**** BEGIN LOGGING AT Wed Jan 15 02:59:57 2020 Jan 15 07:02:23 angelo__: no, i won't be at fosdom, sorry. but lots of other cool folks and content to be found there! Jan 15 07:24:35 New news from stackoverflow: How to remove linker error Problem undefined reference to `VTimeGetActiveHorzWidth' Jan 15 07:27:43 Good morning Jan 15 07:28:39 I've got a linux-ti-staging_%.bbappend file which looks like this: FILESEXTRAPATHS_prepend := "${THISDIR}/files:" SRC_URI += "files://spi.cfg" Jan 15 07:29:24 So just add files/spi.cfg as source file. When building this package, I get this error: Jan 15 07:29:49 bb.data_smart.ExpansionError: Failure expanding variable SRCPV, expression was ${@bb.fetch2.get_srcrev(d)} which triggered exception NoMethodError: Could not find a fetcher which supports the URL: 'files://spi.cfg' Jan 15 07:30:10 ThomasD13: it's file:// Jan 15 07:30:22 arrrrgh! Thanks erbo :) Jan 15 07:30:33 No worries :) Jan 15 07:57:42 Morning all, I'm trying to create a custom image type ( with company crypto stuff) and i'm stuck when I want to copy rootfs from DEPLOY_IMAGE_DIR to an other folder. --> FATAL ERROR: Error reading file into data: Is a directorymkimage ||any idea to fix that ? Jan 15 09:34:21 I want to add kernel feature to my image (CONFIG_SECCOMP_FILTER) how to to that? Jan 15 09:39:24 nacknick: which kernel are you using? Jan 15 09:39:37 (more specifically, which recipe) Jan 15 09:50:01 qschulz: I'm not sure. It's for Raspberry PI 3. How should I check which version? Jan 15 09:58:30 nacknick: bitbake virtual/kernel -e | grep PREFERRED_PROVIDER_virtual/kernel Jan 15 09:58:50 nacknick: there might be a better way to know which one's used but I don't remember on top of my head Jan 15 10:07:55 nacknick: basically the answer depends. If that recipe has "inherit kernel-yocto" somehwere, then you can create a defconfig fragment (.scc or .cfg) and add it to SRC_URI, that should do it I think (never used that mechanism tbh) Jan 15 10:08:52 nacknick: if it does not, then it's recipe specific. Does it have a similar mechanism? usually if not, kernel recipes allow to pass a full defconfig to it in SRC_URI Jan 15 10:13:03 qschulz: PREFERRED_PROVIDER_virtual/kernel="linux-raspberrypi" Jan 15 10:15:29 nacknick: check in linux-raspberrypi*.bb Jan 15 10:15:55 from what I read from the master version of this recipe, there is a `require recipes-kernel/linux/linux-yocto.inc` and in this file you have inherit kernel-yocto Jan 15 10:16:12 so create a defconfig fragment which ends in .scc or .cfg and add it in SRC_URI Jan 15 10:16:18 (in a bbappend in your layer) Jan 15 10:17:09 qschulz: https://gist.github.com/Yatekii/2ff8be70835b820109bc9883cbb73b1d this is my receipe now. hope that's ok. the only thing I don't understand is why I need to set the PV variable and why to that strange value. I read the docs but it still doesn't make sens to me fully. Jan 15 10:18:39 Yatekii: LICENSE="CLOSED" => NOT true. As per https://github.com/rust-embedded/gpio-utils#license it's LICENSE="Apache-2.0 | MIT" Jan 15 10:19:24 and autorev + srcvpv is strange Jan 15 10:20:01 LetoThe2nd: that's what's asked by https://www.yoctoproject.org/docs/latest/ref-manual/ref-manual.html#var-AUTOREV Jan 15 10:20:19 https://www.yoctoproject.org/docs/3.1/dev-manual/dev-manual.html#automatically-incrementing-a-binary-package-revision-number Jan 15 10:20:28 qschulz: hmkay. looks strange to me, but hey. Jan 15 10:20:46 qschulz: I am aware. it was because the license fuckery was so annoying and now I am gonna use my private repo anyways, so I can decide the terms :) Jan 15 10:21:46 qschulz: yeah I read both of those, but somehow the exact internas are still not 100% clear to me. but I have a bleeding edge setup and a fixed tag one working :) Jan 15 10:22:06 now I am gonna try and use our closed git repo ... couldn't get it to work at home Jan 15 10:22:55 Yatekii: you don't decide what the license is. If you base your project on that code, you need to have a license which is compliant with the one from the original project Jan 15 10:23:01 anyway, lawyer stuff, you do you Jan 15 10:23:43 qschulz: yes, but I am not basing my code on the project :D Jan 15 10:23:46 Yatekii: I don't know exactly where PV is used. But, it's used in the sstate-cache, in the name of packages, in the path to the workdir etc... I guess it's so that you can differentiate between two versions of the same recipe and package Jan 15 10:23:56 I just used it for a very basic example :) I already replaced it with my private repo Jan 15 10:24:34 so if you don't set this, it's gonna overwrite the previous recipe at best, at worst fuck up some internals Jan 15 10:24:49 I cannot give more info as I don't have them :) Jan 15 10:25:13 in do_install, use install -d ${D}${bindir}/ instead of mkdir Jan 15 10:25:26 and your cp looks weird Jan 15 10:25:27 yeah :) Jan 15 10:25:34 qschulz: well there is no .deb file Jan 15 10:25:36 why's the file in that directory? Jan 15 10:25:39 so I can't use install Jan 15 10:25:48 because that's the rust build location Jan 15 10:25:54 it's standard for rust Jan 15 10:27:02 what do you mean you can't use install? Jan 15 10:27:19 tasks are executed on your host system Jan 15 10:27:22 qschulz: i guess there's evil magic hidden in the cargo class Jan 15 10:27:25 I'm pretty sure your distro has support for install Jan 15 10:27:31 it tells me it can't find an installation receipe (I mean that does apt install which ofc doesn't work for my rust binary) Jan 15 10:27:46 Yatekii: gah. no. look at other recipes! Jan 15 10:27:53 ? Jan 15 10:27:54 install -m 644 from to Jan 15 10:28:05 instead of cp Jan 15 10:28:26 LetoThe2nd: yes, install for cp and install -d for mkdir basically Jan 15 10:28:34 sorry for Yatekii ^ Jan 15 10:28:34 qschulz++ Jan 15 10:28:46 Yatekii: man install on your host and you'll see :) Jan 15 10:29:14 (I personnally didn't know what it was before using Yocto so no worries) Jan 15 10:29:29 I mean it's an apt thing, so yeah :) Jan 15 10:29:35 no, its no apt thing! Jan 15 10:29:40 or at least it calls apt internally Jan 15 10:29:44 perfect example: https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta-poky/recipes-core/tiny-init/tiny-init.bb#n23 Jan 15 10:30:07 go in a temp directory in your host system, and type install -d foo Jan 15 10:30:12 you'll have a foo directory Jan 15 10:30:15 *this* is no apt thing. i am absolutely sure. https://gist.github.com/Yatekii/2ff8be70835b820109bc9883cbb73b1d#file-raichu-core_git-bb-L14 Jan 15 10:30:33 so s/mkdir/install -d/g, and s/cp/install -m/g Jan 15 10:30:34 LetoThe2nd: I did not say that anywhere Jan 15 10:30:36 :) Jan 15 10:30:56 10:29 < Yatekii> I mean it's an apt thing, so yeah :) Jan 15 10:31:07 LetoThe2nd: the install command calls apt install. Jan 15 10:31:13 no Jan 15 10:31:16 yes Jan 15 10:31:20 I was debugging it yesterday Jan 15 10:31:33 we are totally talking about that very speific "do_install" in your recipe, and that you're doing it wrong. Jan 15 10:31:41 well anyhow Imma try this again Jan 15 10:31:48 then check your .bashrc .zshrc or whatever is ran at login because install is NOT apt Jan 15 10:31:53 I am not saying I'm doing it right Jan 15 10:31:59 no idea what you are talking about, but there is is *no* apt, and i think we gave pretty exact information on how to do it right. Jan 15 10:32:30 Yatekii: or run `alias` and see if there's an entry for install Jan 15 10:32:51 yesterday when I tried to use install (yes I looked at other receipes, even if you donm't believe it) it complained about an apt command it executed which was missing an install receipe. Jan 15 10:32:58 whcih is why I resorted to mkdir and cp Jan 15 10:33:25 not even that cargo.bbclass has anysigns of crazyness with apt: https://github.com/rust-embedded/meta-rust-bin/blob/master/classes/cargo.bbclass#L111 Jan 15 10:33:34 Hi all, I'm trying to build the sphinx package. I've found this recipe http://git.yoctoproject.org/cgit/cgit.cgi/meta-virtualization/tree/recipes-devtools/python/python-sphinx_1.7.8.bb?h=warrior. I have put it in my personal layer in recipes-devtools/python but bitbake is not finding it. What can I do? Jan 15 10:33:44 Yatekii: so no idea what you did or what you looked at, but get it out of your head REAL QUICK! Jan 15 10:33:55 try alias | grep install= on your host Jan 15 10:33:58 LetoThe2nd: I can just tell what the log said. I am not making any assumptions how bitbake really works Jan 15 10:34:24 there is no alias on this system and I don'0t believe there is at home. but i'll try at home again Jan 15 10:34:24 Mat80: why not just using meta-virtualization as-is? Jan 15 10:35:12 I'm really a newbie for yocto, Since that is the only recipe that I need I was thinking to not download the whole layer Jan 15 10:35:32 Yatekii: reading https://github.com/rust-embedded/meta-rust-bin/blob/master/classes/cargo.bbclass, I think you don't even need your do_install in your recipe Jan 15 10:36:01 I was thinking more layer => more parsing time for bitbake => slow Jan 15 10:36:57 Yatekii: https://github.com/rust-embedded/meta-rust-bin/blob/master/classes/cargo.bbclass#L120-L122 it looks for files in WORKDIR/target/[release|debug]/ Jan 15 10:37:03 qschulz: hmm you are right :/ Jan 15 10:37:12 yes, this is the right behavior Jan 15 10:37:19 kk imma try if this works :) Jan 15 10:37:34 but other than perf issue, what I'm doing wrong? can't I just add a recipe in a layer (that is present in bblayers.conf) and be able to find it using bitbake? Jan 15 10:37:36 Yatekii: if that does not work, maybe check if it's not trying to find files in debug instead of release Jan 15 10:38:58 Mat80: while the parsing time assumption is correct, you put the maintenance burden on yourself, and that has to be correlated. Jan 15 10:39:04 Yatekii: also, figure out why the heck install is not working in your recipe, because that's supposed to work and expected to work (except very specific uses cases, install instead of mkdir or cp is used throughout Yocto) Jan 15 10:39:27 Mat80: technically, if you adhere to the recipes-*/*/*.bb scheme then there should be no problem Jan 15 10:39:45 Mat80: if you have the layer on github or such we can have a look Jan 15 10:39:57 LetoThe2nd I know, but this is just for testing, I'm not sure I'll need it later Jan 15 10:40:10 no sorry is not public Jan 15 10:40:50 Mat80: other things in the layer work? Jan 15 10:40:56 yes Jan 15 10:41:08 qschulz: yes I will try as you both suggested and see if I maybe just operated it wrong :) Jan 15 10:42:17 qschulz ok one question tho: I am trying to clone a private repo over http (most likely it's gonna be public in the end but you never know). I have a the user and pw in .netrc as fond on the web and advised by khem. I don't like that solution particularly but it's okay. but what I struggle is to find the right url ... khem told me to use: git://user@host:......;protocol=ssh . I then did: SRC_URI = "git://nh@code.technokrat.ch/project/raichu-core.git;transport Jan 15 10:42:18 =https" which does not work sadly. do I need to add the host twice? or do I need to insert a : in ".../:ch..." ? Jan 15 10:42:18 I also I tried to take a look at the git fetcher docs but it seems to miss to explain that is there maybe more specific docs I am missing? Jan 15 10:42:18 Mat80: then the only thing that comes to my mind is triple checking that your recipe name + path match the scheme just given. Jan 15 10:42:42 ok I sourced again the oe-init-build-env script and not It works Jan 15 10:42:55 thank your for your time Jan 15 10:48:35 Yatekii: https://www.yoctoproject.org/pipermail/yocto/2014-April/019135.html try with this first? Jan 15 10:49:16 otherwise /me shrug Jan 15 10:49:22 I don't use private repos Jan 15 10:50:39 qschulz: seems like I don't have any linux-raspberrypi*.bb file Jan 15 10:52:30 Oh. I've found something Jan 15 10:53:28 qschulz: I can try, but then I rather make the repo public than use this solution ^^ Jan 15 10:53:41 I don't use private repos mostly either. for some company stuff I ned to Jan 15 10:53:43 *need Jan 15 10:54:28 FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}-${PV}:"LINUX_VERSION ?= "4.9.80"SRCREV = "7f9c648dad6473469b4133898fa6bb8d818ecff9"SRC_URI = "git://github.com/raspberrypi/linux.git;branch=rpi-4.9.y"require linux-raspberrypi.inc Jan 15 10:54:49 Yatekii: sure, it's just to test that Yocto can actually connect to your thing :) If it can then at least you know it works within Yocto in some ways Jan 15 10:55:15 nacknick: what is this :) ? Jan 15 10:55:15 qschulz: This is all the content, there is no`require recipes-kernel/linux/linux-yocto.inc` Jan 15 10:55:28 where's this file? what's its name? Jan 15 10:55:37 qschulz: I agree :) Jan 15 10:55:45 meta-raspberrypi/recipes-kernel/linux/linux-raspberrypi_4.9.bb Jan 15 10:55:54 nacknick you need to check inside linux-raspberrypi.inc Jan 15 10:56:45 nacknick: yeah sorry, I read the .inc :) Jan 15 10:56:52 stuom1: You're right. inside it there is `require recipes-kernel/linux/linux-yocto.inc` Jan 15 10:57:49 qschulz: yeah ofc my pw has a slash in it which is considered part of the URL XD Jan 15 11:06:12 hmm basically it tries to fetch from remote with git://user:pwd\/@code.technokrat.ch/ecarup/raichu-core.git and it tells me "unable to look up user:pwd" and my receipe has this: SRC_URI = "git://code.technokrat.ch/ecarup/raichu-core.git;transport=https;user=user:pwd\/" Jan 15 11:09:51 how can i find where a wic image is defined ? Jan 15 11:10:40 angelo__: look for .wks files Jan 15 11:10:44 angelo__: please define "defined" Jan 15 11:12:52 `bitbake -e ` and look for `WKS_FULL_PATH` should tell you which wks file is used Jan 15 11:14:28 kanavin: I've a hack for python 3.8 in -next. Torn on whether its a good idea Jan 15 11:16:22 LetoThe2nd, paulbarker thanks, found it Jan 15 11:17:09 it comes from a lower layer Jan 15 11:17:35 ok so I also can't curl my private repo, qschulz which is odd ... now I don't know how git clones this at all lol Jan 15 11:18:06 can i override it with a different .wks ? Jan 15 11:18:17 angelo__: should be no problem Jan 15 11:20:14 RP, I'd rather get to the bottom of it :-( it's okay to merge, but I'll try to find time to revert that and look into it Jan 15 11:23:17 kanavin: I would rather too, I added it mainly to see whether it was the only problem Jan 15 11:23:53 kanavin: its rather like the locale problem which we had open for years :/ Jan 15 11:24:39 RP, yes, that was an epic issue, and was solved by someone outside of the main contributors circle :) Jan 15 11:25:28 kanavin: indeed, although Jason does work heavily with the project, just not so much on the core Jan 15 11:34:18 I'm trying to figure out if there is a way to get bitbake to write a rootfs to a directory rather than a tarball. does anyone known if that is possible or where I can find that information? Jan 15 11:48:41 jdowner: ownerships wont be right, does that matter? Jan 15 11:48:53 bitbake can't write a file with root ownerships in if it isn't root Jan 15 11:49:24 that makes sense, but I don't think it'll be a problem Jan 15 11:51:52 in the setup I am currently using the rootfs is not owned by root. it is used via a bwrap container, and at that point the ownership is set to root Jan 15 11:52:22 I should mention that this is only something I want to do for active development Jan 15 11:52:46 the creation of the tarball and subsequent extraction is a bit of a bottleneck Jan 15 11:56:54 LetoThe2nd: is there actually a PROPRIETARY license or what is just a `foo`-like way to say you should put your license there? Jan 15 11:57:07 LetoThe2nd: wrt the tweet you sent on @YoctoThe Jan 15 11:58:04 Yatekii: https://git.yoctoproject.org/cgit.cgi/poky/tree/bitbake/lib/bb/fetch2/git.py the code is here (checkout the correct branch beforehand) Jan 15 11:58:21 jdowner: should be easy to make a new imagetype that just copies the rootfs instead of doing anything to it Jan 15 12:06:44 rburton: thanks! I'll take a look into that Jan 15 12:10:37 Has anyone tried to use kernel-yocto.bbclass with a non-git source for the kernel? I'm hoping to use kernel config fragments but the kernel in question is in svn not git Jan 15 12:11:52 paulbarker: you must have done something really bad in life to be forced to use a kernel on svn ;) Jan 15 12:12:49 $CLIENT uses svn for everything. It's not much of a problem except in this instance Jan 15 12:13:43 The nice config fragment and KERNEL_FEATURES handling really should be separated from the machine branch handling and git magic in kernel-yocto.bbclass. I imagine that's a whole chunk of work to pull them apart though Jan 15 12:15:15 qschulz: no, "proprietary" is actually an already existing keyword Jan 15 12:18:34 LetoThe2nd: May I ask what's the difference then? Jan 15 12:18:58 I see this for example: meta/files/common-licenses/Proprietary but it's practically empty Jan 15 12:21:08 qschulz: CLOSED means that bitbake outright ignores the license. whereas for proprietary you have to supply a file containing the very license you have, and bitbake can trigger if it changes. Jan 15 12:21:54 I see... Question then :) Jan 15 12:22:07 Since the license file is already in meta/files/common-licenses/Proprietary Jan 15 12:22:18 is there some override mechanism for that? Jan 15 12:22:35 e.g. we have a bunch of recipes using CLOSED now because... it's straightforward Jan 15 12:23:04 but we don't have licenses in the sources of all recipes Jan 15 12:23:09 qschulz: not exactly override, you have to provide those you refer to. Jan 15 12:23:15 But AFAIK, they should be identical Jan 15 12:23:25 qschulz: meta-atmel can serve as a good example, AFAIK: https://github.com/linux4sam/meta-atmel Jan 15 12:23:46 LetoThe2nd: yup, then you have a specific license, not "Proprietary" Jan 15 12:25:40 qschulz: indeed, but we're only talking about names now. i personally advocate using "proprietary-mylic" instead of just "mylic" because its clearer. Jan 15 12:25:59 qschulz: and you can whitelist easier with the prefix. Jan 15 12:26:23 trying to think of a replacement name for 'closed' Jan 15 12:26:58 rburton: well, if you don't want people to use it, just throw a warning at their face :p Jan 15 12:26:58 qschulz: but the key point is: it has to be defined != closed so bitbake can check it. if its CLOSED, then bitbake will happily shovel it into your MIT & BSD-only image, totally breaking compliance. Jan 15 12:27:18 LetoThe2nd: that's understood. Thx Jan 15 12:27:36 rburton: s/CLOSED/IGNOREBECAUSEIJUSTWANTTOGETRIDOFTHATMESSAGEEVENIFIHAVENOFUCKINGCLUEWHATITACTUALLYMEANS/g Jan 15 12:27:43 sounds good Jan 15 12:27:47 \o/ Jan 15 12:27:49 Well now I'm surprised, my hack actually worked. I just set `do_kernel_checkout[noexec] = "1"` and `do_validate_branches[noexec] = "1"` and fixed what looks like a bug in do_kernel_metadata. I'll see if I can think of how to upstream that as a supported option Jan 15 12:28:00 LetoThe2nd: needs to be a bit more descriptive? :) Jan 15 12:28:06 like some of the unstable gnome libraries used to need you to #define YES_I_KNOW_THIS_LIBRARY_ISNT_READY_YET Jan 15 12:28:15 rburton: i can add more swearing if desired. Jan 15 12:28:52 kanavin: python3 fails reproducibility tests Jan 15 12:29:16 RP: IGNOREBECAUSEIJUSTWANTTOGETRIDOFTHATMESSAGEEVENIFIHAVENOFUCKINGCLUEWHATITACTUALLYMEANSSOEVERYTIMESOMEBODYSPOTSTHISINTHEWILDIHAVETOEITHERBUYLETOTHE2NDABEERORPRESENTONMYSUPIDITYONTHENEXTYOCTODEVELOPERDAY Jan 15 12:29:16 LetoThe2nd: couldn't it be something we explain in the docs? why one shouldn't do this? Jan 15 12:29:32 LetoThe2nd: I hope we don't have a 255-long string issue with that one Jan 15 12:29:39 qschulz: hehe Jan 15 12:29:45 RP: netter? Jan 15 12:29:55 RP: better, even? Jan 15 12:29:57 LetoThe2nd: yes, everyone should buy you beer Jan 15 12:30:08 qschulz: patches to docs welcome Jan 15 12:30:36 LetoThe2nd: I think you're too used to those German compound nouns Jan 15 12:30:48 LetoThe2nd: trouble is you add something like that, then someone complains it needs better line wrapping Jan 15 12:30:57 paulbarker: hrhr Jan 15 12:31:05 RP: it must be painful to see. Jan 15 12:31:10 LetoThe2nd starting a cult: "Bring your bear to the altar". "Please repeat after me: "Recipe data is local, conf data is global."" Jan 15 12:31:20 beer* Jan 15 12:31:21 LetoThe2nd: right, there is a certain irony there Jan 15 12:31:23 then, and only then people will try to avoid it. Jan 15 12:31:29 qschulz++ Jan 15 12:32:43 RP: its like static_cast<>() in c++. its ugly and painful to type. but that way you are reminded that you are doing potentially very bad. Jan 15 12:32:50 LetoThe2nd: how about THISVARIABLEISPURPOSEFULLYLONGENOUGHTOMAKEYOUTHINKTWICEBEFOREYOUCHOOSETHISSPECIFICLICENSEBECAUSEITISNTACTUALLYALICENSEBUTJUSTAHACKANDYOUDONTWANTHACKSDONTYOUOKTHISWASCALLEDCLOSEDBEFORE Jan 15 12:33:06 qschulz: not insulting enough. Jan 15 12:33:33 LetoThe2nd: Sorry my leader. I'll try better next time I promise Jan 15 12:33:44 My licensing talk said "don't use Proprietary or CLOSED" Jan 15 12:34:42 paulbarker: :) Jan 15 12:34:54 LetoThe2nd: I think there might be a niche for comics based on https://webcomicname.com/ for your @YoctoThe Jan 15 12:35:19 qschulz: if in doubt, listen to paulbarkerinstead of me when it comes to licensing. just to be clear. Jan 15 12:35:36 (if you don't know, all comics ends with the same frame: "Oh no" Jan 15 12:36:08 * LetoThe2nd doesn't know. there is only one relevant webcomic, and its called 'xkcd' Jan 15 12:38:35 RP: sure. I'm still wrapping my head around how to write a section on OVERRIDES mechanism and the consequences Jan 15 12:39:23 qschulz: get yourself a webcam and just explain! Jan 15 12:40:21 LetoThe2nd: that's weird man. I'm fine presenting at conferences: you see the people you're talking to. Jan 15 12:40:40 LetoThe2nd: also, I only have a Chromebook so... setting up the thing... meh Jan 15 12:40:55 qschulz: ok imma try and understand the code :) I also thought of cloning the repo separately and then use file:// but that seems not nice Jan 15 12:41:17 Yatekii: Jan 15 12:41:24 qschulz: :( Jan 15 12:41:36 qschulz: xD Jan 15 12:42:16 LetoThe2nd: i'll think about it I promise Jan 15 12:42:46 i'll write a blog post/doc somewhere first and get it reviewed somehow Jan 15 12:43:05 not confident enough that I won't say crap Jan 15 12:43:25 qschulz: there are people who can read/review :) Jan 15 12:43:39 qschulz: hey, its ok. people are different. i personally find it awful to write anything longer than a couple of words (under normal cirrcumstances), but i have to problem in just talking an hour or two about random stuff Jan 15 12:43:40 RP: PEOPLE CAN READ? Jan 15 12:44:04 qschulz: sometimes not as many as would be ideal but yes, some Jan 15 12:44:41 qschulz: RP: rburton: if you have somethign you'd like to have presented, i can certainly also do that. Jan 15 12:45:02 * LetoThe2nd is reminded to make anmouncements! Jan 15 12:45:57 LetoThe2nd: maybe you can provide a mail address or form on a website when you're on Twitch or in YT comments so people can suggest topics? Jan 15 12:47:13 Ok, I have the dumb. kernel-yocto.bbclass works fine if you set a dummy KBRANCH value, do_kernel_checkout sets up a local git repository and all the subsequent tasks work Jan 15 12:47:38 I have a few fixes to upstream though :) Jan 15 12:58:14 In *linux-raspberry.inc* there is a line: `KBUILD_DEFCONFIG_raspberrypi3 ?= "bcm2709_defconfig"` Jan 15 12:59:57 I added the line `CONFIG_SECCOMP_FILTER=y` to *bcm2709_defconfig* file - this is how I should add that feature to the kernel? Jan 15 13:00:52 If that is the way - How to rebuild the kernel to apply changes? Jan 15 13:02:23 * LetoThe2nd hereby declares todays anthem: https://www.youtube.com/watch?v=th4Czv1j3F8 Jan 15 13:02:30 attention: metal, lyrically nsfw. Jan 15 13:04:54 nacknick: what *exactly* did you do? Jan 15 13:06:10 (and no, you can basically do echo 'CONFIG_SECCOMP_FILTER=y' > seccomp-filter.scc and then add it in a bbappend, that should work and be enough) but I can explain how you could have done it the way you wanted to do it if you give more info Jan 15 13:07:15 qschulz: inside *build/tmp/work-shared/raspberrypi3/kernel-source/arch/arm/configs/bcm2709_defconfig* i added the line `CONFIG_SECCOMP_FILTER=y` Jan 15 13:07:25 Is that file relevant for configuration? Jan 15 13:07:44 LetoThe2nd: I think we have a new chant for your sect. Jan 15 13:07:58 nacknick: Do not touch the tmp dir :) Jan 15 13:08:04 OK Jan 15 13:08:36 it could technically work in some cases but as soon as some tasks of the kernel will be triggered, you'll lose your changes Jan 15 13:08:37 qschulz: rule #666: if it exists, there is a metal version of it :) Jan 15 13:08:50 I should create a file called *seccomp-filter.scc* - where?qschulz Jan 15 13:08:52 LetoThe2nd: it was wrt modifying whatever is in tmp :) Jan 15 13:09:02 nacknick: in your layer Jan 15 13:09:39 qschulz: yeah i know Jan 15 13:09:50 nacknick: place it at the correct place. https://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#using-bbappend-files can be helpful if you never done bbappends before Jan 15 13:10:36 If I'm correct, you should need two lines in your bbappend Jan 15 13:10:59 qschulz: Should I add a new layer or only modify an existing recipe? Jan 15 13:11:20 nacknick: do not modify anything that you don't have control over is my advice Jan 15 13:11:45 so you are git cloning a repo? don't modify anything in it. Except for upstreaming the changes to the original repo :) Jan 15 13:12:02 OK Jan 15 13:12:11 So I will add a new layer, what next? Jan 15 13:12:12 for testing, it is *tolerated* IMHO but anyway you'll need your own layer later :) Jan 15 13:13:09 create the scc file the way I told you, put it in the right place, create the bbappend. Rule over the world. Jan 15 13:13:29 (the last two steps are explained in the link I sent you, read it carefully and make sure you don't have any typo) Jan 15 13:14:03 And don't stop reading before the end of the section, there's important info til the end Jan 15 13:14:06 including "rule over the world"? hh Jan 15 13:14:54 Is it necessary to create a recipe? or just a layer? Jan 15 13:19:44 you need the layer because you'll want to share your changes to some other people Jan 15 13:19:54 and it's a recipe, it's an append to a recipe Jan 15 13:20:01 ah, they're gone Jan 15 13:22:05 qschulz wrote: you need the layer because you'll want to share your changes to some other people Jan 15 13:22:07 qschulz wrote: and it's a recipe, it's an append to a recipe Jan 15 13:22:08 qschulz wrote: ah, they're gone Jan 15 13:22:23 nacknick: ^ Jan 15 13:23:08 dev1990: thanks. disconnected :| Jan 15 13:23:16 np Jan 15 13:26:23 dev1990: thx Jan 15 13:33:10 dev1990: can you copy the instructions qschulz wrote before I was disconnected? I have to refresh the page and all's gone Jan 15 13:33:20 sec Jan 15 13:34:01 you need the layer because you'll want to share your changes to some other people Jan 15 13:34:06 and it's a recipe, it's an append to a recipe Jan 15 13:34:24 (and no, you can basically do echo 'CONFIG_SECCOMP_FILTER=y' > seccomp-filter.scc and then add it in a bbappend, that should work and be enough) but I can explain how you could have done it the way you wanted to do it if you give more info Jan 15 13:34:24 nacknick: https://www.yoctoproject.org/irc/%23yocto.2020-01-15.log.html Jan 15 13:34:37 dev1990: WE'RE BEING MONITORED? Jan 15 13:34:49 hehehe Jan 15 13:36:07 most of irc channels are logged, I think Jan 15 13:37:10 qschulz: The link you sent talks about bbappend, but I could not understand what recipe I'm appending to... Jan 15 13:38:26 meta-browser doesn't have branches after morty, is it supposed to use master only now? Jan 15 13:39:24 mckoan: afaik yes. Jan 15 13:40:03 however unfortunately firefox build fails due to rust :-( Jan 15 13:41:26 yeah, i've given it a spin a couple of days back too and got stuck real quick. Jan 15 13:41:38 luckily it was just for experimentation, no real need. Jan 15 13:43:58 maybe khem knows it better Jan 15 13:52:59 qschulz: What recipe should I append to? and where *seccomp-filter.scc* should be reside? Jan 15 14:01:07 nacknick: well, what's the recipe we've been looking into for the kernel? Jan 15 14:01:38 where the file should reside is also explained in the link I sent you Jan 15 14:02:56 qschulz: you mean this one? meta-raspberrypi/recipes-kernel/linux/linux-raspberrypi_4.9.bb Jan 15 14:03:31 is there a better solution than this: FILENAME_WITHOUT_PATH = "${@d.getVar('FILE_WITH_PATH', d, 1).split('/').pop()}" Jan 15 14:04:16 nacknick: linux-raspberrypi yes, I don't know if it's 4.9 you're using but yes Jan 15 14:05:59 dev1990: what's your need first? if you can call a sh function, basename ${FILENAME_WITHOUT_PATH} should do it? Jan 15 14:07:44 Python also has `os.path.basename()` IIRC Jan 15 14:08:45 paulbarker: then somethink like this should work: ${@os.path.basename(d.getVar('FILE_WITH_PATH', d, 1))} ? Jan 15 14:09:40 Yes, assuming os.path is already imported in that context. I don't recall the full list of what bitbake imports for those expressions Jan 15 14:11:15 qschulz: I read all the section. But if I create the *seccomp-filter.scc* using the bbappend recipe, it will create it in the bbappend recipe's directory Jan 15 14:12:12 I haven't said that you should create seccomp-filter.scc Jan 15 14:13:06 if you recall the very beginning of our discussion, I said you can either have .scc, .cfg (if kernel-yocto is inherited) or full defconfig (if implemented by the recipe) passed in SRC_URI of a recipe Jan 15 14:13:20 And I told you also that exactly two lines in your bbappend should be enough Jan 15 14:13:42 hmm I think the git fetcher does not use https Jan 15 14:13:46 which is why it doesn't work :/ Jan 15 14:13:51 Yatekii: protocol=hhtps Jan 15 14:13:55 Yatekii: protocol=https Jan 15 14:14:07 qschulz: that's what I have Jan 15 14:14:16 SRC_URI = "git://code.technokrat.ch:443/ecarup/raichu-core.git;transport=https" Jan 15 14:14:26 ohhh Jan 15 14:14:28 I am stupid Jan 15 14:14:32 pls don't mind me Jan 15 14:14:50 Read carefully when we send docs :) especially typo wise :) Jan 15 14:14:50 port ? :) Jan 15 14:15:19 a ok Jan 15 14:15:20 dev1990: s/transport/protocol/ Jan 15 14:15:51 nacknick: if you are waiting for a highlight to checkout your IRC, read ~10 messages above Jan 15 14:15:57 qschulz: well, when you are convinced that you already added that parameter, it's hard to read properly :/ Jan 15 14:15:57 ah, yes Jan 15 14:16:34 qschulz: sorry, I thought you wrote `echo bla bla > seccomp-filter.scc` inside the bbappend Jan 15 14:16:35 Yatekii: careful reading before coding :) Jan 15 14:17:05 yeah, you are right, I should not do that 3am :/ because I added that yesterday night Jan 15 14:17:06 ah no, it was meant to be run on your host. Just a way to say that the only content in seccomp-filter.scc should be CONFIG_SECCOMP_FILTER=y :) Jan 15 14:17:36 Yatekii: tired you is idiot you :) (I mean that as general you ;) ) Jan 15 14:18:00 haha yeah 100% ssecond that :) Jan 15 14:18:33 hmm I still got the install commands wrong ... https://gist.github.com/Yatekii/c908f8007ff1134d024e303fdb1e97aa it wants a second arg ... but I just wanna create a dir, not move one :/ Jan 15 14:18:47 or the contents Jan 15 14:19:00 Yatekii: -m takes an argument Jan 15 14:19:45 Yatekii: and again, you shouldn't need that in your recipe Jan 15 14:20:00 the do_install is managed by cargo bbclass IIUC Jan 15 14:20:20 qschulz: yep I know :) I just wanted to try if I can get it to work Jan 15 14:20:30 for science reasons you know Jan 15 14:20:43 rburton: that worked apparently :) thx Jan 15 14:20:50 I thought it already dies on the first command Jan 15 14:21:21 qschulz: what is the second line you mean I should add to linux-raspberrypi_4.9.bbappend? :S Jan 15 14:21:42 I mean besides SRC_URI Jan 15 14:23:46 Yatekii: read the error log when that happens, it usually gives you info which line failed I think Jan 15 14:24:28 nacknick: it's explained in the link, re-read again :) Jan 15 14:24:37 it has to do with where Yocto is looking for files Jan 15 14:25:11 OK Jan 15 14:31:06 nacknick: found it? Jan 15 14:31:12 not yet Jan 15 14:31:38 maybe I read it and don't understand that this is the answer Jan 15 14:31:42 I'll read again Jan 15 14:34:02 Hello, I'm using a separate partition for /boot mmcblk0p1 and I wish to have the file boot.scr also installed to mmcblk0p2 which is rootfs.I made a .bbappend for rpi-u-boot-scr ( http://dpaste.com/15N6RJB ) but it doesn't copy the file to mmcblk0p2. does someone have an idea on how to fix this ? Jan 15 14:34:19 Hello. Here a newbie question: When building sdk_ext, how can I add a receipe that is normally not included? Jan 15 14:35:53 nacknick: SRC_URI just gives the path to a file you want to include, but it's a relative path. Yocto has a default for recipes to know where to look for. Since appends are in a different absolute path, Yocto will not look for that path by default so you need to tell Yocto to look for *files* in *extra paths* Jan 15 14:36:25 nemgti-og: https://www.yoctoproject.org/docs/3.0.1/sdk-manual/sdk-manual.html#sdk-adding-individual-packages AFAIK Jan 15 14:36:30 yea. I read about the FILESEXTRAPATHS Jan 15 14:36:46 nacknick: well you need that one then, that's one line Jan 15 14:36:52 and you need SRC_URI, that's a second line :) Jan 15 14:37:03 and then fingers crossed, it works Jan 15 14:39:05 Hi! LetoThe2nd. Thanks! I began watching your tutorials in Youtube yesterday! I like them! Ich mache gerade was du normalerweise am ende machst! :D Jan 15 14:39:14 nemgti-og: dann, prost! Jan 15 14:39:35 I guess I an not that patient to go through all of them before attemting to do some stuff Jan 15 14:39:43 I just could not understand how those things connect together. I did not understand how those line will finally add *CONFIG_SECCOMP_FILTER=y* to *bcm2709_defconfig* Jan 15 14:40:52 nemgti-og: depends on what you mean by "stuff", but the first 3 should be enough to cover entry level tasks Jan 15 14:41:48 LetoThe2nd: but the sdks are not covered in the first three videos Jan 15 14:42:04 actually I am just now realizing it is covered in the fourth Jan 15 14:42:08 nemgti-og: SDKs is #4 Jan 15 14:42:33 jap. thanks! I hope I get to drink a beer with you and to talk about yocto in a pub Jan 15 14:42:42 plus, SDKs are pretty much overrated except for very specific producer-consumer situations. Jan 15 14:43:21 unless you're running a large team or are a HW vendor, its usually way easier to just let everybody use a proper build Jan 15 14:43:34 nacknick: it's the kernel-yocto class which is taking those .scc files and merging them to the defconfig. You provide an .scc file, without those lines, Yocto does not know it has to use this file Jan 15 14:43:57 I see... well to be honest I am just playing around and enjoying myself Jan 15 14:44:08 nemgti-og: forget about SDKs, then. Jan 15 14:44:29 nemgti-og: save the brain capacity for understanding and writing recipes. Jan 15 14:44:40 haha ok. Will do Jan 15 14:44:57 nacknick: AFAICT, https://git.yoctoproject.org/cgit.cgi/poky/tree/meta/classes/kernel-yocto.bbclass#n72 this is the function which is handling scc files Jan 15 14:45:03 nemgti-og: #7 is also quite important Jan 15 14:45:12 noted Jan 15 14:51:03 qschulz: thanks. make much more sense Jan 15 15:03:01 that lame IRC server... always disconnects me >.< Jan 15 15:03:42 by the way question: recently I get the following error while trying to build the image: Jan 15 15:03:44 `ERROR: Unable to start bitbake serverERROR: Last 10 lines of server log for this session (/home/user/yocto/poky/build/bitbake-cookerdaemon.log): self.add_filewatch(self.data.getVar("__base_depends", False), self.configwatcher) File "/home/user/yocto/poky/bitbake/lib/bb/cooker.py", line 287, in add_filewatch watcher.add_watch(f, Jan 15 15:03:45 self.watchmask, quiet=False) File "/home/user/yocto/poky/bitbake/lib/pyinotify.py", line 1924, in add_watch raise WatchManagerError(err, ret_)pyinotify.WatchManagerError: add_watch: cannot watch /home/user/yocto/poky/build/conf WD=-1, Errno=No space left on device (ENOSPC)ERROR: No space left on device or exceeds Jan 15 15:03:45 fs.inotify.max_user_watches?ERROR: To check max_user_watches: sysctl -n fs.inotify.max_user_watches.ERROR: To modify max_user_watches: sysctl -n -w fs.inotify.max_user_watches=.ERROR: Root privilege is required to modify max_user_watches.` Jan 15 15:04:22 after log off and log on it builds with no problem, but after few minutes getting that error again Jan 15 15:04:27 nacknick: well whats not clear about that? Jan 15 15:04:40 I have enough space Jan 15 15:05:03 read on. Jan 15 15:05:03 enough inodes left? Jan 15 15:05:19 qschulz: don't know what those Jan 15 15:05:31 "To check max_user_watches: sysctl -n fs.inotify.max_user_watches." Jan 15 15:06:08 what does it mean Jan 15 15:07:13 nacknick: irccloud is your friend if you keep getting disconnected Jan 15 15:08:09 nacknick: https://unix.stackexchange.com/questions/13751/kernel-inotify-watch-limit-reached Jan 15 15:08:20 nacknick: and Jan 15 15:08:21 https://hub.docker.com/r/wolverine2k/yocto-build-env/ Jan 15 15:08:29 LetoThe2nd: thanks. I'll look at it Jan 15 15:10:22 LetoThe2nd: btw, any Yocto docker to recommend? Jan 15 15:11:21 The one that was linked seems a little old Jan 15 15:11:37 It's easy enough to build a new container, though Jan 15 15:11:56 qschulz: I am using crops/poky Jan 15 15:12:08 qshcultz: https://github.com/garmin/pyrex Jan 15 15:12:09 so far without any issues at all Jan 15 15:12:23 Ack, forgot to tab complete again :) Jan 15 15:12:55 JPEW: calling me names like this. How dare you Jan 15 15:12:59 :) Jan 15 15:13:38 qschulz, if you do look at pyrex, pull from the "next" branch, it has all the latest goodies. Hopefully we'll be release 1.0.0 soon :) Jan 15 15:13:39 hah Jan 15 15:15:34 JPEW: ubuntu-14.04? Jan 15 15:16:03 wait I'm so confused :D Jan 15 15:16:09 qschulz: depends (TM) Jan 15 15:16:30 qschulz, pyrex has ubuntu-14.04, ubuntu-16.04 and ubuntu-18.04 images Jan 15 15:16:42 You can use it all the way back with Yocto 2.1 Jan 15 15:16:49 qschulz: the obvious choices are crops, pyrex and homebrewn, obviously. and kas-docker if you are a kas guy. Jan 15 15:16:56 (hence why we have so many images) Jan 15 15:17:48 Hmm, perhaps I should add a "quickstart" guide to the documentation :) Jan 15 15:18:06 qschulz: given current activity i would probably suggest pyrex these days, but admittedly i've never tried it. Jan 15 15:18:20 JPEW: erm... do you have a webcam? :P Jan 15 15:18:39 LetoThe2nd, yes.... why? Jan 15 15:18:43 I'm gonna have a look at the different options (some colleagues have some docker because running gentoo or arch) Jan 15 15:18:50 JPEW: IT'S A TRAP Jan 15 15:19:06 JPEW: consider getting in front of it and do a live coding session to show it to the world :) Jan 15 15:19:54 LetoThe2nd, Sure! Jan 15 15:20:08 LetoThe2nd, After we get the 1.0.0 release out (which should be really soon here) Jan 15 15:20:15 JPEW: awesome! Jan 15 15:20:32 ^^^ kergoth, moto-timo Speaking of which.... Jan 15 15:26:24 New news from stackoverflow: Start an Electron app at boot on Raspberry Pi 3 with yocto Jan 15 16:26:05 i would like u-boot to display ghash with version at boot. Can i set up this form yocto recipe ? Jan 15 16:35:22 hmm I am trying to do "devtool deploy-target raichu-core user@host" and it tells me "ERROR: No recipe named 'raichu-core' in your workspace" ... when I run "devtool edit-recipe raichu-core" it opens the correct recipe tho. any ideas? Jan 15 17:11:31 nacknick: https://koansoftware.com/yocto-linux-lab-2018-florence/?lang=it Jan 15 17:47:21 Yatekii: devtool modify then devtool build then devtool deploy-target. Jan 15 17:50:20 qschulz: oh so i need modify first :) didnt find that in the guide ... weird Jan 15 17:50:54 Yatekii: what were you trying to do? Jan 15 17:53:32 qschulz: well i built the receipe with bitbake -c build raichu-core and wanted to deploy that to my existing target to test it because i didnt want to flash the entire rootfs again ofc :) Jan 15 17:53:46 if it's just to upload a new package version without reflashing the image, you can do bitbake raichu-core, take the pkg from tmp/deploy/[ipk|deb|rpm] send it to your board and install it with [apt|opkg|dpkg...]. Though you need this to be installed on your current rootfs Jan 15 17:54:38 well deploy target is just very convenient which is why i wanted to use it :) Jan 15 17:55:17 it is very convenient but I also see you forgetting to run devtool reset raichu-core and loosing some hair on why the heck your recipe isn't updated ;) Jan 15 17:55:31 BUT it is actually a good idea when developing an app or debugging it Jan 15 17:55:32 I'd like to append the content of a recipe local file into the final image artefact Jan 15 17:55:39 something like cat cat ${S}/myfile.txt >> ${DEPLOYDIR}/bcm2835-bootfiles/config.txt Jan 15 17:55:53 instead of simply echoing like printf "${RPI_EXTRA_CONFIG}\n" >> ${DEPLOYDIR}/bcm2835-bootfiles/config.txt Jan 15 17:55:59 any idea? Jan 15 17:56:20 http://git.yoctoproject.org/cgit/cgit.cgi/meta-raspberrypi/tree/recipes-bsp/bootfiles/rpi-config_git.bb Jan 15 17:56:24 well.... who's adding this config.txt and how? Jan 15 17:56:44 ok :) Jan 15 17:56:53 why don't you patch this recipe then? Jan 15 17:56:57 qschulz: I'm doing it into my bbappend do_deploy_append() Jan 15 17:57:18 what's the issue then? Jan 15 17:57:44 qschulz: why would i run reset? Jan 15 17:57:58 qschulz: doesn't find the ${S}/myfile.txt Jan 15 17:58:12 Yatekii: because if you don't the sources in your devtool workspace will be used instead of the recipe in your layers Jan 15 17:58:19 so the day comes you want to use your recipe again... Jan 15 17:58:38 mckoan: ${WORKDIR} instead of ${S} I highly suspect is what you want Jan 15 17:59:12 mckoan: http://git.yoctoproject.org/cgit/cgit.cgi/meta-raspberrypi/tree/recipes-bsp/bootfiles/rpi-config_git.bb#n14 Jan 15 17:59:29 but files from SRC_URI are in ${WORKDIR} by default Jan 15 18:00:12 Yatekii: just don't forget to run reset when you are finished with active development or debugging ;) Jan 15 18:02:54 qschulz: doh! it's been a hard day. Thank you! Jan 15 18:23:02 qschulz: hmm sorry, but can you link me some straight forward docs which explain a basic step by step for this? because I was reading the help docs of devtool and also in the yocto guides but I didn't see reset or anything similar mentioned in there :/ Jan 15 18:23:12 and just from the help info it's hard to tell how to go Jan 15 18:23:32 for example I didn't understand how I get into a devtool workspace and whether I am supposed to create a new one Jan 15 18:24:11 I just used bitbake to rebuild my image and just wanted to deploy the binary inside the image via devtool to not flash the entire image. Jan 15 18:35:41 RP: The Milestone tags are only created on poky repo, can this be extended to oe-core and bitbake repos too ? Jan 15 18:36:59 paulbarker: you should have pinged me about the linux-yocto stuff, I've had a queue for what you were looking for, for ages. Jan 15 18:38:30 zeddii: how would one know unless you posted it to ml or some public forum ? Jan 15 18:38:37 it's been posted Jan 15 18:38:43 plus, comon Jan 15 18:38:59 I see, then perhaps it would be based on that ? Jan 15 18:39:02 who the else hacks linux-yocto ? you think you'd ping the maintiner first. Jan 15 18:39:09 * zeddii is annoyed Jan 15 18:39:14 as I suffer through new kernels Jan 15 18:39:18 and get a comment like that. Jan 15 18:39:19 comon Jan 15 18:40:13 sometimes people need to get it out of their way and results in patches Jan 15 18:48:14 true 'dat. Jan 15 18:48:36 * zeddii goes to hunt for food. Jan 15 18:58:29 n/quit Jan 15 18:58:32 ups Jan 15 19:15:00 khem: yes, it could. I think we need to do a bit of a re-eval in that area Jan 15 19:49:38 RP: this will help in testing the mls Jan 15 21:12:22 how would i add a QA for a kernel build? i want to add some specific checks for a kernel FIT image. Jan 15 21:22:54 khem: looks like our nativesdk-gcc doesn't work properly :/ Jan 15 21:24:08 khem: config.log-configure:5841: gcc -o conftest -isystem/media/build1/poky-morty/build/tmp/sysroots/x86_64-linux/usr/include -O2 -pipe -L/media/build1/poky-morty/build/tmp/sysroots/x86_64-linux/usr/lib -L/media/build1/poky-morty/build/tmp/sysroots/x86_64-linux/lib -Wl,-rpath-link,/media/build1/poky-morty/build/tmp/sysroots/x86_64-linux/usr/lib -Wl,-rpath-link,/media/build1/poky-morty/build/tmp/sysroots/x86_64-linux/lib -Wl,-rpath,/med Jan 15 21:24:16 ia/build1/poky-morty/build/tmp/sysroots/x86_64-linux/usr/lib -Wl,-rpath,/media/build1/poky-morty/build/tmp/sysroots/x86_64-linux/lib -Wl,-O1 conftest.c -lmpc -lmpfr -lgmp >&5 Jan 15 21:24:19 config.log://lib/x86_64-linux-gnu/libm.so.6: undefined reference to `__strtof128_nan@GLIBC_PRIVATE' Jan 15 21:24:26 that is morty with a gcc from pyro buildtools-extended-tarball Jan 15 21:50:44 __strtof128_nan is in glibc 2.26+ Jan 15 21:51:01 what version of glibc is being used Jan 15 22:11:53 hm. i added a task but it seems like that didn't cause a rebuild.. is there something extra that needs to be done to make adding/modifying a task invalidate the stamps? Jan 15 22:12:04 khem: its complicated :/ Jan 15 22:12:29 khem: "gcc -lmpc -lmpfr -lgmp " fails. Add -lm and it passes Jan 15 22:12:59 khem: there is the glibc version in the buildtools tarball, the one on the host system and the target Jan 15 22:14:54 khem: when gcc adds internal libs, it searches the paths differently? Jan 15 22:50:36 khem: looks like its a linker behaviour difference in the path searched for -lm vs its interal lib search Jan 15 23:26:14 khem: its a relocation problem in binutils - SEARCH_DIRS doesn't relocate Jan 15 23:26:52 moto-timo: ^^^ Jan 15 23:27:28 khem: hopefully you have some insights for us :) Jan 15 23:33:16 * RP notes we delete the scripts and rely on the builtins Jan 16 00:01:22 moto-timo: this is such a silly problem :( Jan 16 00:01:33 moto-timo: spent hours but I think its simple Jan 16 00:02:07 RP: same here :) Jan 16 00:04:23 RP: strace it and find what all libs are in play, we do too much trickery for all these native and nativesdk quirks I do not trust what we think Jan 16 00:04:52 khem: right, after staring at strace its using /usr/bin/ld Jan 16 00:04:59 khem: good point, we might be missing something in the tarball Jan 16 00:05:03 khem: if I make it use our ld it would be a god start Jan 16 00:05:14 no, I am interested in tracing the .so its using/loading Jan 16 00:05:14 \o/ Jan 16 00:05:33 khem: been staring at strace for a while Jan 16 00:05:47 -Wl,--verbose was also helpful Jan 16 00:05:56 and perhaps -Wl,-Map,xx Jan 16 00:06:01 so we get a linker map Jan 16 00:06:15 that gives some peek into what linker is seeing Jan 16 00:06:57 if its defaulting to system linker thats proobably GCC_EXEC_PREFIX is not right Jan 16 00:06:59 khem: its simpler, its using the wrong linker Jan 16 00:07:49 RP: for tests export GCC_EXEC_PREFIX to point to where the linker you want is ( ld that is ) Jan 16 00:08:18 if that works then it can be worked backwards Jan 16 00:11:38 khem: I can make it to use the right linker but it still has the wrong SEARCH_DIRS :) Jan 16 00:11:48 RP: which package are you building when that error pops up ? it seems gcc from what I can deduce Jan 16 00:12:40 khem: I added in the ldscripts but its not looking at them and using its internal linker script. Any idea how to force it to use external files? Jan 16 00:12:52 RP: your doing :) deleting scripttempl is perhaps biting here Jan 16 00:13:19 khem: its failing in gcc-cross-initial but its trivial to make it break outside that with just the buildtools tarball alone Jan 16 00:13:38 khem: its always my fault :) Jan 16 00:13:56 I see, Jan 16 00:14:38 RP: set LIB_SEARCH_DIRS Jan 16 00:14:42 env var Jan 16 00:14:47 and see if linker picks it Jan 16 00:15:01 I am still trying to poke at where the hole is Jan 16 00:18:36 RP: is it using some sort of nativesdk-binutils too ? Jan 16 00:18:53 khem: yes, we need nativesdk-gcc using nativesdk-binutils Jan 16 00:19:04 where are recipes ? Jan 16 00:19:19 khem: bitbake builttools-extended-tarball Jan 16 00:19:31 khem: let me dump my status into email Jan 16 00:22:16 RP: can you try to disable 0004-Point-scripts-location-to-libdir.patch for nativesdk case in binutils Jan 16 00:27:03 khem: sent email, sorry its a bit convoluted Jan 16 00:27:24 khem: I need to sleep, too tired to keep going with this now. Will quickly try dropping that though Jan 16 00:30:38 is ALTERNATIVE_TARGET_${PN}-foo[name] = "..." supposed to work..? Jan 16 00:30:47 if i try setting it it seems to be ignored Jan 16 00:32:10 https://www.yoctoproject.org/docs/2.5/ref-manual/ref-manual.html#var-ALTERNATIVE_TARGET Jan 16 00:32:26 where are you setting it? Jan 16 00:32:43 in my recipe. Jan 16 00:33:07 i am trying to use the 'ALTERNATIVE_TARGET_pkg[name] = "target"' syntax, as i have one recipe that creates multiple packages Jan 16 00:35:11 i am going to see if doing it in python works, as the psplash recipe does Jan 16 00:37:02 khem: didn't help :( Jan 16 00:39:56 khem: dumped more into email Jan 16 00:39:59 I need tosleep Jan 16 00:40:49 time to yield() Jan 16 00:58:58 sigh, doesn't seem to work Jan 16 01:34:24 hm. Jan 16 01:34:48 ALTERNATIVE_TARGET_pkg[name] = "target" does not work but ALTERNATIVE_TARGET_pkg = "target" does.. but i need multiple alternatives for a single package Jan 16 01:39:00 now i'm really baffled Jan 16 01:39:33 if i use ${PN} inside of it does not work, but if i manually expand ${PN} inside of it does work Jan 16 02:58:33 New news from stackoverflow: How to store pyc files with setuptools3 in bitbake recipe in the same way setuptools does? **** ENDING LOGGING AT Thu Jan 16 02:59:59 2020