**** BEGIN LOGGING AT Sat May 05 02:59:56 2007 May 05 06:01:48 morning all May 05 07:10:06 goooooooood morning all May 05 07:13:53 perl 5.8.7 build is broken.. at least for me May 05 07:24:58 good morning! May 05 07:27:08 Laibsch: the squashfs images has to create indices and stuff for all the files May 05 07:27:24 and zillion small files need a lot of space in the index May 05 07:27:52 I am a bit surprised that it takes ~1GB more than the 7z version May 05 07:28:07 but a 2g card is like 15 euro nowadays May 05 07:29:05 2G would be fine. But I want two wikipedia. So that is a 4G card May 05 07:29:29 While still being affordable the increase in size stunned me May 05 07:30:13 OTOH such an index makes finding files really fast May 05 07:43:40 I certainly agree May 05 07:43:58 The speed is very much necessary on slow embedded devices May 05 07:44:14 And the money for a bigger card is probably well spent May 05 07:44:28 Nonetheless, the increase in size is stunning. May 05 07:45:19 BTW, koen, did you try and gzip or bzip2 the squashfs? If it compressed a lot more then I assume that there was room for improvement in the algorithm. May 05 07:45:38 * Laibsch lacks the CPU cycles to try immediately May 05 07:45:58 bzipping the squashfs defeats the points of it being a filesystem May 05 07:46:09 koen: I know May 05 07:46:28 I do not intend to use it May 05 07:46:35 But I would like to see the numbers May 05 08:19:43 http://www.angstrom-distribution.org/unstable/images/collie/20070505/ May 05 08:25:21 Boys and girls, bug fixing weekend. We already have a mark of 75 fixed bugs for the last week. So let's push that to at least triple digit, OK? May 05 08:25:30 koen: Nice, thank you May 05 08:28:15 koen: There is an entry ZaurusKernels in the component field of the BTS search function. But it is not possible to set the component to that value. May 05 08:28:53 Sorry, Product field May 05 08:30:05 good morning May 05 08:32:54 hey likewise May 05 08:33:11 likewise: http://lists.freedesktop.org/archives/cairo/2007-May/010542.html May 05 08:36:10 hmm, I should look at the code where they do caching and compare this with the patches for <=1.4.2. May 05 09:10:51 Laibsch : Or at least try not to introduce any new ones ;) May 05 09:19:28 * koen completes some TODO items May 05 09:20:52 03koen 07org.oe.dev * re757d66b... 10/ (7 files in 3 dirs): May 05 09:20:52 linux-ezx 2.6.21: Sync with openezx svn May 05 09:20:52 NOTE: the defconfig needs some tweaking May 05 09:59:57 Can anybody think of a way to complete all steps until do_patch but not the following for package X and the packages it depends on? May 05 10:00:05 zecke: Can bittest do this? May 05 10:03:37 try adding a patch_all tasks a la fetch_all May 05 10:04:04 (and yes, bittest can do that) May 05 10:04:23 koen: But that will fetch ALL packages, won't it? May 05 10:04:32 no May 05 10:04:34 I just want package X and dependencies May 05 10:04:37 Nice May 05 10:04:59 bitbake -c fetchall fetches + deps May 05 10:05:11 marcin wrote it for his source mirror May 05 10:07:09 nice May 05 10:08:11 where is it defined? May 05 10:08:30 * Laibsch tries not to be lazy now and uses rgrep May 05 10:38:14 koen: No further compression with gzip. Trying bzip2 now. May 05 11:31:19 Is do_mrproper working and does it do what I think it does? May 05 11:31:35 IOW, does it fix May 05 11:31:40 !oebug 1864 May 05 11:31:41 * * Bug 1864, Status: NEW, Created: 2007-02-11 12:07 May 05 11:31:43 * * werner(AT)almesberger.net: bitbake -c forget May 05 11:31:43 * * http://bugs.openembedded.org/show_bug.cgi?id=1864 May 05 11:34:56 morning, there are rumours, is good. May 05 11:36:16 polyonymous: Are talking to me? May 05 11:37:44 Laibsch, to everyone who shares the morning :) May 05 11:37:45 Laibsch: no, it removes everything from the DL_DIR May 05 11:38:48 polyonymous: Well, I don't understand your allusion to rumours May 05 11:39:11 Laibsch, I'm just not being serious. That was a way to say "hi". May 05 11:39:26 hi May 05 11:40:06 :))) May 05 11:49:13 rumours, therefor are, good. morning May 05 11:52:06 * polyonymous agrees with likewise May 05 11:52:42 Has anyone of you problems to build obexftp lately? May 05 11:53:44 zecke|studying: Sorry to keep you from studying. Can we close May 05 11:53:47 !oebug 829? May 05 11:53:48 * * Bug 829?, Status: InvalidBugId May 05 11:54:01 !oebug 829 May 05 11:54:02 * * Bug 829, Status: NEW, Created: 2006-04-10 10:42 May 05 11:54:03 * * freyther(AT)inf.fu-berlin.de: Meta-Bug: Bitbake "It could suck more" 1.x.y May 05 11:54:04 * * http://bugs.openembedded.org/show_bug.cgi?id=829 May 05 11:54:12 pastebin.co.uk/14055 (i have checked bugz) Opie wouldnt build May 05 11:55:50 hmm May 05 11:56:10 Laibsch: if you want to, sure May 05 11:56:35 goxboxlive, built fine for me... May 05 11:57:00 hmm, strange, i am using a new work directory. May 05 11:57:20 you mean, you just started afresh? May 05 11:57:31 Because I've built from scratch recently. May 05 11:57:33 yes May 05 11:57:49 whole form scratch May 05 11:57:56 Well, maybe if it was a mess before, then it may help. May 05 11:58:36 I don't really understand why libiconv_* I thought the symbols are called iconv? But you never know until you have troubles :) May 05 11:58:46 right :-) May 05 11:59:05 how recent that revision is, btw? May 05 11:59:41 ah, I see. May 05 11:59:57 That's the most recent I'm updating to now :) May 05 12:00:18 ok, i have just pulled so i am using the latest one May 05 12:00:41 yup. I see. May 05 12:01:05 Well, I can build opie-kdepim-image for spitz with the recent tree. May 05 12:01:41 And for poodle too, but I haven't tried the latter. hvontres have, though. May 05 12:03:15 ok May 05 12:07:05 hmm... git on the device should depend on findutils and cpio. May 05 12:09:56 is it also failing for u now after you pulled? May 05 12:10:36 goxboxlive, haven't tried, but according to updated files they do not affect my build. May 05 12:10:43 goxboxlive, what is your failure now? May 05 12:10:59 It's the same, obexftp May 05 12:11:05 goxboxlive, because my tree is dirty like the dirtiest slut, but all the patches are in bugzilla :) May 05 12:11:08 ah... May 05 12:11:13 No, this one I didn't have... May 05 12:11:17 ok May 05 12:11:40 I can try to rebuild obexftp. I don't feel like doing the whole tree from scratch at the moment... May 05 12:13:30 goxboxlive, built fine... May 05 12:13:54 hmm, then it has something to do with libiconv on my build machine May 05 12:14:01 goxboxlive, have you included required opie versions in local.conf? May 05 12:14:19 yes the .3.inc version May 05 12:14:38 require conf/distro/include/preferred-opie-versions-1.2.3-pre.inc May 05 12:14:45 right May 05 12:14:57 ok, you just didn't mention -pre. May 05 12:15:12 Well, maybe it's the host, check your configure log May 05 12:15:24 i followd the one psokolovsky made at angstrom wiki May 05 12:15:35 goxboxlive, yes, that's the one he mentioned. May 05 12:18:41 mtn status|wc -l : 45 ;-) May 05 12:18:59 can I review the incoming diff between mtn pull and mtn update? May 05 12:19:25 likewise, I think you can, by specifying -r ? May 05 12:19:59 likewise: mtn diff -r `something with mtn status` -r `something with mtn automate heads` May 05 12:21:21 polyonymous: if you are foolish enough to build libiconv (not the one internal to glibc), you get what you deserve May 05 12:21:34 koen, that wasn't me! :) May 05 12:22:01 no, but I hate repeating myself, so I prefix a random username to get around that May 05 12:22:16 koen, and for what I understand, he's done the fresh build. May 05 12:22:36 is there already a fix for the libstdc++ RPATH QA error in gcc-cross somewhere? May 05 12:22:51 koen: do you have idea about OE feed invariants? Like, would it be possible for OE to generate feed with 2 same packages? Or put same package twice in Depends: ? May 05 12:23:10 koen, it looks like he has a glibc iconv, but somehow configure came up withe idea of using libiconv. Well, I haven't seen his logs. May 05 12:23:38 psokolovsky: iirc we fixed that double-depends bug May 05 12:23:48 psokolovsky: iirc it was likewise's patch May 05 12:24:33 likewise: I'm still in doubt whether or not to exclude staging/ from the rpath check May 05 12:24:56 koen: so, would think it will be safe to assume no dups in OE feeds? would it be sane to write a tool with such assumption? Would you personally use it? ;-) May 05 12:25:23 koen, I hate to nag, but I'm just wondering if my email went though to you or have I somehow managed to goof it up? :) May 05 12:25:25 psokolovsky: angstrom has all packages in the feeds inside a sqlite db :) May 05 12:25:49 pH5: no, no proper fix yet. I can spend some time on this, don't expect anything though... May 05 12:26:04 likewise: ok, thanks. May 05 12:26:55 koen: ok, nice, thanks. I have ideas on alleviating ipkg's retardedness May 05 12:27:32 s/though/through/ May 05 12:28:41 pH5: the gcc one doesn't propagate into the build May 05 12:34:57 koen: I don't understand. So May 05 12:35:02 ERROR: QA Issue package libstdc++ contains bad RPATH /home/ph5/src/oe/build/angs May 05 12:35:03 trom/staging/i686-linux/lib in file /home/ph5/src/oe/build/angstrom/work/armv5te May 05 12:35:03 -angstrom-linux-gnueabi/gcc-cross-4.1.2-r3/install/libstdc++/usr/lib/libstdc++.so.6.0.8 May 05 12:35:24 ah, heh May 05 12:35:28 is not a problem with the cros libstdc++, but with the fact that it is put into a package at all? May 05 12:35:32 that will propagate into the package :) May 05 12:36:03 I think package may benefit from this library :) May 05 12:36:29 hi May 05 12:38:46 pH5: we don't know if this is a problem (for this particular case), but we would expect no RPATH in there (I will check another cross-compiler to verify). This one is not bad in the sense that it gives overhead or security issues on the target. May 05 12:39:15 pH5: (to be clear: this does not give verhead or security issues on the target) May 05 12:39:52 psokolovsky: a sanity checker for feed(s) consistency using that sqlite db would be nice May 05 12:42:09 koen: yep, I guess, and very likely. I just want to know what people think about exactly such idea - move expensive (awfully expensive as they're implemented now) checks from ipkg, executed on every run, to a separate validator. May 05 12:42:59 what would be the side effect of this assumption failing in ipkg? May 05 12:43:31 polyonymous: I gave examples of assumptions ipkg checks above May 05 12:44:28 psokolovsky, no, I mean what would happen if ipkg relies on this checks passed, but the feed isn't valid in reality? May 05 12:45:15 polyonymous: what would be effect of dup package appearing in feed? now, ipkg always uses just 1st occurance. wih removing unique'ing code, it will still mostly use 1st occurance, but for exampel will dump both instances on "list", which is even more correct May 05 12:45:53 polyonymous: make exercise of tracing what would happen w/o dup checks in Depends: and friends yourself ;-) May 05 12:45:54 then I'd vote for moving the code out of ipkg. May 05 12:46:06 polyonymous: in other words, no segfaults expected ;-) May 05 12:46:20 psokolovsky, yes, that's about what I was thinking of :) May 05 12:46:34 polyonymous: cool, thanks ;-) prepare for beta-testing then (w/o any hurry or ETAs ;-) ) May 05 12:46:51 psokolovsky, "always prepared!" :) May 05 13:02:45 Can I inherit in an include file? May 05 13:06:01 you should probably "require" it May 05 13:06:23 inherit is usually done for .bbclass May 05 13:07:17 I take it Laibsch asked if he can inherit _IN_ include. May 05 13:07:43 oops, I misread, sorry May 05 13:07:44 and my guess is yes, although I have no clue what the proper answer is :) May 05 13:08:04 * CoreDump|home is pretty sure you can do that May 05 13:08:51 yup May 05 13:09:01 grep inherit packages/*/*.inc May 05 13:20:40 koen: the ptxdist toolchain does not have an RPATH in their cross gcc stdlibc++ May 05 13:21:01 koen: so we have got a small bug *and* a known good cross gcc to peek from May 05 13:44:55 * koen mentions the 'running chrpath in qemu' option again May 05 13:45:41 only a few packages need, so the time-cost shouldn't be too much May 05 13:47:15 koen: The solid-pattern cache work has landed in cairo now. May 05 13:57:34 cworth: nice, thanks for the heads up May 05 14:02:30 morning May 05 14:02:59 koen: ever seen this error? http://rafb.net/p/eMPq6y93.html May 05 14:03:06 morning May 05 14:03:25 likewise: yes May 05 14:04:58 koen: do you happen to know the cause? May 05 14:06:00 koen: glib needs some locale support in printf I assume May 05 14:09:48 koen: ok, --enabling-included-printf solves that. May 05 14:10:03 'mtn pull' as well :) May 05 14:11:30 03koen 07org.oe.dev * r01934ad0... 10/ (1 packages/glib-2.0/glib.inc): glib.inc: make printf disable opt-in May 05 14:19:15 koen: with that commit I can check the incoming diff approach May 05 14:20:33 :) May 05 14:21:15 koen, I hate to nag, but I'm just wondering if my email went though to you or have I somehow managed to goof it up? :) (tried that few hours ago, but you probably didn't notice) May 05 14:21:26 which mail? May 05 14:21:51 (so I guess 'no') May 05 14:22:16 * koen ponders about the poodle pointercals in http://linuxtogo.org/gowiki/AngstromPointercals#preview May 05 14:22:44 Hmm... koen the monotone key one? May 05 14:23:24 oh, that one May 05 14:23:30 I'm waiting for mickeyl May 05 14:23:33 ah, ok. May 05 14:23:40 Just wanted to know if I sent it alright. May 05 14:24:08 never knew for sure if the address on monotonephrasebook is valid. May 05 14:24:25 if people didn't edit it, they are May 05 14:24:27 this shows incoming diffs between a pull and an update: mtn diff -r `mtn automate get_base_revision_id` -r `mtn automate heads` May 05 14:24:40 koen, Well, now that I know you've got mail, I think they are :) May 05 14:25:12 likewise: could you add that to the phrasebook? May 05 14:25:37 likewise, how about getting a diff between pull and working tree? one -r less? May 05 14:26:17 I mean from working tree to head that is. May 05 14:26:40 koen: I think USE_NLS_glib-2.0="yes" ===> --enable-included-printf=yes would work in all cases. May 05 14:27:12 polyonymous: I think that is what I meant. May 05 14:27:43 likewise, but does it take your local modifications in account? May 05 14:27:55 polyonymous: ah no. wait. May 05 14:36:20 koen: yes will add that to the phrasebook May 05 14:38:45 polyonymous: mtn diff -r `mtn automate heads` will show the workspace against the recent pull, but that's a strange thing to do, as incoming changes will be shown as reversed, while your own will not. May 05 14:39:27 polyonymous: you're better off showing the incoming changes and outgoing changes seperately, I guess, as patched forward to their common ancestor. May 05 14:56:12 03koen 07org.oe.dev * rad0a43b0... 10/ (4 files in 2 dirs): chrpath: add cross version May 05 14:56:31 hmm, making packaged-staging package-agnostic seems to be more work than expected May 05 14:56:34 hey zecke|studying May 05 14:59:26 * koen beats head against desk May 05 14:59:55 * koen stabs thunderbird autocomplete May 05 15:02:07 likewise, Well, I think in revers is already a good deal of useful info, no matter how odd the idea sounds. May 05 15:03:18 why doesn't clicking on an entry in thunderbird work? May 05 15:03:35 koen, what entry? May 05 15:03:43 in autocomplete May 05 15:03:48 ahh May 05 15:03:59 if I click on it it will close the popup and *not* select the entry May 05 15:04:13 hmm... seems to work for me. May 05 15:04:33 on OSX I have to double click real fast May 05 15:05:15 seems to work in 1.5.0.10 (on plain old linux) May 05 15:05:41 I suspect it's the old block-gui-on-everything bug May 05 15:06:25 somebody can build glibc-2.5-r5? May 05 15:06:56 on x86_64 May 05 15:07:12 yes May 05 15:07:13 r5? May 05 15:07:16 yep May 05 15:07:17 it's at r6 now May 05 15:07:30 hmm... May 05 15:07:31 I have both r5 and r6 built, according to my tmp/work/arm*/ May 05 15:07:34 need update database May 05 15:07:36 hmm, gcc-cross built May 05 15:07:53 Crofton, how about just gcc? :) May 05 15:07:59 libc6_2.5-r6_arm-oabi.ipk May 05 15:08:00 libc6_2.5-r6_armv5te.ipk May 05 15:08:00 libc6_2.5-r6_armv5teb.ipk May 05 15:08:09 Crofton|home: yes May 05 15:14:24 * likewise is testing the --disable-rpath configure option of gcc... May 05 15:14:31 RP, hrw|gone: http://lists.linuxtogo.org/pipermail/openembedded-devel/2007-May/002090.html May 05 15:17:03 * likewise notices that gcc's --disable-rpath does not help. May 05 16:11:54 koen: the rpath in cross gcc is due to gcc-4.1.x/ldflags.patch. I think that is correct and that this RPATH is OK. May 05 16:12:53 koen: you leaked information :) May 05 16:12:53 I have failed to build madplay :) May 05 16:13:33 zecke|studying: yes, I'm a moron May 05 16:13:55 koen: yes, but now you are forced to install QtE on yiur hx47xx ipaq :) May 05 16:14:10 I have qt3 installed for lyx :) May 05 16:14:21 koen: you should have this two pages long corporate disclaimer at the end of each mail May 05 16:17:21 Crofton|home: why, rpath issues? :-) May 05 16:17:27 yeah May 05 16:20:13 likewise: yes, the ldflags patch is correct (I found that out the hard way earlier) May 05 16:20:44 likewise, zecke|studying: so should we special case gcc-cross in insane.bbclass? May 05 16:21:42 koen, zecke|studying: re-adding the /work suffix in the test will make it bypass any cross tooling again, which I think is good. May 05 16:22:13 (although I now have to learn why other toolchains do not per se link to their own libs). May 05 16:22:45 hi all May 05 16:23:09 koen: May 05 16:23:10 - if bad_dir_test in line: May 05 16:23:12 + if bad_dir in line: May 05 16:23:35 would be my proposed fix for now, until we figure out if cross tooling need to RPATH to it's own libs. May 05 16:25:04 ./configure --help has this option May 05 16:25:15 -disable-rpath do not hardcode runtime library paths May 05 16:25:20 for madplay May 05 16:25:43 Crofton|home: interesting, normally "disable" is default. May 05 16:26:24 03koen 07org.oe.dev * r896c4be5... 10/ (1 classes/insane.bbclass): May 05 16:26:24 insane.bbclass: on check for references to WORKDIR in RPATH, instead of WORKDIR and STAGING May 05 16:26:24 [18:23] likewise: would be my proposed fix for now, until we figure out if cross tooling need to RPATH to it's own libs. May 05 16:27:20 EXTRA_OECONF? May 05 16:27:20 likewise: I still support the idea of building in a faked cross-chroot :) May 05 16:29:00 * koen bangs head against desk again May 05 16:29:05 * victor_rx1950 (wall) May 05 16:29:23 need two harddisks more ^( May 05 16:31:33 03koen 07org.oe.dev * rc7fc7834... 10/ (1 classes/insane.bbclass): insane.bbclass: repair last commit May 05 16:32:12 adding the --disable-rpath to EXTRA_OECONF did not solve the problem :( May 05 16:33:33 how do I search for RPATH in a binary? May 05 16:33:39 scanelf May 05 16:33:50 Is it just me or is it not: Package gcc-cross-locale-ru (4.1.2-r3) is installed on root and has the following files: May 05 16:33:50 /home/hacker/OE/angstrom/tmp/cross/share/locale/ru/LC_MESSAGES/gcc.mo May 05 16:35:17 -rw-r--r-- root/root 437802 2007-05-03 14:42 ./data/build/koen/OE/build/tmp/angstrom/cross/share/locale/ru/LC_MESSAGES/gcc.mo May 05 16:35:44 koen, on device that is. May 05 16:36:03 that doesn't matter for the contents of the .ipk May 05 16:36:43 what do you mean? the file /home/hacker/OE/... is packaged in ipk. May 05 16:36:51 And I'd think this is wrong. May 05 16:37:55 -rw-r--r-- 1 hacker hacker 437802 May 3 12:10 tmp/work/armv5te-angstrom-linux-gnueabi/gcc-cross-4.1.2-r3/install/gcc-cross-locale-ru/home/hacker/OE/angstrom/tmp/cross/share/locale/ru/LC_MESSAGES/gcc.mo May 05 16:40:00 and over here /data/build gets packaged May 05 16:40:35 I'm not sure I understand you. May 05 16:40:58 Crofton|home: if a userspace binary has a bad RPATH, this can also be due to a library dependency. The library libtool could for example be bad as well. May 05 16:42:42 Crofton|home: foobar/rootfs$ scanelf -B -R -r -F '%F has RPATH %r' * | grep -v 'has RPATH -' May 05 16:43:09 Crofton|home: that's three spaces between 'RPATH' and '-' May 05 17:05:04 HopsNBarley: hello there May 05 17:05:29 * likewise about to hit the supermarket May 05 17:06:55 hi likewise - how's it going? May 05 18:05:39 likewise, ping? May 05 18:05:54 madplay and gnu_tls both fail on RPATH May 05 18:06:06 both use libtool to do linking May 05 18:06:19 libtool --mode=link May 05 18:06:50 the link line contains a rpath-link, no rpath May 05 18:25:37 Who is the libtool guru? May 05 18:26:52 koen, ping May 05 18:27:16 Crofton|home: pong May 05 18:27:32 Who would be the libtool guru? May 05 18:27:59 I have no idea May 05 18:28:28 I knew you would asy that .. May 05 18:28:36 we have libtool 1.5.10 May 05 18:28:48 the libtool website has 1.5.22 May 05 18:53:59 Crofton|home: pong, read the backlog. I'm not a autoconf/automake/libtool/configure guru myself. (Damn, forgot to order the books on them with my Amazon order just a min ago..) May 05 18:54:19 "read" as in past-tense. May 05 18:54:28 I am windering if there is a pattern here May 05 18:54:44 and I am wondering if updating to the latest libtool will change things May 05 18:54:57 Crofton|home: but libtool gives back no rpath, only rpath-link, which is harmless May 05 18:55:31 look at the do_compile log May 05 18:55:49 will do, let me see May 05 18:59:16 looks like libtool was called in link mode with rpath-link, and then it adds the rpath to the real link command May 05 18:59:46 RFC: proposing in conf/bitbake.conf: EXTRA_OEMAKE += " -w" May 05 19:02:05 "-w Print a message containing the working directory before and after May 05 19:02:05 other processing. This may be useful for tracking down errors May 05 19:02:05 from complicated nests of recursive make commands. May 05 19:02:06 " May 05 19:02:15 that one? May 05 19:02:54 related to that, I still need a volunteer to experiment with -pg to see if our (stripped) binaries get bigger May 05 19:03:23 koen: yes May 05 19:03:53 koen: curious, bigger compared to what? May 05 19:04:25 to the situation now May 05 19:04:26 Crofton|home: If a add crosstools to my PATH, then run that libtool line manually from the madplay src dir, I get an error. May 05 19:04:46 if it doesn't make a difference, compiling with -pg would make debugging and profiling a lot easier May 05 19:06:09 Crofton|home: false report, I screwed up my PATH... :-/ May 05 19:06:39 koen: maybe we should have a DEVELOPER_DEBUG mode in OE for these options? May 05 19:07:31 koen: ah, maybe the -pg symbols end up in the -dbg packages, is that a workable solution? May 05 19:09:09 wb zecke May 05 19:10:25 yep this is bug with glibc May 05 19:12:39 likewise: that's the question :) May 05 19:13:54 koen: ok, I see :-) May 05 19:14:32 koen: my current todo stack is efika 2.6.21 kernel, madplay rpath, -pg (FIFO ordering) May 05 19:14:46 I'd like proper function names and line number is gdb May 05 19:14:51 koen: and LIKEWISE_PARALLEL=2 May 05 19:15:07 koen: use -dbg? May 05 19:15:15 yes May 05 19:15:28 psokolovsky: http://bugs.openembedded.org/show_bug.cgi?id=2054 =( May 05 19:15:30 that gives me function names most of the times May 05 19:15:39 psokolovsky: same problem May 05 19:21:47 Crofton|home: work/ppc603e-angstrom-linux/madplay-0.15.2b-r0/madplay-0.15.2b/powerpc-angstrom-linux-libtool has hardcode_into_libs=yes, strange since the --rpath-disable would typically make this no May 05 19:22:43 likewise: sed -i -e s:hardcode_into_libs=yes:hardcode_into_libs=no:g ${TARGET_SYS}libtool in autotools.bbclass ? May 05 19:22:58 (or something similar that is correct sed) May 05 19:23:14 I think we should move to libtool 1.5.22 May 05 19:23:17 koen: good point. first let's grep where this occurs. May 05 19:23:23 first and see if that changes behavior May 05 19:23:46 Crofton|home: volunteering? ;-) May 05 19:23:57 copying bb files atm :) May 05 19:24:09 I'll see what patches fail May 05 19:24:19 * likewise crossing fingers (and toes) May 05 19:37:07 Crofton|home: libtool is a can of worms, since they rewrite it every monday May 05 19:37:21 heh May 05 19:37:32 fighting patches now ... May 05 19:38:29 * koen touched OE's libtool 2 years ago and is stil recovering May 05 19:43:42 koen: hardcode_into_libs alone seems not enough... May 05 19:44:05 drat May 05 19:49:33 diff -Nurd? May 05 19:50:35 Crofton|home: ? May 05 19:50:46 Crofton|home: as the preferred options to diff? yes. May 05 19:50:48 or -Naur, but Nurd is easier to remember May 05 19:50:49 trying to remember how to make a patch May 05 19:51:45 03polyonymous 07org.oe.dev * r17540e9f... 10/ (1 packages/gpe-autostarter/gpe-autostarter_0.12.bb): gpe-autostarter: fix compilation with recent linux-linux-headers, closes #2176 May 05 19:51:49 03polyonymous 07org.oe.dev * r01df3eb8... 10/ (1 packages/prismstumbler/prismstumbler_0.7.3.bb): prismstumbler: fix build against motif and recent linux-linux-headers, closes #2177 May 05 19:53:15 bother May 05 19:53:28 how do I get diff to ignore files that are only on one of the dirs? May 05 19:55:49 don't pass -N ;-) May 05 19:56:34 Is there any plans at all to clean up the packages/ dir? May 05 19:56:56 NAiL: clean up in what way? May 05 19:57:19 dunno, make "ls" actually usable in the packages directory? May 05 19:57:46 We can't really do that due to BBFILES May 05 19:57:54 we could group stuff a bit better May 05 19:58:14 (gpe-* -> gpe/, opie-* -> opie/, gnome-* -> gnome/, etc) May 05 19:59:46 yeah May 05 20:00:15 well, gnome is only two directories. gnome and gnome-themes May 05 20:00:32 but still, doing that, there's more than 1200 entries in that directory May 05 20:00:59 we could add an alphabet layer May 05 20:01:09 a/, b/, c/ May 05 20:02:01 what about SECTION= May 05 20:02:02 ? May 05 20:02:12 It's just driving me nuts May 05 20:02:19 our sections are still a mess and random May 05 20:02:31 mmh May 05 20:03:30 that would be a great thing to split up May 05 20:03:40 indeed May 05 20:03:48 categorize it/put it in sections somehow May 05 20:04:12 as a mid to long term project goal perhaps ;) May 05 20:04:14 we could do the debian alphabet thing May 05 20:04:24 koen: well we can complicate BBFILES May 05 20:04:26 a/, b/ ... liba/ libb/... z/ May 05 20:04:30 i guess theres major breakage involved? May 05 20:04:39 but it gives no benefits except for filesystem performance, maybe. May 05 20:05:11 polyonymous: filesystem performance, trac performance (ever tried browsing the packages dir through trac? Don't) May 05 20:05:36 NAiL, No, I haven't :) I thought the issue was making 'ls' usable :) May 05 20:05:39 aye it kills our server for about 10 seconds when someone hits that dir May 05 20:05:50 polyonymous: that's one of the issues May 05 20:06:28 NAiL, ah ok. Well, I think categorization, would make more sense. Just my very personal opinion. May 05 20:06:43 Anyway, it's bath time, courtesy of opie-reader it may take a while :) May 05 20:06:49 Well, I'm mostly satisfied any way it's done. May 05 20:07:10 timtimred: xfs? May 05 20:07:31 zecke: wouldn't it basically being saying /packages/*/*/*.bb instead of /packages/*/*.bb? May 05 20:07:43 koen: no, it's ext3. May 05 20:07:57 NAiL: yes May 05 20:07:57 xfs is horrible with the packages dir May 05 20:08:09 thanks psokolovsky, I guess I could rad the man page May 05 20:08:16 NAiL: what is with nonworking, but basicly yes May 05 20:08:57 zecke: hmm... good point. May 05 20:09:31 if done like the debian alphabet thing, add a nonworking dir in each? May 05 20:12:58 "There's plenty of older hardware that doesn't have the processing power to do WPA, and has to rely on WEP. This is especially true for embedded devices (like print servers and bar code scanners) and PDAs." May 05 20:13:05 why do I still read slashdot? May 05 20:18:29 Crofton|home: the libtool in madplay, line 2721 is to blame for the rpath. May 05 20:19:11 I have libtool-native building May 05 20:21:02 If I get the new libtool building, should I push with DEF_PREF set to -1 so likewise can look at it? May 05 20:21:13 yes May 05 20:22:05 curse the number of bb file May 05 20:22:26 grep DEFAULT */*bb errors with to many args now May 05 20:22:34 03koen 07org.oe.dev * rd6b64846... 10/ (1 packages/tracker/tracker_0.5.4.bb): tracker: add file to DEPENDS (for magic.h) May 05 20:25:43 ok, still have bad RPATH May 05 20:26:00 I will go ahead and push newer libtool for testing purposes May 05 20:26:34 likewise: http://lists.freedesktop.org/archives/cairo/2007-May/010543.html May 05 20:27:02 Crofton|home: use DEFAULT_PREFERENCE="-1" on it, I guess? May 05 20:27:23 ah yes, see you wrote that... May 05 20:27:24 yes May 05 20:27:46 Meanwhile, in a secret room, likewise is debugging the current libtool script. May 05 20:27:55 If we try and pass fix upstream, we should use current libtool May 05 20:28:02 Agreed. May 05 20:28:18 hopefully the fix is the same in current :) May 05 20:28:19 I just want to learn why madplay goes wrong, and most of the others fail. May 05 20:28:33 I think fixing madplay will fix gnutls May 05 20:28:34 s/fail/succeed/ - sigh May 05 20:28:45 sice both link using libtool --mode=link May 05 20:29:34 yes, but libtool makes a decision on when to include the rpath, somewhere around line 2720, and I still do not see why other packages have no rpath. May 05 20:29:53 maybe they do not use libtool for linking May 05 20:29:56 is my guess May 05 20:30:13 I am not sure why madplay uses libtool May 05 20:37:15 03crofton 07org.oe.dev * rbcb26d45... 10/ (13 files in 3 dirs): May 05 20:37:15 libtool : Add bb files for libtool-1.5.22. DEFAULT_PREFERENCE set to -1 May 05 20:37:15 until these are better tested. May 05 20:39:55 zecke/koen: but you don't think splitting it up is a bad idea? May 05 20:40:27 ok, I am rebuilding with libtool-1.5.22 May 05 20:41:01 bbl May 05 20:41:22 NAiL: I have no strong opinion about it May 05 20:41:41 NAiL: I want to do it since the very beginning :} May 05 20:42:03 well then May 05 20:42:14 I'll send a mail to the list and see May 05 20:42:53 night May 05 20:42:58 nite May 05 20:54:23 does anyone run beryl? My computer locks up hardcore, beryl CPU usage jumps to 86% when I run bitbake nano, and have beryl running May 05 20:59:16 summatusmentis, what happens when you run find / ? May 05 20:59:29 heh May 05 20:59:49 polyonymous: I'm confused, what do you mean? May 05 21:00:19 summatusmentis, `find /` is a shell command. I wonder if beryl hogs cpu when you run it in terminal. May 05 21:00:20 summatusmentis: I'll give you a clue. This is #oe.. May 05 21:00:34 That's also very much so :) May 05 21:01:01 NAiL: I know it's #oe. however, emerge -DNatuv world doesn't have this happen, which makes me wonder if it's something in the way OE operates May 05 21:01:12 beryl really doesn't like cpu-intensive tasks. However, your system shouldn't lock up May 05 21:01:12 polyonymous: seemingly not May 05 21:01:52 summatusmentis, well, I've also noted that my plain old x-windows setup doesn't do it neither with emerge, nor with paludis or bitbake :) May 05 21:02:06 CoreDump|home: that's what I thought May 05 21:02:22 polyonymous: meh... plain onld x-windows isn't pretty-fied :-D May 05 21:02:39 summatusmentis, well, I find it pretty much fied. May 05 21:02:52 * CoreDump|home can't stand a 2-D WM after using beryl for some time May 05 21:02:58 ^^ May 05 21:03:02 CoreDump|home, seriously? May 05 21:03:12 polyonymous: yeah May 05 21:03:32 My WM is somewhat 3D, because my second head is at an angle with the first one. May 05 21:03:37 I guess I'm gonna potentially have to decide if I want to dev or have pretty X. Maybe switch back and forth May 05 21:04:01 Can't imagine why would I more 3d in my wm... Gotta try it one day. May 05 21:04:35 polyonymous: it's the little things you get used to =) May 05 21:04:40 it's nice, assuming you don't go over the top w/ the settings May 05 21:04:51 CoreDump|home, that's why I'm afraid I will never be able to part with fvwm :) May 05 21:05:11 I've tried to change it on many wms, but always got back :) May 05 21:05:17 that's what I thought for my beloved windowmaker as well =D May 05 21:06:09 CoreDump|home, well, I need to try, of course, before I try to defend my point any further :)) May 05 21:06:31 trying doesn't hurt, but to each his own of course May 05 21:06:50 right. May 05 21:09:02 Crofton|home, koen: interesting background read, well written and deep: http://wiki.debian.org/RpathIssue May 05 21:12:36 koen, I think it was you, who committed prismstumbler and gpe-autostarter? It seems you forgot to add *.patch May 05 21:14:31 (that is - in addition to modifying .bb files) May 05 21:16:25 polyonymous: summatusmentis: This is beryl on my notebook -> http://hentges.net/tmp/screenshots/AIGLX/aiglx_20070116.avi May 05 21:17:24 I'm using aiglx too May 05 21:20:13 CoreDump|home, transparency is the only think I find useful here. Besides, you have Angstrom MLs in OpenZaurus folder! May 05 21:20:24 lol, right May 05 21:39:45 koen, think you haven't got it: polyonymous koen, I think it was you, who committed prismstumbler and gpe-autostarter? It seems you forgot to add *.patch May 05 21:41:15 03koen 07org.oe.dev * rb5599083... 10/ (3 files in 3 dirs): gpe-autostarter: add missing patch May 05 21:41:19 03koen 07org.oe.dev * r30f2bd37... 10/ (1 packages/prismstumbler/prismstumbler-0.7.3/wireless.patch): prismstumbler: add missing patch May 05 21:41:26 ah ok :) May 05 21:42:17 I could get used to OM's revision checker thingy May 05 21:42:25 !praise svnnow May 05 21:42:26 * cdbot2 bows down to svnnow and chants, "I'M NOT WORTHY!!" May 05 21:42:33 what's svnnow? :) May 05 21:42:52 it's a version tag May 05 21:43:06 uhm... May 05 21:43:11 you mean the fetcher improvement, not svnnow May 05 21:43:20 combined w/ the OM patch for bitbake, packages w/ "svnnow" will only e rebuilt when their SVN revision changes May 05 21:43:38 I think I fixed building gcc for arm on amd64. May 05 21:44:06 At least in terms of building the powerful helloworld tool on Z. May 05 21:44:24 !praise polyonymous May 05 21:44:26 * cdbot2 bows down to polyonymous and chants, "I'M NOT WORTHY!!" May 05 21:44:47 cdbot2 itn't worthy. Where is this world going now. May 05 21:46:20 polyonymous: Hell May 05 21:46:22 ? May 05 21:46:31 wait we are allready there May 05 21:46:35 Hell. (o, world!) May 05 21:47:03 heh May 05 21:49:25 polyonymous: what was broken then? (building for arm on amd64 here for ages) May 05 21:49:35 likewise, building gcc? May 05 21:50:07 polyonymous: sorry, I'm staying up too late. target gcc, right. May 05 21:50:12 !oebug 1951 May 05 21:50:13 * * Bug 1951, Status: REOPENED, Created: 2007-03-07 15:45 May 05 21:50:15 * * nail(AT)nslu2-linux.org: gcc fails to build on x86_64 May 05 21:50:15 * * http://bugs.openembedded.org/show_bug.cgi?id=1951 May 05 21:51:23 ok May 05 21:52:09 I almost never cross-build target toolchains, my targets are too small :-) May 05 21:52:41 likewise, I wouldn't say that mine is really big one, but I just wanted to do it for the sake of completeness :) May 05 21:52:53 polyonymous: very well May 05 21:53:15 sometimes it's so relaxing to compile helloworld when you're on the road :) May 05 21:56:20 submitted a patch. May 05 22:09:10 polyonymous: Could you describe the problem you fixed and how you fixed it in the header of the patch please? We have a hell of a time maintaining our older toolchain patches, and such a comment might help us in the future. May 05 22:09:58 likewise, the patch is, basically, a port of the older one in the same bug. May 05 22:10:58 polyonymous: ah yes, comment #3 should be the patch header, more or less, when we commit it. May 05 22:11:11 likewise, yes, that would make sense. May 05 22:12:45 But the maintenance hell is, I'm afraid, due to something wrong with the approach to the whole toolchain thing. Not that I am able to properly elaborate on my thoughts on the matter or propose any improvement at the moment. May 05 22:46:02 polyonymous: it's very entangled, and fixes/works around a lot of issues. May 05 22:47:47 likewise, true. May 05 22:58:37 03likewise 07org.oe.dev * r256d1626... 10/ (1 packages/db/db_4.2.52.bb packages/db/db_4.3.29.bb): db: Use rm -rf instead of rmdir (which broke re-installation). May 05 22:59:24 hmm... no http support in subversion on the device... May 05 23:02:26 well, goodnight everyone May 05 23:04:56 goodnite May 05 23:05:43 'night all May 05 23:37:50 hi all, is anyboy still awake? May 05 23:38:04 I have a question about OE and icecc May 05 23:39:20 It seems that the only configuration supported is to have two (or more) identical CPUs May 05 23:39:25 Is this correct? May 05 23:43:57 well, night all. May 06 02:51:38 I'm using aiglx too May 06 02:51:52 ^^ sorry, wrong window May 06 02:52:26 wil someone look at this error for me? I ctrl+c'd the job once, and this is the following attempt http://pastebin.ca/473157 **** ENDING LOGGING AT Sun May 06 02:59:56 2007