**** BEGIN LOGGING AT Wed Apr 25 03:00:06 2018 Apr 25 05:32:07 Hello there, wanted to add mongoDB to my yocto image, not getting anything on the internet, can anyone help? Apr 25 06:22:24 New news from stackoverflow: bitbake fails at the simplest recipe Apr 25 06:52:30 New news from stackoverflow: do_rootfs function failed in yocto project Apr 25 07:29:17 morning everyone Apr 25 07:36:19 morn Apr 25 08:16:47 otavio: Hi, I see some activity on https://github.com/codeauroraforum/mfgtools, did you try mfgtools on Linux recently? I though it wasn't supported Apr 25 08:19:51 Hi, someone can tell me if in the meta/recipes-extended/tzdata/tzdata/tzdata_2018c.bb (https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/recipes-extended/tzdata/tzdata_2018c.bb?h=morty) there is a typo in the word Bankok ? Shouldn't it be Bangkok instead ? Apr 25 08:23:19 K-Carli: Indeed, you can submit a patch for it Apr 25 09:22:58 New news from stackoverflow: qt4-embedded executable error: No such file or directory Apr 25 09:57:14 anyone tried bitbake -c fetchall core-image-sato Apr 25 09:57:48 giving error as do_fetchall doen't exist Apr 25 09:59:44 anyone please help how to get out of that Apr 25 10:01:03 pk: you can check with -c listtasks which tasks exist for a given recipe. and yes, it seems that there is no fetchall task Apr 25 10:02:55 pk: and here you go: https://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/meta?id=daf59ef368c0a0ad74cbc881ab4933cc10f06e8c Apr 25 10:17:17 LetoThe2nd: thanks for you'r help it's working Apr 25 11:59:26 I've added grub2 to my project, and i can boot my system with grub. But when i'm trying to use env variables in grub, and more advanced script in grub it seems to now working. Do i have to change my compile parameters of grub to be able to use if statements and grub env variables ? Apr 25 12:19:07 Hey Apr 25 12:38:36 hi! anyone is running bitbake-selftest -v Apr 25 12:44:34 bitbake-selftest -v is giving Assertion error, any suggestion please to make it success Apr 25 12:46:06 pk: suggestion: provide proper details. Apr 25 13:01:17 It is giving as ERROR: test_wget_checkstatus_connection_cache (bb.tests.fetch.FetchCheckStatusTest) (url='http://ftp.gnu.org/gnu/autoconf/autoconf-2.60.tar.gz') Apr 25 13:02:11 TypeError: getresponse() got an unexpected keyword argument 'buffering' Apr 25 13:02:32 pk: It's usually best to post the full bitbake output via a pastebin Apr 25 13:02:42 pk: what distro are you running on, which state of poky, have you checked that wget actually is there and working... etc, pp. Apr 25 13:11:40 Hello everybody, does anyone have a working recipe for openjre-8 and openjdk-8 for Morty? Apr 25 13:12:16 rob_gries: Have you searched the layer index? Apr 25 13:12:52 Yes, I should clarify... I'm looking at this layer http://git.yoctoproject.org/cgit/cgit.cgi/meta-java/ and I'm running morty for aarch64 Apr 25 13:13:42 In the morty branch aarch64 seems not to be supported. But in master and sumo they are. Morty appears to be too old to use those recipes. Apr 25 13:15:37 Ah ok. I'd recommend updating if you can then. If you can't you might have to do the backport yourself Apr 25 13:16:21 I'm running on poky-2.5 rc1 Apr 25 13:17:21 paulbarker: The BSP that I'm using only goes up to pyro, do sumo recipes have a good chance of working on pyro? Apr 25 13:18:15 What's the BSP/HW target that's so out-of-date there? Apr 25 13:23:17 I'm using a Snapdragon 410c with this manifest from CAF -- https://source.codeaurora.org/quic/imm/manifest/tag/?h=IMM.LE.1.0-36900-8x16.1 Apr 25 13:26:43 paulbarker: It also seems that I was incorrect it appears that the later manifests do in fact support rocko... Apr 25 13:28:04 I'd say move to rocko and try backporting the meta-java commits that update the openjdk recipe you need then, hopefully not too much work as it's only going back one release Apr 25 13:28:33 paulbarker- please go through the link https://pastebin.com/raw/GiJnaKhD Apr 25 13:37:22 pk: If you run it again does it fail on the same URLs? I'm wondering if it's a flaky network or if it's a repeatable problem Apr 25 13:40:02 Is it possible use Yocto with an MCU without an MMU? A quick search suggest 'no' but that was several yearse ago and I was wondering if it is still the case today? Apr 25 13:42:14 yes ! it's failing again on same url's Apr 25 13:43:08 deva: this is not really a yocto problem. i guess you actually are talking about linux? Apr 25 13:43:20 pk: Weird. Can you wget those URLs from the command line on the machine you're running bitbake? Apr 25 13:43:25 deva: the limitation here is getting the kernel to run on mmu-less system Apr 25 13:43:52 deva: because *technically* you can use some stuff from the yocto project to build some other OS, like zephyr. in which case the answer is yes. Apr 25 13:44:00 yocto is just a fancy set of build scripts Apr 25 13:44:07 varjag: nope. Apr 25 13:44:31 varjag: yocto is just a fancy marketing name for a lot of stuff, no real thing at all :-) Apr 25 13:45:08 fair enough Apr 25 13:47:17 i'm able to fetch url through cmd line Apr 25 13:47:55 deva: so in a nutshell: if you boil down your question to an actually specific thing, it can be anwered with "yes" or "no". Apr 25 13:47:56 Yocto can stand for an umbrella ella ella he he Apr 25 13:48:20 nayfe: it has officially been called "umbrella project" many times AFAIK Apr 25 13:49:54 LetoThe2nd: sorry, everytime i have this in mind, it reminds me that beautiful Rihanna song \o/ Apr 25 13:50:31 s/beautiful/awful/ Apr 25 13:50:36 nayfe: want me to come up with some heavy metal to make up for it? Apr 25 13:50:44 pk: Definitely looks like an issue with bitbake's test on your machine then. What distro are you running? Apr 25 13:50:52 +1 :D Apr 25 13:51:11 sure! ^^ Apr 25 13:51:55 nayfe: here you go: ttps://www.youtube.com/watch?v=DixnKyI9qts Apr 25 13:52:01 erm, https://www.youtube.com/watch?v=DixnKyI9qts Apr 25 13:55:14 LetoThe2nd: thx didnt hear it before, that guy is talented Apr 25 13:55:22 yw Apr 25 13:58:05 LetoThe2nd, varjag, I was thinking about using uLinux (2.6 kernel with patches) which apparently can run on MMU-less systems. Apr 25 13:59:14 deva: technically there's nothing that keeps you from pouring that into a specific build. but its not done yet, its tricky, and will cause a lot of pain. Apr 25 13:59:53 LetoThe2nd, Do you know of anybody who has been working on it already? Apr 25 13:59:59 nope Apr 25 14:00:07 deva: I'd look at http://nommu.org/ for recent work in this area Apr 25 14:00:20 uclinux is mostly obsolete since cortex m3, AIUI Apr 25 14:00:30 ok, so I guess the answer is "maybe" and "if you manage to get it to work we really want to hear about it" :-) Apr 25 14:01:25 deva: nah, the answer is "its probably possible, do so if you like, but its already obsolete and don't complain if it causes you serious pain" Apr 25 14:01:54 Hehe, clear speech Apr 25 14:02:04 Thanks for the input :-) Apr 25 14:02:32 there has been some effort lately to run linux on lower-end cortex m3 and m4. if you really need to go lowest notch, thats where you go to these days. Apr 25 14:02:33 nommu does exist in recent upstream kernels, you might just have to do a lot of research to figure out your options Apr 25 14:03:34 paulbarker: yeah, it would kinda nowhere resemble a "normal" syste Apr 25 14:03:36 system Apr 25 14:03:44 but if you're looking at something without an MMU I'd suggest Zephyr, FreeRTOS or similar as a first option Apr 25 14:04:59 paulbarker, You are probably right Apr 25 14:20:31 there's been an lwn article on getting to run on non-mmu/constrained systems iirc Apr 25 14:20:35 a couple months ago Apr 25 14:21:12 varjag: thats exactly what i mentioned. "shrinking the kernel with a hammer", four parts about the work of nicolas pitre Apr 25 14:24:59 right, that one Apr 25 16:10:51 Is is possible to make an image with arbitrary packages and without any system or kernel? I.e. a collection of packages? Apr 25 16:11:00 yes Apr 25 16:11:10 just list them in IMAGE_INSTALL Apr 25 16:11:45 what is it that pulls in the system normally? Apr 25 16:12:29 see core-image Apr 25 16:12:39 just inherit image instead, ignore the core-image class, and set IMAGE_INSTALL Apr 25 16:12:50 rburton: perfect, thanks Apr 25 16:20:07 oiy, I'm back to work, and that PyNaCl thing is still bugging me... Does anybody have any idea how to get PyNaCl, a python module, to compile properly? For some reason it needs to compile its own C(++) code, but has an entire build/make/make install process isolated from the outside world inside itself, and is very poorly documented Apr 25 16:20:50 sounds like you'll need to understand how it works to write the recipe Apr 25 16:21:11 assuming you've already checked the layers database for existing recipes Apr 25 16:22:07 far as I can tell, most of the other layers just have patches to comment out OS or Distro checks... Apr 25 16:26:17 * paulbarker looks at setup.py for pynacl Apr 25 16:26:29 don't do it paulbarker ! Apr 25 16:26:33 They have their own copy of libsodium in the source tree Apr 25 16:26:35 madness Apr 25 16:26:48 madness all the way down Apr 25 16:27:01 paulbarker, essential madness u,.,u Apr 25 16:27:17 yet.. pip install works fine for it Apr 25 16:27:32 because pip doesnt have to cross build Apr 25 16:27:40 precisely Apr 25 16:27:50 there's no standard/sensible way to do cross building of hybrid python/c Apr 25 16:28:13 and there is supposed to be an _sodium module that gets built by some part, but which part isn't clear to me from the setup.py Apr 25 17:05:59 * armpit yeah.. bitbake rebooted my system Apr 25 17:06:13 feature. Apr 25 17:06:21 typing / coding break Apr 25 17:06:59 DoS Apr 25 17:08:58 * armpit domo arigato mister core-image-sato Apr 25 18:14:11 okay Apr 25 18:16:52 so strange question... Does there exist an example of installing python modules from a wheel? I generated a custom image with all the requirements of PyNaCl, and have generated a wheel of PyNaCl, and want to properly install that as my PyNaCl package. It's not pretty, it's not neat... but the code is for the correct platform and built properly... Apr 25 18:30:20 so i'm looking at 12434 again, and there's a reported smaller example failing with qml-webos-framework. Apr 25 18:30:27 but i can't build that, and i'm not sure why: Apr 25 18:30:30 WARNING: /home/seebs/src/poky/meta-webosose/meta-webos/recipes-upstreamable/gnutls/gnutls_3.3.27.bb: Unable to get checksum for gnutls SRC_URI entry correct_rpl_gettimeofday_signature.patch: file could not be found Apr 25 18:30:41 this looks like i might need a different branch, or something? Apr 25 18:31:22 aarcane: pretty sure distutils builds from wheels Apr 25 18:32:08 aehs29, I'd like to see/find an example of how, since most of the recipes are downloading .tar.gz files Apr 25 18:33:18 huh. and shoving those out of the way, i get a weird python error from qt5/qt5base_git.bb, saying: Apr 25 18:33:26 Exception: bb.fetch2.MalformedUrl: The URL: '${@''' is invalid and cannot be interpreted Apr 25 18:34:40 this feels like an expansion is failing because something else wasn't set/defined as expected. Apr 25 18:40:42 yeah, ${@} wasn't expanded, which likely means there was a syntax error or ${} ref inside that was unexpanded Apr 25 18:40:51 check for a missing quote or paren or something Apr 25 18:41:58 Sadly, I can't tell what to check for it *in*, and I can't do the environment dump either because of the unhandled exception. Apr 25 18:42:11 So, I can't tell what all ended up in SRC_URI. Apr 25 18:44:54 The most likely thing looks like it might be qt5-git.inc, which adds "${QT_GIT}/${QT_MODULE}.git;name=${QT_MODULE};${QT_MODULE_BRANCH_PARAM};protocol=${QT_GIT_PROTOCOL}" Apr 25 18:45:01 seebs: you're probably missing meta-gplv2 in your mix Apr 25 18:45:03 but all the variables seem to be set, and I don't see any obvious quote/paren issues. Apr 25 18:45:24 seebs: and you need to use morty with meta-webosose Apr 25 18:45:47 ... what's "morty" and how do I use it? Apr 25 18:46:05 morty is the name of Yocto 2.2 release Apr 25 18:46:17 if i have a poky clone, is there a specific branch i can/should check out here? Apr 25 18:46:44 you would need to use exactly the revisions defined in build-webos weboslayers.py for it to work Apr 25 18:46:53 lastest morty won't work as well Apr 25 18:47:56 ahh. Apr 25 18:48:09 and qml-webos-framework reproduces it about as often as glibc-locale, the only advantage is that it's smaller/simpler component Apr 25 18:48:26 yeah. but that's a significant advantage given that i'm running on a 4-core laptop. :) Apr 25 18:49:03 I can strip it down a bit like the luna-init in pseudo-test2, but for luna-init I wasn't able to reproduce it after that Apr 25 18:49:13 interesting! Apr 25 18:49:30 wait, pseudo-test2? Apr 25 18:50:07 but yeah, something is weird. do we have any information on, say, whether this requires a given number of cores/parallel processes to happen? Apr 25 18:50:26 http://git.openembedded.org/openembedded-core-contrib/commit/?h=jansa/master&id=f29561bb37f9770a2b202ee941310e4e3f504c4f Apr 25 18:51:10 this is what I was testing last friday and over the weekend, but wasn't able to reproduce it with combination of pseudo-test and pseudo-test2 builds in latest oe-core over weekend (unlike rebuilding glibc-locale) Apr 25 18:52:56 rburton: can you give seebs shell account on your build beast? all I have access to are behind VPN out of my control Apr 25 18:55:34 Okay, the tarball extraction case... Huh. That's weird. That's one of the usual test cases, and shouldn't have had problems even before. Apr 25 18:55:57 Unless we're ending up with the datadir files deleted without the pseudo database being updated, or something. Apr 25 18:56:21 It's *really weird* that your failure rate with glibc-locale was ~5%. Apr 25 18:56:44 That's too low to be something like "every other build fails" or something else. And I'm betting the set of specific files changes. Apr 25 18:57:53 huh. *stares* Apr 25 18:58:52 this is another quite common reproducer: Apr 25 18:58:53 qtquickcontrols-5.9.5+gitAUTOINC+75e9561d4f-r0 do_package_qa: QA Issue: qtquickcontrols: /qtquickcontrols-qmlplugins/usr/lib/qml/QtQuick/Controls/Styles/Base/TreeViewStyle.qml is owned by uid 1101, which is the same as the user running bitbake. This may be due to host contamination [host-user-contaminated] Apr 25 18:59:12 this should be reproducible with latest meta-qt5, let me trigger the build to check failure rate with this one Apr 25 18:59:55 So, if you're looking in pseudo.log, I'm *guessing* you'll see a bunch of "creat for '' replaces existing ['path']" messages. Apr 25 19:02:09 hmm. Apr 25 19:02:13 creat for '/jenkins/mjansa/build-nodistro-master/BUILD/work/i586-oe-linux/glibc-locale/2.27-r0/locale-tree/usr/lib/locale/ru_UA/LC_NUMERIC' replaces existing 73330495 ['/jenkins/mjansa/build-nodistro-master/BUILD/work/i586-oe-linux/glibc-locale/2.27-r0/locale-tree/usr/lib/locale/nl_BE/LC_CTYPE']. Apr 25 19:02:20 temp/log.do_package_qa.21758:ERROR: QA Issue: glibc-locale: /glibc-binary-localedata-nl-be/usr/lib/locale/nl_BE/LC_CTYPE is owned by uid 2001, which is the same as the user running bitbake. This may be due to host contamination [host-user-contaminated] Apr 25 19:03:13 yes, there are loads of tem in various components Apr 25 19:03:43 but note that the host contamination error is for the file which got "replaced". Apr 25 19:03:53 unfortunately, a tarball won't let me check the actual inode numbers of the things. Apr 25 19:04:05 many of them got replaced, but only one of them ends with host-user-contamination Apr 25 19:04:22 and the one with the issue, is the one with fewer hardlinks than others Apr 25 19:04:46 hmm Apr 25 19:05:02 okay, so, their inode change times are within the same second Apr 25 19:05:20 LC_NUMERIC has 13:19:12.657566874, LC_CTYPE has 13:19:12.661566863. Apr 25 19:05:44 So LC_CTYPE was actually created later. The question is, why did LC_NUMERIC stop having that actual inode on disk? Apr 25 19:05:59 Because whatever happened to it, the net result was a CREAT message for it. Apr 25 19:06:52 hmm. doesn't tar not-preserve hardlinks? Apr 25 19:07:25 why would a package re-run do_compile and do_install, but won't re-do do_populate_sysroot? this would result in stale sysroot, even though the package got rebuilt? any pointers how to debug this? Apr 25 19:07:30 Interestingly, there's no corresponding host contamination for *most* of the "replaces existing..." messages. Apr 25 19:07:43 tar does preserve hard links between entries, afaik, unless you pass the arg to dereference the hard links Apr 25 19:10:39 *thinks* oh, i'm thinking of the space thing -- it still uses space for each copy, i think. Apr 25 19:10:47 because that's useful for getting the file contents by looking up the entry. Apr 25 19:12:01 Thing is, this really shouldn't be an issue with a single tar process, I think? Apr 25 19:14:48 I was thinking this might involve unlinks, but the more I look at it, the more I think that's probably not it. Apr 25 19:15:19 But it's totally true that pseudo can't guarantee that it gets messages in order, so if there were *two* processes unpacking things in the same place, I could imagine conflicts like this happening. Apr 25 19:19:20 ... man, and just when i was getting some ideas, my calendar informs me it is Time For Dentistry. Apr 25 19:25:03 New news from stackoverflow: Installing PHP Composer on Yocto/Arm Beaglebone Apr 25 19:55:15 So taking PyNaCl, building a wheel of it ON my target, then copying that wheel out and making a package out of it on the yoct02 host before in turn building an image with the new package... it works. Apr 25 20:38:42 zeddii_home: ping? Apr 25 20:39:56 zeddii_home: i don't think i'm using meta/classes/kernel-yocto.bbclass correctly with a KBUILD_DEFCONFIG Apr 25 20:42:45 after the merge_config.sh of kernel_configme is done, i'm left with no .config file in ${B} Apr 25 20:43:50 so do_configure then starts with a .config that only has the last two lines of do_kernel_configme: "CONFIG_LOCALVERSION="\"${LINUX_VERSION_EXTENSION}\" Apr 25 20:46:55 also, i don't ever seem to be able to find the log file from the output of merge_config.sh :-S Apr 25 20:49:36 as a result, the .config file that gets generated from do_configure has nothing to do with the KBUILD_DEFCONFIG that i want it to be Apr 25 20:50:24 i can work around it by simply copying my KBUILD_DEFCONFIG file to my layer, renaming it to defconfig, and adding it to SRC_URI, but that defeats the purpose Apr 25 21:56:23 tlwoerner: is this with master ? Apr 25 21:56:57 it is just supposed to pick it up, copy it to ‘defconfig’ in WORKDIR and then the normal processing continues. but we have had bugs in that code before .. I know, shocking. Apr 25 22:52:14 * tlwoerner is shocked! Apr 25 22:53:57 zeddii_home: yes with master. the KBUILD_DEFCONFIG file is copied to defconfig and makes its way through a lot of code. in fact, just before merge_config.sh is run it is still there, but merge_config.sh does something that ends up blowing it away Apr 25 23:09:34 any sstate gurus still around? it's too late for RP. kergoth, are you around? need help with some weird problem where a package gets initially installed from sstate mirror, then some dependency triggers its rebuild - do_compile, do_install, do_package. but it never gets do_populate_sysroot called and the dependency ends up using old copy from sstate... shouldn't new do_install invalidate old sysroot and re-do do_populate_sysroot? Apr 25 23:21:54 afaik it should, yes Apr 25 23:26:56 tlwoerner: feel like opening a bug ? I’ll take care of sorting it out. Apr 25 23:27:51 zeddii_home: i'll dig into it a bit more, if you don't mind... i got this far :-S Apr 25 23:28:18 i'm not entirely sure it isn't me who has made a mistake or misconfiguration Apr 25 23:33:07 tlwoerner: ok. sure, let me know. via email is better, since I’m not going to be on IRC for the next few days. but merge_config should never delete it, I’m wondering if the wrong mode is getting chosen or something new has appeared. Apr 25 23:33:25 I have a newly sync’d merge_config.sh ready to go, but I’m sitting on my queue until after the release. Apr 25 23:33:38 ok, will do Apr 25 23:34:51 kergoth: thanks. are there any checks that may decide that new do_populate_sysroot is not needed? the package is quite unique, as it has no bins or libs, but generates some data files in ${datadir}. looking over staging.bbclass, ${datadir} is handled normally along with libs and bins... **** ENDING LOGGING AT Thu Apr 26 03:00:03 2018