**** BEGIN LOGGING AT Fri Nov 17 03:00:03 2017
Nov 17 03:59:55 <yocti> New news from stackoverflow: how to find out what recipe failed <https://stackoverflow.com/questions/47343009/how-to-find-out-what-recipe-failed>
Nov 17 08:54:15 <sagner> nayfe: since it works when using libcrypto from morty in a rocko rootfs (just overwrite the existing one) it really looks like a pure OpenSSL issue
Nov 17 08:55:34 <nayfe> sagner: yep indeed, did you check why file://fix-cipher-des-ede3-cfb1.patch patch is not present in openssl10 ? didn't have time to yesterday night :)
Nov 17 08:55:45 <sagner> nayfe: in my rocko build environment I now reverted openssl to the state it used to be in morty. But that still fails! So... either some dependency of OpenSSL... or compiler?
Nov 17 08:56:30 <sagner> nayfe: with that build fix-cipher-des-ede3-cfb1.patch is present, but still fails.
Nov 17 08:58:38 <nayfe> for now, i'm struggling with sysvinit to systemd migration, don't have much time to investigate more on this issue :(
Nov 17 09:14:56 <Tamis> Does anyone knows why the ant recipe at meta-java compiles the files with commands and does not invoke the provided build.sh which will produces the build?
Nov 17 09:15:53 <nayfe> tamis: maybe to customize some stuff?
Nov 17 09:16:27 <LetoThe2nd> or maybe the build.sh is not properly cross-compile/containment aware
Nov 17 09:17:13 <Tamis> I am trying to upgrade the recipe because the provided version is obsolete.
Nov 17 09:17:57 <Tamis> I succeeded to build the ant version 1.10.1 with the build.sh in devshell
Nov 17 09:18:26 <Tamis> but trying to reuse the recipe with the commands inside and it fails....
Nov 17 09:18:30 <LetoThe2nd> to most common case is simply that the build scripts provided by the projects work well for their usecase, but fail miserably when it comes to cross-compiling or canadian cross situations.
Nov 17 09:18:53 <LetoThe2nd> (just another reason to use a preexisting and tested system instead rolling your own)
Nov 17 09:20:28 <Tamis> hmmm I will have to adjust the recipe accordingly then. I will try to do it.
Nov 17 09:30:46 <yocti> New news from stackoverflow: opkg_conf_parse_file: /etc/opkg/opkg.conf:2: Ignoring invalid line: `lists_dir ext /var/lib/opkg/lists' <https://stackoverflow.com/questions/47346843/opkg-conf-parse-file-etc-opkg-opkg-conf2-ignoring-invalid-line-lists-dir-e>
Nov 17 09:32:00 <nayfe> sagner: maybe compare openssl compile logs between morty yocto and rocko ?
Nov 17 09:32:39 <nayfe> maybe some global distro policy changes
Nov 17 10:14:40 <Tamis> finally I succeed to build ant version 1.10 with the recipe
Nov 17 12:04:59 <sagner> nayfe: built now pure Yocto poky rocko for ARMv7 and started it with emulator, and it works!
Nov 17 12:05:17 <sagner> nayfe, otavio: so it must be something meta-freescale*
Nov 17 12:56:59 <otavio> sagner: uh???
Nov 17 13:05:54 <sagner> otavio: when ssh into a rocko image using OpenSSH it fails with: ssh_dispatch_run_fatal: Connection to 192.168.7.2 port 22: incorrect signature
Nov 17 13:08:32 <sagner> otavio: that is with fslc distro, not with poky... afact fslc uses Thumb2
Nov 17 13:09:37 <nayfe> sagner in openssl10.inc there is a switch case with archs, maybe it goes to the wrong one ?
Nov 17 13:11:25 <nayfe> you confirm that PACKAGECONFIG contains cryptodev-linux  ?
Nov 17 13:16:19 <sagner> nayfe: openssl10.inc switch looks good, both log.do_configure claim target is linux-armv4 (which seems weird, but it used to be the case in morty too
Nov 17 13:17:28 <kanavin> sagner: nayfe: why not bisect ?
Nov 17 13:19:28 <sagner> kanavin: I am not there yet, currently I have not two git hashes in the same repo where it is working/non-working
Nov 17 13:20:03 <kanavin> sagner: but if it worked in morty?
Nov 17 13:20:38 <kanavin> bisection over n-dimensional space of commits is tricky, I grant you :)
Nov 17 13:20:46 <sagner> kanavin: well, poky it works in rocko.. it seems to be related with fsl-community-bsp. And that is repo, I don't think I can bisect with that
Nov 17 13:21:10 <kanavin> sagner: so you have not been able to reproduce with pristine poky?
Nov 17 13:21:23 <kanavin> (e.g. no weird layers)
Nov 17 13:21:51 <nayfe> fsl are not weird :D
Nov 17 13:22:32 <nayfe> kanavin> indeed, sagner "built now pure Yocto poky rocko for ARMv7 and started it with emulator, and it works!"
Nov 17 13:23:41 <sagner> kanavin: not so far no... But meta-freescale* seem not to alter openssl/openssh, so it must be flags or something. fslc uses Thumb2...
Nov 17 13:24:15 <sagner> Yeah first thing I tried was building for qemuarm with fslc, that worked, but that again uses completely different compiler flags...
Nov 17 13:24:32 <sagner> Then I created qemuarmv7, which failed in fslc, and seems to work in pure poky
Nov 17 13:25:50 <kanavin> then you need to find a point in fslc where things still work
Nov 17 13:27:06 <sagner> kanavin: yes... just time consuming.. meanwhile, I am taking bets... :-)
Nov 17 13:27:23 <kanavin> and generally fslc should QA things better... like run our QA tests on real hardware, including this:
Nov 17 13:27:23 <kanavin> http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/lib/oeqa/runtime/cases/ssh.py
Nov 17 13:27:45 <kanavin> ^^^ this is run under qemu arm daily
Nov 17 13:28:55 <otavio> kanavin: patches welcome ;-)
Nov 17 13:29:10 <otavio> sagner: which machine are you using?
Nov 17 13:29:11 <kanavin> otavio: I guess you know who pays my salary :)
Nov 17 13:29:29 <otavio> kanavin: hheheh
Nov 17 13:30:07 <otavio> sagner: which machine and which distro?
Nov 17 13:32:34 <sagner> otavio: well, we noticed it in all our machines (apalis/colibri-imx6/7)...
Nov 17 13:33:03 <nayfe> otavio: same issue with custom imx6dl on rocko custom distro based on fslc
Nov 17 13:33:21 <sagner> otavio: then I built a qemuarm but using arch-armv7a.inc
Nov 17 13:35:00 <sagner> otavio: So initially seen on Angstrom, then I build pure bsp-community-bsp using fslc-wayland distro
Nov 17 13:37:44 <otavio> sagner: so please see if just changing the distro it fixes
Nov 17 13:37:55 <sagner> My bet is on GCC 7.2 + Thumb2 + OpenSSL...
Nov 17 13:38:18 <otavio> it is a long shot
Nov 17 13:38:20 <sagner> otavio: just did that, setting poky fixes it.
Nov 17 13:38:33 <otavio> sagner: so it is not it
Nov 17 13:38:53 <otavio> sagner: see what the distro changes.
Nov 17 13:39:02 <otavio> sagner: is it fslc-framebuffer?
Nov 17 13:40:24 <sagner> otavio: why not Thumb2? That is different: armv7ahf-vfp-poky-linux-gnueabi vs. armv7at2hf-vfp-fscl-linux-gnueabi
Nov 17 13:41:00 <otavio> it might be; but you can prove it removing the tune
Nov 17 13:41:24 <sagner> otavio: hm, or adding it with pure poky
Nov 17 13:41:31 <otavio> sagner: yes
Nov 17 13:42:03 <otavio> if this is the case, forcing arm for openssl might be a workaround
Nov 17 13:43:33 * otavio goes for lunch
Nov 17 14:29:37 <wto> I'm getting a strange issue when building an SDK containing both cmake and qmake that rpm tells me it's conflicting installation files in environment-setup.d.
Nov 17 14:30:21 <wto> Shouldn't this path be writable by more than one package? Normally it will be merged later on, right? They never write the same file, they just create the directory to write a cmake.sh and qt5.sh into it
Nov 17 14:30:33 <wto> This is on pyro btw
Nov 17 15:11:03 <sagner> otavio: it seems to be Thumb2. Removing all the tune from fslc distro makes it work too
Nov 17 15:23:48 <sagner> otavio: ARM_INSTRUCTION_SET = "arm" in openssl10.inc does not help, it seems that distro overwrites it
Nov 17 15:32:57 <sagner> otavio: it seems that because fslc-base.inc uses a distro specific override it actually overrides ARM_INSTRUCTION_SET of a recipe...
Nov 17 15:42:30 <nayfe> With systemd, systemd-hostnamed.service has PrivateNetwork=yes, and it fails to load on my system with this option. Kernel issue ?
Nov 17 15:44:48 <sagner> nayfe: How does it fail to load
Nov 17 15:45:34 <nayfe> Process: 9942 ExecStart=/lib/systemd/systemd-hostnamed (code=exited, status=225/NETWORK)
Nov 17 15:46:19 <nayfe> i think it tries to run it without network rights, but i'm not sure why
Nov 17 15:46:36 <nayfe> i'm running on 4.1.15, does it support this systemd option ?
Nov 17 15:49:04 <otavio> sagner: try ARM_INSTRUCTION_SET_local = arm
Nov 17 16:01:34 <nayfe> sagner> for systemd problem, i didnt compile kernel with namespace support :)
Nov 17 16:14:18 <sagner> otavio, nayfe: its binutils 2.29 which causes problems
Nov 17 16:14:36 <sagner> most likely...
Nov 17 16:22:06 <otavio> khem: ^
Nov 17 16:27:17 <sagner> ok, got it its that: https://github.com/openssl/openssl/issues/4659
Nov 17 16:27:56 <sagner> basically binutils 2.29 broke OpenSSL when compiled Thumb2
Nov 17 16:28:33 <sagner> poky did not compile with Thumb2 so all worked there, and everything before rocko used older binutils so it worked too
Nov 17 16:34:02 <otavio> sagner: great; are you going to prepare the fix for master and rocko?
Nov 17 16:34:19 <sagner> otavio: yes
Nov 17 16:34:37 <otavio> sagner: great; please add me on Cc so I can ack them
Nov 17 16:34:46 <sagner> otavio: will do
Nov 17 16:34:48 <otavio> sagner: for master, bumping openssl seems best
Nov 17 16:34:59 <otavio> sagner: rocko, adding the patch is better
Nov 17 16:35:13 <sagner> otavio: it is not in a release yet..
Nov 17 16:35:15 <otavio> sagner: https://github.com/openssl/openssl/pull/4673/commits/60ae30077da071eaeaf159917519c0a458045757
Nov 17 16:36:09 <otavio> indeed
Nov 17 16:36:18 <sagner> otavio: but there is 1.0.2m, but the patch is after m
Nov 17 16:36:37 <sagner> otavio: so on master I could move to 1.0.2m and add to patch, and only add the patch in rocko..
Nov 17 16:36:41 <otavio> sagner: I saw that too
Nov 17 16:36:45 <otavio> sagner: yes
Nov 17 16:37:00 <otavio> sagner: and also need to patch 1.1.0
Nov 17 16:37:18 <otavio> both branches are affected
Nov 17 16:37:38 <sagner> hm, yes
Nov 17 16:37:41 <sagner> so 4 patches
Nov 17 16:37:45 <otavio> yes
Nov 17 16:40:43 <sagner> otavio: is it ok to just format-patch from upstream and add Upstream-Status: Backport tag in the commit message manually?
Nov 17 16:42:31 <otavio> sagner: sure
Nov 17 16:42:48 <otavio> if you add .patch suffix to the uri it formats it for you
Nov 17 17:24:16 <sagner> otavio: afaik those are pure CVE releases, maybe something we should backport to morty?
Nov 17 18:02:20 <yocti> New news from stackoverflow: Yocto build not including files in sdcard image <https://stackoverflow.com/questions/46833545/yocto-build-not-including-files-in-sdcard-image>
Nov 17 19:40:50 <nayfe> sagner> well done!
Nov 17 20:02:40 <yocti> New news from stackoverflow: How to STANDALONE execute the following bitbake script in DISTRO="poky" context? <https://stackoverflow.com/questions/47358324/how-to-standalone-execute-the-following-bitbake-script-in-distro-poky-context>
Nov 17 20:32:46 <yocti> New news from stackoverflow: Removing cmake files from yocto recipe build <https://stackoverflow.com/questions/47358540/removing-cmake-files-from-yocto-recipe-build>
Nov 17 21:56:35 <batman_> Hello, I am trying to build a build-appliance-image and I am running into " build-appliance-image-15.0.0-r0 do_rootfs: locale-base-en-us not found in the feeds "
Nov 17 21:56:58 <batman_> can someone please point me in the right direction?
Nov 17 22:00:58 <batman_> I have tried adding GLIBC_GENERATE_LOCALES = "1" in my local.conf. But, the build still breaks with the same error
Nov 17 22:50:39 <kergoth> batman_: GLIBC_GENERATE_LOCALES is a list of locales to generate, or 'all', iirc. something like that. it's not yes or no
Nov 17 22:50:56 <kergoth> IMAGE_LINGUAS and GLIBC_GENERATE_LOCALES will need to be aligned, of course
Nov 17 22:59:29 <nerdboy> hey, i made another foss manifest repo for udoo neo
Nov 17 23:00:40 <nerdboy> uses poky/oe with slightly patched meta-udoo and mainline/foss kernel/u-boot/etnaviv (graphics optional ;)
Nov 17 23:02:22 <nerdboy> https://github.com/sarnold/foss-udoo-neo-platform-bsp <= only has morty branch so far
Nov 17 23:03:46 <nerdboy> oh, and imx-rproc driver from patchwork for cortex-M4
Nov 17 23:04:19 <nerdboy> so, yipee hoo-haw for mainline...
Nov 17 23:06:29 <batman_> kergoth: Thanks, I just set IMAGE_LINGUAS  to be empty
**** ENDING LOGGING AT Sat Nov 18 03:00:01 2017