**** BEGIN LOGGING AT Thu Jan 24 02:59:59 2013 Jan 24 03:46:55 Hi there Jan 24 03:46:59 anybody here Jan 24 03:47:01 ? Jan 24 03:47:21 if you have a question, ask it and wait Jan 24 03:47:31 whether anyone is looking at this windowr ight now is irrelevent Jan 24 03:49:53 I want to build core-image-minimal with external toolchain(code sourcery), and i just only want to build the image only, but the OE system always try to build the whole SDK. do i have someway that just build to get image only? Jan 24 03:50:13 i tried search arround but still don't have any answer Jan 24 03:57:40 how are you configuring the use of the external toolchain? Jan 24 03:58:34 TCMODE = "external-sourcery" Jan 24 03:58:50 EXTERNAL_TOOLCHAIN = "" Jan 24 03:59:37 modified tcmode-external-sourcery.bb to fix some install issue. sysroot extraction. Jan 24 04:01:05 the compile is inprogress, but it compile lot of things as SDK. that i don't need. Jan 24 04:02:05 hmm, that should be working. You could try using mentor's meta-sourcery layer. that's what we use for it. https://github.com/MentorEmbedded/meta-sourcery - just add to bblayers and set EXTERNAL_TOOLCHAIN Jan 24 04:02:27 can i tell OE-core just compile image only? don't compile SDK? Jan 24 04:03:02 i don't understand your question Jan 24 04:03:12 it isn't building an sdk, it's just obeying the dependencies of the recipes Jan 24 04:03:17 which includes a crosscompiler Jan 24 04:03:41 the external-sourcery tcmode is supposed to configure the preferences for the providers such that the external-sourcery recipe satisfies that, and it doesn't have to build any toolchain components from source Jan 24 04:03:48 if it's not doing so with the oe-core content, then that's a bug Jan 24 04:12:44 Example i just want to to compile core-image-minimal for powerpc. but in tmp-eglibc/work directory i can see "all-oe-linux ppc603e-oe-linux qemuppc-oe-linux x86_64-linux" Jan 24 04:13:09 and the build process take a long time Jan 24 04:13:58 how can i ignore x86_64-linux build? that's the one i don't need. Jan 24 04:15:28 no, you're incorrect. it's building only what it needs to build. Jan 24 04:15:50 there are many tools which are run during the build process of crosscompiled software, and we can't rely on all of them being on the host Jan 24 04:15:54 so it builds what it needs Jan 24 04:38:36 Yes, i'm wrong, actually Jan 24 04:39:25 if you look more closely you can see exactly what it's building Jan 24 04:39:32 e.g. autoconf-native, libtool-native, .. Jan 24 04:44:09 yes, that just above compile too much. maybe dependencies for qemu, ... Jan 24 08:39:02 good morning Jan 24 09:39:40 morning all Jan 24 09:39:48 morning bl Jan 24 09:39:58 morning bluelightning Jan 24 09:40:06 hey rburton Jan 24 09:49:00 morning all Jan 24 09:55:10 hi RP Jan 24 11:32:47 and a cold morning it is Jan 24 12:33:15 good morning people Jan 24 12:37:23 rburton, I screwed up and sent you an email meant for the list, can you forward it back to the list :) Jan 24 12:42:30 florian: both; JaMa|Off had found some sstate-cache issues; dunno if it has been fixed already or not Jan 24 12:43:23 guys, i'm setting up my x developments tools Jan 24 12:43:26 by the adt tutorial Jan 24 12:43:46 with this tutorial which devel tools i'm setting up? qt stuff? c stuff? Jan 24 12:44:43 florian: otavio: not finished, still WIP in jansa/test branch Jan 24 13:42:58 hey everyone Jan 24 13:43:08 i'm setting up my development tools Jan 24 13:43:12 with the adt tutorial Jan 24 13:43:31 aand after executing the script in /opt/poky/1.3 Jan 24 13:43:44 something happenned to my build environment... Jan 24 13:43:58 could someone explain to me what happens when i execute the script from adt? Jan 24 13:45:22 hm Jan 24 13:45:25 now i understood Jan 24 13:45:28 when i exec that script Jan 24 13:45:54 this one set some environment variables Jan 24 14:05:10 hello Jan 24 14:14:12 I meet a strange issue for PACKAGECONFIG Jan 24 14:14:32 https://gist.github.com/34585bc559aa597f9ead anyone can help me Jan 24 15:52:14 bluelightning, I have understand pkg_postinst, when it's in host ${D} exist, we can do something on host, else we can do something on the real HW(target) But I don't know how it is related with "firs boot" Jan 24 15:52:21 first boot Jan 24 15:53:13 lyang0: so the postinstall is meant to be run right after the package is installed Jan 24 15:54:01 lyang0: if the package is installed into the image, then the system will try to run it when constructing the image (do_rootfs) Jan 24 15:54:27 if it fails at that time, we put a copy of the script into a directory (e.g. /etc/rpm-postinsts/ for RPM) Jan 24 15:55:03 then there is a script that runs on startup on the target, and if any script files exist in that directory they get run and then deleted if they succeed Jan 24 15:55:35 so naturally if there are scripts in that dir they will get run on the first boot, and since they are deleted afterwards (assuming they succeed) then they won't be run again Jan 24 15:55:42 lyang0: does that help? Jan 24 15:56:12 let me think Jan 24 15:56:44 great Jan 24 15:56:51 I understand Jan 24 15:57:22 where delete them in code? Jan 24 16:00:46 lyang0: not sure exactly; for ipk it's in the run-postinsts script that's installed by the run-postinsts recipe Jan 24 16:01:14 thanks Jan 24 16:04:12 which place is used to put the native packages(ie:qemu-native), is there a variable to tell the place, and it export to $path Jan 24 16:44:00 hi, how to create my own branch in poky-contrib, is there a permission require? or register in some place then have the permission? Jan 24 17:03:43 lyang0: I think you can get your answer here: https://wiki.yoctoproject.org/wiki/Poky_Contributions#Poky_Contrib_Branch Jan 24 17:05:21 lyang0: see the Git workflow section Jan 24 17:06:36 i get Jan 24 17:06:39 Permission denied (publickey). Jan 24 17:06:40 fatal: The remote end hung up unexpectedly Jan 24 17:07:19 lyang0: did you read the bit about "sending your keys" Jan 24 17:09:20 do you mean To access the poky-contrib tree please send Richard Purdie or Michael Halstead an ssh key from each person needing access. You will then be able to clone and push to: Jan 24 17:09:48 ok seems I need email key to them Jan 24 17:10:36 hey guys Jan 24 17:11:12 how do i configure my eclipse to cross development with yocto? i couldn't understand the info about "Cross Compiler Options" Jan 24 17:11:24 lyang0: yes Jan 24 17:16:01 :thakns Jan 24 17:16:04 thanks Jan 24 18:12:36 gbrennon: if you installed toolchain using either .sh or adt-installer, choose standalone pre-built toolchain, e.g. I installed my toolchain at /home/jzhang/test-toolchain, so my toolchain root location is that, for sysroot, if you generate your toolchain using bitbake image -c populate_sdk, then you can point your sysroot at the sysroot directory underneath your toolchain, or if you extract your sysroot using one of the rootfs (need to be a -dev or -sdk ima Jan 24 18:26:40 jzhang i installed through .sh Jan 24 18:26:55 jzhang but when i select standalone pre-built it stilll give me error... Jan 24 18:27:47 "You need to specify toolchain location before.." Jan 24 18:29:28 jzhang, i'll reinstall my .sh Jan 24 18:49:17 jzhang it was something with the .sh install thx. i don't think it is 100% functional, but i'm watching some youtube vids about this Jan 24 19:15:12 is there any gtk or qt kit for yocto project? what is the name of the package so i can include in my compilation? Jan 24 20:07:05 holy spamming vivijim Jan 24 20:29:47 sgw1: do you know when eglibc updates are gonna flow into master ? Jan 24 20:34:42 khem in the next week, we are wrapping up the m3 build and I am testing with your changes in MUT right now. Jan 24 20:35:04 ok I ask because I have some systemd and uclibc updates Jan 24 20:35:08 coming in Jan 24 20:35:58 khem, sure I understand Jan 24 20:36:21 and python3 Jan 24 20:36:37 oh well SIGFOOD bbl Jan 24 20:37:00 Hi all. I'm trying to create a new package that installs stuff in /usr/local (for historical reasons clients want it left there). When I try to build the image I get a QA warning about Files/directories being installed but not shipped. What do I need to tune so that the things I've put in /usr/local will get shipped? Jan 24 20:37:38 FILES_ += "/usr/local/..." Jan 24 20:37:48 ah, nice Jan 24 20:38:04 hrmm usr/local Jan 24 20:38:06 note, the packagename is not the recipe name Jan 24 20:38:36 install them in /usr/bin and create symlinks in /usr/local Jan 24 20:38:45 and one day switch over Jan 24 20:39:26 yeah, I might do that, but I still need the /usr/local ones to be packaged :) Jan 24 20:40:12 sure fray gave you the magic for that already Jan 24 20:40:28 yeah, I'm giving that a try now Jan 24 20:40:29 thanks Jan 24 20:54:28 hum, can I set PACKAGE = "recipe_name" and FILES_ = "/usr/local/*" and pick up everything I've installed there? Jan 24 20:55:11 (I'll grant you this isn't the right way to do this, but I'm somewhat restricted in what I can change.) Jan 24 20:56:45 I'd like a single package with everything that I've installed in ${D} Jan 24 20:57:25 FILES_${PN} += / Jan 24 20:57:43 'er.. make that FILES_${PN} += '/' Jan 24 20:58:07 PACKAGES = "${PN}" Jan 24 20:58:18 those two lines will result in producing one package with all the files.. Jan 24 20:58:26 thanks Jan 24 20:59:53 heh, progress :) Jan 24 20:59:58 now I get a bunch of QA errors Jan 24 21:00:50 No GNU_HASH in elf binary; non -dev/-dgb/-nativesdk package contains symlink .so; non debug package contains .debug directory Jan 24 21:01:26 packaging splitting is expected.. which is why "one big package" likely isn't what you want Jan 24 21:01:57 yeah, I know, and the QA stuff looks very helpful in doing it right Jan 24 21:02:29 but my hands are somewhat tied right now until I convince others to do things right Jan 24 21:03:09 is there a way to disable QA for a recipe/package? Jan 24 21:07:43 or globally, if necessary Jan 24 21:09:48 or I guess I can split the packages according the QA errors, and leave everything /usr/local Jan 24 21:10:04 that shouldn't affect the users, and it should satisfy QA Jan 24 21:10:09 the 'PACKAGES = ...' is what changed the splitting.. Jan 24 21:10:32 it's likely better to simply not change PACKAGES =.. and instead add the missing files to "FILES_${PN} =" Jan 24 21:10:50 note, you don't have to list everything individually, you can list directories, and everything in that directory is picked up.. Jan 24 21:10:52 so have PACKAGE="${PN}-dbg ${PN}-..." and split FILES_${PN}-dbg = "..." FILES_${PN}-... Jan 24 21:11:07 if you get errors about static files, debug files, etc.. add the assorted directory to the appropriate FILES_... Jan 24 21:11:18 I'll give that a shot -- thanks Jan 24 21:13:33 If FILES_${PN} contains "/" and FILES_${PN}-dbg contains "/usr/local/bin/.debug" will that still cause a problem for the former? Jan 24 21:14:00 directories and files are processed in the order of PACKAGES Jan 24 21:14:13 ah, so just list the most generic one last Jan 24 21:14:14 so assuming -dbg is processed before ${PN}.. then it'll work fine Jan 24 21:14:21 cool Jan 24 21:14:31 thanks again Jan 24 21:16:20 I'm not familiar with the 'No GNU_HASH' problem, is there a way to resolve that? Jan 24 21:16:40 means the applications were built with a different compiler -- or were built using odd settings.. Jan 24 21:17:24 yeah, that's likely :) Jan 24 21:17:43 you can turn off the QA check if it's an external binary you just have to live with.. Jan 24 21:18:34 hum, how? Jan 24 21:19:02 trying to find an example... Jan 24 21:20:26 INSANE_SKIP_ = "ldflags" Jan 24 21:20:39 need that for each package that reports the warning/error Jan 24 21:20:52 this is in the recipe, or in a .conf file? Jan 24 21:21:05 in your recipe Jan 24 21:21:12 alright Jan 24 21:21:21 INSANE_SKIP_${PN} = "ldflags" Jan 24 21:21:29 INSANE_SKIP_${PN}-dev = "ldflags" Jan 24 21:21:30 .... Jan 24 21:22:11 (the warning is valid, cause it does show that the binary didn't come from the build system -- and performance and compatibility may be impacted... but if you've verified it, then you can disable that test.) Jan 24 21:22:42 yeah, that's part of what I'm trying to do. I want to build an image to test if the binaries will work as expected Jan 24 21:28:40 I did something pretty dirty -- I overrode the do_package_qa function Jan 24 21:29:12 at least long enough to perform this test Jan 24 21:29:28 I'd hate to go to all the work of spliting thing up to find that it won't work anyway Jan 24 21:29:56 you can do that.. or you can do: WARN_QA = "" \n ERROR_QA = "" Jan 24 21:31:01 ah, good to know. I'll take all of the back to the folks providing these files as evidence that they should change how they're doing things :-) Jan 24 21:31:29 best that you can do is provide them an SDK, and ask them to use the compiler from the SDK to build the apps.. Jan 24 21:31:39 then you can package them (and you shouldn't have any errors/warnings) Jan 24 21:31:53 yeah, that's the end goal Jan 24 21:33:08 I really appreciate your help -- thanks again Jan 24 21:33:18 np Jan 24 22:02:59 hmm so I got these rogue packages which just wont work with sstate without a major surgery in its own build system and much more Jan 24 22:03:09 it seems a bigger ask to please sstate Jan 24 22:03:27 what I would prefer is to turn off sstate of this package Jan 24 22:03:31 and this package only Jan 24 22:03:54 is this something someone is already doing ? Jan 24 22:04:17 what would happen is that this package will rebuild all the time when fetched from sstate Jan 24 22:04:25 which is a hit ok to live with Jan 24 22:05:35 what I did is use SSTATE_MIRRORS = "" Jan 24 22:05:42 in the given recipe Jan 24 22:06:56 gm! Jan 24 22:08:00 Just ran into a strange issue building an older kernel (2.6.36) for a client using Yocto 1.3 (target system is OMAP3 based) Jan 24 22:08:48 the build completes successfully, but the boot hangs at "Starting kernel ..." Jan 24 22:09:14 and yes, I took into account the ttyS/O changeover Jan 24 22:10:43 I happened to have an archive of my old OE classic build setup and have verified that a build from the same SRCREV, same defconfig boots properly Jan 24 22:11:05 Is anyone aware of toolchain changes that may have broken older kernel builds? Jan 24 22:18:35 sakoman: there were bunch of changes in USB stack that triggered kernel bugs with gcc 4.7 Jan 24 22:18:45 sakoman: Do this Jan 24 22:18:54 compile just the kernel with say gcc 4.6 Jan 24 22:18:58 from 1.2 Jan 24 22:19:10 and leave the rest with 1.3/gcc 4.7 Jan 24 22:19:16 and see if that gets you any further Jan 24 22:20:18 hmmm . . . changes in USB stack?? Jan 24 22:20:36 doesn't 2.6.36 predate gcc 4.7? Jan 24 22:20:54 so it is pre-existing code that doesn't work! Jan 24 22:21:20 sakoman: here 2.6.31 does not boot... 2.6.38/9 do..but I think it's flaky kernel source :/ Jan 24 22:21:43 this on armv4 with last tc Jan 24 22:23:05 ant, I suppose I can try later kernels and see where they start working again . . . Jan 24 22:24:20 but they want a particular kernel so I may have to do something ugly with toolchains :-( Jan 24 22:31:28 So I just checked out poky, and also oe-core/bitbake, and either way, bitbake core-image-minimal explodes in a festival of "Error during finalize" messages. Anyone else having issues recently? Jan 24 22:31:49 ... wait, what. I just tried it again and no problems. Now I'm super confused. Jan 24 22:32:15 I'm having success with gcc 4.7 and 2.6.34 / .38. Needed a couple cherry-picks to work around build problems, but never anything at runtime. Jan 24 22:34:51 Huh. I must have run my poky-directory build with a stale environment. I still have problems if I do the same thing in an oe-core directory, though. Weird. Jan 24 22:45:20 khem: I suspect this might be the issue: http://seabright.co.nz/2012/06/11/kernel-not-booting-with-linaro-gcc/ Jan 24 22:51:05 sakoman: testing now Jan 24 22:57:24 ant_home: well not likely to work for me since I already have the referenced patch in my 2.6.36 repo Jan 24 22:58:05 I have that 2.6.31 around Jan 24 22:59:26 ah, operator error! my fingers checked out 3.6 not 2.6.36 :-) Jan 24 22:59:32 I do that too often! Jan 24 22:59:47 2.6.36 is missing the fix, will try it now Jan 24 23:06:31 argh..2.6.31 lacks devtmpfs... Jan 24 23:06:40 * ant_home rebuilds the cpio Jan 24 23:15:29 ouch, again, device_table_minimal lacks input and more Jan 24 23:37:09 sakoman: well, still not booting Jan 25 01:54:57 Hi, although this is a very trivial question, What does sato mean? **** ENDING LOGGING AT Fri Jan 25 02:59:58 2013