**** BEGIN LOGGING AT Wed Oct 21 10:59:57 2020 Oct 21 11:10:53 while generating eSDK (on a different machine than the one that produced the sstate) I'm getting systematically getting the single error "ERROR: Task packagegroup-core-x11-xserver.do_populate_lic attempted to execute unexpectedly and should have been setscened", while diffsigs show no diff between that task's sigdata and the one in sstate, any idea ? Oct 21 11:39:17 yann: is hashequiv enabled or not? is there an sstate artefact with that hash? Oct 21 12:05:27 RP: no hashequiv. Do you mean the hash from the do_populate_lic.sigdata filename ? Oct 21 12:09:35 hm should I a) make a separate image bb file for development version of my image, or have conditional includes based on my own DISTRO_FEATURES variable? Oct 21 12:09:51 yann: are the hashes from the sigdata and the one in sstate the same? Oct 21 12:10:26 Ad0: that is one of those "it depends" situations Oct 21 12:11:20 RP: yes, same hash Oct 21 12:12:14 RP: ah I've overlooked the /unix suffix in that tcl, sorry for noise Oct 21 12:12:43 yann: that suggests the eSDK can't find the sstate then as if the hashes match, they should be used :/ Oct 21 12:12:53 JaMa: its ok, I did too at first Oct 21 12:26:55 RP, yeah Oct 21 13:21:08 Is there any way to remove override all _append statements on a variable? I thought I could do this with _forcevariable but it looks like the appends still get applied after that Oct 21 13:30:12 paulbarker: _remove, but that's ugly, if you can change the original recipe then I usually use separate variable which you can then just set to empty if you don't want the _append to add anything Oct 21 13:31:14 2nd stage is to use separate variable also in the _remove so that it can be disabled once you find out that special case or change your mind later in some other layer Oct 21 13:31:39 JaMa: I need to cancel any _append statements performed in machine conf files without knowing what they are Oct 21 13:32:22 https://github.com/webosose/meta-webosose/commit/d0b237b70264e269f75303a11c43b2eea25de991 is indeed quite ugly :/ Oct 21 14:01:37 Hello all, I'm trying to manage several distros but I have probably miss something. The main goal is to select correct version according the selected distro... Oct 21 14:04:49 Ninic0c0: have recipes for all versions, and the distro sets PREFERRED_VERSION. thats the canonical approach for packages/recipes. Oct 21 14:06:10 paulbarker: you can reset things from anonymous python with a setVar iirc Oct 21 14:08:00 @Leta Oct 21 14:08:14 LetoThe2nd Thx for the tip! Oct 21 14:10:01 :) Oct 21 15:17:25 #git Oct 21 15:17:30 it Oct 21 15:51:58 sakoman: Sorry, I am trying to sort something on the autobuilder so your build was cancelled Oct 21 15:52:27 sakoman, demand a refund! Oct 21 15:52:43 Oh, sorry, restarted it thinking I screwed up somehow! Oct 21 15:53:25 just realized that the YPS schedule says its all in the afternoon, late afternoon, even evening for EU timezones. is this correct and intentional? Oct 21 15:54:29 ndec: ^^^^ Oct 21 15:54:48 It's correct, and intentional ;) Oct 21 15:55:06 it's the same 'schedule hours' as ELCe Oct 21 15:56:01 seriously? for some reason i was totally expecting a somewhat 9-5ish thing as conferences usually are. Oct 21 15:57:39 *sigh* so much of an USian organizer taking care of Europe. Oct 21 15:59:06 LetoThe2nd: already ranted about it last week with rburton and someone else Oct 21 15:59:19 but yeah, that's f-ed up Oct 21 15:59:56 yeah for some reason they picked US eastern as a base time despite the point of ELC-E being that it's for europeans-and-further-east Oct 21 16:00:18 genuinely stupid Oct 21 16:01:40 yeah. thats really hilarious. Oct 21 16:01:56 LetoThe2nd: I'm surprised you don't want to watch a talk at 10pm though Oct 21 16:03:47 has anyone ever managed to get qmake to work with yocto sdks? when I tried to use nativesdk-packagegroup-qt5-toolchain-host , qmake did not find qmakespec files. Oct 21 16:04:08 and I am confused what meta-toolchain-qt5 is for. "Meta package for building an installable Qt5 toolchain and SDK" - so it is its own entire SDK? why would one want that? why have an entirely separate SDK just for qt5? Oct 21 16:05:11 No trolling now, https://twitter.com/TheYoctoJester/status/1318931269141528576 Oct 21 16:08:30 how to get git tag from the yocto repo? essentially wan't to do "git describe" Oct 21 16:16:30 what? there is life outside the office?? Oct 21 16:17:45 now I know that I should blame daily calls at 6PM and 9PM :) Oct 21 16:18:59 JaMa: outside office == drinking space Oct 21 16:19:18 I drink in the office Oct 21 16:19:30 hehe Oct 21 17:57:21 lxc: you need to do that externally, I suppose you are trying to get tags for yocto repositories not for the components its building Oct 21 17:59:08 dv_: its about usecase, many folks write QT apps standalone, using QTCreator and other tooling provided by QT itself, so having a SDK address that is what meta-toolchain-qt5 does Oct 21 18:01:18 halstead RP: is AB down ? I am getting 502 Bad Gateway errors Oct 21 18:02:25 khem checking Oct 21 18:04:11 khem, Yes it is. Checking the logs and then bringing up Oct 21 18:06:29 khem, Looks like it was send SIGTERM and exited gracefully. Oct 21 18:06:49 khem, Back up now. Oct 21 18:14:32 khem: ah, okay Oct 21 18:14:41 khem: any way how I can qmake to work with a yocto sdk Oct 21 18:15:13 right now, I get "Cannot find feature spec_pre.prf" Oct 21 18:17:40 Hello...I get this error when trying to start toaster using "source toaster start" Oct 21 18:17:44 Failure while trying to setup toaster: Problem installing fixtures: no such table: orm_build__old Oct 21 18:17:49 More details here --> https://pastebin.com/qfMusX7V Any ideas? Oct 21 18:17:54 All toaster dependencies are installed inside python virtual env ie. I am using python virtualenv for this Oct 21 18:17:56 khem think ${GITVER} may do what I want Oct 21 18:18:28 halstead: sorry, I've been trying to fix something Oct 21 18:18:50 moto-timo: ^^ Oct 21 18:19:57 khem: please hold off builds for the moment, I need to fix something I've broken Oct 21 18:20:01 RP did I mess up your procedure by restarting the controller? Oct 21 18:20:58 kiwi_29: I've never seen that, but perhaps you need to run the migrations... $ source toaster migrate Oct 21 18:21:10 * moto-timo needs to double check syntax Oct 21 18:21:24 thanks alejandrohs for the ping Oct 21 18:21:34 moto-timo ... I did and it gave me error too. :( Oct 21 18:22:00 kiwi_29: what is your host linux distribution and version? Oct 21 18:22:49 ubuntu 20.04 Oct 21 18:22:56 kiwi_29: also, the CROPS toaster-container works so you could go with docker instead https://github.com/crops/toaster-container Oct 21 18:23:07 kiwi_29: ok let me kick the tires in a vm Oct 21 18:24:32 RP: right, I am also doing a bit of AB testing :) Oct 21 18:26:36 kanavin_home: sorry, it will be back shortly Oct 21 18:26:57 What should have been a quick change in idle time earlier turned into a total disaster :( Oct 21 18:27:03 moto-timo the error when I run "../bitbake/lib/toaster/manage.py makemigrations" is https://pastebin.com/FBPruNjU Oct 21 18:27:23 dv_: usually adding nativesdk variant of qmake is all you need Oct 21 18:27:47 halstead: it didn;t help but no harm done Oct 21 18:28:03 nativesdk-qmake ? Oct 21 18:28:25 RP: ok so bring it down was intentional? maybe a good msg instead of 502 would have helped :) Oct 21 18:28:33 dv_:yes Oct 21 18:28:46 khem: we've no way to do that :/ Oct 21 18:29:18 hmm ok Oct 21 18:29:46 usually a standing page to redirect queries maybe Oct 21 18:30:29 kiwi_29: that seems like a permissions issue. Does the sqlite database file exist and does your user have permissions on it? Oct 21 18:31:12 kanavin_home, sakoman, khem, halstead: ok, its back now and the BuildLog archival is now working as expected Oct 21 18:31:17 Nothing RPROVIDES 'nativesdk-qmake' Oct 21 18:31:20 sorry about the interruptions Oct 21 18:31:40 * RP was worried about it breaking the wiki pages Oct 21 18:31:54 my vm is out of disk space meh Oct 21 18:33:16 RP: thanks, I started a-full :) Oct 21 18:34:00 fray: you owe me one http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=akanavin/package-version-updates&id=ae10980f0923e8379e3607bf13d3b69911ab6fb8 Oct 21 18:34:54 moto-timo .. sqlite3 itself was not installed.. I did apt install sqlite3 but still receiving same error Oct 21 18:35:22 let me do the whole thing again and will report Oct 21 18:41:09 I do have write permission on toaster.sqlite file yes Oct 21 18:41:32 same error after redoing the whole thing Oct 21 18:41:38 I followed the toaster manual Oct 21 18:43:33 kiwi_29: it shouldn't be too different, but double check https://docs.yoctoproject.org/singleindex.html#document-toaster-manual/toaster-manual Oct 21 18:46:38 kiwi_29: I started a toaster-container build/test cycle with ubuntu-20.04 base container Oct 21 18:47:10 does the actual build also happen inside the container...or just toaster runs in container? Oct 21 18:47:32 https://travis-ci.org/github/moto-timo/toaster-container/builds/737794458 Oct 21 18:48:03 toaster and bitbake run in the container, your workspace should be on your host and bind mounted in as a docker volume Oct 21 18:48:35 so it will run with all the available cores and so on. Very little performance hit if any to running with Docker Oct 21 18:49:11 got it... let me try to solve my issue..if it is too much of hassle..I will have to shift to docker Oct 21 18:49:20 I have a lot of custom layers and stuff Oct 21 18:49:29 but I also tried on vanilla poky and still got same errors Oct 21 18:50:30 kiwi_29: feel free to file a bug on Toaster and we can dig in deeper. Not much more I can do until after the YP Summit next week Oct 21 18:50:40 time constrained until then Oct 21 18:51:22 no problem...thank you for your time moto-timo .. I will ping again though ;) if I m stuck Oct 21 18:51:49 kiwi_29: happy to help... within the time I can spare right now :) Oct 21 18:54:56 kiwi_29: I pinged a co-worker to see if she has any other insights Oct 21 19:00:34 Do I have `devtool deploy-target` available in an Yocto-created SDK? Oct 21 19:01:07 manuel1985: you need extensible SDK for that Oct 21 19:01:34 manuel1985: but yes, it should work... I haven't done that in a couple years though Oct 21 19:01:36 Oh, yeah, I was thinking about the extensible SDK to be honest :D Great, thank you Oct 21 19:03:11 manuel1985: after the YP Summit next week I plan to revisit extensible SDK and the crops container for it. We as a community need to figure out how to use it with e.g. VSCode Oct 21 19:03:39 manuel1985: including the Debug Console Oct 21 19:04:40 Looking forward to that! And the the YP summit as well. I'm already signed up. Yes, VSCode integration sounds very interesting, and eSDK, and CI/CD... and just about everything :D Oct 21 19:37:21 apparently travis has a huge queue Oct 21 19:48:52 moto-timo: vscode+crops should be super trivial Oct 21 19:51:10 rburton: please blog about it Oct 21 19:51:32 rburton: especially including debugging and deploy to target Oct 21 19:52:09 vscode plus pyrex too Oct 21 19:52:18 JPEW: ^^^ Oct 21 19:59:42 many thanks moto-timo ... let me know if there is any solution to the issue Oct 21 20:16:05 kiwi_29: the travis job errored out, so it might be time for you to file a bug on it Oct 21 20:16:24 got it...will do moto-timo Oct 21 20:24:05 How do I need to write the `install(TARGET)` clauses of my CMakeLists.txt that they play well with Yocto and any other distribution? Oct 21 20:41:49 * derRichard finds recipes-kernel/linux/kernel-devsrc.bb and pukes Oct 21 20:41:53 guys....really... Oct 21 20:42:12 it took me a full hour to figure why the kernel source in my sdk is complete bonkers Oct 21 20:42:21 now i find kernel-devsrc.bb... Oct 21 20:43:39 you assume a specific source layout. $customer kernel has patched in many drivers/headers with a layout which kernel-devsrc.bb does not understand Oct 21 20:47:43 derRichard: You are trying to compile kernel modules in the SDK? Oct 21 20:48:10 yes Oct 21 20:48:24 so, i need a "sane" kernel tree :-) Oct 21 20:48:47 we did an upgrade from sumo to zeus and faced super odd sdk regressions Oct 21 20:49:26 e.g. our kernel makes use of xenomai, kernel-devsrc.bb misses most xenomai headers/c files. so the kernel shipped by the sdk is broken Oct 21 20:49:32 and cannot be used to build modules Oct 21 20:50:11 cp --parents $(find -type f -name "Makefile*" -o -name "Kconfig*") $kerneldir/build Oct 21 20:50:14 cp --parents $(find -type f -name "Build" -o -name "Build.include") $kerneldir/build Oct 21 20:50:29 is clever, but in xenomai not every directory has a Build or Makefile... Oct 21 20:50:49 so whole directories of headers and source files are missing Oct 21 20:51:40 derRichard: Interesting. I've never tried to compile a kernel module with the SDK before Oct 21 20:52:34 why else do you ship a kernel tree with the sdk? Oct 21 20:52:49 SUMMARY = "Linux kernel Development Source" Oct 21 20:52:49 DESCRIPTION = "Development source linux kernel. When built, this recipe packages the \ Oct 21 20:52:52 source of the preferred virtual/kernel provider and makes it available for full kernel \ Oct 21 20:52:55 development or external module builds" Oct 21 20:52:59 sounds like building modules is the purpose :-) Oct 21 20:53:17 anyway, sorry for the rant. i try to find a fix for now Oct 21 20:53:33 derRichard: I don't doubt that the purpose, I just haven't ever found a need to do it :) Oct 21 20:54:49 lucky you Oct 21 20:55:14 But, I also most use monlithic r/o root filesystem images that get replaced with a new image when updated, so building a single module outside of Yocto itself (and having it baked into the rootfs in the process) isn't useful Oct 21 20:57:26 JPEW: yocto produces a full rootfs for us, with all modules. but developers need to work with the sdk to be able to rebuild some parts of the system quickly Oct 21 20:57:53 so, they have the sdk installed and just clone the main software and type make Oct 21 20:58:31 we actually ship a sdk in terms of software development kit to our inhouse developers Oct 21 20:58:50 and only the core team is allowed to mess with yocto directly Oct 21 20:59:02 derRichard: Ya, we do that too (with waf though.... I don't recommend it ;) ) Oct 21 20:59:57 (saldy) a huge part of the main software are kernel modules Oct 21 21:00:05 *sadly Oct 21 21:00:09 derRichard: Ah, that is sad Oct 21 21:03:07 derRichard: Sorry you're running into issues. With some things we can't win with the size/time/completeness conflicts :/ Oct 21 21:03:29 I'd think the kernel-devsrc changes were in the migration guide Oct 21 21:05:58 RP: well, i didn't realize that kernel-devsrc works only with "well known" kernels Oct 21 21:12:08 RP: is it really worth the hassle? git log on meta/recipes-kernel/linux/kernel-devsrc.bb shows that the thing needs fixing for every new kernel version/arch over and over :-( Oct 21 21:12:53 derRichard: it uses huge amounts of space/resources in testing on the autobuilder? Oct 21 21:13:15 JPEW: waf is a curse word in German for a reason Oct 21 21:13:51 moto-timo: It's getting to be a curse word in English for me.... Oct 21 21:13:59 RP: disk space is cheap. developer's mental health not ;-) Oct 21 21:14:10 JPEW: better than bazel I suspect Oct 21 21:14:22 moto-timo: huh? i speak german but don't know "waf" as word. Oct 21 21:14:33 derRichard: its also time. What is developer time in test cycles runs worth? Oct 21 21:14:37 moto-timo: Never used it, but I suspect you are correct Oct 21 21:14:53 derRichard: clearly my mental health is worthless for example Oct 21 21:15:10 RP: we have both good arguments. right now i'm a little grumpy. so... ;-) Oct 21 21:15:11 derRichard: I might have got that wrong but it was one of the names of a server that was supposedly based on German curse words Oct 21 21:15:37 derRichard: I'm just trying to show its not an entirely straight forward issue :/ Oct 21 21:15:47 derRichard: I think the SDK and kernel (and extensible SDK) needs some love and perhaps rethinking things Oct 21 21:15:54 RP: indeed Oct 21 21:16:31 moto-timo: agreed Oct 21 21:18:29 derRichard: The sad thing is I think we have tried all the different approaches. We get complains whichever one we use (its too big, its too slow, its incomplete, etc) Oct 21 21:18:56 and if its configurable, its not obvious and nobody knows :) Oct 21 21:19:00 we just can't win Oct 21 21:19:06 derRichard: what we really need are use cases and the collateral so we can actually write test cases... but so much of the vendor kernel stuff is closed/private/not FOSS Oct 21 21:19:27 moto-timo: this one isn't really about use cases, we know them, we just can't win :( Oct 21 21:19:32 RP: it is indeed hard to please everyone Oct 21 21:19:47 RP: not my normal daily usage, so I'm a bit naive Oct 21 21:20:04 how about offering a kernel-devsrc-minimal, kernel-devsrc-full or kernel-devsrc-default? Oct 21 21:20:14 in my case, kernel-devsrc-full is also fine Oct 21 21:20:34 derRichard: we probably should. Which one do we test? Oct 21 21:20:47 all three? Oct 21 21:24:41 isn't kernel-devsrc a target package though? I guess I don't understand how it fits in the context of the sdk Oct 21 21:25:56 rewitt: IIUC, you want the kernel source in the SDK so you can compile kernel modules against it, just like you would have the -dev package for a target library in the SDK Oct 21 21:29:44 * derRichard has kernel-devsrc in TOOLCHAIN_TARGET_TASK Oct 21 21:42:58 In most cases with the kernel I would just prefer to rebuild everything and not get the target modules/etc, but I suppose if you were just trying to debug using kgdb or the like then it would be an extra step. Oct 21 21:44:50 rebuild everything takes too long. rebuilding a single kernel module is a matter of seconds Oct 21 21:45:08 derRichard: testing all three for all architectures under qemu isn't something we have the time budget for :( Oct 21 21:45:44 RP: sad to hear :-( Oct 21 21:47:51 Would a variable that allows for including specific directories work? i.e. if you have a fork of the kernel with some unique layout KERNEL_DEVSRC_EXTRADIRS = "path/from/src/root/to/unique" Oct 21 21:49:15 rewitt: xenomai adds a *lot* directories :( Oct 21 21:49:49 right now i have bbappend file with a do_install_apppend(): cd ${S} && cp -a --parents $(find -type d -name "xenomai") $kerneldir/build Oct 21 21:49:56 this seems to work. but needs more testing Oct 21 21:50:48 why don't you just copy every direcory what contains either a .{chS} file? Oct 21 21:51:49 hi there! I'm looking for best practices / docs rgd structuring git repos, dir structure and conf files for a custom poky based distro (minimal/full flavors) for 3 different devices, so 2x3 images in total with different BSP layers needed for the devices and distro flavors. I am confused (beginner) - pls help .. Oct 21 21:51:57 derRichard: In general you don't need the .c or .S files if I remember correctly and we were trying to make something of a manageable size Oct 21 21:52:48 derRichard: I guess we thought copying all the source was easy enough for someone to write their own recipe for, its the subset which is harder Oct 21 21:53:07 * RP thinks https://autobuilder.yocto.io/pub/non-release/ might be improved on where it was Oct 21 21:54:13 ah right, just headers should be ok too. Oct 21 22:12:53 tgamblin: I don't mean that reply to be negative, more that its a really hard problem to show information Oct 21 22:16:22 RP: just cloned android tree, it took 4hrs just to clone :) I dont know how long to build Oct 21 22:19:15 RP: No worries, just wanted to make sure it was submitted for wider review before we started carrying it internally :) Oct 21 22:19:33 On the off chance it was useful to others, that is Oct 21 22:20:05 tgamblin: I'm torn on these things in core :/ Oct 21 22:20:32 some data can be better than non, equally people do tend to misunderstand it if we show it :/ Oct 21 22:21:47 derRichard: I hope you have kernel dev manual handy especially https://www.yoctoproject.org/docs/3.1/kernel-dev/kernel-dev.html#working-with-out-of-tree-modules Oct 21 22:23:44 khem: yes, we build all modules also in yocto Oct 21 22:24:24 RP: Yeah, that was a concern I had when putting it together. It's clear to us what the new version would mean, but I can see how it would mislead people Oct 21 22:25:24 tgamblin: I was asked for a long time to add a general sstate usage summary. I did eventually. I've not heard anyone say its actually what they wanted as they don't understand it :/ Oct 21 22:26:09 yeah.. i used to use buildstats-summary as it was better than nothing, but i ended up disabling it for mel as it just didn't tell the whole story, so it was misleading Oct 21 22:35:48 I think in both Buildroot and Yocto it's possible to set if one would like recipes to create shared or static libraries. So this decision seems to be done by the package maintainer, not the developer. Why? I think most people set this in their project. That doesn't mean it's right, though. Oct 21 22:35:56 RP: sounds good. I will respond to the mailing list too just for completeness Oct 21 22:42:08 tgamblin: it is good to share these things as it gets others thinking and if a majority would find something useful... Oct 22 07:55:15 yo dudX Oct 22 08:26:08 Why is ${PN} "PN"? Shouldn't it be "RN" = "Recipe Name"? Oct 22 08:32:07 manuel1985: technically a recipe can provide an arbitralily chose PN by setting it. its a best practise to derive PN automaticall from the recipes name, but its not mandatory. Oct 22 08:38:00 Got it, thank you. Oct 22 10:25:24 manuel1985: the problem with a project having 20 years of history is that some things don't make a lot of sense :) Oct 22 10:28:41 well, 15 Oct 22 10:33:40 i thought it was only 10? so only 8 years left until drinking age! Oct 22 10:36:47 that's yocto Oct 22 10:36:58 bitbake/oe are older, obviously Oct 22 10:38:10 first bitbake commit from kergoth goes to 2003 ".oe file parser function" :) Oct 22 10:38:33 obviously. but we're in the wrong channel then. read: we will be drinking in #oe years before we can start here. Oct 22 10:51:50 rburton: Thought so :) Oct 22 10:52:24 really, PN is package name as it's used for the deployed packages. BPN *is* recipe name. Oct 22 10:53:16 except for native and nativesdk where BPN *isn't* the recipe name :) Oct 22 10:53:45 gar, true Oct 22 10:55:43 morning Oct 22 10:59:36 hrw: o/ **** ENDING LOGGING AT Thu Oct 22 10:59:57 2020