**** BEGIN LOGGING AT Tue Aug 29 03:00:01 2017 Aug 29 05:18:08 Hey everyone -Hello everyone. Aug 29 05:18:14 Anyone alive this time of night? :) Aug 29 07:15:50 tcpdump: for some its now morning :) Aug 29 07:23:41 RP, did you happen to see rakuco's mail on yocto list on friday "Pyro's uninative and libstdc++ symbols"? he was asking me about the issue and I probably wasn't very useful... Aug 29 07:24:23 my guess was to check for host library path leaks while linking the native tool, does that sound possible? Aug 29 07:45:05 morning Aug 29 08:03:33 jku: replied Aug 29 08:03:41 jku: short version, need to upgrade uninative in pyro Aug 29 08:15:30 thanks Aug 29 08:29:27 hi all, i dont see any Packages.gz in my directory how can i generate it Aug 29 08:35:20 Hi, i have a package that should be rebuild when a change to the kernel or dtb was made. Do i just add DEPENDS += "virtual/kernel" to do this? Aug 29 08:39:47 pagios: bitbake package-index Aug 29 08:39:56 fl0v0: yes Aug 29 08:44:49 bluelightning: thank you! Aug 29 08:56:50 bluelightning, is there a way to set a dependency from an image to "package-index" ? Aug 29 08:59:38 aurele: yes, you'd just set it up as a task dependency i.e. do_build[depends] += "package-index:do_package_index", though I think after building the image you should find the package indexes are populated already - at least it was that way at one point Aug 29 09:06:51 Suppose I have an existing recipe and I want to make and test, on my embedded device, some significant changes to the source code of the corresponding application, what would be the recommended workflow to do this? Aug 29 09:09:40 wouterstreamit: http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#devtool-use-devtool-modify-to-enable-work-on-code-associated-with-an-existing-recipe Aug 29 09:14:52 bluelightning, what is the use of the packages.gz file Aug 29 09:15:31 bluelightning: thanks, that's what I was looking for Aug 29 09:17:02 bluelightning, thanks Aug 29 09:23:44 bluelightning, should the packages.stamps be generated too? Aug 29 09:23:50 .flock Aug 29 09:24:28 pagios: package index for the package manager to read from, as far as I'm aware Aug 29 09:24:31 pagios: not sure Aug 29 09:25:11 wouterstreamit: let us know if you have any further issues / questions - I'm particularly interested to hear from people using devtool as a lot of it was written by me :) Aug 29 09:28:49 bluelightning, where can i start with yocto and all this new thing to me Aug 29 09:30:25 pagios: I guess it depends on what you want to do :) presumably you have at least been able to set it up so you have run through the quick start guide? Aug 29 09:31:04 bluelightning, i am prototyping an IoT software to be used on a yocto OS embedded system Aug 29 09:33:33 pagios: ok, you may want to look at the extensible SDK, or at least devtool if you are able to set up a working build environment - this should help you get your code integrated into the image; there's some coverage of it in the development manual Aug 29 09:39:27 bluelightning, what i am concerned with is later when i have like 50 iot devices and want to roll a package update or security update on them, i can do that from a remote package manager so my iot devices need to read from that remote server is this possible using yocto Aug 29 09:44:19 hi, i'm trying to build a custom recipe that's using cmake and cmake depends on matlab (it builds fine without yocto/bitbake). but cmake fails to find matlab once i try to build it with bitbake Aug 29 09:44:51 guess this is because bitbake sandboxes the build? Aug 29 10:12:11 pagios: you can, but typically in a scenario like that you would use something with a bit more management capability like mender.io I would think Aug 29 10:12:17 pagios: this may be of interest: https://www.linux.com/videos/surviving-wilderness-integrity-protection-and-system-update-patrick-ohly-intel-gmbh Aug 29 10:13:13 bluelightning, yea idea is to have these boxes secure all the time Aug 29 10:13:14 crazy_imp: yes it does Aug 29 10:13:16 thanks i will check Aug 29 10:13:58 bluelightning: can i somehow tell it to make an exception for matlab? Aug 29 10:14:59 crazy_imp: depends - is it trying to link to a library or just run a command? Aug 29 10:17:04 bluelightning: i've told cmake to require matlab because it's needed to generate some source files, besides calling matlab for it, there's no linking or such involved Aug 29 10:19:06 crazy_imp: ok - assuming you are using a recent version (pyro/2.3 or master) then it ought to be as easy as adding the matlab command(s) to HOSTTOOLS Aug 29 10:19:59 bluelightning: morty... Aug 29 10:20:24 crazy_imp: right, in that case it is simply not understanding where it needs to look Aug 29 10:20:49 crazy_imp: you'd have to dig into the cmake files to see how it is looking for matlab Aug 29 10:21:37 I have dug into cmake that way before but unfortunately I don't know much about its matlab integration... I guess it would be in a FindMatlab.cmake file as part of cmake itself Aug 29 10:25:34 bluelightning: yes, there's one. see for the error i got and (what i suspect) the relevant part from the FindMatlab.cmake file https://pastebin.com/6LXLEEiC Aug 29 10:28:40 crazy_imp: hmm, can you perhaps set Matlab_ROOT_DIR ? Aug 29 10:29:24 you'd presumably do EXTRA_OECMAKE += "-DMatlab_ROOT_DIR=/path/to/somewhere" or something like that Aug 29 10:32:26 bluelightning: where would i put the latter? inside the recipe or as enviroment variable for bitbake? Aug 29 10:38:13 crazy_imp: in the recipe Aug 29 10:41:15 doesn't change anything Aug 29 10:45:08 oh, wait Aug 29 10:45:19 guess i have to set all 5 variables Aug 29 10:55:54 hi again, would it be possible to have "DATETIME" in the sdk name? I tried but it seems to make some problems Aug 29 11:03:12 I would assume bitbake recipes are parsed in the order they're written? So, for example, if you tried to install a package that had missing dependencies it would fail? Aug 29 11:03:29 Hipefully Im using the right words to describe that. Aug 29 11:07:08 tcpdump: actual parsing order shouldn't really matter (a recipe can't affect another directly). Task execution on the other hand is carefully controlled (so a library is compiled and put into sysroot before an app (that depends on the library) is configured and compiled Aug 29 11:07:54 jku: I think I may actually be asking about taske execution... Aug 29 11:07:55 1 sec Aug 29 11:08:19 tcpdump, If you didn't set a dependency in a package the compile task will fail if no other package pulled the needed package, and I had such kind of missing dependecies, and bitbake kindly warned me that a dependecy was missing in my recipe Aug 29 11:08:54 tcpdump, if a dependecy you set doesn't fit any recipe, then bitbake will end with an error Aug 29 11:09:25 tcpdump, I hope that is what you were looking for ;) Aug 29 11:09:32 tcpdump: In addition to that the runtime dependencies are handled automatically for the common case (normal dynamic linking) so if you add the app package to your image, the library gets added automatically Aug 29 11:10:24 bluelightning: got it running now, thanks for the pointer :) Aug 29 11:10:34 Just so I make sure Im asking the right questions... a .bb file is called a what? Aug 29 11:10:38 Thats not a recipe is it? Aug 29 11:10:46 yes, recipe Aug 29 11:10:52 OK - got it.. Aug 29 11:12:28 So, I have a .bb recipe that calls a lower level (farther down the path) .bb file that includes some more proprietary functions. For the sake of this discussion we will say that a.bb has a line that references b.bb IMAGE_INSTALL_append = " b" Aug 29 11:12:43 bluelightning, bitbake question, i create on my local pc a web directory storing all my ipks, pointed my other embdeed system to my localmachine web directory using opkg.conf, opkg update gives a weird error parse_from_stream_nomalloc missing new line character at end of file! Aug 29 11:13:10 tcpdump: so you have a image recipe that adds things to the image? Aug 29 11:13:19 So when I make a build and have IMAGE_INSTALL_append = " b" included my build fails. When I comment it out the build works, but of course doesn't include any of the functions of b.bb. Aug 29 11:13:31 tcpdump: the things it adds are actually packages, not recipes Aug 29 11:13:42 So the error it gives me is really generic. Is it safe to assume there's something wrong with my b.bb recipe? Aug 29 11:14:16 jku: sure the b.bb file is the recipe that does things like install packages, create symlinks, etc. Got that. Aug 29 11:15:07 parse_from_stream_nomalloc: Missing new line character at end of file! Aug 29 11:15:20 how can i clean obkg Aug 29 11:17:43 tcpdump: just making sure: the "b" you add to image is not a recipe but a package (that might be produced by recipe "b"). You'll have to show the error message for any more detailed advice Aug 29 11:24:50 jku: https://pastebin.com/zctWMrJ3 Aug 29 11:24:58 opkg.conf line: src/gz cortexa9hf-neon http://12.12.12.12/ipk/cortex is this correct as an entry? Aug 29 11:25:02 I sanitized the names of a few things for the sake of identity Aug 29 11:25:07 But theres the generic log. Aug 29 11:25:10 i do opkg update it works but opkg install says nothing to install Aug 29 11:25:25 It doesnt even reference the "b.bb" recipe. Aug 29 11:25:31 In the errors. Aug 29 11:26:57 hello? Aug 29 11:27:33 Does a recipe care about spaces, tabs, etc? For example, one of my devs made the b.bb file "pretty" using tabs/spaces. Aug 29 11:27:35 Ex: https://pastebin.com/zscdyFu3 Aug 29 11:29:31 tcpdump: yikes. Something fails during package installation... Was there anything more useful in log.do_rootfs.28655 that it links to? Aug 29 11:30:01 There was not jku ... Aug 29 11:30:14 so I just did bitbake b Aug 29 11:30:23 and ran just the b.bb recipe, and that didnt fail... Aug 29 11:30:58 kanavin: https://pastebin.com/zctWMrJ3 -- is that dnf giving up with "error: unknown error"? Aug 29 11:31:06 I think the order of where b.bb is referenced in my a recipe may be the issue. Aug 29 11:32:37 Alright, so, Ive changed the order of operation in my higher level recipe (in githuv), and I want to rerun it.. Is there a way to pull down just that change without having to do it all from complete scratch again? Aug 29 11:32:52 jku: could also be one of package scriptlets failing Aug 29 11:33:18 Sorry for my neglence: Aug 29 11:33:18 I do have my distro.conf file with a lot of IMAGE_INSTALL += "foo" Aug 29 11:33:18 I wanted to add image only when ${MACHINE}=XXX is used, so I did: Aug 29 11:33:18 IMAGE_INSTALL_XXX = " spiutils" Aug 29 11:33:18 But then it shown image shrunk from 100 MiB to 4MiB :-) Aug 29 11:33:18 Why I do need to use IMAGE_INSTALL_append_XXX = " foo" to make the image 100 MiB (correct) again? Aug 29 11:34:44 Any help? Aug 29 11:34:58 lukma: because IMAGE_INSTALL_XXX = " spiutils" means you want _only_ spiutils in your image Aug 29 11:35:23 IMAGE_INSTALL_XXX += "spiutils" Aug 29 11:35:27 jku, can you help Aug 29 11:35:38 I've forgotten to write += in the original post Aug 29 11:35:56 IMAGE_INSTALL_XXX += "spiutils" Aug 29 11:36:51 pagios: sorry not too familiar with opkg Aug 29 11:43:06 lukma: ref manual: "Using IMAGE_INSTALL with the += operator from the /conf/local.conf file or from within an image recipe is not recommended as it can cause ordering issues." (apparently += interferes with =? from core-image.bbclass) Aug 29 11:44:09 When I do a full build, for example bitbake my-custom-image it takes about 2 hours.. it pulls down all the source from github, and caches tons of packages. If I want to make a one lone change and rerun how can I do that? I can use nano to edit the local source in my build dir. but it seems to ignore the change when it runs? Aug 29 11:46:46 tcpdump: you just rerun. bitbake will reuse everything you built previously Aug 29 11:46:59 jku: Thanks for clarification :) Aug 29 11:49:32 tcpdump, did you check devtool I use it to modify the sources and do some testing Aug 29 11:52:28 tcpdump, http://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#ref-devtool-reference Aug 29 11:52:30 bluelightning: hm, it it normal that it'll build cmake stuff witl reldebug enabled and no with target type release? (i even set EXTRA_OECMAKE += "-DCMAKE_BUILD_TYPE=Release" ) Aug 29 11:54:52 crazy_imp: not sure, but I assume so as we do our own processing during packaging to take binaries, split the debugging symbols out and then strip them, so they have to start out with debugging symbols included Aug 29 11:56:03 kanavin: When you say re-use, it should include any changes I made in the cached copy locally? Aug 29 11:56:11 aurele: reading now - thanks. Aug 29 11:59:13 tcpdump: no. you generally are not allowed to edit anything in build directory Aug 29 12:10:47 bluelightning: that would work, but compiling with debug enabled, doesn't work well with that code, gcc goes crazy about it and takes > 10 minutes to compile stuff that builds within 30 seconds... Aug 29 12:17:06 crazy_imp: I should imagine there is a way to control that, you'd have to look into our cmake class Aug 29 12:18:54 bluelightning: meta/classes/cmake.bbclass ? Aug 29 12:19:10 crazy_imp: yes Aug 29 12:36:58 crazy_imp: the EXTRA_OECMAKE things should work: we do exactly that with webkitgtk . Maybe double check in the logs that it actually got used Aug 29 12:56:19 hello guys Aug 29 12:57:34 I have one question.. Is it possibile to build an image for multiple machines ? eg setting MACHINE = "MACHINE1" MACHIN += "MACHINE2" Aug 29 12:57:54 and then running bitbake core-image-minimal Aug 29 12:59:35 fberg AFAIK, it is not possible. The way I did it was to make a scrip that changes the machine in conf/local.conf and then calls bitbake Aug 29 13:01:10 yup Aug 29 13:01:13 thank you Aug 29 13:01:49 fberg, melonipoika: there's BBMULTICONFIG http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#platdev-building-targets-with-multiple-configurations Aug 29 13:01:49 fberg: not possible now, just do two builds. use different build dirs and you can still share sstate if they're similar enough. in the future you'll be able to sort of do that though, 2.4 has multiconfig support Aug 29 13:01:58 but never used it Aug 29 13:02:28 thank you guys Aug 29 13:02:34 rburton: what's missing in 2.3 -- I think we have it in the docs already? Aug 29 13:02:53 good question, no idea, only been back for ten minutes ;) Aug 29 13:03:03 (hi everyone) Aug 29 13:04:40 there's a actually a couple of known issues in the dev manual, probably answers my question Aug 29 13:06:13 rburton: hello, that was a long ten days :) Aug 29 13:22:05 I have some apps that require xerver in ARM64 and ARM32. Is there a way to do multiarch on Yocto? Aug 29 13:22:13 x11 Aug 29 13:22:37 tcpdump: look up multilib in the docs Aug 29 13:23:16 http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#combining-multiple-versions-library-files-into-one-image Aug 29 13:23:17 thx Aug 29 13:27:01 * fray notes he has customers that uses multilib arm64/arm32 systems. The only gotcha right now is compiling on the target can have issues as the kernel headers have not been 'merged' upstream.. Aug 29 13:27:23 but if you are simply running the software on the target (vs trying to do development) everything is workign as far as I know.. Aug 29 14:15:48 armpit: around? Aug 29 14:17:58 yes Aug 29 14:19:18 yes sent you an email Aug 29 14:21:09 armpit: that helps a lot, thanks Aug 29 14:22:43 armpit: thanks, replied Aug 29 14:23:02 armpit: I merged in pyro-next into pyro, that seemed to be building and working ok Aug 29 14:23:39 armpit: have asked a question in email, basically just wondering if I should merge some of the changes into morty? Aug 29 14:28:49 RP: replied Aug 29 14:29:33 basically - yes to merging Aug 29 14:52:57 armpit: cool, thanks Aug 29 14:53:07 np Aug 29 15:15:38 hi, i am try adding gnugp recipe into my image, so when i bitbake it with the image , it gives me this error https://pastebin.com/00r2AHyx , my yocto version is pyro-17.0.1 Aug 29 15:16:11 it exists in meta/recipes-support/gnupg/gnupg_2.1.18.bb Aug 29 15:18:12 seems like it is trying to create a link pointing to a existing file Aug 29 15:19:03 have you tried to cleansstate gnupg and automake-native Aug 29 15:20:08 you can remove the file "/home/yahia/Desktop/elinux_dev/bake_factory/poky-pyro-17.0.1/build/tmp/work/i586-poky-linux/gnupg/2.1.18-r0/recipe-sysroot-native/usr/share/aclocal-1.15/python.m4" manually Aug 29 15:20:24 yahiafarghaly Aug 29 15:22:48 ok i will remove and see the result Aug 29 15:22:52 Hi, I am just starting out with embedded linux. I am working through the quick start guide and have the build running at the moment to make sure everything works. When it is finished I wanted to try to build an image for a Orange Pi Zero (H2+). I have been reading the docs but can not work out how to go about it, any pointers? Aug 29 15:23:19 get the right bsp, set machine in local.conf, build image Aug 29 15:24:19 I'm not aware of a meta-orangepi Aug 29 15:24:31 Well I sort of figured that out, but where do I get the bsp from? Aug 29 15:24:36 but you might be able to find a similar bsp for the SoC Aug 29 15:25:14 CTtpollard: I could not find one for it.. but I figured that is because I didn't know what I was really doing Aug 29 15:26:02 Willy--: https://layers.openembedded.org/layerindex/branch/master/machines/?q=orange Aug 29 15:26:21 [igor] : still give the error Aug 29 15:26:37 nice Aug 29 15:26:50 kergoth: thanks very muchly Aug 29 15:26:53 np Aug 29 15:27:27 kergoth: that is the conf file, don't I need the bsp? Aug 29 15:27:36 I don't understand the question. Aug 29 15:27:41 what exactly do you think a yocto bsp *is*? Aug 29 15:27:46 it's a layer with a machine configuration and associated recipes Aug 29 15:27:53 which is exactly what that url directs you to, no? Aug 29 15:28:44 so the bsp IS just the machine details that you put into the local.conf file? Aug 29 15:29:11 that's not what i said, no Aug 29 15:29:39 all you put in local.conf is MACHINE. that tells bitbake to load that machine .conf, which is responsible for picking the correct kernel, bootloader, etc Aug 29 15:29:51 so in a way, yes, and in a way absolutely not. local.conf isn't where the knowledge lives, that's just where you configure your build. Aug 29 15:30:34 any help ? Aug 29 15:34:05 ok, that makes sense. Aug 29 15:34:33 that said, the term 'bsp' is pretty overloaded, it completely depends on context Aug 29 15:34:57 i.e. some folks ship a bsp *based on yocto*, which is an installer or tarball that includes bitbake, oe-core, all associated layers, the bsp / machine layer, binary images, etc Aug 29 15:35:49 but if you want to build from source, not using a bsp sdk from a chip manufacturer, using the open source components, then all you need is the bsp/machine layer, its dependencies, and bitbake Aug 29 15:37:46 I am trying to build a recipe that has a DEPENDS on python3-flask. In the original recipe it had a python3-Flask (capital F) and was thus failing Aug 29 15:38:32 Ive edited the recipe locally and run bitbake -c clean python-flask-httpauth Aug 29 15:38:42 but it still seems to be trying to find the capital F version. Aug 29 15:38:53 Is there a way to make it re-read my local copy? Aug 29 15:39:36 it does automatically Aug 29 15:39:49 i'd just wipe tmp to be safe, you odn't really lose anything anyway Aug 29 15:45:19 kergoth: what is the format for the MACHINE variable for the orange-pi-zero, is it the .conf file name so 'MACHINE ?= "orange-pi-zero"' Aug 29 15:45:48 yes. the value of machine is used to load conf/machine/${MACHINE}.conf Aug 29 15:45:52 still fighting with my recipe using cmake, put -DCMAKE_BUILD_TYPE=Release into EXTRA_OECMAKE like jku suggested - it ends up in the parameters for cmake as expected, but the build is still with debug enabled and only -O2 Aug 29 15:46:03 TY Aug 29 15:58:49 Hmm, I'll have to dig into how progress is implemented, to see if i can wrap multiple git progress classes under a multi stage progress handler for the new gitsm fetcher Aug 29 16:30:56 RP: thoughts on integrating the submodule support directly into the git fetcher for 2.5, potentially enabled by default? i can't imagine many repositories have submodules but don't require them to build.. can add a url parameter to explicitly disable Aug 29 16:31:40 kergoth: you're probably right about that. If it supports all the fetcher modes correctly I'd be ok with that Aug 29 16:31:51 kergoth: sorry, not had a chance to look at the code or do it justice Aug 29 16:32:01 no worries. i know you have a lot on your plate, as always :) Aug 29 16:32:02 well, I did look but not enough to do it justice Aug 29 16:32:27 i'll pursue with the integration route, will probably make it easier to fix the progress reporting anyway Aug 29 16:38:25 isnt there a bitbake recipe search somewhere? Aug 29 16:39:25 layers.openembedded.org Aug 29 16:41:08 kergoth: I only ever made them separate as there was different functionality guarantees Aug 29 17:24:23 if someone can help me with that https://stackoverflow.com/questions/45944947/apt-key-line-352-comm-command-not-found Aug 29 17:37:55 So I have a recipe that installs some scripts to a directory. I updated those script files in the recipe and issued -c clean for both the recipe and the image, but my changes are not coming through to the resultant rootfs. Aug 29 17:38:00 Any questions or comments appreciated. Aug 29 17:43:56 nevermind, bitbake overwrote the expected compilerflags set in cmake... Aug 29 17:46:22 majuk: clean won't do much if it pulls the prebuilt binaries from sstate Aug 29 17:46:25 majuk: use cleansstate instead Aug 29 17:46:35 kergoth: Ah! Thanks. Aug 29 17:46:46 majuk: that said, you shouldn't have to do that at all if bitbake is trcking changes to the files, so it sounds like you're manually pulling them externally instead of usign SRC_URI, you should fix that Aug 29 17:46:56 it picks up changes to sources automatically Aug 29 17:47:57 kergoth: Yea, I tried that to begin with, but it generated issues that I didn't understand and don't have time to address at the moment. That's good to know though. Aug 29 17:48:24 Piece by piece I am getting a better hold on this system. Aug 29 17:48:24 there are ways to explicitly track external files without using SRC_URI, but not worth the hassle unless absolutely necessary Aug 29 17:48:38 i.e. we do that for WKS_FILE, since the user can point that to any path on disk, but it's not in SRC_URI Aug 29 18:11:58 I do source setup-environment and get ERROR: do not use the BSP as root. Exiting... Aug 29 18:12:06 is there a way to allow it to work as root? Aug 29 18:15:53 Neverind Aug 29 18:16:04 commented out the root check from the config script. Aug 29 18:18:45 ugh Aug 29 18:18:48 other stuff sucks now.. Aug 29 18:44:10 cd .. Aug 29 18:44:28 meow Aug 29 19:03:38 tcpdump: really best not to do builds as root. Aug 29 19:03:43 there's a reason the check is there Aug 29 19:03:50 Ya, I made a user. Aug 29 19:04:09 But having said that, its a dedicated build box, so not too bad. Aug 29 19:04:43 still bad Aug 29 19:05:11 its also about containing arbitrary build systems Aug 29 19:10:24 you're one screwed up makefile away from a rm -rf / due to an undefined variable :P Aug 29 19:48:09 we should really do something to reduce the verbosity of runfetchcmd commands, specifically the display of the export commands. imo we should use the process environment to inject htose, and just add a logging message about what that environment is, at Fetch() construction time, or something, so we see it once, not 842 times in the log Aug 29 20:18:09 Sorry for all the dumb questions guys and gals - Im new to this :( Aug 29 20:18:34 I need to install some python things that dont exist... as such, Ive made the following recipe: Aug 29 20:18:38 https://www.irccloud.com/pastebin/qdzDiH2G/ Aug 29 20:18:44 1. Does that look right? Aug 29 20:19:28 The file is recipes-devtools/python/python-flask-cors_3.0.3.bb Aug 29 20:19:39 would I then do bitbake python-flask-cors? Aug 29 20:20:17 that will build it.. but will not include it in any generated images.. Aug 29 20:20:51 either in your customer image you need to tell the system you need a package that recipe produces -- or you can do it globally to all images you generate.. Aug 29 20:21:03 the global way isn't the recommended way, but it's quick and easy if you are experimenting.. Aug 29 20:21:13 edit your conf/local.conf and add something like: Aug 29 20:21:32 INSTALL_PACKAGES_append = " python-flask-cors" (note the leading space) Aug 29 20:22:30 the name you put into INSTALL_PACKAGES is a -package- name, not a recipe name. It needs to match up to one of the 'PACKAGES' that is produced by the python-flask-cors recipe. You can use bitbake -e python-flash-cors to figure out what names it may be producing.. Aug 29 20:26:26 Thanks fray ! Man, Im swimming here.. Aug 29 20:27:34 fray: If I have all my source code local, can I somehow point bitbake to my .bb recipe files to test them before I do an include in my global image? Aug 29 20:27:52 Im just wanting to make sure that my recipe actually works. considering this is my first recipe ever... Aug 29 20:28:17 bitbake will build it.. Aug 29 20:28:35 you would want to use one of the tools to define a local layer (if you havn't already) and include that in your project.. Aug 29 20:28:47 hmmm Aug 29 20:29:09 Did that recipe look OK, at a rudimentary level? Aug 29 20:29:15 I did not look Aug 29 20:29:46 ok, it looks reasonable to me. Aug 29 20:29:56 OK - thanks... sorry to be a bother. Aug 29 20:30:02 Im about to just go for broke... Aug 29 20:30:05 is pypi in oe-core? if not -- then you'll need to make sure you have meta-python for it Aug 29 20:30:30 (pretty sure flash and six are in meta-python, so you probably already have that configured in) Aug 29 20:30:40 not sure.. I just found an example recipe and tried to make it match. Aug 29 20:30:45 How can I check? Aug 29 20:30:51 cat conf/bblayers.conf Aug 29 20:31:27 something like this: Aug 29 20:31:28 BBLAYERS ?= " \ Aug 29 20:31:28 ##OEROOT##/layers/oe-core/meta \ Aug 29 20:31:34 ##OEROOT##/layers/meta-openembedded/meta-python \ Aug 29 20:31:34 " Aug 29 20:31:49 (replace ##OEROOT##/layers with your local path) Aug 29 20:31:57 ${OEROOT}/layers/meta-openembedded/meta-python \ Aug 29 20:32:03 thats in there Aug 29 20:32:05 so looks good. Aug 29 20:32:08 yes Aug 29 20:34:08 Last question - I put all my recipes in meta-qcom/recipes-devtools/python. Aug 29 20:34:12 That should be OK, right? Aug 29 20:34:27 Just made a new dir named python Aug 29 20:34:56 yes.. meta-qcom will need to be in that same conf/bblayers.conf -- and the layer needs a conf/layers.conf (assuming you have all that -- and its configured in a typical way) then it should work Aug 29 20:36:59 yea meta-qcom is in the conf/bblayers.conf Aug 29 20:41:21 https://www.irccloud.com/pastebin/aZYCdRpR/ Aug 29 20:41:33 That went terribly wrong fray lol Aug 29 20:42:01 https://www.irccloud.com/pastebin/00nOwpjr/ Aug 29 20:42:13 Theres my recipe for python-flask-httpauth Aug 29 20:43:03 value our of range means something told the build system either a null (non integer) value or ??? Aug 29 20:43:27 Ahh.. your recipe is missing the SRC_URI = ... Aug 29 20:43:38 (or is pypi setting that?) Aug 29 20:43:57 I think pypi is setting that, from what I understand.. Aug 29 20:44:17 ok.. then I don't know you may have to ask on #oe.. be sure to mention if you are using master or a specific branch of oe-core Aug 29 20:44:58 fray: can you add a repo manager such as dnf and just installed this stuff for arm64? Aug 29 20:44:59 Do you know? Aug 29 20:45:20 you can add DNF, (if you are on 2.3 or master -- it's 'smartpm' prior to that) Aug 29 20:45:22 *install Aug 29 20:45:44 but you still need to build the packages somehow.. the DNF and such may not install Fedora/CentOS packages.. due to difference in dependencies and compilation.. Aug 29 20:45:46 Do you know of any good ARM64 repos? Aug 29 20:45:50 Is there one for yocto? Aug 29 20:45:59 makes sense. Aug 29 20:46:07 The yocto ones I know of are usually specific to an OSV for 'project'.. Aug 29 20:46:12 OSV -or-, not for.. Aug 29 20:46:35 I see Aug 29 20:47:51 As soon as I figured this out Im going to beat my dev that wrote all his crap in python with these frameworks... for an embedded platform. Aug 29 20:48:55 :D Aug 29 20:48:57 ya, python, running on an embedded system, is powerful -- but does consume additional resources ... Aug 29 20:49:17 not necessarily bad, but you can bet you won't have things included in build systems Aug 29 20:49:30 (depends on requirements and hardware) Aug 29 20:49:34 yes Aug 29 20:49:35 If it was just python that would be great... but he wrote it all with these elaborate frameworks and modules.. Aug 29 20:49:52 Yea, if it was going on a server would be perfect. Aug 29 20:51:02 https://www.irccloud.com/pastebin/VHudarO2/ Aug 29 20:51:14 I need to get all these packages integrated... Aug 29 20:51:41 RP: Hmm. If we have multiple specified names/branches/revs, and the repo uses submodules on each, but different ones, should we fetch and unpack only the ones for the rev we're checking out? we can't really fetch and unpack all the ones we might need, because their repos under .git/modules/ for unpack might conflict (i.e. same name, different upstream repo and url). Thinking we just fetch and unpack the bits we're actually going to be unpacking and Aug 29 20:51:42 using, but add methods to the fetcher that a recipe could use to set up the submodules for a non-default branch/rev wherever they want it Aug 29 20:51:46 * kergoth ponders Aug 29 21:12:35 Im trying to install or move an entire directory. Aug 29 21:12:39 install -m 755 ${WORKDIR}/git/client.conf ${D}/etc/openvpn Aug 29 21:12:47 Thats one that works now. Aug 29 21:13:01 But I want to do that with an entire directory tree, rather than a single file... Aug 29 21:13:04 What would that look like? Aug 29 21:21:46 tcpdump: just use cp. you'll have to correct the ownership afterward, though, to make sure it ends up owned by root rather than the build user Aug 29 21:54:00 kergoth: only fetch/unpack what you're using seems sane Aug 29 22:01:41 Hello everyone. I have a question regarding patching. Aug 29 22:02:42 If I have a patch file located in layer_a and a bb file in layer_b, how can i include the patch file in my SRC_URI? Aug 29 22:03:49 I tried SRC_URI = "file:// layer_b/file/some-patch.patch" and it does not work Aug 29 22:06:33 Need a .bbappend file Aug 29 22:07:00 bhargav: ^ Aug 29 22:08:56 Well I was going to pastebin you my dts patch layer, but pastebin is throwing a fit. Aug 29 22:10:25 So I add the patch path in FILESEXTRAPATH in the bbappend file? Aug 29 22:10:36 bhargav: https://pastebin.com/QBMCp6j8 Aug 29 22:11:10 So those two patches live in the directory specified by FILESEXTRAPATHS_prepend Aug 29 22:14:34 majuk: can I add FILESEXTRAPATH in my bb file itself? Aug 29 22:15:13 I am writing a new version of an existing recipe in a different layer and I want my new recipe to pull hte old patches Aug 29 22:15:35 Ah. Yea, I dunno, haven't come close to handling that situation, just patching existing recipes. Aug 29 22:25:10 * Crofton|work washes himself with bleach Aug 29 22:27:21 lol Aug 29 22:27:39 Don't forget to drink some, get that bleach cleanse working for you. Aug 29 22:45:06 how can i submit patch in yocto in proper way ? (especially in pyro version) Aug 29 22:58:15 bhargav: its better that you make a copy of patches in your own layer if you are overwriting a recipe completely Aug 29 22:58:55 otherwise you are at the behest of original layer and patches in there might change and break your recipe Aug 29 23:00:07 matrix808 Aug 29 23:04:06 I am trying to build a new version of a package and applying my own patches on top in my own layer through an `openssl_%.bbappend` file. However, during configure, the configure script is unable to find a file from the original recipe because BBPATH points the the .bbappend file's directory. Is there a way to add a .bbappend to my own layer and have it use the files from the original recipe as well? Aug 29 23:07:34 mtahmed: it'll alwyas be able to find the original files when appending, unless you've incorrectly modified FILESPATH/FILESEXTRAPATH Aug 29 23:07:50 and bbpath is irrelevent in this case, file:// files are found via FILESPATY Aug 29 23:08:00 BBPATH is used to find classes, config files, and .inc files Aug 29 23:13:21 khem: Thanks! I ended up doing the same since I could not figure out a way for my problem Aug 29 23:13:44 What you said actually makes sense Aug 29 23:15:16 kergoth: Oh that makes sense. I think I might have been missing a the file that it required because I was redefining SRC_URI instead of appending to SRC_URI. Aug 29 23:16:17 I still have the problem of FILEEXTRAPATHS_prepend not able to use the ${PV} (presumably because it isn't defined when it's expanded; it just gets expanded to the string "%") but it is able to get ${BPN} just fine. Aug 29 23:20:08 PV should work fine in that context, but worst case you could arrange to immediately expand THISDIR, i.e. thorugh an intermediate variable, but leave PV alone Aug 30 01:15:23 ls -la Aug 30 01:15:43 * tcpdump realizes this is NOT the terminal. Aug 30 01:16:37 kergoth: in re: "kergoth> tcpdump: just use cp. you'll have to correct the ownership afterward, though, to make sure it ends up owned by root rather than the build user" Aug 30 01:16:50 Can you literally use linux commands in your recipe? Aug 30 01:17:00 functions in bitbake files are just shell scripts Aug 30 01:17:11 unless they're explicitly marked as python, anyway Aug 30 01:17:15 so yes, it's just a shell script Aug 30 01:17:32 the 'install' command you were already using is a stock command on linux systems, not bitbake specific Aug 30 01:17:33 wow, thats good to know.. Aug 30 01:17:52 thats insane... Aug 30 01:18:16 so, could I do things like "systemctl enable nginx" rather than making and symlinking unit files... Aug 30 01:18:22 * tcpdump is going to cry... Aug 30 01:18:39 which is why you shouldn't be building anything, as root, ever Aug 30 01:18:50 s/anything,/anything/ Aug 30 01:18:56 yea... Aug 30 01:19:00 wow, it all makes sense. Aug 30 01:19:05 Well, not all, but more. :) Aug 30 01:19:17 also, you don't have to symlink and make unit files anyway, that's what the systemd bbclass is for, it handles all of that correctly for you Aug 30 01:21:37 Man, wish I had this chat earlier... Aug 30 01:21:47 So I can go ahead just cp the dir where it goes. Aug 30 01:21:52 Thats hugely helpful. Aug 30 01:25:04 yep. jsut chown -R root.root ${D}/whatever afterward, since cp retains ownership, and the files are owned by you right now Aug 30 01:25:10 install will correct it automatically Aug 30 01:25:32 let me grab my recipe real quick. Aug 30 01:28:39 tcpdump: http://www.aosabook.org/en/yocto.html and https://www.yoctoproject.org/docs/2.3/bitbake-user-manual/bitbake-user-manual.html may be of interest. the latter covers the file format Aug 30 01:30:35 got it - thanks! Aug 30 01:30:41 np Aug 30 01:31:11 both of those are a lot shorter and potentially less intimidating than the rest of the yocto project docs. not tutorial / howto, though, more background and reference material Aug 30 01:31:15 shouldn't take long to peruse, though Aug 30 01:34:21 install -m 755 ${WORKDIR}/git/api ${D}/home/root/api Aug 30 01:34:49 could that work for moving a directory thats commited to an included git repo to the /home/root/api dir? Aug 30 01:34:57 Would tha tbe recursive? Aug 30 01:35:59 or would I be better to do cp ${WORKDIR}/git/api /home/root ? Aug 30 01:36:04 ugh Aug 30 01:36:27 cp ${WORKDIR}/git/api ${D}/home/root Aug 30 01:40:19 cp isn't recursive unless you cp -a Aug 30 01:40:43 worth noting that there's a variable for root's home directory, you should use that instead of hardcoding /home/root Aug 30 01:41:42 * tcpdump has so much to learn. Aug 30 01:43:01 Im also installing files in /usr/bin Aug 30 01:43:06 looks like I can use ${bindir} Aug 30 01:45:36 if you're curious about most of the common available variables, read meta/conf/bitbake.conf Aug 30 01:45:58 got it! Im quite curious.. Aug 30 01:46:04 I was googling that as we speak. Aug 30 01:46:05 Thx Aug 30 01:54:17 kergoth: sorry for the dumb question.. where exactly is meta/conf/bitbake? Aug 30 01:54:26 oe-core or poky, depending on your setup Aug 30 01:54:33 see also the 'find' command Aug 30 01:55:04 ah, my ba Aug 30 01:55:06 bad Aug 30 01:55:15 find . -name "meta" Aug 30 01:55:41 so it would e in recipes probably Aug 30 01:57:00 i dont' understand what you're referring to Aug 30 01:57:14 i just said it's in the conf folder of meta, not any of the recipes dirs Aug 30 02:02:32 yea, its in the recipes dir.. interesting. Aug 30 02:02:37 Found it Aug 30 02:03:35 ${D}/home/root looks like it could be replaced with ${D}$ROOT_HOME Aug 30 02:03:47 pretty cool Aug 30 02:05:06 use bitbake variable reference syntax, not shell Aug 30 02:05:11 that is, ${ROOT_HOME}, not $ROOT_HOME Aug 30 02:05:27 the latter will only work for variables marked as 'export' Aug 30 02:18:57 ${D}${ROOT_HOME} Aug 30 02:18:58 got it Aug 30 02:19:01 thx Aug 30 02:19:53 np Aug 30 02:35:00 kergoth: so adding spaces/tabs in your install directive is OK? I have an OCD about disorganized code. :D Aug 30 02:38:33 like i said, just a shell script Aug 30 02:38:50 the convention is to use hard tabs in shell and spaces in python, but honestly i hate that and use 4 spaces everywhere in my own layers.. Aug 30 02:45:49 kergoth: well, its all refactored... Aug 30 02:45:54 So to see if it all works now. :) Aug 30 02:59:36 kergoth: mind if I bug you with one more question (besides this one) :) **** ENDING LOGGING AT Wed Aug 30 03:00:00 2017