**** BEGIN LOGGING AT Wed Apr 02 02:59:59 2014 Apr 02 03:16:56 When living on the bleeding edge of OE, instead of doing heavy rebuilds whenever core changes happen, would it be possible for a privat person to leach of the Yocto autobuilders sstate cache ? Apr 02 05:15:15 kroon: http://autobuilder.yoctoproject.org/pub/sstate/ Apr 02 05:15:37 kroon: however, I dont think its available for general consumption Apr 02 06:28:08 khem, aha Apr 02 06:39:16 Hi all! I need tips for own kernel module development in Arago project. I have TI6638 evm and I am trying to enable SRIO interface. So far I can compile and see my own module but can not call /drivers/rapidio/ functions in my module because of unknown symbol error when modprobed. I have added rapidio in Kconfig at /arch/arm and enabled it with menuconfig and can see objects in /drivers/rapidio folder. The problem seems to be that I get kernel compiled with Apr 02 08:48:33 morning all Apr 02 09:12:04 I want to put qt pieces into my image Apr 02 09:12:51 I tried IMAGE_INSTALL += packagegroup-core-qt and it builds all the qt pieces, but fails to build the rootfs with 'ERROR: packagegroup-core-qt not found in the base feeds' Apr 02 09:17:54 thaytan, do you get any *packagroup-core-qt* packages created under deploy/ ? Apr 02 09:18:41 the only rpm is packagegroup-core-qt-demoapps Apr 02 09:18:42 I've had some problems with debian.bbclass auto-renaming package names in the past Apr 02 09:19:05 maybe that'll work better Apr 02 09:20:25 thaytan: the question is what exactly are you looking for? if it's only the libs, packaging your application in the form of a recipe will pull those in automatically when you install the app's package Apr 02 09:20:48 (aside from anything configured as a plugin that is, e.g. qt's image format / database backend plugins etc.) Apr 02 09:21:02 bluelightning, a dev system that includes qt headers et al Apr 02 09:24:44 thaytan: qt4-x11-free-dev might be what you're looking for Apr 02 09:25:40 (+ qt4-x11-free-dbg for debug symbols if you want those as well) Apr 02 09:32:05 bluelightning, thanks - trying that Apr 02 09:32:30 rootfs build takes a fair while for this image / on this machin Apr 02 09:34:26 anyone familiar with the Freescale BSP here? Apr 02 09:36:21 i'm confused about how i.MX23 vs i.MX233 vs MXS is used in the BSP. for example there are two functions to generate an sdcard image: generate_imx_sdcard and generate_mxs_sdcard. i would guess that the i.MX233 CPU would use the first one, but the second one is used... Apr 02 09:44:48 bluelightning, thanks! Apr 02 12:07:51 hi all Apr 02 12:08:15 I am using numpy on a Cortex-A9 board (i.MX6) and dot product is very slow Apr 02 12:08:54 I would like to use ATLAS, as in its 3.10.0 version it supports arm and Neon, but there is no recipe for it in yocto, or maybe didn't I search it correctly... Apr 02 12:11:45 jmleo: looks like you'd need to write one Apr 02 12:13:08 bluelightning, I just started to, but I have some difficulties Apr 02 12:14:01 bluelightning, and I don't know in which recipe/layer I should put it Apr 02 12:14:09 well, I am starting with yocto, though ;) Apr 02 12:18:47 bluelightning, can you help me creating this recipe ? I have written an .inc and a bb file, it fetches the bz2 file, passes the checksum, but I can't execute configure "as is" because when untaring, it creates a directory, here is the tree : Apr 02 12:19:07 jmleo: sounds like you need to set S to point to where the source is extracted Apr 02 12:19:11 within the recipe Apr 02 12:20:42 http://pastebin.com/zY68XeWf Apr 02 12:21:09 mmmh, this is what I didn't understand, this "S" thing... Apr 02 12:22:03 jmleo: basically, what bitbake does is just extract the archive with WORKDIR as the current directory Apr 02 12:22:24 so if the archive contains a subdirectory, that'll of course be created under WORKDIR for the recipe Apr 02 12:22:51 you must therefore set S to point to that directory so the system knows where to find the base of the sources Apr 02 12:23:11 there are lots of examples in the recipes we have Apr 02 12:23:57 jmleo: FYI you may find this new section of the manual useful: http://www.yoctoproject.org/docs/1.6/dev-manual/dev-manual.html#new-recipe-writing-a-new-recipe Apr 02 12:24:13 (note this is the 1.6 manual, but this section should be completely applicable to older versions as well) Apr 02 12:25:22 bluelightning, thx a lot for this pointer, favorited ! Apr 02 12:28:15 bluelightning, in order to fully understand, WORKDIR in my case is pointing to yocto/build/tmp/work/cortexa9hf-vfp-neon-poky-linux-gnueabi/atlas/3.10.1-r0/ right ? Apr 02 12:34:18 ok, this is it... now have another issue, as I need in the configure to create a directory and execute configure from this directory, but if I understand correctly, I can override do_configure... Apr 02 12:41:20 you can see the steps here : http://math-atlas.sourceforge.net/atlas_install/node6.html Apr 02 12:47:50 jmleo: you should just be able to set B = "${WORKDIR}/build" to have a separate build directory for the recipe Apr 02 12:55:40 bluelightning, thx and I can also override configure options, because this is not a real autotool builder, and --build=x86_64-linux for instance is not a recognised option. I see one can pass EXTRA_OECONF but maybe should I just write the correct do_configure Apr 02 12:57:39 jmleo: if this isn't autotools then you shouldn't inherit the autotools class, and instead yes you should define your own do_configure Apr 02 13:03:51 I'm back Apr 02 13:10:34 hi, is there any special way of executing tests in Yocto? For instance I have binutils with whole testsuite. In order to execute them manually I need to 'make something'. Does bitbake do any wrapping around testing commands? Apr 02 13:15:41 Xz: we have ptest which is designed to help with packaging and running tests on the target Apr 02 13:16:41 bluelightning, if I want to pass the compiler path, is there a ${CC} variable I can use ? Apr 02 13:17:00 jm__: yes Apr 02 13:17:39 bluelightning: I'm interested rather in sth like unit-tests on host machine, executed just after compilation. If they fail - i would fail the build Apr 02 13:19:56 Xz: we don't have anything like that at the moment as far as I know, but it would be trivial to add another task to do that, and as long as the task function returns non-zero when it fails (assuming it's a shell function) the build will fail if it fails Apr 02 13:21:01 Xz: we do now have oe-selftest for testing the build system, but that's not quite designed for this kind of thing Apr 02 13:33:24 bluelightning: alright, thanks Apr 02 13:36:28 bluelightning, I have a silly question, but is there a recipe to compile and install on a board a gcc it can execute (arm in my case) ? I see gcc, gcc-cross, etc. Apr 02 13:37:05 jm__: the recipe for the target is just called "gcc" Apr 02 13:37:50 jm__: if you want to do compilation on the target you may want to add packagegroup-core-buildessential to your image since that adds a few other useful pieces Apr 02 13:52:17 is there a way that I can incorporate the version of a package I have in my image as part of the image name? Apr 02 14:26:26 bluelightning, in my SRC_URI, I use a https:// but it should clone a git (this is on github) Apr 02 14:32:35 so I changed SRC_URI and use git:// with a protocol=https at the end, which works Apr 02 14:32:44 but again, it creates a directory named git Apr 02 14:33:42 yes Apr 02 14:33:59 typically you then set S to ${WORKDIR}/git Apr 02 14:35:03 rburton, thx Apr 02 14:54:00 hi may someone help me? I got some trouble with meta-gir ...for g-ir-tools-host Yocto builds the a package with wrongs specs. e.g. g-ir-tools-host-dev.rpm is beeing configured to require g-ir-tools-host.rpm which isnt deployed by the recipe.. I tried to modify the .spec-file but its getting restored to original state once I rebuild the rpms with task package_write_rpm Apr 02 15:09:16 RP: any idea why some failed tasks aren't shown in cooker log anymore? "ERROR: Function failed: .." isn't shown for some (maybe all) failed do_fetch, do_package tasks Apr 02 15:09:29 RP: but they are shown in final summary Apr 02 15:10:11 RP: and I'm using knotty with "| tee log" if that matters here Apr 02 15:11:41 weird Apr 02 15:12:13 * JaMa trying to provide simple reproducer and to test if "| tee" causes this Apr 02 15:12:41 and I think it happens only in last ~2 weeks Apr 02 15:13:16 maybe worker thread can die before sending error message? Apr 02 15:15:46 JaMa: is this the stderr thing? Apr 02 15:15:59 that changed recently I believe Apr 02 15:16:14 bluelightning: you're right Apr 02 15:16:25 * JaMa should expect something simple as that :) Apr 02 15:16:32 that reminds me I had better add that to the migration notes Apr 02 15:17:04 yes please do, I'm doing | tee log everywhere and this is the most important part which is then missing Apr 02 15:20:01 and other unrelated question, does someone remember git fetcher change, that it now removes the git checkout from downloads directory when e.g. checking the revision in branch fails? Apr 02 15:20:44 maybe it was always like this and I haven't noticed, but it's quite sad for very big repos when first thing I want to do is "git branch -a --contains $SRCREV" in downloads/git2/repo and it's already gone Apr 02 15:21:18 only .done stamp is there (which maybe should be removed instead) Apr 02 15:23:37 JaMa: hmm, that sounds like a side-effect rather than something deliberate Apr 02 15:24:34 * JaMa trying to reproduce/confirm on something small like dtc Apr 02 15:27:51 hmm didn't happen now, will dig a bit more.. Apr 02 15:29:22 maybe it happens only for linux-yocto because it is using fetcher in a bit specific way and that's where I was seeing this last time Apr 02 16:00:34 JaMa: ah, do you redirect stdout? Apr 02 16:00:42 JaMa: errors now go to stderr Apr 02 16:01:05 JaMa: perhaps we're not logging stderr Apr 02 16:08:38 can I have two bbappends to one recipe in the same directory? :) Apr 02 16:08:52 no cause the filenames would overlap Apr 02 16:08:59 but you can in two different directories.. Apr 02 16:09:23 applicable order is indeterminate though.. so be careful if you try that Apr 02 16:22:13 Xz: why would you want that? Apr 02 16:41:36 RP: yes that was exactly the case, somehow I didn't notice bitbake change to split them Apr 02 16:51:19 hmm linux-yocto removed the checkout (dtp recipe didn't) Apr 02 16:57:15 ERROR: Running idle function Apr 02 16:57:15 gah Apr 02 16:57:26 always getting odd random failures in this vm Apr 02 17:03:31 kergoth: we have see issues due to broken virtualisation before :/ Apr 02 17:04:32 ah, whew, in this case it's just an OOM kill, at least it's not inexplicable.. now to reserve more ram to myself so our automated builds don't steal everything ;) Apr 02 17:05:40 i'm gonna say the latest virtio/kvm drivers have been pretty fast and solid Apr 02 17:06:41 as long as you keep all the parts in sync it seems to work pretty well Apr 02 17:42:27 JaMa: still around? Do you have a prototype of the "world-image" recipe? Apr 02 17:46:09 sgw_: I have actually 2 different implementations, but it's blocked by this issue: http://lists.openembedded.org/pipermail/openembedded-core/2013-January/073968.html Apr 02 17:46:38 I should retry with newer oe-core and bitbake Apr 02 17:49:28 JaMa: I can give it a whirl locally also if you want Apr 02 17:57:51 sgw_: I've rebased all 3 branches and added links to all in the bugzilla if you want to try it, probably will need small update (e.g. I've noticed old LIC_FILES_CHKSUM) Apr 02 19:42:00 * kergoth grumbles Apr 02 19:47:49 gah, this makes no sense Apr 02 19:47:49 fatal: Unable to find remote helper for 'http Apr 02 19:48:00 git from my buildtools-tarball is acting like RUNTIME_PREFIX isn't enabled, in some cases Apr 02 19:48:01 weird Apr 02 19:48:06 hmm **** ENDING LOGGING AT Thu Apr 03 02:59:59 2014