**** BEGIN LOGGING AT Sat Jun 29 02:59:59 2013 Jun 29 09:38:36 rburton: I've fix for mesa, I'll send it to upstream and then update mesa_git.bb Jun 29 10:48:56 bluelightning: do I remember correctly that you wrote the functionality for checksuming local files to include them in do_fetch task checksum? I think I've found issue Jun 29 10:49:14 JaMa: I did, yes Jun 29 10:49:18 JaMa: what did you find? Jun 29 10:50:27 I have mesa recipe in oe-core layer, then newer version meta-webos-backports layer and then .bbappend in another layer Jun 29 10:51:03 FILESPATH correctly search in FILESEXTRAPATH set in .bbappend, then in meta-webos-backports where the appended .bb is and then in oe-core Jun 29 10:51:25 but that checksuming logic is looking only in FILESEXTRAPATH and then in oe-core Jun 29 10:51:47 I'll put it in some better form and pastebin it before creating bug Jun 29 10:58:15 bluelightning: http://bpaste.net/show/110683/ Jun 29 10:58:47 actually FILESPATH doesn't have oe-core paths only meta-webos-backports Jun 29 11:02:35 I've added BBPATH and FILESEXTRAPATHS, something else I should include in bug report? Jun 29 11:10:14 https://bugzilla.yoctoproject.org/show_bug.cgi?id=4790 Jun 29 11:10:15 Bug 4790: normal, Undecided, ---, paul.eggleton, NEW , get_file_checksums(filelist, pn) is searching in different directories then FILESPATH used in do_fetch task Jun 29 12:08:06 JaMa: thanks, I'll look into it Jun 29 12:08:25 bluelightning: thanks Jun 29 12:10:16 bluelightning: btw: I was looking into reading PKGSIZE from buildhistory_get_installed() and haven't seen simple way to map installed libewebkit0 to buildhistory/packages/xscalete-oe-linux-gnueabi/webkit-efl/webkit-efl/latest, did I overlook something? Jun 29 12:10:51 JaMa: the runtime-reverse subdirs within pkgdata would probably be the only way Jun 29 12:11:12 you would also need the pkgarch from the package Jun 29 12:11:56 so to extend scripts/oe-pkgdata-util and use list_installed_packages arch from populate_sdk_ipk.bbclass, right? Jun 29 12:12:20 right, using oe-pkgdata-util would be useful here yes Jun 29 12:12:39 OK, thanks Jun 29 13:51:19 Hello there. I have a problem creating a recipe for a CPAN module. It fails with "... perl/5.14.3/auto/Data/Dumper/Dumper.so: wrong ELF class: ELFCLASS32". Any idea of how to get around this? I'm trying to create a system for i.MX6, on a x86_64 platform. It seems that it uses the x86_64 version of Perl to run "perl Build", and then fails because the Dumper.so is compiled for ARM. Jun 29 16:34:29 kspr: are you inheriting the cpan class? Jun 29 16:41:00 bluelightning, yes, thanks, I found another solution, it is working now. Jun 29 16:41:09 kspr: ok, great Jun 29 22:49:33 bluelightning: ping Jun 29 22:49:47 JaMa: pong Jun 29 22:52:01 bluelightning: that functionality for reading PKGSIZE (or any variable from pkgdata) is almost done: http://bpaste.net/show/110816/ Jun 29 22:53:27 bluelightning: only remaining issue is TARGET_OS used, because e.g. hicolor-icon-theme is all-oe-linux not all-oe-linux-gnueabi as set in image recipe which inherits that .bbclass Jun 29 22:53:38 bluelightning: and in image.bbclass is used the same: Jun 29 22:53:50 for vendor in '${TARGET_VENDOR}' ${MULTILIB_VENDORS} ; do Jun 29 22:53:50 oe-pkgdata-util glob ${TMPDIR}/pkgdata $vendor-${TARGET_OS} ${WORKDIR}/installed_pkgs.txt "$GLOBS" >> ${WORKDIR}/complementary_pkgs.txt Jun 29 22:53:53 done Jun 29 22:53:59 yes I know, there is a bug filed for that Jun 29 22:54:22 https://bugzilla.yoctoproject.org/show_bug.cgi?id=4227 Jun 29 22:54:23 Bug 4227: minor, Low, 1.5, paul.eggleton, NEW , Fix naive complementary package installation handling of TARGET_OS Jun 29 22:56:01 bluelightning: ah good to know Jun 29 22:58:18 bluelightning: whould it be so bad hack to drop last - element from TARGET_OS when it's not found in previous one? Jun 29 22:58:43 bluelightning: because I don't see how PKGMLTRIPLETS whould help me in code where I already know which architecture I want Jun 29 22:58:51 just don't know what suffix it had Jun 29 22:59:00 hmm... I can't help feeling there should be a better solution Jun 29 23:04:02 if not os.path.exists(revlink): Jun 29 23:04:02 # [YOCTO #4227] try to drop -gnueabi from TARGET_OS Jun 29 23:04:02 multimach_target_sys = '-'.join(multimach_target_sys.split('-')[:-1]) Jun 29 23:04:02 revlink = os.path.join(pkgdata_dir, multimach_target_sys, "runtime-reverse", pkg_name) Jun 29 23:04:05 is my hack for now Jun 29 23:09:37 hmm spliting pkgfile by '_' also doesn't work very well :/ kernel-module-x-tables_3.8.13+git3+f121c06ae8_f20047520a-r4.2.2_qemux86_64.ipk breaks it twice (extra _ in version and architecture **** ENDING LOGGING AT Sun Jun 30 02:59:58 2013