**** BEGIN LOGGING AT Mon Mar 21 02:59:58 2016
Mar 21 03:33:00 <riz__> How do I add EGLFS plugin to my poky image?
Mar 21 06:25:05 <Ulfalizer> how does yocto differ between explicitly set and automatically added dependencies, e.g. for QA checks? won't both of those just set/extend (R)DEPENDS?
Mar 21 06:48:51 <niteshnarayanlal> hi is it possible to set an environment variable which is visible across multiple yocto recipes
Mar 21 08:32:52 <mckoan> good morning
Mar 21 08:57:03 <zero_note> hi guys, I'd like to use a static-populated /dev on my target, so any of you have ever used systemd without udev?
Mar 21 08:59:14 <zero_note> despite it may seem a systemd related question, I think that here there are more people that have worked about linux fastboot
Mar 21 11:06:32 <belen> joshuagl: ty
Mar 21 11:54:43 <niteshnarayanlal> Hi, I am getting an error while doing this
Mar 21 11:54:45 <niteshnarayanlal> http://fpaste.org/343302/58561266/
Mar 21 11:54:58 <niteshnarayanlal> Files/directories were installed but not shipped
Mar 21 11:55:24 <niteshnarayanlal> I looked about it on internet and found an option to skip this check but in that case the empty directories which I am trying to create
Mar 21 11:55:27 <niteshnarayanlal> are not getting created
Mar 21 12:02:31 <boucman_work> niteshnarayanlal: installed but not shipped means that the "make install" part of your recipe correctly creates those files/directories
Mar 21 12:02:48 <boucman_work> but none of the packages you create says that it wants them
Mar 21 12:03:13 <boucman_work> you probably want to add them to the FILES_${PN} list of files to install
Mar 21 12:03:36 <boucman_work> i.e your do_install is correct :)
Mar 21 12:07:30 <townxelliot> belen: can you remember the (recent) bug number for Toaster being unable to run builds?
Mar 21 12:08:44 <townxelliot> belen: don't worry, I located it
Mar 21 12:09:07 <belen> townxelliot: there were 2. 9240 and 9252
Mar 21 12:09:41 <townxelliot> belen: thanks
Mar 21 12:09:47 <belen> townxelliot: btw, I am trying the patch we sent to bitbake-devel for those on top of master, and doesn't seem to fix the problem for me. Did you have a chance to try it?
Mar 21 12:12:55 <niteshnarayanlal> boucman_work,  thanks a ton :)
Mar 21 12:13:00 <boucman_work> np
Mar 21 12:16:57 <btooth> hi.. how to remove FILES from a package installation in a bbappend? I mean something like FILES_${PN} -= "myfile"
Mar 21 12:17:44 <lq> http://linkcash.co/2Zn
Mar 21 13:59:04 <zero_note> hi guys, I've a systemd_%.bbappend which I use to add my custom services to the final image, but in the final image these services will result disabled (no symlink)
Mar 21 13:59:59 <zero_note> how can I set these services as enabled?
Mar 21 14:00:52 <zero_note> I've just used SYSTEMD_AUTO_ENABLE = "enable" in my bbappend, but it did not work
Mar 21 14:01:58 <rburton> don't use a systemd bbappend to control services in other recipes
Mar 21 14:05:37 <zero_note> I've to write my own bb for adding and managing my custom systemd services?
Mar 21 14:06:51 <rburton> do the service control in the recipes that add the services
Mar 21 14:07:31 <zero_note> I use the bbappend to add the services
Mar 21 14:08:59 <rburton> to the systemd  recipe?
Mar 21 14:09:05 <rburton> well should work then
Mar 21 14:10:57 <zero_note> yes, I use systemd_%.bbappend to add systemd services and other stuff (like config files, etc...) and I've no problem at all, excluding the fact that these services will be disabled
Mar 21 14:12:22 <boucman_work> zero_note: if there is nothing confidential, you might want to pastebin your .bbappend somewhere for us to look at...
Mar 21 14:19:05 <karobar> anyone else have problems installing packagegroup-core-x11-utils with debian packaging (jethro branch)?
Mar 21 14:19:13 <zero_note> boucman_work: here it is http://pastebin.com/xzAsGsnb, thank you :-)
Mar 21 14:20:44 <riz__> After including meta-qt5 in my build and creating the toolchain I am having trouble pointing to "qmake" in QtCreator. It is saying that it does not exist or is not an executable. Here is some information on it: http://pastebin.com/mBEUTJND
Mar 21 14:22:48 <boucman_work> zero_note: nothing unusual that I can see...
Mar 21 14:25:23 <boucman_work> the only thing I can see is that according to the systemd.bbclass documentation, you shouldn't have to install the service file yourself, the systemd.bbclass should do that for you
Mar 21 14:25:42 <boucman_work> it might interact badly with the auto-enable feature, but I wouldn't bet on it
Mar 21 14:31:07 <zero_note> boucman_work: really don't know...I'll try in some other ways and let you know about my mistake
Mar 21 14:32:13 <niteshnarayanlal> boucman_work,  if I want a package to install the headers and libraries in the standard path of build so that other recipes could use that
Mar 21 14:32:29 <niteshnarayanlal> but I don't want it to be part of the rootfs
Mar 21 14:33:41 <niteshnarayanlal> in that case just using install or cp command to copy them to ${libdir} /${incdir} would do?
Mar 21 14:35:22 <boucman_work> niteshnarayanlal: normally you have nothing special to do...
Mar 21 14:35:40 <boucman_work> your package will install the include at the normal place during its "do_install"
Mar 21 14:35:54 <boucman_work> then the files will be part of a "-dev" package
Mar 21 14:36:10 <niteshnarayanlal> yes ideally the packages install them on its own but this package is not doing that
Mar 21 14:36:14 <niteshnarayanlal> :/
Mar 21 14:36:15 <boucman_work> when another package DEPENDS on your first package, yocto will automatically install the required files in the sysroot
Mar 21 14:36:32 <boucman_work> where your second package will find it, but it's not part of the rootfs
Mar 21 14:36:56 <boucman_work> niteshnarayanlal: the add the require calls to install in the do_install section
Mar 21 14:37:34 <niteshnarayanlal> you mean something like cp -rf ${S}/out/include ${STAGING_INCDIR}
Mar 21 14:37:36 <boucman_work> see how zero_note does it in his pastebin earlier (the do_install_append part)
Mar 21 14:38:13 <boucman_work> i'm not sure what exactly the target is, but something like that...
Mar 21 14:39:01 <boucman_work> ${D}${includedir} IIRC
Mar 21 14:39:51 <boucman_work> and no STAGING_INCDIR is not what you want
Mar 21 14:40:21 <riz__> When I IMAGE_INSTALL_append qtbase and qtbase-plugin I get the following error: ERROR: Task 400 (/home/ryan/minnowboard/poky/meta-qt5/recipes-qt/qt5/qtbase_git.bb, do_configure) failed with exit code '1'
Mar 21 14:40:26 <riz__> Any idea why this might happen?
Mar 21 14:40:52 <niteshnarayanlal> ah ok thanks :)
Mar 21 14:42:50 <boucman_work> riz__: there is usually more usefull error message higher in your log when that happens... anything suspicious ?
Mar 21 14:43:20 <riz__> http://pastebin.com/Lp9hkFN6
Mar 21 14:43:26 <riz__> That is from the log
Mar 21 14:44:39 <riz__> I am thinking qtbase-plugins are needed to add EGLFS in my image. As of now I cant get Qt5Cinematic Experience working without EGLFS so I am trying to add qtbase-plugins in hope of it working. IT is saying that the EGLFS plugin is not installed.
Mar 21 14:44:41 <boucman_work> lots of errors in that log
Mar 21 14:45:40 <riz__> I am having trouble understanding what the core issue is so I can resolve it.
Mar 21 14:46:41 <boucman_work> does your target support OpenGL in general ? is it enabled ? (I think it's a distro feature)
Mar 21 14:46:55 <boucman_work> if you changed it, you need to clean/rebuild everything IIRC...
Mar 21 14:48:08 <riz__> I am using poky jethro on a minnowboard
Mar 21 14:48:14 <riz__> I believe it does
Mar 21 14:48:26 <riz__> How would I clean it?
Mar 21 14:50:04 <boucman_work> riz__: is opengl in your DISTRO_FEATURES ? (bitbake -e | less, then look for opengl)
Mar 21 14:54:30 <riz__> I dont see it in the include history
Mar 21 14:55:45 <boucman_work> riz__: what do you mean ? it's not an include, it's a DISTRO_FEATURE (a variable)
Mar 21 14:57:06 <riz__> Sorry,, no it is not
Mar 21 14:59:07 <boucman_work> ok, then you need to add it to your distro. is your distro poky or a custom distro ?
Mar 21 15:01:22 <riz__> poky
Mar 21 15:03:28 <riz__> This is done in build/conf/local.conf?
Mar 21 15:03:38 <boucman_work> hmm, i'm a bit suprised poky doesn't have opengl by default... try adding DISTRO_FEATURES +="opengl" in your local.conf
Mar 21 15:04:00 <boucman_work> riz__: it's one of the possible places, but it's the most convinient place when we are trying to figure out what's going on
Mar 21 15:04:09 <riz__> I am looking another way
Mar 21 15:04:09 <boucman_work> we might consolidate it somewhere else later...
Mar 21 15:04:12 <riz__> using bb
Mar 21 15:05:07 <rburton> poky has GL by default
Mar 21 15:05:09 <rburton> oe-core doesnt
Mar 21 15:07:09 <boucman_work> well, riz__ says he is using poky but opengl is disable, can something disable opengl after the fact ?
Mar 21 15:07:19 * boucman_work is not very good at opengl+yocto
Mar 21 15:07:45 <riz__> I got these errors now: http://pastebin.com/edYHRq3i
Mar 21 15:10:17 <rburton> yeah don't use DISTRO_FEATURES +=
Mar 21 15:10:48 <rburton> i'd want proof that DISTRO_FEATURES doesn't contain opengl if poky is being used
Mar 21 15:10:51 <rburton> pastebin will do nicely
Mar 21 15:11:11 <rburton> but you want _append for DISTRO_FEATURES, as it's lazily-assigned using ?=
Mar 21 15:11:25 <rburton> so += assigns  value and you don't get the distro features you expect
Mar 21 15:12:01 <riz__> can I use bb to confirm if opengl is available?
Mar 21 15:12:14 <riz__> bb show DISTRO DISTRO_FEATURES?
Mar 21 15:13:06 <rburton> bitbake -e |grep DISTRO_FEATURES= works for me
Mar 21 15:13:08 <riz__> I saw something like that
Mar 21 15:13:53 <rburton> yeah bb show DISTRO_FEATURES works
Mar 21 15:13:55 * rburton had to update bb
Mar 21 15:15:27 <jose1985> hello!
Mar 21 15:15:49 <riz__> ryan@hp-pc:~/minnowboard/build$ bitbake -e |grep DISTRO_FEATURES= POKY_DEFAULT_DISTRO_FEATURES="largefile opengl ptest multiarch wayland" DISTRO_FEATURES="alsa argp bluetooth ext2 irda largefile pcmcia usbgadget usbhost wifi xattr nfs zeroconf pci 3g nfc ipv4 ipv6 libc-backtrace libc-big-macros libc-bsd libc-cxx-tests libc-catgets libc-charsets libc-crypt libc-crypt-ufc libc-db-aliases libc-envz libc-fcvt libc-fmtmsg libc-fstab libc-ft
Mar 21 15:16:02 <jose1985> can anyone help me with successfully creating a recipe for python-scipy?
Mar 21 15:16:16 <riz__> It is a poky default
Mar 21 15:16:16 <jose1985> i'm already working on this for 3 weeks with no success
Mar 21 15:16:39 <jose1985> current state is: gfortran + libgfortran done
Mar 21 15:16:49 <jose1985> lapack done
Mar 21 15:17:07 <jose1985> can't complete scipy  compilation due to linking problems
Mar 21 15:17:18 <jose1985> is there any protocol to ask for help here?
Mar 21 15:17:28 <rburton> riz__: your paste was truncated
Mar 21 15:17:47 <rburton> jose1985: ask concrete questions, and for complex problems you might do better on the lists
Mar 21 15:17:58 <riz__> http://pastebin.com/nvpiahw5
Mar 21 15:18:12 <rburton> riz__: so you have opengl then
Mar 21 15:18:27 <riz__> Now I see it, yes
Mar 21 15:18:59 <riz__> So, what could be the issue as before: http://pastebin.com/Lp9hkFN6
Mar 21 15:19:32 * rburton never used qt
Mar 21 15:20:04 <riz__> It is saying upengl test failed and it is disabled
Mar 21 15:20:13 <riz__> opengl
Mar 21 15:20:20 <rburton> what MACHINE are you using
Mar 21 15:20:58 <riz__> intel-corei7-64
Mar 21 15:21:13 <rburton> maybe the qt recipes are broken, try bitbake opengl first
Mar 21 15:21:15 <riz__> building for minnowboard
Mar 21 15:21:18 <rburton> i mean bitbake mesa
Mar 21 15:21:54 <riz__> bitbake mesa was successful
Mar 21 15:22:46 <rburton> oh
Mar 21 15:22:50 <rburton> you removed x11 from DISTRO_FEATURES
Mar 21 15:23:05 <rburton> libGL is opengl-on-x11
Mar 21 15:23:20 <rburton> basically, the qt recipe is broken
Mar 21 15:23:29 <riz__> The tutorials for wanboard said it needed to be removed for eglfs to work
Mar 21 15:23:56 <rburton> why are you reading documentation for the wandaboard when you're working on a minnow?
Mar 21 15:24:37 <rburton> this is driver-specific so what applies for the freescale drivers may not be true for x86 with mesa
Mar 21 15:24:38 <riz__> I figured most of it was general case
Mar 21 15:24:41 <rburton> not at all
Mar 21 15:24:43 <riz__> I see
Mar 21 15:24:50 <rburton> with mesa, egl always works
Mar 21 15:25:34 <riz__> So I dont have to do anything to add an eglfs plugin?
Mar 21 15:25:45 <rburton> now we're back to "i have never used qt"
Mar 21 15:26:56 <riz__> In general, how are plugins added to poky?
Mar 21 15:27:06 <rburton> in general, it depends on what plugins you're talking about
Mar 21 15:27:15 <riz__> I removed the removal of x11 and at least it built
Mar 21 15:27:24 <_4urele_> hi everyone
Mar 21 15:27:46 <riz__> There is no general procedure? I looked at documentation and dont see anything specific
Mar 21 15:27:49 <rburton> if you dont want or need x11  then removing it makes sense, but you want to speak to someone who understands qt about making the egl bits actually build
Mar 21 15:28:05 <rburton> riz__: by definition, a plugin isn't generalisabe
Mar 21 15:28:14 <rburton> i suggest you ask on the meta-qt5 mailing list
Mar 21 15:28:18 <riz__> understood
Mar 21 15:28:19 <_4urele_> is there a way to build every dependencies of a package but not the package (to build dependencies of the image but not build the entire rootfs which is long...)
Mar 21 15:28:41 <jose1985> @rburton ok, fair enough
Mar 21 15:28:53 <rburton> _4urele_: build the image and control-c the rootfs? :)
Mar 21 15:28:58 <jose1985> i will check the lists one more time and come back with a concrete question
Mar 21 15:28:59 <jose1985> thanks
Mar 21 15:30:19 <_4urele_> rburton, ;) let's continue this way, thanks
Mar 21 15:36:49 <Guest_98765> Allah is doing
Mar 21 15:37:00 <rburton> the dishes?
Mar 21 15:37:01 <Guest_98765> sun is not doing Allah is doing
Mar 21 15:37:14 <Guest_98765> moon is not doing Allah is doing
Mar 21 15:37:28 <Guest_98765> stars are not doing Allah is doing
Mar 21 15:37:40 <Guest_98765> planets are not doing Allah is doing
Mar 21 15:37:52 <Guest_98765> galaxies are not doing Allah is doing
Mar 21 15:38:05 <Guest_98765> oceans are not doing Allah is doing
Mar 21 15:38:20 <Guest_98765> mountains are not doing Allah is doing
Mar 21 15:38:23 <rburton> DOING WHAT
Mar 21 15:38:32 <rburton> the housework, i hope
Mar 21 15:38:34 <rburton> that would be awesome
Mar 21 15:38:37 <Guest_98765> trees are not doing Allah is doing
Mar 21 15:38:49 <Guest_98765> mom is not doing Allah is doing
Mar 21 15:39:04 <Guest_98765> dad is not doing Allah is doing
Mar 21 15:39:11 <rburton> man i should have done a your mom instead of housework joke
Mar 21 15:39:16 <Guest_98765> boss is not doing Allah is doing
Mar 21 15:39:28 <rburton> your mom
Mar 21 15:39:30 <Guest_98765> job is not doing Allah is doing
Mar 21 15:39:43 <Guest_98765> dollar is not doing Allah is doing
Mar 21 15:39:54 <Guest_98765> degree is not doing Allah is doing
Mar 21 15:40:05 <Guest_98765> medicine is not doing Allah is doing
Mar 21 15:40:18 <Guest_98765> customers are not doing Allah is doing
Mar 21 15:40:37 <Guest_98765> you can not get a job without the permission of allah
Mar 21 15:40:56 <Guest_98765> you can not get married without the permission of allah
Mar 21 15:41:17 <Guest_98765> nobody can get angry at you without the permission of allah
Mar 21 15:41:26 <Guest_98765> light is not doing Allah is doing
Mar 21 15:41:37 <Guest_98765> fan is not doing Allah is doing
Mar 21 15:42:18 <rburton> thanks RP
Mar 21 15:42:23 <boucman_work> RP: thx
Mar 21 15:42:38 <RP> np :)
Mar 21 15:42:43 <JaMa> how evil :)
Mar 21 15:42:49 <RP> took me a while to remember the commands...
Mar 21 15:43:20 <fledermaus> bizarre troll.
Mar 21 15:43:31 <JaMa> is there plan to merge pending dizzy patches?
Mar 21 15:43:42 <boucman_work> yeah, not sure if bot or real person
Mar 21 15:43:46 <RP> JaMa: remind me which ones are pending?
Mar 21 15:43:59 <JaMa> RP: glibc security
Mar 21 15:44:12 <JaMa> RP: http://lists.openembedded.org/pipermail/openembedded-core/2016-March/119125.html
Mar 21 15:48:48 <jose1985> concrete question: Scipy depends on a 'lapack' package (BLAS and LAPACK) in order to build. This actually means I must define my dependency as a DEPEND, am I wrong?  And, as for as I know, you need to make it a 'native' package in order be available to gfortran-cross.
Mar 21 15:48:56 <jose1985> This workflow still puzzles me a bit
Mar 21 15:49:05 <RP> JaMa: done
Mar 21 15:49:35 <RP> JaMa: they'd ended up being filtered in a weird way sorry
Mar 21 15:49:44 <JaMa> RP: thanks
Mar 21 15:50:10 <boucman_work> jose1985: what does "scipy depends on lapack" means in non-yocto term ? does lapack provide binaries that scipy needs to build ? does lapack create a library ?
Mar 21 15:50:20 <boucman_work> or is it something more unorthodox than that ?
Mar 21 15:51:10 <jose1985> yes, lapack is a Linear Algebra library needed by scipy on compile time
Mar 21 15:52:00 <jose1985> I wrote the recipe for this, which also wasn't easy because it's Fortran code
Mar 21 15:52:11 <boucman_work> ok, if lapack is correctly packaged, you just need to DEPEND on it, everything else should happen automagically
Mar 21 15:52:28 <boucman_work> you wrote the recipe for lapack ?
Mar 21 15:52:57 <jose1985> yes, it's actually quite simple and I already wrote a couple of recipes for my use case.
Mar 21 15:53:08 <jose1985> altough the Fortran depedency was tricky
Mar 21 15:53:57 <boucman_work> ok... if the recipe follows the correct model, then DEPEND on it is enough... and you shouldn't need to build the native version
Mar 21 15:54:39 <jose1985> took the liberty of doing a pastebin: http://pastebin.com/Acwr7P1z
Mar 21 15:54:50 <boucman_work> thx
Mar 21 15:55:13 <boucman_work> that is a simple recipe indeed :P
Mar 21 15:55:18 <jose1985> :)
Mar 21 15:55:49 <boucman_work> do you even need the FILES_${PN} line ? I would think cmake.bbclass would take care of that for you...
Mar 21 15:55:55 <jose1985> one thing I haven't understood is why i needed to explicit the FILES_${PN}
Mar 21 15:56:00 <jose1985> lol
Mar 21 15:56:02 <boucman_work> :P
Mar 21 15:56:17 <jose1985> yeah, it didn't for some reason
Mar 21 15:56:21 <boucman_work> hmm
Mar 21 15:56:47 <jose1985> It only created the dbg and staticdev packages
Mar 21 15:57:14 <jose1985> I'm guessing this is due to the name/nature of the library files
Mar 21 15:57:29 <boucman_work> oh ?
Mar 21 15:57:44 <boucman_work> I don't know enough about fortrant, what does it generate ?
Mar 21 15:57:58 <boucman_work> maybe the automated package split doesn't know how to deal with fortrant
Mar 21 16:00:10 <boucman_work> jose1985: maybe you could paste somewhere the content of ${WORKDIR}/image and ${WORKDIR}/package-split ? that would help
Mar 21 16:01:12 <jose1985> ok, 1min
Mar 21 16:04:16 <jose1985> http://pastebin.com/pxE8XKVK
Mar 21 16:05:10 <jose1985> this is package-split
Mar 21 16:05:22 <jose1985> I didn't know this existed
Mar 21 16:05:26 <boucman_work> hmm, ok...
Mar 21 16:05:39 <boucman_work> yeah, it's pretty awesome to know a little bit about that when debugging recipes
Mar 21 16:06:12 <jose1985> I don't have an image dir
Mar 21 16:06:28 <jose1985> but have a package that contains the contents of the default package
Mar 21 16:06:55 <boucman_work> please run "bitbake -f -c do_install lapack"
Mar 21 16:07:03 <boucman_work> that should recreate the image directory
Mar 21 16:09:37 <jose1985> http://pastebin.com/gePrcmzb
Mar 21 16:10:46 <boucman_work> ok, so your recipe does not provide include files, nor dynamic libraries, only static...
Mar 21 16:10:55 <boucman_work> is that normal for fortran libs ?
Mar 21 16:11:24 <jose1985> Don't know the first thing about fortran
Mar 21 16:12:00 <jose1985> Just trying to fulfill all the dependencies for installing scipy
Mar 21 16:12:05 <boucman_work> ok, I thought you said lapack was fortran, nevermind...
Mar 21 16:12:31 <jose1985> and it is, and you need it to compile SciPy from source
Mar 21 16:13:07 <boucman_work> ok. lapack is a pure-static, no-include library... let's assume this is normal
Mar 21 16:13:23 <boucman_work> what is scipy complaining about ? what is it looking for that it doesn't find ?
Mar 21 16:15:06 <jose1985> Scipy originally was complaining about the absence of these static libraries even tough I DEPEND on lapack
Mar 21 16:16:05 <boucman_work> at that point, I'm not sure what DEPENDS does for static libraries... maybe someone else here knows...
Mar 21 16:16:11 <jose1985> http://pastebin.com/55n6Eq5z
Mar 21 16:16:20 <jose1985> this is the current state of my recipe
Mar 21 16:16:35 <jose1985> is find the libs now becasue of the LAPACK and BLAS export
Mar 21 16:16:42 <jose1985> but I know it is a hack
Mar 21 16:16:47 <kergoth> RP: does https://github.com/MentorEmbedded/poky/compare/master...kergoth:task-level-exports seem reasonable for 2.2? Just want to make sure I didn't miss anything. I can send it as an RFC too if that's better than irc :)
Mar 21 16:18:09 <boucman_work> jose1985: is it really a hack ? is the scipy build tool able to find these libs automatically ?
Mar 21 16:18:34 <RP> kergoth: I've often wished we had that in. Please do send and I'll flag it for 2.2
Mar 21 16:18:44 <RP> kergoth: kept wondering what happened to it!
Mar 21 16:19:31 <kergoth> k, will do. thanks
Mar 21 16:19:45 <jose1985> boucman_work: I feel like it is a hack because I have to point to the targe machine sysroot  for the /usr/lib
Mar 21 16:19:45 <kergoth> ran of our tasks for my work sprint so started poking at my personal backlog
Mar 21 16:19:54 <jose1985> target*
Mar 21 16:20:18 <jose1985> isn't it supposed to be available at the host machine sysroot?
Mar 21 16:21:34 <boucman_work> I'm not sure what you mean by host-machine-sysroot... it is a target library (i.e compiled to target-machine-instructions) so it should be in the target sysroot
Mar 21 16:21:36 <RP> kergoth: I'd love to do that!
Mar 21 16:21:56 <boucman_work> the host sysroot should only contain native binaries, but that is not what we have here
Mar 21 16:21:58 <kergoth> RP: heh :) it's a pretty rare occurrance
Mar 21 16:22:42 <jose1985> that part is what is really fuzzy for me
Mar 21 16:23:11 <boucman_work> ok... i'll try to explain, but anybody in the chan : feel free to correct me if I say anything stupid
Mar 21 16:23:25 <jose1985> ok, thanks a lot
Mar 21 16:26:18 <boucman_work> so, a normal recipe will generate "stuff" that is specific to the target (.so .a, binaries etc...)
Mar 21 16:26:39 <boucman_work> part of those are only needed in the target rootfs (.so, and binaries)
Mar 21 16:27:06 <boucman_work> but others are needed by the cross-compiler to build other recipes further down the line (includes, .a files, .pc files etc...)
Mar 21 16:27:18 <boucman_work> these need to be stored "somewhere" where the compiler will find them
Mar 21 16:27:59 <boucman_work> that somewhere is the target sysroot. it is organized in a way similar to a root file system (includes are in ${sysroot}/usr/include)
Mar 21 16:28:37 <jose1985> ahh ok. I tought that this 'somewhere' was the host sysroot
Mar 21 16:28:46 <boucman_work> so -dev and -staticdev packages are typically installed in the cross-compiler's sysroot at build time, so other recipes can find the stuff
Mar 21 16:29:00 <boucman_work> ok, moving on to the host sysroot
Mar 21 16:29:02 <boucman_work> :P
Mar 21 16:29:15 <boucman_work> some recipes need stuff to be installed on the host to work
Mar 21 16:29:33 <boucman_work> note that I am speaking of "recipes" not packages
Mar 21 16:29:50 <jose1985> duly noted :)
Mar 21 16:30:10 <boucman_work> for instance, a recipe could have a weird build system coded in perl, instead of using automake/cmake etc...
Mar 21 16:30:32 <boucman_work> in that case, yocto needs to build perl FOR THE HOST and install it somewhere to be able to use it later
Mar 21 16:31:18 <boucman_work> it can't just install the perl binary in /usr/bin because it would be ugly and bitbake doesn't have root privileges on the host machine in the first place
Mar 21 16:31:42 <jose1985> yes, obviously :)
Mar 21 16:31:53 <jose1985> for instance, the python setuptools or distutils
Mar 21 16:31:53 <kergoth> I wonder if we should consider setting `set -u` in our shell scripts. we'd have to fix up a few things, but it'd make it easier to catch 1. typos, and 2. fallthrough of undefined bitbake vars that might not have been intended. given it's easy to add a default value to a variable, either in bitbake or shell..
Mar 21 16:31:55 * kergoth ponders
Mar 21 16:32:00 <jose1985> need python to run
Mar 21 16:32:06 <boucman_work> so, bitbake has this other rootfs-like place where it puts HOST stuff (libs, binaries) that it needs to do its stuff that are native
Mar 21 16:32:33 <kergoth> course an alternative would be to disable bitbake to shell fallthrough and make undefined var references fatal
Mar 21 16:32:38 <boucman_work> that's the host-rootfs. Typically a package named perl-native (that is built by the perl recipe) is installed there
Mar 21 16:33:05 <boucman_work>  /host-rootf/host-sysroot
Mar 21 16:33:45 <boucman_work> so <package>-native means a package built by yocto, but instead of cross-compiling, we are native-compiling and keeping it for our personal use
Mar 21 16:33:51 <boucman_work> is that all a bit more clear ?
Mar 21 16:33:57 <jose1985> yes, a lot clearer
Mar 21 16:34:03 <boucman_work> yay, success :)
Mar 21 16:34:44 * kergoth checks to see how painful `set -u` is, as having it set would make it easier to diagnose any problems when removing global exports in favor of task-level
Mar 21 16:35:13 <jose1985> one thing: the cross compiler shoud look on the target rootfs for the libraries that he needs to build something
Mar 21 16:35:43 <boucman_work> yes, since it needs target-libraries to build target-binaries
Mar 21 16:35:45 <kergoth> we pass --sysroot to it in CC, as long as the buildsystem obeys our CC var, that's handled
Mar 21 16:36:12 * boucman_work needs to run, or he'll miss his bus
Mar 21 16:36:21 <boucman_work> sorry
Mar 21 16:36:27 <jose1985> no worries
Mar 21 16:36:30 <jose1985> thanks a lot
Mar 21 16:37:14 <jose1985> kergoth: that was for me? :P
Mar 21 16:37:27 <kergoth> jose1985: yes, indeed
Mar 21 16:38:23 <jose1985> ohh ok, that's what --sysroot means
Mar 21 16:38:27 <jose1985> one less mistery
Mar 21 16:38:38 <jose1985> thanks
Mar 21 16:41:30 <zero_note> boucman_work: so the last way that I hope should give me the possibility to "enable" my systemd services in final image is to manually create links in do_install_append, but I've never do this before
Mar 21 16:41:57 <zero_note> here are my attemp http://pastebin.com/y9Ffbs5v
Mar 21 16:44:33 <zero_note> because my build end with "WARNING: QA Issue: Symlink /etc/systemd/system/multi-user.target.wants/macaddrs.service in systemd points to TMPDIR [symlink-to-sysroot]"
Mar 21 16:44:42 <arkver> zero_note: http://stackoverflow.com/questions/32902003/enable-systemd-service-in-yocto-build
Mar 21 16:45:16 <arkver> Anders' comment at the bottom looks like what you need.
Mar 21 16:47:10 <arkver> "[A] better way would be to use the systemd.bbclass. You'd do this by inherit systemd in your recipe, and then specify the name of the service file in SYSTEMD_SERVICE = "name_of_the_file". This will automatically enable the service for you.", for the record.
Mar 21 16:47:29 <zero_note> arkver: thanks, this link is open in my browser from hours :/
Mar 21 16:48:29 <zero_note> arkver: I know, but if you look at my original bbappend http://pastebin.com/xzAsGsnb
Mar 21 16:49:43 <zero_note> you can see that, *in theory*, all configs are correct...but systemd services will not be enabled in the final image
Mar 21 16:50:13 <zero_note> so my last trick is to create links "manually"
Mar 21 16:50:14 <arkver> Dunno - can't help further. Don't personally use systemd built by yocto. Sry.
Mar 21 16:50:32 <zero_note> arkver: thank you anyway :-)
Mar 21 16:53:12 <kergoth> zero_note: sounds like you didn't set SYSTEMD_AUTO_ENABLE to enable
Mar 21 16:53:20 <kergoth> though it should be set that way by default nowadays
Mar 21 16:53:26 <kergoth> worth checking its value with bitbake -e <recipe>
Mar 21 16:54:58 <zero_note> kergoth: I had already set explicitly SYSTEMD_AUTO_ENABLE to enable, without any effect
Mar 21 16:55:55 <kergoth> if you want to fix the QA failure with your manual links, then fix the link destination to be either relative, or absolute with the target path. that is, don't include ${D} in the link destination, only the link path
Mar 21 16:56:10 <kergoth> i.e. ln -s foo ${D}/bar, not ln -s ${D}/foo ${D}/bar
Mar 21 17:03:32 <arkver> zero_note: SYSTEMD_SERVICE_${PN} += "gpio-pwr.service"
Mar 21 17:03:32 <arkver> SYSTEMD_SERVICE_${PN} += "macaddrs.service"
Mar 21 17:03:36 <arkver> Whitespace?
Mar 21 17:03:58 <arkver> Between the two services?
Mar 21 17:04:54 <zero_note> kergoth: thanks I'll try with your suggestion
Mar 21 17:05:26 <zero_note> arkver: do you mean something like  SYSTEMD_SERVICE_${PN} += "macaddrs.service gpio-pwr.service" ?
Mar 21 17:05:32 <arkver> yeah
Mar 21 17:05:46 <arkver> just guessing
Mar 21 17:06:36 <zero_note> at this point, I'll try everything
Mar 21 17:16:03 <zero_note> kergoth: yeah, your fix about link destination does the trick, finally my services starts at boot, thank you!
Mar 21 17:29:41 <kergoth> zero_note: np. still shouldn't have had to do the manual link, but at least you're unblocked now
Mar 21 18:20:39 <igor4> how can I see the final content of PACKAGE_INSTALL, I need to debug why a package is been installed instead other?
Mar 21 18:27:24 <riz___> How could I get "file" or "readelf" working on my poky build?
Mar 21 18:39:43 <riz___> How do I include "ld-linux-x86-64.so.2" in my image?
Mar 21 18:41:46 <kergoth> riz___: you don't need to do anything to include it. what problem are you trying to solve, exactly?
Mar 21 18:43:59 <riz___> I am tryin to run a Qt app on my minnoboard, but it is saying it cannot find the executable
Mar 21 18:44:15 <riz___> My readelf gives the following: http://pastebin.com/hWCeYcf8
Mar 21 18:44:45 <hs89> stupid quedstion that I'm hoping someone can help with -- is it possible to get my kernel recipe to point to the master branch instead of a release branch?
Mar 21 18:45:15 <riz___> I assumed it had something to do with ld-linux-x86-64.so.2
Mar 21 18:47:07 <kergoth> you can't just grab a binary from some random place and run it on your board and expect it to work
Mar 21 18:47:28 <kergoth> binary compatibility between linux distros is quite often non-trivial
Mar 21 18:48:23 <riz___> I built the sdk and am using meta-qt5
Mar 21 18:48:38 <riz___> I setup the environment on QtCreator and deployed it
Mar 21 18:48:49 <riz___> What am I missing?
Mar 21 18:49:16 <kergoth> sounds like your sdk doesn't line up with what you're running on the board. were they from the same build dir with the same configuration?
Mar 21 18:50:19 <riz___> I thought so but now I am having doubts
Mar 21 18:51:33 <riz___> I bitbakes core-image-minimal
Mar 21 18:51:41 <riz___> meta-qt5 was included
Mar 21 18:51:51 <riz___> Then I executed bitbake core-image-minimal -c populate_sdk
Mar 21 18:52:08 <riz___> That gave me  my sdk in opt/poky/2.0.1...
Mar 21 18:52:24 <riz___> I used that
Mar 21 18:53:02 <riz___> It gave me two folders in my sysroot actually
Mar 21 18:53:23 <riz___> corei7-64-poky-linux and x86-64-pokysdk-linux
Mar 21 18:53:50 <riz___> I used the latter for the toolchain for QtCreator
Mar 21 19:06:31 <kergoth> you sourced the environment setup in the sdk before using qt creator?
Mar 21 19:06:42 <riz___> Yup
Mar 21 19:06:47 <riz___> I put it in qtcreator.sh
Mar 21 19:07:58 <riz___> Am I supposed to point the qtcreator to the toolchain in corei7-64-poky-linux or x86-64-pokysdk-linux
Mar 21 19:08:00 <riz___> ?
Mar 21 19:08:58 <riz___> The problem is that there is no qmake in x86-64-pokysdk-linux/usr/bin
Mar 21 19:26:18 <riz___> Does anyone know what the meta-qt5 mailing list is?
Mar 21 19:26:25 <riz___> I cant find it
Mar 21 19:26:53 <riz___> Does anyone know what the meta-qt5 mailing list is?
Mar 21 19:29:46 <mborzecki> riz___: readme mentions openembedded-devel@lists.openembedded.org so I'd try there
Mar 21 19:30:03 <riz___> Thanks\
Mar 21 20:16:50 <kergoth> Hmm, limiting var access to a specific set of vars will be interesting, i'll have to brush up on the datasmart internals
Mar 21 20:17:06 <kergoth> write a little subclass
Mar 21 20:48:18 <arkver> I'm doing kernel dev on a couple of different v4.5 based branches at the moment, and I have two recipes which differ only in the branch they check out.
Mar 21 20:48:54 <arkver> When I try to switch from one to the other in a bitbake core-image-minimal I get an error in do_populate_sysroot.
Mar 21 20:49:15 <arkver> ERROR: The recipe linux-tx6-sl-git is trying to install files into a shared area when those files already exist.
Mar 21 20:50:02 <arkver> If I rm tmp/sysroots/<machine>/sysroot-providers/virtual_kernel then I get a bunch of other errors.
Mar 21 20:50:20 <arkver> The only way I know of to cleanly fix it is rm -rf tmp.
Mar 21 20:50:31 <arkver> Can anyone suggest anything less drastic?
Mar 21 20:51:50 <arkver> bitbake -c cleanall virtual/kernel doesn't do the trick.
Mar 21 21:03:23 <bluelightning> arkver: clean, switch back to the original, clean, then switch back to the new one and continue
Mar 21 21:12:05 <arkver> bluelightning: That works, thanks!
Mar 21 21:14:53 <bluelightning> np
**** ENDING LOGGING AT Tue Mar 22 02:59:58 2016