**** BEGIN LOGGING AT Tue Mar 30 02:59:56 2021 Mar 30 07:41:54 good morning Mar 30 07:42:46 yo dudX & mckoan Mar 30 08:10:15 Hi there, my name is Eren Atas from Liberty Global. We are having a problem while trying to reach to https://eula-downloads.yoctoproject.org/index.php since yesterday at ~6.30PM CEST (Amsterdam time) and the issue still persists. I checked the IRC logs but I could not see anything. Mar 30 08:10:16 When I try to ping it, it gives ping: unknown host eula-downloads.yoctoproject.org Mar 30 08:10:16 Does anybody know about the issue? Mar 30 08:10:17 Cheers! Mar 30 08:22:48 erenatas: it has been mentioned on the mailing list already. there is no ETA on it being back at the moment. Mar 30 08:27:08 LetoThe2nd thank you for letting me know! Mar 30 08:30:01 what triggers the connection there anyways? i can't find a reference to the eula downloads in poky ad hoc. Mar 30 08:38:01 Good Morning! Could you tell me what I need to add to IMAGE_INSTALL_append for having the shell command "nice" in my yocto distro? Mar 30 08:39:25 or a map of linux program to yocto package, respectively Mar 30 08:42:40 ... I mean a link to such map Mar 30 08:42:47 ...would also be good Mar 30 08:44:15 fbre: you could try oe-pkgdata-util which can search for a package that provides a certain path, for example /bin/nide or /usr/bin/nice Mar 30 08:44:55 fbre: but there is no such map, at least not at the moment because it would be quite complicated and oriject dependent. Mar 30 08:45:03 *project. Mar 30 08:46:28 fbre: good guesses for sure are usually coreutils or util-linux Mar 30 08:47:35 Thanx erenatas and LetoThe2nd. Is there a link to find out which programs are in coreutils and util-linux Mar 30 08:47:45 ? Mar 30 08:48:24 LetoThe2nd the URL is in CONNECTIVITY_CHECK_URIS on some of our bitbake builds. Mar 30 08:48:25 ERROR: OE-core's config sanity checker detected a potential misconfiguration. Mar 30 08:48:28 fbre: oe-pkgdata-util. and the recipes, which you can find through the layerindex. Mar 30 08:49:10 erenatas: can you tell me where they get injected? e.g. look at bitbake -e of those variables and find where they get set? Mar 30 08:49:17 LetoThe2nd: ok thanx Mar 30 08:52:43 LetoThe2nd the link is in meta-rdk/conf/distro/rdk.conf Mar 30 08:52:44 We will temporarily override the value until it gets back up Mar 30 08:53:28 Thank you very much for your help! Mar 30 08:54:15 Looks like it is in coreutils. It's mentioned here anyway https://git.yoctoproject.org/cgit.cgi/poky/plain/meta/recipes-core/coreutils/coreutils_8.32.bb Mar 30 08:54:58 erenatas: ah. i grepped poky as well as meta-openembedded and wondered. Mar 30 08:56:19 So I should add       IMAGE_INSTALL_append += " coreutils"    in my distro.conf, right? Mar 30 08:59:58 fbre: are you really asking now how to add a package to your build now? Mar 30 09:02:55 LetoThe2nd: I'm asking because I'm not 100% sure because sometimes it's in IMAGE_INSTALL_append, sometimes in EXTRA_IMAGE_FEATURES_append. Mar 30 09:03:16 LetoThe2nd: any chance eula-downloads.yoctoproject.org is actually down? Mar 30 09:03:18 * qschulz chuckles Mar 30 09:03:32 qschulz: hrhrhr Mar 30 09:03:58 ...or some other variables I not remember right now ;) Mar 30 09:04:09 fbre: nope. certainly not. the one is about packages, the other is about features. so if this confuses you, i *strongly* suggest to start watching/reading the introductory stuff. Mar 30 09:05:01 fbre: hint: IMAGE_INSTALL is a good start, _append maybe too, distro is usually wrong in that context. Mar 30 09:06:43 I've read the docs several times but it wasn't exactly clear when it's a package and when a feature. Mar 30 09:08:18 My feeling is coreutils is a package :) Mar 30 09:09:30 ...that I want to add to my distribution. That's why I guess IMAGE_INSTALL_append +" coreutils" has a chance of 87,346% to be right ;-) Mar 30 09:10:08 += Mar 30 09:12:55 fbre: i seriously doubt the "read several times". and in that case, it probably should be DISTRO_EXTRA_RRECOMMENDS or DISTRO_EXTRA_RDEPENDS, it... depends a bit :) Mar 30 13:41:40 halstead: once you are awake, can you maybe have a look at the eula-downloads thing? Mar 30 14:06:00 Hello, I'm looking for the script yocto-compat-layer.py which is named here:  https://www.yoctoproject.org/ecosystem/branding/compatible-registration/ Mar 30 14:09:06 Hello, Im trying to run GUI application in core-image-minimal (qemu). How can I achieve that? I added some qt applications but they fail to run on xcb. What is the proper way to do that? Do you have any materials? Mar 30 14:09:58 i dont need window managers etc, just single app running Mar 30 14:12:43 Hi, Does a toto.bb recipe with PACKAGES=+${pn}-tata and BBCLASSEXTEND="native" should be able to generate a toto-tata-native pakage ? Mar 30 14:15:22 clementp[m]: seems reasonable to me (provided it\s ${PN} and not ${pn}) Mar 30 14:17:52 qschulz: it's to me but i'm trying to do nodejs-npm-native from http://cgit.openembedded.org/meta-openembedded/tree/meta-oe/recipes-devtools/nodejs/nodejs_12.20.2.bb Mar 30 14:17:57 Don't see where is my error Mar 30 14:18:01 or error in the recipe Mar 30 14:18:28 got an ERROR: Nothing PROVIDES 'nodejs-npm-native' Mar 30 14:19:43 clementp[m]: because depending doesn't work on packages, but on recipes. Mar 30 14:20:28 LetoThe2nd: you're stealing my support tickets now :o? Mar 30 14:20:48 qschulz: i try to be faster than you at least one time in march '21 Mar 30 14:21:31 Thanks so my recipe depends on nodejs-native but how do I say i need the NPM binary ? Mar 30 14:22:10 clementp[m]: think again. if you depend on a complete recipe, what will you probably get? Mar 30 14:22:39 clementp[m]: recipes produce packages. Recipes are for build time, packages are for runtime. DEPENDS is for build time, RDEPENDS ir for runtime. Mar 30 14:23:33 clementp[m]: if nodejs-native builds npm package, it is likely DEPENDS'ing on nodejs-native will bring the npm binary in the native sysroot of your other recipe. Not guaranteed, because it depends on SYSROOT_DIRS_NATIVE but very likely Mar 30 14:24:56 although... now that I'm thinkin about it... what do you want to do? don't we have an npm class to inherit for npm recipes? Mar 30 14:25:35 qschulz: Thanks got it, I'm trying to do a "deep" nodejs package => no nodejs on the target Mar 30 14:25:45 just building a .html/js to have static web files Mar 30 14:26:38 I saw someone calling it "deep native" but not sure it's the good term Mar 30 14:26:54 I would like to npm install / npm run build and then install the outputs on /var/www/ Mar 30 14:27:10 clementp[m]: I helped as much as I could, nodejs is very far from my knowledge base :) maybe other can help, good luck :) Mar 30 14:28:51 *sigh* Mar 30 14:29:11 if you do DEPENDS = "nodejs-native", then you should totally be able to run npm. Mar 30 14:30:01 qschulz: thanks for your help! I have rerun and it's working.... Mar 30 14:30:20 Not sure what i have done but I did a clean and indeed the npm is working -_- Mar 30 14:31:11 clementp[m]: everybody experiences brain fart sometimes, don't worry :) Mar 30 14:31:47 Thanks also LetoThe2nd, I didn't know the difference between package/recipe in depends/rdepends Mar 30 14:32:10 At the beginning Yocto was very complicated compare to buildroot but when you start to understand each feature it's really nice :) Mar 30 14:49:35 do_kernel_metadata does not tell which bsp scc file it's using ? Mar 30 14:51:38 I'm adding an additional kmeta tree with a voluntary error inside and it's not complaining about it, how can I see what it's really doing ? Mar 30 14:57:37 is it even possible to overlay my own BSP files from a separate tree over yocto-kernel-cache ? Please tell me it's not *necessary* to fork the repo :) Mar 30 14:58:28 building poppler with qt5 gives me this warning, AUTOMOC for target poppler-qt5: Could not find moc executable target Qt5::moc Mar 30 14:58:32 any suggestions? Mar 30 15:07:57 answering to my last question: file://bsp;type=kmeta;subdir=kernel-meta/ does the trick for kmeta merging Mar 30 15:08:52 (but still noone complains about my syntax-error BSP, so I guess I'm still missing something) Mar 30 15:09:56 tlwoerner: btw, I'm working on that rockchip kmeta Mar 30 15:10:43 first things first, understanding how kmeta is supposed to work :/ Mar 30 15:32:54 yann: Are you trying to add kmeta for meta-rockchip? We use kmeta internally, but we have our own repo for it Mar 30 15:34:01 Hello, Im trying to run GUI application in core-image-minimal (qemu). How can I achieve that? I added some qt applications but they fail to run on xcb. What is the proper way to do that? Do you have any materials? Mar 30 15:36:27 JPEW: right Mar 30 15:37:27 sigh, bitbake just hand if all networking is off inside a container. need to rethink about off line builds... Mar 30 15:37:35 /hand/hangs/ Mar 30 15:42:41 yann: i demonstrate this in the WIP work i pushed to github, did i forget to provide you with a link? Mar 30 15:42:53 gotcha, finally grasped how KMACHINE works, and spp now finds my .scc Mar 30 15:43:02 yann: in recipes-kernel/linux i create a rockchip-kmeta directory Mar 30 15:43:14 yann: and under there i put all my rockchip-specific things Mar 30 15:43:18 tlwoerner: yeah, but it only has a placeholder Mar 30 15:43:51 BinBamBim: seems like core-image-minimal only includes stuff required to boot and prompt you with a console. you could try core-image-sato which has x11. I think qt outputs xcb errors since core-image-minimal doesn't have x11 by default Mar 30 15:43:56 yann: ah right, because i didn't have it working yet, but i left the mechanism there Mar 30 15:44:25 yann: the next 3 needed things are: Mar 30 15:44:42 1. adding ${THISDIR} to FILESEXTRAPATHS_prepend Mar 30 15:44:58 2. adding file://rockchip-kmeta;type=kmeta;name=rockchip-kmeta;destsuffix=rockchip-kmeta to SRC_URI Mar 30 15:45:43 tlwoerner: I started to wonder whether we need RK_KERNEL_CONFIG_TYPE after all, AFAICT we could do what we want with LINUX_KERNEL_TYPE, or did I miss something ? Mar 30 15:45:50 BinBamBim: or core-image-x11 might be simpler than core-image-sato. in any cases, you need some graphical stack (x11, wayland, or check out qt eglfs) Mar 30 15:46:34 3. enabling settings with Mar 30 15:46:36 my plan has become: 1. get a kmeta to work for nanopi-m4 (including linux-yocto-tiny) 2. see if we still miss something Mar 30 15:47:38 BimBamBim: ^ sorry, misspelled your username. Mar 30 15:47:44 and if we need RK_KERNEL_CONFIG_TYPE, I have a 10-line patch sparing the need for anonymous python Mar 30 15:48:09 yann: LINUX_KERNEL_TYPE is what differentiates between the various linux-yocto flavours, i don't know if it's wise to overload someone else's variables (?) Mar 30 15:48:16 champagneg: but is it possible to ommit x11 and use just wayland or qteglfs on qemu? Mar 30 15:48:59 yann: personally i find the anonymous python easier to understand what it is doing and how it is going to work than using overrides Mar 30 15:49:13 tlwoerner: I don't think it's forbidden to add new values for LINUX_KERNEL_TYPE, but I'd like to challenge the idea that even that would be needed Mar 30 15:51:22 yann: in an ideal world we could forego the defconfig and just build up a random kernel config piece by piece, the same way we build images :-D Mar 30 15:51:30 tlwoerner: to me any short-but-still-clear declarative is more readable than 10 times the same amount of code - and when it brings extensibility when the code would have to be changed, the winner is clear to me :) Mar 30 15:52:00 BimBamBim: I would think so, there's core-image-weston for wayland. I think you'll need meta-qt5 or meta-qif you want to run a qt app. But weston has samples clients that should work correctly within core-image-weston Mar 30 15:52:24 s/meta-qif/meta-qt4/ Mar 30 15:52:38 tlwoerner: yeah, that's what I tried to do, but then we need to manually activate tons of essential options, which are already set in kmeta snippets Mar 30 15:53:15 anyway I shall have something to post today or tomorrow, that should provide a concreate basis for further discussion Mar 30 15:53:16 yann: oooh nice :-D Mar 30 15:53:43 yann: let me be clear. i think this is an awesome conversation to be having and i look forward to improvements in this area :-) Mar 30 15:54:06 tlwoerner: great :) Mar 30 15:54:26 Bruce has done a ton of work in this area and i'm really looking forward to leveraging it in a non-core BSP Mar 30 15:55:59 champagneg: I use meta-qt5 and its `cinematicexperience` example app, which works good, but only on vnc. I would like qemu to automatically open gui window when I run app, without the neccessity to run separate vnc client... Mar 30 16:00:22 BimBamBim: to launch the app right after boot, _something_ like a sysvinit init script or a systemd unit would need to start your app when the system boots. It maybe also be configurable at the compositor (weston for example) level to launch something right after it starts Mar 30 16:01:25 but well, I can just manually run it by commandline, right...? Mar 30 16:01:36 yes Mar 30 16:02:14 im one step behind, I cant run the app ;) for xcb it fail and I dont have other working backends Mar 30 16:04:53 if you want to try the wayland route, you could try to build an image based on core-image-weston add `qtwayland` Mar 30 16:06:14 LetoThe2nd, Yes I took it offline as recommended at https://bugzilla.yoctoproject.org/show_bug.cgi?id=14174. I suppose we need to find a way to take the site offline and keep those builds working. Mar 30 16:06:32 then, i think, you'll be able to launch a qt app for wayland with `myapp -qpa wayland` Mar 30 16:10:41 halstead: heh interesting. thanks for the heads up. Mar 30 16:12:29 halstead: care to also drop a note on the ml question? Mar 30 16:13:01 LetoThe2nd, I went ahead and put the site back online. The downloads are still unavailable as they have been for over 6 months. Mar 30 16:13:27 halstead: funny. Mar 30 16:14:00 halstead: lets see what happens: https://twitter.com/TheYoctoJester/status/1376915161861218305 Mar 30 16:14:10 LetoThe2nd, I see the mailing list. I suppose I'll keep the site up with a notice to keep those checks working. Mar 30 16:14:33 halstead: maybe take it to the call. (i gotta run now, sorry) Mar 30 16:14:47 * halstead nods. Mar 30 16:21:09 so I added EXTRA_OECMAKE = "-DOE_QMAKE_PATH_EXTERNAL_HOST_BINS='${OE_QMAKE_PATH_EXTERNAL_HOST_BINS}'" to my poppler bbappend. Should this just be patched into the actual recipe? I'm not sure if I'm the only one seeing the poppler-qt5 issue. Mar 30 16:21:56 error that I was seeing: AUTOMOC for target poppler-qt5: Could not find moc executable target Qt5::moc Mar 30 16:25:01 do .conf files HAVE to be under a conf/machine directory? or can i place them in another directory in my custom layer? Mar 30 16:31:50 why would you put them somewhere different? every other machine layer puts them in conf/machine Mar 30 16:32:06 but: Mar 30 16:32:08 meta/conf/bitbake.conf:include conf/machine/${MACHINE}.conf Mar 30 16:32:15 ^^^ that is where they go Mar 30 16:47:58 ok, thanks rburton. Mar 30 17:21:00 there will be a virtual Yocto Project Summit last week of May: https://www.yoctoproject.org/yocto-project-virtual-summit-2021/ Mar 30 17:21:17 the call for papers is now open: https://pretalx.com/yocto-project-summit-2021/cfp Mar 30 17:21:23 registration is not open just yet Mar 30 17:22:57 tlwoerner: \o/ Mar 30 17:24:45 mckoan: thanks Mar 30 17:26:25 I'm puzzled by a kernel config issue: I have tons of kconfig options turned to "m", with do_kernel_configcheck telling me my config does request "y" but the "IC2=m" dependency seems to cause the changes. However, I can't see any file setting CONFIG_I2C=m, and do_kernel_configcheck does not even report a mismatch. Mar 30 17:27:41 Looks like I'm back to square one: how do I see what kmeta files are getting used ? Looks like we have some missing information here, right ? Mar 30 17:42:15 damned, it's just really not finding a BSP definition 'cause I did not write one for "tiny". I'll send a minimal patch to at least get aware of this, though I'm really not happy with the information level I can get Mar 30 17:45:53 Who attends the Ubuntu vs. Yocto stuff? Mar 30 17:46:47 "YP just supports a few archs" - Ubuntu core supports x86 and arm ;) Mar 30 17:48:13 RobertBerger: where is that? Mar 30 17:48:46 @rburton: Hehe it's about to finish now. We are at QA Mar 30 17:49:18 @rburton: They recorded it. As soon as the link is out I'll let you know. Mar 30 17:49:31 Should be on youtube, I guess, eventually Mar 30 17:50:13 RobertBerger: what event is that ? Mar 30 17:50:35 It's some UB20 promotional stuff. Mar 30 17:50:35 rburton: https://www.brighttalk.com/webcast/6793/470692?utm_campaign=communication_reminder_starting_now_registrants&utm_medium=email&utm_source=brighttalk-transact&utm_content=button Mar 30 17:50:46 @Jama thanks Mar 30 17:51:14 "This webinar will help you make this decision" "presented by Canonical" Mar 30 17:51:23 not convinced they'll be helping you make a decision Mar 30 17:51:49 hehe - they really try to help to convince ;) Mar 30 17:51:59 short version is, that you need snaps for everything, because they are great and solve everything automatically Mar 30 17:52:03 I asked a few questions on the chat there, but so far they were ignored ;) Mar 30 17:52:16 @#$@# ubu and their snaps Mar 30 17:52:22 I think I should start counting the word snap Mar 30 17:52:47 Reminds me a bit of an angry crocodile - snap snap ;) Mar 30 17:52:51 the only place they get traction on that, is when they talk to business folks / execs Mar 30 17:53:00 * zeddii is living that in his corporate world Mar 30 17:53:06 so I speak that statement from experience. Mar 30 17:53:09 if you want more unfair build system comparisons, then see Guix talk https://archive.fosdem.org/2020/schedule/event/ggaaattyp/attachments/slides/3965/export/events/attachments/ggaaattyp/slides/3965/talk.pdf Mar 30 17:53:36 They funny thing is, that one of the people there is former Wind River who knows how YP/OE works Mar 30 17:53:52 But I believe he forgot many things ;) Mar 30 17:54:15 RobertBerger, what was the name ? there were lots of people at Wind River that only claimed to know OE/YP Mar 30 17:54:27 Diego Bruno - Italy Mar 30 17:54:33 hahahahah. 'er no. Mar 30 17:54:41 I know him in person. Mar 30 17:54:46 me too :D Mar 30 17:54:48 Played piano with him ;) Mar 30 17:55:10 JaMa: I had missed that talk @ FOSDEM when I was there. interesting Mar 30 17:56:31 BTW the example of building a simple system is "bitbake world" ;) Mar 30 17:56:45 * zeddii does that every day! Mar 30 17:57:04 RobertBerger: this has the merit of being simple :D Mar 30 17:57:34 Hehe - you need to see the presentation. That was in the context of "it takes time to build" Mar 30 17:57:37 * zeddii goes back to his meeting. the mention of snaps brought me up from the dead. Mar 30 17:58:02 @zeddi - codeword "crocodile" Mar 30 17:59:17 one of this recent build system talks also claimed that Yocto still need python2 and that it doesn't support many reproducible builds, pity these people don't update their slides more than once in 2 years.. Mar 30 18:00:08 i should do a yocto vs guix presentation and slide 2 will be "uses scheme, not continuing" Mar 30 18:00:51 FML, that I didn't notice. Mar 30 18:01:01 rburton: :) Mar 30 18:01:11 now you brought me back from the dead again. scheme. I had to code in it for two years. nearly died. Mar 30 18:01:14 like any good gnu project, scheme is the preferred language for config file Mar 30 18:01:43 https://guix.gnu.org/cookbook/en/guix-cookbook.html#A-_0060_0060Hello-World_0027_0027-package Mar 30 18:01:44 when I talk buildsystems, I stick to the advice of don't mention others, don't compare. Mar 30 18:01:50 'As you can see, most of it is rather straightforward.' Mar 30 18:02:13 dear lord Mar 30 18:02:27 that looks like my emacs config I've carried around for 10 years and am afraid to touch. Mar 30 18:02:34 oh i solved that problem Mar 30 18:02:38 vscode ftw! Mar 30 18:03:07 * zeddii looks for a rock Mar 30 18:04:03 "GNU Hello is a dummy project that serves as an idiomatic example for packaging." Mar 30 18:04:54 guix does have some neat ideas and its got the advantage of not having over a decade of cruft, but jesus scheme Mar 30 18:05:22 "he spoke fluent, idiomatic guixish" Mar 30 18:06:37 rburton: you could have kept that to yourself ;-) Mar 30 18:18:41 scheme is awesome, when applied to the right domain. For DSSSL it was really good. Just some people fancied that programming using XML instead was kind of neat... Mar 30 18:18:54 * yann is still grieving after all these years... Mar 30 18:30:31 ugh, this stupid "via lists." from address thing when i send-email from my gmail is irritating, anyone know how to resolve that? Mar 30 18:31:46 bluelightning: just sent a buildhistory patch but forgot to cc you, sorry about that Mar 30 18:32:08 ugh, now to dig through my backlog of stuff to submit.. Mar 30 18:32:30 also, somehow i didn't know vardepvalueexclude existed until 5 minutes ago Mar 30 18:32:43 that's handy, better than override the entirety of vardepvalue Mar 30 18:38:09 Hi, I am looking to get some training with board bringup, u-boot and kernel configuration for Xilinx FPGA boards. Ideally this would be 1 on 1 sessions via zoom. Does anybody have recommendations for who can offer these services? Thanks! Mar 30 18:40:16 yann: patch 1/3 is definitely going in sooner rather than later :-) Mar 30 18:56:19 felix_inst: I am not of much help, but since Xilinx maintain meta-xilinx, I assume they offer something like that. Or check: https://elinux.org/Experts for embbeded soft. consultants Mar 30 19:09:47 champagneg, Thanks, that list looks promising! Mar 30 19:16:46 kergoth: is that for your gmail address? I know for me, my list mail comes out ok and just the copies back to me are modified Mar 30 19:20:26 ah, good point, yeah, i think i'm seeing that same behavior Mar 30 19:26:56 It would be nice to get only one copy of every e-mail I send to the list :) Mar 30 19:32:12 +1 Mar 30 19:33:09 moar mailz! Mar 30 19:35:14 hello, I am on master and am getting an error running 'yocto-check-layer' - my layer doesn't ask for 'projucer' but somehow it's being required so then... x11 dist feature is required... hoping for some guidance here .. https://gist.github.com/rpcme/be464a77fdccafa3d995b9a68be77f9e Mar 30 19:36:44 rpcme: send patch for meta-openembedded/meta-multimedia/recipes-multimedia/packagegroups/packagegroup-meta-multimedia.bb to include projucer only with x11 in distro_features Mar 30 19:37:18 ok thanks whew I thought maybe I was doing something awkward ;) Mar 30 19:37:51 it's just too many combinations and not all of them are properly tested Mar 30 19:38:00 should I send that patch as I'm already in that file? Mar 30 19:40:21 rpcme: https://lists.openembedded.org/g/openembedded-devel/message/90461 Mar 30 19:40:45 yes I just tweaked it manually Mar 30 19:40:55 if you already did it that's all good :) Mar 30 19:41:48 yes checked the patch and I did the exact same thing. Mar 30 19:50:00 JaMa: you should only get one and only if you ask for it. Have you multiple addresses subscribed? Mar 30 19:50:59 RP: I see one as my sent e-mail (sending from gmail) and then I receive another one from the list (with this via lists) Mar 30 19:51:54 JaMa: on all lists? oe or yp? Mar 30 19:52:44 JaMa: you could check your settings in groups.io as I suspect some things are configurable. I definitely only get the one (with the via bit added) Mar 30 19:52:45 oe-devel and oe-core for sure (since the migration from mailman), not sure about yp list as I don't send e-mails there so often Mar 30 19:52:56 I can confirm, all of them. Mar 30 19:53:19 I send an email. wait. mark it read. wait more. mark the next copy read as well. Mar 30 19:53:26 RP: but you're not sending from gmail, right? Mar 30 19:54:35 RP: and if I disable groups.io to send copy of my own e-mail, then when sending e.g. PATCHv2 I won't see the Message-Id for Reply-To: in my client Mar 30 19:55:10 JaMa: it is from a google account Mar 30 19:55:37 JaMa: hmm. I had a glance at the settings and I can't spot anything. The echo has to come from somewhere :/ Mar 30 19:56:54 the first one is stored by gmail I believe, see https://paste.ubuntu.com/p/nm5wj3nQPD/ Mar 30 19:57:36 JaMa: is this because you're viewing sent items as well as incomming? Mar 30 19:57:55 JaMa: that must be something gmail and settings... Mar 30 19:58:05 I have the same behaviour: one original-looking apparently saved by gmail, and one furstrating "via" on which the replies are done :) Mar 30 19:58:20 the echo (with via lists) is the one from ML https://paste.ubuntu.com/p/Xw7YMBR5k5/ so it looks like issue on gmail side (and maybe wasn't happening with mailman, just because mailman didn't send my own back by default) Mar 30 20:00:30 and in gmail rules I have just this: Mar 30 20:00:33 Matches: list:"yoctoproject.org" Mar 30 20:00:34 Do this: Skip Inbox, Apply label "OE_core" Mar 30 20:00:46 I am hoping that someone can point me to documentation for the 'software layer signature check' in 'yocto-check-layer' so hopefully I can school myself enough to overcome the failures Mar 30 20:01:00 JaMa: absolutely no rules on my side, fwiw Mar 30 20:01:04 it's possible that it does apply this rule also for e-mails being sent to list Mar 30 20:01:20 yann: ah, thanks, so not the rules Mar 30 20:05:09 in tune-riscv.inc we see (e.g.) TUNE_ARCH_tune-riscv64 = "riscv64" Mar 30 20:05:30 then in arch-riscv.inc there is TUNE_ARCH = "${TUNE_ARCH_tune-${DEFAULTTUNE}}" Mar 30 20:05:51 what purpose does the TUNE_ARCH serve? Mar 30 20:07:19 grep meta/conf and meta/classes for it, it's used in a number of variables Mar 30 20:07:26 (where DEFAULTTUNE = "riscv64") Mar 30 20:09:24 strange, when running yocto-check-layer to test the issue rpcme reported I've noticed that it now fails due to AVAILABLE_LICENSES being extended by meta-multimedia layer, that's a bug right? There were some changes in this are recently and extra LICENSE_PATH shouldn't change the signatures (at least unless the value is being used by that recipe), see https://paste.ubuntu.com/p/469g4z8C9n/ Mar 30 20:09:30 in that particular case the TUNE_ARCH value is the same as the tune, but that's not always the case Mar 30 20:11:07 ack kergoth Mar 30 20:24:27 I wonder if this is related to duplicate e-mails https://groups.io/helpcenter/faq/1/group-member-faq/q-why-do-i-get-a-spoofing my e-mails have: Mar 30 20:24:30 From: "Martin Jansa via lists.openembedded.org" Mar 30 20:24:33 X-Google-Original-From: Martin Jansa Mar 30 20:24:54 while I don's see this in RPs e-mails as he isn't sending from @gmail.com even when it's an gmail account Mar 30 20:28:55 I don't see how to disable this in groups.io setting (other than writting them an e-mail) Mar 30 20:41:41 I am hoping someone can point me to documentation to understand: 'AssertionError: Adding layer XXX changed signatures.' in yocto-check-layer.  I also have another strange error with the check signature routine saying ntp rdepends cannot be found but meta-networking is defined as a dependency Mar 30 20:42:34 but the first error is just nebulous, I have no idea where to even start hunting down the problem and I can't find documentation on this check.  I'm sure it's out there, but I've been hunting for the last hour or so. Mar 30 20:43:58 rpcme: the signature change means exactly that, just including the layer changed something. if it's a bsp layer, make sure everything in a bbappend is done using a MACHINE override, not unconditional. same for distro and distro layer. Mar 30 20:44:10 the output of yocto-check-layer will show exactly what changed.. Mar 30 20:44:55 i'm looking at this section of code in siteinfo.bbclass: http://paste.ubuntu.com/p/FvGHdkWbfR/ Mar 30 20:44:58 ok... so.. I think I'm getting it now... I can't modify kernel then right? Mar 30 20:45:32 you can modify anything, just use an appropriate override so it only happens when MACHINE is set to your machines Mar 30 20:45:46 we dont want anythign to change solely based on layer inclusion, generally Mar 30 20:45:56 we want to require an opt-in to the changes, onew ay or another Mar 30 20:46:04 ok this is not a BSP - it's an application layer so I must be doing something illedgal Mar 30 20:46:08 *illegal Mar 30 20:46:27 it is complaining on linux-yocto Mar 30 20:46:28 is this formating a "call" as a string (in call), as "extra_site_info_call_function" as the function, rand then calling via bb.utils.better_eval(call, locs)? Mar 30 20:46:39 software layers are fuzzier, and yocto-check-layer has an argument to control it. that is, it's less strict, but still ideally should require an opt in Mar 30 20:46:54 for example, meta-virtualization is a software layer but everything is based on the `virtualization` distro feature Mar 30 20:47:11 in other cases i've operated based on a machine feature, or a completely different variable Mar 30 20:47:15 it's really case by case for software layers Mar 30 20:48:24 this is really useful.  I think I have reading to do on the options.  thank you Mar 30 20:48:28 for years mentor's meta-mentor-staging layer unconditionally altered the build, knowingly not complying with this, as that layer specifically exists to fix bugs, and is intended to only stage changes that i'm submitting upstream but haven't gotten to yet Mar 30 20:48:40 but recnetly i switched it to obey a new distro feature, and made our mel distro enable it Mar 30 20:49:13 I want it to be clean and "not surprising" to users so better to make it an opt-in Mar 30 20:49:22 thank you - I have homework :) Mar 30 20:50:19 rpcme: you said you're altering a kernel build, but it's not a bsp layer, so you'll have to consider whose responsibility this is -- is it a machine feature or a distro one? for example, we have something like that where we enable kernel options for virtualization, it's a bsp recipe, but that's arguably a distro feature, so we do it based on the latter, not the former Mar 30 20:50:43 oe-core has a bbclass to ease changes based on distro features, distrooverrides.bbclass Mar 30 20:51:31 I also have https://github.com/MentorEmbedded/meta-mentor/blob/master/meta-mentor-common/classes/feature_overrides.bbclass to add an override for other conditions, like machine or combined features. with the right setup in my layer.conf, i can then go SRC_URI_append_feature-my-machine-feature = " foo" rather than mucking about with inline python Mar 30 20:51:53 ping me if you have a need for that, it sounds more likely you'll use a distro feature, but that's conjecture, so i'll put it out there :) Mar 30 20:53:10 rpcme: nicely done in actually thinking about this stuff, that's more than most layer maintainers do. giving thought to usability, yocto compliance, and convention shows an impressive amount of care Mar 30 20:55:15 machine or distro feature based overrides could be useful for the baremetal stuff I'm doing Mar 30 21:00:01 yes the problem is one of the recipes requires kernel features that were not default in older kernels (cgroups, namespaces)... I might even be able to pluck those modifications Mar 30 21:00:22 probably zeus or dunfell onward... will need to check Mar 30 21:31:28 Anyone know why I get this on the latest master: "libgomp: Thread creation failed: Resource temporarily unavailable" Mar 30 21:31:34 It's been happening for a few weeks Mar 30 21:35:54 JPEW: no, but curious if you figure it out. master does seem to be a bit more memory sensitive all of a sudden Mar 30 21:43:38 is meta-boot2qt not being upgraded to newer releases? Latest I see is zeus Mar 30 21:46:17 fray: https://github.com/MentorEmbedded/meta-mentor/blob/master/meta-mentor-common/classes/feature_overrides.bbclass#L1-L9 -- added usage example. incredibly simple implementation, but i like the flexibility vs distrooverrides.bbclass Mar 30 21:52:46 Having a hard time figuring out how to do that with kernel changes.  I think I attempted this before - maybe a year or so ago without success. Mar 30 21:53:02 probably need to sleep on it. Mar 30 21:57:00 regarding the bb.utils.better_eval(call, locs), so do i basically define a function in my own layer taking arguments (archinfo, osinfo, targetinfo, d) where archinfo, osinfo, and targetinfo are the dict's and should be extended with my cpu/arch? Mar 30 21:58:42 yates: def my_datafunc(archinfo, osinfo, targetinfo, d): Mar 30 21:58:51 then SITEINFO_EXTRA_DATAFUNCS += "my_datafunc" Mar 30 21:58:58 right Mar 30 21:59:09 thanks Mar 30 21:59:11 np Mar 30 21:59:35 so short answer: yes :) Mar 30 22:00:10 i'll take any answers you'll give me! Mar 30 22:00:23 short, long, fuzzy, wrong, .... Mar 30 22:01:36 is there a special place my_datafunc() should reside? Mar 30 22:01:49 does it have to be under /conf? Mar 30 22:01:51 (e.g.) Mar 30 22:02:21 or in layer.conf itself? Mar 30 22:03:45 in a conf file or bbclass file? Mar 30 22:04:22 yates: you can't define a function in a conf file at all, syntax limitation Mar 30 22:04:41 yates: i'd suggest createing a class for it and adding that class to your INHERIT in an appropriate config file Mar 30 22:05:47 good info! Mar 30 22:06:40 INHERIT will ensure it's globally inherited, so can affect all classes. you could also put it in a .inc under conf/ and include that from a config file, but i think the class is cleaner. regardless it needs to be global to ensure its available in every context where siteinfo is available Mar 30 22:06:56 er, affect all recipes Mar 30 22:07:55 gotcha Mar 30 22:51:22 RP: rpmdeps isn't respecting BB_PARALLEL_MAKE and is instead falling back to some automatic detection mechanism? Mar 30 22:51:25 maybe Mar 30 22:57:28 JPEW: i used to get those all the time Mar 30 22:57:43 (resource temporarily unavailable) Mar 30 22:58:07 JPEW: entirely possible Mar 30 22:59:01 at the time the thinking was it was related to pixz Mar 30 22:59:15 (or pigz) ? Mar 30 22:59:45 I'm wondering if we need to be setting OMP_NUM_THREADS=${BB_PARALLEL_MAKE} to keep libgomp from going crazy Mar 30 22:59:47 and yes, there was that same component of something not respecting -j or parallel or whatever Mar 30 23:03:58 * JPEW runs a test Mar 30 23:12:41 Does anybody know of a source that talks about generating a minimal device tree? I am looking for what one typically needs at a bare minimum. Zynq / ZynqMP are the platforms I am looking for Mar 30 23:36:31 JPEW: you mean BB_NUMBER_THREADS? Mar 30 23:45:58 RP: Yes.... and it worked (sort of) Mar 30 23:47:07 I got a bunch of signature changed errors when I did `export OMP_NUM_THREADS = "${BB_NUMBER_THREADS}"`, but when I did a test with just `export OMP_NUM_THREADS = "10"` it passed Mar 30 23:48:21 JPEW: you would do. Take a look at the places that BB_NUMBER_THREADS is referenced in bitbake.conf Mar 30 23:48:57 JPEW: it'll need some work with task hashes if you have to export it Mar 30 23:49:44 JPEW: maybe fixing it's value with a vardepvalue would work best Mar 31 00:03:00 oh my... finally passed yocto-check-layer.  It is much better now though, seriously that was a good exercise Mar 31 00:07:26 rpcme: congrats Mar 31 00:08:54 kergoth: thanks for the buildhistory fix Mar 31 00:08:59 np Mar 31 00:20:11 RP: should dummy-sdk-package besides RPROVIDES also set PROVIDES? seeting WARNING: nativesdk-texinfo-6.7-r0 do_package_qa: QA Issue: nativesdk-texinfo rdepends on nativesdk-perl, but it isn't a build dependency, missing nativesdk-perl in DEPENDS or PACKAGECONFIG? [build-deps] Mar 31 01:16:47 RP: Like: `OMP_NUM_THREADS[vardepvalue] = ""` ? Mar 31 01:17:37 RP: I'm pretty sure the variable has to be exported; we don't really know where some tool might be using native OpenMP Mar 31 01:19:25 JPEW: we have XZ_DEFAULTS where we control number of threads xz uses we could have something similar for openmp as well Mar 31 01:20:32 khem: OK, ya Mar 31 02:00:55 make -j $(nproc) ... what could go wrong? lol **** ENDING LOGGING AT Wed Mar 31 02:59:56 2021