**** BEGIN LOGGING AT Fri May 06 02:59:58 2016 May 06 03:00:00 maybe with extra D's ? May 06 03:00:31 bitbake -S printdiff is usually the best bet May 06 03:00:47 once it's built from source, though, it'll be too late to do that, since it'll be comparing against the new sstate archive rather than the old May 06 03:03:30 on a different topic, is there a ptest patch somewhere for krogoth? May 06 03:03:42 what do you mean? May 06 03:03:48 to do what? May 06 03:04:29 ERROR: sed-4.2.2-r0 do_package: QA Issue: split_and_strip_files: 'file /home/sarnold/fsl-community-bsp/build-imx233mini/tmp/work/armv5e-poky-linux-gnueabi/sed/4.2.2-r0/package/usr/lib/sed/ptest/testsuite/utf8-2.inp' failed [split-strip] May 06 03:04:46 a whole bunch of those May 06 03:04:55 sed, libxml2, and strace May 06 03:05:09 Summary: There were 28 ERROR messages shown, returning a non-zero exit code. May 06 03:05:43 both meta-ti/poky and fsl stuff May 06 03:06:38 at least it's not 28 different packages, but it only takes one... May 06 03:07:10 from what someone was talking about on irc, it was a bug in the file command on the host, on certain host distros May 06 03:07:20 i can't recall the details, though. what are you building on? May 06 03:08:24 BUILD_SYS = "x86_64-linux" NATIVELSBSTRING = "Gentoo" May 06 03:08:28 hardened May 06 03:10:35 been some changes in util-linux tools lately May 06 03:10:47 haven't seen anything weird with file May 06 03:13:12 trying to find a mention of it on the lists, but not finding anything. could try a build in a container or chroot of another distro to try to verify its host related, otherwise probably someone tomorrow will remember. i think most of europe is still asleep, and it's late evening in the US, so it'll be dead for a while yet May 06 03:13:22 course trying to search for 'file' isn't a fun exercise May 06 03:15:41 build-wandboard2/BUILD/work/cortexa9hf-vfp-neon-mx6qdl-poky-linux-gnueabi/chromium38/1.0-32-r10/git/src/third_party/binutils/Linux_x64/Release/bin/ld: error: obj.host/third_party/libvpx/source/libvpx/build/make/libvpx_obj_int_extract.obj_int_extract.o: unsupported reloc 42 against global symbol stderr May 06 03:15:46 does this ring a bell for anyone May 06 03:16:04 this ld binary is shipped by chromium May 06 03:16:15 I don't get why this fails May 06 03:17:19 non pic maybe? May 06 03:18:12 there is a separate libvpx that should work, it's not configured that way i take it? May 06 03:18:47 not sure chrome can do that... firefox can May 06 03:22:18 Task chromium38:do_compile couldn't be used from the cache because: May 06 03:22:20 We need hash c8625a78d9f37d0264e3909c36477cc2, closest matching task was eb1e7451534667639a6b2415a6af5bc9 May 06 03:22:22 Dependency on task chromium38chromium38.bb.do_configure was added with hash 49e8c068432f1cd58d9b019784ef5d00 May 06 03:22:24 Dependency on task chromium38chromium38.bb.do_configure was removed with hash aa82464f19a5e2ebb3b687b4f91ab123 May 06 03:22:34 can someone explain these hashes May 06 03:23:01 what is hashed - the compile output or configure output or the whole of the working directory of that package May 06 03:23:28 none of those May 06 03:23:35 we checksum the input, the metadata May 06 03:23:58 can I see in a human readable format the difference May 06 03:24:06 in this case the do_configure task changed, either due to one of its dependent tasks changing, or its task itself, or a variable it uses May 06 03:24:10 -S printdiff does that May 06 03:24:17 elsewhere in teh output May 06 03:24:28 that is what I see with printdiff May 06 03:24:29 it'll show the difference in variables May 06 03:24:58 printdiff doesn't output 4 lines, no May 06 03:25:44 that is all I can se May 06 03:27:15 http://hastebin.com/zafawoduli.hs May 06 03:28:44 kergoth: that was it May 06 03:29:10 brand new set of file patches i didn't have... May 06 03:40:32 ah May 06 05:32:56 heeen: it seems to me that your build of chromium is using internal version of binutils that chromium bundles May 06 05:33:19 heeen: and gcc is expecting newer binutils May 06 05:33:53 heeen: so you need to pass -Dlinux_use_bundled_gold=0 to chromium build May 06 07:21:51 khem: perhaps this is something that can be added to the recipe? May 06 08:20:01 Marex, ping, may I PM you? May 06 10:53:28 ant_work: yeah, sure May 06 12:50:53 dv_: patches are welcome, May 06 12:52:36 ok, I'll add this in a followup patch to what I already posted May 06 12:52:45 (currently looking at JaMa's issue) May 06 12:52:49 ok May 06 14:07:18 JaMa: the patch merged on master, for opencv, is right. I sent a proposal for backport, on top of what is at krogoth-next, for review. May 06 14:10:20 otavio: ok, thanks May 06 14:21:25 JaMa: when krogoth-next will be merged? May 06 14:27:28 soon May 06 14:35:37 otavio: how well does gold work with imx6 builds? May 06 14:35:51 dv: it works fine, why? May 06 14:36:02 I'll try to use it by default then May 06 14:36:12 since it works faster and uses much less RAM May 06 14:36:15 dv: as khem suggested, it is better to not use the internal one May 06 14:36:27 I mean in general, not just for chromium May 06 14:36:40 btw I am patching the recipes to force chromium to use the OE binutils & gold May 06 14:36:43 dv: I would love to learn what are the general advantages of it May 06 14:36:53 dv: good; makes sense May 06 14:37:05 dv: also it ensures our fixes are used May 06 14:37:07 lol May 06 14:37:13 heh, yes May 06 14:37:35 I'm still amazed that they bundle binutils.. May 06 16:34:52 JaMa: ping May 06 16:48:13 JaMa: FYI i was also looking to upgrate pigin recipe and I noticed one thing that the configure script is checking native python May 06 16:48:37 as you are upgrading piding so I thought I should share that info with you May 06 16:48:53 by native python I mean host python May 06 16:49:02 it was not picking python from sysroot May 06 16:50:12 checking for python May 06 16:50:12 configure:20729: found /usr/bin/python May 06 16:50:12 configure:20742: result: /usr/bin/python May 06 16:51:01 JaMa: kergoth recommended me to inherit from pythonnative.bbclass May 06 16:51:19 I thought you are working on upgrade so this info will help you to fix the problem May 06 17:10:54 Noor: well it's supposed to check native python, why do you think it needs python-native? May 06 17:11:55 Noor: the is an issue when trying to find python libs and headers, which was fixes I believe in that patch to use pkg-config May 06 17:12:10 Noor: but python as executable is just used in pidgin build to generate something May 06 17:12:31 and any native python should be good enough for that (it doesn't have to be the one from python-native recipe) like in other recipes May 06 17:13:20 noor was telling me the other day that it was using include paths retruned by 'python-config' and those were pointing at host paths May 06 17:13:30 if your pkg-config change addressed that, then we're probably fine May 06 17:13:49 noor went home for the night, i think. he's in lahore May 06 17:14:11 http://lists.openembedded.org/pipermail/openembedded-devel/2016-May/107341.html May 06 17:15:53 * kergoth nods May 06 17:15:56 should be good hten May 06 17:16:17 think he just wanted to make sure it was addressed one way or another May 06 17:16:43 I was just confused by what he said, because configure is also checking python as executable used during the build May 06 17:17:05 and that's fine to be /usr/bin/python May 06 17:17:43 btw: your nick is causing a lot of typos since Yocto 2.1 release :) May 06 17:20:16 cantarell-fonts_git.bb is also missing an inherit pkgconfig May 06 17:24:04 JaMa: heh, that doesn't surprise me, they possibly should have picked a different name :) i keep having to do a double-take when i see krogoth in email subject lines :) May 06 18:03:39 on second thoughts may be kergoth was more familiar name for release name :) May 06 18:18:01 JaMa: kergoth: I am back for some time :) May 06 18:18:34 JaMa: yeah I saw your commit May 06 18:20:01 I was thinking that as we are builiding python native so configure script should check that instead of host python May 06 18:21:05 if it is perfectly fine to use the host python then I am fine with it May 06 18:21:21 I never saw such behaviour that is why I brought it up May 06 19:16:54 Noor: I actually think we could avoid needing python-native quite so much, originally it was only added to allow the cross-compilation of python itself. we could use PYTHONPATH or similar to point at the sysroot.. but there would be more dependency on the sanity of teh host python that way.. course, it needs to be sane for bitbake anyway, so.. May 06 20:06:30 hmm this wandboard has two sd slots May 06 20:07:00 in u-boot it boots from one, but then expects root on the other May 06 20:52:31 Does https://github.com/kergoth/meta-kergoth-wip/blob/master/meta-minimal-exports/conf/export.inc seem reasonably sane? breaking it into categories makes it easier to bring back certain ones in recipes to fix any issues there. May 06 21:22:20 heeen: I don't think it happens when using meta-fsl-arm-extra, does it? If so, please prepare a fix for U-Boot May 06 23:55:22 otavio: it seems after the kernel comes up it has a different number May 06 23:55:32 mmcblk0 becomes mmcblk2 **** ENDING LOGGING AT Sat May 07 02:59:58 2016