**** BEGIN LOGGING AT Thu Aug 22 02:59:58 2013 Aug 22 05:39:56 Hi yocto. I successfully compiled toolchain-qte and i installed it in opt directory.But when i try to run qmake from/opt/poky/1.4.1/sysroots/x86_64-pokysdk-linux/usr/bin/qmake2 , i got "sh: 1: -list: not found" Aug 22 05:40:32 My host is ubuntu_12.04 x64. Aug 22 05:41:27 I tried sample qt applications. That also results with same error. Aug 22 07:27:49 hi, i want to build the cross-compiler for fortran, so i took a look at the gcc-cross recipe and figured out that it requires gcc-configure-common.inc, there i set the FORTRAN variable to ",fortran", everything builds fine, (with bitbake -f -c compile gcc-cross & crosssdk) but when i try to compile a package which need gfortran as a compiler, the gfortran compiler is not found Aug 22 07:28:38 i took a look at http://www.yoctoproject.org/docs/1.4/dev-manual/dev-manual.html#cross-development-toolchain but somehow i don't really understand what i have to do to get the gfortran cross compiler working Aug 22 07:29:17 in the sysroots directory i found a cross-gfortran compiler, which is executable (tried with arm-*something-gfortran -v) Aug 22 07:42:29 thesignal: is the pacakged using autotools? Aug 22 07:43:29 tf: which one? the one which needs gfortran? Aug 22 07:44:15 yes Aug 22 07:44:42 no it isn't Aug 22 07:45:32 you might have to patch it, it's probably looking for gfortran, not arm-bla-bla-gfortran Aug 22 07:46:24 do you know a recipe where this has been done? so i can take an idea how to do this Aug 22 07:46:54 thesignal: no, sorry Aug 22 07:47:08 kk, but thanks for the hint Aug 22 07:47:24 thesignal: you should find more details in the error log Aug 22 07:49:08 tf: it just says "gfortran: command not found" but i'll take a look at the run.do_configure Aug 22 07:55:08 thesignal: probably the makefile is referring to gfortran, it should be using a variable; I don't know what that would be with fortran, but for C, you would call the compiler as $CC, not gcc Aug 22 07:59:57 tf: thanks, i already took a look at the makefile and it's exactly like you thought :) Aug 22 08:00:33 and the toolchain and everything needed should be compiled with bitbake -f -c compile gcc-cross, right? Aug 22 08:01:59 that will just compile it, not install it Aug 22 08:02:38 'bitbake gcc-cross' will build and install, and package it Aug 22 08:03:48 morning all Aug 22 08:28:15 Hi everybody,does anyone know how to modify the root file system before it's built in the ramdisk image? thks Aug 22 08:39:08 I want to add my own sysroot in meta-toolchain. Is it possible? Aug 22 09:08:37 vicky_: you mean an additional separate sysroot? or do you mean customise the target part of the SDK to add additional components? Aug 22 09:28:23 Hi all Aug 22 09:28:51 currently UBOOT_ARCH was exported from meta/classes/kernel-arch.bbclass Aug 22 09:28:55 export UBOOT_ARCH = "${@map_uboot_arch(d.getVar('ARCH', True), d)}" Aug 22 09:29:14 so it always get UBOOT_ARCH= $ARCH} Aug 22 09:29:28 in my case i want UBOOT_ARCH difference from ARCH Aug 22 09:29:52 then i comment out this export statement in kernel-arch.bbclass Aug 22 09:30:02 but i think it is not the right way Aug 22 09:30:31 could any one give me an idea? Aug 22 09:32:29 what are you trying to do? Try just setting UBOOT_MACHINE Aug 22 09:33:55 i want to generate u-boot image. and in image_types_uboot.bbclass using mkimage -A ${UBOOT_ARCH} -O linux -T ramdisk -C $2 -n ${IMAGE_NAME} Aug 22 09:34:55 but because UBOOT_ARCH was set by ARCH so it can not create as i expected Aug 22 09:44:36 see, you are using mkimage *native* on your host..what is wrong with that? Aug 22 09:45:04 you ar ebuilding for a target belonging to ARCH, isn't? Aug 22 09:46:02 ant_work: my work right now is ARM64 Aug 22 09:46:19 but for uboot, ARM64 also use the same arch with ARM Aug 22 09:46:36 so linux, uboot use difference $ARCH Aug 22 09:47:32 that's why UBOOT_ARCH=$ARCH make me to problem. other architecture will not have that problem Aug 22 09:50:34 ah, I see, arm64/aarch64 is new stuff Aug 22 09:51:24 as far as I see Khem is already working on it Aug 22 09:51:34 [Angstrom-devel] [meta-angstrom] angstrom-next: add versions to build for aarch64 Aug 22 09:52:08 besides u-boot support you need a proper ARCH set in OE/Yocto Aug 22 09:53:18 yes, right now i can build uboot, linux, core-image-minimal and all work. but i just want to ask the right way to set UBOOT_ARCH Aug 22 09:53:46 because i set in local.conf it will be overrwite by kernel-arch.bbclass Aug 22 11:31:26 hi i got this error http://pastebin.com/UsrgybKq while executing the bitbake at91bootstrapn on poky atmel layer,even i changed the appropriate changes in conf files.can you tell me what is that issue Aug 22 11:41:24 bluelightning: yes. I need the additional sysroot Aug 22 11:58:29 bitbake image is not updating kernel modules into the image Aug 22 12:05:50 anyone can tell me why the tunctl recipe has an include file ? Aug 22 12:06:18 for convenience, here is the link: https://github.com/joeythesaint/meta-networking/tree/master/recipes-support/tunctl Aug 22 12:07:06 probably for when there were multiple versions Aug 22 12:07:26 there were? Aug 22 12:11:53 will configure run automatically before make without a dedicated section if the software requires that? Aug 22 12:17:58 vicky_: I don't think we currently support that Aug 22 12:18:25 do_configure () { oe_runconf Aug 22 12:18:25 } Aug 22 12:18:32 do I need to put that into the file explicitly? Aug 22 12:18:38 to get ./configure run before make? Aug 22 12:18:56 lpapp: if you inherit autotools, no, but note that running oe_runconf is not the same as what autotools.bbclass does by default Aug 22 12:19:10 lpapp: oe_runconf will just run configure, skipping out the normal autoreconf step Aug 22 12:19:24 what i did was bitbaked meta-toolchain-qte but it doesnt included my custom packages at layer meta-atmel Aug 22 12:19:49 vicky_: right, it won't; you may wish to use bitbake -c populate_sdk instead Aug 22 12:20:32 vicky_: but then you'd also need to add the tools back into the host side of the SDK; for that you can append the names of the appropriate packages to TOOLCHAIN_HOST_TASK Aug 22 12:20:32 bluelightning: I am not using autotools. Aug 22 12:20:41 lpapp: right then oe_runconf won't help you Aug 22 12:20:49 bluelightning: hmpf Aug 22 12:20:52 so what will ? Aug 22 12:21:02 lpapp: you'll need to define do_configure and in it run the configure script yourself Aug 22 12:21:07 I mean ./configure is autotools. Aug 22 12:21:31 (or, do whatever is needed for the particular piece of software to configure it before building, if it doesn't use a configure script) Aug 22 12:23:22 bluelightning: it uses a configure script. Aug 22 12:23:30 but in general, I saw "configure scripts" without autotools. Aug 22 12:23:35 anyway, so what then? Aug 22 12:23:40 so, as above Aug 22 12:23:44 how to run ./configure? Aug 22 12:23:56 without autotools stuff like m4, ac etc? Aug 22 12:24:25 oe_runconf should run ./configure, no? Aug 22 12:24:29 don't inherit autotools, and define do_configure that calls the configure script with whatever arguments are appropriate Aug 22 12:25:01 oe_runconf is designed for autotools and is in fact defined in autotools.bbclass so you will not be able to use it for non-autotools configure scripts Aug 22 12:25:19 bluelightning: why ? Aug 22 12:25:41 actually, I had taken a look into it, and it does not do much else than running the configure script. Aug 22 12:25:53 so it looks feasible to me for running the configure script in the source folder ... Aug 22 12:26:14 because there's no commonality in other likely hand-coded configure scripts that we can support in some hypothetical non-autotools oe_runconf function Aug 22 12:26:27 ? Aug 22 12:26:35 just run the configure script with the appropriate arguments from your own do_configure Aug 22 12:26:40 I only need to run ./configure, and based on the autotools class, it just does that. Aug 22 12:26:49 with the convenience of printing the config log in case of failure. Aug 22 12:27:04 that's autotools-specific behaviour Aug 22 12:27:34 why would it Aug 22 12:27:43 although it does not support arguments for configure. Aug 22 12:27:50 not that I need those, but still. Aug 22 12:29:56 bluelightning: what is autotools specific ? Aug 22 12:30:15 does every non-autotools configure script output a config.log? no Aug 22 12:30:24 ? Aug 22 12:30:33 that is what people do when there is an error, yes. Aug 22 12:30:38 regardless autotools. Aug 22 12:30:40 no Aug 22 12:30:44 sure, yes. Aug 22 12:31:43 | ERROR: Function failed: do_configure (see ... Aug 22 12:32:06 oe_runconf: not found Aug 22 12:32:09 why not? Aug 22 12:32:20 because as I already said above it's defined in autotools.bbclass Aug 22 12:32:28 if you're not inheriting autotools you won't have access to it Aug 22 12:35:03 what is the point of inherit? Aug 22 12:35:54 why do we even need to inherit rather than just being available by default? Aug 22 12:36:27 anyone know of a yocto-poky specific uEnv.txt file for beaglebone black? It doesn't get created by the bitbake for beaglebone. Aug 22 12:36:46 lpapp: as above, if it's not autotools we can't expect any kind of commonality that would merit such a common function Aug 22 12:36:56 needed for u-Boot to work properly. Aug 22 12:37:01 lpapp: why is it a problem to just run ./configure? Aug 22 12:40:29 bluelightning: I am not sure if it is autotools or not, to be honest. Aug 22 12:40:40 I was merely referring to that, it does not use any other autotools stuff. Aug 22 12:40:46 m4, ac, libtools, etc Aug 22 12:40:55 other than that, I am not sure what makes a configure script "autotools". Aug 22 12:41:19 I have not used autotools (luckily) the last 5-6 years. Aug 22 12:41:56 oh, it has configure.ac and Makefile.am Aug 22 12:42:00 maybe, it is autotools then. Aug 22 12:42:09 and aclocal.m4, meh. Aug 22 12:42:15 anyway, it does not build. Aug 22 12:42:54 bluelightning: http://paste.kde.org/~lpapp/p901001d9/ Aug 22 12:45:16 right, you should be inheriting autotools then Aug 22 12:45:54 can't tell what the error is there Aug 22 12:47:02 me neither ... Aug 22 12:47:18 can I simply pass arguments to oe_runconf ? Aug 22 12:47:59 just add whatever arguments you need to EXTRA_OECONF Aug 22 12:48:48 hmpf Aug 22 12:48:52 k. Aug 22 12:50:54 fwiw, it works if I build on the host. Aug 22 12:51:01 sure it is not a recipe issue? Aug 22 12:51:32 here is the recipe: http://paste.kde.org/~lpapp/p80838cd0/ Aug 22 12:53:49 lpapp: well you should definitely start without your do_configure there Aug 22 12:56:03 ? Aug 22 12:57:25 drop you definition of do_configure, autotools.bbclass provides it already Aug 22 12:57:35 sure, but that should not cause such issues. Aug 22 12:57:39 and yes, it is the same without. Aug 22 12:58:05 sure, I'm just making a recommendation Aug 22 12:58:34 lpapp: ok, any recommendation for the error? ;-) Aug 22 12:59:31 it's a compilation failure with stunnel itself, you'll have to debug it like any other compilation failure Aug 22 13:00:03 I'd start by making sure there are no clues in the do_compile (such as errors not shown in the snippet you pasted) or anything unusual in the do_configure log Aug 22 13:00:16 s/do_compile/do_compile log/ Aug 22 13:01:06 I dropped an email to the stunnel-users mailing list in the meantime ... Aug 22 13:02:56 You could try to go those where the source is downloaded and manually try to compile it Aug 22 13:03:09 and see what errors it produces Aug 22 13:03:19 true Aug 22 13:04:04 when you know correction then you can make a patch and include that to SRC_URI Aug 22 13:04:30 bluelightning: Guest26644 this is suspicious at the end of configure, http://paste.kde.org/~lpapp/pe0a7d211/ Aug 22 13:04:33 at least I have done a a couple of times Aug 22 13:04:58 mayhaps I am missing something? Aug 22 13:05:47 that is odd... never seen that before Aug 22 13:06:19 kinda me neither ... Aug 22 13:06:30 /bin/bash: ../-libtool: No such file or directory Aug 22 13:06:49 I think running "make" manually will not try to use the cross autotools? Aug 22 13:06:54 anyone know the status of icecc support? saw an old mailthread for aug/sept 2012 where someone had issues but nothing after that Aug 22 13:07:07 so it is not simply going into the work folder, and "./configure && make", right ? Aug 22 13:07:23 you need to do the settings first Aug 22 13:08:21 I'm not sure about running the configure steps.. I was just talking about compiling the code Aug 22 13:11:50 but anyway in order to get my code to compile I needed to modify the .am file to include one setting and I done that with patch Aug 22 13:12:12 Guest26644: I do not know how to do any settings in there... Aug 22 13:12:15 and what exactly, etc. Aug 22 13:12:53 Guest26644: archlinux seems to have a makefile patch, but it is about install, so I believe it is not related, https://projects.archlinux.org/svntogit/community.git/tree/trunk/Makefile.patch?h=packages/stunnel Aug 22 13:13:42 You can do a patch for any file. That was just example Aug 22 13:14:09 of course... no doubt in there. Aug 22 13:14:10 if you can somehow find the what the error is .. Aug 22 13:14:14 I am just trying to solve this stuff. :) Aug 22 13:14:30 then you can try to google that and try patch to fix it Aug 22 13:15:49 well, packager life is not easy I guess. Aug 22 13:16:02 lpapp: Gentoo has similar patch http://xrl.us/bppi3g Aug 22 13:16:29 you could add it in do_configure_prepend Aug 22 13:16:57 ant_work: what patch? Aug 22 13:29:59 ant_work: related to this issue I am facing, or you mean the Makefile am stuff? Aug 22 13:43:21 my yocto kernel keeps getting stuck at "Starting kernel ..." on beaglebone black using core-image-minimal. Anyone run into this? http://pastebin.com/yv98C20a has boot log from BBB. Aug 22 13:43:55 lpapp: obvioulsy I was referring to the arch linux patch you showed Aug 22 13:44:18 ant_work: ok, that would not help in here, I believe. Aug 22 13:44:31 nor did EXTRA_OECONF += "--disable-fips" Aug 22 13:44:32 lpapp: have you tried removing your do_configure ? Aug 22 13:44:43 ant_work: yes Aug 22 13:44:45 ah, ok, this was th enext step: customize it Aug 22 13:45:09 ? Aug 22 13:45:14 customize what exactly? Aug 22 13:45:32 cfo215: could you verify that your console output is on ttyS0? Aug 22 13:46:04 lpapp: one way is as you did, with Aug 22 13:46:04 EXTRA_OECONF Aug 22 13:46:20 ant_work: no, I mean what flags, etc. Aug 22 13:46:28 anyway, I am waiting for the authors to reply, I believe. Aug 22 13:46:30 I have no clue in here. Aug 22 13:46:50 I wonder if anyone else could test my recipe on desktop. Aug 22 13:46:58 without external toolchain, arm board etc. Aug 22 13:47:19 simar_: sorry I'm a newb... how do I go about that? I got the output for the pastebin from gtkterm connected to the console port on the black. Aug 22 13:47:26 this is the latest variant, http://paste.kde.org/~lpapp/pfb1888d8/ Aug 22 13:47:40 file name is stunnel_4.56.bb Aug 22 13:47:51 it only depends on openssl. Aug 22 13:48:06 simar_: http://pastebin.com/yv98C20a Aug 22 13:50:05 simar_: i'm pretty sure I'm not understanding the question? I followed the instructions at https://www.yoctoproject.org/download/texas-instruments-arm-cortex-a8-development-board-beagleboard-2 Aug 22 13:51:45 it seems your console line isn't correct. Aug 22 13:51:46 I don't have SERIAL_CONSOLE_beagleboard = "115200 ttyS2" in my local.conf. Should I add "SERIAL_CONSOLE_beaglebone = "115200 ttyS0"? to local.conf and rebuild the image. Aug 22 13:52:22 hi, I'm trying to compile a package (octave) but the compilation fails, in the log i found "error: ::gmtime has not been declared", i think i have a mistake with my sysroots configuration, any ideas? Aug 22 13:52:27 cfo215: I see it as bootargs=console=ttyO0,115200n8 in the pastebin that you sent me. Aug 22 13:52:58 simar_: ic. so I need to change my uEnv.txt file. Aug 22 13:54:32 thesignal: you probably miss in DEPENDS some package it depends on Aug 22 13:54:38 the error refers to /usr/include/c++/ctime Aug 22 13:55:12 tf: how do i find out which packages that are? Aug 22 13:56:50 thesignal: libc++ or whatever it's called? Aug 22 13:57:48 hm, good point, i'll try some Aug 22 13:58:25 simar_:Should I add "SERIAL_CONSOLE_beaglebone = "115200 ttyS0"? to local.conf and rebuild the image. Aug 22 14:00:39 cfo215: console=ttyO2,115200n8 Aug 22 14:03:53 simar_ do I use beagleboard or beaglebone in the local.conf? my target is beaglebone. Aug 22 14:04:13 cfo215: you can also check in meta-yocto-bsp/conf/machine/beagleboard.conf to see what console line you have. Aug 22 14:04:44 simar_: thanks. Aug 22 14:05:12 cfo215: beaglebone surely - but then of course you'll need a layer providing the beaglebone machine (e.g. meta-beagleboard - confusing naming perhaps) Aug 22 14:13:50 Hi there, how do I extend uclibc to add libstdc++ ? Aug 22 14:18:06 blueligtning: so I need to download meta-beagleboard; change it to dylan; update my local.conf to include 'SERIAL_CONSOLE_beaglebone = "115200 ttyO2"' and re-run bitbake and recreate the SD card. Aug 22 14:19:03 is there a better irc chat program than empathy on ubuntu? I really don't like it... Aug 22 14:20:09 cfo215: xchat is the usual Aug 22 14:20:09 empathy isn't great for IRC Aug 22 14:20:58 rbuton: thanks, installing it now. Aug 22 14:22:04 irssi best of all :) Aug 22 14:22:41 cfo215: you can use completion with tabs. it is "rburton". :) Aug 22 14:27:49 bluelightning: I wonder if the updated libX11 will help marko's issue, I am sending email. Aug 22 14:28:06 sgw_: not sure... Aug 22 14:28:54 rburton: thanks for your help. did you happen to get my last message before my empathy crashed? Aug 22 14:29:32 cfo215_: last i saw was thanks, installing it now Aug 22 14:31:18 rburton: I need to git clone meta-beagleboard; change it's branch to dylan; add it to bblayers.conf; update local.conf with ttyO2 stuff and rerun bitbake for my image. Aug 22 14:31:29 cfo215_: that sounds reasonable Aug 22 14:33:05 bluelightning, rburton : I'll try those.. and recreate the SD card of course. Aug 22 14:37:30 bluelightning: http://www.stunnel.org/pipermail/stunnel-users/2013-August/004310.html Aug 22 14:37:39 bluelightning: how can I get the proper include path? Aug 22 14:38:13 lpapp: figure out how that path is getting in there and fix it Aug 22 14:38:26 lpapp: it'll be specific to this particular piece of software Aug 22 14:38:56 bluelightning: really ? :O Aug 22 14:39:29 bluelightning: what is the "proper way" for upstream? Aug 22 14:40:29 bluelightning: first of all, what should be the include path? Aug 22 14:41:21 lpapp: ${STAGING_INCDIR} I would think Aug 22 14:41:39 bluelightning: that variable is empty if I print it on the command line... Aug 22 14:41:57 which might be expected, so is it not a relatively clear path which should be the same from the projectroot? Aug 22 14:43:28 bluelightning, rburton: to be clear, do I then remove meta-ti from bblayers.conf Aug 22 14:45:24 cfo215_: does meta-beagleboard require meta-ti? Aug 22 14:45:36 if not, then remove it Aug 22 14:46:04 tf, not according to the README.md Aug 22 14:50:54 ERROR: Unable to parse /media/toshiba-usb3/work/poky/meta-beagleboard/conf/layer.conf: [Errno 2] No such file or directory: '/media/toshiba-usb3/work/poky/meta-beagleboard/conf/layer.conf' Aug 22 14:51:22 anyone building android systems with yocto here? Aug 22 14:51:54 do I need to copy one from the common-bsp/conf/ or meta-beagleboard-extras/conf? Aug 22 14:51:56 bozojoe: systems? Aug 22 14:51:58 bozojoe: haven't heard of anyone doing it, but it should be perfectly doable if you fancy the challenge Aug 22 14:53:03 "android systems" i mean using yocto for commercial android phone or tablet kernels/rootfs/etc Aug 22 14:54:05 cfo215_: no, each layer needs to provide a layer.conf, it should be there Aug 22 14:55:06 bozojoe: I do not think oe-core, or even oe would have the packages to do so. Aug 22 14:55:54 lpapp: sure, any android support would be a separate layer Aug 22 14:56:05 its been mulled over numerous times by various people, but nothing (afaik) has actually happened Aug 22 14:56:08 including bionic Aug 22 14:56:22 yep Aug 22 14:56:24 so it would be quite a bit of investment. Aug 22 14:56:29 yep Aug 22 14:56:55 It doesn't. I used git clone git://github.com/beagleboard/meta-beagleboard.git Aug 22 14:57:32 so where are the cross-compiler include files with Yocto ? Aug 22 14:57:46 I'm very frustated now. It's supposed to be "The official OpenEmbedded/Yocto BSP layer for Beagleboard.org platforms." Aug 22 14:57:51 what is STAGING_INCDIR falling to, compared to the TOPDIR for instance, or well, even build dir. Aug 22 14:58:04 thanks all... i learned of yocto yesterday... i need to ask around our org to see if anyone has seriously contemplated this. Aug 22 14:58:30 lpapp: STAGING_INCDIR is a variable in .bb scope Aug 22 14:59:06 bozojoe: FWIW I like the idea but have never had the time to look at it. The system itself should be able to support it quite well Aug 22 14:59:09 rburton: can you try to verify the cross compilation issue with stunnel? Aug 22 14:59:31 lpapp: not at the moment, sorry. Aug 22 14:59:39 bozojoe: if you are not google, I would not probably bother... Aug 22 14:59:51 it is not using GNU Linux. Aug 22 15:00:01 no busybox, etc. Aug 22 15:00:36 lpapp: lol... i know... if you != google then all_bets = off Aug 22 15:00:37 well, google or some google affilated partner. Aug 22 15:01:07 cfo215_: the layer is in common-bsp folder, you need to add it in bblayer.congf Aug 22 15:02:47 ndec, change /poky/meta-beagleboard to /poky/meta-beagleboard/common-bsp in bblayers? Aug 22 15:02:54 yes. Aug 22 15:03:02 ndec, I'll try that. Aug 22 15:04:08 bozojoe: I would not like to discourage you ... go ahead by all means if you like, but well, it is not a simple job. Aug 22 15:04:29 ERROR: ParseError at /media/toshiba-usb3/work/poky/meta-beagleboard/common-bsp/recipes-kernel/linux/linux.inc:7: Could not inherit file classes/machine_kernel_pr.bbclass Aug 22 15:04:30 ERROR: Command execution failed: Exited with 1 Aug 22 15:04:33 bozojoe: even getting a lot simpler stuff is painful with Yocto at times, like external cross compiler. Aug 22 15:04:54 here, you would replace the gnu userspace. Aug 22 15:08:39 lpapp: luckily, my dev environment and target are both x86... hopefully that will make things easier Aug 22 15:09:01 I'm ready to give up on Yocto/OE.. it shouldn't be this hard to set up. I've wasted so much time chasing down "undocumented" features. All you people have been very helpful but it's been a few weeks now and still don't have a bootable machine. Aug 22 15:09:21 bozojoe: you cannot avoid the non-gnu userspace... or you cannot call it android otherwise. :) Aug 22 15:09:25 maybe gnudroid. Aug 22 15:09:34 cfo215_: iirc machine_kernel_pr is in meta-oe. i'm surprised, if it needs that, the readme should list it Aug 22 15:11:13 my mistake... I have the folder. don't have it in my bblayers Aug 22 15:14:50 cfo215_: you might have better luck in the beagleboard support channels as they'll know how their layer is meant to work Aug 22 15:14:51 i added .../meta-oe/meta-oe to my bblayers.conf and it's now parsing recipes.. Aug 22 15:16:10 seems one day beagle* will use linux-yocto Aug 22 15:17:01 ant_work, thats why I'm here... trying to use poky to build an image for my BBB. Aug 22 15:17:19 ERROR: No recipes available for: Aug 22 15:17:19 /media/toshiba-usb3/work/poky/meta-oe/meta-oe/recipes-core/busybox/busybox_1.21.1.bbappend Aug 22 15:17:19 ERROR: Command execution failed: Exited with 1 Aug 22 15:18:00 at least with meta-ti I was able to "build' an image. it just wouldn't boot on the BBB. Aug 22 15:18:03 layer branch mismatch? Aug 22 15:18:13 * kergoth shrugs Aug 22 15:18:28 the point is, main kernel dev is done outside Yocto but talking with one lead devel I can say there is interest Aug 22 15:18:51 I'm not going to help or comment - it's unappreciated job here, apparently... Aug 22 15:19:36 it's a bit uncomfortable to have a beagleboard machine listed in compatibility and then needing extra-layers... Aug 22 15:20:05 ant_work: beagleboard is supported in yocto directly Aug 22 15:20:43 ant_work: don't confuse the matter with beaglebone Aug 22 15:20:47 yes, in linux-yocto Aug 22 15:21:48 denix, how about bone? is it or isn't it doable out of the box" .i.e. fresh download of poky? Aug 22 15:22:11 ant_work: and if needing extra-layers is a problem, then you shouldn't use yocto at all Aug 22 15:23:37 poky? don't know. I'm sure it's not. it's a _reference_ distro - don't expect it to support everything and a kitchen sink Aug 22 15:23:52 denix: I was just talking with Koen about how you can quickly put together kernel and share patches without git branches and stuff for a pool of machines (SOC) Aug 22 15:23:54 denix, ant_work: if the bottom line is that I can't use Yocto to build an image, extra-layers or not, then I need to move on and quit wasting my time trying. Aug 22 15:24:19 * jkridner would still like to see Yocto switch from BeagleBoard to BeagleBone Black to make the entry cost lower, especially since the stock Yocto multimedia support isn't very strong. Aug 22 15:24:38 cfo215_: it's pretty simple; for a machine, you need a bsp that supports that machine; for bone that's meta-beagleboard Aug 22 15:24:53 all can be in a linux-yocto_x.y.bbappend Aug 22 15:25:07 cfo215_: you can definitely use Yocto with extra layers to get BBB up and running - either meta-ti or meta-beagle is needed Aug 22 15:25:14 and a proper dir structure Aug 22 15:25:15 What are the cmake parameters that needs to be included in my recipe? Aug 22 15:25:26 can anyone try to build my recipe? Aug 22 15:25:30 ant_work: why centralized place? Aug 22 15:25:35 lpapp: sure, later home Aug 22 15:25:49 ant_work: I will not be available later. Aug 22 15:25:57 tf, denix: i tried with meta-ti and got an image built but it wouldn't boot off the SD card. Aug 22 15:26:14 got stuck at "Starting kernel ..." Aug 22 15:26:18 denix: you would still need a BSP layer Aug 22 15:26:40 cfo215_: there is a mailing list for meta-ti. as well as the one for meta-beagle - have you tried asking there? Aug 22 15:27:20 denix, no but I will. Aug 22 15:27:50 cfo215_: there are many variables - have you preped your SD properly? MLO, u-boot, uImage, DTB? Aug 22 15:29:07 jkridner: I think zeddii is planning on refreshing the reference hw soon, but it's almost certainly too late for 1.5 Aug 22 15:29:15 I followed (partially) the instructions for beagleboard on Yocto-Progject website. Aug 22 15:29:30 cfo215_: beableboard is not beaglebone Aug 22 15:30:00 tf, yeah, got that. I have one of each. Aug 22 15:30:15 cfo215_: here, from our point of view it is the same..kernel + patches Aug 22 15:31:29 ant_work: and from another point of view it's just a PCB with some components on it... Aug 22 15:32:57 denix: see how different machines can convive: http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-yocto-3.8 Aug 22 15:33:34 denix, do you have a working SD card for a BBB that uses yocto-poky-linux? Aug 22 15:34:04 ant_work: I'm not arguing it's possible. I'm saying there are many reasons to not do it that way (mostly political, that you don't have control over...) Aug 22 15:34:21 denix, i'm rebuilding my image using meta-ti. I'll work on finding an answer to my SD card issue. Aug 22 15:36:45 cfo215_: I don't currently have an image from poky - I don't use that distro. again, yocto != poky Aug 22 15:44:41 denix, thanks. I'd be happy with any distro that works on BBB with Qt4 or Qt5 on it. Console only. No X11 stuff here. Aug 22 15:45:15 i was hoping to save time with Yocto for building images. Aug 22 15:48:35 bluelightning: does that also mean moving the x86 reference to MinnowBoard? Aug 22 15:49:47 jkridner: I'm not sure... I'm not really involved in that side of things, I just know Bruce has mentioned refreshing the reference hardware Aug 22 15:54:58 cfo215_: and it should work. for example, our Arago distro works on all supported TI platforms from meta-ti with Qt4, no X11 Aug 22 15:55:29 cfo215_: although, meta-beagle might have better cape support, if you require that Aug 22 16:03:10 cfo215_: take advantage of koen help on #beagle. Keep the logs ;) Aug 22 16:05:04 ...and ignore the funny people :D Aug 22 16:10:05 ant_work: ? Aug 22 16:12:41 denix: he's getting red carpet...plenty of help...a #beagle chat could have been much worse... ya know Aug 22 16:13:21 I get my daily dose of humor there Aug 22 16:13:48 ant_work: ah, you mean regular trolls there... :) Aug 22 16:14:23 yes, #beagle could be very intimidating for newbies! :) Aug 22 16:14:25 professionals Aug 22 16:19:47 ant_work: I will try to be available later ... feel free to ping me. Aug 22 16:19:53 if I am up... Aug 22 16:20:04 sorry I don't have toolchain here Aug 22 16:20:09 just Gentoo x86 Aug 22 16:20:30 no, at home Aug 22 16:20:37 sure Aug 22 16:21:06 but I might not be able to provide the recipe, so better if you can back it up Aug 22 16:21:08 seems (as always) straigthforward at first sight ;) Aug 22 16:21:29 what does? Aug 22 16:21:50 I'd have expected it to compile flawlessy Aug 22 16:22:07 well, it depends on upstream, I guess. Aug 22 16:22:08 I'm familiar with stubborn recipes anyway Aug 22 16:22:21 http://paste.kde.org/~lpapp/pccd6cbd6/ -> here is the recipe again. Aug 22 16:22:48 the first step would be to figure out where Yocto looks for the cross-compilation includes. Aug 22 16:22:51 yes, is the same version as Gentoo has Aug 22 16:22:55 which path compared to the build dir, etc. Aug 22 16:23:52 lpapp: no, you get that wrong way round, you need to work out where the upstream package looks for the includes Aug 22 16:24:14 no no Aug 22 16:24:25 that one clear already, see the build output. Aug 22 16:24:35 and it is apparently wrong as it is /usr/include Aug 22 16:24:44 which probably should be somewhere inside the build directory. Aug 22 16:24:50 include failing is almost always down to two things: a missing dependency from DEPENDS or hardcoded path in the upstream project Aug 22 16:24:53 but I would like to know where exactly. Aug 22 16:26:02 lpapp: have fun :) Aug 22 16:26:04 lpapp: my workflow is usually to patch the Makefile an then if possible to transfer the changes in the recipe via EXTRA_OE.. or custom tasks Aug 22 16:26:25 does anybody know how to add libstdc++ to uclibc based image? Aug 22 16:27:18 Krz: there is uclibc++ iirc Aug 22 16:27:30 ant_work: I am interested in any workaround. :) Aug 22 16:27:51 tf: what's the channel name? #uclibc ? Aug 22 16:27:55 but it works for arch-arm fwiw which is also using arm 32. Aug 22 16:28:10 Krz: iirc != irc. :) Aug 22 16:28:20 Krz: ping Khem in some hour. He is UT-9 iirc Aug 22 16:28:42 ok, what is iirc then? :) Aug 22 16:28:56 Krz: check urban dictionary. Aug 22 16:29:06 Krz: if i recall correctly Aug 22 16:29:12 Krz: http://www.urbandictionary.com/define.php?term=iirc Aug 22 16:29:36 ant_work, "koen to me: cfo215_: what the fuck are you touching local.conf and bblayers.conf for?" Aug 22 16:29:56 heh Aug 22 16:30:04 ant_work, very helpful. Aug 22 16:30:09 yikes, he's even worse than I used to be back in the day Aug 22 16:30:19 cfo215_: don't put up with that kind of behaviour... I wouldn't Aug 22 16:30:28 if you don't know exactly what the implications are, just follow the steps listed in Angstrom instructions Aug 22 16:30:35 they use a strange wrapper Aug 22 16:30:40 ah, koen. Aug 22 16:30:56 but surely, once you discover oe-core / Yocto under the skin you can customize it Aug 22 16:31:23 ant_work: https://www.stunnel.org/pipermail/stunnel-users/2010-August/002722.html -> looks like configure needs to get the --host=arm-linux stuff? Aug 22 16:31:26 ant_work, I've done that. I get an image. But it's not exactly what I need. Aug 22 16:32:06 --host=arm-linux-gnueabi to be precise. Aug 22 16:32:09 cfo215_: do you have a list of packages youwant to install ? Aug 22 16:32:37 define needed image Aug 22 16:32:58 it's there that you have toplay, not in the .conf files Aug 22 16:33:03 bluelightning, "it is better to stay quiet and be judged a fool, than to open your mouth and eliminate all doubt." I let that stuff slide off. I'm too old to care. Aug 22 16:33:07 that's what Koen was politely saying Aug 22 16:33:19 lpapp: if you have a look at your log.do_configure you'll see that will already be being sent to the configure script by default Aug 22 16:33:23 * lpapp is trying EXTRA_OECONF += "--host=arm-linux-gnueabi --disable-fips" Aug 22 16:33:41 bluelightning, it's certainly not addressing my questions. Just useless jabber. Aug 22 16:33:48 cfo215_: I mean, there are polite ways of pointing out when someone is doing something wrong, and that is not politeness Aug 22 16:33:50 lpapp: the autotools class already does that Aug 22 16:33:51 bluelightning: by? Aug 22 16:33:58 cfo215_: language like that is not acceptable in my book Aug 22 16:33:59 tf: that is bad Aug 22 16:34:04 lpapp: just grep the recipes for EXTRA_ and see the examples Aug 22 16:34:05 because it should not be arm-linux Aug 22 16:34:05 cfo215_: at least, not on public IRC Aug 22 16:34:09 it should be arm-linux-gnueabi Aug 22 16:34:14 bluelightning, you're correct there. Aug 22 16:34:19 lpapp: it does whatever is right Aug 22 16:34:19 so if it does the wrong thing by default, it is not wonder it does not owrk. Aug 22 16:34:20 work* Aug 22 16:34:28 lpapp: check what arguments actually are being supplied Aug 22 16:34:38 bluelightning, so he "opened his mouth"... ;) Aug 22 16:36:33 bluelightning: it is passing, yea. Aug 22 16:36:54 cfo215_: anyway, the Angstrom script does create local.conf for you and stacks up and fetches many layers Aug 22 16:37:08 cfo215_: you can obviously do that by yourself Aug 22 16:37:40 cfo215_: subtle difference is you have to pass MACHINE to the script, because none is defined in local.conf Aug 22 16:37:41 Krz: I don't think there is uclibc++ recipe in current Yocto, but there is one in oe-classic, http://git.openembedded.org/openembedded/tree/recipes/uclibc++ Aug 22 16:37:42 I'm not there to see it but I hope he's not saying meta-beagle must be used with Angstrom and you're doing it wrong if you're not using that distro... Aug 22 16:37:45 (bad imho) Aug 22 16:38:39 ant_work: cfo215_ so finally I don't see anything wrong in the local.conf you posted, just the lack of a MACHINE Aug 22 16:38:48 Krz: obvkously, you would need to fix that up for Yocto, but it's a start Aug 22 16:38:49 tf: I didn't find it either Aug 22 16:39:06 tf: can't I just c++ support to uclibc? Aug 22 16:39:25 well now, it's a different library Aug 22 16:39:26 Krz: I'm sure khem had a comment about uclibc++ a while back but I don't recall exactly what he said - khem? Aug 22 16:39:27 cfo215_: obviously the IMAGE_INSTALL_append are unforgivable ;) Aug 22 16:39:59 bluelightning, I'm just trying to get any image on my BBB. All in need is to support the lcd/touchscreen (LCD CAPE7), boot into my Qt 4 app, and allow my Qt 4 app to access the GPIO and Analog imputs, also, support for wlan. Aug 22 16:40:34 ant_work, must be... though it works :) Aug 22 16:40:44 bluelightning: that cape stuff and it's bunch of patches are the matter Aug 22 16:41:05 ant_work: that should be provided by meta-beagleboard surely Aug 22 16:41:12 I have had Poky images working on the Bone White in the past, but always using meta-ti for kernels Aug 22 16:41:13 that change solves a similar issue that I have ... http://patches.openembedded.org/patch/44919/ Aug 22 16:41:18 cfo215_: sounds like a reasonable expectation Aug 22 16:41:32 bluelightning: yes Aug 22 16:41:51 I recently built an image for my beaglebone (white) as well; unfortunately X doesn't start, I keep meaning to report that as a bug Aug 22 16:42:16 https://github.com/openembedded/meta-oe/blob/master/meta-oe/recipes-support/onig/files/do-not-use-system-headers.patch Aug 22 16:42:19 here is the change. Aug 22 16:42:34 bluelightning, I don't need X11, just shell into "my-qt-application -qws". Aug 22 16:42:54 cfo215_: ok, was just mentioning a data point Aug 22 16:43:07 lpapp: nice catch Aug 22 16:43:12 bluelightning, np Aug 22 16:43:14 ant_work: ? Aug 22 16:43:52 https://github.com/kergoth/vim-bitbake/pull/5 Aug 22 16:43:53 it looks like same thing happens Aug 22 16:43:59 cfo215_: the image did boot at least and the LCD part of the cape seemed to be working at least (lcd3 though in my case) Aug 22 16:44:17 includedir = /usr/include Aug 22 16:44:23 inside the generated Makefile Aug 22 16:44:32 now, the question where it is emanating frmo? Aug 22 16:44:35 from?* Aug 22 16:44:38 * kergoth is happy to no longer see Error highlights in his recipes and classes Aug 22 16:44:40 anyone with autotools experience? Aug 22 16:44:56 check the initial sources Aug 22 16:45:01 Makefile.am Aug 22 16:45:25 Makefile.am does not have any includedir entry. Aug 22 16:45:31 kergoth: awesome :) Aug 22 16:45:57 * ant_work heading home, bye Aug 22 16:46:01 that is pretty much it: http://paste.kde.org/~lpapp/p66b5ad78/ Aug 22 16:46:01 lpapp: that's what Makefile is generated from Aug 22 16:47:19 lpapp: src/Makefile.am, no? Aug 22 16:47:54 stunnel_CPPFLAGS += -I$(SSLDIR)/include Aug 22 16:48:02 WINCPPFLAGS = -I$(OPENSSLDIR)/include Aug 22 16:48:40 the latter is presumably for windows, but the former is your problem, I think Aug 22 16:49:23 why is that a problem? Aug 22 16:49:27 fuck Aug 22 16:49:28 SSLDIR will be worked out in some way that does not work when crosscompiling Aug 22 16:49:30 fuck Aug 22 16:49:48 cristianiorga: channelling the spirit of koen? Aug 22 16:49:53 oooops Aug 22 16:49:59 tf: the issue is not related to openssl Aug 22 16:50:07 it fails at /usr/include/features.h Aug 22 16:50:14 doesn't matter Aug 22 16:50:31 if the makefile injects /usr/include into the include path, you are screwed Aug 22 16:50:36 no, it is just that I tried to search for the term, and entered text in the wrong place :-) Aug 22 16:50:39 tf: http://paste.kde.org/~lpapp/pdf856e5b/ Aug 22 16:50:45 tf: SSLDIR is not /usr/include/ Aug 22 16:50:54 cristianiorga: lol Aug 22 16:51:00 oooops Aug 22 16:51:07 lpapp: what is it then? Aug 22 16:51:16 tf no clue, sorry. Aug 22 16:51:33 so how do you know it's not /usr/include ? Aug 22 16:51:42 and it happened to me twice :-) oops again Aug 22 16:51:50 because it is followed by /include Aug 22 16:52:01 and it is not looked as /usr/include/include for the features.h Aug 22 16:52:47 lpapp, you work from the Makefile back to Makefile.am, but most likely you will end up in configure.ac, where there will be some home-grown check for some include location that is broken when crosscompiling Aug 22 16:53:15 tf> the weird thing is the fact it seems to work for others on arm Aug 22 16:54:22 and what does that prove? :) Aug 22 16:54:41 buildsystem is not b0rked. Aug 22 16:54:48 no, it does not Aug 22 16:54:49 btw, there is also an option called --with-ssl. Aug 22 16:55:01 where the SSLDIR comes from. Aug 22 16:55:06 maybe I should supply the poky thingie ? Aug 22 16:55:34 it is also used on gentoo what ant_work linked before. Aug 22 16:55:38 you could try that, but you really need to work out where in the Makefile the /usr/include comes from, it's nto that hard Aug 22 16:55:41 to be more complex, it is not used on arch though. Aug 22 16:55:52 tf: you are welcome to help with that. Aug 22 16:55:58 nothing I found the last 10-20 minutes. Aug 22 16:57:10 lpapp: you look into the Makefile and find where the -I/usr/include is Aug 22 16:57:15 nope Aug 22 16:57:20 I already grepped for that, nothing really. Aug 22 16:57:24 it is more complex than that. Aug 22 16:57:48 what might be related from configure.ac is this: http://paste.kde.org/~lpapp/pc09933c6/ Aug 22 16:58:15 lpapp: which is what I told you ages ago Aug 22 16:58:30 your best option is to use the --with-ssl argument then Aug 22 16:58:47 which I mentioned above. Aug 22 16:58:53 sure Aug 22 16:58:57 however, that is not a solution itself. Aug 22 16:59:09 as it still needs to receive the proper path which I would not be sure what is. Aug 22 16:59:34 bluelightning, that's hopeful. I'm trying a cfo215_ customized angstrom image right now. preping the sd card. I'll see how that goes. Aug 22 17:00:10 bluelightning, i'd still like to get the poky booting though. Aug 22 17:00:13 lpapp: see bitbake.conf. STAGING_EXECPREFIXDIR might be the right value for that test. Aug 22 17:00:35 I do not find ant_work's gentoo log anymore. :( Aug 22 17:01:19 http://xrl.us/bppi3g Aug 22 17:01:33 gentoo does: --with-ssl="${EPREFIX}"/usr \ Aug 22 17:01:35 lpapp: probably ${STAGING_INCDIR}/ Aug 22 17:01:46 what is the equivalence for Yocto, STAGING_EXECPREFIXDIR ? Aug 22 17:02:14 STAGING_INCDIR = STAGING_EXECPREFIXDIR/include ? Aug 22 17:02:51 tf: forward slash ? Aug 22 17:03:42 EXTRA_OECONF += "--disable-fips --ssl-dir='${STAGING_INCDIR}'" ? -> something like this ? Aug 22 17:04:13 EXTRA_OECONF += "--with-ssl='${STAGING_INCDIR}' --disable-fips" Aug 22 17:04:16 sorry, that one. Aug 22 17:05:44 NOTE: Executing RunQueue Tasks Aug 22 17:05:44 NOTE: Tasks Summary: Attempted 568 tasks of which 555 didn't need to be rerun and all succeeded. Aug 22 17:05:47 khem: hi there, do you know how to add libstdc++ to uclibc based image? Aug 22 17:05:56 but I cannot find the stunnel binary anywhere ? Aug 22 17:06:09 so why am I not getting errors if I cannot find that anywhere either? Aug 22 17:07:03 Krz: is there anything linking against it or you just want to add the library? Aug 22 17:07:22 file ./tmp/pkgdata/armv5te-linux-gnueabi/stunnel Aug 22 17:07:33 ./tmp/pkgdata/armv5te-linux-gnueabi/stunnel: Quake I or II world or extension, 1953701946 entries Aug 22 17:07:36 Quake ?? Aug 22 17:07:50 * lpapp is full of confused Aug 22 17:10:12 ./tmp/sysroots/foo/usr/lib/stunnel/libstunnel.{la,so} Aug 22 17:10:15 strange, I have those. Aug 22 17:10:18 but not the binary itself. Aug 22 17:10:29 do I need to have a do_install for get everything just right? Aug 22 17:13:56 RP: I want library, since we develop some code in C++ Aug 22 17:14:12 bluelightning_, angstrom booted. now to put my Qt app on it and see how well it plays with the touch screen. Aug 22 17:14:16 Krz: just add the lib to the IMAGE_INSTALL line? Aug 22 17:14:41 seems stuff is built. Aug 22 17:15:05 RP: Easy one. I thought there is some magic behind it. Aug 22 17:15:09 but how can I get everything into the right place? Aug 22 17:15:20 i.e. including bin, not just the library Aug 22 17:15:22 RP: don't I have to change uclibc flags? Aug 22 17:15:23 I thought it would be automatic? Aug 22 17:15:45 Krz: I don't think so, IMAGE_INSTALL += "libstdc++" should work Aug 22 17:15:53 Krz: could be wrong though Aug 22 17:16:04 RP: ok, will give it a go Aug 22 17:16:09 RP: will let you know if it fails, thanks Aug 22 17:16:51 Krz: there is also libstdc++-dev in case you want that Aug 22 17:17:47 RP: no, we just want to run binaries on target compiled outside Aug 22 17:20:02 here you can see the built stuff, and then the output into the sysroot: http://paste.kde.org/~lpapp/p37c02b92/ Aug 22 17:20:08 tf: rburton bluelightning_ ^ Aug 22 17:21:02 oh, I have ./tmp/deploy/rpm/armv5te/stunnel-4.56-r0.armv5te.rpm Aug 22 17:21:06 how can I check what it contains? Aug 22 17:21:28 lpapp: one way is to look in packages-split/stunnel/ Aug 22 17:21:57 bluelightning: packages-split where ? Aug 22 17:22:26 lpapp: under the workdir for the recipe, it's shown in your paste Aug 22 17:22:29 work ? Aug 22 17:22:30 ok Aug 22 17:22:51 ok, then all fine, thanks. Aug 22 17:22:54 submit time, I guess. Aug 22 17:23:30 I guess it is ok if I submit it created in the read only repo. Aug 22 17:23:35 will there be problem out of that ? Aug 22 17:31:57 sent against the write stuff; thanks again everyone involved. Aug 22 18:58:59 hi, I wonder why packages do this all around: "${SOURCEFORGE_MIRROR}/tunctl/tunctl-${PV}.tar.gz Aug 22 18:59:02 " Aug 22 18:59:14 rather than using ${PN} for /tunctl/ etc Aug 22 19:00:36 ${PN} can change when using multilib builds or native/nativesdk builds.. Aug 22 19:00:50 ${BPN} will not change, and likely could be used as a substitute Aug 22 19:01:08 well, the point is why hard code? Aug 22 19:01:50 likely so if someone renames the component the right file is still downloaded.. but it's convention vs necessity Aug 22 19:02:27 ? Aug 22 19:02:35 why would you rename foo to bar? Aug 22 19:02:35 you can also you ${BP} which is ${BPN}-${PV} Aug 22 19:02:53 sgw_: yeah, I know. Aug 22 19:03:01 no, someone might rename tunctl_1.0.bb to tunctl-1.0_1.0.bb so they could have multilib versions.. Aug 22 19:03:06 the point is why there are so many hard coded recipes around Aug 22 19:03:24 fray: that would not make so much sense Aug 22 19:03:25 lpapp, because that is how the original writer of the recipe did it.. and it doesn't really matter Aug 22 19:03:51 lpapp sure it does.. if I need to have both version 1.0 and 2.0 of a package (usually a library) thats the way you do it, via rename method.. Aug 22 19:04:12 I do not follow Aug 22 19:04:15 point is, there is nothing stopping a switch from explicit name to using BPN or BP or anything else.. it's just not what the authors of the individual recipes have done. Aug 22 19:04:28 if you rename due to version ${BP} will just work (tm). Aug 22 19:09:17 a partial example.. Aug 22 19:09:32 we have two copies of libpng that we want both installed.. one is called libpng_1.6.3.bb, the other is libpng12_1.2.50.bb Aug 22 19:09:45 if both were simply libpng_ver.bb then only one could be built and installed at a time.. Aug 22 19:10:11 so there are cases where the recipe name is NOT the same as the package name being downloaded.. that used to be a lot more common then it is today.. Aug 22 19:10:42 and yes, many of the cases could be changed to ${BP} and work just fine.. on an uprev, that would probably be an acceptable change Aug 22 19:11:02 mayhaps Aug 22 20:18:44 sgw_: so I have only a few days left to get u-boot in to make the update for the next release? Aug 22 20:20:30 lpapp: right to make 1.5 updates and feature freeze is Sunday Midnight PST. Aug 22 20:20:50 lpapp: your last version you withdrew because of some issues, right? Aug 22 20:21:24 yes, me and the u-boot guys have had no any idea why fw-utils fails. Aug 22 20:21:39 so I am about to update it leaving that out Aug 22 20:21:49 only u-boot and the mkutils Aug 22 20:21:58 to be honest I do not even know what the fw-utils was meant to do Aug 22 20:22:14 usually people need the u-boot image and the mkutils stuff to build the kernel for uImage Aug 22 20:22:17 nothing more. Aug 22 20:22:32 also, someone had a minor nitpick. Aug 22 20:23:19 sgw_: will there be any consolidated pull until Sunday? Aug 22 20:25:10 lpapp: There will be one tomorrow and then Monday after I do initial testing with everything that comes in for our build Tuesday Aug 22 20:25:34 well, I am pushing it now then Aug 22 20:25:37 without fw-utils Aug 22 20:26:08 tommorows is based on what I am building now, so this will be in monday's likely. Aug 22 20:26:28 but I will review it after I get back from my run! Aug 22 20:26:48 ok for me if that does not mean it misses 1.5 Aug 22 20:34:03 JaMa: ping Aug 22 20:34:10 pong Aug 22 20:34:30 JaMa: meta-qt5 seems to be broken, http://paste.kde.org/~lpapp/p56d7bafb/ Aug 22 20:36:21 lpapp: it looks you've downloaded only HTML redirect, right? Aug 22 20:36:57 you should try with newer meta-qt5.. Aug 22 20:37:55 well, it is ten days old. Aug 22 20:38:52 well structure on qt-project.org was changed few days ago and was immediately updated in meta-qt5.. Aug 22 20:39:10 niiice Aug 22 20:39:31 sgw_: submitted, but it is bloated now due to the removals. Aug 22 21:00:40 does anyone one how to specify rpm encoding? means utf8 or ISOxxx" Aug 22 21:02:04 how does the PACKAGECONFIG variable work? from reading libav's recipe, it looks like there is a mechanism to build it without libx11, but i don't know where/how to set it Aug 22 21:03:31 BCMM: it is mapping to the config options. Aug 22 21:04:18 lpapp: yeah, i mean, where do i set it? Aug 22 21:04:25 BCMM: https://www.yoctoproject.org/docs/current/poky-ref-manual/poky-ref-manual.html Aug 22 21:04:28 see the description. Aug 22 21:04:42 BCMM: you set it in the recipe, for instance .bbappend Aug 22 21:04:59 lpapp: so sorry, i thought i'd searched that page, but clearly i hadn't Aug 22 21:05:29 the description at the bottom in the glossary. Aug 22 21:05:45 lpapp: actually, that just describes what the variable does... Aug 22 21:05:56 lpapp: are you saying that editing the recipe is the correct way to change it? Aug 22 21:06:30 yes Aug 22 21:06:55 well, .bbappend to be precise, in your layer. Aug 22 21:07:28 if you wanna modify an existing package... otherwise if it is new, I guess you can do in your .bb Aug 22 21:07:28 BCMM: lpapp: well, generally the default packageconfig rely on DISTRO_FEATURES. especially for x11, i really think libav should not do that, and instead do something like here http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-multimedia/alsa/alsa-tools_1.0.26.1.bb#n25 Aug 22 21:07:49 ndec: in this case, it says it can't build ffmpeg without x11 in DISTRO_FEATURES Aug 22 21:09:03 BCMM: why so? Aug 22 21:09:24 ndec: sorry, i mean libav (which provides ffmpeg) Aug 22 21:09:34 to me it says, i want x11 by default. if your distro does use it, make a .bbappend... Aug 22 21:09:44 ndec: i guess libav's default PACKAGECONFIG doesn't rely on distro features... Aug 22 21:09:52 my point is that it should ;-) Aug 22 21:09:58 like every other recipe. Aug 22 21:09:59 yeah, i'll use a bbapend Aug 22 21:10:35 ndec: i quote: PACKAGECONFIG ??= "bzip2 x264 x11" Aug 22 21:11:05 so the right way to modify a recipe from somebody else's layer is to make my own layer with a bbappend containing the changes? Aug 22 21:11:14 yes Aug 22 21:11:19 thanks, i'll do that Aug 22 21:11:48 unless you plan to upstream it asap. Aug 22 21:11:57 i started a build with libx11 stuff before realising i probably didn't have to do that. is there a clean way to avoid that stuff ending up in my final image? Aug 22 21:11:57 which might be the case in here, actually. Aug 22 21:12:43 lpapp: i'm a complete newbie to yocto/oe - guess you're talking about what ndec said, that it shouldn't want x11 by default when DISTRO_FEATURES doesn't have x11? Aug 22 21:16:00 lpapp: do i need to basically duplicate the directory structure to make the bbappend apply to the right recipe? Aug 22 21:17:41 no. Aug 22 21:17:54 you only need to make sure that the .bbappend is parsed. Aug 22 21:18:06 e.g. conf/layer.conf file must reference it. Aug 22 21:18:31 and the file name of the .bbappend is important, it must be the same as .bb Aug 22 21:19:14 ndec: "conf/layer.conf file must reference it" - the layer, or the specific bbappend? Aug 22 21:19:50 to make your own layer, you need to create conf/layer.conf in your layer. Aug 22 21:20:09 that file has the list of all recipes in your layer (through regexp). Aug 22 21:20:35 you don't even need to make any directory structure in fact. you could have a flat directory of recipes, it's just a convention. Aug 22 21:20:47 only the filename is important, not the dirname. Aug 22 21:26:40 lpapp: just wanted to check you have removed some older versions of u-boot, are you sure that none of the bsps need those specific versions? Aug 22 21:50:35 seebs: are you here? Aug 22 21:50:47 * seebs exits, pursued by a bear. Aug 22 21:50:50 ... yes Aug 22 21:51:40 lol ... still continue on yesterday's issue about snappy. I feel the generated rpm shouldn't be used with ascii decoder. Is there anyway to specify the encoding is utf8? Aug 22 21:52:38 I don't know. Thinking about it, I think there's a couple of separate questions here. The first is what exactly is triggering the encoding check -- what text is it that has this non-ASCII value, and where does it come from? Aug 22 21:53:09 Then we get the question of whether to just try to work around it by editing something so there's no non-ASCII value happening, or to try to change the policy. Aug 22 21:53:45 My guess would be that the policy comes from one of two sources: The first would be "it just happens to be the default and no one ever thought about it before", the second is "there are specific things we care about which break if there are non-ASCII values in certain fields". Aug 22 21:54:00 non-ASCII in that created rpm I don't know ... it's mystery to me too, but using utf8 decode shouldn't have that error at least Aug 22 21:54:52 i checked the package itself, it doesn't specify any encoding or that 0xc1 byte, it's hard to edit the rpm either Aug 22 21:55:01 Hmm. Aug 22 21:55:55 So is this an error you get from a specific RPM which was successfully created? If so, maybe someone who knows more about RPM than I do could stare at it a bit to try to determine whether it's contrary to the spec or whether this is a bug in the utility that's throwing the error. Aug 22 21:58:08 seebs: yes, it's from libsnappy.x.x.x.rpm created ... Aug 22 21:58:29 error happening while rpm trying to install it into rootfs tarball Aug 22 21:59:29 I saw backends/rpm/header.py has different encodings defined like utf8 or iso-8859-1, just don't know how to specify it at the beginning Aug 22 22:01:07 Hmm. Aug 22 22:04:23 I do not know this code at all, but it seems to me that what's happening involves the reader/writer being initialized around line 271 of pm.py, in which case just adding ENCODING = "utf8" right after the existing setting of ENCODING would likely show you a possibly-different set of behaviors. Aug 22 22:06:04 It appears that it's trying to determine the "preferred" encoding, and defaulting to ASCII. Without adding some classy debugging prints, I can't tell which one of those is happening. It looks like it might well be harmless to try changing it. Aug 22 22:07:59 This does look intentional though, so it may well be that tweaking that will result in strange failures elsewhere. Aug 22 22:08:08 seebs: ok, I'll give it try but it's just some concept prove, the real fix might need to be fine tuned Aug 22 22:08:18 yep Aug 22 22:08:42 You could also, probably, change reader(open(rpmoutpath)) to reader(open(rpmoutpath), errors = "ignore") Aug 22 22:09:28 seebs: that sounds better Aug 22 22:13:56 Hmm. This has been this way forever, it seems. Aug 22 22:16:59 Okay, this code appears to come from revision 393 of smartpm, and the revision log is "Changed some bits in the rpm output parsing trying to solve encoding bugs." From 2005-01-19. Aug 22 22:17:39 seeb: it went rhough with the change. To fix it permanently, shall I directly patch python for pm.py? Aug 22 22:18:43 I would not recommend that without finding out why this was set up this way. The key point seems to be revision 389, "Use converting codecs when grabbing rpm output through sys.std{out,err}." Aug 22 22:19:10 So if you can find out why that change was needed, or what it was trying to solve, that'll give you a good starting point in figuring out what the "right" solution is. Aug 22 22:19:35 I will say... My basic intuition is that it is almost certainly wrong for the interpretation of RPM output by smartpm to be contingent on the user's locale settings. Aug 22 22:19:36 alright, fair Aug 22 22:30:03 i've managed to successfully build an image (rpi-hwup-image). what's the correct way of doing it again, with an extra package (mpd) included? if i just do bitbake mpd then bitbake rpi-hwup-image again, will it be included in the new image? Aug 22 22:31:36 BCMM: no. Aug 22 22:31:50 the package would be built, but not added in the image. Aug 22 22:32:08 to add a package in an image, you need to explicitely request that. Aug 22 22:32:35 see http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#usingpoky-extend-customimage Aug 22 22:34:29 ndec: so what exactly does the sysroot represent? Aug 22 22:34:52 it's where each package puts .h files and other development files. Aug 22 22:35:12 so that when building B, if B needs .h files from A, it will pick them there. Aug 22 22:35:21 assuming A has 'installed' the .h files properly. Aug 22 22:35:38 ah, it's somewhere for include paths, not install paths, to exist? Aug 22 22:35:43 the .h files are not used from their orginal $WORKDIR Aug 22 22:35:50 yes. Aug 22 22:36:01 thanks Aug 22 22:36:10 well, 'native' tools go there too for example. Aug 22 22:36:43 and it's only two lines to make my own image... bitbake seems to make basing stuff on other stuff pretty easy Aug 22 22:36:54 hehe Aug 22 22:37:08 on the shoulders of giants ;-) Aug 22 22:56:40 I am trying to make changes to u-boot and then recompile u-boot source. It looks like it is not taking any of the changes and it is generating the same u-boot binary all the time. My changes are not reflected in the u-boot binary Aug 22 23:17:45 Can anyone help me with this> Aug 22 23:25:19 Hari_: with what? Aug 22 23:25:52 u-boot is not compiling properly, my changes are not reflected in the binary Aug 22 23:26:09 I did a clean build Aug 22 23:28:02 totally OT, but anyone had success with the rt2800usb driver? getting failed to init hw with my rt2870 adapter Aug 22 23:29:18 kergoth: pastebin? Aug 22 23:31:53 gah, ejected and re-inserted, and now getting a spinlock lockup, down for the count.. Aug 22 23:32:02 * kergoth glares at his board **** ENDING LOGGING AT Fri Aug 23 02:59:59 2013