**** BEGIN LOGGING AT Tue Apr 30 02:59:58 2013 Apr 30 07:33:26 good morning Apr 30 07:35:14 hi apelete Apr 30 07:36:46 hi woglinde Apr 30 07:36:48 morning Apr 30 07:37:33 good morning Apr 30 07:50:13 hi ka6sox and mckoan and jama and tasslehoff Apr 30 07:50:13 hi ka6sox, hi mckoan Apr 30 07:50:47 morning all Apr 30 07:55:15 hi woglinde and the rest of you :) Apr 30 07:57:51 hi bluelightning Apr 30 08:00:12 morning tasslehoff, woglinde, all Apr 30 08:01:18 morning all, bluelightning , woglinde Apr 30 08:01:27 hi silvio_ Apr 30 08:40:39 i've added a custom staging phase to one of my recipes, but building always fails with the error ' do_populate_sysroot) failed with exit code '1''. Logfiles don't reveal any useful information. What could be the problem? An empty do_stage() function gives the same problem. Apr 30 08:54:12 hi tasslehoff, bluelightning, silvio_, all Apr 30 08:54:21 hi apelete Apr 30 09:08:48 hi apelete Apr 30 09:25:50 morning all Apr 30 09:28:53 hi pb_ Apr 30 09:37:10 Hi everybody. I'm trying to open a devshell on my server (so no Konsole, xfterm, etc), but bitbake is complaining. It seems it doesn't speak to tmux correctly; can someone help me? Apr 30 09:37:28 Here are the logs: http://pastebin.com/UkSkn7UV and http://pastebin.com/iZc1cCig Apr 30 09:38:28 It looks like bitbake is passing some wrong parameters to tmux Apr 30 09:38:54 because tmux is returning its usage options Apr 30 09:39:49 hi bluelightning Apr 30 09:46:09 hi all Apr 30 09:46:44 hi JaMa Apr 30 09:58:28 I've interesting issue with local fetcher (file://) Apr 30 09:59:22 we have R/O samba mount of PREMIRROR, SOURCE_MIRROR_URL ?= "file:///mnt/blah/blah" Apr 30 10:00:02 so most of files in downloads folder are just symlinks to /mnt/blah/blah Apr 30 10:00:57 but with --enable-generate-mirror-tarballs it's a problem because downloads/git2_abc.git.tar.gz is also symlink to R/O file system Apr 30 10:04:59 in a recipe I'm trying to DEPEND on net-snmp-server-snmpd but it says that the package can't be found, however I include that package in my image recipe with using IMAGE_FEATURES without a problem, any ideas why? Apr 30 10:06:46 jackmitchell: instead of DEPENDS I think you want RDEPENDS_${PN} Apr 30 10:06:57 jackmitchell: also I guess you meant IMAGE_INSTALL rather than IMAGE_FEATURES ? Apr 30 10:08:50 bluelightning: yes IMAGE_INSTALL, I need a header file from net-snmp will RDEPENDS ensure it is built before the package is built? Apr 30 10:09:32 I thought RDEPENDS, was Runtime Depends, so it would just ensure that net-snmp-server-snmpd is included in the image Apr 30 10:09:46 I'm confused about the last discussion about ${libexecdir} Apr 30 10:09:48 jackmitchell: no, for that you need DEPENDS; however it would be DEPENDS on just net-snmp not a runtime package name Apr 30 10:10:12 bluelightning: ok, is there a way to tell between the two, I must be missing something Apr 30 10:10:22 as far as I understand, libexecdir should not contain user executables, only libraries and executables to be used by other programs Apr 30 10:10:42 how does OE use libexecdir right now? And how it will be changed? Apr 30 10:10:50 bluelightning: ah, no worries I get it Apr 30 10:11:07 jackmitchell: DEPENDS is build time, so it's a recipe (or matching PROVIDES); RDEPENDS is runtime so it's a package (or matching RPROVIDES) Apr 30 10:11:36 net-snmp is the build package, packaes defined in PACKAGES are extra runtime "grouped" packages Apr 30 10:11:46 meta-packages almost Apr 30 10:12:05 jackmitchell: best not to use the term package at the build time end, we have "recipe" for that Apr 30 10:15:23 eren: libexecdir used to be /usr/libexec/; in danny it was changed to /usr/lib/${BPN} Apr 30 10:15:49 eren: or rather, ${libdir}/${BPN} Apr 30 10:17:02 BPN is package name, then? Apr 30 10:17:24 eren: recipe name Apr 30 10:17:31 oh, okkie Apr 30 10:17:42 although the "P" did stand for package since there was confusion of the term "package" in the old old days Apr 30 10:20:03 now it's back to original Apr 30 10:20:26 it doesn't create subdirectories Apr 30 10:27:29 eren: libexecdir is like bindir; it is a place where binaries (not libraries) are to be placed. Difference is, that these binaries are not to be called by users but by other programs only (e.g. because fds have to be set up) Apr 30 10:29:22 argl.... kernel build with gcc 4.8 fails with "internal compiler error: in advance_target_bb, at sched-rgn.c" :( Apr 30 10:34:44 ensc|w: ok, so libexecdir in OE is not decided yet? I mean, Saul proposed some change at the end, and Encrico opopsed Apr 30 10:34:48 opposed* Apr 30 10:36:05 * ensc|w is Enrico Apr 30 10:39:40 oh :) Apr 30 10:40:04 I deduce it is not decided yet, then Apr 30 10:46:02 woglinde_: how do you modify patches against openjdk, it seems using the usual quilt edit/refresh is not possible since patches are applied at configure stage manually? Apr 30 11:11:35 RP, you are really planning to go on vacation? Apr 30 11:16:29 suihkulokki its possible, but you have to add them Apr 30 11:16:33 in the recipes Apr 30 11:16:53 under ICEDTEA patches Apr 30 11:34:13 hmm is there a way to run depmod prior to booting, aka on the host Apr 30 11:39:24 Crofton|work: yes ;-) Apr 30 11:43:28 has somebody an unpatched gcc 4.8 available and can check whether http://pastebin.com/raw.php?i=EgNEUPDA ices with it too? Apr 30 11:43:44 woglinde_: that applies to icedtea patches, but I think I can't do that on patches against openjdk under build/ directories? Apr 30 11:44:29 RP, good Apr 30 11:46:01 suihkulokki you can but you have to adjust the paths Apr 30 12:01:02 ensc|w: heh, it would be slightly sad if that testcase does ice in 4.8. I think I have a 4.8 tree somewhere, I'll have a go. Apr 30 12:03:29 pb_: accordingly http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56732 this might be arm specific Apr 30 12:04:14 yeah, probably, the implementation of naked functions is quite target-specific. Apr 30 12:04:27 ah, I see from the bugzilla that it does indeed ice in 4.8 so I won't bother testing. Apr 30 12:11:44 pb_: hi Apr 30 12:12:50 fwiw someone was fighting with 4.8 on #beagle (CSS tc maybe?) Apr 30 12:13:09 hi ant_work Apr 30 12:13:17 sorry, I haven't tested that kexec thing yet but I will try to do it this afternoon Apr 30 12:13:36 no hurry, thx Apr 30 12:16:43 build of gcc-4.8 does not seem to be smp safe either; here, it fails from time to time because bconfig.h can not be found :( Apr 30 12:18:08 So far gcc 4.8 seems to be a turd Apr 30 12:19:44 I'm not sure it's much worse than any other x.y.0 release has been. Apr 30 12:20:07 The bconfig.h thing is weird though, I'm surprised none of the gcc developers has noticed that. Apr 30 12:20:22 pb_: my memory is bad, so I'm just going with this once is worse :) Apr 30 12:32:46 pb_: are the autobuilders running with PARALLEL_MAKE = '-j '? Apr 30 12:41:58 the gcc autobuilders? I don't know (the gcc-testresults emails don't generally say) but I would have expected so. Apr 30 12:48:06 pb_: you are using MIPS devices for work? Apr 30 12:48:33 yeah, most of our current products are MIPS Apr 30 12:49:56 pb_: do you deal with ralf? E.g. I have this (http://patchwork.linux-mips.org/patch/2698/ and 2699) waiting for review since 08/2011. :) Apr 30 12:55:15 no, we haven't needed to deal with Ralf very much (at least, not lately). We aren't doing a whole lot of kernel work to be honest. Apr 30 12:56:15 pb_: no; I mean the oe autobuilders Apr 30 12:56:45 oh, right. I don't think oe itself has any autobuilders, do you mean the yocto ones? Apr 30 12:57:34 if so, I have no idea at all how those are configured or what they do. you could ask the h4x0rs in #yocto I guess. Apr 30 12:58:49 ensc|w or ask jama he builds world whith high -j too Apr 30 13:00:56 fwiw, I think the jenkins setup we use here is using the default oe-core values for -j. That seems to give -j16 on our build host. Apr 30 13:01:07 I routinely use -j32 Apr 30 14:39:48 ensc|w: woglinde: not very high -j, just 8 Apr 30 14:48:47 jama okay Apr 30 14:58:26 JaMa: and things like systemd build for you? systemd's do_install() fails here nearly 100% with -j8: https://bugs.freedesktop.org/show_bug.cgi?id=63474 Apr 30 14:58:46 ensc|w: don't you have unpached make in your host distro? Apr 30 14:58:58 ensc|w: there is thread about it on oe-core ML Apr 30 15:01:35 I am using Fedora18's make Apr 30 15:02:08 and this error is definitively within systemd's makefile Apr 30 15:02:53 ensc|w: build make-native first and it should build fine Apr 30 15:03:30 ensc|w: http://patches.openembedded.org/patch/48853/ Apr 30 15:06:49 JaMa: fedora18 has it already http://pkgs.fedoraproject.org/cgit/make.git/commit/?h=f18&id=3c30cb13ffa3b415a43614c3aa177ea9dde93026 Apr 30 15:07:36 hi. i am implementing some sort of a distro based on OE. so far i was tracking 'master' branch, but wanted to switch to 'dylan' to avoid large changes for now. I can see 'dylan' branch on oe and meta-oe, but not on bitbake. 'as a distro' should I simply track bitbake/master branch? Apr 30 15:08:13 ensc|w: ok, this is all I can think of.. I've ubuntu on jenkins and gentoo on my home machine and both build systemd fine with -j8 Apr 30 15:11:59 ndec there was no release? Apr 30 15:12:50 woglinde: the last tag is 5 months old. Apr 30 15:12:54 is that what you mean? Apr 30 15:14:10 there is a branch called 1.18. is that what 'dylan' is using? Apr 30 15:15:10 ndec: see what commit the Yocto Project use Apr 30 15:15:36 they do a bundle with oe-core, meta-yocto, bitbake so have a look and see which commit it is, if you can Apr 30 15:16:21 jackmitchell: indeed, it's the HEAD of 1.18 branch. Apr 30 15:16:29 which is imported in poky/dylan Apr 30 15:16:33 ndec: yes bitbake-1.18 is right branch for dylan Apr 30 15:16:40 1.16 for danny Apr 30 15:16:44 master only for master Apr 30 15:16:46 ok, thanks. i will track that branch. Apr 30 15:16:47 hm they should have made a release Apr 30 15:41:01 make[1]: *** No rule to make target xamples/01-swallow.edc', needed by ll-am'. Stop. Apr 30 15:41:05 :-/ Apr 30 15:41:27 ERROR: Task 6467 (/home/rah/phones/shr/shr-core-gta04/meta-openembedded/meta-efl/recipes-devtools/python/python-edje_1.7.0.bb, do_compile) failed with exit code '1' Apr 30 15:42:35 rah: "xamples"? some kind of off-by-1 or typo? Apr 30 15:44:13 it's some race.. I've seen it like 1 in 20 builds on jenkins Apr 30 15:46:32 bluelightning: bad paste through screen Apr 30 15:46:43 make[1]: *** No rule to make target `examples/01-swallow.edc', needed by ll-am'. Stop. Apr 30 15:46:57 I guess PARALLEL_MAKE = "" is needed then? Apr 30 15:47:05 (in this particular recipe) Apr 30 15:48:08 make[1]: *** No rule to make target `examples/evas-demo/01-app_launcher/01-app_launcher.edc', needed by `all-am'. Stop. Apr 30 15:48:16 that's with MAKEOPTS="" Apr 30 15:48:49 I don't think that'll do anything Apr 30 15:49:16 it's kill the build Apr 30 15:49:57 in fact, it looks like the original error doesn't go away with MAKEOPTS="" Apr 30 15:50:04 yup Apr 30 15:50:09 (I modified the Makefile) Apr 30 15:50:16 ? Apr 30 15:50:18 make[1]: *** No rule to make target `examples/01-swallow.edc', needed by `all-am'. Stop. Apr 30 15:50:22 wrong solution Apr 30 15:50:26 that's with MAKEOPTS="" Apr 30 15:50:32 have you tried setting PARALLEL_MAKE = "" in the recipe? Apr 30 15:50:46 no Apr 30 15:50:50 bluelightning: why would I do that? Apr 30 15:51:07 rah: this is how we handle upstreams that have broken makefiles Apr 30 15:51:29 er, hm Apr 30 15:51:37 actually that's probably not what's happening here is it Apr 30 15:51:44 bluelightning: how does this "handle" a broken makefile? Apr 30 15:51:55 rah: are you perhaps using make 3.82 on your host machine? Apr 30 15:52:13 bluelightning: no, make 3.81 Apr 30 15:52:42 hmm, not sure then :( Apr 30 15:54:48 are there tools which can generate a dot-style depgraph for gnu makefiles? Apr 30 15:55:02 hmm this is new: Apr 30 15:55:03 | DEBUG: Executing python function do_package_ipk Apr 30 15:55:03 | *** Warning: The following files have a UID greater than 99. Apr 30 15:55:03 | You probably want to chown these to a system user: Apr 30 15:55:03 | -rw-r--r-- 3 xuser xuser 1874 Oct 2 2010 ./usr/src/debug/libnotify/0.6.0-r1/libnotify-0.6.0/libnotify/internal.h Apr 30 15:57:04 why shoud xuser own such files? Apr 30 15:57:29 it shouldn't... Apr 30 15:58:32 uhm Apr 30 15:59:01 we should check for strange user owning files in / Apr 30 15:59:25 buildhistory can be used to detect when these things change, FWIW Apr 30 15:59:34 I've seen this warning only because the same python function failed later with cannot allocate memory Apr 30 15:59:48 so maybe this warning should be promoted to bb.warn too Apr 30 16:00:05 JaMa: I don't think this is a warning from the metadata, I think it's opkg Apr 30 16:01:05 ah right.. opkg-build:You probably want to chown these to a system user: " >&2 Apr 30 16:01:08 so opkg-utils Apr 30 16:06:22 JaMa: if you bitbake -c devshell imagename and then look in ../rootfs how many files are owned by xuser? Apr 30 16:06:51 (i.e. this will be checking under pseudo, which is what counts) Apr 30 16:07:51 it will build for few more hours but I'll look later Apr 30 16:11:00 how do I add arguments to ./configure in a recipe? Apr 30 16:12:49 rah: EXTRA_OECONF += Apr 30 16:19:33 bluelightning: thanks Apr 30 16:25:52 * rah has patched python-edge.inc Apr 30 16:25:55 err Apr 30 16:25:57 * rah has patched python-edje.inc Apr 30 16:44:49 rah: does it help? is some patch comming? Apr 30 17:02:35 JaMa: regarding your llvm thing that you posted on the mailing list, you can't do "pop {lr}" in thumb, the instruction encoding just doesn't allow it. Apr 30 17:10:41 pb_: this part was taken from khem's patch in llvm2.8 http://git.openembedded.org/meta-openembedded/tree/meta-oe/recipes-core/llvm/llvm2.8/0019-issue6065.patch Apr 30 17:11:18 zecke_: the qt5.1 CI you blogposted about doesn't use OE, right? Apr 30 17:15:56 JaMa: and does the llvm2.8 one actually work? That patch looks broken to me. Apr 30 17:16:48 JaMa: no. It is a crazy shit broadcom MIPS/uclibc toolchain Apr 30 17:17:13 zecke_: ah, heh, one of those Apr 30 17:17:36 we use oe on our broadcoms :-} Apr 30 17:17:40 JaMa: they used to be a "key-account" for Trolltech.. but Broadcom has decided to use Apr 30 17:18:30 JaMa: Android and Chrome.... everybody knows how much Google cares about MIPS. Apr 30 18:02:59 pb_: it builds fine, woglinde said that there were some issues with it and that it's currently disabled in openjdk Apr 30 18:03:27 zecke_: ah, OK I was just wondering if there is another meta-qt5 layer I don't know about.. Apr 30 18:04:08 JaMa: sadly not. Lay-offs happened before I got around to not use the crappy toolchain Apr 30 18:05:23 after a lot of time spent trying to fix meta-qt5 I agree, it's very sad that you didn't have enough time to fix it before me :) Apr 30 18:11:04 JaMa: why do you need it? for WebOs/QtWebKit? Apr 30 18:11:37 yes Apr 30 18:11:58 but it's almost fixed Apr 30 19:54:44 I have only used oe/bitbake once years ago and need to create a custom distro for a project I have. I am having a bit of trouble getting started. the yocto project already has a distro and some of the docs mention it was integrated into the new openembedded-core. I tried the oe-made easy but it seems to also make critical decision already and changing is unclear. Ther is a lot of old... Apr 30 19:54:45 ...documentation. I need to know what is the current best practice. When I source the oe-init script in oe-made-easy, it seems to create a conf folder in the wrong spot so I cannot bitbake the target I need. Apr 30 19:56:25 ie. there is only a conf directory in the raspberry-pi folder created by oe-init no other things like the pseudo it complains about **** ENDING LOGGING AT Wed May 01 02:59:58 2013