**** BEGIN LOGGING AT Wed Mar 04 02:59:57 2020 Mar 04 08:09:03 i actually ran into the guy behind https://github.com/EugenWiens/vscode-bitbake yesterday :) Mar 04 08:19:14 New news from stackoverflow: Why is yocto build for intel core i7 crashing? Mar 04 08:21:49 Hello guys, someone here has good tips and/or reference to understand how to debug custom tasks/classes ? I'm not able to see where bb.note should be printed :S Thank you Mar 04 08:49:21 New news from stackoverflow: Yocto image is not adding the files to rootfs Mar 04 08:58:25 Hello! Im running zeus branch with meta-intel, does anyone know why CONFIG_DEVMEM=y doesnt propagate to the kernel config on target? It still says "CONFIG_DEVMEM is not set" in config.gz on target. I need it to get /dev/mem, a watchdog needs it Mar 04 09:06:25 hello im using https://github.com/Freescale/meta-freescale and it depends on "git://git.openembedded.org/openembedded-core" does it means I have to replace poky/meta for openembedded-core/meta in bblayers.conf ? Mar 04 09:11:43 Hi everyone. I'm trying to build a receipe that requires python3. I tried `inherit python3native` and `RDEPENDS_${PN} = " python3 python3-numpy"` but bitbake complains that: `aise RuntimeError("Python version >= 3.5 required.")` Mar 04 09:12:10 Why is python3 not being used here? Any hints? Mar 04 09:21:57 Anyone? Mar 04 09:22:03 mauz555: No, you can keep poky/meta. It's the same content as in oe-core/meta, Mar 04 09:38:52 RP: ping - who's the mastermind behind tinfoil? Mar 04 09:59:53 Sorry, I got disconnected. Has anyone the email history here: https://yocto.yoctoproject.narkive.com/9fJuNa0I/scipy-recipe I'm looking for the attachment by matthias Schoepfer Mar 04 10:12:02 LetoThe2nd: RP or bluelightning Mar 04 10:14:20 PinkSnake: bitbake -D (multiple Ds, I never remember how many) should give you the notes. Otherwise, I think they might make it to the log.do_ in WORKDIR/temp? The first works for sure, the second I don't know :) Mar 04 10:16:20 rburton: i ran into the guy behind https://github.com/EugenWiens/vscode-bitbake here, and at the moment he's parsing bitbake output Mar 04 10:16:38 woo Mar 04 10:17:24 rburton: gave him a very *gentle* nudge to look at tinfoil, hence the question Mar 04 10:17:57 it would be quite neat if vscode-bitbake was a language server that could start up a memory resident bitbake Mar 04 10:18:08 would be able to fire builds off then too Mar 04 10:20:56 rburton: he's still around so can certainly relay bits and pieces. so that would be starting the memory resident, and getting stuff in and out via tf? Mar 04 10:28:06 qschulz Thx double D make the job Thank you :) Mar 04 10:31:31 Hi there, Mar 04 10:32:42 Does anyone can help me in installing a simple file into rootfs (like a motd), that is already provided by base-files? How to choose like a "prevered install provider for file X"? Mar 04 10:33:24 watchd0g: easiest is to write a bbappend for base-files and provide your own file there Mar 04 10:40:54 hum ok Mar 04 10:41:25 i'll try Mar 04 10:57:03 hi, im trying to build a project that has: unix: PKGCONFIG += mm-glib if i put #include in the header file it cant finde libmm-glib.h. if i only put #include in the cpp file it builds Mar 04 11:02:19 Someone here is able to tell me which variable can I use to get (inside tmp/work ) .cpio rootf path ? Mar 04 11:05:19 PinkSnake: bitbake -e and look for a variable with the path you want. Note that it might not always be a good idea to try to get files from some given paths (i'm thinking searching in other recipe's sysroot, or looking into the image workdir too early in the building stage) Mar 04 11:10:23 Hey, I'm having a problem with devshell - when I try to run, for example, bitbake -c devshell bash, terminal window pops up, and immediately closes. Any hints why is that? Mar 04 11:19:53 New news from stackoverflow: Any one hit same compiling failed issue for Yocto project? Mar 04 11:28:22 hmw1: sounds like a problem with your build tool and not yocto Mar 04 11:32:43 qschulz Thx a lot I never think about -e ^^ Mar 04 11:35:27 rburton: yes, but im using a "yocto" sdk or do you think its the ide Mar 04 11:35:55 qschulz In fact I'm writing a bbclass to create encrypted FIT image and i need the .cpio.gz path :) Mar 04 11:36:10 hmw1: i think your code is wrong, be it the C or the makefiles Mar 04 11:37:01 kpo maybe try ** bitbake -c devshell **only Mar 04 11:37:35 kpo: what terminal? Mar 04 11:37:42 and what yocto release Mar 04 11:38:11 modern terminal apps like to be client/server and just tell another instance to pop up a window which goes against how devshell wants to work Mar 04 11:39:01 rburton: I'm using a .pro file that PKGCONFIG adds the libs ( if i add the include in the cpp file everything works) if i add it to the header it fails to finde the lib Mar 04 11:39:12 so put it in the cpp file? :) Mar 04 11:39:26 is the header file actually being included in your source? Mar 04 11:39:42 again sounds like your source or .pro is broken Mar 04 11:40:28 rburton : the header is include as header file Mar 04 11:41:59 rburton: HEADERS += modem.h SOURCES += modem.cpp Mar 04 11:42:21 did you actually #include modem.h in modem.cpp? Mar 04 11:42:32 yes Mar 04 11:43:01 so assuming you've done what you said you've done, there's literally no difference to the preprocessor Mar 04 11:43:25 which suggests that your code is broken somewhere Mar 04 11:49:45 rburton: btw if i do a #include in the header it "works" but then it brakes in the internal sdk lib modemmanager Mar 04 11:53:55 rburton: I've tried different ones, mainly I use konsole, tested even on xterm. Yocto thud Mar 04 11:59:07 hmm, I changed some time ago default interpreter from bash to fish, anyway I always first change to bash, then run source yocto-thud/oe-init-build-env ./build and then bitbake Mar 04 11:59:30 could it be because of the default shell change? Mar 04 11:59:42 I'll try to revert to fish Mar 04 12:03:14 Yep, it was because of the default shell - when I changed back to bash it started working :) Mar 04 12:05:47 rburton: ping sorry just so i understand it right: the best way to do the plugin would starting a memres bitbake, and then getting stuff in and out how? is tinfoil really the way to start? Mar 04 12:06:03 LetoThe2nd: dunno :) Mar 04 12:06:12 tinfoil is a nicer api than the old api though Mar 04 12:08:27 rburton: which is the old api? Mar 04 12:08:52 rburton: i just need some pointer to hand on, and i know little more than that the buzzwords exist. Mar 04 12:11:44 same entrypoint but tinfoil is better Mar 04 12:12:06 there's a load of scripts in oe-core/scripts or even kergoth's bb that hit tinfoil to dig into the data store Mar 04 12:12:44 rburton: ah! is bb somewhere in mainline repo? Mar 04 12:13:27 no, its on github Mar 04 12:13:46 rburton: got it, thanks. Mar 04 13:17:00 Hi, my u-boot does not have the tftp command. I remember there's a config switch for the yocto build to enable that feature. Do you remember where to switch that tftp feature for u-boot on? Mar 04 13:24:25 I'm on the u-boot=> prompt. I type a ? and get a list of all commands. But I can see tftp is not offered there. Mar 04 13:26:32 CONFIG_CMD_TFTPBOOT needs to be enabled in u-boot config Mar 04 13:35:44 Is it possible to use addtasks with a python function ? Mar 04 13:37:02 AFAIK with a python task yes. It's really one line, so try it and you'll know for sure Mar 04 13:38:03 qschulz I have spent 3 days on "trying" ^^ Mar 04 13:38:44 where's your python function (which file, where on the FS), what did you try, what's the expected behavior, what's happening etc. Mar 04 13:39:10 And more specifically, what are you trying to achieve (the actual root of your problem that you're trying to fix with a python function) Mar 04 13:39:20 (also python function != python task IIUC) Mar 04 13:40:54 I'm writing a class to create encrypted FIT image, my .bb image recipe inherit of the custom class. I'm not able to call python TASK from do_blabla_appen() Mar 04 13:42:46 ${@pythontask()} in a bash task should work Mar 04 13:49:30 qschulz ${@python_code} works but ${@pythontask} not :( Mar 04 13:50:25 PinkSnake: we're using it, so I'm pretty sure this works :) Mar 04 13:50:55 qschulz feel free to share ;) Mar 04 13:51:00 So the problem is somewhere else I guess. where's your python function (which file, where on the FS), what did you try, what's the expected behavior, what's happening etc. Mar 04 13:52:53 qschulz As I have already said before, i want to create an encrypted FIT image. So I have created bbclass, my image recipe inherit from this class, I'm not able to call a python function form do_unpakc_append() function Mar 04 13:53:46 foo.bbclass: def my_python_function(d): blabla; myrecipe.bb: inherit foo; do_install_append() { ${@my_python_function(d)} } Mar 04 13:54:20 thanks, joergho Mar 04 13:59:04 qschulz Exception during build_dependencies for do_write_manifest_file :( doesn't matter, Thx for support Mar 04 14:03:46 PinkSnake: maybe someone else will be able to help you because I don't think we're able to understand each other and communicate correctly. Please try again later, usually people are responsive, I'll not hijack the question so you've your chances on having someone else answer your question. Mar 04 14:04:21 qschulz: Tkt pas pour moi va ;) merci Mar 04 14:06:40 JaMa, I ran the test with meta-qt5 25d0fa605 and everything looks fine (simple gui application generated by gnuradio-companion works) Mar 04 14:07:11 However, to fix the fonts problem, I needed to create a symlink /usr/lib/fonts -> /usr/share/fonts/ttf Mar 04 14:07:33 Could we add a recipe / packet to meta-qt5 which provides this link? Mar 04 14:18:35 hi - how to PATCH https://github.com/Freescale/meta-freescale/blob/zeus/recipes-graphics/drm/libdrm_2.4.91.imx.bb so Package "dri" can be found through pkg-config ? Mar 04 14:43:12 mauz555: I'm not sure that recipe provides dri in the first place Mar 04 14:44:11 JPEW: IMX_LIBDRM_SRC ?= "git://source.codeaurora.org/external/imx/libdrm-imx.git;protocol=https;nobranch=1" Mar 04 14:44:24 JPEW: IMX_LIBDRM_BRANCH = "libdrm-imx-2.4.91" Mar 04 14:45:05 JPREW: I guess the recipe fetches/install this git repo/branch : https://source.codeaurora.org/external/imx/libdrm-imx/tree/?h=libdrm-imx-2.4.91 Mar 04 14:45:44 Correct... and that repo doesn't provide "dri"... did you mean "drm" perhaps? Mar 04 14:46:38 JPEW: here is the original bug im confronting: https://pastebin.com/rPMqQk79 Mar 04 14:47:21 JPEW: "Package dri was not found in the pkg-config search path." Mar 04 14:48:58 JPEW : at the top of this recipe https://github.com/Freescale/meta-freescale/blob/zeus/recipes-graphics/drm/libdrm_2.4.91.imx.bb it says "Direct Rendering Infrastructure\" (DRI). DRI is required for many hardware Mar 04 14:55:40 mauz555: Sure, but I don't think libdrm provides dri.pc: https://pastebin.com/DxsQGEqb Mar 04 14:56:03 I'm not sure what would provide that Mar 04 14:56:30 qschulz after 3 days all my stuff works :) Thx for your patience ;) Mar 04 15:00:31 PinkSnake: could you share what it was so people can benefit from those 3 days :) ? Mar 04 15:00:37 Happy it works now Mar 04 15:01:17 qschulz I have juste made some changes from this bbclass https://git.phytec.de/meta-phytec/tree/classes/fitimage.bbclass?h=zeus Mar 04 15:05:34 JPEW: I have found a dri.pc there : https://packages.debian.org/search?arch=amd64&mode=exactfilename&searchon=contents&keywords=dri.pc Mar 04 15:06:34 mauz555: Right looks like that's provided by mesa, which might not be present if you are using the freescale OpenGL drivers instead Mar 04 15:08:32 JPEW: my goal is to run chromium on wayland for my platform. should I post an Issue to OSSystems/meta-browser or tweak the chromium /recipebuild so it uses freescale OpenGL ? Mar 04 15:09:01 mauz555: I'm not sure, sorry, I don't know much about chromium Mar 04 15:09:10 JPEW: ty Mar 04 15:20:46 New news from stackoverflow: Missing files in Yocto SDK Mar 04 15:25:29 joergho: is that really an the issue in pyqt5 or in gnuradio? Mar 04 15:33:58 Hi all, is possible to add systemd scripts within a python3 recipe for custom apps? Mar 04 15:36:57 RP: JPEW: Increasing my paranoia level to try and figure out why I can't get reproducible tests running correctly. Trying with a completely-new set of repos and build environment Mar 04 15:37:12 s/environment/folder Mar 04 15:41:29 tgamblin: Ok... it is *really* weird Mar 04 15:41:57 tgamblin: One option might be to hack up the reproducible build test to just run bitbake -e and see how the variables are being set Mar 04 15:42:19 JPEW: yeah, I am puzzled. That's not a bad idea Mar 04 15:42:45 tgamblin: I wonder if there is some variable override being applied... Mar 04 15:43:21 argh this is driving me nuts Mar 04 15:44:08 libc-package.bbclass arranges to have glibc-binaries, glibc-gconvs, glibc-localedatas, and glibc-locales rrecommend on the individual packages, but it's not actually doing it. the resulting packages don't have the recommends Mar 04 15:44:44 JPEW: That would make sense, but I don't see anything obvious in my env variables. You'd think that this clean slate build should come back with the expected behavior in that case, but I guess we'll see once the test finishes... Mar 04 15:45:22 If I bb.warn() in populate_packages_append, the values are as expected, but the resulting ipk isn't Mar 04 16:00:24 yocto layers are a nightmare to assemble, who expected git repositories (layers) to work together solely based on the fact that they share the same branch name. Mar 04 16:12:48 JPEW: out of curiosity, can you paste the command you run to run the reproducibility tests? Mar 04 16:14:58 Apparently nxp have retired their freescale git server: is there a method to define a mirror for all git.freescale uris without having to create individual .bbappend files? I have tried defining PREMIRROR as the following but it doesn't appear to be working: "git://git.freescale.com\/.* git://github.com/NXP/" I also tried escaping the / and . Mar 04 16:14:59 characters. Mar 04 16:18:07 tgamblin: oe-selftest -r reproducible Mar 04 16:20:46 mauz555: It heavily depends on how actively supported the layers are; some are better than others Mar 04 16:21:22 mauz555: But, it can be a bit of pain to make sure you're up to date with all the layers if you have a lot of them Mar 04 16:23:34 jrdn__: I understood some of NXP repos are now on https://source.codeaurora.org/external/imx/, hope it helps Mar 04 16:25:54 Regarding the "Kent Dorfman" posting on the yocto list, I will buy the first person to address him as "Flounder" a beer. Mar 04 16:27:27 Crofton|road: As far as pseudonyms go, that's one of the better ones I've seen :) Mar 04 16:32:40 Oh I agree, and the horror story of how he gets his build info is amusing Mar 04 16:41:14 mauz555: how else would you expect it to work than by branch? when oe-core makes incompatible changes, there's only one sane way to handle that as a maintainer of another layer Mar 04 16:44:57 JPEW: Crofton|road: I find it very sad that wherever he works at has been overrun by lawyers to this degree Mar 04 16:48:42 He has a unique set of challenges. This is why people rewrite BSP's if they can Mar 04 16:50:45 I mean that he is too scared to use his real name (let alone actually contribute to yocto). Something about the industry he's in. Mar 04 16:51:06 New news from stackoverflow: Bitbake: "Nothing PROVIDES" error Mar 04 16:56:01 why would anyone write an other answer to an already answered (and accepted answer) thread almost two years later with the exact same answer but poorly rewritten? Life is a wonder sometimes Mar 04 17:43:12 * zeddii chuckles at the LTS thread Mar 04 18:31:12 Can you DEPENDS a "subpackage" of another recipe? Mar 04 18:32:14 fullstop: no Mar 04 18:32:43 Is there any way to get the package into recipe-sysroot? Mar 04 18:32:48 fullstop: DEPENDS lists recipes, not output packages Mar 04 18:35:25 fullstop: afaik, all files built by a recipe get deployed to the sysroot when it is in DEPENDS. What are you seeing that is missing? Mar 04 18:35:51 smurray: It's likely my fault somewhere here, but I'll explain what I am trying to do. Mar 04 18:36:33 I have a kernel recipe which is based on the meta-atmel kernel recipe. I need the kernel headers in order to build a package which requires >= kernel 5.5 and this is not satisfied by the libc headers. Mar 04 18:37:05 So this got me linux-at91 and linux-at91-headers Mar 04 18:37:40 This, I believe, should actually be linux-at91 and linux-at91-dev Mar 04 18:38:53 So, in short, I'm trying to get the headers from the recipe which provides virtual/kernel in my sysroot. Mar 04 18:39:23 Hi! I am having a problem with my final image - it does not contain some files I was expecting to find in it. Those files are part of FILES_${PN}-dev and ${PN}-dev is part of PACKAGES. I can see the files in place in the "image/" corresponding folder (i.e. under build/tmp/work/x86-64xxx/${PN}-${PV}" and I also find the files under /build/tmp/sysroot-components. I have alredy try including the ${PN}-dev Mar 04 18:39:29 to my image by means of ... Mar 04 18:39:32 ... IMAGE_INSTALL_append and didn't work. I also included the files in the ${PN} (i.e. no -dev") but they are still not to be found in the final image... HELP!! Mar 04 18:39:43 fullstop: just linux-at91, you can't specify -dev in DEPENDS. I suspect it still may not work, as it's dependent on do_deploy_sysroot, which is probably not defined to do what you want Mar 04 18:39:55 smurray: that gets me something to go on, thanks Mar 04 18:40:12 I can monkey around with do_deploy_sysroot Mar 04 18:40:38 fullstop: you can see some explanation in https://www.yoctoproject.org/docs/current/ref-manual/ref-manual.html#var-DEPENDS Mar 04 18:40:45 I guess I should mention that the headers are not present in the -dev package Mar 04 18:41:22 fullstop: if whatever you're trying to build needs internal kernel headers, it's probably going to take some work Mar 04 18:42:11 smurray: I assumed as much. :-( Mar 04 18:42:36 libgpiod with kernel >= 5.5 lets you specify line bias Mar 04 18:43:27 fullstop: does it need some header that's outside include/uapi for that? Mar 04 18:46:10 smurray: I'm pretty sure that it only includes gpio.h Mar 04 18:46:13 fullstop: if it's just that you effectively need linux-libc-headers to be 5.5, then that can be bumped somewhat easily. Mar 04 18:46:39 and version.h Mar 04 18:46:45 or just bbappend the recipe, do a make headers_install to a custom location, and then deploy that dir to the sysroot. Mar 04 18:47:05 changing libc-headers for just that, is overkill. Mar 04 18:47:19 true, but it's pretty easy ;) Mar 04 18:47:28 if you like full rebuilds :D Mar 04 18:47:37 and being on your own for all interfaces to the kernel for all packages. Mar 04 18:47:44 zeddii: that's where I was going initially.. but I was having difficulty getting the headers included in the -dev package Mar 04 18:47:49 but yah, easy enough to do. Mar 04 18:48:16 yeah, there is some chance of breakage if you're not already close to 5.5, indeed Mar 04 18:48:53 I'm on zeus, which I _think_ is 5.3 Mar 04 18:50:20 I do: make headers_install INSTALL_HDR_PATH="${D}/usr" in do_install_append Mar 04 18:50:33 and FILES_${PN}-dev += " /usr/include/" Mar 04 18:50:39 not a good idea to go over the existing ones. Mar 04 18:51:04 Hi! I am having a problem with my final image - it does not contain some files I was expecting to find in it. Those files are part of FILES_${PN}-dev and ${PN}-dev is part of PACKAGES. I can see the files in place in the "image/" corresponding folder (i.e. under build/tmp/work/x86-64xxx/${PN}-${PV}" and I also find the files under /build/tmp/sysroot-components. I have alredy try including the ${PN}-dev Mar 04 18:51:05 qa fails Mar 04 18:51:08 it is normally some alternate include path, that you add to the application that needs them. you can then just deploy them to the sysroot and use them that way. Mar 04 18:51:10 to my image by means of ... Mar 04 18:51:13 ... IMAGE_INSTALL_append and didn't work. I also included the files in the ${PN} (i.e. no -dev") but they are still not to be found in the final image... HELP!! (now a bit more desperate! :D) Mar 04 18:51:29 zeddii: understood Mar 04 18:51:53 nemgti-og: -dev packages are typically not in the final image Mar 04 18:51:58 fullstop: I have an old patch around here somwhere that does that. I should double check the deploy semantics, I haven’t needed to do it in ages. Mar 04 18:53:42 nemgti-og: if you need headers or things like that you will need to explicitly package them. Mar 04 18:53:48 fullstop: thanks. That expains a lot, but if I add the files to FILES_${PN} shouldn't I get the files in the final image? I have tried that already but the files are still not there. Is there anywhere to look at? Mar 04 18:55:15 nemgti-og: are the files actually installed by the recipe's do_install? i.e. do they end up in ${WORKDIR}/image? Mar 04 18:57:04 smurray: Yes. The files are present in ${WORKDIR}/image Mar 04 18:57:17 RP: I'm not sure who maintains yoctoproject.org/docs, but the Quick Build link at the top seems broken Mar 04 18:57:37 nemgti-og: okay, if you look in packages-split, do they appear in one of the package directories? Mar 04 18:59:01 smurray: Yes. They are part of the -dev package Mar 04 18:59:06 I was thinking about this the other day, would it be beneficial to add something to robots.txt to get google to stop indexing the 1.x documentation? Mar 04 19:00:26 nemgti-og: okay, adding that package to the image should be resulting in them being on target. Check the image manifest in tmp/deploy/image//.manifest and see if the package is making it in Mar 04 19:00:58 nemgti-og: what type of file is this which is not included? Mar 04 19:02:56 smurray: also, any idea why this: do_configure[depends] += "virtual/kernel:do_shared_workdir" (as referenced from the linux-libc-headers recipe) doesn't do what I expect it to? Mar 04 19:03:24 hold off on that until I look at kernel.bbclass and the yocto kernel. Mar 04 19:04:47 fullstop: what do you expect it to do? ;) That'll result in tmp/work-shared//kernel-source being populated by fetch/unpack Mar 04 19:05:12 smurray: I expected kernel headers somewhere in my sysroot. ;-) Mar 04 19:05:22 maybe not in the right place, but there Mar 04 19:07:45 fullstop: definitely look at kernel.bbclass for the shared workdir stuff, that's an optimization to explicitly prevent multiple copies of the kernel tree ;) Mar 04 19:08:41 fullstop: I think you'll just need to use different variables than you're expecting to get at the kernel tree Mar 04 19:10:19 likely. I'll let you know when I figure it out! Mar 04 19:15:38 smurray: sorry - I had to attend something at the door. So, the package is not in the manifest :/ But some of the file of the package (i.e. lib*.so*) are present in the final image. Does that make any sense? Mar 04 19:15:56 fullstop: the files I am missing are *.cmake files Mar 04 19:16:57 smurray: fullstop I also may be using the incorrect terms here. Final image do not mean the target'S image. I am actually talking about the eSDK that I am generating Mar 04 19:17:30 ah, that's different Mar 04 19:18:20 xS.... sorry for the missleading information... I think I need to sleep a bit Mar 04 19:18:36 okay, no worries Mar 04 19:33:23 I'd like to extend the kernel-devicetree package to install additional files. I hoped I could use do_install_append() in a bbappend file, but since the kernel-devicetree stuff is in a bbclass, I take it that's not possible? Mar 04 19:36:34 smurray: Do you think I need to investigate more why those packages are not part of the .manifest files? Do you have any other hint? Mar 04 19:38:28 nemgti-og: is the non-dev package part of the image that you specify when you run 'bitbake -c populate_sdk_ext '? Mar 04 19:39:01 yes Mar 04 19:39:39 xyzzy42: I did some stuff with a devicetree bbclass to install it in the output directory only to find that the bbclass did that already. Mar 04 19:39:40 nemgti-og: does anything else in the image use that -dev package to build? Mar 04 19:39:59 smurray: No Mar 04 19:40:12 xyzzy42: There's a kernel recipe that inherits kernel-devicetree, you can bbappend that recipe and add do_install_append(). Mar 04 19:40:55 But you can't bbappend a bbclass, so you'd have to do it per kernel recipe Mar 04 19:41:48 erbo, I see, that's just fine, my change is only appropriate for the kernel recipe it's intended to modify anyway Mar 04 19:42:23 smurray: A slight correction. In the image (not the eSDK), the package is there, but those specific files are also not found in it Mar 04 19:42:43 nemgti-og: the non-dev package, you mean? Mar 04 19:43:21 * LetoThe2nd has just finished the iot conf... its really crazy. they tell fancy stories about all apps and clouds and stuff, and when you ask what they *really* use to get stuff done in the end, they say "yocto" Mar 04 19:43:55 fullstop, yes, it's install the dtbs fine, but want to make a symlink for backward compatiblity with old bootloaders, as the name of the dtb has changed Mar 04 19:44:10 LetoThe2nd: :D Mar 04 19:44:19 smurray: If by non-dev package you mean the *.cmake files, yes. The rest of the fils of the -dev package (i.e. /usr/lib/*.so.*) are there. The *.cmake files I am trying to include should be in /usr/lib/cmake// Mar 04 19:44:48 erbo: i kid you not. Mar 04 19:45:23 which worked! it's not possible to append the bbclass itself, but it is possible to append to things to bbclass did in the recipe using it Mar 04 19:45:27 nemgti-og: I'm guessing you're mistaken. Only .so (note no .[0-9]*) files are in -dev packages, the versioned ones are in the non-dev package Mar 04 19:45:56 nemgti-og: so I believe the SDK population isn't just the -dev packages of everything in the image, it's more likely the -dev packages from the superset of all the DEPENDS of what's in the image. Hence, you won't see the -dev package in this case, as nothing else needs it to build. Mar 04 19:46:03 "azure sdk in a container here, and data flow from to there provisioning yaddayadda" - "and how are you building the real system that runs the containers?" - "oh, umm, thats yocto" Mar 04 19:46:51 nemgti-og: you can try appending the -dev package to the variable TOOLCHAIN_TARGET_TASKS to explicitly add it Mar 04 19:47:34 err, TOOLCHAIN_TARGET_TASK, no 'S', see https://www.yoctoproject.org/docs/3.0.1/ref-manual/ref-manual.html#var-TOOLCHAIN_TARGET_TASK Mar 04 19:48:05 LetoThe2nd: lol Mar 04 19:48:46 smurray: Ok. Let me try that. But... currently TOOLCHAIN_TARGET_TASK contains my package. Should I add *also* the package-dev? Is that what you meant? Mar 04 19:48:58 nemgti-og: yes Mar 04 19:49:49 smurray: 3 or 4 speakers, in a row. Mar 04 19:50:27 LetoThe2nd: at least Microsoft thankfully didn't create their own NIH build system for Azure Sphere Mar 04 19:50:30 smurray: Ok. I'll try that.. Just to see if I understood. Even if my package is added in TOOLCHAIN_TARGET_TASK, those files might not be added because no other package needs them to be built (they are no other package dependency)? Mar 04 19:51:39 nemgti-og: no, that's the opposite of what I said. They won't be added to the default value of that variable when you run 'bitbake -c populate_sdk_ext', hence you need to manually add them Mar 04 19:52:18 nemgti-og: in an ideal world, you never touch TOOLCHAIN_TARGET_TASK or HOST_TASK manually Mar 04 19:52:37 smurray: yeah. hum. you're just putting something in my head. Mar 04 19:54:12 So the special Microsoft Linux, or whatever they call it, is just yocto? Mar 04 19:54:25 smurray: Ok. I get it. But it sounds like adding manually packages to those variables is not the recommended practice though Mar 04 19:55:42 nemgti-og: yeah. It tends to get messier in practice, for stuff like you're seeing Mar 04 19:56:07 xyzzy42: no, but obviously a lot of folks are running thw microsoft middleware on "yocto linux" Mar 04 19:56:21 smurray: Thanks. I will have to learn how to do this the cleaner way in the future. Mar 04 19:56:25 nemgti-og: it's typical to end up with a small toolchain recipe that inherits populate_sdk or populate_sdk_ext and adds the extra stuff Mar 04 19:57:51 The press release seemed to imply microsoft had created a new linux distro and that only this distro would work with azure Mar 04 19:57:54 smurray: haaaaaa ok! That means that -in theory- I should be able to easily do this in a cleaner way. I hope I still get some time for this next week. Mar 04 19:58:55 xyzzy42: I believe it's just their own distro config, tbh Mar 04 20:00:34 I'm actually surprised they didn't create an entirely new build system from scratch by leveraging code that is microsoft specific Mar 04 20:02:09 xyzzy42: heh, exactly. They're even contributing security fixes to oe-core Mar 04 20:02:38 yeah. strange as it sounds, MS is really playing nice with us at the moment. Mar 04 20:05:29 now they just need to sign up as a Platinum member ;) Mar 04 20:06:02 i think we all agree that this would be a very good thing "TM" Mar 04 20:06:15 +1 Mar 04 20:06:26 * RP looks at the computer and wonders whether he's missing emails Mar 04 20:15:07 xyzzy42: you are asking for yocto they dont even mention Linux here https://www.microsoft.com/security/blog/2020/02/24/azure-sphere-microsoft-answer-iot-threats-reaches-general-availability/ Mar 04 20:15:20 but read this https://azure.microsoft.com/en-us/blog/azure-sphere-s-customized-linux-based-os/ Mar 04 20:15:59 unless it was completely thrown out of window, it is based on "Yocto Linux" Mar 04 20:16:03 :P Mar 04 20:19:17 khem: :) Mar 04 20:26:18 the talk on the azure sphere yocto usage from ELC for the curious: https://www.youtube.com/watch?v=-T7Et5qfqQQ&list=PLbzoR-pLrL6ol7Cf1g_4rsCda23OiLh8d&index=83&t=0s Mar 04 20:26:27 good ad for MULTICONFIG ;) Mar 04 20:28:23 smurray: trying to promote :) Mar 04 20:29:36 JPEW: RP: I can get the reproducibility test to work properly if I use "oe-selftest -r reproducible -j 1" instead of just "oe-selftest -r reproducible" Mar 04 20:30:54 tgamblin: then its my recent code changes :( Mar 04 20:33:36 RP: :( surprised no one else is running independent reproducible on master to catch it too. At least I can avoid tearing apart my setup though Mar 04 20:34:29 RP: which change(s) in particular are you referring to? Mar 04 20:35:03 https://git.openembedded.org/openembedded-core/commit/?id=50238b0717b04e0a1fa69d618e8c8aa8445a80b0 ? Mar 04 20:38:06 Huh, I wonder if the bitbake.conf file that the reproducible build is generating doesn't get put in the right spot for some reason Mar 04 20:38:46 RP: is there a contact for the yp.org website? all the default links on https://www.yoctoproject.org/docs/ for 3.0.2 are 404ing... Mar 04 20:39:54 halstead: 20:38 < smurray> RP: is there a contact for the yp.org website? all the default links on https://www.yoctoproject.org/docs/ for 3.0.2 are Mar 04 20:39:57 404ing... Mar 04 20:40:26 LetoThe2nd: heh, I wasn't sure if halstead got saddled with that on top of his many other responsibilities Mar 04 20:41:42 smurray: even if hes not, he knows whom to poke. Mar 04 20:44:23 LetoThe2nd: true Mar 04 20:51:04 smurray: I'd mention to halstead Mar 04 20:51:37 tgamblin: yes, that one Mar 04 20:52:05 smurray: Its halstead, ndec or send email Mar 04 20:53:40 RP: thanks Mar 04 21:08:55 khem, even in the 2nd release, they seem pretty careful to say they are using Microsoft's operating system. Which happens to have a Linux base. But it's not Linux, it's Microsoft. Mar 04 21:50:36 Since I wasn't using busybox for anything (systemd, util-linux, etc.), I turned it off. But it turns out the "resize" program, to set the serial console size, was from busybox Mar 04 21:50:58 The only alternative I can find is a version that's in xterm. Mar 04 21:51:30 It seems the latter is not in poky? Mar 04 21:53:02 because it's on openembedded. But it requires the whole X11 library suite to get this one console program. Mar 04 22:17:04 xyzzy42: yeah, that's why it's in busybox, to avoid needing to pull in X Mar 04 22:23:30 xyzzy42: if you can live with busybox in the image, but just don't want all the applets, could over-ride the defconfig to turn everything else off. Otherwise, I think you'd be stuck re-implementing, perhaps by grabbing the code from X11 and pruning it. It's a pretty simple utility, just some termios.h tc*attr calls in the busybox version Mar 04 22:45:49 Wow, a green a-full build! :) Mar 04 22:46:10 amazing what you can do with css Mar 04 22:48:11 background: #0f0; ? Mar 04 22:52:09 RP, nice! Mar 04 23:23:41 Hello...anybody encountered this error " Exception: TypeError: join() argument must be str or bytes, not 'NoneType'" on do_unpack Mar 04 23:25:15 my SRC_URI="git://GITURL;protocol=ssh;branch=master" Mar 04 23:25:22 in the recipe Mar 04 23:26:15 smurray, thanks, did that it, works ok. Made a busybox defconfig with mostly just resize and the console is working. It's amazing how awful "vi and less don't work" is. Mar 04 23:27:10 xyzzy42: heh, indeed. journalctl is the one that usually I need to resize for Mar 04 23:30:42 Oddly, the xterm console doesn't work correctly in minicom for inserting text on a command line. Does work in vt100. Doesn't in vt102. Mar 04 23:31:11 I could have sworn that used to work Mar 04 23:34:08 it did, the problem appears to be "-R UTF-8" as a minicom option. But without that, the UTF-8 stuff in systemd doesn't display. And the dots and line drawing looked so nice.. Mar 05 00:22:38 New news from stackoverflow: Why is my app a different size on my yocto build server compared to what is installed on target? Mar 05 00:49:40 is there a way to get poky's systemd-conf recipe to NOT install one of its snippets, without entirely replacing the do_install target? Mar 05 00:50:28 write a bbappend and delete it from staging Mar 05 00:54:51 khem, thanks, that's worked. I was trying to find a way to not install it, but removing it after the fact is easier. Mar 05 00:55:26 The file I'm removing is the journald config that directions journald to log to syslogd. I don't see the point of having two syslogds on the system. Mar 05 01:02:25 xyzzy42: why not just make an override Mar 05 01:02:31 journald has a drop-in dir for its conf Mar 05 01:03:11 https://www.freedesktop.org/software/systemd/man/journald.conf.html Mar 05 01:06:48 mischief, that was my plan if not installing the original override was too hard. But I hate decoy config files. There'd be a 00-systemd.conf that says we're logging to syslogd, but ha, not realy! Our config is in another castle. Mar 05 02:02:20 the default config has all options commented.. pretty obvious if you have an explicit dropin somewhere. Mar 05 02:54:09 mischief, poky puts an explicit dropin to turn on forwarding to syslog in the dropin dir. It's not the default config with all the options commented out. So if I add another dropin, now there are two different dropin files with uncommented options that conflict. If you just see one, you might think you know how something is configured and would be wrong. **** ENDING LOGGING AT Thu Mar 05 02:59:56 2020