**** BEGIN LOGGING AT Tue Nov 08 03:00:01 2016 **** BEGIN LOGGING AT Tue Nov 08 05:17:46 2016 Nov 08 08:48:05 has somebody any experience with building the ligfortran library? Nov 08 08:48:20 in the newest yocto-master it is not building Nov 08 08:54:56 i can only say, i tried to look at the fortran stuff for writing a recipe for R some time back. I failed so totally and miserably that I suspect fortran in OE to be the incarnation of pure evil now. Nov 08 08:59:27 :DD I can fully understand you. I really don't see a problem on my site. You habe the libgfortran_6.2.bb recipe and you have to add the following lines in the local.conf to actually build it: Nov 08 08:59:33 FORTRAN_forcevariable = ",fortran" Nov 08 08:59:38 RUNTIMETARGET_append_pn-gcc-runtime = " libquadmath" Nov 08 09:00:51 then it actually starts building, but it says it is missing header files, which are actually provided in the WORKDIR (backtrace-supported.h) Nov 08 09:01:29 I just don't know why is somebody releasing such a recipe, if it doesn't work Nov 08 09:02:23 I always think there is maybe a little line of code I missed to add or an append file, but libgfortran should work out of the box Nov 08 09:03:07 i guess it at least worked for some use case or point in time. but seeing so little attention and testing, a certain 'decay' is inevitable Nov 08 09:04:24 Yeah, but I mean even the 5.4 version is not building Nov 08 09:05:51 can't offer much beyond understanding for your problem, sorry. Nov 08 09:08:45 no problem, thx for your heartfelt sympathy Nov 08 09:30:22 benz: fortran isn't tested by many people so it can break easily. patches obviously welcome, and if you want to own fortran support in oe then you're welcome - i don't think anyone else does Nov 08 09:36:03 would one really want to own that? ;-) Nov 08 09:37:25 Probably not.. Nov 08 09:38:34 Especially in this project which is aimed at embedded devices .. Fortran is still mostly being used for in the HPC world.. Nov 08 09:39:35 i.e. most of the people who are left who still use Fortran are in the business of HPC Nov 08 09:40:28 actually if i had the time to properly takcle it, i even might be interested. but given real life, it would be a hack-into-wahtever shape, which would then only be frustrating for everybody involved. Nov 08 09:40:42 on the other hand... Nov 08 09:40:51 maybe i'll have another look. just because i can :=) Nov 08 09:41:09 If you have success let me know it ;) Nov 08 09:41:32 I got the 5.4 recipe to work with some little appends :) Nov 08 09:42:21 benz: doing what exactly? Nov 08 09:44:20 I used the 2 lines I mentioned in the local.conf and added the not shipped files to the FILES_${PN} variable Nov 08 09:45:16 patch please :) Nov 08 09:49:18 I think somebody should first patch 6.2, that would be way better since this is the latest recipe Nov 08 09:58:55 benz: you mean morty? or master? Nov 08 09:59:10 master Nov 08 09:59:25 ok, lets see Nov 08 09:59:40 hi Nov 08 10:48:49 hello guys, could somebody help me with this issue: http://ix.io/1Dpv ? Nov 08 10:50:48 Ox4`: mind giving a hort description before pointing people to random links they shall open just because your nickname gives such a trustworthy impression? Nov 08 10:52:03 ok, I have a problem whilst running bitbake command here ^^^^ Nov 08 10:53:08 * LetoThe2nd decides not to open, then. Nov 08 10:56:29 seriously, i'd at least expect enough effort to know if some custom recipe of yours doesn't build, some layer is broken, bitbake crashes, you are trying to run it on windows.... whatever. Nov 08 10:57:14 you also don't walk into a car shop and tell the mechanic: "hey here is my car, i have aproblem with it" and then expect it to be properly handled. Nov 08 10:57:18 Nov 08 11:55:38 benz: with the said modifications to local.conf and IMAGE_INSTALL_append = " libfortran" a build of master succeeded here. do i need to extend the testcase somehow? Nov 08 12:52:00 rburton: maybe ask cisco for an accounting number :-) Nov 08 13:03:22 LetoThe2nd: :) Nov 08 13:03:43 LetoThe2nd: Did you try it with qemux? I'm building for "intel-cori7-64". So the command is: MACHINE=intel-corei7-64 bitbake libgfortran Nov 08 13:03:46 rburton: no bashing, justsaying ;-) Nov 08 13:04:08 benz: target is qemux86 Nov 08 13:05:07 benz: benz ah ok, bitbake libgfortran breaks. Nov 08 13:06:21 LetoThe2nd: What did you try? core-image-minimal? Nov 08 13:06:28 benz: yep Nov 08 13:08:19 Can you try it with bitbake libgfortran? I'm trying right now MACHINE=qemux86 bitbake libgfortran Nov 08 13:09:10 seems i had some misconception about IMAGE_INSTALL. bitbake libgfortran breaks with something backtrace, looking at it now. Nov 08 13:10:02 yeah with backtrace-support.h Nov 08 13:10:22 right? Nov 08 13:10:29 exactly. now lets see where that comes from. Nov 08 13:12:48 I found the header in the workdir: work/.../libgfortran/6.2.0-r0/gcc-6.2.0/build.x86_64-poky-linux.x86_64-poky-linux/libbacktrace Nov 08 13:13:20 benz: i'd rather suspect that it only gets built conditionally Nov 08 13:13:25 When I added in CFLAGS it failed at another library... Nov 08 13:14:08 Could be, let me know if you have something new or questions Nov 08 13:15:24 benz: see https://github.com/gcc-mirror/gcc/blob/master/Makefile.def#L592 Nov 08 13:16:33 benz: i'm giving it a shot now with RUNTIMETARGET_append_pn_gcc-runtim = " libquadmath libbacktrace" Nov 08 13:17:22 I think I tried the same, but it also failed Nov 08 13:17:41 lets see :) Nov 08 13:20:58 I tried it again, but the same error occured Nov 08 13:28:14 awww crap Nov 08 13:28:26 :( Nov 08 13:28:37 oh i get to blame josh! Nov 08 13:28:41 joshuagl: ^ Nov 08 13:29:41 I am once more in need of some brain power on this side. How does one add a user to the sudoers file properly, I've managed to add user but that user has no sudo power. Nov 08 13:30:02 Strike5150: no helpers, you'll have to write to /etc/sudoers manually Nov 08 13:30:04 The sudoers file seems to be generated so I can't append it Nov 08 13:30:40 if you have a specific package that installs the user, you could also write to sudoers then Nov 08 13:31:20 rburtoncan you simply echo / append the sysroot version of sudoers to bake it into the image? Nov 08 13:31:28 rburton: can you simply echo / append the sysroot version of sudoers to bake it into the image? Nov 08 13:32:10 rburton: I sometimes see recipes play with ${D} files, is this what your suggesting? Nov 08 13:32:18 Strike5150: in a pkg_postinst would work better Nov 08 13:32:37 rburton: Ok that makes more sense Nov 08 13:32:53 rburton: thanks! Nov 08 13:35:58 rburton: I realize now the best way is to add a sudoers file under /etc/sudoers.d, the reference image is configured to grab drop in files from there. Nov 08 13:50:44 LetoThe2nd: And, did it work out for you? Nov 08 13:53:39 benz: nope, same error. looking at the recipes suggest that there is some work needed, so i'll look into that laters. Nov 08 14:33:35 dummy question : I get the warning saying debian8 is not validated with my version of bitbake, it's been working great so far, is there some ongoing work to validate debian? Should I better switch to ubuntu, or maybe it doesn't matter much Nov 08 14:38:24 mathieu_la: which yocto version? Nov 08 14:38:53 current one shouldn't complain about Debian 8 Nov 08 14:39:03 weird thing is I see "Debian-8.*" under SANITY_TESTED_DISTRO and I still get the warning (using krogoth) Nov 08 14:39:13 in poky.conf i mean Nov 08 14:41:23 mathieu_la: interesting... in any case debian jessie is fine Nov 08 14:43:12 OK thanks Nov 08 14:44:00 oh actually my Debian is called "DebianGNULinux-8" which does not match the regex Nov 08 14:44:36 so nevermind Nov 08 14:50:46 I have patches on the list for this Nov 08 14:51:31 it depends which of the sources of distro identification you have available, we currently prefer `lsb_release -ir` and fall back to /etc/os-release, it sounds like your system is using os-release mathieu_la Nov 08 14:52:08 my patches should a) prefer os-release and b) be more consistent regardless of source on most distros (not openSUSE, unfortunately) Nov 08 14:52:56 it's still a bit weird, I don't see the warning (on jessie, although yocto master)... Nov 08 14:55:48 jku: do you have lsb_release installed? Nov 08 14:55:56 I just tried Nov 08 14:56:08 joshuagl: yes Nov 08 14:56:09 and the warning disappears if I install lsb_release Nov 08 14:56:32 ok not weird anymore :) Nov 08 14:56:48 more details in https://bugzilla.yoctoproject.org/show_bug.cgi?id=10591 and the patch series Nov 08 14:56:49 Bug 10591: normal, Medium, 2.3, joshua.g.lock, IN PROGRESS REVIEW , oe.lsb.distro_identifier() doesn't always return the same value for CentOS 7 Nov 08 14:57:19 is /etc/os-release, I have NAME="Debian GNU/Linux" and ID=debian Nov 08 14:57:37 maybe looking for ID instead of NAME could solve it Nov 08 14:57:44 mathieu_la: see the patch series :-p Nov 08 14:58:02 http://lists.openembedded.org/pipermail/openembedded-core/2016-November/128532.html Nov 08 14:58:49 yes nice! Nov 08 15:55:48 Hi, Is there a criteria for which machines/configurations get built by the Yocto Autobuilder? Nov 08 15:56:06 I was wondering if I could add a configuration or two Nov 08 15:58:18 stephano: I've got those manuals, and I've repeatedly tried to use "devtool add" on the proprietary repo I have to package. It gives me a CMake-based recipe, which is correct, but during the build process it downloads CppUTests, compiles CppUTestTests, and misplaces the file so that it then fails its own testing phase with "no such file or directory". Nov 08 15:58:19 ZubairLK: suggestions welcome if the impact isn't too bad. adding whole new machines would be something to discuss with the advisory board. Nov 08 15:58:59 Which is the same error I get from trying to build in the root of the source tree rather than in a "build/" subdir, but the recipe seems to ignore any B="..." directives I add and behave just the same way. Nov 08 16:07:55 rburton: thanks Nov 08 19:36:55 hi there Nov 08 19:37:23 I was trying to create a core-image-x11 image with "morty" in Ubuntu 16.04 Nov 08 19:37:55 but I'm getting an error in the last step when creating the image Nov 08 19:38:15 something related with prelink Nov 08 19:38:57 it is trying to prelink some folders that are not part of the image Nov 08 19:39:10 /lib/xtables , /lib/init for instance Nov 08 19:39:25 any idea? Nov 08 19:51:32 Spulit: does it work if you remove 'image-prelink' from USER_CLASSES in local.conf Nov 08 19:51:44 yap Nov 08 19:52:07 khem: yes, but is prelink really needed? Nov 08 19:52:46 it is not Nov 08 19:53:00 it speeds up your load time Nov 08 19:53:12 so it might improve your boottime a bit Nov 08 19:53:16 its an optimization Nov 08 19:53:31 however if you care for ASLR and hardened systems you should not use it Nov 08 19:57:57 khem: nevertheless this should be fixed somehow...I don't know if it is due to the fact of using Ubuntu 16.04 Nov 08 19:59:36 Spulit: agreed, I think you can file a ticket Nov 08 19:59:51 in bugzilla with intruction on reproducing it Nov 08 20:27:57 anyone with a recent ubuntu? is ld actually gold? Nov 08 21:30:18 rburton: not on 16.10 Nov 08 21:30:20 /usr/bin/x86_64-linux-gnu-ld -> ld.bfd Nov 08 22:07:36 are the errors about missing sstate *.siginfo files expected? sstate seems to work fine, but some tasks don't have siginfo and then I see errors trying to fetch/access those files... Nov 09 00:00:07 I started a custom board and now I have no kernel in /boot and I'm trying to figure out why. It looks like it's being built and placed in /tmp/deploy/images, but not added to the rootfs Nov 09 00:00:38 It's just a clone of the Beaglebone device for a similar device Nov 09 00:13:55 I think I got it Nov 09 00:14:32 I was missing .bbappends with srcrev etc Nov 09 00:43:36 nope, still nothing Nov 09 02:47:01 denix: there always should be .siginfo files Nov 09 02:47:07 for a give sstate Nov 09 02:58:15 khem: I don't seem to have them for some tasks... wonder why **** ENDING LOGGING AT Wed Nov 09 03:00:00 2016