**** BEGIN LOGGING AT Tue Aug 27 02:59:58 2013 Aug 27 05:05:58 hi Aug 27 05:06:13 i have build wayland-xcomposite plugin Aug 27 05:06:39 when i run my apllication it is throwing "XComposite required" Aug 27 05:06:52 im trying to get GUI from putty (ssh) Aug 27 05:06:59 any help in this regarding please Aug 27 05:08:06 in normal standalone system i'm able to get GUI but through putty ssh protocol i'm not getting..the output Aug 27 05:56:18 Morning. Any good advice on how to debug this building error on libgles-omap3? http://pastebin.com/bCDEJJ0B Aug 27 06:01:54 seems I'm not the only one http://comments.gmane.org/gmane.linux.embedded.yocto.meta-ti/2434 Aug 27 06:14:42 The command it struggles with runs without errors in devshell Aug 27 08:48:03 iam having issues not to be able to use this SRC_URI = "hg://code.google.com/p/picolisp;module=default;rev=${SRCREV};proto=https" .. Aug 27 08:48:09 what do i do wrong here ? Aug 27 08:49:25 rob_w: why do you need rev=${SRCREV} ? Aug 27 08:49:35 rob_w: shouldn't the hg fetcher be taking care of that automatically? Aug 27 08:49:50 to pull a certain version ? Aug 27 08:50:36 hi, is it normal that OECONF has "--prefix=/usr" in it? shouldn't it be pointing to my sysroot instead of the hosts "/usr"? Aug 27 08:53:28 rob_w: yes I can see that, but setting SRCREV should be enough Aug 27 08:53:43 ah Aug 27 08:54:08 thesignal: no, prefix is not the same as sysroot; that prefix value is the expected default Aug 27 08:55:36 bluelightning: ok. because in my case, octave is trying to install to "prefix" which it shouldn't, also, during compilation it's looking for files in /usr/include instdead of the sysroot i guess Aug 27 08:56:07 thesignal: right, in which case the sysroot isn't being passed through correctly Aug 27 08:56:17 thesignal: does octave use autotools? Aug 27 08:56:46 bluelightning: jep Aug 27 08:57:24 thesignal: are you specifying your own do_configure or just using the default one provided by "inherit autotools" ? Aug 27 08:58:07 bluelightning i am new to hg .. can you tell me the correc SRC_URI of that project ..?? i dont seem to get it Aug 27 08:58:07 bluelightning: i have specified my own do_configure Aug 27 08:58:45 thesignal: hmm, you may wish to try commenting that out (and specify any needed additional options in EXTRA_OECONF) Aug 27 08:58:58 bluelightning: i used an old recipe for octave as a template, i haven't thought of it yet. my do_confiure does "gnu-configize and oe_runconf Aug 27 08:59:05 nobody else having issues with libgles-omap3? or am I the only one still building this..? Aug 27 08:59:22 bluelightning: thanks i'll try commenting it out Aug 27 09:00:03 thesignal: right, I just saw that in the OE-Classic octave.inc; it's likely that our autotools support has improved since those days Aug 27 09:01:42 rob_w: I don't have any experience with hg unfortunately Aug 27 09:03:40 rob_w: it looks OK based on other examples I have seen FWIW Aug 27 09:04:05 the module= parameter screws me up Aug 27 09:04:16 nvm i try n error it out Aug 27 09:11:18 bluelightning: it seems like it worked :) at least i get no errors during compilation Aug 27 09:11:28 bluelightning: thanks a lot Aug 27 09:11:30 thesignal: great! :) Aug 27 09:11:53 thesignal: good that an old ugly recipe hack we had is no longer necessary :) Aug 27 09:12:00 bluelightning: xD Aug 27 09:12:51 bluelightning: but i stell have an issue: it tells me "files/directories were installed but not shipped". how does the default do_install look like? Aug 27 09:13:21 thesignal: it basically calls make install as generated by automake Aug 27 09:13:43 (when inheriting autotools, that is) Aug 27 09:14:22 bluelightning: and where does it usally put the binaries etc, which will be packaged later? in "image" right? Aug 27 09:14:57 yes, D (where everything is installed under in do_install) defaults to ${WORKDIR}/image Aug 27 09:15:58 bluelightning: so overwrting do_install shouldn't be necessary? Aug 27 09:16:19 in theory, no Aug 27 09:17:27 ok. then i'll try to stick with the default do_install before changing anything Aug 27 09:17:39 (why haven't i thought earlier of this :P ) Aug 27 09:28:15 hm.. bluelightning: i now get Issue: non -dev/-dbg/-nativesdk package contains symlink .so: octave path '/work/armv7ahf-vfp-angstrom-linux-gnueabi/octave-3.6.4-r1/packages-split/octave/usr/lib/octave/3.6.4/libcruft.so' Aug 27 09:28:24 for multiple packages as error Aug 27 09:28:54 thesignal: right, that .so symlink needs to go into the -dev package Aug 27 09:28:55 and also some "non debug package contains .debug directory Aug 27 09:29:10 similarly that .debug dir needs to go into the -dbg package Aug 27 09:34:20 bluelightning: is it necessary to describe what goes into a package with FILES_libcruft for example ? or is openembedded clever enough to figure this out? Aug 27 09:38:14 thesignal: there are defaults for all the packages that appear in the default value of PACKAGES (see PACKAGES and FILES_* definitions in meta/conf/bitbake.conf) Aug 27 09:38:37 thesignal: for packages you add yourself you will need to specify your own FILES_ value as appropriate Aug 27 09:40:48 bluelightning: thanks, i'll dig into it Aug 27 10:05:10 hi all Aug 27 11:13:22 pb_: hi Aug 27 11:14:57 pb_: it seems many of the default vars do not expect collapsing /usr in / Aug 27 11:15:36 morning pb_, ant_work Aug 27 11:16:23 hi bluelightning Aug 27 11:16:33 ant_work: which ones? I think most of them are ok nowadays. Aug 27 11:17:20 bluelightning: i got rid of all the package errors except one, and i don't know what to do: "QA Issue: octave rdepends on libcruft.dev" (and some similar) i haven't defined the rdepens for octave anywhere, also i have libruft + what goes inside it defined in my recipe Aug 27 11:17:25 any ideas? Aug 27 11:18:14 you've probably ended up with a shared library in libcruft-dev when it should be in libcruft. Aug 27 11:18:15 thesignal: usually that indicates that you have packaged something you shouldn't have in the -dev package, e.g. a shared library that one of the executables depends upon rather than just the .so symlink Aug 27 11:22:43 hm... if i don't put libcruft (in this example) into the libcruft-dev, i recive an error about an "non -dev/-dbg package containing a symlink .so, but the octave rdepends for this case disappears Aug 27 11:23:46 sry, i meant libcruft.so* into FILES_libcruft-dev Aug 27 11:24:20 so, that's your error. you should put libcruft.so (only) into libcruft-dev, and libcruft.so.(everything else) into libcruft. Aug 27 11:25:11 or you can write those as "libcruft${SOLIBSDEV}" and "libcruft${SOLIBS}", which might make it more obvious what's happening. Aug 27 11:25:48 ... but, if you just leave FILES alone and use the defaults from bitbake.conf then everything should probably "just work" unless libcruft is particularly unusual in some way. Aug 27 11:28:06 libgles-omap3: debugedit fails in do_package, but the command works fine when I run it manually, both in devshell and in a normal shell. any ideas how I can debug this? Aug 27 11:28:50 what's the error you get during do_package? Aug 27 11:29:10 i already tried it, leaving FILES alone, but that didn't work. i put everything as you suggested and now it builds the packages without any errors :) Aug 27 11:29:41 bluelightning & pb_: thank you guys, finally i figured it out Aug 27 11:31:06 pb_: it's about the debian split of klibc: libklibc goes into /lib and headers under /usr/lib/klibc normally Aug 27 11:31:26 pb_: http://pastebin.com/bCDEJJ0B Aug 27 11:40:33 that looks like https://bugzilla.redhat.com/show_bug.cgi?id=304121, though if that was the problem you'd expect it to happen when running by hand as well. Aug 27 11:42:32 Is there a way to make linux try a second rootfs if the first becomes corrupt or unmountable? Aug 27 11:42:47 sure, your initramfs can do whatever it likes. Aug 27 11:43:08 mount /dev/firstroot || mount /dev/secondroot ... Aug 27 11:44:27 Hmm yes, this is a cleaner solution, long as the initrd says intact Aug 27 11:44:29 thanks Aug 27 11:46:54 well, the initramfs is (typically) part of your kernel, and if the kernel gets damaged then you have a different set of problems to worry about. Aug 27 11:47:12 obviously you can store a primary and fallback kernel as well, but then your bootloader needs to be able to figure out that the kernel is broken and fall back to the secondary one. Aug 27 11:51:08 pb_: yeah. found a post on meta-ti about it, so I'm not the only one Aug 27 11:51:50 hello. is there a way to disable rm_work for a specific set of packages? Aug 27 11:52:06 in other words, let OE do rm_work, except for certain packages Aug 27 11:52:18 yes, do set an empty task in the recipe Aug 27 11:52:19 dv_: define rm_work() { \n } in recipe Aug 27 11:52:31 hi hrw Aug 27 11:52:36 hmm. any way to do that without touching the recipe? Aug 27 11:52:42 ideally something I set in the local.conf Aug 27 11:54:23 yes, you can set RM_WORK_EXCLUDE Aug 27 11:55:39 oh, yes Aug 27 11:55:54 nice, that should do. thanks Aug 27 11:55:58 rm_work.bbclass ;) as always the code can talk Aug 27 12:22:45 tasslehoff: ok, replying again in less hostile environment here (especially since woglinde is out) - that is a known issue of debugedit, we are currently looking for a workaround Aug 27 12:24:04 denix: I sensed there was some tension :/ Aug 27 12:25:05 if this is the double-slash thing, can't you just fix your makefiles to use fewer slashes? Aug 27 12:25:07 tasslehoff: just plain NIH syndrome Aug 27 12:25:18 or more slashes, come to that :-) Aug 27 12:25:29 pb_: there are no obvious double slashes Aug 27 12:25:36 denix: I actually managed to build by removing x11_prog from the recipe Aug 27 12:25:46 I don't use X anyway, and those were giving the issues Aug 27 12:26:14 tasslehoff: yeah, I need to rework the X/Raw split better - it's on my list Aug 27 12:27:06 denix: think this will work for me. this is (hopefully) a last release on an omap3530-based platform. Aug 27 12:27:18 let me know if you want me to test something on my setup Aug 27 12:27:22 denix: in the output from readelf, you mean? Aug 27 12:27:54 tasslehoff: as for #beagle - there's only one god there and he's name is koen, if you are not following his word, you are a heretic! :) Aug 27 12:28:22 pb_: in the error message Aug 27 12:29:39 denix: right, but (assuming we are talking about the same debugedit problem that tasslehoff had earlier) the error message doesn't actually mention the problematic string; you'd need to inspect the eglinfo_x binary to see what's wrong with it. Aug 27 12:30:11 pb_: ah, good point. let me look into that. Aug 27 12:30:12 :) Aug 27 12:30:21 gotta run for now, be back later Aug 27 12:30:25 yeah, that was the tip I saw on some fedora bugs as well. Aug 27 12:31:05 "readelf -wl .../eglinfo_x" is probably what you need, look for lines containing "//" Aug 27 12:32:12 pb_: yup. there was a culprit there Aug 27 12:32:25 denix: /home1/prathap/imgtiddk16_4117/ti_references/sources/GFX_Linux_DDK/src/eurasia/eurasiacon/binary_omap3430_linux_debug//usr/local/XSGX/include/X11 Aug 27 12:33:53 pb_, tasslehoff: thanks! Aug 27 12:43:39 thesignal: great stuff... if you wouldn't mind, it would be really awesome if you could send a patch to add your newly refreshed recipe to meta-oe ;) Aug 27 12:43:41 * mckoan hates jet-lag side effects :-( Aug 27 12:43:57 good morning all Aug 27 12:44:40 bluelightning: i already put it on github, but it's not clean, enough i think Aug 27 12:44:49 i'm not using shared libraries Aug 27 12:45:13 bluelightning: but you can take a look: https://github.com/thesignal/octave-openembedded-recipe Aug 27 12:48:12 thesignal: the --with-*= including references to ${D} look a bit odd Aug 27 12:48:57 thesignal: also RDEPENDS should be set with a package override (e.g. ${PN} if it's the main package that needs the dependency) Aug 27 12:50:21 ok, so i'll change RDEPENDS to? = ${PN} and libfftw3 ? Aug 27 12:51:03 bluelightning: i'll delete the --with part, as it is using static libraries anyway, one octave developer already complained Aug 27 13:00:28 thesignal: RDEPENDS_${PN} = "..." Aug 27 13:02:39 bluelightning: updated, thanks for the hints Aug 27 14:53:42 JaMa: re the LAMP series; should I pull it all in or just the meta-webserver parts and then you pull in the meta-oe mysql -> mariadb change (assuming it's OK)? Aug 27 15:00:11 bluelightning: looks ok, feel free to pull them all Aug 27 15:01:15 JaMa: ok thanks Aug 27 15:21:30 JaMa: log.world.20130826_032659.log is definitely ok Aug 27 15:25:10 ant_work: yes, looks good, thanks a lot for finding that bug Aug 27 15:25:30 rather horrible bug Aug 27 15:25:34 * ant_work hides Aug 27 15:56:23 JaMa: about navit failures, maybe worth to try fresh svn and use cmake Aug 27 16:25:57 hi. i am trying to understand xxx_machine=, e.g. i think it's called machines overrides. so is that correct that for *any* variable XYZ, we can set XYZ_machine="foo", and XYZ will be set to "foo" only if building for 'machine'. do i get that right? Aug 27 16:26:40 ndec: correct Aug 27 16:27:16 and what does something like that mean in a .bbappend: Aug 27 16:27:16 COMPATIBLE_MACHINE_palmpre = "palmpre" Aug 27 16:27:48 ndec: if using overrides such as these within a recipe, you usually need to be careful about ensuring PACKAGE_ARCH is set to "${MACHINE_ARCH}" so that the package is correctly marked as being machine-specific Aug 27 16:28:16 ndec: assuming you want to be able to build for multiple machines in the same build dir that is Aug 27 16:28:25 or you maintain package feeds Aug 27 16:28:26 yes, that's indeed my problem Aug 27 16:29:02 that's just ensuring COMPATIBLE_MACHINE is valid for that machine... it does look a bit odd to override it for the one machine but the idiom works Aug 27 16:29:14 hmm, well, no in fact. i don't want to build multiple machines in same , but i have multiple machines/BSP layer that are parsed, but 1 only being used. Aug 27 16:29:39 it's because otherwise you need to make sure that appending to regexp still produces valid regexp Aug 27 16:29:45 the .bbappend is from ./meta-palm/recipes-bsp/x-load/x-load_git.bbappend Aug 27 16:29:47 using COMPATIBLE_MACHINE_palmpre is easier Aug 27 16:30:05 and because palmpre is also in MACHINEOVERRIDES it matches also MACHINE palmpre2 Aug 27 16:33:50 ok... it's still a bit confusing... but i am getting there, slowly. Aug 27 16:35:16 and if i have several 'similar' machines, for which i need to set variables to the same value, i need to dupplicate the XYZ_machine lines i guess, right? Aug 27 16:37:41 you can define common MACHINEOVERRIDE for them Aug 27 16:37:51 see how MACHINE_CLASS or SOC_FAMILY are used Aug 27 16:38:42 ok. **** ENDING LOGGING AT Wed Aug 28 02:59:58 2013