**** BEGIN LOGGING AT Thu Jan 29 03:00:00 2015 Jan 29 07:01:56 does core-image-minimal support ip networking, java throws an exception that it can't get ip of localhost Jan 29 08:35:30 good morning Jan 29 08:35:57 morning Jan 29 10:01:38 good morning Jan 29 11:39:55 hi all Jan 29 11:44:47 hi bluelightning Jan 29 11:44:55 hi lpapp Jan 29 12:07:08 Hi All, I've built gstreamer-vaapi-1.0_0.5.8.bb fine and then created a new file called gstreamer-vaapi-1.0_0.5.9.bb which also builds fine. However, when putting PREFERRED_VERSION into my local.conf to build the earlier version I'm hitting "Package gstreamer-vaapi-1.0 was not found in the pkg-config search path." for a later recipe. Any ideas on how to investigate this? Jan 29 12:08:44 I've tried doing a "-c cleanall" on the recipes involved but still hitting the same error Jan 29 12:37:51 why is md5sum reporting different hashes for a build that was incrementally created after a slight change in a recipe and the build created with the same change, but from scratch? Jan 29 12:37:59 I thought it ought to be the same. Jan 29 12:38:23 I just updated the SRCREV to include a build fix Jan 29 12:38:35 (in our own software) Jan 29 12:40:18 nvm, I know why... timestamps are different Jan 29 13:04:40 bluelightning: hi, For some time I have been trying to follow the guidelines provided in the reference manual regarding use of INITRAMFS_IMAGE (and friends) variables to generate a ramdisk.but each time I try I hit upon circular dependency issues. Jan 29 13:07:17 bluelightning: googling around, I find references to core-image-minimal-initramfs and I believe I am doing as suggested. still doesn't work. I currently have a work around but it's ugly and causing me problems at runtime. Is there a "clear" set of instructions that I can follow? the documentation seems to be patchy on the subject Jan 29 13:17:53 darkhorse: it is very patchy on that area I'm afraid, it doesn't really help but FWIW I've filed a documentation bug about it (bug 7096) Jan 29 13:17:55 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=7096 enhancement, Medium, 1.8, bruce.ashfield, NEW , Document how to use initramfs features Jan 29 13:28:23 bluelightning: just had a breakthrough NOW!! Jan 29 13:29:41 bluelightning: the problem was I was doing everything okay, but then I had an extra task that I run on most of my images, as soon as I disabled that task, circular deps are not happening anymore ;) Jan 29 13:29:56 ah right, I see Jan 29 13:30:11 bluelightning: but surely there is a lack of documentation on the issue Jan 29 13:30:33 darkhorse: don't know if this is helpful, but I have two separate local.conf files I use. One for regular image and the other for initramfs. Jan 29 13:31:09 FileNotFound: that's a bit of an extreme solution, it's pretty easy to specify configuration on a per-recipe basis, even from a conf file Jan 29 13:31:42 e.g VARNAME_pn-yourrecipe = "somevalue", or to append, VARNAME_append_pn-yourrecipe = "somevalue" Jan 29 13:31:49 FileNotFound: hmm.. I actually define two different machines and hence two machine configs - one for initramfs based kernel and one for standard kernel - i gues that is very similar to what you are doing Jan 29 13:32:25 darkhorse: indeed Jan 29 13:33:43 the only things in the initramfs local.conf not in the regular one is I define INITRAMFS_IMAGE, INITRAMFS_IMAGE_BUNDLE, and IMAGE_FSTYPES Jan 29 13:34:31 FileNotFound: that's right Jan 29 13:40:06 yo Jan 29 13:40:29 i have created core-image-minimal Jan 29 13:40:37 it's bootinhg with qemui Jan 29 13:40:49 but i'd like to change the keyboard layout Jan 29 13:40:55 is it possible ? Jan 29 13:41:12 (of course, loadkeys isn't installed) Jan 29 13:41:31 hi every one good evening Jan 29 13:41:50 im a new on embedded linux Jan 29 13:42:08 so i wantt to compile kernell Jan 29 13:42:23 but i dont know detail about this Jan 29 13:42:52 anybody can tell me how its do Jan 29 13:42:53 ? Jan 29 13:44:42 shinobi: well your platform usually should have some kind of documentation. Jan 29 13:45:15 shinobi: but in general, getting started with a kernel cross compile is no different from a standard compile, you just need a proper toolchain available Jan 29 13:52:52 shinobi: pelase keep the conversation in-channel, so others can learn and/or help too, ok? Jan 29 13:54:51 shinobi: and there actually is a lot of documentation and in most cases even step-by-step instructions out there on how to compile a kernel for the most popular platforms. Jan 29 13:55:50 shinobi: so once you picked the board you want to work on, there usually is a good starting point already there (if you don't use something highly special or customer specific) Jan 29 13:56:44 shinobi: especially for the beaglebone black and the raspberry pi, there should be boatloads of resources. Jan 29 13:57:49 shinobi: and besides cross-compilation, only the standard linux knowledge in terms of compiling etc. applies. so if you're having troubles with the terminology, no need to go embedded - just start by learning your way on a standard linux box. Jan 29 16:49:18 bluelightning: i have c++ apps in my image which require libstdc++.so.6 but the build system can't track this dependency? Jan 29 16:51:17 darkhorse: probably because it doesn't need a linker flag Jan 29 16:54:34 LetoThe2nd: how do you mean? I am not passing any linker flags to the C++ apps Jan 29 16:55:31 darkhorse: exactly thats the point. i guess (*really* *guess*) that the dependency tracking works with the flags that the linker gets passed. but libstdc++ is implicitly available, so it does not get tracked. Jan 29 16:57:34 LetoThe2nd: but then libc is the same which is traced Jan 29 16:58:32 darkhorse: more guessing: maybe libc is just always there unless you explicitly make everything statically linked? i mean, even busybox needs it. Jan 29 16:58:49 (but of course everything i just typed can be plain wrong) Jan 29 16:59:16 LetoThe2nd: sure - but i do appreciate what you said. thanks Jan 29 17:05:41 LetoThe2nd: darkhorse: it should work based upon the shared library dependencies, the system doesn't look at linker flags Jan 29 17:07:16 bluelightning: ah ok, thanks for the correction! Jan 29 17:07:34 so I'm not sure why libstdc++ wouldn't be picked up Jan 29 17:07:56 dlopened libraries won't be detected, naturally, but libstdc++ isn't handled that way AFAIK Jan 29 17:14:55 am I mistaken or wic is not able to create gpt partitions ? Jan 29 17:16:43 tomz: ^ Jan 29 17:17:29 bluelightning: then clearly something is not right. my app depends on libstdc++ and it crashes when unable to find it. I tried to add an RDEPENDS += "libstdc++" which didn't work. then i added it to IMAGE_INSTALL and then give an build error Jan 29 17:23:04 tomz: other huge issue with wic, it is not able to create partitions smaller than 1MB Jan 29 18:05:59 kergoth: hi, i am using meta-sourcery and when i say "bitbake meta-sourcery-toolchain" it all goes fine. but the packages don't all seem to be populated. specifically listdc++ package is empty Jan 29 18:06:10 there's no such thing as meta-sourcery-toolchain Jan 29 18:06:44 kergoth: but a pacakage with this name gets built into my work area! Jan 29 18:06:48 no, it doesn't Jan 29 18:06:57 there's never been a package called that in meta-sourcery in the history of its existence Jan 29 18:07:03 perhaps you mean external-sourcery-toolchain? Jan 29 18:07:22 what branch are you using of both meta-sourcery and poky/oe-core, and what version of the Sourcery G++ toolchain are you using? Jan 29 18:07:28 kergoth: sorry - yes Jan 29 18:08:49 kergoth: i am on poky-dizzy-12.0.1 and am using 2014.12-async1 Jan 29 18:10:20 kergoth: i am using a pre-built sourcey toolchain versioned 4.7.3 Jan 29 18:11:32 insufficient information Jan 29 18:11:39 4.7.3 is the gcc version, not the sourcery toolchainv ersion Jan 29 18:13:48 kergoth: to be honest, i don't know. this is what is installed on the company build machine. i don't think i can get the sourcery version info from installed binaries Jan 29 18:14:05 yes, you can, it's included in the gcc —version output Jan 29 18:14:09 or gcc -v, or whatever it is Jan 29 18:14:27 that version info shows both the gcc version and the vendor version information in parentheses Jan 29 18:14:36 meta-sourcery runs it to extract the version information for its packages Jan 29 18:14:55 kergoth: that's what i tried. here's the outcome: ./g++ --version g++ (GCC) 4.7.3 Copyright (C) 2012 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. Jan 29 18:14:58 alternatively, you may be able to just examine CSL_VER_MAIN or SOURCERY_VER_MAIN in bitbake -e Jan 29 18:15:04 that's not the external sourcery toolchain. Jan 29 18:15:42 you need to be running /path/to/sourcery/install/bin/-gcc --version Jan 29 18:17:51 kergoth: oops! sorry I am an idiot. there's a native version of the same Jan 29 18:18:05 kergoth: here's the actual cross-toolchain: mipsel-unknown-linux-gnu-gcc (crosstool-NG 1.19.0) 4.7.3 Copyright (C) 2012 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. Jan 29 18:18:35 meta-sourcery is a layer for support of the mentor graphics sourcery g++ toolchain specifically. no other external toolchains are guaranteed to work at all. they might, but no guarantees Jan 29 18:18:48 otherwise it serves as an example of how to support a given vendor external toolchain Jan 29 18:18:55 e.g. wind river has their own layer for their toolchains Jan 29 18:19:16 in other words, it's no surprise that crosstool-ng's toolchain isn't necessarily doing what it should with meta-sourcery, particularly of that branch Jan 29 18:20:44 kergoth: hmm. but when i look under the image directory, it does seem to populate all libraries including libstdc++. it's only at the packaging stage, it decides to keep it empty :) Jan 29 18:21:05 most likely its filename doesn't perfectly match the FILES entries in the recipe Jan 29 18:21:09 again because its an unsupported toolchain Jan 29 18:21:26 you might be able to make adjustments to get it packaging it without too much difficulty Jan 29 18:22:12 kergoth: thanks a lot! I will have a look again. Jan 29 18:22:18 np Jan 29 18:23:17 the master branch of meta-sourcery has actually been tested with crosstool-ng, though i'm not sure if it's actually packaging libstdc++ offhand Jan 29 18:23:30 of course the master branch would need master of poky/oe-core unless you adjust the appends Jan 29 18:26:20 abelloni: the recent massive cleanup of wic removed everything unused and untested that was inherited from the base (mic), and the inherited gpt support fell into that category. Support for gpt should be (re-)added, please file a bug so it gets done (or I can do that if you want). Jan 29 18:27:19 abelloni: same thing with the < 1Mb partition sizes - we'll need to add the ability to specify something other than Mb for --size Jan 29 18:28:25 kergoth: ahh.. that's why when i tried meta-sourcery master branch with poky-dizzy-12.0.1 it gave me weird "variable references itself!" type errors Jan 29 18:28:38 layer branches always need to be matched Jan 29 18:28:43 meta-sourery is no exception to that Jan 29 18:28:56 though its contents are probably closer to being independent of core version than most Jan 29 18:30:10 kergoth: is 2014.12-async1 is a good choice to use with poky-dizzy-12.0.1? I couldn't find a tagged release for dizzy in meta-sourcery Jan 29 18:30:47 that's probably closest, yeah, i dont' think we ever started ad izzy branch, but the 2014.12 mel release was dizzy based Jan 29 18:31:04 iirc anyway Jan 29 18:32:38 kergoth: i see.. thanks for all the help. I think I have enough pointers to move forward Jan 29 18:32:52 np Jan 29 18:58:55 tomz: ok, how would you go to switch from MB to kB? Jan 29 18:59:14 switch every user at once or provide some kind of compatibility ? Jan 29 19:17:29 abelloni: I guess it would just be --size=200k while the default would be in Mb if no units were specified, and Mb would also work i.e. --size=100 <=> --size=100Mb? Jan 29 19:18:12 abelloni: technically it would be KB and MB rather than lowercase (bits) Jan 29 19:26:12 tomz: I meant internally Jan 29 19:26:35 currently, the size is in MB (despite being stated that it is in Mb) Jan 29 19:26:54 would you switch all the prepare_rootfs_* at the same time ? Jan 29 19:27:16 or still let them provide the size in MB ? Jan 29 19:28:32 I'm talking about that: Jan 29 19:28:33 http://git.openembedded.org/openembedded-core/tree/scripts/lib/wic/kickstart/custom_commands/partition.py#n250 Jan 29 19:43:32 abelloni: internally, it looks like it would make sense to redefine size in terms of 1k blocks. So that particular du command would have to remove the 'm' and use just -Lbs. Similar internal conversions throughout. Jan 29 19:58:38 Any grub experts around? Jan 29 19:59:23 I am trying to boot an image created with mkefidisk.sh on a fairly standard PC-like hardware (VTC1010). Jan 29 20:00:24 I do get the "Welcome to GRUB!" message on the monitor after starting bootia32.efi in the EFI shell, but then it's stuck. Jan 29 20:01:00 I already removed the "serial --unit=0 --speed=115200 --word=8 --parity=no --stop=1" from grub.cfg because I don't have a serial console attached. Jan 29 20:01:15 Do I understand that line right in that it would redirect grub to the serial console? Jan 29 20:14:52 tomz: ok, I'll try to propose a patch Jan 29 20:21:10 abelloni: ok, sounds great, looking forward to seeing it Jan 29 20:25:52 Regarding grub: do I only need bootia32.efi and grub.cfg, or also something else? Other guides describe using grub-install (which I haven't tried yet) to get grub installed, whereas mkefidisk,sh just copies what's already in the .hdimg. Jan 29 20:33:17 Okay, I found grub-mkimage in grub-efi_2.00.bb. That combines everything into one file, like bootia32.efi. Next question: which modules do I need to get standard console support on PC hardware? The default GRUB_BUILDIN is "boot linux ext2 fat serial part_msdos part_gpt normal efi_gop iso9660 search". Jan 29 20:34:50 Looks like something like "all_video" or "efi_gop efi_uga ..." is missing. Jan 29 20:35:11 No, wait, efi_gop is there. Jan 29 20:35:33 hello all, has anyone done any work on the logicpd, specifically the torpedo dm3730? I am trying to get yocto working on it Jan 29 22:24:41 Hello. I am trying to add a recipe to an image but the recipe does not result in any binaries (it is tclap, which is a header library). All tutorials tell me to use IMAGE_INSTALL_append, but then I get an error: tclap not found in the base feeds... I think because there is no tclap binary. How do I add this recipe the right way? Jan 29 22:35:58 I have two sets of kernel modules. module in one set exports symbols needed by the other. how can I share the Module.symvers with the dependant set? Jan 29 22:55:16 darkhorse: you don't need to do anything, execution of depmod is automatic whenever necessary Jan 29 22:57:52 kergoth: well that's not happening in my case. I have a recipe that inherit module.bbclass to build a bunch of out-of-tree kernel modules. some of those export symbols which are to be used by another completely independent set of modules being build by another recipe Jan 29 22:58:14 thats just fine, depmod willb e run when the package is installed Jan 29 22:59:43 kergoth: hmm..and depmod will update the kernel Module.symvers? Jan 29 23:07:14 hello, does anyone have a good resource or reference for cross compiling software for arm? Jan 29 23:08:23 kergoth: BTW, I am not using yocto kernel recipe. I simple inherit kernel.bbclass and use module.bbclass for external modules Jan 29 23:45:14 kergoth: i see that you were referring to depmodwrapper earlier which should run after towards the end of an image recipe. is that correct? Jan 29 23:46:18 when the package is installed Jan 29 23:51:24 abelloni: is there any documentation about depmodwrapper? i couldn't find anything in the ref/dev/kernel manual **** ENDING LOGGING AT Fri Jan 30 03:00:00 2015