**** BEGIN LOGGING AT Thu Jun 25 02:59:59 2015 Jun 25 06:43:01 Hello Jun 25 06:43:24 When building an image with IMAGE_TYPES = "iso", I've an error : Jun 25 06:43:26 Missing or unbuildable dependency chain was: ['core-image-minimal', 'syslinux'] Jun 25 06:43:36 How do I add syslinux ? Jun 25 06:44:16 (IMAGE_FSTYPES, not *TYPES) Jun 25 06:47:56 Metanaou: its not about building syslinux, its about finding out why its not found or cannot be built. Jun 25 06:49:40 Metanaou: and a quick check in the layer index suggests that you need oe-core Jun 25 06:51:21 LetoThe2nd: oe-core isn't in Poky ? Jun 25 06:53:58 Metanaou: i can properly explain it, just got that bit from http://layers.openembedded.org/layerindex/branch/master/layers Jun 25 06:54:06 s/can/can't/ Jun 25 06:54:30 Reading this : https://github.com/openembedded/openembedded-core Jun 25 06:54:45 It seems that oe-core is in Poky. Jun 25 06:55:04 So I could add it, but I'll have two times the same thing. Jun 25 06:56:36 Metanaou: yeah you're right syslinux is in recipes-devtools Jun 25 06:57:00 so next step: find out why it can't be built. are you targetting something other than x86, maybe? Jun 25 06:57:13 Yeah Jun 25 06:57:19 It's for ARM. Jun 25 06:57:20 no syslinux for you, then. Jun 25 06:57:26 so no iso Jun 25 06:57:28 Aww, shit. Jun 25 06:57:40 c'mon, how did you expect it to boot? Jun 25 06:57:50 So if I want to have a deployable image, how can I manage that ? Jun 25 06:58:08 Maybe syslinux was used only to built it. :D Jun 25 06:58:10 depends on your way of deployment. Jun 25 06:58:36 For Freescale, I use .sdcard image, it's pretty cool. Jun 25 06:58:47 Metanaou: you are aware that the boot process in ARM land is heavily machine specific, in contrast to x86's bioses and such? Jun 25 06:58:53 dd if=image.sdcard of=/dev/sdX, done. Jun 25 06:59:16 LetoThe2nd: no, I wasn't. Jun 25 06:59:29 well then you are now. Jun 25 06:59:30 I'm pretty new in embedded. Jun 25 07:00:24 plus, the sdcard thing isn't freescale specific, it's specific to a line of boards that share a common hardware setup in terms of sd card interfacing and boot settings. Jun 25 07:00:35 <_4urele_> Metanaou, what board do you have? (beaglebone black, raspberry pi, sabre sd...) Jun 25 07:00:43 i can easily crank up a fsl board where that "sdcard" thing won't do anything at all. Jun 25 07:01:12 Metanaou: so your first step is: specify the board in use, and determine the way of deployment. Jun 25 07:01:24 _4urele_: the fsl is an i.mx6q sabresd. Jun 25 07:01:32 I have also an Olimex lime2 Jun 25 07:02:59 LetoThe2nd: with the sabresd, I boot on u-boot and I download a minimal image on RAM (rootfs is .cpio.gz.u-boot). Then I download the real system and flash it on eMMC. Jun 25 07:03:30 And I separatly flash a specific u-boot in the boot partition, in order to enable the boot on eMMC. Jun 25 07:03:49 <_4urele_> Metanaou, then you are pretty aware of the boot process! Jun 25 07:04:00 Metanaou: to me, that does not specify how you deploy, but rather a manual reflashing Jun 25 07:04:13 With the Olimex, it's way harder, 'cause I don't have a depployable image. Jun 25 07:04:33 LetoThe2nd: so what do you mean by "how you deploy" ? Jun 25 07:05:38 what do you want to hand out? some blob that people have to reflash manually as you described? or some blob that automatically reflashes a board upon boot? if so, what medium are you using? Jun 25 07:06:57 the sdcard image that you named for fsl is an example of the latter, with sd card being the medium Jun 25 07:07:29 The first boot is done using the sd card, in order to have u-boot. Jun 25 07:07:39 But after, I'd like to do it all by network. Jun 25 07:08:32 So the simplest is to have a deployable image (like the .sdcard, that I use without sd card) that I can flash on the eMMC. Jun 25 07:09:09 and having a short look at the olimex board, i'm pretty convinced one could also create a sdcard image comparable to the fsl one for it. its just hte the sdcard magic you know is provided by fsl, and specific to them. you might be able to beat it into shape, of course. Jun 25 07:09:41 It's seems interesting. Jun 25 07:09:45 Metanaou: ok, so you're actually looking for a way to deploy a *new* blob to devices that are already initially flashed and functional, right? Jun 25 07:09:54 I'll have a look into it, thanks. Jun 25 07:10:03 LetoThe2nd: almost. Jun 25 07:10:25 Metanaou: in that case, have a look at meta-swupdate by sbabic. Jun 25 07:10:27 On the first boot, I have a functionnal system booted on RAM? Jun 25 07:10:31 s/?/. Jun 25 07:10:59 But then, yes, I want to deploy another blob on the device. Jun 25 07:11:09 so you need the deployment-receiver mechanism to be present in the blob AND then initial boot. Jun 25 07:11:27 which again is of course possible. Jun 25 07:13:12 swupdate could probably be used to archieve that, so my advice would be to have a proper look at it. Jun 25 07:15:24 LetoThe2nd: can't I use a simple bash script ? Jun 25 07:15:53 Which download the sdcard image and u-boot binary and flashs them in the eMMC ? Jun 25 07:16:13 of course you can, but then your and my defintion of "simple" won't share too much intersection Jun 25 07:16:55 of course if leave out any kind of fault-tolerance, power interruption, broken rom and all those fun parts, then it becomes almost "simple" Jun 25 07:17:48 been there, done that. now we're using swupdate, as its just more bang for the buck - we don't have to replicate functionality that somebody else already has properly implemented and tested Jun 25 07:18:53 for something in private tinkering, no problem with a small script. you can fix it whenever it fails. Jun 25 07:19:34 if you want to ship a real product, i seriously advice you to evaluate existing, proven and well-tested mechanisms first. Jun 25 07:19:54 (with swupdate just being one of them, i'm not trying to sell you anything.) Jun 25 07:21:11 <_4urele_> LetoThe2nd, I didn't knew swupdate, I'm very interested about it! are you using it? Jun 25 07:21:22 LetoThe2nd: the same, it seems really cool. Jun 25 07:21:31 Too bad I haven't found it before. :( Jun 25 07:21:39 _4urele_: we're using it, yes. Jun 25 07:22:16 <_4urele_> is it possible to update the current used rootfs with it? Jun 25 07:22:17 I can't share too many details as i'm really just a user, not the architect - but it works, and it certainly is vastly superior to our previous homebrew mechanism Jun 25 07:22:28 _4urele_: don't think so. Jun 25 07:22:50 you always need some kind of twin or intermediate system. Jun 25 07:23:36 <_4urele_> LetoThe2nd, that's make sense ;) just wanted to be sure, but it seems pretty awesome! Jun 25 07:24:52 i repeat, don't get me wrong: i do *not* advocate using that particular thing, i advocate properly evaluating exiting things before wasting time on a limited homebrew solution. Jun 25 07:33:36 LetoThe2nd: it doesn't seem to be supporting yocto 1.7+ Jun 25 07:38:13 Metanaou: ok, so? Jun 25 07:38:42 So there are some awful errors. :p Jun 25 07:39:14 how can i check if my bbapped is read or loaded? Jun 25 07:46:41 Metanaou: can always be fixed :-) Jun 25 08:37:15 miandonmenmian_, bitbake "myrecipe" Jun 25 08:40:54 from what I can tell, if it doesnt complain, then it is loaded. I suppose what I'm doing wrong has to be somewhere whiting the priority of my layer Jun 25 08:41:07 because in spite that my layer includes some extra packages, those packages are not loaded Jun 25 08:41:12 on the image Jun 25 08:41:43 does your image include the dependency on the recipe? Jun 25 08:42:32 miandonmenmian_, how are you testing for "loaded"? Jun 25 08:47:38 the recipe require dependencies, so i suppose it would complain is it isnt there. however one of the dep might not be included on the image conf Jun 25 08:47:55 im testing for loaded, because its not on the image when the image is run Jun 25 08:48:08 i'd rather find a faster approach to this kind of tests Jun 25 08:51:00 miandonmenmian_, have you done this before? Jun 25 08:51:17 was working before i updated my yocto environment Jun 25 08:51:48 but i did use the bbappend before for the included packages, instead i edited other layer directly Jun 25 08:51:50 did you check that your layer is on the new environment? Jun 25 08:52:00 yes, layer is there Jun 25 08:52:09 just realized that not all packages are built Jun 25 08:52:26 i dont understand why some are built, but no error output Jun 25 08:53:00 miandonmenmian_, is there a reason you customized another layer rather than creating your own? Jun 25 08:53:29 i created my own, but seems the append didnt work, so just added the steps of my layer to the other layer on top Jun 25 08:53:57 Now im trying to get my layer really working Jun 25 08:54:01 the bbappend will only get applied if the original recipe is used Jun 25 08:54:21 both my own and the original recipes are loaded Jun 25 08:54:56 bitbake has them when compiling image. bitbake-layers also includes them Jun 25 09:05:22 redengin: does the append have to name the main source on it? i see no reference to this, but i saw some appends have a similar line Jun 25 09:05:54 like this: require conf/distro/poky.conf Jun 25 09:13:49 confs are different than recipes Jun 25 09:15:23 miandonmenmian_, is there any way to see what you're working on? Jun 25 09:16:47 redengin: i'll try to clean it up and post it later on. but i was trying to make some sense of it before doing that, i'll try to put it somewhere Jun 25 09:24:36 I'd start by comparing your local.conf between environments Jun 25 09:25:20 did you create a custom image, or are you using local.conf to define your image dependencies? Jun 25 09:26:37 and yes, if you have just "x.bbappend" in your layer, you'll still need the layer that provided the original "x.bb" Jun 25 09:27:52 but you should see errors if there are unmet dependencies Jun 25 09:43:19 hello everyone, I'm trying to compile an arm distr on ubuntu 15.04, and the only thing i can't compile is mesa(warning treated as errors", someone has an idea at which point i need to fix that? Jun 25 09:45:17 wrock: helpful if you pastebin the actual errors Jun 25 09:50:28 sec Jun 25 09:58:18 here are the last couple of line rburton: http://pastebin.com/vJTMSp2P Jun 25 10:00:19 i tried to delete all -werror flags in the package Jun 25 10:00:23 but that didnt work Jun 25 10:00:37 somehow the changes didn't apply Jun 25 10:06:56 also, in run.do_compile i can't find an Werror flag Jun 25 10:07:14 hey guys Jun 25 10:07:41 is there a documentation about creating my own rootfs and building busybox? Jun 25 10:09:06 zloy`, wouldn't you just bitbake busybox? Jun 25 10:09:59 redengin: I have my own config file for busybox Jun 25 10:10:04 zloy`, what do you mean by rootfs? busybox won't boot Jun 25 10:10:21 redengin: furthmore, I'd like to create my own rootfs with init scripts Jun 25 10:11:11 redengin: by rootfs I mean the directories structures /bin, /sbin, /root, /dev, /sys, etc. Jun 25 10:11:50 redengin: and why busybox won't boot? Actually I built busybox and created rootfs manually and it boots perfectly Jun 25 10:12:22 zloy`, did you make the init busybox? Jun 25 10:12:43 redengin: yes, I did Jun 25 10:13:03 zloy`, ah, then I understand Jun 25 10:13:20 redengin: so, what about the documentation? :) Jun 25 10:13:41 I just want to create my own meta for building busybox and rootfs Jun 25 10:13:48 zloy`, I'm not sure I understand what documentation there would be Jun 25 10:14:28 zloy`, as in you just want an image that defines busybox as the init? Jun 25 10:16:02 redengin: and creates my own rootfs folders structure Jun 25 10:16:17 and scripts of course Jun 25 10:17:10 zloy`, then you'd make a (or many recipes) to add those folders/scripts Jun 25 10:18:11 redengin: as far as I understood this (https://www.yoctoproject.org/docs/1.8/dev-manual/dev-manual.html#understanding-and-creating-layers) document is that what I want, right? Jun 25 10:18:56 somewhat, but really you just need a recipe for your customization, that recipe needs to be in a layer Jun 25 10:19:44 and then that layer needs to be added to your build environment Jun 25 10:20:20 should I create my own layer? Jun 25 10:20:34 or I can customize an existing one? Jun 25 10:20:51 zloy`, layers just make it easier to reuse what you make Jun 25 10:21:21 zloy`, that way you won't have to push your changes onto the parent layer Jun 25 10:22:10 zloy`, inside your build environment it doesn't care about the git heritage Jun 25 10:23:59 zloy`, make sense? Jun 25 10:25:22 redengin: I think yes :) I am creating my own layer anyway :) Jun 25 10:26:27 layers are a best practice, but once you've added it to your build environment (bblayers.conf) its all one big pool Jun 25 10:27:17 understood, thank you Jun 25 10:28:26 also, per previous confusion, if you add a layer of bbappend, you still need the layer that includes the bb Jun 25 10:29:48 ok, I will create a layer :) Jun 25 10:29:52 thank you for the support Jun 25 10:51:02 redengin: btw, if I create by own layer, where I want to build busybox and rootfs, should I include meta, meta-yocto and meta-yocto-bsp to bblayers.conf file in the build/conf/ directory? Jun 25 11:11:10 you only need meta-yocto if you're using poky, and meta-yocto-bsp if you're using a bsp from there Jun 25 11:37:39 rburton, thanks for watching stackoverflow Jun 25 11:39:49 do postint run at first boot on oe-classic? Jun 25 11:51:01 If I remember correctly they all run both at install time and at first boot Jun 25 11:52:22 Crofton: what florian said, but there is a way to test if it currently runs at first boot. you want it? Jun 25 11:53:58 Crofton|work: got a rss feed for "yocto", should probably add more keywords ;) Jun 25 11:54:29 some scripts that contain bits which will fail not running at boot time test if they executed during install to an offline root Jun 25 11:54:38 yeah, I think that is correct Jun 25 11:54:39 dbus e.g. Jun 25 11:55:13 http://stackoverflow.com/questions/31001815/debugging-bitbake-pkg-postinst-pn-append-to-config-file-installed-by-other-r Jun 25 11:56:05 Crofton|work: he used ${D} in a postinst, which will get expanded during the build Jun 25 11:56:12 so have the wrong value Jun 25 11:56:47 pkg_postinst_stuff() { Jun 25 11:56:47 #!/bin/sh -e Jun 25 11:56:47 if [ x"$D" = "x" ]; then Jun 25 11:56:47 # do something at first boot here Jun 25 11:56:47 else Jun 25 11:56:47 exit 1 Jun 25 11:56:47 fi Jun 25 11:56:48 } Jun 25 11:57:12 someone reply on Stackoverflow :) Jun 25 11:57:20 Will do :) Jun 25 11:57:23 we need to build reputation there Jun 25 11:57:33 so we can vote down bad answers Jun 25 11:57:59 rburton, I marked openembedded yocto and poky Jun 25 12:01:08 oh, thanks to jku for monitoring stackoverflow Jun 25 12:01:36 I answered, UPVOTE PLS 😄 Jun 25 12:01:52 jk Jun 25 12:02:34 zwerch: runs at firstboot if rootfs fails Jun 25 12:02:46 if rootfs run exits 0, firstboot doesn't happen Jun 25 12:02:56 i'd put that in a comment but i don't have enough reputation Jun 25 12:03:09 (firstboot is basically fallback) Jun 25 12:03:18 earning reputation seems hard Jun 25 12:03:53 especially when you can't earn reputation with "well, actually" comments :) Jun 25 12:04:11 I edited it, better like this? Jun 25 12:05:52 would be good to say that the behaviour is run all postinsts at rootfs time, and any that fail run again on first boot. Jun 25 12:05:58 which leads to your example Jun 25 12:06:09 seems right, yes Jun 25 12:06:16 *but* the question was about doing something at rootfs time, so he just needs to use $D properly Jun 25 12:06:30 oh, unless there's an ordering problem Jun 25 12:08:43 does rpm ensure all depends are installed when it runs the postinst? Jun 25 12:16:36 what do people use to view rss feeds these days? Jun 25 12:17:31 i used feedkit on my macbook ;) Jun 25 12:18:37 looking at the feedly chrome thing Jun 25 12:30:45 Reeder on iOS / OS X Jun 25 12:46:48 I'm setting up fedely, see if this helps me Jun 25 13:14:44 hey, how can you modify the the bitbake core-minimal / core-sato? (I'd like to have some modules from sato in the minimal image) Jun 25 13:15:06 since the sato image is too load it on the zedboard Jun 25 13:19:31 wrock: create new image recipe, start with the core minimal one and add the stuff you want Jun 25 13:19:32 wrock: making your own image is not difficult. You can copy e.g. meta/recipes-sato/images/core-image-sato.bb and then remove/add image features, and install extra packages and packagegroups Jun 25 13:24:22 so minimal is everything in meta/ but without recipes-sato? jku Jun 25 13:25:21 what determines the order of layers from which bbappend files are parsed? I have some trouble overriding a variable in a bbappend file from an upstream layer. Jun 25 13:25:51 Oh and how can I best debug this without building the complete image and flashing it to see what changed? Jun 25 13:26:06 wrock: minimal is what meta/recipes-core/images/core-image-minimal.bb defines (definitely not everything in meta) Jun 25 13:30:21 hellerbarde: "bitbake -e " is very useful for debugging variables (lots of output though, pipe to a file and search there) Jun 25 13:31:04 jku: okay. thanks a lot. Jun 25 13:33:13 i have something like that there,were are they defined? IMAGE_INSTALL = "packagegroup-core-boot ${ROOTFS_PKGMANAGE_BOOTSTRAP} ${CORE_IMAGE_EXTRA_INSTALL}" Jun 25 13:33:24 i guess its the best i lloking for a manual.. Jun 25 13:47:56 hellerbarde: bbappend application order is done in layer priority order, so check your layer priority in your layer.conf. Jun 25 13:48:45 ohhhh, right. I thought i remembered that from somewhere. kergoth thanks so much. So much to remember when getting into this system. Jun 25 14:04:00 hi guys Jun 25 14:04:31 I'm trying to build an image for x86 but I can't seem to make it get rid of the initramfs Jun 25 14:04:46 I do not want a live image, I want a normal installation on an usb key Jun 25 14:05:00 so first step would be to get rid of the initramfs Jun 25 14:06:34 if I try to do INITRD_IMAGE = "", build is complaining that: ERROR: Nothing PROVIDES '' Jun 25 14:07:18 then it just says my target has no buildable providers Jun 25 14:07:23 and that's it Jun 25 14:07:41 somehow, somebody drags in the initramfs Jun 25 14:07:50 how can I debug that kind of dependency? Jun 25 14:18:46 manderke: define "normal install". the default image type for x86 is live, so change that instead of trying to hack it. ie make it build a ext3 file system that you can put in a partition you created, or instead use wic to create a disk image from the ext3 Jun 25 14:19:51 rburton, I've changed IMAGE_FSTYPES, it is not live anymore Jun 25 14:20:49 my problem is happening when IMAGE_FSTYPES is set to my own class Jun 25 14:20:57 but only on x86 device Jun 25 14:22:28 I've no idea why live is even considered Jun 25 14:23:08 folks - tricky one Jun 25 14:23:24 I need to pull multiple git repos inorder to build a project. Jun 25 14:23:44 I need to pull repo into A Jun 25 14:23:57 I then need to pull repo B in A/B Jun 25 14:24:10 I then do a make inside A/ Jun 25 14:24:16 make sense? Jun 25 14:24:35 These are two seperate git repos - not branches on the same repo. Jun 25 14:25:25 so far so usual Jun 25 14:25:27 rburton, just gave a try with IMAGE_FSTYPES set to ext3 and I get the same error Jun 25 14:25:52 manderke: don't force initrd_image to '' Jun 25 14:26:05 that's unrelated to "i dont' want a live image" Jun 25 14:26:11 raykinsella78: what's the problem? Jun 25 14:26:18 rburton, if I don't do that, the initramfs ends up embedded in the kernel Jun 25 14:26:55 rburton - I well I specify two git:// in my SRC_URI, it just ignores the second one. Jun 25 14:27:10 RC_URI = " \ Jun 25 14:27:10 git://github.com/tianocore/edk2.git;protocol=https;branch=svn/branches/UDK2014.SP1;name=UDK2014 \ Jun 25 14:27:10 git://github.com/tianocore/edk2-FatPkg.git;protocol=https;branch=master;name=FatPkg \ Jun 25 14:28:04 manderke: i thought you wanted a non-live image? why is using a initramfs so bad? Jun 25 14:29:13 raykinsella78: see recipes-multimedia/gstreamer/gstreamer1.0-omx_git.bb for an example of pulling multiple git repos Jun 25 14:29:31 legend thanks Jun 25 14:29:51 (basiscally set destsuffix so it doesn't want to unpack them both to git/ Jun 25 14:30:18 rburton, because the provided initramfs does some things I don't need; for starters I want the rootfs to be mounted from one of the usb's key partition while the init script from the initramfs tries to loop mount a rootfs image file Jun 25 14:44:34 bluelightning_: hey Jun 25 14:44:49 hello Jun 25 14:44:57 I'm not really here btw ;) Jun 25 14:45:33 ah, ok. I will not ask then. Jun 25 14:45:50 ask the channel, not the GUY WHO SHOULD NOT BE ON IRC Jun 25 14:46:00 bluelightning_: don't make me find an op to kick you Jun 25 14:46:07 DO IT!! Jun 25 14:46:08 rburton: I wasn't paying any attention until I was pinged Jun 25 14:46:10 but it is related to our last conversation Jun 25 14:46:17 btw, I asked the channel afterwards and no one knew :) Jun 25 14:46:23 it is about the sdk generation (stripping). Jun 25 14:46:29 I'm dropping now ;) Jun 25 14:46:38 for some reason I cannot get the files-in-sdk.txt generated Jun 25 14:46:52 even though the buildhistory variable looks correct to me. Jun 25 14:46:55 I am using 1.6.3. Jun 25 14:47:45 what is the name for 1.6.3? Jun 25 14:47:46 and I am not sure how to pin this issue down. Jun 25 14:47:49 daisy Jun 25 14:47:51 dizzy daisy etc? Jun 25 14:48:02 hmm, I do not have a build for that to check, sorry Jun 25 14:48:21 it is probably about my environment, yet I am not sure what to check. Jun 25 14:48:29 <_4urele_> rburton, what a leadership! Jun 25 14:48:32 based on the manual, it ought to be generated when the SDK is generated. Jun 25 14:48:53 given that the buildhistory variable is set up properly. Jun 25 14:49:54 I can be sure about that with bitbake -e -c populate_sdk myimage | grep ^BUILDHISTORY Jun 25 14:50:54 BUILDHISTORY_FEATURES="image package sdk" Jun 25 14:51:00 BUILDHISTORY_COMMIT="1" Jun 25 14:52:53 perhaps it is better to ask on the mailing list as it seems to be a bit specific knowledge? :) Jun 25 14:53:00 and also, it is summer time. Jun 25 14:55:12 yeah, you need to find someone with a daisy build and compare notes, also, tst on a later build to see if it works as you expect. Jun 25 14:55:18 I see the file in my master build Jun 25 14:57:00 so any idea on how I can get rid off of the initramfs? Jun 25 14:57:36 Crofton|work: hmm, I can try that with the latest poky, but I cannot test it with my layer :) Jun 25 14:57:45 but even that is a good start, indeed. Jun 25 15:19:45 how do I override the SRCREV of a named branch in local.conf Jun 25 15:19:46 ? Jun 25 15:21:13 SRCREV_pn-edk2[branch] isn't working Jun 25 15:28:37 Hi guys ! Jun 25 15:29:01 I have a question, regarding the use of an external toolchain in yocto Jun 25 15:29:05 raykinsella78: ? Jun 25 15:29:24 raykinsella78: SRCREV means some hash, which is available only in one branch Jun 25 15:29:32 you will not get the same hash for a commit in a different branch. Jun 25 15:29:38 so I am not getting your question. Jun 25 15:29:43 nicolas_: ok Jun 25 15:29:57 I'm currently trying to produce an armhf image for an iMX6 board Jun 25 15:30:43 Using an external toolchain from linaro (2013.10), and it seems to be alsmot working Jun 25 15:30:56 raykinsella78: it uses overrides not flags, try SRCREV_branch_pn-edk2 Jun 25 15:31:24 System boots, but I notice that some of my binaries got prefixed with "arm-linux-gnueabihf-" Jun 25 15:31:40 nicolas_: by "some" do you mean "the toolchain" Jun 25 15:32:02 Nope, I mean the package nano for example Jun 25 15:32:16 now that's special Jun 25 15:33:13 Yes ! I started digging, but I am now a little bit hopeless as I can't ssem to find where this would happend Jun 25 15:35:02 And I also can't figre out with some packages would be prefixed and others not Jun 25 15:35:30 I should probably mention that I am using deb packages format Jun 25 15:37:26 rburton: And this is probably all connected, but ldconfig is not found on the target Jun 25 15:40:31 Any ideas :) ? Jun 25 15:47:02 thanks .. rburtion Jun 25 15:48:28 is there a way to mark a bb as permanently dirty, to force a re-download and a rebuild on every run? Jun 25 15:48:48 use -c clean maybe? Jun 25 15:53:17 so does bitbake -c cleanall delete all the things only for the specified recipe or for _all_ dependencies as well? Jun 25 15:55:40 cleanall will remove work tree, sstate, and downloads for the recipe Jun 25 16:06:08 heh, that is confusing, since fetchall runs fetch against the recipe and its deps :) Jun 25 16:06:30 yeah, i guess Jun 25 16:15:15 inconsistent, i should say Jun 25 16:23:18 thanks guys. Jun 25 16:23:54 RP: https://gist.github.com/d6736d5c777d007ab4fc Jun 25 16:24:41 kergoth: i had a custom class that caused that too Jun 25 16:24:42 hmm Jun 25 16:24:49 kergoth: you've event handlers that cause events to be fired Jun 25 16:25:24 i added an event mask to make my handler work, not sure if that's a good enough fix for you Jun 25 16:26:04 rburton: sanity.bbclass fires events from its handler :P Jun 25 16:26:04 alternatively, we basically reference count the assignment of d in execute_handler Jun 25 16:26:27 i added a mask and it stopped happening, which is when i stopped caring about the problem... Jun 25 16:26:53 sanity has eventmask already :) Jun 25 16:27:21 kergoth: throw a debug statement before it executes the handler so you know what is causing the problem Jun 25 16:28:20 kergoth: I think its events which trigger other events that do that Jun 25 16:28:29 yep, see above Jun 25 16:28:38 kergoth: I'm just a bit puzzled why some cases hit it and others generally dont Jun 25 16:29:35 none of my sanity tests failed, so i'm assuming its the success event in this case Jun 25 16:30:08 kergoth: well my timer class caused that exact error, until i gave it event masks Jun 25 16:30:46 (https://github.com/rossburton/meta-ross/commit/22a0a4058a9359bb223f60da2e16f5c13eee2dde#diff-c491088205b10cd3b0a70eb575192607) Jun 25 16:30:48 i don't fire events anywhere in any of my local layers Jun 25 16:31:02 kergoth: log messages are events (e.g. bb.note) Jun 25 16:31:06 i currently have poky, meta-oe, meta-virtualization, meta-oe, and meta-mentor & meta-sourcery. Jun 25 16:31:48 i expect we should revert this until the kinks are worked out, if we can't print messages from event handlers, that's a serious problem :) Jun 25 16:32:34 Hmm, it's kind of too bad that event handlers are run immediately in caller context, rather than queued up for execution in an event loop or something Jun 25 16:32:39 oh well Jun 25 16:32:51 kergoth: well, what puzzles me is the autobuilder works, all my builds work etc Jun 25 16:33:01 kergoth: I'd like to understand how to trigger this Jun 25 16:33:35 currently all i have configured is the poky layers, meta-oe, meta-sourcery, and the meta-mentor layers. Jun 25 16:33:39 i'll try to pare down to a test case Jun 25 16:33:56 it's happening on every bitbake execution in this configuration Jun 25 16:35:23 kergoth: I did comment on https://bugzilla.yoctoproject.org/show_bug.cgi?id=7668 earlier btw. If we remove the predefined function capability of register we could implement this like you suggest instead Jun 25 16:35:24 Bug 7668: enhancement, Medium, 1.9 M1, richard.purdie, IN PROGRESS REVIEW , Feature request: inject 'd' as well as 'e' into event handlers Jun 25 16:35:40 saw that, i'd forgotten about the direct function handling.. Jun 25 16:35:48 afaik no one is using that, but i guess some external layer could be Jun 25 16:35:51 hmm Jun 25 16:36:08 kergoth: I'm not aware of anything either fwiw Jun 25 16:36:31 kergoth: the other way is to only remove d from builtins if we added it Jun 25 16:37:13 * kergoth nods Jun 25 16:37:33 i wish there was a cleaner way to do this.. can't just inject something into the context of a function object other than through builtins :\ Jun 25 16:40:03 huh, https://stackoverflow.com/questions/2749655/why-are-closures-broken-within-exec is rather interesting Jun 25 16:40:07 always wondered about that Jun 25 16:41:37 I always rather liked being able to pass a function into bb.event, as it made it more general purpose, but our event handling is pretty specific to use with our metadata anyway in most cases, so.. Jun 25 16:48:23 hmm... python3 in poky seems... incomplete Jun 25 16:49:15 RP: i think rburton is right, it seems related to event handlers with no eventmask (i didn't think i had any left, but i was wrong, should have paid closer attention), so the same event is fired from both the outer and inner events when events are fired from other events.. but i'm not sure why that would cause this failure offhand Jun 25 16:49:28 haven't verified that adding eventmasks fixes it yet though, testing further Jun 25 16:49:47 tripzero: did you just install python3-core and none of the modules (python3-modules will pull them all in)? Jun 25 16:50:21 specifically, in this case, if i print messages before and after the event function call, i'm seeing multiple events fired after check_sanity_eventhandler starts but before check_sanity_eventhandler exits Jun 25 16:50:36 rburton: probably! Jun 25 16:50:59 rburton: is there a way to get bitbake to spit out a list of packages for a recipe? Jun 25 16:51:25 yep, rburton called it, fixing those resolved the problem Jun 25 16:51:42 kergoth: now you can figure out *why* Jun 25 16:52:06 tripzero: oe-pkgdata-util can do that i think Jun 25 16:52:09 its possible i had no event handlers that catch the sanity success or failure messages Jun 25 16:52:19 so nothing was run in the nested event handling Jun 25 16:52:28 but ahving a non-eventmasked event handler changed that Jun 25 16:52:49 i expect adding a handler for the sanity success or failure message would exhibit the same behavior, but that's just a theory Jun 25 17:05:23 so did anybody ever ran a yocto kernel on x86 without the initramfs? Jun 25 17:13:19 or to put it differently, did anybody try running a regular (installed) usb yocto os? (as opposed to a live usb yocto image?) Jun 25 17:18:54 zeddii, awake? Jun 25 17:18:55 | /home/balister/src/oe-core/build/tmp-glibc/work-shared/ettus-e1xx/kernel-source/arch/arm/include/asm/irqflags.h:11:5: warning: "__LINUX_ARM_ARCH__" is not defined [-Wundef] Jun 25 17:18:55 | #if __LINUX_ARM_ARCH__ >= 6 Jun 25 17:19:13 any tips on that, with linux-yocto-custom Jun 25 17:20:50 Crofton|work, hmm. was just building lots of random kernels yesterday. that's with an older kernel or 4.1 ? Jun 25 17:21:01 older 3.0 era Jun 25 17:21:36 3.5 era builds, just doesn't work :( Jun 25 17:36:18 zeddii, might be self inflicted .... Jun 25 17:36:35 no .... Jun 25 17:38:29 hmm, wonder how difficult it'd be to make recipetool more extensible by pulling plugins from lib/recipetool/* from all entries in BBPATH Jun 25 18:34:35 grr. even with python3-modules I still don't have "imp": ImportError: No module named 'imp' Jun 25 18:34:47 which means I can't run "get-pip.py" Jun 25 18:35:05 rburton: are there any other essential python3 packages besides python3-modules? Jun 25 18:44:23 sounds like somebody forgot to include it in the python-3.4-manifest.inc for somer eason Jun 25 18:46:59 http://lists.busybox.net/pipermail/buildroot/2015-June/130833.html Jun 25 19:03:37 tripzero: what kergoth said - please fix the manifest Jun 25 19:31:55 rburton: I missed what kergoth said. which manifest? Jun 25 19:32:05 kergoth: can repeat directive? Jun 25 19:41:23 see scripts/contrib/python/generate-manifest-3.4.py and meta/recipes-devtools/python/python-3.4-manifest.inc which is generated by it Jun 25 19:41:28 imp must be missing from the former, add it Jun 25 19:54:33 y Jun 25 20:05:41 kergoth: thx Jun 25 20:16:04 np Jun 25 21:15:41 Crofton|work: someone should see if we could share a test framework ;-) Jun 25 21:38:13 RP, everyone should write their own Jun 25 21:54:25 kergoth: I quickly wrote up a patch to fix that event handler issue, at least in theory Jun 25 21:54:33 kergoth: whether we should do something different remains to be seen Jun 25 22:03:16 * kergoth nods, he doesn't have a very strong opinion on it either way, though modifying __builtins__ feels slightly uglier **** ENDING LOGGING AT Fri Jun 26 02:59:58 2015