**** BEGIN LOGGING AT Thu Apr 10 02:59:58 2014 Apr 10 07:15:53 moin Apr 10 09:05:47 bluelightning: remember yesterday's PLATFORM_ARCH = ${MACHINE} conversation? it looks like it doesn't work out for me, because the packages are built in-place (source code is not downloaded) Apr 10 09:06:02 which means the build dir is the same for all machines Apr 10 09:07:09 so I'm back to square one -- even if I clean the package and rebuild it, bitbake considers it the same because the recipe version and filename match Apr 10 09:07:32 there's no notion that the package can be built for different machine targets :( Apr 10 09:07:35 xerent: PACKAGE_ARCH not PLATFORM_ARCH - you are using the former in the recipe right?# Apr 10 09:07:53 PACKAGE_ARCH, I mean, yes. Sorry. Apr 10 09:08:33 ah and it's MACHINE_ARCH not MACHINE, but I guess you have that right in the recipe as well? Apr 10 09:09:54 My colleague thinks that PACKAGE_ARCH is set implicitly anyhow, regardless of whether I do it explicitly in the application recipe. Is he correct? Apr 10 09:10:40 I'll check the MACHINE vs MACHINE_ARCH thing, if that's causing it Apr 10 09:10:43 he's right in that it's set implicitly to "${TUNE_ARCH}" Apr 10 09:10:56 but that is the same for machines that share the same base architecture Apr 10 09:11:15 MACHINE_ARCH contains the machine name itself so it is different per machine Apr 10 09:12:07 ah! Apr 10 09:25:29 <_valle_> I have built an image for imx53 using yocto. At boot an image of tux is shown. I cannot find anything in uboot. Where is this set? I want to remove it Apr 10 09:27:47 _valle_: its in your kernel config, i just don't remember the exact switch Apr 10 09:30:01 <_valle_> LetoThe2nd: Thank you Apr 10 09:30:21 something in the boot category, IIRC Apr 10 09:47:02 Hi everybody! Is there a way to list the SRC_URIs of all available recipes? background: I need to give somebody access to all git repos that are hosted at our git server so he can build an image himself. Apr 10 10:05:57 hsychla: Why not provide a mirror instead? Apr 10 10:06:17 hsychla: did you try bitbake -e | grep SRC_URI? :> Apr 10 10:06:45 hsychla: https://wiki.yoctoproject.org/wiki/How_do_I#Q:_How_do_I_create_my_own_source_download_mirror_.3F Apr 10 10:07:26 You can then produce a consistent offline build without needing access to the git server Apr 10 10:07:53 Xz: that's not going to grab the values for all recipes ;) Apr 10 10:08:15 the mirror includes a snapshot of the servers repo. Assuming you dont have scary recipes that track head it'll work fine Apr 10 10:10:46 we don't want to make the sources publicly available. we want to give external developers a build environment so they can build and test their changes Apr 10 10:10:50 So... My morning question for all - I'm putting together a new build for some custom hardware. In my layer I've added cases to build for both the reference platform target and my custom board. What's the generally accepted way to build for *both* targets? Apr 10 10:11:37 but i think I got my list with a little find/cut/grep magic Apr 10 10:11:43 I assume I cant use the same build dir as they could clash? So should I re-call oe-init-build-env with a new MACHINE and new build dir to swap between the two? Apr 10 10:13:19 hsychla: someone can correct me if Im wrong but I'm pretty sure that means you'll need to make some new binary recipes for your source then... Apr 10 10:14:36 pev: why? if their key is granted access in our gitolite installation, they can access the sources Apr 10 10:14:44 hsychla: I don't quite get the contradiction of "I need to give somebody access to all git repos that are hosted at our git server" with "we don't want to make the sources publicly available" Apr 10 10:15:24 hsychla: OK, so using ssh access...? Apr 10 10:15:34 pev: yes Apr 10 10:16:37 hsychla: Something like Apr 10 10:16:40 find . -name "*.bb*" | xargs grep -sI SRC_URI | grep git://github.com Apr 10 10:16:42 for example Apr 10 10:16:59 assuming you're in the root of all your layers... Apr 10 10:17:06 yes, that's what i meant by "find/cut/grep magic" Apr 10 10:17:11 :-) Apr 10 10:17:22 thanks anyway Apr 10 10:18:06 It seems to be generally easiest to use your own crafted find/greps for such things Apr 10 10:18:37 rburton: Do you know anything about building for multiple machines in the same world? Apr 10 10:56:54 <_valle_> With CONFIG_LOGO_LINUX_CLUT224=n I got rid of the logo during boot. Apr 10 10:57:15 <_valle_> I also set CONFIG_FRAMEBUFFER_CONSOLE=n Apr 10 10:57:55 <_valle_> However this gives me another problem... At boot, instead of getting a black screen I get colors randomized Apr 10 10:58:59 <_valle_> Does someone know how to get a black screen during bootup after CONFIG_FRAMEBUFFER_CONSOLE=n is set? Apr 10 11:00:28 _valle_: we simply use the coprocessor to disable display backlight during boot - but of course that requires a coprocessor that controls backlight ;) Apr 10 11:06:12 <_valle_> xerent: Okay :) but that won't work for me. Apr 10 11:12:51 _valle_: Can you not configure it then blank it in your bootloader Apr 10 11:13:01 i.e. a black bootsplash? Apr 10 11:15:46 <_valle_> I guess I could make my own black logo and replace the logo linux_clut224 with mine. Apr 10 11:16:04 <_valle_> pev: Was that what you meant? Apr 10 11:17:25 _valle_: yea Apr 10 11:18:06 _valle_: Another option is to modify the fb driver so that when it initialises the framebuffer memory it memset's it first. Easy for RGB Apr 10 12:13:59 <_valle_> I made a patch for the logo_linux_clut224.ppm with my own logo. How should I apply this? Should I write a recipe just to apply this patch? Apr 10 12:17:08 <_valle_> Or should I use linux-imx_2.6.35.3.bbappend to change this? Apr 10 12:21:35 _valle_: in your own layer it's up to you; bbappend would be fine for this kind of thing I think Apr 10 12:26:35 <_valle_> bluelightning: Okay, trying to learn how things is done usually :) Apr 10 12:46:09 <_valle_> pev: Where should I start to modify the fb driver? Apr 10 12:46:10 hi. i'm trying to get SPI working on i.MX233 Olinuxino Maxi using Freescale BSP. I've got an oscilloscope connected to the SPI pins, but I'm getting absolutely nowhere ... Apr 10 12:46:39 I have enabled SPI (spidev etc) in the kernel, but there is no /dev/spi* to be found.. Apr 10 12:46:57 i have no idea where to go from here ... :( Apr 10 12:50:28 _valle_: Which kernel /driver are you using? Apr 10 12:51:07 <_valle_> pev: Kernel linux-imx_2.6.35.3 Apr 10 12:52:30 _valle_: Ive not used that before - which repository does it come from? Apr 10 12:52:38 https://github.com/Freescale/linux-mainline ? Apr 10 12:53:01 git://git.freescale.com/imx/linux-2.6-imx.git ? Apr 10 12:54:24 <_valle_> pev: I don't know. I'm using the layer meta-fsl-arm Apr 10 12:54:53 _valle_: What do you enable driver wise in your kernel .config? Apr 10 12:55:43 Looks like it might be imxfb? Apr 10 12:56:48 My best guess not having looked at your build would be that you're using : Apr 10 12:56:48 http://git.freescale.com/git/cgit.cgi/imx/linux-2.6-imx.git/tree/drivers/video/imxfb.c?h=imx_2.6.35_maintain Apr 10 12:57:04 Have a look in : imxfb_probe() Apr 10 12:57:20 Looks like fbi->map_cpu is the base address of the FB Apr 10 12:58:34 you could probably add memset(fbi->map_cpu, 0, fbi->map_size); after the check for "Failed to allocate video RAM" and see if that works Apr 10 12:59:10 and if it does, play nicely and wrap it in your own specific CONFIG_FB_IMX_ZERO or something Apr 10 12:59:42 <_valle_> pev: Thank you! I shall try it Apr 10 13:00:56 Of course check that that *is* the fb your board is using first though! Apr 10 13:01:13 and maybe add a printk to prove that you're hitting it. Apr 10 13:05:37 <_valle_> pev: Hm how do I check what fb that is being used? Apr 10 13:32:28 _valle_: Look at what objects have been built? Apr 10 13:54:31 ok. so. i've created a recipe for a software package built using autotools Apr 10 13:54:40 libnl-3 to be precise Apr 10 13:55:03 when building i get this error: Apr 10 13:55:09 ERROR: QA Issue: non debug package contains .debug directory: libnl path /work/armv5te-oe-linux-gnueabi/libnl/3.2.24-r0/packages-split/libnl/usr/lib/libnl/cli/qdisc/.debug/blackhole.so Apr 10 13:55:14 any ideas? Apr 10 13:56:03 my .bb follows the example from the development manual almost precisely... Apr 10 14:02:15 ionte: I'm pretty sure there is a libnl3 recipe already though... Apr 10 14:02:53 under meta/recipes-support/libnl/ Apr 10 14:03:39 oh. i couldn't find it before, but there it is. thanks! Apr 10 14:04:13 what's the best way to make a package that doesn't build, but just deploys a bunch of files into the file system? Apr 10 14:04:21 make a package, or edit image preprocess? Apr 10 14:07:05 xerent: typically make a separate recipe, unless you have to construct the file contents completely dynamically at image creation time that is Apr 10 14:08:18 do I need an empty Makefile and what not, or can I just override do_compile in the recipe to do nothing Apr 10 14:14:51 xerent: no need for a Makefile - in fact do_compile won't do anything (and won't error) if there is no Makefile in the source dir Apr 10 14:15:26 OK thanks :) Apr 10 14:25:19 xerent: There're loads of examples in the tree - look for stuff that deploys scripts or init scripts for examples Apr 10 14:27:35 right, the basics are just mention the files in SRC_URI, have do_install install them to the appropriate locations under ${D} and then if needed add to FILES_${PN} if the files are installed to locations other than the "standard" ones Apr 10 14:52:41 hi, I'm trying to prevent a kernel module from loading at boot. I tried a file in /etc/modprobe.d, and modprobe.blacklist as a kernel option, but neither worked. What's the recommended way to do this in Yocto? Apr 10 14:53:04 (I still want the module to be available for manual loading, but it doesn't work when loaded early in the boot process) Apr 10 14:54:15 PulkoMandy: you could try module_autoload_ = "", but that'll only work if that is how it's being loaded Apr 10 14:55:58 PulkoMandy: I think you can blacklist on the kernel commandline too? Apr 10 14:56:38 https://wiki.archlinux.org/index.php/kernel_modules#Using_kernel_command_line_2 Apr 10 14:57:25 yes, I tried that but it doesn't seem to work Apr 10 14:57:28 If its not working, it's probably a syntax thing no? Apr 10 14:57:57 modprobe.blacklist=mwifiex_sdio Apr 10 14:58:11 I have this in the kernel command line, seems hard to make a syntax error there Apr 10 15:01:40 hmm, if I have two packages that install the same file, and I want one to have priority Apr 10 15:02:50 PulkoMandy: Have a look in your busybox defconfig Apr 10 15:03:03 PulkoMandy: Whats the state of CONFIG_FEATURE_MODPROBE_BLACKLIST? Apr 10 15:04:12 xerent: What file? Apr 10 15:04:25 a text file that I put into a random folder Apr 10 15:04:29 xerent: Is there a good reason for that? Normally you try to avoid that kind of contention... Apr 10 15:04:50 package A is included in two images, and package B is included in one image, where that file should be overridden Apr 10 15:05:11 so I guess I'll have to split package A into A1 and A2, and only install A1 on the second image with B Apr 10 15:05:19 but it seems like a mess Apr 10 15:05:20 poky/meta/recipes-core/busybox/busybox-1.21.1/defconfig:506:# CONFIG_FEATURE_MODPROBE_BLACKLIST is not set Apr 10 15:05:27 I guess I can customize that? Apr 10 15:05:33 Normally I'd put the file in it's own package and have the other two depend on it maybe? Apr 10 15:05:39 PulkoMandy: For sure! Apr 10 15:06:31 PulkoMandy: Just add a bbappend that modifies it's state Apr 10 15:06:50 if I made package B dependent on package A, will package A be built first, and then the file is overwritten when B is installed? Apr 10 15:06:55 PulkoMandy: I do that in some of my layers. I add a : do_prepare_config_append() Apr 10 15:07:19 PulkoMandy: and in that do a sed e.g. sed -i 's/# CONFIG_MPSTAT is not set/CONFIG_MPSTAT=y/' ${S}/.config Apr 10 15:08:08 PulkoMandy: There's probably a better / 'proper' way to do this these days (someone once mentioned config fragments to me) but this is the only way I could get to work... Apr 10 15:13:14 Anyone know how to work out dependencies? i.e. my current build decides it wants to build perl for some reason I can't explain which really extends the build but I don't know what specifically has pulled it in? Apr 10 15:14:16 xerent: no, most package managers will abort installation entirely when doing that unless forcing the install, or unless package B indicates it replaces package A entirely (RREPLACES, potentially with RCONFLICTS/RPROVIDES). the only other possibility that comes to mind is using CONFFILES, not sure if the package manager allows those to be replaced or not. afaik. Apr 10 15:25:44 is there a way to update package list after building a single recipe (bitbake package) instead of an image (bitbake image)? Apr 10 15:26:00 what do you mean by 'package list'? Apr 10 15:26:17 If you do a full release, what is the most important thing beside keeping the steps repeatable (by using a local mirror, using fixed repository version/hashes)? Apr 10 15:26:19 deploy/ipk e.g. Apr 10 15:26:21 How helpful is the buildhistory option? Apr 10 15:26:42 you mean the package feed/index? see the package-index recipe Apr 10 15:27:04 kergoth: in the deploy/ipk folder there are Package description files, but when you do a bitbake package, these are not updated Apr 10 15:27:46 kergoth: ok Apr 10 15:29:40 volker-: enable it and take a look for yourself. personally, i'd never do a build without it Apr 10 16:05:25 kergoth: i looked at it and checked also the documentation about it (thats how I found it). But I don't really get the sense/meaningfulnes of it Apr 10 16:05:48 it gives you a ton of useful information, i'm not sure how you wouldn't see the usefulness Apr 10 16:06:01 what packages are in the image, what sizes they are, what packages were emitted, what metadata they have Apr 10 16:06:14 but combined with the ability to store buildhistory in a git repository, it becomes even more useful Apr 10 16:06:24 since you can track what changes to the build occurred when you make a change to the metadata Apr 10 16:06:35 and monitor changes in the image contents or size over time, to watch for regressions Apr 10 16:08:03 kergoth: hmmm. maybe it makes more sense with scm Apr 10 16:08:35 kergoth: I also see package information that are not even added to the image Apr 10 16:08:37 set BUILDHISTORY_COMMIT ?= "1" Apr 10 16:08:44 the package info is about all packages built Apr 10 16:08:48 the image info is what's limited to the image Apr 10 16:08:52 kergoth: I have to check if it supports p4 :( Apr 10 16:08:56 why? Apr 10 16:09:34 my company does not support git at the moment and they don't see a reason for it Apr 10 16:10:35 its up to you whether you put your buildhistory repository somewhere persistent. even just local it's useful, though less for monitoring long term changes. presumably you could just check in the buildhistory git repo into your p4 repo, no? :) Apr 10 16:10:36 * kergoth shrugs Apr 10 16:11:09 kergoth: look how frequent .git changes ;-) Apr 10 16:11:38 buildhistory uses its own git repository Apr 10 16:11:39 kergoth: how do you add the build history accross versions? Don't you have a branch per release? Apr 10 16:11:43 it'll change when buildhistory does a new commit Apr 10 16:12:55 kergoth: what is your benefit out of the postrm snippets in the build history? Apr 10 16:13:25 as i said, it includes all the metadata about the packages, including their install scripts, again, presumably to monitor changes over time Apr 10 16:13:29 i didn't write the thing, i just use it Apr 10 16:13:49 kergoth: so you don't use all parts of the output? Apr 10 16:14:14 buildhistory satisfies multiple use cases Apr 10 16:18:04 kergoth: true. how do you mange buildhistory accross versions? using one repo for all releases and one repo for all development builds? Apr 10 17:40:24 * kergoth thinks we should deprecate the wildcard handling in file: uris in SRC_URI. using a subdir instead is trivial Apr 10 17:50:04 bah, even if i modify qemu.bbclass to obey a variable to disable it, there's no such mechanism for the intercepts Apr 10 17:54:37 kergoth: I kind of view them as mostly the same amount of pain Apr 10 17:55:30 at least we've now got to the point where checksums should mostly work properly with them Apr 10 18:03:37 I just dislike the special casing, considering the feature doesn't really buy us anything we can't already do in other ways Apr 10 19:32:10 I'm trying to use the yocto plugin for eclipse/kepler/cdt with the run target being an IMX6 based board. I have a simple hello world program that I can run remotely via ssh/tcf-agent. However, when I try to debug it I get the error "Could not determine GDB version using command: gdb --version". "gdb --version" works on the board and on the host/eclipse system. Anyone have any idea what could Apr 10 19:32:10 be wrong? I am using a custom "bitbake -c populate_sdk" image as the sysroot/toolchain. Maybe the problem is that gdb is missing from there? Apr 10 20:52:51 how often will the git snapshots/tar.gz files in the yocto folder be updated? Apr 10 21:45:02 mhh, whats the best way to add a perl-module ? Apr 10 21:57:57 mikejanzen: you can add packages to the sdk produced using bitbake -c populate_sdk. I'm not saying that's the problem -- I don't know -- but if it is, it's resolvable Apr 10 22:01:07 see TOOLCHAIN_HOST_TASK, adding something like nativesdk-gdb (assuming that the gdb recipe supports that), would include gdb in the native sdk. Apr 10 22:04:48 regorianer: there is a cpan.bbclass, see http://www.yoctoproject.org/docs/1.5.1/ref-manual/ref-manual.html Apr 10 23:04:56 volker-: thanks, this works fine, but I'm stucked with another error since more than half an hour now :( - http://pastebin.com/qcDuwp4q and it's about this module http://search.cpan.org/~cook/Device-SerialPort-1.04/SerialPort.pm Apr 10 23:05:49 i read a lot about it and some guys are telling that these are device mount issues, but if i do a normal ./configure it works, just not with the run.do_configure sets Apr 10 23:14:35 regorianer: there must be a config.log file Apr 10 23:14:43 in your build tree Apr 10 23:14:59 that should give u more info Apr 10 23:15:45 yeah Apr 10 23:15:51 it's that he cannot run the a.out Apr 10 23:16:45 khem: if I add manually "--host arm-poky-linux-gnueabi" it works. but i cannot accept this as a solution Apr 10 23:17:33 regorianer: that should be already done for you if u inherit autotools Apr 10 23:19:47 uff, i only inherit cpan ... didnt know i also need autotools Apr 10 23:20:11 and there i dont understand why i need autotools then Apr 10 23:21:36 cpan is for perl stuff, it doesn't magically know that you're using autotools Apr 10 23:21:44 not all perl proejcts use autotoolss as their buildsystem, do they? Apr 10 23:23:12 mhh i dont get it Apr 10 23:23:37 autotools are fine, if i have a project containing a configure.ac and a makefile.am Apr 10 23:23:59 but in this point i only have the configure.ac Apr 10 23:24:26 there is no makefile, but the makefile.PL which will be run by the cpan function Apr 10 23:24:53 so ... if i inherit autotools, it fails cause autotools want to make install, but there is no make install routine in the perl-module Apr 10 23:25:09 maybe i am thinking a bit wrong, pls then correct me. :) Apr 10 23:25:55 regorianer: what does configure.ac looks like ? Apr 10 23:26:02 it seems someone did half the job Apr 10 23:26:11 http://search.cpan.org/~cook/Device-SerialPort-1.04/SerialPort.pm Apr 10 23:26:33 what i could do is to overwrite do_compile Apr 10 23:26:51 but i dont know if that's a solution one should do Apr 10 23:39:08 regorianer: don't know. with autotools you can define EXTRA_OECONF, might work with cpan, too Apr 10 23:41:39 regorianer: if you look at other tools you often see such overwrites of do_configure, patches or similar Apr 10 23:42:58 regorianer: What I always do is checking on http://layers.openembedded.org/ for similar projects. Actually there is a meta-perl that might be helpful Apr 10 23:46:47 regorianer: what happens if you add EXTRA_OECONF="--host=test"? does it show up in the error output? if so, you can inject the --host flag this way Apr 10 23:48:36 volker-: no, doesnt show up Apr 10 23:48:51 regorianer: and if it works this way, look into the autotools.bbclass at variables CONFIGUREOPTS Apr 10 23:49:08 take a look into cpan.bbclass: perl ${EXTRA_PERLFLAGS} Makefile.PL ${EXTRA_CPANFLAGS} Apr 10 23:49:37 try EXTRA_CPANFLAGS (don't know how cpan/perl Make works) Apr 10 23:51:43 well, EXTRA_CPANFLASGS does not work either, but i take a look at the cpan.bbclass Apr 10 23:53:39 regorianer: looks like it wont help if you look in http://cpansearch.perl.org/src/COOK/Device-SerialPort-1.04/Makefile.PL Apr 10 23:54:20 regorianer: what gives me an indication that something injects something somehow is the arm-poky-linux-gnueabi-gcc -march=.... --sysroot=... line. Apr 10 23:55:04 regorianer: sorry, I am not a specialist at it, but hope to have provided some hints (where at least I would look) Apr 10 23:55:44 what do you mean by that line, in which file ? Apr 10 23:55:59 regorianer: look into your pastbin, the "checking for gcc" line Apr 10 23:56:17 ah, well there. Apr 10 23:56:54 grep -r -- "--sysroot" meta*/classes Apr 11 00:02:18 compiling tcpflow on ARM problems; Error: the autoconf log indicates errors, it looked at host include and/or library paths while determining system capabilities.. Apr 11 00:02:48 I'm new to yocto. Anyone have any pointers how I should begin to solve this? Apr 11 00:03:34 gateforks: I don't know what the detailed error is. Usually the output gives a good clue what it is. Apr 11 00:04:20 what I see in yocto is the issue I saw in the past most times when doing direct cross compiling (without a framework around it). Apr 11 00:06:24 gateforks: if you use the autotools class, you can modify the ./configure flags with EXTRA_OECONF. What I did in my cases was running "./configure --help" and disabling everything possible to get it stripped down to a minimum and/or looking directly in the configure.* files Apr 11 00:09:01 could you repeat that? My connection logged me out while I was reading your response Apr 11 00:09:48 gateforks: see query Apr 11 00:10:48 gateforks: and 'bitbake myrecipe -c configure' can also speed trial-and-error up Apr 11 00:22:55 well, volker- thanks for the help and hints. learned a bit. but at the end if just patch the Makefile.PN. obviosly yocto is fucking me by telling me Device-SerialPort-dbg contains illegal characters, (other than [a-z0-9.+-]) .... nothing just can work instantly ... :D but okay, i go on resolve next issue ... Apr 11 00:31:56 is there a list of the var names for usr, etc, lib ... as ${sysconfdir} and the others ? Apr 11 01:05:40 there should be a doc section on recipe stuff Apr 11 01:06:44 but im pretty sure it's the standard definitions, ie, same as you would see in an rpm spec file Apr 11 01:08:54 they're macros in rpm, so not a one-to-one correspondence Apr 11 01:12:02 found it Apr 11 01:12:05 http://www.embeddedlinux.org.cn/OEManual/directories_installation.html Apr 11 02:16:28 gah, i wish this random qemu-native ICE on rhel6 would go away Apr 11 02:16:32 anyone else run into this one? Apr 11 02:44:47 thx nerdboy **** ENDING LOGGING AT Fri Apr 11 02:59:58 2014