**** BEGIN LOGGING AT Sun Aug 23 02:59:56 2009 Aug 23 07:01:02 hm, i have the dreaded Aug 23 07:01:03 fatal: cannot pread pack file: No such file or directory Aug 23 07:01:04 fatal: index-pack failed Aug 23 07:01:09 message again when i do a git pull Aug 23 07:01:31 anyone an idea on how to fix this, last time I had this I ended up starting from scratch and wasting several hours Aug 23 07:02:00 it might be that this happened because on my previous pull the network had issues Aug 23 07:19:08 hmmm yeah Aug 23 07:19:27 may be we should build git native with NO_PREAD=1 Aug 23 07:19:39 which package is it complaining on Aug 23 07:19:51 it immediately does so after a git pull Aug 23 07:20:03 on OE ? Aug 23 07:20:10 yes Aug 23 07:20:30 oh then I guess you have to reclone Aug 23 07:20:41 or recompile git :) Aug 23 07:20:54 with NO_PREAD set Aug 23 07:20:59 which distro is it ? Aug 23 07:21:04 i'm in my openembeeded dir, do a git pull, it says receving objects 100% .... done, then I get the fatal Aug 23 07:21:14 angstrom Aug 23 07:21:20 on your host Aug 23 07:21:26 oh sorry opensuse Aug 23 07:22:13 well if I nuke everything it works, it worked in the past too, but I had a dns issue during my latest pull so somewhere it got something wrong, maybe a truncacted file or so Aug 23 07:22:22 (and that is opensuse 11.1 to be precise) Aug 23 07:23:00 I know discarding everything and starting from scratch will get me back into business, but that will cost quite some time due to the rebuild Aug 23 07:23:04 probably you pulled from a mirror which was not cloned fully Aug 23 07:23:20 I dont know any other way to solve it Aug 23 07:23:29 if packfile is corrupt Aug 23 07:23:48 where does the packfile live ? Aug 23 07:24:06 i have .git/index Aug 23 07:24:14 about 2.2 mb Aug 23 07:25:51 I dont think its just one file Aug 23 07:27:12 you can just delete metadata and leave your tmp intact and then reclone Aug 23 07:27:23 looking at .git/objects, but also getting some help in #git now Aug 23 07:28:07 good Aug 23 07:31:51 03Khem Raj  07org.openembedded.dev * rcb55e66397 10openembedded.git/recipes/u-boot/u-boot_git.bb: Aug 23 07:31:51 u-boot_git.bb: Dont reset machine SRC_URIs by using common SRC_URI. Aug 23 07:31:51 We have this SRC_URI which forces bitbake to abandon the machine Aug 23 07:31:51 specific SRC_URI override if it happens to be before this in Aug 23 07:31:51 parse order. Anyway this SRC_URI is not needed too. Aug 23 07:31:53 Signed-off-by: Khem Raj Aug 23 07:47:48 03Khem Raj  07org.openembedded.dev * rebbe06e97f 10openembedded.git/conf/distro/ (3 files in 2 dirs): Aug 23 07:47:48 minimal, sane-toolchain: Use eglibc if LIBC is not selected. Aug 23 07:47:48 Currently we fall back to glibc if LIBC is not set. Aug 23 07:47:48 We now will use eglibc as default lib for minimal Aug 23 07:47:48 and sane-toolchain Aug 23 07:47:50 Signed-off-by: Khem Raj Aug 23 07:47:52 03Khem Raj  07org.openembedded.dev * rb2247b97bb 10openembedded.git/conf/distro/micro.conf: Aug 23 07:47:57 micro: Use eglibc as default lib if LIBC is not set. Aug 23 07:47:59 Signed-off-by: Khem Raj Aug 23 07:48:01 03Khem Raj  07org.openembedded.dev * r8043677ecf 10openembedded.git/conf/distro/include/ (sane-toolchain-uclibc.inc sane-toolchain.inc): Aug 23 07:48:04 sane-toolchain: Define eglibc PREFERRED_VERSIONS Aug 23 07:48:06 Set preferred versions for eglibc to be 2.10. Aug 23 07:48:08 Bump up uclibc PREFERRED_VERSION to 0.9.30.1 Aug 23 07:48:12 and for uclibc add more PREFERRED_PROVIDERS. Aug 23 07:48:14 Signed-off-by: Khem Raj Aug 23 13:59:07 staging/i686-linux is very useful, keeps tools from being interfered by my regular installation. Aug 23 14:00:31 * Kero also see that auto-dep finding can be bad. is C depends on B and B depends on A (from linking .so s), C will depend on A. If C strictly uses B, that's not desired. Can I remove the auto-dep on A by hand in the .bb file? Aug 23 14:20:26 Kero: Well, the auto-dep is correct in that way Aug 23 14:20:36 Kero: if you don't like that, link with -Wl,--as-needed Aug 23 14:23:00 it's not wrong, but it'd be nice when B is rebuilt depending on D, C would not have to be rebuilt; with package C only depending on B, that should work. Aug 23 14:23:16 Kero: i don't follow on this. Aug 23 14:23:41 Kero: package dependencies != build dependencies Aug 23 14:24:08 * Kero is talking about package deps. Aug 23 14:26:25 Kero: Okay, I'm not able to follow then. The basic so rule is. If you use a symbol x from libx.so, you will need to link to libx.so. Now sometimes the lib list is too excessive this is why --as-needed exists. but as I don't understand the example it might total rubbish Aug 23 14:27:59 maybe I'm talking rubbish, and maybe I should read docs to understand exactly what is going on. Aug 23 14:28:55 so let me rephrase the Q Aug 23 14:29:06 regardless of auto-deps. Aug 23 14:36:02 been building ruby/gnome2 packages today; libgdk-pixbuf-ruby1.8 uses pango, gdk-pixbuf, cairo, libpango-ruby, rcairo and of course ruby. Aug 23 14:36:17 libpango-ruby also uses rcairo and ruby. Aug 23 14:37:03 in theory (it's a bad idea, o I' won;t do it), dropping package dependency of libpango-ruby on ruby will not hurt, rubyb will be installed because of rcairo. Aug 23 14:37:44 would something like RDEPENDS -= "ruby" work? (suppose not) Aug 23 14:38:45 Kero: -= not, oe_filter_out is present... the question is what do you win? Aug 23 14:38:56 Kero: if that library needs header files staged by ruby you will lose Aug 23 14:38:58 nothing in this case. Aug 23 14:39:19 Kero: or actually... RDEPENDS = "ruby" and the "ruby" package contains the interpreter? Aug 23 14:39:36 Kero: in that case this would be wrong... it might/should RRECOMMENDS the interpreter Aug 23 14:40:01 was looking for a better example when I realized that example to be not depending anyway: libgtk-ruby does not depend on anything X related, which is good. Aug 23 14:41:24 so I learned something more about OE/bitbake again. educational weekend. Aug 23 14:44:03 do you know if DEPENDS_sub_package does anyhting? I had it in old bb files, but it does not see to create any (build) dependency Aug 23 14:44:11 *seem Aug 23 14:44:25 Kero: that sounds right Aug 23 14:47:03 don't remember whether I made it up back then; as long as RDEPENDS_sub_package and DEPENDS work, things do seem to work as I need them to work. Aug 23 14:48:38 so now I have ruby 1.8.6 recipes & patches and a ruby/gtk binding; where do I send those recipes & patches? Aug 23 14:49:32 technically I made them for shr-unstable, but esp the ruby1.8.6 recipe should be good for any OE distro. Aug 23 14:50:17 Kero: oe-dev is a good place to mail them to Aug 23 14:50:37 Kero: use git-format-patch, and one patch per email, so our patchwork.openembedded.org keeps track of them Aug 23 15:01:53 git-format is just diff -ru, I think? will look that up in the oe-dev archives to be sure ;) Aug 23 15:02:12 Kero: more like, use git commit to commit them, and then send the commits to us Aug 23 15:02:46 can do, too! :) Aug 23 15:36:17 gm Aug 23 15:38:37 likewise: hey! Aug 23 15:38:41 how are you? Aug 23 15:48:22 I am getting gzip: broken pipe errors due to tar -xzf from unpack routines (on karmic ) Aug 23 15:48:38 Kero: yeah Aug 23 15:48:46 khem`: yeah, me too on poky :) Aug 23 15:48:59 khem: read the gzip FAQ on SIGPIPE and shells Aug 23 15:49:15 we do os.system() Aug 23 15:49:24 in OE classes Aug 23 15:50:05 khem: read the gzip FAQ on SIGPIPE and shells Aug 23 15:50:48 oops wrong tab completion Aug 23 15:51:14 zecke: I changes my nick a moment ago Aug 23 15:51:26 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rd36dfc7937 10openembedded.git/recipes/linux/linux-2.6.30/at91/2.6.30.2/patch-sets/linux-2.6.30.2-at91-008-TCLIB-Warning-for-SAM9G45.patch: add dependency on xlib-gpe-ximage modified: recipes/tftpboot/tftpboot.bb to tftpboot Aug 23 15:51:27 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r39160f04bf 10openembedded.git/recipes/tftpboot/tftpboot.bb: And add recipe fix Aug 23 15:51:27 someone keeps stealing khem whenever I logoff Aug 23 15:51:27 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rf03c320b26 10openembedded.git/recipes/at91bootstrap/ (5 files in 3 dirs): Add trap handlere to at91bootstrap Aug 23 15:51:29 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r75e8ac2e35 10openembedded.git/recipes/at91bootstrap/ (7 files in 3 dirs): Remove at91bootstrap-2.13-rc1/-rc3 in favour of -rc4 Aug 23 15:51:32 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rd5519fba5e 10openembedded.git/recipes/linux/linux-2.6.30/at91/2.6.30.2/ (boards/at91sam9g45ek/defconfig patch-sets/SRC_URI_append.inc): remove tclib from g45 defconfig, broken Aug 23 15:51:51 zecke: hmm the FAQ says that it can be safely ignored in some cases Aug 23 15:53:06 zecke: may be we should also ignore it in bitbake/classes if there is no other error reported Aug 23 15:53:10 khem: in poky I just ignored it... Aug 23 15:53:24 zecke: cool. Push it to OE too Aug 23 15:53:42 khem: the tricky thing is how should we know that everything got extracted? where is this coming from? bash? linux 2.6.31? Aug 23 15:53:54 khem: oh, I just witnessed it on a poky build and patched it locally :) Aug 23 15:56:53 khem: do you have any better idea than ignoring? Aug 23 15:58:57 http://www.mail-archive.com/universe-bugs@lists.ubuntu.com/msg115722.html Aug 23 16:01:27 http://www.chiark.greenend.org.uk/ucgi/~cjwatson/blosxom/2009-07-02-python-sigpipe.html Aug 23 16:02:14 okay, mystery solved? Aug 23 16:37:16 zecke: I change base.bbclass yesterday to use Popen Aug 23 16:37:19 it did not help Aug 23 16:38:10 but I did not install the signal handler Aug 23 16:42:50 hi Aug 23 18:47:40 jo kergoth Aug 24 01:01:30 Is there a list of recipes and DESCRIPTION fields somewhere on the web? Aug 24 01:01:56 Or an easy way to generate such a list, perhaps for the preferred versions only? Aug 24 01:04:19 (many DESCRIPTION fields are multiline, so a simple grep does not do what I want) **** ENDING LOGGING AT Mon Aug 24 02:59:57 2009