto that does the rest Nov 29 17:28:13 well, that's not quite true Nov 29 17:28:45 i added this so binconfig_sysroot_preprocess(): sed -i -e 's:_ldflags="-L$libdir":_ldflags="-Wl,-rpath-link=$libdir -Wl,-Y$libdir":g;' $config Nov 29 17:29:15 s/so/to/ Nov 29 17:29:48 the -Y option is an old solaris option that says "add this path to the default paths" Nov 29 17:30:28 still, i don't see why this would hose zlib Nov 29 17:32:06 should i change it to: sed -i -e 's:_ldflags="-L$libdir":_ldflags="-Wl,-rpath-link=$libdir -Wl,-rpath-link=$libdir/../../lib -Wl,-Y$libdir":g;' $config Nov 29 17:35:15 that's in the wx-config script Nov 29 17:35:41 i suspect this is getting really over-engineered but i'm also cooking right now Nov 29 17:36:33 all binconfig does is change all references of eg $libdir to $STAGING_LIBDIR Nov 29 17:36:45 so instead of getting -L/usr/lib it returns -L/path/to/sysroot/lib Nov 29 17:36:58 the problem is the -L option itself Nov 29 17:38:16 if wx-config is assuming that zlib is in $libdir then its wrong, and just hack it to add STAGING_BASELIBDIR too Nov 29 17:38:23 erm, with a bonus _ Nov 29 17:39:04 STAGING_BASELINELIBDIR is basically /lib? Nov 29 17:39:36 yes Nov 29 17:39:41 STAGING_BASELIB_DIR Nov 29 17:39:50 libdir = ..../usr/lib Nov 29 17:39:55 baselibdir = .../lib Nov 29 17:40:02 right Nov 29 17:40:16 from the last 10 seconds sounds like wx is assuming zlib is in the same prefix Nov 29 17:40:21 same dir, even Nov 29 17:42:48 if that assumption is wrong, why does it work on a native build? Nov 29 17:46:06 is there a $baselibdir as STAGING_BASELIB_DIR is toe STAGING_LIB_DIR? Nov 29 17:46:30 have a look in bitbake.conf :) Nov 29 17:46:54 i've heard of that before.. :) Nov 29 17:50:14 the problem with -L is that it was causing the link to pick up the wrong version of libpthread. from the ld man: "Directories specified on the command line are searched before the default directories. All -L options apply to all -l options, regardless of the order in which the options appear." Nov 29 17:50:49 so without a -L, the correct one was picked up Nov 29 17:51:47 at least that's my story and i'm sticking to it... Nov 29 17:51:51 ymmv Nov 29 17:54:40 RP: I sent a v2 for musl update, I am not yet able to reproduce the issue locally, I am doing bitbake -ctestimage core-image-sato-sdk for qemux86-64, please give this a shot it has three more commits see https://git.musl-libc.org/cgit/musl/log/?qt=range&q=79f653c6bc2881dd6855299c908a442f56cb7c2b..39ef612aa193cc6e954ac5a01574300ccd4b7ef9 Nov 29 18:26:51 why is this happening? https://paste.fedoraproject.org/paste/n5b1Lq3y0VGqw7lvIG6vRw Nov 29 18:27:12 (running the app that DEPENDS on wxwidgets on the target) Nov 29 18:27:35 yeah, different ABI versions, but why are they different? Nov 29 18:29:23 different flags used between the library and application build? Nov 29 18:29:32 iirc c++ abi can be specified.. Nov 29 18:31:19 you mean via a compiler flag? Nov 29 18:33:32 things like -mfloat-abi=hard ? Nov 29 18:34:28 yes, i.e. -fabi-version=0 or -std=c++0x or whatever Nov 29 18:34:58 in fact, from a quick google about it, https://www.reddit.com/r/Gentoo/comments/8x5e9m/incompatible_c_abi_correct_way_to_handle_in_ebuild/ looks to be the very same error Nov 29 18:35:02 but not -mfloat-abi, that would show different error abotu VFP registers Nov 29 18:35:17 that said, i'd expect wx-config to return the correct cxxflags Nov 29 18:35:35 so perhaps you just need to ensure you're using those Nov 29 19:27:41 wxGTK 2.8.x and 3.0 seems to incompatible Nov 29 19:29:15 another issue could be flags used to compile wxGTK 3.x Nov 29 19:30:27 gcc8 defaults to c++14 Nov 29 19:33:15 it seems one of the involved packages is forcing -std flag can you post verbose do_compile for both the app and lib Nov 29 19:42:58 see this is the issue https://github.com/bmwcarit/meta-ros/blob/master/recipes-extended/wxwidgets/wxwidgets_2.9.5.bb#L28 Nov 29 19:43:20 delete that Nov 29 19:53:24 is there a way to specify the -std= option for the c++ when bitbaking a library (wxwidgets)? what does yocto use by default? my application is using -std=c++11 Nov 29 20:02:56 yates: it's up to the underlying buildsystem Nov 29 20:03:05 bitbake doesn't fuss with how to build some particular cpp file Nov 29 20:05:50 is there a g++ option to emit the default c++ abi? Nov 29 20:09:45 Has anybody managed to build wifi into the yocto image for raspberry pi? I've added IMAGE_INSTALL_append = " linux-firmware-bcm43430" to local.conf, but it didn't work Nov 29 20:16:07 kergoth: ok. i'm just trying to see what XY in std=c++XY is being used when the autotools compile the wxwidgets source. Nov 29 20:16:20 ...trying to find out what XY... Nov 29 20:26:39 can i do something like "export CXXFLAGS="$CXXFLAGS -std=c++11" in a do_compile_prepend? Nov 29 20:30:18 oe doesn't set abi, which means you'd run into this outside of it too, so it should really be fixed there Nov 29 20:30:35 where? Nov 29 20:30:44 in the underlying buildsystem, as mentioned before Nov 29 20:31:03 autoconf? make? qmake? i have no idea what that recipe's project uses Nov 29 20:31:12 either way, abi selection isn't handled at the oe level Nov 29 20:31:18 generally speaking, anyway Nov 29 20:38:09 since this is autotools-based, are you saying that i should see where the abi is specified there? Nov 29 20:38:23 e.g., in some .m4 files or somesuch? Nov 29 20:38:56 beyond "./configure; make; make install" i'm pretty ignorant of autotools... :( Nov 29 20:45:26 and like i asked before, what if the autotools doesn't specify it? then isn't the compiler default used? and the compiler is NOT part of the autotools, but part of yocto Nov 29 20:45:39 the SDK, if i have that term correct. Nov 29 20:45:41 no? Nov 29 20:46:10 or am i confused? Nov 29 20:46:13 it's not relevant, because the same toolchain is used for both wxwidgets and your app Nov 29 20:46:19 if both were using the default, the mismatch wouldn't occur Nov 29 20:46:23 so one or both are specifying it Nov 29 20:46:44 yes, my applicaiton is specifying c++11, i know that for a fact. Nov 29 20:47:07 and that's not matching how wxwidgets is being compiled. Nov 29 20:47:10 so change one or hte other to match Nov 29 20:47:11 what if i really needed that ABI for my app? Nov 29 20:47:56 maybe i should rethink that - maybe i don't need it for the app. Nov 29 20:58:44 but yes you could manually adjust CXXFLAGS, assuming it's being obeyed properly. you could append to it right it he recipe. CXXFLAGS += "..." Nov 29 21:02:16 la_croix: use Yoe distro :) https://github.com/YoeDistro/yoe-distro Nov 29 21:03:47 yates: if you need c++11 then c++14 will work too, you dont need to lock to c++11 Nov 29 21:04:43 yates: unless you tell me what feature from c++14 is regressing your app, I would suggest to use defaults from compiler Nov 29 21:04:55 unless you are using gcc < 5 Nov 29 21:07:15 khem: hmm, meta-yoe seems not to be in the layer index... Nov 29 21:14:56 hadn't heard of yoe before, interesting Nov 29 21:18:34 I guess I will add it... I typically don't, people should submit their own layers, but I doubt Cliff would object Nov 29 21:28:17 does "devtool build " run all the tasks to the end including rpms in the tmp/deploy? Nov 29 21:36:53 any poor soul also using 2.2 Morty (or any pre-RSS version) with multilib? Nov 29 21:37:04 yes Nov 29 21:38:06 do you also see errors/warnings like ERROR: QA Issue: gcc-cross-canadian-i686 rdepends on nativesdk-elfutils, but it isn't a build dependency? [build-deps] ? Nov 29 21:38:48 it's caused by wrong PKGDATA_DIR set by multilib.conf, but I'm wondering why nobody was complaining about it before Nov 29 21:39:19 yates: no, just up to do_populate_sysroot & do_packagedata Nov 29 21:39:53 yates: however, it's not documented, but you can configure that in conf/devtool.conf Nov 29 21:44:33 bluelightning: yes, maybe I should submit it Nov 29 21:44:46 khem: already taken care of :) Nov 29 21:44:53 khem: I also added meta-riscv Nov 29 21:44:55 cool ! thanks Nov 29 21:45:02 yes I got email :) Nov 29 21:45:16 ah yes I forgot we now email the maintainers Nov 29 21:45:37 Cliff added you after the fact hence why you wouldn't have gotten one for meta-yoe Nov 29 21:46:27 i see Nov 29 21:47:12 strictly speaking I wonder if we should email maintainers when they get added if they aren't adding themselves Nov 29 21:47:23 seems like a courteous thing to do Nov 29 21:47:45 * bluelightning adds todo item Nov 29 21:47:57 you should Nov 29 21:56:49 Hey guys, I have a problem running a simple python3 script on my target. I wrote a recipe for it and try to figure out all dependencies, which I then included in the image. I also included 'gobject-introspection-data' into my DISTRO_FEATURES. Still - if I try to execute the script on my target it fails with ' File "/usr/lib/python3.5/site-packages/gi/importer.py", line 127, in find_module 'introspection typelib not found' % names Nov 29 21:56:55 on my host machine Nov 29 21:57:29 Are there any packages I could have missed which are not obvious? For the record, I am kinda new to python, never needed it before. Nov 29 22:04:06 is there a way to bitbake my recipe using modified files in the fetch folder? it seems the recipe will overwrite the mods there with the pristine original recipe source sometime during bitbake Nov 29 22:04:28 not sure what you mean by fetch folder Nov 29 22:04:36 but devtool modify is the easiest method to work with modified source Nov 29 22:05:52 kergoth: yeah, but i need the rpm's to generate too. Nov 29 22:06:05 and? Nov 29 22:06:11 and they're not Nov 29 22:06:15 devtool modify foo; bitbake foo -> now you have rpms Nov 29 22:06:25 you don't have to use devtool build just to use devtool modify Nov 29 22:06:37 doh! Nov 29 22:07:09 well why didn't you say so ... Nov 29 22:07:11 * yates ducks Nov 29 22:07:13 devtool modify makes a persistent change by creating the workspace layer, adding it to BBLAYERS, and generating a bbappend for the recipe in that layer which points to the pre-extracted/patched sources. any subsequent bitbake command will continue to work there Nov 29 22:13:17 does bitbake yes Nov 29 22:13:41 great. Nov 29 22:21:50 is the proper way to invoke my application's clean function as part of a -c clean to define a do_clean_append() with my clean command(s)? Nov 29 22:22:52 it's not working: https://paste.fedoraproject.org/paste/QAP4whUhu4Yk2pe~vVB1kA Nov 29 22:23:22 i don't see what's wrong with this syntax - most of it is copied from the do_compile() Nov 29 22:23:24 https://paste.fedoraproject.org/paste/JzwGb~f3DfNzFDeWuKSTPQ Nov 29 22:28:19 does do_clean_xyz have to be python ? Nov 29 22:28:37 it can't be shell like the other tasks can be??? Nov 29 22:29:08 yates: appends have to be whatever the original function is in Nov 29 22:29:17 doh! Nov 29 22:29:31 two years' in now and just finding this out... Nov 29 22:30:57 there is a lot of STUFF to know in this tool!!!!!!!! Nov 29 22:31:06 it hurts Nov 29 22:34:12 building OS isn't for the weak :) Nov 29 22:36:16 things really start to get interesting when you begin forgetting faster than you learn/refresh... Nov 29 22:36:45 * yates cracks the mental whip at himself!!!! "Hya!" Nov 29 22:47:47 does morty use python 2? Nov 29 22:49:17 better "does morty's bitbake use python 2"? Nov 29 22:50:30 ah, 2.6 or 2.7, right? Nov 29 22:50:34 https://www.openembedded.org/wiki/Required_software Nov 29 22:57:01 morty uses python3 Nov 29 22:57:18 but bitbake itself is still py2 Nov 29 23:01:48 yates: if you can, do out of tree builds instead of needing a custom clean command Nov 29 23:02:50 so e.g. the python functions inside a .bbclass are python3? "python do_listtask() {...}" Nov 29 23:04:13 no if bitbake is running in py2 then the functions need to be py2 Nov 29 23:05:10 rburton_: i don't understand the term "out of tree" although i've heard it used a lot. what do you mean by "out of tree bilds"? Nov 29 23:05:23 builds Nov 29 23:06:31 if S=WORKDIR/foo-1.2/ then build tree can be B=WORKDIR/build. no built objects are written to S, but all written to B Nov 29 23:07:07 so does a standard do_clean remove the whole B? Nov 29 23:07:14 no, because by default S=B Nov 29 23:07:31 *but* autotools will set B=workdir/build and delete $B Nov 29 23:07:48 ditto cmake, meson, and a few other build systems that do out of tree builds properly Nov 29 23:07:53 of course, morty, so maybe not that old Nov 29 23:16:58 is there no way to call a shell function from python? Nov 29 23:17:53 so "python do_clean_append() {call myshellclean}" and "myshell_clean () {make clean}" Nov 29 23:18:17 bb.build.exec_func() will let you run any bitbake function from a py function Nov 29 23:18:32 you could just override do_clean entirely Nov 29 23:18:48 presumably you're not actually running make clean, right Nov 29 23:20:00 not exactly but close. "make -f makfile.mak target= platform= clean" Nov 29 23:20:33 why are you doing that in do_clean? Nov 29 23:21:04 what's wrong with doing it there? seems obvious from the naming convention Nov 29 23:21:11 do_clean deletes ${WORKDIR} Nov 29 23:21:31 so your source tree disappears Nov 29 23:22:01 i need a command i can run to clean my project's output files. how to do? Nov 29 23:22:17 lets back up: what are you trying to do Nov 29 23:22:46 make clean, basically, at the right directory in my project files, and with an environment variable set Nov 29 23:23:31 because my build puts the output outside of yocto (in /tmp) i need to run this to really clean Nov 29 23:24:26 i thought it would be easy to splice that in to yocto's clean process... Nov 29 23:24:35 somewhere, somehow.. Nov 29 23:24:48 it does what Nov 29 23:25:03 why does it put stuff in /tmp instead of under $WORKDIR Nov 29 23:25:43 this is a make system i've had since before yocto existed Nov 29 23:25:51 it's the way i did things Nov 29 23:26:50 do_clean[postfuncs] = "my_horrible_clean_hack" Nov 29 23:27:08 hm, no, how would you run make clean after workdir has been deleted Nov 29 23:27:12 maybe a prefuncs Nov 29 23:27:20 then you can just write a shell function that runs make Nov 29 23:27:24 before workdir disappears Nov 29 23:27:41 what prefuncs? Nov 29 23:28:57 do_clean[prefuncs] = "my_horrible_clean_hack" Nov 29 23:29:12 my_horrible_clean_hack() { make clean } Nov 29 23:29:19 ok, cool Nov 29 23:29:31 +=, actually. Nov 29 23:29:47 lemme give it a whirl.. Nov 29 23:31:52 and cleansstate invoke clean, so it'll run then too? Nov 29 23:31:59 --invokes Nov 29 23:33:47 ok seems to be working.. cool. Nov 29 23:45:08 is readelf part of some tools recipe? Nov 29 23:46:14 yates: yes, binutils Nov 29 23:46:21 yates: thanks man! Nov 29 23:46:27 teamwork! Nov 29 23:49:54 rburton_: why aren't you gone? you usually exit hastily around 5:07 Nov 29 23:50:00 EST Nov 29 23:50:16 i should shutup and enjoy the treat! Nov 29 23:50:19 haha Nov 29 23:50:25 because i just finished my 11pm staff meeting Nov 29 23:50:33 ? Nov 29 23:51:27 hooray working in a team which has people across the globe, and being the only one in europe. 11pm calls Nov 29 23:53:08 to California? Nov 29 23:53:27 that'd be what, 3pm Nov 29 23:53:47 nm. Nov 29 23:53:52 mind going... fading... Nov 29 23:55:44 west coast Nov 29 23:55:56 and malaysia Nov 29 23:56:24 i hope they dialled in from bed as its 7am for them! Nov 30 00:01:24 rburton_: are you in the uk? Nov 30 00:01:29 yes Nov 30 00:01:35 cool Nov 30 00:01:51 never been except to sit on the runway at heathrow Nov 30 00:01:58 not our finest moment Nov 30 00:02:13 and i was born there.. really. dad was USAF stationed there in 1957 Nov 30 00:02:31 what base? Nov 30 00:02:39 Oglethorpe? Nov 30 00:02:42 i think? Nov 30 00:02:43 great name Nov 30 00:02:48 RAFB Nov 30 00:03:01 is that near Kings Lynne? Nov 30 00:03:14 Lynn Nov 30 00:03:18 they stayed there too. Nov 30 00:03:21 oh, sorry.. Nov 30 00:03:50 not aware of that base but i used to live sort-of-near kings lynn, and yeah there's several bases still there **** ENDING LOGGING AT Fri Nov 30 03:00:00 2018