**** BEGIN LOGGING AT Tue Mar 22 02:59:58 2016 Mar 22 04:30:41 Hi. What is the correct way to add a line in the target image to /etc/rc.local? Do I need to patch the recipe that creates it or can I do it from a different recipe? Mar 22 04:45:25 otay: typically you'd patch the recipe that provides the thing, using a postinstall script (pkg_postinst_${PN} () { ... }) Mar 22 04:45:31 otay: but rather than using rc.local wouldn't you be better off using an actual initscript? Mar 22 04:46:02 if you can provide the initscript we have pretty robust support for ensuring it gets enabled Mar 22 04:52:55 okay, good info on the patch. Was wondering about that. Yes, initscript would be better. But even better would be to modify the actual software itself. However, I'm not totally sure of the future of it so don't want to spend too much effort. Mar 22 04:53:36 To clarify, the build I'm doing comes with smartpm. I just need to set a default channel. Mar 22 04:55:00 So, yeah, adding the channel the proper way would be better but figured if I can add the channel (repo) easily it will buy me some time. Sorry to sound like a hack. Mar 22 04:55:54 There are a lot of layers in this build. Feels like I'm jumping in to the deep end. Mar 22 04:56:39 But you just gave me another idea Mar 22 04:57:39 Correction: you gave me the answer. Mar 22 04:57:44 post install script Mar 22 04:59:13 So I can patch the thing that makes rc.local. 2) do a post install script in a different package or same package or 3) cry a lot.. jk;3) set the channel in the code somewhere. Mar 22 05:01:08 Is it uncommon to modify the destination image from a recipe? I mean, couldn't I do echo foo >> ../../dest/image/etc/rc.local from a totally different recipe? Mar 22 05:01:25 Or is that weird? Mar 22 05:02:45 Either way, I like the post install script idea. At least I can handle that. Thank you Mar 22 05:03:12 * otay :) Mar 22 05:03:40 btw; we have a fork of yocto. Mar 22 05:04:09 (I mean the guy who hired me does ;) ) Mar 22 05:14:37 otay: you can't poke files into the image like that from another recipe Mar 22 05:15:06 otay: in an image recipe, you can add a call to a function in ROOTFS_POSTPROCESS_COMMAND and then have that function do whatever you want to the image though Mar 22 05:15:20 having said that there ought to be a means to set up feeds without needing to do this Mar 22 05:15:41 which version of the build system are you using (or rather, which is your fork based on) ? Mar 22 05:15:47 ahh, interesting. Mar 22 05:15:53 It's the Wind River one Mar 22 05:16:06 Well, one of them I should say Mar 22 05:16:55 The one for IoT. Mar 22 05:17:01 Hi, I am facing an issue while writing a recipe. This recipe depends on another package not just for its libraries and headers Mar 22 05:17:10 but also for its makefiles Mar 22 05:17:30 so I can't simply import its headers and libraries to std path Mar 22 05:17:41 any suggestions? Mar 22 05:18:14 bluelightning but I agree, that doing something like ROOTFS_POSTPROCESS_COMMAND is a sign I'm doing something wrong. I like your idea of post-install. And actually, just want to make an rpm that does it. Sound clean Mar 22 05:20:23 (sounds clean: An rpm that just runs a command in a post install script. I wish I would have thought of that. :) ) Mar 22 05:21:14 * otay goes to it Mar 22 05:25:59 (smart channel --add in pkg_postinst_${PN} in OurRepo.rpm) +1 to bluelightning. Mar 22 05:45:54 niteshnarayanlal I had that problem before also but ended up doing something different so I'm no help Mar 22 06:07:07 i get some annoying "WARNING: QA Issue: foo has an incompatible license. Excluding from packaging. [incompatible-license]" warnings where there doesn't seem to be any real problem since the packages in question are never included in any images. it should be safe to silence them with INSANE_SKIP, right? Mar 22 06:14:47 otavio, yeah the best possible solution I think is to build that package from the same recipe first Mar 22 06:14:49 and then use it Mar 22 06:15:00 although its not a very refined way :/ Mar 22 06:19:00 Ulfalizer I was getting a similar QA warning and did: do_qa_configure () {} Mar 22 06:19:50 override. Not sure if it will help in your case. Mar 22 06:24:57 seems a bit hackish Mar 22 06:27:07 looks like the INSANE_SKIP_* logic is coded specifically for each case in package.bbclass Mar 22 06:28:13 but it hasn't been coded for incompatible-license :/ Mar 22 06:29:39 maybe the license type given doesn't match the license file? Mar 22 06:31:18 Does anybody know if it's possible (from a layer) to make an archive (rpm) that only has a post install script in it? Mar 22 06:32:12 otay: nah, the problem is just that one of the packages generated from the recipe has a blacklisted license Mar 22 06:32:24 however, that package isn't used, so it's okay Mar 22 06:33:26 ahh Mar 22 08:13:45 good morning Mar 22 08:14:21 morning Mar 22 08:22:02 howdy! :-) Mar 22 08:23:06 goood morning Mar 22 08:32:01 Hi all. I'm having an issue with my own distro. I've copied poky.conf to my own layer, renamed and modified it: https://gist.github.com/Anticom/1bd29c45f002a0b96a84 Mar 22 08:32:30 The issue i'm faced with now is that i can't see, where default-distrovars.inc is included now Mar 22 08:32:55 Any recommendations how to track down the issue? Mar 22 08:51:32 Okay i've tracked it down to bitbake.conf Mar 22 08:51:50 How would i change, what that includes? custom bb.conf in my layer aswell? :O Mar 22 08:52:17 Or would you just not use IMAGE_FEATURES_DEFAULT at all? Mar 22 08:55:55 Anticom: what is it that you're aiming to do? Mar 22 08:58:35 mborzecki: Well 1st currently i'm having a lot of IMAGE_FEATURES_remove in my image target to get rid of various features comming from IMAGE_FEATURES_DEFAULT and 2nd i wanted my own distro anyway so i can set stuff like update channels there. I think also the choice of init manager would be appropriate there, wouldn't it? Mar 22 08:59:49 Anticom: yes, this is the right place to setup systemd as init Mar 22 09:01:25 mborzecki: However i don't want to reinvent the wheel (which would end up being square with my skills anyway) but only tune certain aspects of poky.conf (e.g. the IMAGE_FEATURES) Mar 22 09:02:45 But i'm still having a hard time figuring out what is included or configured where and why Mar 22 09:02:58 And (probably most important) some bb internals Mar 22 09:03:09 Anticom: keep in mind that distro should only setup a policy, image decides what gets installed (and that may partially depend on distro features) Mar 22 09:03:53 mborzecki: But poky.conf is setting some IMAGE_FEATURES i don't need. Why would i have x11 as distro feature if my target device doesn't have a screen? :> Mar 22 09:04:04 or opengl or alsa if i don't need audio Mar 22 09:05:06 Anticom: you would never provide your own bitbake.conf, unless you were doing something with bitbake completely other than OpenEmbedded that is Mar 22 09:05:13 Anticom: i don't recall poky setting image features directly, iirc it's more like poky would enable x11 in distro features Mar 22 09:05:24 mborzecki: it is Mar 22 09:05:26 Anticom: just override the values you want to in your custom distro config, or even local.conf temporarily Mar 22 09:05:50 remember poky is a sample, if it doesn't fit what you want (which is expected) then create your own distro with your own DISTRO_FEATURES Mar 22 09:06:00 poky explicitly enables lots of features Mar 22 09:06:04 Anticom: you're familiar with bitbake -e which shows a trace of where variables get set? Mar 22 09:06:28 bluelightning, rburton and how are the DEFAULT_ _DEFAULT meant to be handled? Just throw them away in a custom distro? Mar 22 09:06:47 Because their values are burried deep in some include hierarchy which i don't get how to override Mar 22 09:07:08 bluelightning: yup i am :) Mar 22 09:07:16 Anticom: the intention is that you provide your own value for DISTRO_FEATURES, or add/remove items from it - DISTRO_FEATURES_DEFAULT is meant to remain at the default in case you need that Mar 22 09:07:47 for most cases the hierarchy is irrelevant - just set the value you want to set in your distro config Mar 22 09:08:10 bluelightning: So how wDISTRO_FEATURES_DEFAULT living in meta/.../default-distrovars.conf which along the way is included by bitbake.conf? Mar 22 09:08:23 because its the default? Mar 22 09:08:35 how would* Mar 22 09:08:35 just assign your own Mar 22 09:08:37 Anticom: you're not meant to change DISTRO_FEATURES_DEFAULT Mar 22 09:08:48 change DISTRO_FEATURES, that's what actually has an effect... Mar 22 09:08:53 bluelightning: well that was my question :) Mar 22 09:09:50 so just set DISTRO_FEATURES ?= "a b c ${DISTRO_FEATURES_LIBC}" discarding any default stuff is the way to go? Mar 22 09:10:34 a distro should use = Mar 22 09:10:34 i'd say DISTRO_FEATURES = ".." rather than ?= Mar 22 09:10:35 Anticom: pretty much except you might as well use = not ?= Mar 22 09:10:49 unanimous ;) Mar 22 09:11:02 it's great when we all agree :) Mar 22 09:11:03 but why is it done that way in poky.conf? Mar 22 09:11:22 like what would set DISTRO_FEATURES prior to poky.conf? Mar 22 09:11:22 Anticom: it's meant to make it slightly easier to override if you want to keep using poky.conf itself Mar 22 09:11:29 local.conf Mar 22 09:11:47 so local.conf is evaluated first and then poky.conf? Mar 22 09:11:51 it actually just makes it more fragile, we should change that to = imho Mar 22 09:11:52 yes Mar 22 09:12:02 well wouldn't it be better to keep ?= in that case so i can still override DISTRO_FEATURES in local.conf? Mar 22 09:12:05 people who do DISTRO_FEATURES += in local.conf suddenly find glibc wont build Mar 22 09:12:11 Anticom: if you're curious about the order, look at the bunch of require... lines in conf/bitbake.conf Mar 22 09:12:27 Anticom: you'll have to overide the entire declaration, if you want to alter then use _remove and _append - then it doesn't matter Mar 22 09:12:46 bluelightning: That's the only thing im missing in the docs: More pretty drawings like in android docs for example Mar 22 09:12:52 Anticom: FWIW, local.conf *must* be parsed before the distro config, since before local.conf is parsed we don't actually know what the value of DISTRO is ;) Mar 22 09:12:54 Anticom: this is the order in which *.conf are included https://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/meta/conf/bitbake.conf#n696 Mar 22 09:13:09 bluelightning: good point :) Mar 22 09:13:53 Thank's for all the information yet again (: Mar 22 09:14:07 with great comunity comes great projects ;) Mar 22 09:14:16 I agree we should document the order, the question is where would be the best place in the docs to do it... Mar 22 09:15:03 bluelightning: either a seperate "internals" section or isn't there something explaining how bitbake works in a high-level view already? Mar 22 09:15:11 Like where there are some charts explaining it already Mar 22 09:22:01 Anticom: there's a bitbake manual, but that mostly covers the fine details of parsing and effects of +=, _append, ?=, = and so on Mar 22 09:23:11 mborzecki: I know but thanks :) Mar 22 09:23:18 Have to reboot, brb Mar 22 09:35:01 bluelightning, rburton: One more question: Why is graphical and audio stuff set in the defaults? I mean talking embedded devices (thinking towards IoT devices like "simple" connected temperature sensor etc.) they offten don't have a display and/or speakers. Is there no more "elegant" way to split those concerns into seperate configs / units? Mar 22 09:35:39 I'm imagining something like IMAGE_FEATURES packagegroups where you could say "i want audio" or "i have a display" Mar 22 09:35:52 Anticom: it's entirely expected that you change your DISTRO_FEATURES for what you're working on Mar 22 09:35:54 Or is poky meant to be more feature packed? Mar 22 09:36:06 *poky* enables almost everything, for testing Mar 22 09:36:20 rburton: Of course it is but from my personal experience it's tedious to get that sorted as a newbie :) Mar 22 09:36:28 rburton: okay get it Mar 22 09:36:46 if poky disabled audio and video, we'd have nothing testing those recipes Mar 22 09:37:10 Anticom: clearly the bits of the docs where it says "and now you create your own distro" needs to be clearer Mar 22 09:37:13 rburton: so the idea is if you're new just build poky; it has everything and can do anything Mar 22 09:37:41 rburton: well there's a long weekend ahead. I might fiddle with a draft for you guys if that might help Mar 22 09:37:59 That would also help me prove that i got everything right so far Mar 22 09:38:46 another big thing is when to define your own machine Mar 22 09:39:34 mborzecki: i.MX6 ftw :p Mar 22 09:40:07 i'm not responsible for the BSP part but afaik keith-koep ships some bsp stuff so it's rather straight forward Mar 22 09:42:33 rburton: Btw is there some way to find out more about what a certain distro feature does? For example can i search for distro features on layers.openembedded.org ? Mar 22 09:42:50 Anticom: they're often deliberately vague Mar 22 09:42:59 (opengl for example) Mar 22 09:43:12 rburton: is there some bb flag to find out more about it? like -e ? Mar 22 09:43:16 no Mar 22 09:43:19 grep? Mar 22 09:43:21 bummer Mar 22 09:43:26 what flag in particular? Mar 22 09:43:51 fair enough but grep can take forever if you don't know based in which directory you have to search Mar 22 09:44:15 rburton: Well i was asking for a "debugging /tracing" thing where you could provide an image feature instead of a target much like bb -e target Mar 22 09:44:38 or recipe to get my terms correct Mar 22 09:44:52 Anticom: search meta/ Mar 22 09:45:27 rburton: probably the distro features often refer to some linux googable stuff like "argp" for example i suppose Mar 22 09:46:20 pretty sure that one isn't used anymore Mar 22 09:46:29 (grep agrees) Mar 22 09:46:56 we have a standalone argp recipe for glibcs without it these days Mar 22 09:47:20 oh uclibc respects it still Mar 22 09:54:38 Oh one last thing i've asked myself a couple of times now. I sometimes see @bb.utils.foobar helpers here and there. Are they documented in the bb manual? I actually never really coded python so i don't even know, whether that's some inlined python or some custom bb DSL and where to find the API Mar 22 09:56:07 ${@ is inline python Mar 22 09:56:28 bb.utils are cunningly in bitbake/lib/bb/utils.py :) Mar 22 09:56:40 Is there a list of commonly used helpers? Mar 22 09:56:43 there's also meta/lib/oe/ with some stuff in Mar 22 09:57:16 bb/utils.py is a convenient list of utility functions :) Mar 22 09:57:18 I'd like to write more sophisticated recipes but i'm having a hard time to understand what the params are and what some of them are good for Mar 22 09:57:24 Okay, i'll check that out Mar 22 09:57:26 cheers Mar 22 10:04:16 yeah, the python API would need some love on the documentation side... Mar 22 10:17:31 along with defined variables, like `d` Mar 22 10:26:34 there's a little bit of documentation in the bitbake manual at least Mar 22 10:33:01 call me crazy - but does file-rdep do an ldd or readelf -d to figure out the dependencies of an executable, and then check that something provides them. Mar 22 10:43:17 I don't get where pulseaudio is added to my DISTRO_FEATURES. bb -e recipe yields: # append utils.py:128 [features_backfill] \\\ # " pulseaudio bluez5" Mar 22 10:43:41 But what has utils.py got to do with my DISTRO_FEATURES and what is features_backfill? Mar 22 10:46:51 nvm, think i've figured it out myself Mar 22 10:53:58 Can someone tell me where do I use a custom defconfig for uboot in my uboot.bbappend? I used UBOOT_MACHINE="defconfig", but it gives me no errors and it seems that the defconfig is not really used. Any help very appreciated. thanks Mar 22 11:58:12 hi i am trying to modify this recipe Mar 22 11:58:13 https://github.com/flihp/meta-measured/tree/0c8b4e24335a2fbb24472d27093d823f613f7375/recipes-tpm/trousers Mar 22 11:58:44 I have few patches which I am including and then trying add CFLAGS_append and LDFLAGS_append to Mar 22 11:58:57 add the path of required library and headers Mar 22 11:59:10 but it is still not able to find those Mar 22 11:59:31 I am wondering if I am doing that at wrong location. Mar 22 11:59:44 as it is using something init.sh Mar 22 11:59:53 not sure if those flags will impact its compilation Mar 22 13:06:37 When I use "bitbake core-image-minimal -c populate_sdk" is there a way of changing the directory it builds into? It asks me the first time, but no other time after that. Mar 22 13:08:20 it asks you where to *unpack* but after that you don't need to run the installer Mar 22 13:09:41 It isnt asking me anymore Mar 22 13:09:44 Just the first time Mar 22 13:10:24 I updated my poky build, so I wanted to update the SDK, but in a different directory Mar 22 13:14:02 maybe its because you need to run bitbake -c cleanall core-image-minimal before populate_sdk again Mar 22 13:14:47 riz_: do you mean the directory the installer unpacks to? Mar 22 13:14:57 Yes Mar 22 13:15:46 try a fresh terminal, maybe its finding an env var and updating the one you're currently using? Mar 22 13:19:36 hello everyone Mar 22 13:19:56 New terminal did not work Mar 22 13:20:44 i've a strange behavior with eGalax touch screen, recognized by my system as HID, so i'm trying to bind/unbind the USB from HID to touchscreen driver but getting No such file or directory error whenever i try to echo -n "1-2.4:1.0" > /sys/bus/usb/drivers/usbtouchscreen/bind Mar 22 13:21:19 anyone know in wihc situation this kin of error happends (u checked that my device is unbound before) Mar 22 13:22:09 * CTtpollard is also having problem with an eGalax touchscreen Mar 22 13:23:17 riz_: try to run the sdk with the option -d Mar 22 13:23:22 is that what you want;/ Mar 22 13:23:23 ? Mar 22 13:23:33 install the sdk in a different path? Mar 22 13:24:16 ohmy: udev is currently picking my screen as a mouse and as a tablet, not as touch Mar 22 13:27:43 ohmy: depends on the hardware but have you verified that its mean to appear as a touchscreen? Mar 22 13:34:35 rburton: well actually it's a touch screen, and the touch works when i connect it to my ubuntu desktop, but on my embeded device its jut recognized as hid :( Mar 22 13:34:55 ohmy: sounds like you need to check that you've actually build and loaded the touchscreen driver Mar 22 13:35:13 or that ubuntu doesn't have patches to make it work in their kernels Mar 22 13:35:23 CTtpollard: well for me (udevadm monitor) outputs KERNEL[66625.251575] add /devices/pci0000:00/0000:00:15.0/usb1/1-2/1-2.4/1-2.4:1.0/0003:0EEF:7200.0014 (hid) Mar 22 13:35:38 rburton: i load the touchscreen driver manually Mar 22 13:35:48 so I type bitbake core-image-minimal -d " populate_sdk? Mar 22 13:35:55 rburton: modprobe usbtouchscreen Mar 22 13:36:22 riz_: -d when running the installer. bitbake -c populate_sdk just builds the installer. Mar 22 13:37:12 riz_: do you want to build a new SDK or want to install a new SDK? Mar 22 13:37:15 this is my log http://pastebin.com/mY3Yw6d8 Mar 22 13:37:30 I want to build my sdk again in a different directory Mar 22 13:37:39 when you run bitbake -c populate_sdk it will build a installer on the deply dir Mar 22 13:37:44 that makes no sense Mar 22 13:37:51 you pick the directory on sdk install Mar 22 13:38:23 the installer is a script that will install the SDK on your machine Mar 22 13:38:52 I am getting confused. I have only built the sdk using bitbake. Then I setup the environment and point to the SDK using QtCreator Mar 22 13:39:16 you might want to install the sdk then Mar 22 13:39:18 rally stuck if anyone have a begining of idea Mar 22 13:39:30 as populate_sdk just builds a tarball in your deploy directory Mar 22 13:39:33 tried echo -n "1-2.4:1.0" > /sys/bus/usb/drivers/usbhid/unbind Mar 22 13:39:38 than modprobe usbtouchscreen Mar 22 13:39:44 and after echo -n "1-2.4:1.0" > /sys/bus/usb/drivers/usbtouchscreen/bind Mar 22 13:40:01 but the last commands says write error: No such device Mar 22 13:41:43 riz_: first go to the build/tmp/deploy/sdk Mar 22 13:42:15 there is two files, .manifest and .sh Mar 22 13:42:21 so run the .sh file Mar 22 13:42:34 it will prompt you where you want to install the SDK Mar 22 13:43:14 then it will install the SDK on your machine Mar 22 13:43:19 Yes. OK Mar 22 13:43:20 Got it Mar 22 13:43:33 Thanks Mar 22 13:43:41 youre welcome Mar 22 13:44:33 Hi there ! Mar 22 13:44:43 I have to add "INHERIT += "populate_sdk_qt5" in my local.conf, correct? Mar 22 13:44:53 For qt5 that is Mar 22 13:45:18 I have ntp in my image, and I only have ntpd on my target. How can I get ntpq and ntpdate ? Mar 22 13:48:33 :( Mar 22 13:52:19 I can see a ntp-utils rpm file, but how should it be installed in the image ? Mar 22 14:01:58 jmleo: depends on what runtime package manager you've got Mar 22 14:02:23 oh, it's an rpm Mar 22 14:02:49 well just copy the rpm somewhere on your targets filesystem and run rpm -i /path/to/ntp-utils Mar 22 14:04:26 ok, but I would like to have it when I compile the ntp recipe... I have "ntp" in my image but ntp-utils does not exist Mar 22 14:05:15 is because ntpdate is a different package Mar 22 14:05:34 jmleo: you need to install on your image the package ntpdate and ntp-utils.. Mar 22 14:05:46 you should always use rpm -U --not-- -i unless you specifically know what you are doing.. Mar 22 14:05:46 if you look at ntp recipe Mar 22 14:05:57 you will see the packages that it builds Mar 22 14:05:59 PACKAGES += "ntpdate sntp ${PN}-tickadj ${PN}-utils" Mar 22 14:05:59 -i will allow the system to install multiple copies of a package, which is almost certainly what you never want Mar 22 14:07:52 fray, igor3 thx :) Mar 22 14:20:24 jmleo: add it to IMAGE_INSTALL for your image Mar 22 14:20:36 boucman_work, yes, thx, I just did it :p Mar 22 14:20:41 IMAGE_INSTALL is the list of packages you want on your image (not recipes) Mar 22 14:21:02 that was my confusion (package / recipe) Mar 22 14:21:10 ok, you're good, then :) Mar 22 14:23:45 people, I made two images on a distro that one image has the tar.bz2 of the other image inside. Mar 22 14:25:08 to do it, I put the tar.bz2 of the image on the src_uri of a package that is installed on the other image Mar 22 14:26:02 and I made a task dependecy of the do_fetch of this package depending on the do_build of my image Mar 22 14:26:49 The question is: The IMAGE_INSTALL become a dependency of my other image? Mar 22 14:27:00 instead, make the container image depend on the first image's do_deploy Mar 22 14:27:59 sorry, but how can I do that? I dont know what is a container Mar 22 14:28:39 one image contains another image Mar 22 14:28:48 so the outer image needs to depend on the inner image's do_deploy task Mar 22 14:29:24 it is exactaly what I want Mar 22 14:30:26 But how I will put the image inside my other image? via SRC_URI? Mar 22 14:30:40 no Mar 22 14:30:45 you'd write some code to do that Mar 22 14:31:09 have a dig around meta/classes, some of the image types there embed other images inside them Mar 22 14:31:10 Now when I am installing my SDK I get "Setting it up...ls: cannot access /opt/poky/2.0.1/environment-setup-*: No such file or directory". Never happened before. Any ideas? Mar 22 14:32:10 riz_: it seems like you do not have permission to write in this path Mar 22 14:32:36 I recommend you to install the SDK on you home directory. Mar 22 14:32:47 OK Mar 22 14:32:53 Will try now Mar 22 14:33:10 thanks rburton Mar 22 14:33:29 I will take a look on it Mar 22 14:33:31 igor3: (initramfs, for example) Mar 22 14:34:07 nice, thank you sir Mar 22 14:38:52 Same thing when installing in home directory Mar 22 14:39:01 This is odd. Mar 22 14:39:09 Could it be that the SDKPATH is broken? Mar 22 14:40:00 its weird, I have never seen it before Mar 22 14:40:47 the SDKPATH and other enviroments will be set only after you source the SDK/environment-setup-*... is how the SDK works Mar 22 14:40:49 Doesnt it create the environment setup script? Mar 22 14:41:05 That is what I am not understanding Mar 22 14:41:20 Why does it want to see it there when it is supposed to create it Mar 22 14:41:34 the SDK installer will create a path with a toolchain and a sysroot path Mar 22 14:41:43 there will be a script that you will source Mar 22 14:42:15 then when you source this script, you will have your environment to build anything Mar 22 14:42:32 but your machine will be the same with the SDK installed Mar 22 14:42:34 Yes, but installingthe SDK provides the script Mar 22 14:42:40 yer Mar 22 14:42:41 yes Mar 22 14:42:53 So why is it saying it cant find it if it needs to create it? Mar 22 14:43:00 this script that you are having problem, is exactly the script you have to source Mar 22 14:43:11 Exactly Mar 22 14:43:17 its the weird stuff Mar 22 14:43:24 I dont know why Mar 22 14:43:32 maybe you can run it with sh -x Mar 22 14:43:39 and see what is going on Mar 22 14:43:49 try to install the SDK with sh -x Mar 22 14:44:17 rburton: When adding PACKAGE_FEED vars to my distro.conf it's wise to assign them using += since local.conf is parsed first, right? Mar 22 14:44:32 Just want to make sure i'm missing something Mar 22 14:44:35 basically the script will unpack the toolchain and stuffs on the path, only changing the paths with the path you choose on the installation Mar 22 14:46:34 Anticom: depends, really. you could just use = in the distro and if someone wants to extend they can use _append in local.conf. Mar 22 14:46:58 gnah... i forgett about _append way too often. Thank's for pointing that out Mar 22 14:47:15 wait what? but i thought _append is evaluated at parse time Mar 22 14:47:57 so it's added to PCK_FEED_FOO in local.conf and then will be set (--> overwritten) in distro.conf wouldn't it? Mar 22 14:48:40 Think i have to read that chapter in bb manual yet again .__. Mar 22 14:54:31 _append happens *after* = += etc Mar 22 14:58:25 kergoth explained it to me differently iirc Mar 22 14:59:22 _append/_prepend/_remove are lazy and happen at the end of parsing or at getVar()/expansion time, depending on bitbake version. =+/+=/.=/=. are immediate operations, at that exact line during the parsing Mar 22 15:01:17 kergoth: and ?= ??= are... complicated :P Mar 22 15:01:37 there's only one difference between the two Mar 22 15:01:47 the last ??= which is defined wins. the first ?= which is defined wins Mar 22 15:01:58 as ?= sets the var right there, whereas ??= is a lazy default Mar 22 15:03:40 rule of thumb: ??= is usually the right answer in config metadata, ?= or = is usually the right answer in recipes/classes Mar 22 15:04:03 but, if both are in use on the same variable, of course you need to consider that. ?= will always win over ??=, since ?= sets the var Mar 22 15:05:23 kergoth: hm but didn't you say that _append etc. was evaluated at parse time? Mar 22 15:05:27 or was that something different? Mar 22 15:06:09 as i just said, those are evaluated at the *end* of parsing in old bitbake versions Mar 22 15:06:23 not when encountering that line, the way the others do Mar 22 15:06:30 append/prepend/remove are lazy, the others are immediate Mar 22 15:11:26 Okay get it, thanks for clarification Mar 22 15:13:39 kergoth: iirc if I have FOO ?= "x", and FOO += "y" then in the end I'll get FOO = "y" right?, but if I had FOO_append = " y", the result would be FOO = "x y" Mar 22 15:17:46 nope. ?= is immediate. the line after the ?=, FOO is x Mar 22 15:18:02 if FOO was ??=, then yes, the += would immediately set the value and the default wouldnt' be used Mar 22 15:18:06 so i think you had them mixed up :) Mar 22 15:19:08 seems so ;) Mar 22 15:21:08 sometimes i think we should either pick one, be lazy or not, otherwise there's confusion about what happens when Mar 22 15:21:37 * kergoth tries removal all the global exports but PATH and adding task level exports for them all on configure, compile, and install, and fires off a world build Mar 22 15:21:42 s/val/ving/ Mar 22 15:23:13 given the time it takes for bitbake world, offworld build would seem more appropriate Mar 22 16:21:04 So I think I found out an old problem I have been having regarding not being able to run a Qt app on my poky image. When I readelf -l my image I see that there is a request for the program interpreter: /lib64/ld-linux-x86-64.so.2 Mar 22 16:21:34 I have found in "lib" in many places, but there is no such "lib64" that carries this Mar 22 16:22:10 Thus it can't find it and my Qt app cannot un on the target (presumably) Mar 22 16:22:57 My question is, how do I fix this? I am not even sure where in the build it is looking for this dependency Mar 22 16:24:03 riz_: the app that doesn't work, is it a binary blob, or did you build it with yocto ? Mar 22 16:24:48 I built it through QtCreator using the built and installed SDK Mar 22 16:24:52 (typically this happens when the build system compiles for a multilib x86 but your target is x86_64 only) Mar 22 16:25:16 hmm, then that's beyond my comprehension of cross-compilation Mar 22 16:25:33 In your first scenario, what would the fix be?> Mar 22 16:25:45 use the same distro for your sdk and your target Mar 22 16:25:51 in this case they don't match Mar 22 16:26:11 That is what I figured, but I did use the same distro Mar 22 16:26:18 How can I double check the match? Mar 22 16:27:44 Also, where are the files that it is pulling the lib from? I feel that if I just make a lib64 folder and throw the file in there it will work. Mar 22 16:28:48 you could symlink lib64 to lib and cross your fingers, but thats a terrible hack Mar 22 16:29:05 better to work out why your sdk thinks files are in a different place to the image on your target Mar 22 16:34:31 in my case it was a binary blob, I couldn't recompile... so my .bb for the blob made a symlink from lib64 to lib. ugly but it did fit my use-case Mar 22 16:34:40 (of course, I never pushed that upstream :P ) Mar 22 16:35:09 How did you do the symlink? Mar 22 16:35:15 I am unfamiliar with it Mar 22 16:36:52 And is that done in the installed SDK files? Mar 22 16:49:07 I did it manually by adding a lib64 folder and copied contents of lib into it. That fixed it Mar 22 16:49:36 How do I make it so that a lib64 folder is prebuilt in the image? Mar 22 16:54:15 riz_: unless you've found a really bad bug in the sdk, this is a configuration problem at your end Mar 22 16:55:36 rburton: my use case is still a valid one, though... (not asking for anything here, just mentionning it if someone ever pops up with a similar problem) Mar 22 16:56:47 if you had a single random binary you could fix the binary directly… Mar 22 16:59:02 yeah, I never figured out how, though... Mar 22 17:03:23 I guess my question is what configuration is really needed? I bitbake my poky image, then I bitbake the SDK. Not sure what I can configure or mess up in between. Mar 22 17:03:34 boucman_work: scripts/relocate-sdk.py does that Mar 22 17:03:54 riz_: either your image thats booting isn't hte one you built, or the sdk is broken, or something else is wrong Mar 22 17:04:12 fired off a sdk build to see Mar 22 17:08:25 rburton: with latest patches I posted today, now we can build world for qemux86, qemux86-64, qemuarm64 on musl, Can we enable these on builder ? Mar 22 17:09:16 assuming the ab agrees, then yes Mar 22 17:09:20 libunwind is still broken on mips* ppc* Mar 22 17:09:27 ppc has few more issues Mar 22 17:19:30 but I guess we can still start doing world build for three Mar 22 17:19:34 so we dont regress Mar 22 17:20:58 hey Mar 22 17:21:30 I need to override a Python distutils package gfortran path Mar 22 17:21:59 i can't get it to stop using my host native one Mar 22 17:22:44 can't find a way to override this setting Mar 22 17:41:10 anyone know of a backlight util for adjusting backlight (brightness) on LCD display without restoring to 'echo x >/sys/class/backlight... blabla Mar 22 18:31:46 what's necessary to build or install -dev packages? Mar 22 18:33:56 they are built as part of the main recipe build Mar 22 18:34:25 you can add them to the IMAGE_INSTALL.. variables or use the dev-pkgs IMAGE_FEATURE Mar 22 18:35:41 hmm, i've added it to the IMAGE_INSTALL but I don't have dev-pkgs Mar 22 18:41:06 dev-pkgs will install all -dev packages into the image Mar 22 18:41:14 makes them a bit large :) Mar 22 18:58:59 even with dev-pkgs, it's not being added. Could there be a common reason why a recipe wouldn't provide a -dev package properly? There's a FILES_${PN}-dev from a .inc file (https://github.com/bmwcarit/meta-ros/blob/master/recipes-ros/catkin/catkin.inc) Mar 22 19:03:00 hmm Mar 22 19:03:10 enable buildhistory and look at the output? Mar 22 19:04:55 i'll take a look Mar 22 19:07:37 that will show you what files are in what packages Mar 22 19:07:55 sometimes there are packaging issues and files get packaged in the wrong package Mar 22 20:41:54 Hmm, fray, you know why the rpm recipe sets %_libexecdir to %{libdir}/%{name} rather than obeying our libexecdir? was that needed because the previous default value included PN/BPN? Mar 22 20:42:08 just curious, i was fixing recipes to stop relying on the target paths being exported Mar 22 20:44:46 i seem to spend half my time on stack overflow correcting other peoples answers Mar 22 20:45:06 kergoth: seems like a fair assumption, fix it :) Mar 22 20:45:17 heh :) Mar 22 20:46:12 i hate breaking changes like removing global exports. there's never any way to know if it'll break external layers. worse yet, it can end up not breaking anything, just result in something not being packaged correctly Mar 22 20:46:14 rburton, there is a lesson there Mar 22 20:46:41 Crofton: stay away from SO? Mar 22 20:47:04 no Mar 22 20:47:19 we just suck at communicating Mar 22 20:47:34 i lack the patience for SO. i'm a lot more patient than i used to be, but still a ton of those questions are answered all over the docs, annoys me Mar 22 20:47:35 heh Mar 22 20:48:34 we are trying to build good answers on SO Mar 22 20:48:38 it is hard though Mar 22 20:48:50 I wonder if the BUILD_ vars should always be exported, or if recipes should opt-in if they need to build native tools as a part of their build process. plenty of such recipes have to define them to different variables anyway, as afaik there's no official naming (i.e. CC_FOR_BUILD) Mar 22 20:49:35 _FOR_BUILD is moderately standard in autotools Mar 22 20:49:48 at least it's suggested and autostars has macros Mar 22 20:50:17 (http://www.gnu.org/software/autoconf-archive/ax_prog_cc_for_build.html) Mar 22 20:50:49 in which case it's even less likely that defaulting to exporting *our* names is useful.. Mar 22 20:50:54 might prototype it, see how much explodes Mar 22 20:51:17 trying to do test migration to use the task level exports i'm working on, it's surprising just how much we export Mar 22 20:52:25 kergoth: autoconf exports CC_FOR_BUILD, agree that BUILD_CC is likely unused by anyone else Mar 22 21:35:27 Question: is it good idea to use function in local.conf ? I want do PREMIRROR_append for list of urls and for all protocol. I would pass them as bash array. If not what are the options ? Mar 22 21:48:08 mastier: create a .bbclass or .inc which uses either anonymous or inline python to manipulate PREMIRRORS, then either add the former to INHERIT or require the latter Mar 22 21:55:49 alternatively just set PREMIRRORS like a normal variable and use \n to split lines Mar 22 22:31:08 kergoth: ok, that's an idea, thanks, but maybe overkill Mar 22 22:31:45 i prepared local.conf so it uses variables from environment Mar 22 22:32:03 but sometinh like PREMIRROR_APPEND = Mar 22 22:32:10 *_append Mar 22 22:32:40 http://.* {BB_EXT_PREMIRROR_URL1} \n" Mar 22 22:32:50 the again _append Mar 22 22:33:15 http://.* {BB_EXT_PREMIRROR_URL2} \n" Mar 22 23:44:42 i want to override a kernel module (replace with out of tree module), but I get PR errors. I've tried adding the module via "PREFERRED_PROVIDER_virtual/kernel-module-" to my local.conf, but can't get it to work Mar 22 23:44:52 how can I accomplish this task? **** ENDING LOGGING AT Wed Mar 23 02:59:58 2016