**** BEGIN LOGGING AT Tue Sep 21 02:59:57 2010 Sep 21 03:01:33 ffmpeg is using both dsp + neon ? Sep 21 05:58:41 gm Sep 21 05:59:05 03Roman I Khimov  07org.openembedded.dev * r3e9ff12c2c 10openembedded.git/recipes/bzip2/ (3 files in 2 dirs): Sep 21 05:59:05 bzip2: update 1.0.5 to 1.0.6 Sep 21 05:59:05 * fixes CVE-2010-0405 Sep 21 05:59:05 Signed-off-by: Roman I Khimov Sep 21 05:59:05 Acked-by: Frans Meulenbroeks Sep 21 05:59:14 03Roman I Khimov  07org.openembedded.dev * r6cb49edc28 10openembedded.git/recipes/bzip2/ (3 files in 2 dirs): Sep 21 05:59:14 bzip2: remove 1.0.2 Sep 21 05:59:14 Very old, not pinned, has known bugs and security issues. Sep 21 05:59:14 Signed-off-by: Roman I Khimov Sep 21 05:59:14 Acked-by: Frans Meulenbroeks Sep 21 05:59:14 03Roman I Khimov  07org.openembedded.dev * r416da30aea 10openembedded.git/recipes/clamav/ (3 files in 2 dirs): Sep 21 05:59:15 clamav: update 0.96.2 to 0.96.3 Sep 21 05:59:15 * bugfix release, considered as safe upgrade Sep 21 05:59:16 Signed-off-by: Roman I Khimov Sep 21 06:41:04 ericben|away: yesterday you asked me to see if I could build gs for you, i kicked off a build then forgot about it so never peeked at the outcome Sep 21 06:41:24 ericben|away: this is what I get Sep 21 06:41:26 NOTE: package gs-8.64-r2: task do_compile: Started Sep 21 06:41:26 ERROR: Task 13 (/home/frans/workspace/openembedded.git/recipes/gs/gs_8.64.bb, do_compile) failed with exit code '1' Sep 21 06:41:26 ERROR: __init__() takes at most 3 arguments (4 given) Sep 21 06:42:18 eFfeM_work, update your bitbake Sep 21 06:44:15 grg, will try, one sec Sep 21 06:48:37 I'm trying to bundle libmpfr.so.4 with my SDK, but how should I tell the SDK-GCC where to look for it? Sep 21 06:51:40 ericben|away: btw when tyring to build gs, I also ended up with QA issue errors in cups (about cups.desktop) Sep 21 06:52:05 grg my bitbake is from git sep 10 Sep 21 06:52:37 I tried setting LD_LIBRARY_PATH and LD_RUN_PATH, but none of those worked Sep 21 06:53:13 eFfeM_work, The most recent commit should fix the __init__() takes at most 3 arguments (4 given) error. http://cgit.openembedded.org/cgit.cgi/bitbake/commit/?id=085e66f9c14123ea2c0f1e34f7737cf77071f86d Sep 21 06:53:22 grg only one change compared to git head in lib/bb/build.py, pulled that, retyring now Sep 21 06:53:58 you'll get a proper error message now :) Sep 21 06:54:14 yup Sep 21 06:55:15 ericben|away: now get the error Sep 21 06:55:15 | ./obj/echogs: ./obj/echogs: cannot execute binary file Sep 21 06:55:30 and a warning that I do not like: | ./psi/dscparse.c:2183: warning: array subscript is above array bounds Sep 21 07:07:46 khem: ping Sep 21 07:23:58 morning Sep 21 07:35:16 hi hrw, JaMa Sep 21 07:36:46 hi eFfeM_work Sep 21 07:37:08 JaMa had a good vacation ? Sep 21 07:39:52 eFfeM_work: yes Sep 21 07:39:58 :-) Sep 21 07:40:07 but now I'm 2,5 hour late to work :) Sep 21 07:42:06 hi eFfeM_work : thank you, this is the same error I had yesterday Sep 21 07:42:11 I'll send a fix Sep 21 07:42:18 yw Sep 21 07:42:24 & appreciate the fix Sep 21 07:44:25 03Eric Bénard  07org.openembedded.dev * r56fa8d7ad3 10openembedded.git/classes/utils.bbclass: (log message trimmed) Sep 21 07:44:25 utils.bbclass: fix is_machine_specific Sep 21 07:44:25 * base.bbclass can set a package's PACKAGE_ARCH to MACHINE_ARCH if Sep 21 07:44:25 it finds file:// in the SRC_URI and the corresponding file is machine Sep 21 07:44:25 specific (recipes/foo/foo-1.0//) Sep 21 07:44:26 * but is_machine_specific actually fails to detect these files as Sep 21 07:44:27 urldata.path is used instead of urldata.localpath thus preventing the Sep 21 07:50:05 is this channel logged? Sep 21 07:54:44 tasslehoff: yes Sep 21 07:58:48 ~logs Sep 21 07:58:48 All conversations are logged to http://ibot.rikers.org/channel, where "channel" is replaced by the URL-encoded channel name, such as %23freenode for #freenode. Lines starting with spaces are not logged. logs are updated daily Sep 21 07:58:55 tasslehoff: ^^^ Sep 21 07:59:42 !logs Sep 21 07:59:43 Channel logs for #oe are archived at: Sep 21 07:59:44 http://hentges.net/tmp/logs/irc/%23oe Sep 21 07:59:45 Live-logs are available at Sep 21 07:59:47 http://hentges.net/tmp/logs/irc/livelogs/%23oe.livelog Sep 21 07:59:49 See ?? help-logs for usage instructions Sep 21 08:00:28 eFfeM_work: thanks. Sep 21 08:00:39 yw Sep 21 08:01:23 Tartarus, khem, did you decide wether to link libmpfr.so.4 and friends statically with GCC (and if so, when to do the change)? Sep 21 08:01:42 tasslehoff: check the patch khem sent yesterday on the ml Sep 21 08:01:51 ericben: ah. ok Sep 21 08:02:26 tasslehoff: http://patchwork.openembedded.org/patch/2979/ Sep 21 08:03:26 ericben: cool. thanks. I'll try that one out. Sep 21 08:23:24 hm, that didn't work out on my system. gcc still can't find libmpfr.so.4 Sep 21 08:25:27 maybe I have to trigger a rebuild of gcc-cross Sep 21 08:39:45 hi, I updated my openembedded distro and udev doesn't generated /dev/mtdblock* Sep 21 08:40:03 did something happen with udev between 124 and 151? Sep 21 08:55:45 hm. seems libmpfr.so.4 is now present in my sdk, but gcc still complains. perhaps some environmental issues? Sep 21 09:05:54 khem *ping* Sep 21 09:09:18 tasslehoff: try using strace on the gcc, and find out where it's looking for mpfr Sep 21 09:15:19 tasslehoff: what is your exact problem ? Sep 21 09:15:48 ericben: from the beginning, or the exact error message I get now? Sep 21 09:16:11 what you get now Sep 21 09:17:58 ericben: ./angstrom/arm/libexec/gcc/arm-angstrom-linux-gnueabi/4.3.3/cc1 gives "error while loading shared libraries: libmpfr.so.4: cannot open shared object file: No such file or directory" Sep 21 09:18:45 this is inside OE or when using the sdk ? Sep 21 09:19:01 ericben: when using the sdk Sep 21 09:19:06 ok same thing here Sep 21 09:19:11 tasslehoff: what did you did to your host system? Sep 21 09:19:24 or was sdk built on other system? Sep 21 09:20:03 hrw: even on the same system we get this (and also for libtinfo when running gdb) Sep 21 09:20:07 hrw: I unpacked the SDK to a folder and ran the environment-setup Sep 21 09:20:22 host has /usr/lib/libmpfr.so.1 , sdk is using .4 Sep 21 09:20:56 .4 in in sysroots/armv6-angstrom-linux-gnueabi/usr/lib/libmpfr.so.4 Sep 21 09:21:00 so it looks like sdk does not got all libraries Sep 21 09:21:06 hrw: I had a discussion with khem and Tartarus about this yesterday, and khem did a patch to link the libs statically in GCC. Sep 21 09:21:13 sorry sysroots/i686-linux/usr/lib/libmpfr.so.4 Sep 21 09:21:36 hrw: yes the sdk lacks some host libraries Sep 21 09:21:45 hrw: I have libmpfr.so.4 in my SDK after compiling with the patch Sep 21 09:21:48 shouldn't LD_LIBRARY_PATH contain the directories which the cross-gcc needs Sep 21 09:22:35 in /angstrom/arm/arm-angstrom-linux-gnueabi/usr/lib/libmpfr.so.4 Sep 21 09:22:48 tasslehoff: oh yes it's here sorry Sep 21 09:24:26 from strace it seems cc1 only looks in my host libraries to find it Sep 21 09:24:30 it's tasslehoff : the workaround (dirty) is : cd /usr/lib ; sudo ln -s libmpfr.so.1 libmpfr.so.4 ; sudo ln -s libncurses.so libtinfo.so.5 Sep 21 09:25:45 ericben: yeah, but I'd rather not ship the SDK to my colleagues with those instructions :) Sep 21 09:26:04 tasslehoff: add the directory to LD_LIBRARY_PATH e.g.: export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/angstrom/arm/arm-angstrom-linux-gnueabi/usr/lib/ Sep 21 09:26:17 tasslehoff: yes, maybe janp_ suggestion to use LD_LIBRARY_PATH is the better solution in that case Sep 21 09:27:37 janp_: I've tried that. but it doesn't work for me. Which is odd... Sep 21 09:35:27 tasslehoff: which distribution is on your PC ? Sep 21 09:36:56 if ubuntu : https://bugs.launchpad.net/ubuntu/+bug/366728 Sep 21 09:55:33 tasslehoff: what you could also try is prepending your gcc invokation with LD_LIBRARY_PATH=/angstrom/arm/arm-angstrom-linux-gnueabi/usr/lib/ Sep 21 09:56:07 tasslehoff: e.g. LD_LIBRARY_PATH=/angstrom/arm/arm-angstrom-linux-gnueabi/usr/lib/ /angstrom/arm/libexec/gcc/arm-angstrom-linux-gnueabi/4.3.3/gcc Sep 21 10:12:09 janp_: hmm. could this be a pebkac as well... Sep 21 10:13:20 tasslehoff: pebkac? Sep 21 10:15:36 janp_: problem exists between keyboard and chair :) Sep 21 10:17:03 LD_LIBRARY_PATH=/home/tassle/src//toolchains/angstrom/arm/arm-angstrom-linux-gnueabi/usr/lib/ ./gcc -Werror -Wall -DLINUX -march=armv7-a -O2 -g /home/tassle/Desktop/main.c Sep 21 10:17:52 this fails. Sep 21 10:20:44 hi Sep 21 10:26:44 though all the libs I need are in the mentioned folder. Sep 21 10:28:23 tasslehoff: ./gcc would need to be done in the directory where the sdk is installed in the gcc path Sep 21 10:29:25 janp_: yes, I was in that dir, and it is the libmpfr.so.4-error I get Sep 21 10:30:30 janp_: sudo ln -s libmpfr.so.1 libmpfr.so.4 && sudo ln -s libgmp.so.3 libgmp.so.10 , makes it work for now, so if that's a safe move I'll do that for now. Sep 21 10:30:40 in /usr/lib, that is Sep 21 10:33:13 tasslehoff: I don't really recommend that, certainly not for a development environment for multiple developers Sep 21 10:33:51 janp_: understandable Sep 21 10:34:10 tasslehoff: if you export LD_LIBRARY_PATH and then run gcc with strace, will the loader search for the correct directory at least? Sep 21 10:38:47 I tried mdev from busybox to generate the /dev/mtdblock devices and that works, however udev still won't create them Sep 21 10:39:51 janp_: it seems to look in the directory, yes Sep 21 10:42:49 tasslehoff: what is the line from strace about the libmpfr.so.4 file? do you as your user have read access to it? Sep 21 10:47:33 janp_: http://pastebin.com/tstZah8t Sep 21 10:47:51 I have read access Sep 21 10:52:17 tasslehoff: could you try strace with the -f option because cc1 is a subprocess of gcc. just before the error message, there is a vfork, another process. strace -f will also trace any subprocesses. Sep 21 10:55:37 janp_: from vfork and down: http://pastebin.com/w4sWmdV0 Sep 21 11:17:41 http://marcin.juszkiewicz.com.pl/2010/09/21/i-am-not-a-filesystem-expert/ Sep 21 11:25:00 hrw, lol! Sep 21 11:25:45 lol but true actually :) Sep 21 11:31:24 it happens so often Sep 21 11:31:26 ... Sep 21 11:31:31 hehehe Sep 21 11:31:36 tasslehoff: the libmpfr.so.4 isn't the arm version right? file libmpfr.so.4 should tell you Sep 21 11:34:09 janp_: it is the arm-version.. Sep 21 11:38:41 tasslehoff: that's wrong, it should be the libmpfr.so.4 that is build from the native libmpfr.so.4 from oe not the arm version. cc1 indeed searches correctly but the library itself is wrong Sep 21 11:57:24 janp_: ah. Sep 21 12:00:24 gm Sep 21 12:11:37 gm Sep 21 12:19:41 tasslehoff: how did you generate the sdk? Sep 21 12:24:56 could someone be so kind and have a look at http://patchwork.openembedded.org/patch/2865/ ; since I am not the maintainer I need ack/nak and this completelly blocks the igep0020 machine Sep 21 12:27:52 mlip: done Sep 21 12:30:36 tasslehoff: what is your email to try a patch to fix this problem ? Sep 21 12:32:28 ericben: thx Sep 21 13:15:35 hi i am having problem using bitbake Sep 21 13:15:44 i am behind proxy Sep 21 13:15:50 awhat to do? Sep 21 13:16:29 Connecting to sources.openembedded.org|140.211.169.165|:80... failed: Connection timed out. Sep 21 13:16:38 this is the error i am getting every time Sep 21 13:16:53 anybody there?? Sep 21 13:17:31 ?? Sep 21 13:18:33 ???? Sep 21 13:18:51 03Martin Jansa  07org.openembedded.dev * r8887667384 10openembedded.git/recipes/e17/elmdentica_svn.bb: elmdentica: temporary bump SRCREV above EFL_SRCREV to fix build after 55b3912f14b1919bf2cd330e48546f11c179794f Sep 21 13:19:43 ?? Sep 21 13:20:52 03Paul Menzel  07org.openembedded.dev * r2592c12ae9 10openembedded.git/docs/usermanual/reference/var_src_uri.xml: Sep 21 13:20:52 docs/usermanual: s/are be copied/are copied/ Sep 21 13:20:52 Signed-off-by: Paul Menzel Sep 21 13:22:35 r0ck_: check these scripts http://gitorious.org/angstrom/angstrom-setup-scripts/trees/master they handle the proxy configuration Sep 21 13:22:45 ok Sep 21 13:27:38 ericben: pm'ed you the mail now. Sep 21 13:28:10 tasslehoff: ok I'm finishing the tests but it seems to work here Sep 21 13:29:40 ericben: cool. I gotta run now, but I'll test it tomorrow morning. Sep 21 13:34:58 tasslehoff: ok I'll post it to the list (but I think the fix is quite a big hack but this may bring opinions on how to fix this in a clean way) Sep 21 13:52:16 Hi, if I call in configure.ac $PKG_CONFIG --cflags, all parts are prefixed with proper SYSROOTS patch, but if I call ${PKG_CONFIG} --variable=includedir then it doesn't prefix it with SYSROOT, is there better way then fixing it in configure.ac by prefixing manually? Sep 21 14:04:50 console-image is building pango ... Sep 21 14:08:01 cbrake: right, the little-known ascii art backend Sep 21 14:10:35 how can I build angstrom distro with the latest codesourcery external toolchain? Sep 21 14:32:26 morning Sep 21 14:32:33 hi kergoth Sep 21 14:32:46 thanks for looking into that machine specific thing, appreciate it Sep 21 14:33:07 yes, diging into the code help to better understand how things work :) Sep 21 14:33:43 :) Sep 21 14:33:45 any opinion concerning the sdk patch serie I just sent ? Sep 21 14:34:12 ideally, it'd be smart enough to also know if any machine specific overrides were applied, or something -- we should be able to get that once bitbake gets metadata checksumming integrated and stamps go away Sep 21 14:55:11 I'm really digging this 'awesome' window manager Sep 21 14:55:14 reminds me a lot of ion3 Sep 21 14:59:43 yeah, one of my ex-colleagues was a big fan of awesome. Sep 21 15:00:54 though, whenever I used to go over his desk and ask him to show me something, the response was always "sure, let me just switch to that window... [ five minutes passes ] ... right, here we are, what did you want again?" Sep 21 15:01:28 ha Sep 21 15:01:48 hah Sep 21 15:16:06 I like using tiled workspaces for terminals with it, makes it very easy and fast to do work without distractions anywhere near the workspace Sep 21 15:18:13 damnit, why the hell is vim/gtk2 stupid about fonts. in other setups, setting an invalid font fails and it falls back to the next font in the line Sep 21 15:18:20 but here it just ends up displaying really oddly instead Sep 21 15:18:21 hmm Sep 21 15:18:27 wonder if its specific to my setup Sep 21 15:25:43 khem: ping? Sep 21 15:50:37 fraxinas: yes Sep 21 16:00:28 khem: 4.4.4-initial built, but intermediate breaks during compile Sep 21 16:00:48 it happens when compiling multilib which is strange already because it's configured to disable-multilib Sep 21 16:01:17 plus, it fails when trying to include std c lib files from a non-existing incdir and i dunno where it got that idea from either Sep 21 17:12:06 fraxinas: do you know when the branch was created Sep 21 17:17:44 not sure Sep 21 17:18:25 fraxinas: pastebin the errors Sep 21 17:19:58 stefan_schmidt: how did bitbake world go Sep 21 17:20:44 khem: had to blacklist some packages and are at 15000/70000 atm Sep 21 17:21:05 * stefan_schmidt checks the latest state Sep 21 17:21:41 stefan_schmidt: ok Sep 21 17:22:06 Actually I wonder if -k makes some downloads fial Sep 21 17:22:07 fail Sep 21 17:22:17 it should not Sep 21 17:22:21 I often have top clean up the downloaded tarball and fetch it again Sep 21 17:22:26 fetch is another task Sep 21 17:22:37 yeah, must be some other glitch I have there Sep 21 17:23:29 Running task 40348 of 80575 Sep 21 17:23:33 :) Sep 21 17:23:44 cool :) Sep 21 17:23:51 khem: what distro/machine? Sep 21 17:23:59 minimal/efikamx Sep 21 17:24:09 ok, need to keep that in mind Sep 21 17:24:25 angstrom and bug20 might be strange due to all this TI tools Sep 21 17:24:36 stefan_schmidt: yeah Sep 21 17:24:40 normally I don't use them at all but during world they all turn up Sep 21 17:24:44 manual fetching Sep 21 17:24:53 I see them too Sep 21 17:25:03 actually I hit some bugs too Sep 21 17:25:11 we have been fixing them slowly Sep 21 17:27:10 khem: yup, need to add some patches instead of blacklisting the packages. Its just easier right now. :) Time is a bit short right now Sep 21 17:27:42 stefan_schmidt: sure you can also keep bitbake -k going and collect errors Sep 21 17:28:08 khem: doing that as well Sep 21 17:30:12 I ran out of space Sep 21 17:30:24 but then put a new HD with 256G into laptop Sep 21 17:30:28 it was 100G Sep 21 17:30:39 even with rm_work I ran our of space Sep 21 17:30:44 I think due to pstage Sep 21 17:30:48 urgs Sep 21 17:31:04 should have enough space at the buildhost. At least I hope :) Sep 21 17:31:05 but I had 53000 tasks completed Sep 21 17:32:11 way better then what I have atm Sep 21 17:32:24 anyway, we will getting there. No rush needed :) Sep 21 17:32:36 yes its a back burner thing Sep 21 17:33:12 bugzilla is up please let me know if you see any issues Sep 21 17:33:27 khem: will do Sep 21 17:33:45 hi khem Sep 21 17:33:49 khem: biggest problem is to filter out the smaller issues or strange fetch problems I have over here Sep 21 17:34:17 hmm are you behind a firewall Sep 21 17:35:04 khem: not really. Its NAT'ed but all fetching works fine normally Sep 21 17:36:47 khem: do you think the way to solve the sdk's libs problem I proposed is correct ? http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024438.html Sep 21 17:38:33 ericben: I see no issues Sep 21 17:38:41 seems ok Sep 21 17:40:17 ok, except the PR increase in gcc,gmp & mpfr packages, is there anything else I should add? Sep 21 17:41:45 when I try yo use external toolchain csl with angstrom, TARGET_SYS seems to be wrong Sep 21 17:42:36 ericben: you will also need it in libmpc and libelf if you use gcc 4.5 Sep 21 17:43:10 actually it's gcc 4.4. Do you think I add libelf & libmpc to the gcc-cross-sdk ? Sep 21 17:43:21 you should Sep 21 17:43:33 actually try to generate the sdk with gcc 4.5 Sep 21 17:43:50 that covers a superset of libraries Sep 21 17:43:59 ok will try this but with angstrom any versions are more or less locked Sep 21 17:44:03 IOW if you get 4.5 right previous ones will be right too Sep 21 17:44:18 ericben: use minimal Sep 21 17:44:20 :) Sep 21 17:44:34 yes, I think I'll use minimal for testing gcc 4.5 Sep 21 17:44:51 in minimal you set the versions in your conf file or it automatically get 4.5 ? Sep 21 17:45:03 PREFERRED_GCC_VERSION_local = "4.5" Sep 21 17:45:11 in your local.conf will do it Sep 21 17:45:13 how does oe decide which kernel patches to download from kernel.org? Sep 21 17:45:15 what means the local override ? Sep 21 17:45:19 minimal uses 4.4.4 default Sep 21 17:45:29 it mean local.conf Sep 21 17:45:46 and its powerful Sep 21 17:45:53 it overrides others Sep 21 17:46:13 ah ok Sep 21 17:46:58 ok I see in conf/bitbake.conf it's the first one Sep 21 17:47:12 thanks, I'll try this later Sep 21 17:47:56 need to reboot this box brb Sep 21 17:48:52 btw if you have opinion on this other patch serie : http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-September/024422.html feel free to send comments :) Sep 21 17:49:11 diner time, bbl Sep 21 17:52:26 khem: http://pastebin.com/sqNDafJh Sep 21 17:58:57 khem: http://pastebin.com/sqNDafJh Sep 21 18:00:53 fraxinas: ok it did not build glibc-initial Sep 21 18:00:59 or staged it properly Sep 21 18:01:12 I would say sync back changed in glibc too Sep 21 18:01:17 or eglibc whichever you use Sep 21 18:02:16 ah Sep 21 18:02:18 okay Sep 21 18:02:25 that's possible! i cleaned glibc Sep 21 18:02:43 i forgot about it Sep 21 18:02:54 you should then do bitbake glibc-initial Sep 21 18:03:07 then it will stage the stuff needed for gcc-cross-intermediate Sep 21 18:03:20 aye! thanks Sep 21 19:53:03 is there a way to override rm_work for some specific package? Sep 21 20:13:41 khem: yay, works Sep 21 20:13:53 ok np Sep 21 20:35:04 khem: with your patch which ask gcc-sdk to use stati libmpfr, gmp & co, my patches for these libs are no more needed if I understand well ? Sep 21 20:37:41 ericben: yes Sep 21 20:38:08 ok I'm testing you patches and will update my patchset to the remaining necessary patches Sep 21 20:38:30 *sigh* gcc-cross-intermediate worked but now it fails with gcc-cross_4.4.4 Sep 21 20:39:08 /dream/oe1.6/dm8000/build/tmp/cross/mipsel/mipsel-oe-linux/bin/ld: /dream/oe1.6/dm8000/build/tmp/staging/mipsel-oe-linux/usr/lib/libc.a(abort.o): relocation R_MIPS_HI16 against `a local symbol' can not be used when making a shared object; recompile with -fPIC Sep 21 20:39:08 /dream/oe1.6/dm8000/build/tmp/staging/mipsel-oe-linux/usr/lib/libc.a: could not read symbols: Bad value Sep 21 20:40:01 ericben: gdb needs something similar to what khem did for mpfr & co I think Sep 21 20:41:14 Tartarus: you ok with that patch then I will push it Sep 21 20:41:26 just ack'd v2 Sep 21 20:43:41 oki doki Sep 21 20:43:51 * khem kicks fetchmail Sep 21 20:45:23 Tartarus : is adding ncurses-sdk and zlib-sdk in sdk's lib not a good thing ? Sep 21 20:46:27 it works fine this way (same for libmpfr & co in fact) Sep 21 20:46:36 It sucks for relocation Sep 21 20:46:44 Unless you want to investigate $ORIGIN mangling Sep 21 20:46:49 LD_LIBRARY_PATH sucks Sep 21 20:47:01 ah ok Sep 21 20:47:26 when you relocate the sdk, you only change the environment setup script and that works as is ? Sep 21 20:47:59 We don't rely on that script Sep 21 20:48:15 03Khem Raj  07master * r03ba2d4f86 10openembedded.git/recipes/ (11 files in 5 dirs): Sep 21 20:48:15 gcc-cross,gcc-cross-sdk: Link libmpfr, libgmp, libmpc and libelf statically into gcc Sep 21 20:48:15 * Generate static version of these libraries Sep 21 20:48:15 Signed-off-by: Khem Raj Sep 21 20:48:15 Acked-by: Tom Rini Sep 21 20:48:16 But yes, the SDK directory can move around wherever Sep 21 20:48:17 03Khem Raj  07master * rf990853e39 10openembedded.git/recipes/iphone/ (4 files): Sep 21 20:48:17 iphone: Mark recipes with COMPATIBLE_MACHINE=iphone Sep 21 20:48:17 * Remove EXCLUDE_FROM_WORLD Sep 21 20:48:18 Signed-off-by: Khem Raj Sep 21 20:48:21 03Khem Raj  07master * r5486dcbf08 10openembedded.git/recipes/corosync/ (corosync_1.2.7.bb files/corosync-docs.patch): Sep 21 20:48:21 corosync: Fix compilation Sep 21 20:48:21 Update to 1.2.8 Sep 21 20:48:21 Add corosync-docs.patch to fix html doc compilation Sep 21 20:48:21 Signed-off-by: Khem Raj Sep 21 20:48:22 03Khem Raj  07master * rd83521d8d0 10openembedded.git/recipes/groff/ (groff_1.19.2.bb groff_1.20.1.bb): Sep 21 20:48:22 groff_1.20.1.bb: Add recipe. Sep 21 20:48:23 Add groff-native with BBCLASSEXTEND Sep 21 20:48:23 Signed-off-by: Khem Raj Sep 21 20:48:25 03Graham Gower  07master * r9aad9078dc 10openembedded.git/recipes/ (8 files in 3 dirs): (log message trimmed) Sep 21 20:48:25 tapioca, farsight: move to obsolete. These recipes are old and unbuildable. Sep 21 20:48:32 ok there is also libopkg.so and libz.so in the sdk's lib dir Sep 21 20:49:25 Yes Sep 21 20:49:34 We also don't care about opkg :) Sep 21 20:49:40 But all of that stuff should be origin mangled Sep 21 20:49:49 cc & co is hrader due to funky burried paths Sep 21 20:50:03 (libexec/.../../cc1) Sep 21 20:51:33 by origin mangled you mean the base directory can change ? Sep 21 20:52:38 what package manager are you using ? deb or no package at all ? Sep 21 20:58:58 We install to the sdk with opkg Sep 21 20:59:04 But no, no pkg mgr in the SDK Sep 21 21:00:08 ok this makes sens as packaging is done in oe Sep 21 21:00:23 will see this tomorrow Sep 21 21:00:31 bye Sep 21 21:53:35 hi, im doing the getting started oe tutorial Sep 21 21:53:50 i am doing the bitbake nano part Sep 21 21:53:55 at one line it says Sep 21 21:53:57 cp: cannot stat `/stuff/build/tmp/work/i686-linux/stagemanager-native-0.0.1-r13/sysroot-destdir//stuff/build/tmp/*': No such file or directory Sep 21 21:54:21 anybody knows what conf line did i missed? Sep 21 21:54:35 miss* Sep 21 21:54:43 aafuentes: do you have prerequisites Sep 21 21:54:47 on your distro Sep 21 21:54:54 installed Sep 21 21:55:08 yep Sep 21 21:55:25 whats your distro Sep 21 21:55:29 ubuntu Sep 21 21:55:47 it seems its trying to join to paths badly Sep 21 21:56:20 with those two bars // Sep 21 21:56:34 what does your TMPDIR look like Sep 21 21:56:39 in local.conf Sep 21 21:56:44 does it have a traling / Sep 21 21:58:25 i dont have tmpdir defined in my local.conf Sep 21 21:58:43 what does your local.conf look like Sep 21 21:58:48 can you paste it somewhere Sep 21 21:59:11 it created a tmp on the folder i lauched bitbake tho Sep 21 21:59:27 yep Sep 21 22:01:27 http://pastebin.com/9ni6CdCT Sep 21 22:02:25 it gave that error, but it seems the recipe continues to do its stuff :S Sep 21 22:15:44 khem, u there? Sep 21 22:16:02 yes Sep 21 22:17:21 define TMPDIR = "/stuff/tmp" Sep 21 22:17:44 thanks khem ;) Sep 21 22:19:03 another question... khem. Is there a way to distribute compile openembedded recipes? it seems the builds takes forever Sep 21 22:19:41 packaged staging is there Sep 21 22:20:08 i have a farm of computers... is there a way to use them? Sep 21 22:20:24 yes icecc Sep 21 22:20:26 or distcc Sep 21 22:21:15 mmm... ive used distcc with gnu projects in the past Sep 21 22:21:35 ill take a look at it with bitbake... thanks khem ;) Sep 21 22:21:48 sure Sep 21 22:21:57 you can ask any questions you have here Sep 21 22:22:41 ill dig a little deeper in bitbake... and then come back with lots of questions Sep 21 22:22:46 thanks ;D Sep 21 22:24:35 man, the do_fetch should be parallelized lol Sep 21 22:24:53 i think i can go to bed and come back tomorrow Sep 21 22:29:42 is there a way I can add something to a package's postinst without overriding the recipe? a 'foo_pkg_postinst' defined outside of recipe doesn't override foo's pkg_postinst Sep 21 22:34:11 foo_pkg_postinst_append in your amend.inc or local.conf or whatever outside the recipe you're modifying? Sep 21 22:40:21 Tartarus, thats what I tried but it didn't seem to take - should that be valid? Sep 21 22:40:52 are you seeing it not append or not exist? Sep 21 22:41:03 and where are you doing it from/ Sep 21 22:42:15 was doing it from local.conf was not seeing it exist but I'll check again if that syntax should work Sep 21 22:45:20 amend.inc is how we do stuff Sep 21 22:45:35 so, same toplevel as you have conf/local.conf, recipes/foo/amend.inc Sep 21 22:46:31 Tartarus, ah... thats useful, I wasn't aware of amend.bbclass - thx! Sep 21 22:46:40 tharvey: from local.conf, you'd want pkg_postinst_whatever_append_pn-somerecipe, most likely Sep 21 22:46:51 or just _pn-somerecipe or _local rather than append Sep 21 22:47:07 using an override will ensure it overwrites the recipe's contents, but the pn- one will only apply to that one recipe, which is probably what you want Sep 21 22:47:34 ah... ok Sep 21 23:10:01 hi , when i try to compile sgx with "bitbake omap3-sgx-modules" , i got error on_compile bc_cat.c:490: error: implicit declaration of function 'omap_rev_lt_3_0' , how to fix this problem ? any solution ? Sep 21 23:19:05 ugh Sep 21 23:19:10 * kergoth stabs bitbake a few times Sep 21 23:20:39 kergoth: thrice convicted you go to jail for life :) Sep 21 23:22:39 keep finding things that annoy me Sep 21 23:23:09 kergoth: btw I was reading the post from RP about varref Sep 21 23:23:19 kergoth: is it in poky now Sep 21 23:37:00 hmm... its no longer obvious to me why bitbake -k world has halted Sep 21 23:37:14 grg: same issue here Sep 21 23:37:18 :) Sep 21 23:37:23 fun Sep 21 23:37:26 it quit without any notice Sep 21 23:37:42 last one i got was Running task 58315 of 80464 Sep 21 23:37:45 NOTE: Running task 47837 of 80575 (ID: 46, /home/kraj/work/oe/openembedded/recipes/libaio/libaio-oracle_0.3.0.bb, do_rm_work_all) Sep 21 23:37:48 NOTE: package libaio-oracle-0.3.0-r1: task do_rm_work_all: Started Sep 21 23:37:51 NOTE: package libaio-oracle-0.3.0-r1: task do_rm_work_all: Succeeded Sep 21 23:39:43 libaio-oracle failed for me in do_qa_staging, due to imlib.pc Sep 21 23:41:00 but it is right at the end of the log Sep 21 23:41:11 (well, not quite the last thing) Sep 21 23:42:54 there seems to be some issues with packages used in meta-toolchain: NOTE: gcc contains dangling symlink to /lib/libgcc_s.so.1 Sep 21 23:42:54 NOTE: g77-symlinks contains dangling symlink to /usr/bin/i486-angstrom-linux-g77 Sep 21 23:42:55 NOTE: gfortran-symlinks contains dangling symlink to /usr/bin/i486-angstrom-linux-gfortran Sep 21 23:42:55 NOTE: libstdc++-dev contains dangling symlink to /usr/lib/libstdc++.so.6.0.10 Sep 21 23:42:55 ERROR: QA Issue with gcc: non -dev package contains symlink .so: gcc path '/work/i486-angstrom-linux/gcc-4.3.3-r17.1/packages-split/gcc/usr/lib/gcc/i486-angstrom-linux/4.3.3/libgcc_s.so' Sep 21 23:42:56 if I restart then it executes rm_work and staging for all Sep 21 23:43:03 not very helpful Sep 21 23:45:41 | /mnt/oe/tmp/sysroots/mipsel-oe-linux/usr/lib/libbfd.a: could not read symbols: Bad value Sep 21 23:45:51 i get a couple of those, which don't look good Sep 21 23:48:04 hey. is it possible to specify an overlay dir for conf/machine like for the recipes folder? Sep 21 23:48:32 CMoH-notebook, no you can overlay that as well Sep 21 23:48:38 grg: huh can you dump /mnt/oe/tmp/sysroots/mipsel-oe-linux/usr/lib/libbfd.a Sep 21 23:48:55 tharvey: those qa issues are with gcc cross or target gcc Sep 21 23:48:56 tharvey, sorry, i don't understand what you mean Sep 21 23:49:28 CMoH-notebook: overlay is mirror copy of oe Sep 21 23:49:31 is it "no, you can't overlay..." or "no, you can overlay" Sep 21 23:49:35 CMoH-notebook, your trying to create a collection that adds new recipes and machine confs? thats common Sep 21 23:49:39 you can overlay Sep 21 23:49:39 think of oe metadata also as an overlay Sep 21 23:50:12 so you can have a, b, openembedded which will be three overlays Sep 21 23:50:14 khem, I get this (which does not appear to be fatal to bitbake world): | /mnt/oe/tmp/sysroots/x86_64-linux/usr/mipsel/lib/gcc/mipsel-oe-linux/4.4.4/../../../../mipsel-oe-linux/bin/ld: /mnt/oe/tmp/sysroots/mipsel-oe-linux/usr/lib/libbfd.a(archures.o): relocation R_MIPS_26 against `a local symbol' can not be used when making a shared object; recompile with -fPIC Sep 21 23:50:30 and each overlay looks like openembedded Sep 21 23:50:44 hmm, i think of an overlay as where specified in BBRECIPES="$OEBASE/openembedded/recipes/*/*.bb", which basically specifies a glob for files. Sep 21 23:50:54 how can i define an overlay for the whole openembedded? Sep 21 23:51:24 khem, right, QA issues with gcc, are people just ignoring them? my build script sees the error and craps out Sep 21 23:51:29 grg: issue is that someone installed into /mnt/oe/tmp/sysroots/mipsel-oe-linux/usr/lib/libbfd.a which is a non pic version Sep 21 23:51:47 khem, tharvey, i already have the folders having the same structure as oe Sep 21 23:51:52 works fine on some arches but arches like mips x86_64 or arm will cry Sep 21 23:52:07 I think this base_do_unpack error is more likely the culprit: http://pastebin.com/jt94DSdi Sep 21 23:52:14 CMoH-notebook, some examples here: http://elinux.org/BeagleBoardOpenEmbeddedDevelopment Sep 21 23:53:41 tharvey, khem, thanks - i'll have to read these Sep 21 23:53:45 CMoH-notebook, and in the Getting_Started guide: http://wiki.openembedded.net/index.php/Getting_Started - specifically there, see how BBPATH is setup to reference /stuff/build first then /stuff/openembedded Sep 21 23:54:29 CMoH-notebook: http://sakrah.homelinux.org/node/2 Sep 21 23:54:32 the beagleboard setup is more complicated and shows how you would prioritize collections (overlays) Sep 21 23:54:48 that will help with latest OE technology of bblayers and bbappend Sep 21 23:55:34 khem, ah... bblayers and bbappend look nice - seems like there are now so many different ways to overlay/override/append heh Sep 21 23:55:52 tharvey: the one I blogged works for me Sep 21 23:55:57 and is something I like Sep 21 23:56:35 why would someone put SRC_URI += "foo", and then put foo in files/$TARGET_MACHINE/ ? Sep 21 23:56:46 just breaking the package for everyone else? Sep 21 23:57:25 eh, yeah - now i have to choose between 3 solutions, which makes it a bit harder :D Sep 21 23:57:33 grg: that will make the package machine specific Sep 21 23:57:40 anyway, it's great to have these at hand Sep 21 23:58:04 so basically bitbake reads all conf/*.conf files within the $PWD ? Sep 21 23:58:26 if i put conf/site.conf near conf/local.conf they'll both be parsed Sep 21 23:58:30 CMoH-notebook, no, it uses BBPATH Sep 21 23:59:20 ah, ok; but it does glob if i understand the guides correctly Sep 22 00:01:45 it globs but it uses BBPATH not PWD Sep 22 00:02:04 khem, about the QA Issue with gcc - that causes bitbake to error out - am I missing something? Sep 22 00:03:05 no Sep 22 00:03:15 I think you have QA strict Sep 22 00:04:27 warning as errors you mean? looking for where that would be defined Sep 22 00:04:31 yeah Sep 22 00:04:42 whats in your INHERIT Sep 22 00:05:42 yup, bblayers are really cool as well as bbappend Sep 22 00:06:01 are they stable? Sep 22 00:06:22 my INHERIT is package_ipk, debian, sanity, recip_sanity, devshell, insane Sep 22 00:06:24 CMoH-notebook: yes Sep 22 00:06:36 you saying most people don't inherit insane? Sep 22 00:07:17 tharvey: actually that error makes sense Sep 22 00:07:22 but you can live by Sep 22 00:07:55 its just a QA check so the packages build, but it causes bitbake to return an error code that my build script catches Sep 22 00:08:12 so I guess I should disable insane for those packages or fix the packages Sep 22 00:08:25 was just surprised to find them fail the QA check like that Sep 22 00:08:25 yeah Sep 22 00:08:49 QA checks are nice to err on Sep 22 00:24:48 03Khem Raj  07master * r466148c26e 10openembedded.git/recipes/chumby/chumby-firmware_1.2.bb: Sep 22 00:24:48 chumby-firmware_1.2.bb: Fix typo in COMPATIBLE_MACHINE Sep 22 00:24:48 Signed-off-by: Khem Raj Sep 22 00:27:51 khem, python-numpy needs a COMPATIBLE_ARCH=arm, but it doesn't look like it exists. Any hints? Sep 22 00:31:59 and firefox is compatible with i{3,4,5,6}86,powerpc,arm Sep 22 00:33:02 grg: there is something called COMPATIBLE_HOST Sep 22 00:33:36 you could do COMPATIBLE_HOST = "arm.*" for numpy Sep 22 00:33:56 khem, cool Sep 22 00:34:37 and COMPATIBLE_HOST = '(powerpc|arm|i.86.*)' for firefox Sep 22 00:35:27 COMPATIBLE_HOST = '(powerpc|arm|i.86).*' Sep 22 00:35:32 seems to be right Sep 22 00:42:03 numpy is another recipe that has been broken recently Sep 22 00:42:58 on some machines Sep 22 00:44:15 rt Sep 22 00:44:16 er Sep 22 00:44:30 grab the mips patches I can only assume exist? :) Sep 22 00:44:32 for firefox Sep 22 00:44:53 to me, compatible and "happens to build right now" aren't the same thing Sep 22 00:45:10 stuff that's very specific to a given HW platform due to being ties to the HW makes sense Sep 22 00:46:42 well, the fixes i just sent are for fixing do_unpack problems due to machine/arch specific patch dirs Sep 22 00:47:10 Yeah Sep 22 00:47:25 still, what I'm saying applies. Someone else can argue the other direction Sep 22 00:47:38 But to me, compatible* is a heavy tool Sep 22 00:48:01 and means "never going to be work outside of ..." Sep 22 00:48:17 i agree with what you're saying, but there isn't really a good alternative. i don't have an interest in fixing every package for my arch Sep 22 00:48:35 i have no plans to run firefox on a 300mhz mipsel, for instance Sep 22 00:48:47 So, why did you hit this then? Sep 22 00:48:50 bitbake world? Sep 22 00:48:53 yup Sep 22 00:48:58 So don't do that. Sep 22 00:49:04 i want bitbake world as a regression test Sep 22 00:49:17 it will be a very powerful thing Sep 22 00:49:52 in what way? Sep 22 00:49:56 Tartarus: I think its fine if some package can work but does not work yet on a given arch Sep 22 00:50:04 its better to state that clearly Sep 22 00:50:18 khem: That's fine too, but that's not what COMPATIBLE* means to me anyhow Sep 22 00:50:25 well if i do a bitbake world on the testing branch each week then its rather obvious to see what has broken in the past week Sep 22 00:50:38 if some mips folks need firefox they can fix it and take it out Sep 22 00:50:43 its just casting a wider net for testing Sep 22 00:50:53 build testing Sep 22 00:50:56 Tartarus: that I would agree Sep 22 00:50:57 not usable testing Sep 22 00:51:05 its a start Sep 22 00:51:20 Tartarus: may be there could be some other var which says on what arch it works Sep 22 00:51:31 i cant test usability without buildability. and buildability uncovers large numbers of problems already Sep 22 00:51:41 semantically it will be exactly same as comaptible_host Sep 22 00:51:49 grg, right Sep 22 00:52:15 So I'd say you've run into software that should be fixed to compile Sep 22 00:52:19 Or proven it can't be Sep 22 00:52:32 world will build a lot of stuff you'll never build on your 300mhz mipsel Sep 22 00:52:44 But otoh maybe you would try fennec Sep 22 00:52:45 i have finite resources. i cant get distracted by everything. Sep 22 00:52:48 So you'd need to fix firefox anyhow Sep 22 00:53:06 grg, So fix firefox as it's a rather easy one to get past your current problem Sep 22 00:53:25 * Tartarus did it before, for mips Sep 22 00:53:45 ideally the approach should be to fix the package Sep 22 00:53:56 is such cases Sep 22 00:54:06 Tartarus, well if i do that, i should then add mips.* to compatible_host. Because sh4,blackfin,etc is not supported Sep 22 00:54:25 So fix sh4 too Sep 22 00:54:31 And maybe you do want to COMPATIBLE_OS it Sep 22 00:54:39 Of course everything just about might need that :) Sep 22 00:54:45 :) Sep 22 00:55:43 * Tartarus wonders if jsautoconf.h could be generated with qemu-$arch Sep 22 00:56:47 I guess what I'm trying to say is that if you think bitbake world should build something useful, I don't like the idea of saying "oh, we masked out N recipes too rather than port them", without some plan to get back to actually porting them Sep 22 00:57:02 You can argue that maybe nios2 is never going to be "fast" enough Sep 22 00:57:05 But not MIPS Sep 22 00:57:45 I understand your point, but i'm not going to fix every error i come across Sep 22 00:58:13 So please come up with a plan to fix them later or try and get someone else to fix them later Sep 22 00:59:20 Debian for example looks to have that file for mipsel already Sep 22 00:59:43 http://packages.debian.org/hu/lenny/mipsel/libmozjs-dev/filelist Sep 22 01:00:10 03Cliff Brake  07master * rf20e70e0a4 10openembedded.git/recipes/nodejs/nodejs_0.2.1.bb: nodejs: add native support Sep 22 01:00:26 (So I guess I'm trying to say if we want a model where bitbake world works, we should try for the debian version of everything builds everywhere as much as is possible) Sep 22 01:00:52 hi , when i try to compile sgx with "bitbake omap3-sgx-modules" , i got error on_compile bc_cat.c:490: error: implicit declaration of function 'omap_rev_lt_3_0' , how to fix this problem ? any solution ? Sep 22 01:03:10 do bbappend recipes only work within a bblayer? or is it parsed when simply placed in BBFILES? Sep 22 01:04:11 it doesn't have to be a layer, the two pieces of functionality are independent, it uses the filename alone to determine which are appended to which, not the directory structure Sep 22 01:04:51 hmm... when was it implemented? i have the git from 2010-09-08 Sep 22 01:05:12 git log --grep=append Sep 22 01:05:15 is your friend Sep 22 01:05:19 kk :) Sep 22 01:07:11 yeah, i don't have it Sep 22 01:08:09 its in master, not 1.10 Sep 22 01:08:12 just so you know Sep 22 01:08:54 ty - i'll fix things **** ENDING LOGGING AT Wed Sep 22 02:59:57 2010