**** BEGIN LOGGING AT Thu Mar 21 02:59:57 2019 Mar 21 07:47:12 tz you could try to create a dummy recipe requiring only RDEPENDS Mar 21 07:53:55 New news from stackoverflow: Yocto /boot partition not using opkg/ipk packages Mar 21 08:17:07 hi Mar 21 08:20:23 hi guys Mar 21 08:20:41 i have a question if someone can help... i am using bitbake and i need some header but i dont need sozrce code where functions are declared. but until i add src uri for that files i get undefined reference. i need some way to include them but i don't want to build them as part of my recipe cause they dont belong logically there...and also they generate warnings when sourced... Mar 21 08:21:39 i am maybe bad at explaining cause i complicate things, also i lack some technical knowledge 😌😳 Mar 21 08:44:44 Absentee: sounds like you jsut want a DEPENDS on the package that provides the headers Mar 21 09:04:00 hi guys i have a question if someone can help... i am using bitbake and i need some header but i dont need sozrce code where functions are declared. but until i add src uri for that files i get undefined reference. i need some way to include them but i don't want to build them as part of my recipe cause they dont belong logically there...and also they generate warnings when sourced... Mar 21 09:05:19 if you had stayed: Mar 21 09:05:26 08:44 < LetoThe2nd> Absentee: sounds like you jsut want a DEPENDS on the package that provides the headers Mar 21 09:06:20 but as you are being pretty much unclear about which headers, its hard to offer a better answer. Mar 21 09:06:48 headers are generated by me from proto file so they are not part of some librarY Mar 21 09:07:17 then you either have to add them to your source, or automate the generation Mar 21 09:07:47 you can totally have those "proto" tools as a native dependency and do the generation as a task before compiling Mar 21 09:08:45 and bitbake will then "see" .cc files for those headers without sourcing them? Mar 21 09:09:14 bitbake will hopefully see no .cc or .h files at all. Mar 21 09:09:21 :D Mar 21 09:09:24 it up to your build system to see those things Mar 21 09:09:35 bitbake can just trigger the tools and procedures you need. Mar 21 09:10:15 kergoth: thanks for the explanation about fetcher implementation yesterday Mar 21 09:10:47 u1106: he'll probably not be around for the next 4-5 hrs (timezone) Mar 21 09:11:09 huh...interesting... i probably should read more about that :D Mar 21 09:11:36 and I left while he was there :) Mar 21 10:35:35 hello Mar 21 10:39:53 Does someone know why there is udev in my rootfs even though I use mdev. It adds more than 10MB to my rootfs :(... Mar 21 10:41:23 And I've upgraded my environment from Krogoth to Sumo Mar 21 10:42:02 Toub: bitbake -g $YOURIMAGE, then inspect package-depends.dot Mar 21 10:42:34 there is also some oe-something tool to assist, but the dot file is easily search- and readable in itself Mar 21 10:46:06 I found a package called "eudev" so I suppose it's udev Mar 21 10:48:23 it seems eudev is a dependency of usbutils and packagegroup-base recipes Mar 21 10:48:53 then this is probably your answer. Mar 21 10:49:58 so there is no way to remove it? I didn't need it when I was using the krogoth branch :( Mar 21 10:50:10 Toub: there is always a way Mar 21 10:51:34 you basically have to walk the dependency chain and then break it at the point thats most suitable for you Mar 21 10:53:49 ok thanks :) Mar 21 10:54:47 maybe its even enough to tweak some virtual provider, but thats beyond my expertise then. Mar 21 11:33:58 LetoThe2nd: thanks to you I managed to get rid of udev. thx again Mar 21 11:34:33 Hi, can someone help me identify my problem, I created a recipe that installs my qt application binary only with help of "Inherit=qmake5". Now when im trying to recompile i get the following error https://pastebin.com/3e7xVzaM Mar 21 11:35:02 Which is wierd since i only used a runable application in the first place Mar 21 11:36:16 The recipe is very simple : https://pastebin.com/UfSK0Dd1 Mar 21 11:49:18 Toub: glad to hear. mind to give a short outline, so we can give better advice if the question arises again? Mar 21 12:00:53 Or does qmake5 require the source files to be able to install qt applications ? Mar 21 12:12:52 you need source files to build the source files, yes Mar 21 12:13:15 looks like that tarball doesn't contain what it expected Mar 21 12:15:39 rburton: So you need to compress the entire project folder and include it, I compiled and put the binary in the recipe Mar 21 12:16:07 willie2: easier to just point it at your source repo Mar 21 12:16:34 if you want to just install the binary you built, that's weird, but just remove the qmake bit and just copy the file in do_install Mar 21 12:45:24 hi Mar 21 12:46:14 help pls. How to install kernel-modules package on live image? Mar 21 12:47:33 i tried CORE_IMAGE_EXTRA_INSTALL and IMAGE_INSTALL Mar 21 12:47:36 the same way as any other package. put "kernel-modules" in your IMAGE_INSTALL variable (or a packagroup that is being installed). Mar 21 12:47:48 if that didn't work, you aren't building an image that uses those variables. Mar 21 12:50:09 rburton: thanks, can you help me understand something else, the application is 7,1M in /work/cortex..../application/image../application but it is only 2,5M on the target system? Mar 21 12:50:11 kernel-modules ipk is empty and depends on all modules. then i open image there is only modules for installed package groups Mar 21 12:50:28 willie2: because we strip binaries Mar 21 12:51:22 rbutron: Okay thanks, I tought something was wrong :) Mar 21 12:53:32 alex_eri: then maybe you are building a kernel config that does not compile any modules? Mar 21 12:54:10 there is modules ipks Mar 21 12:54:12 alex_eri. kernel-modules is a meta package, it should just hold the dependencies. Mar 21 12:54:35 when you install it, the image assembly will bring all the individual packages in. Mar 21 12:54:50 ok, i know. trying to install it with deps Mar 21 12:59:04 rburton: I'm curious, what exactly is beeing stripped? I had two versions, A=6,1M B=7,5M both 2,5M @ target Mar 21 13:00:46 i have https://gist.github.com/alex-eri/4ca44ac056b855e18160fd47f8f83b0f Mar 21 13:02:53 modules in second file installed to image, ipks not installed with metapackage... trying with MACHINE_EXTRA_RDEPENDS Mar 21 13:13:21 willie2: the debug symbols Mar 21 13:25:00 New news from stackoverflow: How to make loadable file system over the read only rootfs? Mar 21 13:55:05 New news from stackoverflow: One of Application on my Yocto 2.0 Linux board is NOT going to background Mar 21 14:37:42 LetoThe2nd: of course ;) What do you want to know? Mar 21 14:38:47 So... I find myself in charge of a build where QA is being *really* pedantic about proving that we can build our Yocto-based system without touching the Internet. Mar 21 14:38:55 I'm running into some problems with git submodules. Mar 21 14:39:28 Can somebody confirm that git submodules completely bypass SOURCE_MIRROR_URL? Mar 21 14:40:00 That's what I'm seeing, and it makes sense. Just wanted to confirm, and... maybe solicit ideas for workarounds. Mar 21 14:40:39 (I mean: workarounds for the fact that I'm not 'allowed' to touch the Internet during the build.) Mar 21 14:41:19 evadeflow: kergoth and fray are the best people to answer that Mar 21 15:00:41 otavio:rtti should work fine I have not used it myself so dont know first hand Mar 21 15:11:25 i was going to suggest that maybe own-mirrors doesn't have a gitsm:// entry, but they left.. Mar 21 15:17:11 I guess this is confirmation enough: https://bugzilla.yoctoproject.org/show_bug.cgi?id=11831 Mar 21 15:17:12 Bug 11831: enhancement, Medium+, Future, kergoth, ACCEPTED , Proposal to replace/rewrite gitsm Mar 21 15:18:43 evadeflow: the gitsm rework in master, currently in the queue to backport into thud, fixes that Mar 21 15:20:14 Good to know. I'm on sumo. Not sure whether to try pulling ahead, or trying 'backporting the backport'. Mar 21 15:20:25 Or... something else. Mar 21 15:21:39 As usual, I'm in a hair-on-fire hurry, so I may clone the 'problem' repos---along with their submodules---then create a new 'mega fork' that 'un-submodules' everything. Mar 21 15:21:46 evadeflow: if you look on the bitbake-devel list, it's a series of 10 patches from mhatle, all but one are to one file. I've not looked to see how hard it is to apply them to sumo Mar 21 15:21:57 But... it makes me question my career choices. `:-/ Mar 21 15:22:29 smurray: that sounds promising Mar 21 15:24:08 evadeflow: you could try dropping the newer gitsm.py into the sumo tree and trying it, but I suspect there may be enough differences that it won't work without some tinkering Mar 21 15:25:28 Sadly, this BSP is currently on *morty*. But the vendor (Variscite) just updated the layers we based the work on to sumo, so... sumo is achievable without too much effort. Mar 21 15:25:43 I guess I better do that sumo port sooner rather than later! Mar 21 15:27:53 armpit: https://autobuilder.yocto.io/pub/non-release/20190223-15/testresults/testresult-report.txt Mar 21 15:36:52 ugh, trying to fix my scriptutils log improvements series from a couple weeks ago, but i have yet to see the recipetool and devtool self tests pass here even without it applied Mar 21 15:37:49 i was going to suggest the same thing, provide a new gitsm.py in your layer and inject it to replace the existing one in the fetcher, much like how the npm layer added the npm fetcher from the metadata Mar 21 15:37:57 hack, but you're stuck with hacks no matter wha tyou do at this point Mar 21 15:38:26 * kergoth yawns Mar 21 15:38:48 Interesting. Mar 21 15:38:54 also, this error message stating that this selftest won't pass without running runqemu-gen-tapdevvs is not helpful when you have no idea what arguments to pass to the thing.. Mar 21 15:38:56 ugh Mar 21 15:39:59 https://github.com/imyller/bitbake-npm/blob/master/classes/npm-fetch.bbclass + https://github.com/imyller/bitbake-npm/blob/master/lib/npm.py -- of course, .append() won't do for you, since you'd wnat to replace the existing one, but you should see the idea Mar 21 15:40:19 When you say "inject it" (a new gitsm.py)... Mar 21 15:40:29 Ah, okay, looking at those links... Mar 21 15:43:55 anyone know how to enable persistent systemd logs on pyro? Mar 21 15:44:07 kergoth: runqemu-gen-tapdevs should tell you how to run it. could be better, could be worse Mar 21 15:44:33 kergoth: I guess I don't quite 'get it' yet. So... the existing gitsm.py is in `poky/bitbake/lib/bb/fetch2/gitsm.py`. To override that, I create a `lib` folder in my layer, and then... Mar 21 15:44:50 Actually, have I got *that* part right? Mar 21 15:44:59 I need a `lib` folder with the new gitsm.py? Mar 21 15:45:57 RP: 'the numeric user id the tap devices will be owned by' isn't useful unless you know the tests run as the build user. ' is the number of tap devices to create' is only useful if you have a clue how many it needs. etc. every argument assumes knowledge you don't necessarily have Mar 21 15:46:58 evadeflow: yes, then copy it over, and in a globally inherited class, use anonymous python to replace the instance of bb.fetch2.gitsm.GitSM with an instance of gitsm.GitSM Mar 21 15:47:00 basically Mar 21 15:47:09 kergoth: the example should at least print sensible defaults. We should perhaps say that Mar 21 15:47:10 (copy gitsm.py from bitbake to your layer, that is) Mar 21 15:47:34 it prints an example, but it just has ids like 1004 1001 and a num of 4. is that right for me? probably not Mar 21 15:47:38 * kergoth shrugs Mar 21 15:48:12 kergoth: well, it did figure out your user and group which is what most people would use Mar 21 15:48:31 it used to hardcode 1000 1000 4 Mar 21 15:48:43 kergoth: patches to make it more understandable very welcome Mar 21 15:48:47 it did, but there's no way to know, looking at it, that it's not just a fixed example, unless you already know your uid and gid :) Mar 21 15:49:36 kergoth: I agree we should actually say that Mar 21 15:50:05 kergoth: Thanks, I think I get it. The example you linked does `bb.fetch2.methods.append( npm.NPM() )`. Do I need to *remove* the existing impl, or will call append() Just Work™? Mar 21 15:50:45 evadeflow: yeah, exactly, you'd have to remove the existing one, or replace it, otherwise bitbake will just use the first one it finds that supports the url scheme in question Mar 21 15:51:10 kergoth: Cool—*thank you*!! Mar 21 15:51:46 evadeflow: i.e. for i, m in bb.fetch2.methods.enumerate(): if isinstance(m, bb.fetch2.gitsm.GitSM): break; else: bb.fatal("Failed to find GitSM?"); bb.fetch2.methods[i] = gitsm.GitSM() Mar 21 15:52:28 of course that's assuming the copied fetcher module *works* with the older bitbake and possibly python you're using, but that's a different issue you'd run into trying to backport it as well Mar 21 15:52:35 and you could fix those on a case by case basis Mar 21 15:52:37 if any Mar 21 15:53:27 Gotcha. Thanks for filling in the override details there, that saves me a lot of finger-mumbling around. Mar 21 15:53:37 in any case, that's just one suggested possibility for dealing with it, no guarantees Mar 21 15:53:39 good luck Mar 21 15:53:45 np Mar 21 15:53:59 er, that should be enumerate(bb.fetch2.methods) Mar 21 15:54:11 that's what happens when i write untested code without enough coffee :) Mar 21 15:54:16 LOL> Mar 21 15:54:20 No worries. Mar 21 15:54:55 enumerate() just lets you get both the index and the value when iterating, so we know what index in the list to replace with the new entry in a pythonic way Mar 21 15:55:01 * kergoth wanders off to get some actual work done Mar 21 15:57:46 ERROR: gitsm: submodule unpack failed: UnpackError Unpack failure for URL: 'gitsm://github.com/arfoll/doxygen2jsdoc.git;protocol=https;name=doxygen2jsdoc;subpath=doxygen2jsdoc;barecl Mar 21 15:57:46 one=1;nobranch=1'. No up to date source found: clone directory not available or not up to date: /data/kergoth/mel/yocto/scriptutils/build.selftest/downloads/git2/github.com.arfoll.do Mar 21 15:57:46 xygen2jsdoc.git; shallow clone not enabled Mar 21 15:57:49 * kergoth bangs head against desk Mar 21 16:24:30 http://build.webos-ports.org/sources.openembedded.org.access.html this is interesting, looks like the sources.oe.org domain was moved few months ago (which matches with info from halstead) Mar 21 16:25:52 someone has badly configured mirrors it seems.. /git2_${REPOSITORY_SERVER2_PATH}.msrtool.git.tar.gz Mar 21 16:27:31 4 from 11 top hits on sources.oe.org had this unexpanded REPOSITORY_SERVER2_PATH variable Mar 21 16:29:24 anyone from fujitsu here in this channel? see ^ Mar 21 16:31:39 well it stopped around Jun/2016, but in just 2 months it generated 10% of all time requests for sources.openembedded.org Mar 21 16:31:49 yikes Mar 21 16:32:06 counting only those with REPOSITORY_SERVER2_PATH Mar 21 16:45:34 JaMa, moving git2_code.qt.io.qt.qtwebengine-chromium.git.tar.gz aside and letting rsync run did avoid the timeout. I've set the timeout to 6000 and I'm seeing if it will complete. Mar 21 16:45:52 JaMa, libqmi-1.22.2.tar.xz and everything else is mirrored. Mar 21 16:46:16 halstead: cool, thanks Mar 21 16:55:57 And it did complete. Mar 21 17:15:30 Anyone know how to enable persistence for systemd logs? I tried to create a symlink from /var/log/journal to a location on the sdcard, but that really messed things up Mar 21 17:52:01 nate02:you have to do mount magic Mar 21 17:52:25 symlinks might not help if the location is not mounted Mar 21 17:53:24 khem, thanks. Can you explain a bit more? Mar 21 17:54:19 mount /var/log somewhere /dev/sda2 or where ever you have persistent storage Mar 21 17:55:50 I did think about that. But is it possible to just moun the journal folder? I'm a little worried about write cycles on my sdcard. The journal has settings, but there are other log files in /var/log Mar 21 17:56:41 btmp, dnf, hawkey, lastlog, messages, wtmp Mar 21 18:02:14 well journal will do the most damage anyway Mar 21 22:35:59 if anyone needs to use gcc-7 with latest oe-core, then it needs few more changes than just restoring the recipes: removing gcc-cross-initial, updating DEBUG_PREFIX_MAP and small change in libxcrypt. You can check https://github.com/webOS-ports/meta-webos-ports/commit/61d6c22f31e98a4e63bae06c215c0140ea164ed7 Mar 22 01:14:56 anybody have any idea how to debug init scripts without changing behaviour? I've got a bunch of scripts that seemingly don't execute at startup, but if I add "#!/bin/bash -ex\nexec >> /var/log/$0-init 2>&1" at the top they all work fine :s Mar 22 01:15:41 sounds like you made them run under bash instead of /bin/sh, which could be an entirely different shell Mar 22 01:16:20 yep, the exec line wasn't working under /bin/sh Mar 22 01:16:31 although this is a later iteration, might rollback to it to see if it works Mar 22 01:16:43 (the exec line with sh) Mar 22 01:16:55 the ones that are failing are things like networking, which shouldn't really be an issue afaik Mar 22 01:17:49 and syslog... which makes things a bit annoying obviously Mar 22 01:18:55 still works with /bin/sh and the exec line, hmmm **** ENDING LOGGING AT Fri Mar 22 02:59:57 2019