**** BEGIN LOGGING AT Thu Sep 21 03:00:02 2017 Sep 21 03:54:05 How do I know the kernel name to call when I do "bitbake -c menuconfig kernelname" ? Sep 21 04:15:38 User12345678: you don't need to, just run bitbake -c menuconfig virtual/kernel Sep 21 06:01:19 quick question as all this never sticks into one brain... Sep 21 06:01:32 When I want to declare an arch specific variable in a recipe Sep 21 06:02:09 like e.g.: PONIES_armv7a = "--enable-flying-ponies" Sep 21 06:02:37 Where is this coming from again ? the specific arch names I mean ? Sep 21 06:03:12 Specifically I'm curious about aarch64 and i586, I think 'i586' will be correct there Sep 21 06:05:33 hmmm, guess it is the "tune" ? Sep 21 06:05:53 coming deep from the bowels of conf/machine/include ? Sep 21 06:09:36 scraping through http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/conf/machine/include/x86/arch-x86.inc ... does this make the correct name for 64bit intel 'x86_64' ? Sep 21 06:09:59 So: PONIES_x86_64 = "--64-flying-ponies" Sep 21 06:10:21 and then later ./configure ${PONIES} ... , like that ? Sep 21 06:11:10 * gtristan saw x86-64 somewhere else and is head scratching wondering what the correct name is Sep 21 06:15:18 oh well that's weird Sep 21 06:16:24 so line 8 http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/conf/machine/include/x86/arch-x86.inc#n8 calls X86ARCH64 x86_64, and then line 43 says x86-64 http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/conf/machine/include/x86/arch-x86.inc#n43 Sep 21 07:00:48 I have sources build by CMAKE. there are kernel module sources installed with the package. how can I build my sources and then build installed kernel module? Sep 21 07:01:09 I have sources build by CMAKE. there are kernel module sources installed with the package. how can I build my sources and then build kernel module from the installed sources? Sep 21 08:01:14 somday I'll write a patch to properly handle separate /var and/or /usr... Sep 21 08:01:33 but it's tricky, it interacts with image generation, package installation etc... Sep 21 08:19:27 boucman_work: what is the use case ? Sep 21 08:20:04 nefethael: / is read-only, /var is read-write on a separate partition, packages install stuff in /var so you can't just start from an empty /var Sep 21 08:20:36 I also had customers that wanted a separate /usr, but their justification was more flacky, it was more a case of "do what the customer asks" Sep 21 08:21:12 the /var case can also be generalized depending on the system, I had people that wanted a r/w /config etc... Sep 21 08:21:47 We are going there pretty soon with our HW products. We'd like to have RO fs for most parts, and separate RW for the writeable stuff. For the purpose of reliablity. Sep 21 08:22:35 And we're not going to use unionfs or aufs or those things. I know of customers who have had tremendous problems with these in the field Sep 21 08:22:36 sveinse: right now I have a dozen purpose-built FSTYPE to do that... it's ugly but it works. Sep 21 08:23:06 boucman_work: one nice solution is overlayfs :) Sep 21 08:26:07 yes, but /var needs to be in a separate partition, so it's not overwritten at upgrade time (we overwrite the whole rootf partition) Sep 21 08:26:30 sveinse: strange that layered FS aint work for you, overlayfs is stable for my concern Sep 21 08:28:12 boucman_work: with overlayfs, you can put the overwriten files into separate partition, so that you can provide an minimal /var and user can customize it at will Sep 21 08:28:57 boucman_work: but /usr /var can always be usefull in some case i admit :) Sep 21 08:28:58 hmm Sep 21 08:29:09 +separate Sep 21 08:29:32 that would be a bit klunky with the upgrade use-case, but it could work... Sep 21 08:29:45 ;) Sep 21 08:30:22 I should also cleanup and upstream my FSTYPES which creates a single ubi containing multiple volumes (in my case : two volumes with the rootfs and one with /var) Sep 21 08:30:36 we use that kind of FS arch, i'll need to port it into yocto sooner or later ... Sep 21 08:31:08 did you have a look to swupdate ? Sep 21 08:32:05 nefethael: Have you stressed the overlayfs, e.g. by cutting power at random times? Because that is the use case for all this: A system that can survive powerkills reliably Sep 21 08:33:10 we have capacitors, so we don't have hard shutdown Sep 21 08:33:24 nefethael: right Sep 21 08:34:23 never had any issue, plus its mainline in kernel ;) Sep 21 08:34:25 nefethael: yes, but swupdate is another brick in the update problem, it's about downloading and installing. Sep 21 08:38:22 boucman_work: yep, i looked at https://github.com/sbabic/meta-swupdate-boards but there is no example for i.MX6, hard for me to understand what to do regarding SPL/wic/u-boot integration... (even if there is good doc on it) Sep 21 08:39:14 nefethael: there is no good doc on how to do that for i.MX6, afaik Sep 21 08:40:16 and I tend to not use wic, I never managed to get it to work with u-boot... it seems very EFI specific (or it was last time I looked into it) i.MX6 needs its own tool to install a SPL, I'm not sure if you can "just flash" an image to NAND... Sep 21 08:40:38 I should go back to trying wic at some point... Sep 21 08:42:39 i use it to create sdcard image with SPL, u-boot and rootfs, but didn't try for more complex arch for now, wic is quite powerful Sep 21 08:43:40 there are great improvements on serial download stuff in u-boot those days Sep 21 08:43:46 sdcards work ok, it's the nand which afaict need the spl to be installed with kobs-ng Sep 21 08:44:16 right now, i'm looking into installing the nand rootfs on the sdcard but it's a bit tricky... Sep 21 08:44:25 the sequence is Sep 21 08:44:29 1) build the nand image Sep 21 08:44:49 2) build a package that takes the nand image from DEPLOYDIR and adds it to /boot Sep 21 08:44:56 3) build the sdcard image Sep 21 08:45:05 without confusing yocto with dependencies :P Sep 21 08:45:15 ahah ;) Sep 21 08:45:46 with the nand image having a multi-volume ubi containing the rootfs.ubifs twice :P Sep 21 08:46:00 it all works pretty well, all things considered :) Sep 21 09:07:22 What is the relationship between -native and nativesdk- packages? Sep 21 09:08:55 The thing is that I'm lacking protobuf from our sdk, but I see that the sdk already contains nativesdk-protobuf, but no protoc for host. If I build protobuf-native I find the tool Sep 21 09:10:48 I'm not quite getting the nomenclature for TOOLCHAIN_*_TASK Sep 21 09:15:46 I meet a problem with building weston, I try to use the opengl from the ARM mali https://paste.fedoraproject.org/paste/SLw9Adw5EheBDPYAY44n-A Sep 21 09:16:30 I have wrote some pkg-config files for the mali package, why it still can't find the package at QA stage? Sep 21 09:27:02 what is meta-openembedded ? Sep 21 09:33:41 I have a meta-toolchain-sp which contains inherit populate_sdk, sets TOOLCHAIN_HOST_TASK+="nativesdk-packagegroup-sp". Then nativesdk-packagegroup-sp.bb inherits nativesdk packagegroup, sets RDEPEDNS_${PN} += "nativesdk-protobuf" Sep 21 09:34:12 Are there then any reason why protobuf is missing from the sdk? Am I doing this right? Sep 21 09:34:45 User12345678: basically, a collection of several openembedded layers grouped roughly by functionality Sep 21 09:35:45 Do I need to download meta-openembedded and use it with poky, or poky already include meta-openembedded? Sep 21 09:36:32 User12345678: poky is just oe-core Sep 21 09:36:55 meta-openembedded has several other layers in you can use if you want, such as meta-multimedia or meta-python Sep 21 09:37:11 User12345678: poky does not include all of it, but the core part. you do not *need* it, but it can provide a lot of packages that you might want to pull in when going further Sep 21 09:37:30 I see Sep 21 09:39:14 Does anyone have any good examples on how to make sdks and how to include things into them? Sep 21 09:40:06 I would need to use a cross-compiler, do I need to do anything with poky, or do I just download poky then bitbake my image? Sep 21 09:40:50 User12345678: by default, poky takes care of all that. Sep 21 09:41:00 User12345678: the first thing bitbake does is build the cross compiler Sep 21 09:41:10 I see Sep 21 09:41:14 User12345678: you might want to go through the yocto project quick start document to see what it actually needs and outputs. Sep 21 09:41:30 sveinse: easy way is to add TOOLCHAIN_HOST_TASK extensions to the image that you want the sdk to be based on Sep 21 09:41:53 sveinse: what command do you use to build your sdk ? Sep 21 09:42:22 (and i'm not sure how to tell yocto that a -native package should also go in the sdk... there is probably a way to do it) Sep 21 09:42:46 rburton: did you see my question above about TOOLCHAIN_HOST_TASK and that nativesdk-protobuf does not end up in my sdk? Sep 21 09:42:51 you can't put native in a sdk, that's what nativesdk is for Sep 21 09:43:52 From searching the codebase in Pyro, I think that is still the way to do it Sep 21 09:44:07 or am I mistaking? Sep 21 09:45:07 LetoThe2nd, I have followed the Quick start guide already and built the image. Now I'm following another guide but it says to clone linaro after cloning poky. Why? Sep 21 09:45:22 User12345678: because you're following the linaro guide? Sep 21 09:45:48 User12345678: no idea, as i do not know that specific guide. probably you skipped the introductory paragraph that mention the document being for some specific case. Sep 21 09:45:55 No, it's a yocto built for renesas r-car arm processor Sep 21 09:46:08 yeah so you need a specific bsp for that, which is probably in meta-linaro Sep 21 09:46:41 User12345678: read "for some specific soc" as "some specific use case" Sep 21 09:47:05 sveinse: that should work without seeing actual recipes. are you sure nativesdk-protobuf contains the binaries you're after? i'd be using -e to verify the additions you expect are happening. Sep 21 09:47:13 Can I still not use the linaro and just bitbake poky directly? Sep 21 09:47:45 User12345678: poky alone certainly does not include board support for your thing Sep 21 09:48:20 User12345678: if you're targetting a specific machine then you'll need to use the bsp for that machine. poky itself only supports qemu machines, and a few minimal machines for QA purposes. you're *meant* to glue pieces together for your purposes, poky is just an example. Sep 21 09:49:38 So, Poky doesn't have cross compiler for ARM included? Sep 21 09:50:01 yes, it does. but it doesn't know the precise details of a renesas r-car arm processor Sep 21 09:50:13 User12345678: it does build a cross compiler if needed. it just does not know about your specific renesas soc Sep 21 09:50:19 what config flags does the kernel need. what arm tune is it. how does it want a file system laid out. Sep 21 09:51:05 I see. It's clearer now Sep 21 09:51:32 rburton: hopefully a heavy metal tune Sep 21 09:54:53 But how would meta-Linaro knows about the details of the r-car arm processor? Unless Linaro is affiliated with Renesas and got the details of the implementation, I would think it should be a layer that renesas provide instead, isn't it? Sep 21 09:55:47 yes, unless linaro have done it for renesas. you're the one that said you've got linaro for the r-car processor. Sep 21 09:56:41 you always need a Board Support Package, oe-core only comes with virtual processors, poky adds some basic real ones for QA. meta-intel has proper x86 ones, meta-ti has some TI boards, meta-freescale has the freescale boards. you'll need one for your target. Sep 21 09:59:18 I see. Clearer now. Thanks Sep 21 09:59:19 rburton: given that description, one would not need a bsp when not targetting a processor. Sep 21 09:59:37 (which i think is a valid usecase too) Sep 21 10:00:10 how would you build something when not targetting a processor? code has to run on something... Sep 21 10:01:12 rburton: i admit i was taking the wording a bit literally - i'm thinking of container creation. where you *basically* just target an arch abi Sep 21 10:03:39 so yeah they exist - BSPs that target epic xeon processors with all the optimisations and kernel tweaks for super speed Sep 21 10:05:20 i know, i know. just happen to be tinkering with some stuff in that direction, and the bspness i'm currently using is basically just generic x86-64 + linux-dummy Sep 21 10:09:35 you really want a custom bsp, you want the virtio drivers and support for the target hardware Sep 21 10:09:41 genericx86 has quite a low CPU bar Sep 21 10:25:28 hello guys ! I have a problem with an aoutotoll recipe. I have created a recipe that fetch the source code from a git repo, and use the "inherit autotools" directive to configure/compile/install the code. BY the way, running "bitbake -c configure thisprogram" ends up with the following error: we can access libc without dlopen .. no, we can access libc with libc.so....no, we can access libc with /usr/lib/libc.so*... no, error: cannot figure Sep 21 10:25:28 out how to access libc Sep 21 10:25:34 i'm amused Sep 21 10:26:12 sysroot=/home/.../build/tmp/sysroot/raspberrypi0 Sep 21 10:27:24 and if "$ find /home/.-../raspberrypi0 -name libc.so" returns "/home/.../build/tmp/sysroot/raspberrypi0/usr/lib/lib.so" Sep 21 10:27:57 fberg: sounds like something in the configure.ac is hardcoded. Sep 21 10:29:05 from the log file I can see that the prefix variable is set to: prefix=/usr Sep 21 10:29:37 fberg: OE does that, and imho it is unrelated. Sep 21 10:29:57 exec_prefix bindir sbindir etc etc are all set as standard Sep 21 10:30:09 fberg: i'd look into the configure.ac and check if there is something hardcoded there. Sep 21 10:30:47 I'm sorry, I can understand what you mean but can you give me an example ? Sep 21 10:31:11 fberg: open the file and see if you can find a written out path :) Sep 21 10:32:30 i have a configure.in file only and I think it's the same Sep 21 10:33:12 fberg: nope, configure.in gets automagically created in the process, usually. Sep 21 10:33:24 fberg: is this some publicly visible source? Sep 21 10:34:17 yes: https://github.com/mariusae/trickle Sep 21 10:36:44 fberg: https://github.com/mariusae/trickle/blob/master/configure.in#L182 Sep 21 10:37:04 fberg: there is some non-sysroot aware hackery Sep 21 10:38:11 mmmm Sep 21 10:38:35 f*** Sep 21 10:38:38 fberg: you basically have to path configure.in Sep 21 10:38:43 s/path/patch/ Sep 21 10:39:30 still don't know what to patch :-) Sep 21 10:39:52 i'm pretty new to the world.. power to the noob Sep 21 10:40:33 LetoThe2nd: no configure.in is the source, old name for configure.ac Sep 21 10:40:49 oh god that configure is horrible Sep 21 10:40:50 fberg: i personally would try to find out why it contains this manual testing anyways Sep 21 10:41:03 rburton: to 1) ah thanks 2) exactly Sep 21 10:41:16 3) I agree on both Sep 21 10:41:25 *at least* wrap AC_TRY_RUN in a cache check so you can trivially seed the right value Sep 21 11:44:26 hi all Sep 21 11:44:31 i have an issue Sep 21 11:44:40 i ran bitbake u-boot Sep 21 11:45:15 and after that i can't find anything in the git dir under tmp/work ... Sep 21 11:46:09 mmmm Sep 21 11:46:56 during the test I see that there is the AC_TRY_RUN directive, so I suppose it wiil try to run on my host the compiled code for the target and hence the error Sep 21 11:47:00 Costin1: what are you looking for? Sep 21 11:47:02 Is it correct ? Sep 21 11:47:22 fberg: AC_TRY_RUN will fail if it knows its cross-compiling, the test should cater for that Sep 21 11:47:37 fberg: (which is why i said its horrible and autoconf says don't do this) Sep 21 11:47:44 lol Sep 21 11:47:48 i look to build u-boot and get the source code for it Sep 21 11:48:03 i need it for debugging purpose of U-boiot Sep 21 11:48:27 Costin1: and you've looked in tmp/work/[machine arch]/u-boot/[version]? Sep 21 11:48:33 yes Sep 21 11:48:41 Costin1: do you have rm_work enabled? Sep 21 11:49:07 i am not sure Sep 21 11:49:15 how do i enable/disable it? Sep 21 11:49:16 if you inherit rm_work in local.conf you are Sep 21 11:49:31 let me check Sep 21 11:49:39 the other possibiltiy is that it just pulled the uboot packages from sstate so didn't need to build anything Sep 21 11:50:00 so 'bitbake u-boot -c patch -f' will force it to unpack and patch u-boot Sep 21 11:50:23 i don't have rm_work Sep 21 11:51:00 i also get this warning Sep 21 11:51:01 WARNING: /home/testfarm/QorIQ-SDK-V2.0-20160527-yocto/sources/meta-nxp-npi-1703/recipes-bsp/u-boot/u-boot-qoriq_2016.09.bb.do_patch is tainted from a forced run Sep 21 11:51:17 thats because you passed -f, ignore it Sep 21 11:51:23 ok Sep 21 11:52:11 i got it now Sep 21 11:52:20 i mean sources are in git dir Sep 21 11:52:25 one more things Sep 21 11:52:48 when doing a build i should also get the elf file Sep 21 11:52:54 is this in git dir too? Sep 21 11:53:56 build results are probably in a build/ directory alongside, but if it did pull from sstate then they won't exist as it didn't build anything Sep 21 11:54:25 so they'll be in tmp/deploy Sep 21 11:55:21 i have the .bin indeed Sep 21 11:56:00 but i should have had an elf file or something to import in eclipse and do debug on it Sep 21 11:58:53 e.g for kernel i have in linux-qoriq/4.1-r0/build the vmlinux file Sep 21 12:15:51 do we really need rpcgen in HOSTTOOLS? it was added in original RP's commit fa764a403da34bb0ca9fa3767a9e9dba8d685965 but I haven't found any explanation why we need it Sep 21 12:16:11 some distributions seem to disable building /usr/bin/rpcgen with glibc-2.26 Sep 21 12:42:57 JaMa: I'm pretty sure something depended upon it or I'd not have added it. I would be happy to remove it if we can Sep 21 12:49:23 hi guys, i fired a script in /etc/rc.local and now my system does not show the console login anymore, i cant login to it from the network too, i can go into the boot prompt is there a way to recover from that instead of formattin? Sep 21 12:50:10 pagios: pass the kernel init=/bin/sh or something alike Sep 21 12:50:28 LetoThe2nd, how can i do that ? i am at boot:> Sep 21 12:50:50 i have ubifsload- command etc Sep 21 12:51:03 pagios: i have no idea where your propt and what your bootloader is ;-) Sep 21 12:51:52 in uboot, the arguments that are passed to the kernel are in bootargs Sep 21 12:52:06 but often the scripts generating it are a bit convoluted Sep 21 12:52:09 boot bootargs=init=/bin/sh? Sep 21 12:52:35 if i type boot it boots normally but hangs Sep 21 12:53:23 more like setenv bootargs init=/bin/sh, then boot. Sep 21 12:53:31 setenv bootargs init=/bin/sh Sep 21 12:53:35 nice Sep 21 12:53:58 but like i said, the boot command can certainly include magic that overrides it. check the kernel log, it tells you the given arguments right at the beginning Sep 21 12:54:54 yea it doesnt owrk Sep 21 12:58:42 RP: ok, will try to build world without it Sep 21 13:00:10 JaMa: the creation of the original list was really about documenting what we needed 'today'. I do want to try and reduce the list where it makes sense Sep 21 13:00:23 JaMa: sadly I haven't had the time to actually do that... Sep 21 13:00:30 (yet) Sep 21 13:02:12 LetoThe2nd, no good :/ Sep 21 13:02:44 pagios: boot from an sd card and mount the filesystem, that would be another way Sep 21 13:03:03 no sd card :/ Sep 21 13:03:10 only the built in flash Sep 21 13:04:01 pagios: its basically things that can happen to you when you develop stuff in-target, instead of an outomated build, or at least boot from nfs ;-) Sep 21 13:04:16 pagios: boot from nfs then maybe... Sep 21 13:05:27 no network card Sep 21 13:06:05 *shrug* Sep 21 13:06:25 then either dig through your u-boot stuff, or do a reflash Sep 21 13:09:52 LetoThe2nd, seems i have OTG Sep 21 13:10:02 i flash the ubi file to usb using dd? Sep 21 13:10:24 pagios: this is highly board specific, i don't have the slightest clue. Sep 21 13:13:52 LetoThe2nd, can i install an imx6 imahe on usb and boot from it ? Sep 21 13:13:56 like something generic Sep 21 13:15:05 pagios: highly board- and bootloader specific. and the bootloader magic is probably more complicated than just changing the bootargs Sep 21 13:25:59 i need help with errors when building with dora 1.5 on ubuntu 16LTS https://pastebin.com/MsCajZkY seems like compiler version issues to me Sep 21 13:26:21 zarzar, dora is ancient Sep 21 13:26:27 i'm building core-image-sato\ Sep 21 13:26:29 likely going to be a source of ongoing pain Sep 21 13:26:45 for the reasons you are seeing Sep 21 13:27:07 i know but i need to match a system config including yocto and kernel and compiler for FIPS compliance Sep 21 13:27:15 ouch Sep 21 13:27:40 zarzar, I have vague understanding of FIPS, but that is insane Sep 21 13:27:54 do you have anyone actually supporting dora? Sep 21 13:27:58 yes very painful but i need to get close as possible, yocto 1.5, kernel 3.10, gcc 4.8.1 Sep 21 13:28:04 FullyInsaneProtocolSpecification? Sep 21 13:28:08 lol Sep 21 13:28:19 haha it is insane yes Sep 21 13:28:22 zarzar, use an older linux distro to do the build Sep 21 13:28:28 I use docker in such situations Sep 21 13:28:41 and warn customer they are risk for CVE's Sep 21 13:28:52 yea i could use older ubuntu, but i like my 16LTS VM :( Sep 21 13:29:03 well Sep 21 13:29:13 chose your pain Sep 21 13:29:15 LetoThe2nd, https://en.wikipedia.org/wiki/Federal_Information_Processing_Standards Sep 21 13:29:17 yea, its a no win situation for me Sep 21 13:29:38 exactly Sep 21 13:29:40 Crofton|work: hrhr Sep 21 13:29:46 i actually really only need the compiler for application debugging Sep 21 13:29:52 which FIPS btw Sep 21 13:29:59 140-2 Sep 21 13:30:33 i have not been able to find the same version of 4.8.1 armhf cross toolchain anywhere else Sep 21 13:32:02 well, look slike your choices are clear Sep 21 13:32:10 charge extra for this Sep 21 13:34:42 does yocto build the compiler from source? hopefully it does not Sep 21 13:34:56 by default, it does. Sep 21 13:35:13 and its openembedded/poky, not yocto ;-) Sep 21 13:36:37 bummer Sep 21 13:36:56 there is way to use an external toolchain, though. should be in the docs. Sep 21 13:37:01 any idea if the built version is available somewhere? Sep 21 13:44:31 is this the armhf cross toolchain for 1.5? http://downloads.yoctoproject.org/releases/yocto/yocto-1.5/toolchain/x86_64/ Sep 21 13:45:38 you could try it Sep 21 13:46:55 looks like it is not cross toolchain, hard to tell from the write up Sep 21 14:02:09 If I get ERROR: Feature 'xcomposite-glx' was enabled, but the pre-condition 'features.wayland-client && features.opengl && !features.opengles2 && libs.xcomposite && libs.glx' failed. while doing: bitbake -k meta-toolchain-qt5 https://pastebin.com/APbQZbjs while trying to a Qt SDK for the Raspberry Pi on an image where I don't want X11 or Wayland, with default DISTRO_FEATURES , does this mean that I either need to exclude X11 and Wayland with Sep 21 14:02:11 DISTRO_FEATURES_remove = " x11 wayland" and then qtwayland won't be build in the Qt SDK ? Sep 21 14:04:17 ubuntu 12 for dora? or ubuntu 10? Sep 21 14:12:46 hi everyone Sep 21 14:14:18 i was wondering, how do i select only some recipes from an already publicy-available layer? like, i need to have docker inside my embedded system image, but the meta-virtualization layer that includes it carries also loads of other packages (like python or xen) that i definitely don't need Sep 21 14:15:02 shoukd i use bbappend? or just create an empty layer and copy-paste the docker recipe? (i don't like the second option) Sep 21 14:16:54 SonicPenguin: they won't get built unless you ask them to be built Sep 21 14:17:22 like adding meta-python with his 100 python modules won't add all of them to your image, just means you *can* build them Sep 21 14:17:54 rburton, ok thank you Sep 21 14:18:44 rburton, i know i can find the answer in the docs, but i'll ask you: where do i define the recipes in that layer that i want to build? Sep 21 14:19:35 edit the image recipe that you're building Sep 21 14:22:24 SonicPenguin, http://opensdr.com/posts/building-small-gnuradio-images/ Sep 21 14:22:27 might help Sep 21 14:23:02 not exact steps for yuo, but should give you ideas Sep 21 14:24:02 Crofton|work, thank you, that's what i needed. frankly even if there's a lot of documentation for this project, i still feel lost Sep 21 14:25:00 there are a lot of knobs Sep 21 14:25:14 I guess why should think about some more how to get real work done kind of things Sep 21 14:27:28 exactly, some kind of quick starting guide or cookbooks for the most common cases. the mega manual is great but feels like a university book Sep 21 14:28:12 SonicPenguin: quickstart? Sep 21 14:28:28 mega manual is all of the docs together in one epic bundle specifically for searching fulltext Sep 21 14:28:33 otherwise, drill into the right subsection Sep 21 14:35:06 rburton, yes, it's the same anyway. quick start is pretty useless because just shows the installation and the building of a premade image Sep 21 14:35:24 development manual has more "next step" sections Sep 21 14:36:03 http://www.yoctoproject.org/docs/2.3.1/dev-manual/dev-manual.html#usingpoky-extend-customimage Sep 21 14:36:45 What is poky? Sep 21 14:36:56 Crofton|work: shuttup Sep 21 14:37:29 lol Sep 21 14:37:41 Crofton|work: ... and dance! Sep 21 14:38:09 Crofton|work: in fairness its only in the anchors, the text doesn't say it Sep 21 14:38:14 https://twitter.com/evilbmcats/status/910628639523340290 Sep 21 14:38:50 do any sections do a good job describing how MACHINE, DISTRO and image should be orthogonal Sep 21 14:41:53 rburton, ok this is something. but it should be on top of guide! Sep 21 14:42:52 SonicPenguin: patches welcome for the docs, but http://www.yoctoproject.org/docs/2.1/yocto-project-qs/yocto-project-qs.html#qs-next-steps links to http://www.yoctoproject.org/docs/2.1/dev-manual/dev-manual.html#extendpoky Sep 21 14:59:59 I am looking at a component that fails when building the nativesdk version. In particular there is a bbappend that performs a do_install_append() and installs files but are not shipped (packaged) in the SDK build, and that then fails on QA. What's a good way to build the native component but NOT apply this append (or the do_install_append) for the SDK build only (still apply for target build)? Sep 21 15:07:46 gunnarx: use a class override Sep 21 15:08:00 eg do_install_append_class-target will only run in target builds Sep 21 15:08:18 ah perfect, makes sense. Sep 21 15:08:23 thanks, I'll try Sep 21 15:08:45 there's a similar override for each class, -target -native -nativesdk etc Sep 21 15:25:19 Yeah, I expected as much. I'm used to MACHINE overrides also. Problem is I do only like 10-20% code development nowadays and can't keep all the features in my head. :) Sep 21 15:26:09 armpit: next time you fire a build on the AB there's a dropdown where you can select a release name and it'll pre-populate all of the branch fields Sep 21 15:26:30 joshuagl, k, thanks Sep 21 15:27:21 * rburton cheers agai Sep 21 15:28:54 rburton: have you tried it yet? Sep 21 15:29:36 not yet Sep 21 15:33:06 tsk Sep 21 15:33:14 2.4 M4 M+ my foot Sep 21 15:39:33 RP: rburton: do you have AB runs planned for this evening? Sep 21 15:39:40 i'll fire a mut shortly Sep 21 15:39:42 * joshuagl would like to try and repro the eclipse-poky-neon issue Sep 21 15:39:44 joshuagl: I've found a bug in those dropdowns Sep 21 15:39:53 RP: good, what is it? Sep 21 15:40:01 joshuagl: trigger that build and kill everything but neon? Sep 21 15:40:13 joshuagl: meta-gplv2 only has pyro and master branches Sep 21 15:40:35 RP: gah! Sep 21 15:40:45 RP: didn't realise, will add some special casing Sep 21 15:41:14 joshuagl: meta-qt3 won't get a rocko branch afaik Sep 21 15:41:36 I can see this is going to be "fun" to maintain :-) Sep 21 15:41:56 joshuagl: thankfully the exceptions get fewer going forwards Sep 21 15:46:08 it will only cause issues with morty Sep 21 15:46:15 armpit: quite the collection of pyro patches, thanks. Is there a bitbake branch anywhere with the bitbake patches in? Sep 21 15:47:04 RP, err nope Sep 21 15:47:31 armpit: ok, np Sep 21 15:47:56 working on morty kernel updates now Sep 21 15:53:35 RP: thanks for testing, I have a patch. Will test and deploy later/tomorrow Sep 21 15:56:11 joshuagl: thanks! Sep 21 15:56:27 armpit: can you check I've merged everything in pyro please? Sep 21 15:57:35 is the cross toolchain for armhf built into binaries during a yocto bitbake build for an armhf target/ Sep 21 15:57:44 * armpit looks Sep 21 15:58:12 RP looks good. thanks Sep 21 15:59:37 armpit: thanks Sep 21 16:17:17 armpit: I've missed patches :( Sep 21 16:20:21 armpit: sorted now I think Sep 21 16:35:24 Hi, Im looking into setting up a ptest framework, does anyone know how this chart is populated? https://wiki.yoctoproject.org/wiki/Ptest_6964efddd31c479386d1643c1025bc102710392f Sep 21 16:35:38 if there are already scripts available for this? Sep 21 17:02:08 rburton: I want to have git versions of recipe for gcc and friends, using the new incantation you added is there some quick howto ? Sep 21 17:04:23 is it devupstream.bbclass ? Sep 21 17:51:16 hi, is there a mechanism/best practice for initial system preparation on first boot? e.g. setting hostname etc. Sep 21 17:56:41 Guest51513: I dont think theres a single mechanism for that, you may use a combination of classes to achieve that though, like useradd.bbclass + rootfs_postinstall.bbclass and such Sep 21 18:01:45 aehs29: iirc there was a runonce example for sysvinit. sounds like I have to write my own stuff for something I expected that many others have the same issue Sep 21 18:17:06 Is it possible to define or call python functions in local.conf, site.conf or distro.conf? Sep 21 18:17:52 The use case is to read and parse a central given VERSION file loading a value into the SP_VERSION Sep 21 18:18:21 Can I include a bbclass from these files and use these functions? Sep 21 18:18:27 you can call them from inline python in a .conf, but the function definition itself would have to be placed in a bbclass in INHERIT, or in a .inc file required from the .conf Sep 21 18:19:15 So the use of "inherit" in e.g. distro.conf is ok? Sep 21 18:24:09 you need to do use INHERIT Sep 21 18:42:04 I'm attempting to make my very first distro based on poky, and based on the mess of local.conf, I have put this into distro/sp-poky.conf: https://bpaste.net/show/9874aeed86f1 . Anything that stands out NOT being directives for a distro and should remain in local.conf? Sep 21 18:59:01 Is it possible to conditionally append to bitbake variables based on an environment variable? Sep 21 18:59:45 I have an environment variable that changes the flavor of our build. Depending on the flavor I need to append an extra install file. Sep 21 19:00:10 There's probably a better architecture but I'm not the right person to re-architect it... Sep 21 19:00:44 I ended up pre-parsing local.conf prior to running, bitbake, but that is certainly not an ideal way of doing it Sep 21 19:13:13 sveinse: thanks, I hope to avoid that ;-) Sep 21 19:16:17 Can a distro definition sit in multiple layers? By that I mean, have a base distro in one layer and amend (bbappend?) to it in another? Sep 21 19:31:27 How can I access a file (a VERSION file which is located in the root of the layer) from a bbclass file? Sep 21 19:31:50 Does the bbclass file have a THISFILE path or similar? Sep 21 19:36:09 khem: yes, devuptream. doesn't work for native yet though, so doubt its useful for gcc Sep 21 19:45:28 Interesting. There seems to be no SRCDIR or similar. I actually have to parse through each entry in BBLAYERS to find the path to my layer :o Curious Sep 21 19:47:56 Is there any particular reason for not having a variable reference to the current (bitbake) source file? Sep 21 20:09:28 any comments on ADT bug? https://bugzilla.yoctoproject.org/show_bug.cgi?id=12095 Sep 21 20:09:30 Bug 12095: normal, Medium, 2.4 M4, brian.avery, NEW , Toolchain.Host.Mismatch problem Sep 21 20:10:45 rburton: gcc doesnt have native but all other variants are there Sep 21 20:39:10 I got a task here which results in ERROR from sstate_task_postfunc which is a py function. It does not log anything why or when. I expect an exception being raised behind the scenes. How can I enable bitbake to print the exception? Sep 21 20:39:19 I've tried -DDDDDDD in vain Sep 21 20:40:56 yeah python functions are notorious Sep 21 20:42:27 Hello folks Sep 21 20:45:50 hello Otavio Sep 21 20:46:33 Hi, Im looking into setting up a ptest framework, does anyone know how this chart is populated? https://wiki.yoctoproject.org/wiki/Ptest_6964efddd31c479386d1643c1025bc102710392f if there are already scripts available for this? Sep 21 20:47:35 Guest23341, maybe a Yocto bug would be in order Sep 21 20:59:26 RP: have you seen https://www.python.org/dev/peps/pep-0538/ ? Sep 21 20:59:47 was just reading https://docs.python.org/3/whatsnew/3.7.html Sep 21 21:00:36 oh, that's interesting. new argparse feature "The parse_intermixed_args() supports letting the user intermix options and positional arguments on the command line, as is possible in many unix commands." Sep 21 21:06:19 khem: nice, just saw the lwn article about 4.4 and 4.9 kernels having patches to build with clang. good stuff Sep 21 21:07:11 kergoth: The Google Android guys have been busy upstreaming LLVMLinux patches. Sep 21 21:07:20 very cool Sep 21 21:07:24 And fixing new issues. Sep 21 21:07:47 kergoth: Both google and Linaro are finally kicking butt. Sep 21 21:08:02 Which is nice. Though I wish they had listened to me 5 years ago. Sep 21 21:09:19 Both camps have apologized to me that they didn't help LLVMLinux project back when we started... They get it now. Sep 21 21:09:35 I'm just glad that value is finally understood. Sep 21 21:14:30 kergoth: interesting, I had not Sep 21 21:16:22 Android and ChromeOS are both pretty close to being clang-only these days. Sep 21 21:17:27 Along with FreeBSD, Mandriva, and of course iOS and MacOS... Sep 21 21:47:29 n Sep 21 22:23:56 khem: well actually, if it doesn't use bbclassextend you'll be okay (it can't extend an extended recipe yet) Sep 21 22:26:02 rburton: ok Sep 21 22:26:12 then i am out of luck Sep 21 22:26:32 rburton: btw. I am sending a llvm ver bump to final 5.0 release Sep 21 22:28:19 kergoth: yes clang community is very easy going.. no snotty people, and very responsive Sep 22 02:43:13 does anyone here have experience with building qt for imx based boards? I have a board I am building qt5 for, and a qt app, when I try and use qt media player from the qtmultimedia recipe, I get an error complaining that the service isn't found. Wondering if I am missing something in my yocto config **** ENDING LOGGING AT Fri Sep 22 03:00:00 2017