**** BEGIN LOGGING AT Thu Aug 18 02:59:58 2016 Aug 18 07:20:37 Hi all, here is my question : I want to have "brctl" command to bridge ethernet and bluetooth, what is your advise to make it ? I tried to use "bride-utils" that is in "meta-networking", but it doesn't want to compile (the dependencies are ok)... Is there a simplier solution than to use meta-networking ? Thanks ! Aug 18 07:22:40 Kakounet: meta-networking is probably the way to go... why doesn't it compile ? Aug 18 07:22:50 (and what version of yocto are you using ?) Aug 18 07:22:55 Kakounet: although bride-utils is a wonderful typo, meta-networking should be fine. have you checked that you are not mixing up branches? what is the compile error you see? Aug 18 07:23:07 LetoThe2nd: hehe Aug 18 07:23:52 ERROR: Failed to parse recipe: /home/architech/architech_sdk/architech/hachiko/yocto/poky/meta-openembedded/meta-oe/recipes-support/libsoc/libsoc_0.8.1.bb Aug 18 07:24:18 that's weird... Aug 18 07:24:20 ERROR: Error executing a python function in : Aug 18 07:24:26 ...and a lot of lines Aug 18 07:24:52 could you pastebin those lines somewhere ? what version of yocto and meta-networking are you using ? Aug 18 07:26:03 Do you have the command to know the version of yocto plz :) Aug 18 07:26:13 Kakounet: so you're either on krogoth or master, right? on purpose? Aug 18 07:26:59 Kakounet: plus, the path suggests that you cloned meta-oe beneath your poky directory, with is certainly a bad practise. Aug 18 07:27:37 http://pastebin.com/UFPPAsZg Aug 18 07:28:16 14.04 not validated? that sounds like you are some super ancient thing.... Aug 18 07:28:22 * Kakounet is kind of a yocto begginner :) Aug 18 07:28:51 Kakounet: let me guess, and you are starting out with some tarball blob that you received from your board distributor? Aug 18 07:29:38 I'm using a virtual machine that has been given by a board distributor yes :) It's Architech Aug 18 07:29:46 Hachiko board :) Aug 18 07:29:58 Kakounet: first of all, i'd suggest carefully walking through the yocto quick start tutorial to learn the super basics Aug 18 07:30:32 Kakounet: you're basically running into a version mismatch between the poky version that your supplier baked into that VM and meta-oe, which you just cloned into master Aug 18 07:31:15 Kakounet: so the fix is to find out the poky version they are supplying, and then switch to the corresponding branch on the meta-oe checkout. Aug 18 07:31:18 my guess would be meta-oe (master) uses python3 whereas your provider's yocto uses an older version (python2) Aug 18 07:31:40 I read a white book made by french guys (I'm french) the "Smile" company. It's super basic... But ok I will read the yocto quick start Aug 18 07:32:24 ah ok, cool, it's a version mismatch, Ok I can dig in that direction Aug 18 07:33:04 Kakounet: and to get your terminology right: you are not using yocto, you are using poky, which is provided by the yocto project. Aug 18 07:33:18 * boucman_work works there, the author is my neighbour :P Aug 18 07:34:05 boucman_work: ok, now you're basically FLS. Aug 18 07:34:38 FLS ? Aug 18 07:34:54 boucman_work: First Level Support Aug 18 07:35:01 hehe Aug 18 07:35:32 Okay ! Thanks for those elements, it's really helpful. And ok I will get my terminology right ;) Aug 18 07:36:51 LetoThe2nd: TBH, that's one of the services my company provides, but most board providers do a horrendous job of supporting Yocto. Old versions that they absolutely refuse to upgrade, most of them don't know what a recipe is... their documentation is minimalist (on board compilation, wtf) Aug 18 07:37:55 boucman_work: ack. Aug 18 07:39:37 boucman_work: suggesting customers to compile in a VM sounds like abuse on purpose anyways. Aug 18 07:39:50 Hello everyone Aug 18 07:39:51 yeah, but it's very common Aug 18 07:40:09 I have a quick question regarding the SDK Aug 18 07:40:20 as is providing binaries in a tar.gz, unpacking the rootfs, overwriting and repacking Aug 18 07:40:45 I'm building chromium inside a yocto build env, that needs Google's depot_tools obviously Aug 18 07:41:08 boucman_work: It's really my situation, I have no support from them and I try as much as I can alone but it's a really big subject. You made a huge and super powerfull thing, but it's hard for a begginner :D Aug 18 07:41:13 Now I would like to install the depot_tools (mostly pyhon scripts) to the populated SDK Aug 18 07:42:38 Kakounet: yeah, I know how you feel, but once you get it, yocto is awesome Aug 18 07:56:26 hi all Aug 18 08:05:22 hi Aug 18 08:08:40 boucman_work: s/yocto/openembedded/ :) Aug 18 08:09:04 LetoThe2nd: yocto just overheated my laptop, I probably have missed your last sentence :P Aug 18 08:09:51 boucman_work: i just meant that once you get the hang of it, openembedded is awesome. not yocto, as yocto is just a fancy name in the end. Aug 18 08:10:19 hehe Aug 18 08:10:42 agreed, though yocto is more and more the "commercial name" for the whole OE galaxy Aug 18 08:12:04 boucman_work: like in so many places, something just doesn't magically become more true if repeated often enough. Aug 18 08:12:49 (please don't take it personal :-) ) Aug 18 08:13:16 agreed again, but you need to adapt to your audience, most people with beginner questions don't have the OE-background to really understand the difference between yocto and OE, and they don't really care since they are working on yocto anyway :P Aug 18 08:15:08 boucman_work: the whole naming construct is unfortunate, that is oh so true. Aug 18 08:16:26 boucman_work: yet i still think that ironing out a couple of things here and there does good in the long run. i mean, there is no way to properly respond to somebody who want to know where to download rpms to install on his machine that is running the "yocto distribution" Aug 18 08:17:26 boucman_work: either that person gets the terminology, and the meaning behind it right - or there is no help. Aug 18 08:17:37 (can be no help) Aug 18 08:18:07 "yocto distribution" <- more shadyness, don't you mean "poky" ? :) Aug 18 08:18:13 LetoThe2nd: yeah, pedagogy is a subtle balance Aug 18 08:18:30 jubr: see context :) Aug 18 08:18:55 LetoThe2nd: ok, ok, in context: true. Didn't read too far back Aug 18 08:19:03 sometime you have to lie to your students so that they understand... being pedantic is usefull when your students understand the underlying concepts, but in that case most beginners don't. I save that for later Aug 18 08:19:19 but overall, I think we agree anyway Aug 18 08:19:44 and even if not, there's nothing bad about civilized disagreement. Aug 18 08:20:16 :) Aug 18 08:23:53 thats what oh so many people mix up. just because i disagree with you on a professional topic (happens often enough with my customers), theres no personal problem. i still think you are a nice guy. Aug 18 08:24:23 boucman_work: (again, not you specifically. speaking to a 'virtual customer') Aug 18 08:37:45 hello guys, I managed to get a core-image-minimal NAND boot on my imx28evk but I encountered a kernel panic https://paste.ee/p/oDSL9 Aug 18 08:38:02 any idea what it is about ? Aug 18 08:45:49 kimo: its unable to mount your rootfs Aug 18 08:46:51 kimo: line 175ff sugest that your filesystem is not what you think it is. Aug 18 08:48:38 LetoThe2nd: Yes but I just flashed the ubifs image using update_nant_filesystem cmd Aug 18 08:49:20 kimo: "i did something" does not necessarily mean "i did something correctly" or "i did something successfully" Aug 18 08:49:51 kimo: my personal advice would be to set up nfs root, make that boot, and from there inspect the ubi/ubifs Aug 18 08:51:42 LetoThe2nd: right, I was expecting someone would have this issue, which is maybe related to imx28evk.conf Aug 18 08:51:42 kimo: or, as a slightly simpler stage maybe - stay in uboot and inspect from there, it might have some means of doing that. Aug 18 08:53:55 kimo: my gut feeling says that you just did something slightly wrong, and hence the nand content is not exactly what it should be. but thats just guessing Aug 18 08:54:02 LetoThe2nd: do you have any knowledge of how the yocto autotest/builder infrastructure works, by any chance ? Aug 18 08:54:33 boucman_work: only very superficial. AFAIK its a buildbot that has been beaten into shape Aug 18 08:54:47 boucman_work: pidge is the one to poke about. Aug 18 08:55:15 (or at least she was the last time i touched that topic) Aug 18 08:55:42 ok, thx Aug 18 08:55:44 boucman_work: which parts of it? autobuilder, yes. the automated testing? it's been a while, but yeah. Aug 18 08:55:55 pidge: ah you wake? cheers! Aug 18 08:56:18 LetoThe2nd: Yup, ireland time now. Aug 18 08:56:21 pidge: I am trying to build yocto with no gcc in PATH (i.e using BUILD_CC variable and checking that no software hardcodes the name gcc) Aug 18 08:56:33 pidge: holiday or relocated? ;-() Aug 18 08:56:40 and i've found quite a few problems just by building core-image-minimal. Aug 18 08:56:53 LetoThe2nd: relocated um... almost 2 years now. Aug 18 08:57:04 I'll submit a patch serie for that at some point, but I don't have the manpower nor time to check with a bitbake world Aug 18 08:57:43 so the only way to guarentee that such bugs don't reapear would be to have some project-level testing. I don't know how to do that, but the autobuilders seemed like a reasonable place to start Aug 18 08:58:01 pidge: ah ok... last time we met there was too much talk on D&D and gaelic, little of RL value ;-) Aug 18 08:58:18 boucman_work: ok, yeah, that would be good to get that into the oe testing framework. Aug 18 08:58:31 LetoThe2nd: I will give a try changing the min-io-size to 2048 as it says line175 Aug 18 08:58:49 LetoThe2nd: Ha, I don't remember this. Aug 18 08:59:08 pidge ok, nice to see i'm not too much off the line, can I come back to you when my main patch is more mature on the subject ? Aug 18 08:59:36 pidge: OEDEM dublin, lunch. Aug 18 08:59:39 boucman_work: sure. pidge@toganlabs.com Aug 18 09:00:02 pidge: thx, or IRC Aug 18 09:00:21 LetoThe2nd: ahh, yes! Aug 18 09:00:54 pidge: probably you just don't link my face and nick. Aug 18 09:02:17 LetoThe2nd: Yeah, I know another Leto I got you confused with. Aug 18 09:03:05 pidge: np :-D i often get confused with some other Leto. Jared, methinks. Aug 18 09:25:07 I'm trying to do a versionned packagegroup, i.e a packagegroup that depends on specific version of recipes.... I can add version information in RDEPENDS, but that only works at rootfs time, it doesn't make bitbake compile the right version of the software Aug 18 09:25:35 is there an easy way to do that ? (I tried to add a version to DEPENDS, bitbake seems to accept the syntax but does not compile the proper version anyway) Aug 18 09:29:09 boucman_work: that's PREFERRED_VERSION I think Aug 18 09:29:23 but it lives at the .conf level, not at .bb level Aug 18 09:30:40 yeah, that's my problem... Aug 18 09:31:03 Don't think bb can do that... Aug 18 09:31:40 I need to set PREFFERED_VERSION for each program in the packagegroup, I'd like to just set it for the packagegroup... Aug 18 09:31:47 multiple versions can't co-exist in the same sysroot|deploy/ipk anyways Aug 18 09:32:37 true, but bb can have multiple versionned recipe, so versionned depends could be theoretically possible (with the correct error messages in cas of contradictory dependencies) Aug 18 09:32:57 but thx, if it's not possible, I won't look further on how to do it :P Aug 18 09:33:12 RDEPENDS still guarentees version consistence, which is good Aug 18 09:45:10 hrm. is there a way in combo-layer to specify a revision to peg an entry to. so when you combo-layer update, it won't update just that repo? Aug 18 09:53:48 RP, I guess you'd know the answer to this maybe? ^^^ Aug 18 09:56:52 pidge: offhand I don't remember Aug 18 09:57:33 if there is it's not documented. I'll poke through it I guess. Aug 18 10:14:31 pidge: the script itself is pretty simple Aug 18 10:18:09 By default the apache2 recpie out of yocto poky meta-www seems not to enable its systemd server Aug 18 10:18:15 How can I enable this service? Aug 18 10:18:45 Yeah I know with systemctl enable apache2. But I talk about the package generation runtime Aug 18 10:29:38 HyP3r: http://www.yoctoproject.org/docs/2.1/ref-manual/ref-manual.html#ref-classes-systemd Aug 18 10:29:44 "Services are set up to start on boot automatically unless you have set SYSTEMD_AUTO_ENABLE to "disable". " Aug 18 10:30:03 presumably the apache2 recipe sets that variable? Aug 18 11:25:26 alexlarsson: hey alex Aug 18 11:26:44 alexlarsson: the tune etc i wasn't aware was in CC but the sysroot certainly always is. this has bitten me before though and i have wondered if we should change CC to be literally the gcc name even if its entirely useless without CFLAGS (to get the sysrot) Aug 18 11:28:02 rburton: hey Aug 18 11:28:23 do you know where I should put a patch that affects ALL version of linux ? Aug 18 11:29:01 (it is not possible to overload the hardcoded name "gcc" used to compile the kconfig stuff, thus breaking BUILD_*) Aug 18 11:38:44 boucman_work: you can patch everything using linux-yocto by simply adding to SRC_URI right? Aug 18 11:40:19 yes, I guess that will be good enough for my goal. other meta will hav to fix their stuff themselves... Aug 18 12:18:06 joshuagl: you were right inside this apache2 script stands: SYSTEMD_AUTO_ENABLE_${PN} = "disable" Aug 18 12:19:04 How can I enable this service? Aug 18 12:19:10 with a bbapend file? Aug 18 12:22:34 yup Aug 18 12:23:34 HyP3r: or setting SYSTEMD_AUTO_ENABLE_pn-apache2 = "enable" somewhere in your distro's conf should work Aug 18 12:23:49 (I think) Aug 18 12:24:25 Well I have anyway allready a meta layer for my project and also a apache2.bbappend which should fix this Aug 18 12:24:34 But I'm still wondering why its by default disabled Aug 18 12:28:11 rburton: I don't really know unfortunately, it was always various points of jethro Aug 18 12:28:18 And I'm wondering why yocto is not using any chroot enviroment each complex package has its patches to fix path problems \o/ Aug 18 12:28:27 rburton: but, it used to work, and at some time broke Aug 18 12:28:43 rburton: not sure exactly what caused the change Aug 18 12:29:14 alexlarsson: hmm Aug 18 12:29:49 rburton: the host didn't change (centos7) Aug 18 12:29:58 so, shouldn't be the host compiler Aug 18 12:30:20 but we did bump the jethro rev a few times Aug 18 12:30:37 unfortunately i was on vacation so i don't really know when it appeared, etc. Aug 18 12:31:18 didn't spot anything in jethro which would cause this, i'll look again Aug 18 12:32:07 Its kinda weird though Aug 18 12:32:26 It happens in g-ir-scanner, but only when called from webkit Aug 18 12:32:31 alexlarsson: i though CC had the sysroot in for a long time, so maybe distutils changed? Aug 18 12:32:52 alexlarsson: maybe you could do some debugging with 'bitbake -e' if you search for "^export CC" Aug 18 12:33:19 i'd suggest patching distutils to split on whitespace before doing basename Aug 18 12:33:54 though as i said, i've seen a few other pieces moan if CC isn't just a binary name, so i do wonder how much breaks if we move the sysroot from CC to CFLAGS Aug 18 12:33:56 (my guess: lots) Aug 18 12:34:15 several packages failed to build when we made not listening to LDFLAGS throw a warning Aug 18 12:37:30 rburton: distutils is part of python, right? Aug 18 12:38:29 rburton: the weird thing is that all my other packages build fine, and i'd be surprised if webkit is the first that uses g-ir-scanner Aug 18 12:38:59 rburton: which is very confusing, because the distutils issue is very clear, looking at the distutils code Aug 18 12:39:34 hi guys Aug 18 12:42:08 aaand glibc git refuses to compile :( Aug 18 12:42:08 alexlarsson: oh webkit build from oe-core? Aug 18 12:42:20 rburton: no, in flatpak Aug 18 12:42:24 ah right Aug 18 12:42:44 i was going to say, g-i should be disabled everywhere in jethro Aug 18 12:42:47 python -c "import sysconfig; print sysconfig.get_config_var('CC')" Aug 18 12:42:48 x86_64-unknown-linux-gcc -m64 -march=core2 -mtune=core2 -msse3 -mfpmath=sse Aug 18 12:42:58 that is on my build Aug 18 12:44:18 can anyone tell me where the syslog file should be? i'm trying to output all systemd logs to syslog but it's not working Aug 18 12:45:14 ForwardToSyslog=yes is defined in my journald.conf Aug 18 12:49:00 rburton: do you know where the sysconfig is stored? Aug 18 12:49:42 no, sorry Aug 18 12:49:53 knowing python, blasted into the module directly Aug 18 12:53:29 rburton: /usr/lib/python2.7/_sysconfigdata.py it seems Aug 18 12:54:33 when attaching to syslog process with strace, i can see all the logs coming from journald, but there is no output Aug 18 12:55:39 Hm another strange thing: I have added openvpn to my core image. And openvpn has this line: "RRECOMMENDS_${PN} = "kernel-module-tun"". But I don't have a kernel module 'tun' Aug 18 12:56:08 Here the modinfo: "modinfo: ERROR: Module tun not found." Aug 18 12:58:39 HyP3r: my understanding : RRECOMMENDS is not mandatory, so if it's not found, it won't be an error at build time Aug 18 12:59:01 so if noone builds kernel-module-tun, it won't be build Aug 18 12:59:17 you need to add kernel-module-tun to IMAGE_INSTALL yourself, I guess Aug 18 12:59:50 rburton: hmm, i think i may have an idea Aug 18 13:00:20 boucman_work: ok I'll try Aug 18 13:01:00 rburton: looking at my build, it seems like the _sysconfigdata.py is "right" (has no --sysroot) Aug 18 13:01:11 rburton: but the _sysconfigdata.pyc *does* have it Aug 18 13:01:40 rburton: and then we run the entire thing through ostree, reseting the mtimes, and we have some python mtime fixup thing for this in flatpak Aug 18 13:02:04 rburton: so, i wonder if yocto rewrites this file after the .pyc is written Aug 18 13:03:03 huh Aug 18 13:03:32 yes Aug 18 13:03:41 py_package_preprocess in python.bb Aug 18 13:03:59 though that is meant to be removing the sysroot... Aug 18 13:04:00 python_2.7.9.bb: sed -i -e 's:--sysroot=${STAGING_DIR_TARGET}::g' -e s:'--with-libtool-sysroot=${STAGING_DIR_TARGET}'::g \ Aug 18 13:04:10 yes, and it does Aug 18 13:04:15 in the .py file Aug 18 13:04:25 but its left in the .pyc file Aug 18 13:04:38 which admittedly should be invalid due to the timestamp Aug 18 13:05:01 so, the fault is mine Aug 18 13:05:10 so its a comedy interaction between our recipe expecting the pyc to be regenerated/ignored, and you adjusting timestamps? Aug 18 13:05:20 yeah Aug 18 13:05:28 boucman_work: I added this to my IMAGE_INSTALL but there is no such package :( Aug 18 13:06:25 surely we can remove those bits earlier than that preprocess hook Aug 18 13:06:42 well, its a general problem for me though Aug 18 13:06:49 i should handle it as such Aug 18 13:06:55 or i will run into similar issues later. Aug 18 13:07:06 recompile all py before trashing timestamps? Aug 18 13:07:51 Or just drop all .pyc files that have an invalid timestamp Aug 18 13:08:41 presumably there's a good improvement in loading .pyc over .py so especially for the core library a forced compile might be sensible Aug 18 13:10:06 boucman_work: then you have to make sure your kernel config even builds that module Aug 18 13:11:10 LetoThe2nd: yeah, true Aug 18 13:11:23 rburton: http://nedbatchelder.com/blog/201310/finding_stale_pyc_files.html Aug 18 13:11:24 boucman_work: ah sry Aug 18 13:11:31 HyP3r: that was for you ^^^^^^^ Aug 18 13:11:44 HyP3r: what LetoThe2nd said... I'm not sure the proper, yocto, way to enable a module, but that's what you need to do Aug 18 13:12:05 boucman_work: as usual, it depends. Aug 18 13:12:15 alexlarsson: find all .pyc which are older than .py? Aug 18 13:12:19 would be a sensible sanity check Aug 18 13:12:24 * rburton adds to insanitier Aug 18 13:16:29 boucman_work: yeah I guess I have to add this kernel module Aug 18 13:16:49 HyP3r: boucman_work: there are the various fragments that can also be depended on IIRC or be set up through DISTRO, or if its a simpler setup, just modify the kernels .config Aug 18 13:18:34 To the apache thing back, while compiling the kernel module tun: I have created a apache2.bbapped (which im already using to patch some configurations) and wrote the line: SYSTEMD_AUTO_ENABLE_${PN} = "enable" Aug 18 13:18:44 But this service is still not eanabled :( Aug 18 13:20:26 did you misspell bbappend as you did above? :-) Aug 18 13:20:49 HyP3r: the name of the bbappends has to completely match the name except for the extension Aug 18 13:21:12 HyP3r: there are a couple of wildcards, but apache2.bb... is certainly not correct Aug 18 13:21:37 The exact name of my bbappend: "apache2_2.4.10.bbappend" Aug 18 13:21:52 The original file: apache2_2.4.10.bb Aug 18 13:22:51 bitbake apache2 -e | grep ^SYSTEMD_AUTO_ENABLE ? Aug 18 13:23:04 HyP3r: then first check if the systemd unit is not packed into a seperate package. second, evaluate the recipe as joshuagl just said, he was faster on typing :) Aug 18 13:23:20 k Aug 18 13:23:29 not sure since I only saw some of this discussion, but maybe this will be helpful: https://wiki.yoctoproject.org/wiki/Technical_FAQ#I.27ve_created_a_bbappend_for_a_recipe_but_what_I.27m_setting_there_isn.27t_having_any_effect.2C_how_do_I_fix_this.3F Aug 18 13:23:56 bluelightning: any documentation is useful if correct, be it only for later reference :) Aug 18 13:24:10 LetoThe2nd: good point :) Aug 18 13:33:20 joshuagl: LetoThe2nd http://pastebin.com/Hcd6REH3 looks good right? Aug 18 13:33:46 I have now enabled TUN/TAP inside the kernel configuration. How can I check if its enabled Aug 18 13:34:37 HyP3r: build the kernel and then see if the package is generated Aug 18 13:34:58 ok, its enough to save the config as '.config' right? Aug 18 13:36:14 HyP3r: yeah, that looks OK. Did `bitbake apache2` result in a new package being generated? (That variable is used in a postinst that runs at first boot) Aug 18 13:37:29 HyP3r: erm.... no, not really. especially not if you want things to be reproductible and you are using OE to build the kernel Aug 18 13:38:12 HyP3r: you have to make sure that the new config file ends up at the place where your former one was fetched from Aug 18 13:43:49 joshuagl: well I gues I have to to a clean for that Aug 18 13:49:26 hello. How can I find out which u-boot defconfig is used to build my u-boot.img? Aug 18 13:50:51 the u-boot-var-6ul_2015.10.bb recipe file that is apparently being used does not directly provide such info Aug 18 13:56:08 eduardas_m: IIRC uboot does not rely on defconfigs, but its own machines. thats what the uboot machine name in the recipe selects. with that information, look at the uboot sources. Aug 18 14:01:34 LetoThe2nd: this is not to be confused with linux kernel defconfig which is a separate matter entirely Aug 18 14:01:59 when building u-boot one can do something like the following: Aug 18 14:02:12 make ARCH=arm CROSS_COMPILE=${CC} wandboard_defconfig Aug 18 14:02:28 ins the u-boot sources there is a "configs" folder Aug 18 14:02:43 eduardas_m: i know. Aug 18 14:02:55 and the file names there end with "defconfig" Aug 18 14:04:03 LetoThe2nd: the config in kernel-build-artifacts seems ok Aug 18 14:04:08 so I have to use this file when I compile from u-boot sources separately, but Yocto does not utilise them? Aug 18 14:04:19 and yes... the apache2 has just to be recompiled :S Aug 18 14:04:24 or just repacked Aug 18 14:04:52 eduardas_m: yocto does not utilize anything, as it is no software :-P Aug 18 14:05:27 eduardas_m: but the usual bitbake recipes for u-boot just trigger the very same config step as a manual build does. Aug 18 14:06:24 but how do I trace that from the .bb file? Aug 18 14:07:06 LetoThe2nd: I want to do manually what Yocto does for me automatically Aug 18 14:07:34 eduardas_m: doesn't your recipe say something like U_BOOT_MACHINE = "something"? or one of the includes it pulls in? Aug 18 14:09:18 eduardas_m: http://www.yoctoproject.org/docs/2.2/dev-manual/dev-manual.html#platdev-appdev-devshell might be helpful Aug 18 14:10:23 eduardas_m: and http://www.yoctoproject.org/docs/2.2/ref-manual/ref-manual.html#usingpoky-debugging-viewing-logs-from-failed-tasks will also show you how to see exactly what commands bitbake runs Aug 18 14:10:26 LetoThe2nd: https://github.com/varigit/meta-variscite-mx6ul-mx7/blob/imx_3.14.52_6ul_ga-var01/recipes-bsp/u-boot/u-boot-var-6ul_2015.10.bb Aug 18 14:10:37 this is the .bb file Aug 18 14:10:53 there is no U_BOOt_MACHINE Aug 18 14:11:58 The docs suggests SSTATE_MIRRORS ?= "file://.* ...". What is the significance of this file URL? Aug 18 14:12:25 eduardas_m: you could run 'bitbake -e u-boot-var-6ul' and search for U_BOOT_MACHINE. might be from u-boot.inc or fsl-u-boot-localversion.bbclass if it exists. Aug 18 14:12:37 Ulfalizer: ++ Aug 18 14:12:38 'bitbake -e' also shows you where variables are set Aug 18 14:12:51 * Ulfalizer tries giving general advice with zero context :P Aug 18 14:13:35 That's 42 isn't it? Aug 18 14:13:58 42 is the answer Aug 18 14:17:24 Ok great apache2 starts now automatic but the kernel module 'tun' is not aviable :/ Aug 18 14:18:17 under kernel-build-artifacts I have the .config file for my kernel (I gues this is the correct directory). And inside this file I have the line "CONFIG_TUN=m" Aug 18 14:18:36 (I created it as kernel module, maybe not so good idea) but 'modprobe tun' does not work Aug 18 14:19:08 sveinse: guessing it's consistent with e.g. PREMIRRORS where you can have things like "git://./.* https://some.com/location" to map the former to the later, only file:// is the only thing that makes sense for the sstate cache Aug 18 14:19:32 and yeah, that should be explained in the documentation, imo Aug 18 14:21:32 HyP3r: you have to check for: http://www.yoctoproject.org/docs/2.1/mega-manual/mega-manual.html#var-UBOOT_MACHINE Aug 18 14:23:57 Ulfalizer, LetoThe2nd : it is now apparent that in poky sources u-boot.inc actually uses UBOOT_MACHINE for make...now I only need to know where UBOOT_MACHINE is set Aug 18 14:24:54 LetoThe2nd: UBTOO? Aug 18 14:26:21 HyP3r: sorry, wrong nick Aug 18 14:26:27 eduardas_m: ^^ Aug 18 14:26:34 LetoThe2nd: no problem ^^ Aug 18 14:26:46 LetoThe2nd: I guess I missed after menuconfig the savedefconfig \o/ Aug 18 14:29:49 HyP3r: that might work as a single shot, but its really important that the config finds its way back to the place where the original one came from. otherwise you will be stuck instantly again the next time you want to modify anything Aug 18 14:31:48 ok? But whats the better way? Aug 18 14:34:50 eduardas_m: 'bitbake -e' tells you where variables are set too, not just their value Aug 18 14:34:56 HyP3r: *sigh* like i said: make sure the config file you created ends up exactly where the original file was before. e.g. in the layer, in the repo, whatever. Aug 18 14:36:02 well I gues the kernel.bbclass is more complex than simply coping my config file (which contains every configuration point of linux) to there, and your answer is really universal Aug 18 14:36:32 Ulfalizer, LetoThe2nd: thank you for your help...was really informative... Aug 18 14:37:19 I read now something about that I have to create a 'linux-toradex_4.1.bbappend' with a SRC_URI Reference to a config file which is only configuring those things which are nessercary (in my case: tun.cfg) then it should work Aug 18 14:37:20 eduardas_m: when going through the manuals, a tip is to always read the latest version btw Aug 18 14:37:22 HyP3r: not at all. it really, really depends on your specific kernel setup. it can be as simple as just replacing some defconfig file. Aug 18 14:37:39 HyP3r: or that, the cfg way is also possible. Aug 18 14:37:40 guess you risk seeing something version-specific, but it's a small risk compared to updated descriptions Aug 18 14:38:01 And inside the tun.cfg stands: "CONFIG_TUN=y" Aug 18 14:38:06 HyP3r: always keep in mind - what would a coworker need to find tomorrow if you get hit by a bus tonight. Aug 18 14:38:07 that way is really scalable Aug 18 14:38:38 LetoThe2nd: a good documentation :). Each recpie of my meta layer has some lines documentation Aug 18 14:39:06 hopefully enough :3# Aug 18 14:39:25 HyP3r: and if that defconfig ends up in your layer, then all is fine. thats wahat i want to say. Aug 18 14:40:55 LetoThe2nd: yep this bbappend and tun.cfg is inside my layer ;) Aug 18 14:42:57 HyP3r: perfect! Aug 18 14:45:01 LetoThe2nd: only by the way: my cowork which left the company and gave me this huge pile of missconfiguration. He simply compiled the kernel out of the whole yocto toolchain and copied the zImage into the rootfs before do_rootfs Aug 18 14:47:38 HyP3r: yeah - thats why i'm saying to get it done right from the beginning Aug 18 14:59:47 LetoThe2nd: it seems like its not working. I still have "ERROR: Cannot open TUN/TAP dev /dev/net/tun: No such device" Aug 18 14:59:52 LetoThe2nd: can you help me? Aug 18 15:00:55 LetoThe2nd: I don't know how to troubleshoot this problem Aug 18 15:06:48 HyP3r, tunctl -b -u $USER Aug 18 15:07:07 that'll make you a tap device Aug 18 15:07:37 nisha: I don't have the depending kernel module integrated and also not this tunctl binary Aug 18 15:11:21 HyP3r, heh, seems I might have a similar problem with a kernel config file Aug 18 15:14:45 I have followed this tutorial: http://www.yoctoproject.org/docs/2.1/dev-manual/dev-manual.html#creating-config-fragments Aug 18 15:14:58 nisha: and I should add this tunctl which maybe useful troubleshoot this Aug 18 15:17:34 HyP3r, so you already have a .bbappends file in your recipe-kernel folder? Aug 18 15:18:39 well I have a meta layer 'meta-foo' with the directory 'recpie-bar' insside this folder I have the file 'linux-toradex_4.1.bbappend' with this as conent: http://pastebin.com/Km8bsGTK Aug 18 15:18:56 then I have inside this ' Aug 18 15:19:24 then I have inside this 'recpie-bar' folder the folder 'linux-toradex' with those two files listes in SRC_URI Aug 18 15:19:41 And tun.cfg contains 'CONFIG_TUN=y' Aug 18 15:22:02 that looks right Aug 18 15:22:51 yeah, but something is missing. I'm not sure if its allowed to use 'y' or should I use 'M'? Aug 18 15:24:12 if you use m it just means you will have to manually load it Aug 18 15:24:48 you can try doing an insmod tun.ko after you boot Aug 18 15:30:10 cgroup are awesome... I finally found how to not have bitbake eat my cpu :) Aug 18 15:35:05 Wow what a bunch of shit. I told you my configuration ~17:19 and now my system has again booted: '# CONFIG_TUN is not set' Aug 18 15:35:31 Yeah. And then people are wondering why there are so many mass shootings Aug 18 15:39:57 alexlarsson: WARNING: python-2.7.12-r1 do_package_qa: QA Issue: package python-core contains stale pyc work/corei7-64-poky-linux/python/2.7.12-r1/packages-split/python-core/usr/lib64/python2.7/_sysconfigdata.pyc [stale-pyc] <— i'll push this new test into oe-core to ensure we don't do it again Aug 18 15:43:32 HyP3r: what kernel are you building? Aug 18 15:46:40 if you are building linux-yocto you can/should make use of kernel config frags as you seem to be attempting from what I see Aug 18 15:47:36 but if you are inserting a different kernel then config frags are not applied and the 'seeded' defconfig is all that is used Aug 18 15:47:45 per http://www.yoctoproject.org/docs/2.1/kernel-dev/kernel-dev.html Aug 18 15:47:47 omg, this thing might be working. rock and roll. Aug 18 15:51:23 I think someone needs to by HyP3r a drink after this week Aug 18 15:51:26 *buy Aug 18 15:51:34 Ulfalizer, LetoThe2nd: you were of great help today..finally was able to build modified u-boot for my board with the proper configuration Aug 18 15:51:56 so, I tried to add a kernel config but I get this build error: Computing transaction...error: Can't install packagegroup-distro-base-1.0-r83@qemux86_64: no package provides kernel-module-virtio-gpu Aug 18 15:55:53 nope. Aug 18 16:03:02 so, this multiple git repot with a cmake build in bitbake is oh so close to working Aug 18 16:03:19 https://gist.github.com/netskink/643bdfbc92017b265881e600670ca906 Aug 18 16:03:34 it pulls the three repots, puts them in a single git subdir Aug 18 16:04:00 and possibly attempts to build from the src dir in one of the repots Aug 18 16:04:16 but it might actually be trying to build in git/ and not git/pcmx Aug 18 16:04:55 well you put S=git/pcmx so it should be doing cmake in there Aug 18 16:05:32 yes, that was the goal Aug 18 16:05:33 looking at cmake.bbclass shows you that it has a OECMAKE_SOURCEPATH to tell it where the cmakelists is Aug 18 16:05:38 which defaults to ${S} Aug 18 16:05:51 ahh so that is an override? Aug 18 16:06:21 well you shouldn't need it if you set S to the right place Aug 18 16:06:29 you know actually saying what breaks really would be helpful Aug 18 16:06:40 im trying to specify just in case. Aug 18 16:07:01 when I did the simple cmake tutor build in bitbake, i used jsut workdir/git Aug 18 16:07:05 and that worked Aug 18 16:08:01 i wonder if at this point its a cmake problem and I know very little about it Aug 18 16:08:58 davis: the only recipe I have worked on with multiple git repos is the recipe for spice Aug 18 16:09:03 http://git.yoctoproject.org/cgit/cgit.cgi/meta-cloud-services/tree/recipes-support/spice/spice_git.bb?h=master Aug 18 16:09:28 let me look at it, one sec Aug 18 16:09:35 not sure if that will help but having a reference is never a bad thing Aug 18 16:09:54 davis: seriously, pastebin how it breaks and what you expect it to do, it may be something trivial Aug 18 16:10:06 but just saying "it doesn't work" means nobody can help Aug 18 16:10:11 :) Aug 18 16:10:58 lol, but i'm so verbose. Aug 18 16:11:06 i updated the gist to have an error Aug 18 16:11:51 btw, if i use OECMAKE_SOURCEPATH is generates same error as without it specified Aug 18 16:12:02 yeah because its running cmake fine and its finding the cmakelists Aug 18 16:12:06 so where are the files its looking for Aug 18 16:12:28 fwiw, when i look at the error, it looks to me like its in this dir: Aug 18 16:12:36 /pcmx/1.0.0+gitAUTOINC+d3_d3v16_pcmx-r0/build/ Aug 18 16:12:45 pcmx is the name of the recipe Aug 18 16:12:51 yes, we do out of tree builds, and tell cmake where the cmakelists is Aug 18 16:13:01 look at the cmake invocation in the log you pasted Aug 18 16:13:13 the actual code would be in pcmx/xxxxgitxxxr0/git/pcmx Aug 18 16:14:06 right, and its finding the cmakelists fine Aug 18 16:14:13 because its reading that and then saying it can't find other pieces Aug 18 16:14:22 so where are they and where is it actually looking Aug 18 16:15:10 its hard for me to say, let me show in pastebin. it looks like it trying to build pcmx seperate Aug 18 16:15:13 one sec Aug 18 16:15:52 basically we're out of "yocto" and into "your code" Aug 18 16:16:12 yeah that might be it Aug 18 16:16:14 i'm mistaken Aug 18 16:16:29 the other subdir there is not for pcmx src its for timestamps and such Aug 18 16:17:14 so, if I make a subdir say foo and then cd to foo and git pull the three repots Aug 18 16:17:41 do a cd foo/pcmx/ and then do cmake . it builds Aug 18 16:17:57 so try mkdir foo/build ; cd build ; cmake ../pcmx Aug 18 16:18:21 (which is all the class is doing) Aug 18 16:18:58 tenner says that fails and your cmakelists has false expectations about file layout Aug 18 16:19:05 yes,, if I do cd to tree in bitbake and do cmake . in same place as where I do this manually it fails. Aug 18 16:19:18 with the error shown in bitbake Aug 18 16:22:01 so we now know its nothing to do with yocto Aug 18 16:22:10 but your software not handling out-of-tree builds Aug 18 16:22:50 so after inherit CMAKE set B=${S} and add a comment along the lines of "fix the stupid cmakelists" Aug 18 16:25:21 Hello .. How can I mask a MULTILIB version of a recipe, instead of its whole? Aug 18 16:25:37 BBMASK does not understand multilib prefexes it seems :-( Aug 18 16:26:04 what exactly are oyu trying to achieve with bbmask? Aug 18 16:26:11 bbmask should only be used to work around parse errors Aug 18 16:28:57 kergoth, I have a recipe (provided by board manufacturerer in their own layer) which extracts binary .so files (related to graphics). These .so files are only 64-bit version (under /usr/lib64). Meanwhile MULTILIB 32-bit support is enabled, and Bitbake tries to build the 32-bit version of it Aug 18 16:29:52 And when building 32-bit version, do_install() tries to install libraries from ${libdir} (/usr/lib in 32-bit mode). This folder does not exist, and thus the build fails :-( Aug 18 16:30:40 why is bitbake trying to build the 32 bit version of it? what's pulling it into your build? if something is pulling it in, bbmask isn't going to magically fix that, it'll just error out instead Aug 18 16:31:16 kergoth, I guess that's the right question to ask .. Is there a way to know who's pulling the 32-bit version? Aug 18 16:31:17 presumably you could also disable multilibs for that recipe, so the variants don't get created for it Aug 18 16:31:30 bitbake -g is your best bet, either with or without -u depexp Aug 18 16:32:31 kergoth, great! and the second part: "disable multilibs for that recipe" .. how to do that? check for ${MLPREFIX} in a .bbappend ? Aug 18 16:49:32 marka: I've compiling the kernel linux-toradex its from the meta layer meta-toradex but this kernel also inherits the class kernel. You are posting link to the documentation. I followed this topic http://www.yoctoproject.org/docs/2.1/kernel-dev/kernel-dev.html#changing-the-configuration Aug 18 16:50:00 CTtpollard: there are so many recpies and different which makes it really hard. Aug 18 16:51:04 CTtpollard: and the customer is crying everyday into my ears why I'm not finished Aug 18 16:51:13 But ok, I'll get that today ;) Aug 18 16:56:49 This fyi: Aug 18 16:56:50 http://git.toradex.com/cgit/meta-toradex.git/tree/recipes-kernel/linux Aug 18 17:05:20 marka: well thats the answer: the recpie linux-yocto inherits kernel and kernel-yocto. This bbclass searches for cfg files and applies them. My linux-toradex only inherits kernel.bbclass, thats the reason why its not working Aug 18 17:05:43 So now for all: my custom kernel script (from toradex) which is not using that way of adding configurations. How should I do it then? Aug 18 17:05:59 With classic patches? should I just patch the .config file? Aug 18 17:06:03 (lol classic) Aug 18 17:18:05 HyP3r, a defconfig will work with something that just inherits kernel.bbclass. Aug 18 17:18:31 linux-tordax allready applies a defconfig http://git.toradex.com/cgit/meta-toradex.git/tree/recipes-kernel/linux/linux-toradex-4.1/defconfig Aug 18 17:18:54 I guess I use this :S -> http://git.toradex.com/cgit/meta-toradex.git/tree/recipes-kernel/linux/linux-toradex_git.bb#n24 Aug 18 17:19:10 aha. yes. Aug 18 17:19:29 echo "foobar" >> .config <- what a nice solution Aug 18 17:19:32 in that case, I'd say copy their defconfig, add your lines and then add it to your layer Aug 18 17:19:37 Every time I see a Dr. Dobb's article I get sad :( Aug 18 17:19:40 at least the archive is still up Aug 18 17:19:56 zeddii: wouldn't it then collied while fetiching phase? Aug 18 17:20:00 nope Aug 18 17:20:17 the layer ordering and priority will take care of it. Aug 18 17:20:33 I can't say the order precisely from memory .. so double check. Aug 18 17:21:17 HyP3r. that's why my latest series for master is a prep to extend the fragments to all kernel types. but there is a LOT of history in the config of the kernels, so it has been a multi year effort to get something reasonable. Aug 18 17:21:31 zeddii: yup, I'll copy that defconfig into my directory and run bitbake -c patch linux-toradex, I'll see whats there Aug 18 17:21:32 * zeddii has to transition, but will reappear as zeddii_home in a bit. Aug 18 17:21:59 cool. I'll be around if it doesn't work. we can sort it out and avoid any really nasty hacks :P Aug 18 17:22:03 zeddii: good driving or so :) Aug 18 17:37:53 RP: did anyone add 'unset' to the bitbake vim syntax? Aug 18 18:55:05 if i want to force a repull from git and build of the target foo, $bitbake -c do_cleanall foo will do it correct? Aug 18 18:55:49 ie. $ bitbake foo will then pull, configure and build afterwards Aug 18 19:01:41 no, cleanall doens't prevent use of sstate Aug 18 19:01:50 it'll re-run the tasks, but it might just pull from sstate rather than building Aug 18 19:02:50 bitbake -c cleanall foo; bitbake -C fetch foo; would be my recommendation. you could also do cleansstate along with cleanall, but sstate can still be used after a cleansstate when SSTATE_MIRRORS is involved. -C will force the matter regardless Aug 18 19:04:49 kergoth: many thanks Aug 18 19:04:54 np Aug 18 19:10:46 zeddii zeddii_home its colliding, and ignoreing my file -.- Aug 18 19:12:08 kegroth, sorry for resurrecting an older discussion ;-) You've mentioned earlier you can disable multilib for a certain recipe? Aug 18 19:12:15 davis: there's also bitbake's --setscene-only and --no-setscene arguments, but that'll affect any task bitbake runs, not just the ones for this recipe, so -C would be best there Aug 18 19:12:34 kegroth, thanks a lot for your advice earlier btw. I've indeed now know who pulls the undersied 32 bit versions of the library :D Aug 18 19:13:08 kergoth: rocks so much they need to rename that branch, s/krogoth/kergoth/ Aug 18 19:13:43 darwish: in theory the recipe could override MULTILIBS = "", but i've never tried it Aug 18 19:13:50 davis: hah Aug 18 19:15:25 kergoth, I see .. thx :-) Aug 18 19:15:30 np Aug 18 19:15:41 hmm. ok i had an error in my cmake project based source. Aug 18 19:16:03 oddly enough when i was building at cmdline it was using something left around from before Aug 18 19:16:54 i fixed that. Now I can definitely issue the git pull outside bitbake, do cmake -G "Unix Makefiles" and then cmake . followed by make and all builds fine Aug 18 19:17:56 however, if I go to bitbake pulled source with corrected source. at least I did -C fetch and clean, it fails. Aug 18 19:18:18 it fails for diff reason than before so at least i'm making forward progress, yay! Aug 18 19:25:36 ok, it appears that the cmake build for this proj is putting files in my home dir which it uses to build. that ain't gonna work. Aug 18 19:25:56 ha, nice! Aug 18 19:26:09 that's an all-new level of screwedup that i hadn't considered Aug 18 19:26:35 well there was some errors before where the gstreamer receipe you pointed out got me to here. Aug 18 19:27:50 Hi all. Is there a standard bitbake module recipe formula that addresses when a module needs symbols from another module? I pointed KERNEL_SRC to the kernel build's WORKDIR/build and it works. It's just very hack-y Aug 18 19:28:42 paging zeddii for IanCoolidge's question Aug 18 19:32:47 HyP3r: I ran a test here, and works as I’d expect. Aug 18 19:33:12 my 2nd defconfig is the file in ${WORKDIR} and hence fed into the kernel configuration step. Aug 18 19:34:05 in ${WORKDIR} is allready a defconfig :/ Aug 18 19:34:38 sure Aug 18 19:34:41 that’s not the point. Aug 18 19:35:02 And my defconfig is not at this position Aug 18 19:35:07 depending on the order in the SRC_URI and how you bbappended the file://defconfig, the last one is going to remain there. Aug 18 19:36:02 Can you tell me how can I get the information which bb file is exectued (or runng) when I call bitbake -c config linux-toradex Aug 18 19:36:25 In my stuff folder (with the layers inside) I have more than 5 linux-toradex*.bb files Aug 18 19:36:33 I'm slowly getting confused around that Aug 18 19:36:58 highest version or the PREFERRED one gets ran Aug 18 19:37:03 either the latest / highest version. or the one that is tagged as compatible with your $MACHINE. Aug 18 19:37:32 bitbake -e has all the info about what is being processed. Aug 18 19:38:15 None of the linux-toradex files is 'PREFERRED' set but I gues its working with MACHINE Aug 18 19:41:20 Thats a huge load of data '-e' Aug 18 19:41:33 Thats so depressing Aug 18 19:41:59 you could set it in local.conf like: PREFERRED_VERSION_linux-toradex = "1.2.3" make sure linux-toradex_1.2.3.bb is chosen. Although its redundant if you're using the highest version Aug 18 19:42:23 or tmp/work/{machine}/linux-toradex/ would show the version being built Aug 18 19:42:36 or just temporarily remove all but one :) Aug 18 19:42:55 Yeah I'm a often in this directory to see whats happens e.g. while configure Aug 18 19:42:59 there I have the 4.1-r0 Aug 18 19:43:05 But I have two recpies which is build 4.1 Aug 18 19:43:12 But with different checksums Aug 18 19:43:44 go into git and do git log Aug 18 19:43:54 or have one recipe instead Aug 18 19:44:58 I did bitbake -c configure linux-toradex went into the working diretory and took a look into the current checkout and it was fitting with another file (o.O) wait I post it here Aug 18 19:45:57 https://github.com/Freescale/meta-fsl-arm-extra/blob/daeab6d460a0163c6ae69982180cd8aa5a2c1a7c/recipes-kernel/linux/linux-toradex_4.1.bb Aug 18 19:46:33 But thats not all... wait o.O Aug 18 19:47:53 http://git.toradex.com/cgit/meta-toradex.git/tree/recipes-kernel/linux/linux-toradex_4.1.bbappend?id=eca5937408c75a997b160bd1b92e1496b07421cf Aug 18 19:48:09 Mind the directory about: http://git.toradex.com/cgit/meta-toradex.git/tree/recipes-kernel/linux?id=eca5937408c75a997b160bd1b92e1496b07421cf Aug 18 19:48:44 I guess files like linux-toradex_git.bb and so on are ignored. Only this file from Freescale is loaded but then Version changed with this file from Toradex. Aug 18 19:49:06 Because the "e6d111cd909551cec5902358db1e25dcaa8c86bb" <- hash is matching with my current version Aug 18 19:49:42 So NOT the Scripts of Toradex are working the Script from Freescale (which are more obvious) are working Aug 18 19:53:39 But over all: zeddii_home here is my current bbappend (please mind there is allready another bbappend from toradex loaded): http://pastebin.com/QPmLXaRu Aug 18 19:59:49 Or anthoer path: linux-toradex_git.bb can't be append with linux-toradex_4.1.bbappend? Aug 18 20:00:17 kergoth: no, I suspect not Aug 18 20:00:34 make it linux-toradex_%.bbappend Aug 18 20:01:01 RP: k, figured, i'll see about updating it Aug 18 20:01:35 My append file is loaded but wins not the race with defconfig anthoer defconfig (out of 3) is feteched :/ Aug 18 20:02:15 do FILESEXTRAPATHS_append += "${THISDIR}/${PN}:" Aug 18 20:02:30 Take a look at my pastebin Aug 18 20:02:35 IanCoolidge: I added this Aug 18 20:02:41 IanCoolidge: or I have this :) sorry Aug 18 20:02:59 IanCoolidge: that will not work. FILESEXTRAPATHS is : separated, not space, and += adds a space Aug 18 20:03:12 that needs to be =, and even then, _append is wrong, it'll be the lowest priority, not highest, you want prepend Aug 18 20:03:38 s/needs to be =/needs to be :=/ Aug 18 20:03:59 := forces immediate expansion, which makes THISDIR resolve to the subdir of the append, otherwise it'll be the subdir of the recipe Aug 18 20:04:58 I gues I don't win the race because of that http://pastebin.com/QPmLXaRu vs. http://git.toradex.com/cgit/meta-toradex.git/tree/recipes-kernel/linux/linux-toradex_4.1.bbappend?id=eca5937408c75a997b160bd1b92e1496b07421cf Aug 18 20:05:01 AH, you're right Aug 18 20:05:18 yeah, I use FILESEXTRAPATHS_append := "${THISDIR}/${PN}:" Aug 18 20:05:25 i'm so used to just +='ing everything Aug 18 20:05:28 and my meta folder has the name 'meta-racechip' and the other is 'meta-toradex' so the with toradex wins -.- Aug 18 20:05:59 HyP3r: layer priority is what matters here. Aug 18 20:06:10 Where can I configure that? Aug 18 20:06:15 layer.conf Aug 18 20:06:18 k Aug 18 20:06:36 http://www.yoctoproject.org/docs/2.1/dev-manual/dev-manual.html#prioritizing-your-layer Aug 18 20:06:54 indeed Aug 18 20:08:04 Only to make sure: the layer racechip and toradex is loading defconfig who who should have the higher prioroty to win? Aug 18 20:09:31 highest priority gets preferred over lower priority Aug 18 20:10:04 whats a high priority a low or high value? I've saw systems where 0 is the highest prio. Aug 18 20:10:46 that wouldn't make sense here. in yocto we have no idea what future layers might exist, which may well need to override the core. it needs to grow toward higher priority, not toward lower Aug 18 20:10:48 high = high Aug 18 20:11:06 okay thanks :) Aug 18 20:14:16 Well I guess this night I will run a bitbake -c cleanall and a full new build, after this priority change I'm feeling not really well Aug 18 20:14:50 if you want to know which defconifg will be used, just bitbake -e yourrecipe | grep FILESEXTRAPATHS=, make sure the layer you want to be used is before the other in the list Aug 18 20:18:05 Yeah. Priority fixed that. Now its working thanks kergoth IanCoolidge and zeddii_home :) Aug 18 20:18:16 good news. Aug 18 20:18:19 I had looked away. Aug 18 20:18:23 glad you got it sorted. Aug 18 20:18:38 What a akward include strucrute, but I see that the developer has cleaned that Aug 18 20:19:26 http://img.pr0gramm.com/2016/08/10/8cef4d8d3406e16d.jpg Aug 18 20:19:50 there’s lots of rope out there. and lots of questionable layer structures out there. but the flexibility is the power and bane of the system :D Aug 18 20:23:17 I gues it make sence to build a network - device - flashing infrastructure, I have the bad feeling that this is not the last time yocto compiling my kernel and so on Aug 18 20:54:07 Hello Aug 18 20:54:18 Is there a way to remove stuff from IMAGE_INSTALL .. rather than just append to it? Aug 18 21:01:14 darwish: you can copy the recpie and remove the lines? I know thats not a nice solution Aug 18 21:02:17 HyP3r, yeah .. I wished there would be a way to do so like the _append mechanism :-( Aug 18 21:02:50 ah ok Aug 18 21:02:59 OK .. Now imagine I'm depending on a silicon provider recipe & image .. and that image contains known buggy packages that I want to remove .. what's the best way to not include them? Aug 18 21:03:09 BBMASK the hell out of them? :D Aug 18 21:03:29 IMAGE_INSTALL_remove in local.conf, maybe Aug 18 21:03:35 IMAGE_INSTALL is just a variable Aug 18 21:03:41 so yah, what IanCoolidge says will work. Aug 18 21:03:56 oh .. didn't know there's a _remove suffix .. will recheck the docs! Aug 18 21:14:03 so it appears this cmake build needs fully qualified pathnames. It builds stuff in /home/foo and /home/goo which it then uses to do the rest of its build. I've tried to dork with it to do relative makes but it fails. I give up on that. I have managed to get it work with /tmp/foo type mechanism. However this seems to be a problem in bitbake. Is this even possible? have a fully qualifed dir to build things, Aug 18 21:14:09 then use that dir to build other things? Aug 18 21:15:45 Can you describe what you're trying to do? I can't figure it out from what you wrote. Aug 18 21:16:14 cmake handles out of tree builds just fine. patch that buildsystem to do something that makes sense and move on Aug 18 21:16:35 all cmake.bbclass does is runs cmake. there's no magic. Aug 18 21:18:19 well, its not working as is. Aug 18 21:18:54 if I use the source in the build/tmp... dir and do $ cmake -G "Unix Makefiles"; make it will build Aug 18 21:18:57 as i said, the buildsystem is doing something stupid, patch cmakelists to behave the way every other sane buildsystem does Aug 18 21:19:07 but its using files in /tmp/foo Aug 18 21:19:45 yah, I am just trying to patch up something which works outside bitbake to be bitbake compliant Aug 18 21:20:10 and I get an error in bitbake which I don't otherwise. Aug 18 21:20:50 i think if I can get it to stop doing fully qualified subbuilds this will work. Aug 18 21:21:55 again, all bitbake does is runs cmake. there's no magic. but it does expect everything to build in ${B} against source in ${S}. if it's assuming stupid ahrdcoded paths, then that needs to be fixed Aug 18 21:22:13 kergoth: I agree. i'm 100% on that. Aug 18 21:22:26 because using /tmp is not working. Aug 18 21:23:23 so one question though. i was trying to see if cmake was doing a $cmake . or a $camke -G "Unix Makefiles" though Aug 18 21:23:42 i could not, but I found a variable in cmake.bbclass Aug 18 21:24:06 EXTRA_OECMAKE Aug 18 21:24:34 I tried to set that to do this, but that did not change my situation. Aug 18 21:25:25 the log.do_configure shows the results of cmake command, but not the command which generated the file. Aug 18 21:26:30 thre is a run.do_configure script Aug 18 21:26:37 I'm guessing its in there somewhere. Aug 18 21:26:50 https://github.com/openembedded/openembedded-core/blob/master/meta/classes/cmake.bbclass#L107 Aug 18 21:27:00 https://github.com/openembedded/openembedded-core/blob/master/meta/classes/cmake.bbclass#L2 Aug 18 21:27:05 it's running cmake ${S} Aug 18 21:27:19 one sec, let me look at your links Aug 18 21:27:26 it cds to ${B}, then runs cmake ${S} Aug 18 21:27:32 same as you'd do outside of bitbake to build in a separate build dir Aug 18 21:27:39 you cd to the build dir, then cmake /path/to/source/tree Aug 18 21:27:41 then make Aug 18 21:27:44 exactly what bitbake does Aug 18 21:27:54 its' not passing -G "Unix Makefiles", presumably because that's default behavior Aug 18 21:28:07 the links are just lines in cmake.bbclass.. Aug 18 21:28:34 EXTRA_OECMAKE += "-G 'Unix Makefiles'" would work fine, if you think that's a factor, but i doubt it Aug 18 21:29:05 yeah, well in this cmake setup ive tried to do $cmake . and outside of bitbake and it fails. however, if I do $cmake -G "Unix Makefiles" it does. Aug 18 21:29:19 you removed '.' from the latter? Aug 18 21:29:20 so perhaps that is something which will also need to be fixed. Aug 18 21:29:28 no, i removed nothing. Aug 18 21:29:37 you just wrote two commands, but -G wasn't hte only change Aug 18 21:29:43 if its out of tree that is breaking then just set B=${S} *after* the inherit cmake Aug 18 21:29:54 'cmake .' and 'cmake -G "Unix Makefiles"' -> the latter doesn't have the source tree path Aug 18 21:29:55 indeed Aug 18 21:29:56 just in an regular shell. this works $ cmake -G "Unix Makefiles"; make Aug 18 21:30:13 you removed .. so you made two changes to the command, you added the -G, and you removed the path to teh source tree Aug 18 21:30:27 but yes, as rburton says, if the problem is the out of tree build, just set B = ${S} and move on Aug 18 21:36:32 hhmm. if I do "inherit pkconfig cmake" and on next line B="${S}" as well as remove EXTRA_OECMAKE variable mod it fails. After I make the change, I'm doing this to test it. $ bitbake -c cleanall pcmx; bitbake -c fetch pcmx; bitbake pcmx Aug 18 21:38:14 https://gist.github.com/643bdfbc92017b265881e600670ca906.git Aug 18 21:38:19 in more concise terms Aug 18 21:40:49 i presume the comment about how the build should be in d3 is wrong as you set S to pcmx Aug 18 21:41:27 so now bitbake is just going into git/pcmx and running "cmake ." Aug 18 21:41:29 literally that's it Aug 18 21:42:06 if you can make that work outside of bitbake then i guess you want to start looking at the toolchain file the class writes and how that could be breaking your cmakelists Aug 18 21:42:07 yes, that is an error in the comment. Aug 18 21:42:15 it should be pcmx Aug 18 21:42:27 by moving that B= line after inherit Aug 18 21:43:21 it has the same error, but now when I try to build in shell ouside bitbake environment in the build dir it complains about c compiler Aug 18 21:43:31 so that is doing something Aug 18 21:44:15 so important test is do some clean checkouts, go into pcmx, run cmake . Aug 18 21:44:54 im in middle of clean/fetch/build now. when completes. ill try Aug 18 21:44:57 also, try out of tree build from clean with a build alongside the clones, cd into it and run cmake ../pcmx Aug 18 21:45:40 if those work then out-of-tree is fine and basically you've either got a really bad cmakelists, or our toolchain file (which tells cmake what compiler to use) is upsetting your cmakelists (which implies you've a bad cmakelists) Aug 18 21:46:28 so you guys have been saying out of tree builds. I might not have clear understanding. I think you mean, if you are using bitbake and it puts files in build, get a new shell and cd to build/tmp-glibc/....pcmx/../git/pcmx and do the command there. Aug 18 21:46:59 an out of tree build is simply where you make puts the compiled objects into a directory that isn't where the sources are Aug 18 21:47:34 so the WORKDIR/build directory that cmake.bbclass uses, where the cmake generated files and eventually the binaries end up Aug 18 21:47:36 this is the soruce dir ~/setup-scripts/build/tmp-glibc/work/corei7-64-oe-linux/pcmx/1.0.0+gitAUTOINC+d3_d3v16_pcmx-r0/git/pcmx$ Aug 18 21:47:45 they're good because it means we can delete it automatically on rebuilds Aug 18 21:47:50 i thought it was the build dir as well. Aug 18 21:48:18 that is where i was trying to do build outside of bitbake Aug 18 21:48:40 i'd do tests outside of bitbake with entirely fresh checkouts Aug 18 21:49:32 outside of bitbake just in a unix shell, $cmake . fails. $cmake -G "Unix Makefiles" works. Aug 18 21:49:45 impressively broken cmakelistss Aug 18 21:49:49 so it looks like bitbake is trying to do cmake . Aug 18 21:49:52 what about just cmake Aug 18 21:49:54 without the -G Aug 18 21:50:31 (because that is the default value) Aug 18 21:51:00 $cmake says usage error. i belive that will fail on the cmake tutorial source as well. Aug 18 21:51:18 ok, i stand corrected Aug 18 21:51:27 now if I do $cmake . it works. Aug 18 21:51:30 i *bet* your cmakelists expects the build dir to be a subdirectory of the top level of the sources Aug 18 21:51:43 oh right Aug 18 21:51:48 then again, i've dorked with the cmake file today in this sample repot Aug 18 21:52:00 trying to fix things. Aug 18 21:52:19 cmake defaults to doing builds inside a directory under the source tree Aug 18 21:52:36 we move it for ease of cleaning and no sane cmakelists will care Aug 18 21:53:00 im just trying to make sure, i don't have any artifiactgs in /home/davis and /tmp Aug 18 21:53:02 but if the cmakelists is assuming that ../ is the source tree then all sorts will fail Aug 18 21:53:05 they all look clean Aug 18 21:53:18 let me get a new shell as well, since it might have something in environment Aug 18 21:53:55 literally impossible to debug software via irc and pastebin, hooray open source! Aug 18 21:53:57 * rburton -> bed Aug 18 21:54:34 sweet, so yes indeed if I open new terminal, ensure /tmp and /home have no artifacts. cmake . works so that is good. Aug 18 21:54:39 rburton: goodnight Aug 18 21:56:09 try an out of tree build to see if that works. mkdir build alongside your checkouts, cd into it, cmake ../pcmx Aug 18 21:56:43 if that works then congratulations you've reduced it down to bad interaction between our toolchain file generated by cmake.bbclass and your source Aug 18 21:56:55 yes, if I do clean git checkout, cmake .; make it builds Aug 18 21:57:09 ie. seperately from bitbake Aug 18 21:57:19 it will put stuff in /tmp/jfdinstall though Aug 18 21:57:43 cmake will put files in /tmp/jfdinstall that it uses to complete cmake step. Aug 18 21:57:59 for no sensible reason as there's a perfectly good build directory to use Aug 18 21:58:02 anyway Aug 18 21:58:13 have fun Aug 18 21:58:15 yes, when they told me they were using cmake Aug 18 21:58:26 i said, we can strip out cmake and make our life easier Aug 18 21:58:36 cough feel free to leak the software online cough Aug 18 21:58:44 yah, have a good night Aug 18 21:58:57 sleep well. im going to bolt to do homework soon. Aug 18 22:00:26 davis: cmake works fine if it's used correctly, it's incorrect use of cmake that puts stuff in /tmp/jfdinstall; no sane CMakeLists.txt will do that. Aug 18 22:00:41 neverpanic: i agree. Aug 18 22:01:27 i tried to change the cmakelist.txt to use relative build settings for some of the tools. ie. XXX=../jfdinstall and it did not work Aug 18 22:01:53 why is it that I set up my environment yet I am constantly missing packagegroups Aug 18 22:02:08 however the deeper parts of this build work, it seems to require fully qualifed pathnames. Aug 18 22:02:11 one of the packagegroups I needed I had to copy and paste a changelog because I couldn't actually find the file Aug 18 22:02:38 davis: you shouldn't try to use relative paths, use ${CMAKE_BINARY_DIR} and other appropriate CMake variables Aug 18 22:07:40 why do I need a packagegroup from openembedded-core? I thought I only needed the meta-openembedded but core-image-minimal is saying I'm missing dependencies from it Aug 18 22:09:49 it's saying the dependency is unbuildable... Aug 18 22:11:50 neverpanic: the cmakelists.txt file i'm editing is setting CMAKE_BINARY_DIR and corresponding instal dir to be /home/foo Aug 18 22:12:16 i changed that to /tmp/foo since ../foo and ./foo did not work. Aug 18 22:14:14 does anyone know anything on getting unbuildable error on a packagegroup. I can't find anything Aug 18 22:15:12 davis: A CMakeLists.txt should never set CMAKE_BINARY_DIR; CMAKE_BINARY_DIR should always be the current working directory when calling CMAke Aug 18 22:15:30 A CMakeLists.txt should also never set CMAKE_INSTALL_PREFIX, that's for users to set. Aug 18 22:15:46 SET(CMAKE_INSTALL_PREFIX ./tmp/JFDInstall/PCMX) Aug 18 22:15:47 SET(CMAKE_BINARY_DIR ./tmp/JFDInstall/development/PCMX/build) Aug 18 22:16:03 i hear what you are saying. i'm just working with an existing setup Aug 18 22:16:16 if I use ./tmp it fails. /tmp it works. Aug 18 22:16:19 Girafferson: You're not providing any of the information required to debug this. You didn't pastebin an error message. Nobody can help you with this. Aug 18 22:16:22 im trying to figure it out. Aug 18 22:16:28 davis: Just delete those two lines. Aug 18 22:16:40 And tell whoever wrote them to read the CMake documentation Aug 18 22:16:47 neverpanic: lol Aug 18 22:16:51 oh man, lol Aug 18 22:17:10 neverpanic, http://pastebin.com/hQDTTB55 Aug 18 22:17:27 But then again, since they set those two files, it's likely that they actually rely on those paths elsewhere :/ Aug 18 22:18:31 yah cmake works, but make immediately tries to create a dir /usr/local/foo Aug 18 22:19:55 davis: that's likely because the CMakeLists.txt uses CMAKE_INSTALL_PREFIX where it should be using CMAKE_BINARY_DIR Aug 18 22:21:17 Girafferson: crda is in meta-networking, see http://layers.openembedded.org/layerindex/branch/master/recipes/?q=crda Aug 18 22:21:29 You're right that this is part of meta-openembedded. Aug 18 22:21:44 However, poky master doesn't have this dependency in its packagegroup-core-boot: http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/recipes-core/packagegroups/packagegroup-core-boot.bb Aug 18 22:22:16 So either some other layer sets one of the variables in this recipe to include crda, or the file looks different for you (because you're on a different branch for example) Aug 18 22:22:16 i don't understand why I have to include each meta from meta-openembedded Aug 18 22:22:56 Use bitbake -e packagegroup-core-boot and figure out where the crda in RDEPENDS_${PN} comes from Aug 18 22:23:38 You shouldn't have to; it might be your bsp that forces you to. Aug 18 22:26:32 neverpanic: imm tryinhg your suggestion Aug 18 22:26:45 neverpanic: btw are you German? Aug 18 22:28:29 Yes; which means it's half past midnight, and since I have to work tomorrow I'll likely leave soon. Aug 18 22:28:45 ok thank you; I just added in meta-networking and meta-python. I still don't understand the meta-openembedded layer but I'll have to look at it once I finish the build Aug 18 22:29:06 ok very good. i worked in germany. schon platz. Aug 18 22:29:34 have a good night. i'm fixing up the enumerous make+directory entries now. Aug 18 22:30:59 lol, its building out of source now, with relative dir. Aug 18 22:31:13 its getting closer, many thanks. Aug 18 22:31:27 you're welcome. Good night and happy debugging. Aug 18 22:31:43 ghuten nach Aug 18 22:31:54 dankeschön Aug 18 22:32:11 bitteschon Aug 18 23:12:10 2 Aug 18 23:13:16 abelal: this patch looks ok Aug 18 23:13:25 ross will pick it up Aug 18 23:14:41 jubr: no, it wont work from config context Aug 18 23:35:34 so, a heads-up for people, yes, I'm aware of the Exit Status 4 messages from pseudo, I think the message is a harmless diagnostic and probably doesn't need to be displayed (the startup code had some fancy stuff in it because I was debugging it, and since it didn't show up under light load, I didn't look too closely), but I'm gonna try to double-check. **** ENDING LOGGING AT Fri Aug 19 02:59:58 2016