**** BEGIN LOGGING AT Fri Feb 03 03:00:01 2017 Feb 03 08:07:33 Hmm, the speedup I'd hoped for from the patchelf change didn't appear :( Feb 03 08:18:10 good morning Feb 03 08:20:42 'morning Feb 03 08:23:21 Hi guys! Feb 03 08:24:52 I've built my own kernel, but i can't pass next step with vmlinux. I got error - " No rule to make target 'vmlinux'. Stop." in Step "make -j 4 vmlinux vmlinux.bin CC=mips-deby-linux-gcc -fuse-ld=bfd LC=mips-deby-linux-ld.bfd" Feb 03 08:25:36 could you help me please with this issue? how to fix it? Feb 03 08:26:40 I run " bitbake -k linux-mykernel" and added new .bb in meta/recipes-kernel/linux/linux-mykernel_3.19.bb Feb 03 08:30:28 how can I get shared libs symlinks created with a non *-dev ipk? I have a cmake based recipe, which works fine, but the symlinks are provided by the *-dev ipk package Feb 03 08:36:53 mrpelotazo: which symlinks exactly -- libname.so is (almost always) supposed to be in -dev package Feb 03 08:38:10 jku: they are libname.so. I would like to not having to install the *-dev package just to get the program running... Feb 03 08:40:02 mrpelotazo: if your library has a versioned so file, you should not usually need the non-versioned .so to run a program Feb 03 08:42:58 jku: so I should rework cmake to link against the versioned one and use the symlink only for development, right? Feb 03 08:49:06 sounds good to me. You can of course workaround the packaging rules but for the normal case this should be correct Feb 03 08:53:19 jku: OK. tnx Feb 03 09:05:40 RP: whats the chances of merging Armin's morty staging branch? Feb 03 09:34:49 nrossi: looking at the patches, good? Feb 03 09:44:43 maybe a dump question but is there a way to get a somehow usable dependency graph with bitbake -g ... I have a really simple image with "just" around 40 packages but the dot files are already some mb's and the accourding pngs doesn't show anything? Feb 03 09:47:09 RP: Armin indicated he was just waiting for you to get around to merging it on this thread here: https://www.mail-archive.com/openembedded-core@lists.openembedded.org/msg91135.html Feb 03 09:47:57 nrossi: I guess I was waiting on builds but those haven't happened. I'll just merge it, there are other important things missing from morty atm though :/ Feb 03 09:48:32 mdnneo: our task graph is complicated and visualising it using conventional viewers doesn't work well sadly :( Feb 03 09:48:54 has anyone gotten meta-darwin to work lately? Feb 03 09:48:58 RP: ah ok, im just getting repeat reports which are fixed by the patch in the linked thread which Armin merged. So just looking to nudge the process along :) Feb 03 09:49:06 I get: ERROR: No recipes available for: [...] sources/meta-darwin/recipes-devtools/binutils/binutils-cross-canadian_2.24.bbappend Feb 03 09:49:27 RP: I somehow also think there are the host package in as well ... is there some way of really just get the image relevant part? Feb 03 09:50:15 mdnneo: it will include the tools it needs to build them, yes. You can usually navigate those files ok with a text file viewer fwiw Feb 03 09:52:28 nrossi: merged Feb 03 09:52:52 mdnneo: it's usually easier to just look through the .dot file with a normal text editor as RP said Feb 03 09:53:00 then you can easily search for the packages etc. Feb 03 09:53:07 or grep Feb 03 09:55:55 RP: thanks! Feb 03 10:00:17 mdnneo: also, if you weren't aware, the depexp ui is ok for some purposes "bitbake -g -u depexp " Feb 03 10:00:43 :q Feb 03 10:01:13 wrong window :D jku ... I will have a look Feb 03 10:16:46 I'm trying to understand why gstreamer1.0-plugins-bad has libx11 in it's recipe-sysroot? gstreamer1.0-plugins-bad does not DEPEND on libx11... Feb 03 10:18:11 jku: plugins-base seems to depend on it? Feb 03 10:21:45 and that's enough to bring it into plugins-bad sysroot? I must have misunderstood something at some point Feb 03 10:23:46 jku: sstate.bbclass, setscene_depvalid() "# Target populate_sysroot need their dependencies" Feb 03 10:24:30 jku: I guess stuff in base can link against libx11 so we have to have it present for things which depend on base else things can go wrong... Feb 03 10:36:18 hi, I am trying runqemu session with raspberrypi2, but libvirt is not done and the kernel has no "virt" configs, so I think I get this error """No 'PCI' bus found for device 'virtio-net-pci'""", anyone can advise? thx. Feb 03 10:38:15 rburton: the aarch64 failure isn't my patch but the one from QA, I think we suddenly enabled more testing of the sdk images Feb 03 10:40:03 rburton: http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=3eef8f7955d09063b1ebfd7ea5cc74ee67fcc4a2 was the guilty change Feb 03 10:49:27 didn't we purposefully remove a bunch of those slower ones from the defaults? Feb 03 10:50:29 ah, we still do that later on with a _remove for mips and aarch64 Feb 03 10:50:35 joshuagl: We removed them from mips, right. I did check Feb 03 10:57:57 jku: guess I will have to force creating a libname.so symlink. the program dlopens a plugin passed as a parameter on the CLI: $program Feb 03 10:59:29 jku: but the plugin lib is installed with version: libname.x.y.z and the program expects it to be just "*.so" Feb 03 11:00:21 jku: so how could I get that plugin lib symliked when installing the non-dev-ipk? Feb 03 11:00:36 Hello, I have a question regarding OPKG package manager, I'm changing the way ipk packages are created inside yocto, to support extended attributes inside tar files creation, by editing opkg-buil shell script in poky. Now I succedded to have the ipk file supporting XATTRS inside, but this changes the package header, and opkg-cl is complaining installing this package in rootfs because it has unknown header format. BTW option format=gn Feb 03 11:01:27 get_header_tar: Unknown typeflag: 0x78: No such file or directory. Feb 03 11:29:48 mrpelotazo: your library sets soname. _usually_ that should be e.g. "libname.so.x" Feb 03 11:30:17 mrpelotazo: you should dlopen that name Feb 03 11:31:24 (again, usually) libname.so.x is a symlink to libname.so.x.y.z. Both are in main package Feb 03 11:35:19 jku: I can't change the dlopen code. All I'm allowed to do now is to get the current code running as is Feb 03 11:37:35 jku: the prgram gets the plugin_name to load as a paramter and it expects the according lib to be "*.so". So I would need to get the ipk to create .so links for the installed plugin-libs Feb 03 11:46:53 mrpelotazo: in that case you probably really need to put the .so in the main package . It shouldn't be difficult but I can't quite remember details... maybe try settings SOLIBSDEV = "" and adding the .so file to FILES_${PN} ? Feb 03 11:48:19 mrpelotazo: https://wiki.yoctoproject.org/wiki/TipsAndTricks/Packaging_Prebuilt_Libraries#Non-versioned_Libraries Feb 03 11:48:28 right that's be done for the day, time to collapse again Feb 03 11:48:39 rburton: :( Feb 03 11:49:16 * kanavin is ready to collapse as well Feb 03 11:49:27 the week of intense bugfixing has taken its toll Feb 03 11:49:48 oe-selftest has been really useful to reveal issues with dnf I haven't thought of Feb 03 11:49:55 heh, and I'm sitting here gently shaking :/ Feb 03 11:50:05 and the AB hasn't even been involved yet at all :) Feb 03 11:50:05 kanavin: selftest was great for rss too Feb 03 11:51:51 RP: yeah, investing in a test suite (which may seem like not the most important thing to do) *always* pays off majorly in the future Feb 03 11:53:08 okay, enough, I'm staring at an error, and literally can't focus Feb 03 12:32:10 jku: that works! Is it possible instead of setting an empty SOLIBSDEV, just to exclude the one plugin lib so I can still get the other "*.so" symlinks in the *-dev package? Feb 03 12:36:44 mrpelotazo: yes you should be able to set SOLIBSDEV AND SOLIBS so that they catch the only the correct ones... Feb 03 12:37:42 er no, looking at the how it's used that might not actually work Feb 03 12:49:55 mrpelotazo: there is a FILES_SOLIBSDEV (see e.g. meta/recipes-support/db/ recipes) but I've never tried it. In the worst case you can just define FILES correctly for main package and -dev and then silence the QA warnings once you know they don't apply to you... Feb 03 12:56:14 i am trying to build my kernel. it built. but failed on step do_package_qa with error " QA Issue: Endiannes did not match ( 0 to 1) . how to troubleshoot it ? Feb 03 12:56:40 I use receipt from meta/recipes-kernel/linux/linux-mykernel_3.19.bb Feb 03 12:57:53 my build system arch is x86_64 Feb 03 12:57:57 my target arch is mips Feb 03 12:58:34 am i on right way with using meta/recipes-kernel/linux/linux-mykernel_3.19.bb for this task? Feb 03 12:59:34 "ERROR: QA Issue: Endiannes did not match ( 0 to 1) on work/qemumips-dby-linux/..../crypto/seqiv.ko [ arch ] Feb 03 13:00:39 I added TARGET_CC_ARCH += "${LDFLAGS}" into linux-mykernel_3.19.bb but it doesn't help Feb 03 13:01:01 guys, any clues ? Feb 03 13:01:17 Olegt: how is your local.conf? Feb 03 13:01:32 Olegt: how is your linux-mykernel_3.19.bb? Feb 03 13:02:18 Olegt: are you using any mips metalayer? Feb 03 13:02:55 Olegt: https://www.yoctoproject.org/docs/current/yocto-project-qs/yocto-project-qs.html Feb 03 13:04:25 my local.conf Feb 03 13:04:26 BB_NUMBER_THREADS ?= "${@oe.utils.cpu_count()}" PARALLEL_MAKE ?= "-j ${@oe.utils.cpu_count()}" MACHINE ??= "qemux86" DISTRO ?= "deby" PACKAGE_CLASSES ?= "package_deb" EXTRA_IMAGE_FEATURES = "debug-tweaks" USER_CLASSES ?= "buildstats image-mklibs image-prelink" PATCHRESOLVE = "noop" BB_DISKMON_DIRS = "\ STOPTASKS,${TMPDIR},1G,100K \ STOPTASKS,${DL_DIR},1G,100K \ STOPTASKS,${SSTATE_DIR},1G,100K \ ABORT,${TMPDIR},100M,1K \ Feb 03 13:05:51 Olegt: I think you need to read the manual before Feb 03 13:06:19 Olegt: you are saying you want to build for MACHINE ??= "qemux86 Feb 03 13:10:02 mckoan: thanks. I tried to build for qemumips. I override it ( Feb 03 13:11:21 mckoan: I read a few manual but some points still unclean for me Feb 03 13:11:50 jku: adding the libs to FILES_SOLIBSDEV works. Now I get the expected .so links in the -dev package (exluding the one defined in FILES_${PN}). As you mentioned now I get a QA error: ERROR: QA Issue: non -dev/-dbg/nativesdk- package contains symlink .so Feb 03 13:12:22 mckoan: Do I have to create new metalayer or able to use recipes-kernel/linux instead of it? Feb 03 13:13:25 jku: could I suppress it just for the one .so defined in FILES_${PN}? or will the QA be disabled generally? Feb 03 13:14:23 Olegt: i never used mips, but I think that if you set MACHINE=qemumips you won't need to add any custom recipe, just try Feb 03 13:14:25 mrpelotazo: you can skip per package, grep for INSANE_SKIP in meta/ for examples Feb 03 13:22:36 mckoan: I use https://github.com/meta-debian/meta-debian. gcc4.9 doesn't support ARCH=mips32r5, so I updated its receipt. and I have to other task to install custom kernel. As first step I'm going to build qemumips, after that export images. Feb 03 13:22:53 mckoan: http://pastebin.com/kr4PWve7 local.conf and linux-my-kernel one more time Feb 03 13:27:17 jku: its actually mentioned in the doc: http://www.yoctoproject.org/docs/2.0.3/ref-manual/ref-manual.html#var-INSANE_SKIP Feb 03 13:28:01 jku: works fine, and btw, thank you! Feb 03 13:29:12 Olegt: I sugegst you to ask at meta-debian@googlegroups.com Feb 03 14:43:16 Hello all, does anyone have any tips on debugging ipk rootfs install? I get the following error and it really doesn't make sense calculate_dependencies_for: Cannot satisfy the following dependencies for evince: * libgl * * opkg_solver_install: Cannot install package evince. Feb 03 14:46:49 when I check the info on the package evince I get the following depends: gstreamer1.0 (>= 1.8.3), hicolor-icon-theme, lcms (>= 2.7), libatk-1.0-0 (>= 2.20.0), libc6 (>= 2.24), libcairo-gobject2 (>= 1.14.6), libcairo2 (>= 1.14.6), libexpat1 (>= 2.2.0), libffi6 (>= 3.2.1), libfontconfig1 (>= 2.12.1), libfreetype6 (>= 2.6.5), libgcc1 (>= 5.4.0), libgdk-pixbuf-2.0-0 (>= 2.34.0), libglib-2.0-0 (>= 2.48.2), libglib-2.0-utils, libgnome-des Feb 03 15:59:13 I switched to using RPM and encountered the same issue but now with libgtk, I get the following error: Computing transaction...error: Can't install libgtk-3.0-3.20.9-r0@corei7_64: no package provides libgl Feb 03 15:59:46 I don't understand because I have created a nvidia recipe that provides virtual/libgl Feb 03 16:01:22 virtual/libgl is a build time provider. libgl is a binary package Feb 03 16:05:31 kergoth: do you mean evince is looking for a pacakge named libgl provides /usr/lib/libGL.so? Feb 03 16:10:25 kergoth: do I just need to add libgl as a provider to my nvidia recipe? Feb 03 16:18:28 tanuk: bonus points for license clarity :) Feb 03 16:19:27 Thanks :) Feb 03 16:20:32 hello Feb 03 16:35:07 tanuk: is the adrian canceller something we should put the effort of adding the license for? (adding a new license isn't hard) Feb 03 16:37:24 Maybe. It's not nice to remove functionality that used to be there, but on the other hand, my gut feeling is that very few people use it. Feb 03 16:37:42 god knows why fftw was disabled Feb 03 16:38:25 I haven't tried any of the echo cancellers myself, but from what I've heard the adrian canceller isn't very effective. Feb 03 16:40:18 I can make a patch for enabling the adrian canceller. Feb 03 16:43:21 kergoth: Thanks for the definition. I ended up adding libgl and others as a rprovides and that fixed my issue. I was just confused because I am updating from fido to morty and in fido I didn't have this issue. Feb 03 16:44:37 khem: Morning to you, I am looking at some MUSL related issues with meta-intel and found that strace (in OE-Core) is not building for x86-64, it seems to build for x86. Also are you aware of any work-arounds for argp.h usage? Feb 03 16:47:58 tanuk: well if there's a better option then leaving it off seems totally sensible Feb 03 16:48:10 sgw_: there's argp-standalone Feb 03 16:53:44 rburton: yes, there is also 5 years old and not maintained it appears Feb 03 16:56:06 halo Feb 03 16:56:21 can i receive istruction to install boost? Feb 03 16:56:31 there is somewhere a package? Feb 03 16:57:11 sgw_: for argp support depend on http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-support/argp-standalone/argp-standalone_1.3.bb?h=master Feb 03 16:57:18 [runqemu][raspberrypi2] ok I answer my own-self with this issue """No 'PCI' bus found for device 'virtio-net-pci'...""", virtio somewhat works better with "-machine virt" which could be stuffed inside QB_MACHINE (not forgetting QB_SYSTEM_NAME = "qemu-system-arm"). Feb 03 16:57:21 argp-standalone should provide what is needd Feb 03 16:57:36 sgw_: and for strace/x86_64 is there any discussion Feb 03 16:58:44 sgw_: it was building fine few weeks ago. So it must have regressed, I think we only do x86 in world builds on AB and it probably was uncaught Feb 03 16:58:52 khem: Ok, will look at argp-standalone, for strace I am seeing size_t and wchar_t conflicting types when building x86-64 but not x86 Feb 03 16:59:21 khem: likely, since I am starting to build world for meta-intel and saw that. Feb 03 17:00:26 sgw_: are there any starce modifications done via meta-intel ? Feb 03 17:00:36 is it normal that krogoth generates a meta-toolchain named *.2.0.1.sh istead of 2.1? Feb 03 17:00:43 khem: nope, I got the same failure with qemux86-64 Feb 03 17:02:16 khem: meta-intel itself has a couple of additional issue, which I am digging into Feb 03 17:02:23 and it announces at boot : Poky (Yocto Project Reference Distro) 2.0.1 beaglebone /dev/ttyO0 Feb 03 17:03:59 rburton: Ok, I won't make that patch then. Feb 03 17:06:44 * mckoan maust have any branch mismatch... Feb 03 17:06:54 s/maust/must Feb 03 17:16:58 Hi. Poky doesn't support gstreamer1.0 (v1.10), the latest version is 1.8, what can I do to get it to build to 1.10? Feb 03 17:18:53 deivd___, increase the recipes version, build and hunt the errors Feb 03 17:27:03 rob_w sorry. I'm learning poky Feb 03 17:27:19 on my recipe I have "http://gstreamer.freedesktop.org/src/gstreamer/gstreamer-${PV}.tar.xz" but $PV is not set there Feb 03 17:27:21 no worries , thats how you learn bet Feb 03 17:27:24 best Feb 03 17:27:37 its the name of the recipe file .. Feb 03 17:27:50 which is ? Feb 03 17:27:56 gstreamer1.0_1.8.3.bb Feb 03 17:28:18 so copy taht to like gstreamer-1.0_1.10.0.bb Feb 03 17:28:31 I jsut did Feb 03 17:28:31 bitbake-layers show-recipes gstreamer1.0 Feb 03 17:28:32 Parsing recipes..done. Feb 03 17:28:32 === Matching recipes: === Feb 03 17:28:32 gstreamer1.0: Feb 03 17:28:32 meta 1.10.2 Feb 03 17:28:34 lovely Feb 03 17:28:40 have fun Feb 03 17:29:02 Thanks. The documentation hit me like a truck, there's too much to learn and no 'easy' starting point Feb 03 17:29:39 its like always, go slow but then u will get the hang of it ,, its very well structured Feb 03 17:30:02 and its all scripts, so you can grep and read into them Feb 03 17:31:08 although gstreamer is a rather big pile of packages .. it might just work .. but there is a fair chance that patches against 1.8 wont apply against 1.10 Feb 03 17:32:00 I'll see. For now I'll fix up the checksums Feb 03 17:33:06 I'm tired of building my images by hand, so I'll see if I can get my iamge working with poky Feb 03 17:35:50 if you get poky under your hood , trust me you will prevail !! using it for years now and it keeps me employed, fed and happy !! ,.. need to run now , have fun Feb 03 17:47:36 sgw_: what error message do you see Feb 03 17:51:03 khem: I will email you the cooker log Feb 03 18:03:24 sgw_: pastebin it Feb 03 18:06:26 khem: http://pastebin.com/bK5wqSgU Feb 03 18:11:51 sgw_: I see Feb 03 18:12:33 I think I have perused this issue in past, the problem is really in strace how its trying to grok x32 and 32bit syscalls along Feb 03 18:13:04 cant find the IRC discussion we had on it on musl channel Feb 03 18:16:11 size_t really depends on bitness of compiler Feb 03 18:16:30 and our headers are not multilibbed Feb 03 19:43:06 tslib 1.4 released Feb 03 19:43:07 * kergoth yawns Feb 03 19:54:53 Friday dilemma. Too late for another coffee? Too early to head to the pub? ;-) Feb 03 19:56:08 to early? Feb 03 19:56:14 only if it's 7am Feb 03 19:56:19 and you've just woken up Feb 03 20:16:59 not sure there is a TZ where it is 7am currently... Feb 03 20:17:44 at least one that is not in the middle of an ocean. Feb 03 20:20:37 paulg: its pubtime somehwere so just follow it Feb 03 20:24:23 who am I to disagree.... Feb 03 21:02:16 Any meta-intel folk around, by chance? specifically with experience with minnowmax Feb 03 21:04:07 kergoth: I'm here, what do you need? Feb 03 21:04:52 ever run into one of the cpus/cores not being online? cpu1 is online, cpu0 is not, via examining /sys/devices/system/cpu/cpu[01]/ Feb 03 21:05:47 kergoth: outside of purposely disabling a core, I haven't, sorry Feb 03 21:05:55 np Feb 03 21:06:09 just wanted to make sure it wasn't a known issue or something, so figured i'd run it by folks Feb 03 21:06:22 going to check master and see if the behavior is the same as with morty, next Feb 03 21:10:32 kergoth: hmm, my minnowboard is showing the same thing, might be a kernel bug because I can turn core1 off and it still runs Feb 03 21:10:43 echo 0 > /sys/devices/system/cpu/cpu1/online Feb 03 21:11:59 kergoth: nevermind, its likely by design that we can't turn 0 off Feb 03 21:12:01 https://www.kernel.org/doc/Documentation/cpu-hotplug.txt Feb 03 21:13:25 Can somebody remind me which package provides /etc/host in a target rootfs, or how to find out? Package management is off in my current rootfs, so I can't use opkg or similar. Feb 03 21:14:46 kergoth: try adding cpu0_hotplug to the kernel command line Feb 03 21:16:09 clsulliv: ah, makes sense. thanks. Feb 03 21:16:23 It's not base-files :-/ Feb 03 21:17:08 Bingo! meta/recipes-core/netbase/netbase/hosts Feb 03 22:46:31 sgw_: I have made a gcc patch to fix this musl strace issue. see https://patchwork.openembedded.org/patch/136766/ Feb 03 22:46:41 sgw_: I would appreciate some testing Feb 03 22:50:37 khem: will do, thanks Feb 03 22:53:18 khem, have you seen any issues with SIGEV_THREAD_ID or major/minor declarations? Feb 03 23:16:06 sgw_: yes SIGEV_THREAD_ID is internal in musl Feb 03 23:18:11 sgw_: what errors do you see related to SIGEV_THREAD_ID Feb 03 23:27:22 khem: error: 'SIGEV_THREAD_ID' undeclared (first use in this function) Feb 03 23:27:22 sev.sigev_notify = SIGEV_SIGNAL | SIGEV_THREAD_ID; Feb 03 23:27:22 Must mean I need to add some missing header file, thanks for that info Feb 04 00:14:17 sgw_: yeah thats common outcome of making it private Feb 04 00:14:31 which package is showing this ? Feb 04 00:16:56 khem: intel-gpu-tools Feb 04 00:37:06 so firstly its non-posix code Feb 04 00:37:21 to access SIGEV_THREAD_ID Feb 04 00:44:00 sgw_: it should do something like this http://git.alpinelinux.org/cgit/aports/plain/main/qemu/fix-sigevent-and-sigval_t.patch Feb 04 00:44:40 khem: thanks for the pointer, I will work on this for next week, running out of steam today! Feb 04 00:46:59 np have a good weekend **** ENDING LOGGING AT Sat Feb 04 03:00:02 2017