**** BEGIN LOGGING AT Wed Jun 12 02:59:58 2013 Jun 12 03:08:06 otavio: http://autobuilder.yoctoproject.org/pub/nightly/20130611-2/machines/imx6qsabresd/ Jun 12 03:08:14 should be all there in a moment. Jun 12 03:08:19 I'll patch it in the morning Jun 12 03:08:35 pidge: found the root cause? Jun 12 03:11:05 yeah Jun 12 03:11:17 misspelled the board name Jun 12 03:11:41 missing the q Jun 12 03:30:38 heh Jun 12 03:30:51 is it possible to convince the git fetcher to fetch from a private repo on github? Jun 12 03:51:01 ;protocol=git+ssh might work, so it passes along a git+ssh:// uri Jun 12 04:10:50 kergoth: thanks, will give that a try Jun 12 07:46:23 good morning Jun 12 08:11:21 morning all Jun 12 08:15:39 mornin Jun 12 08:20:20 hi bluelightning Jun 12 08:20:26 hi ant_work, panda84kde Jun 12 08:21:40 for anybody wondering how to deal with strange licences, here's the reply I recived. https://lists.yoctoproject.org/pipermail/yocto/2013-June/016542.html Jun 12 08:23:21 I also replied :) Jun 12 08:23:34 but Beth had the key bit Jun 12 08:25:12 bluelightning: Oh, you're Paul. Nice to meet you! :) Thanks. That part was already clear to me, however thank you for your contribution. Jun 12 08:25:53 ok, I just figured you asked about LICENSE so that ought to be covered as well :) but good that you got the information you needed Jun 12 09:35:00 does anyone have a mips or ppc environment with a sdk to hand? Jun 12 09:35:14 (or be able to tell me alignment values for a given type) Jun 12 09:51:48 rburton: you mean you didn't build qemu images overnight? ;-) Jun 12 10:20:15 https://www.linux.com/news/featured-blogs/200-libby-clark/723487-qaa-u-boot-creator-wolfgang-denk-on-the-state-of-embedded-linux/ Jun 12 10:20:24 Yocto praise from mr Denk Jun 12 10:24:11 hi bluelightning, Jun 12 10:24:18 hi fahad Jun 12 10:24:19 i just sent a patch for libmtp Jun 12 10:24:34 fahad: indeed, yes Jun 12 10:24:42 will it be okay to send it to meta-oe/recipes-connectivity Jun 12 10:24:53 thanks for your response Jun 12 10:25:38 RP: i did download the sdk images, booting one now Jun 12 10:25:40 or maybe, recipes-extended Jun 12 10:39:25 fahad: recipes-connectivity should be OK Jun 12 10:39:27 fahad: thanks Jun 12 10:40:03 bluelightning: thanks Jun 12 12:57:45 anyone got a kernel git clone to hand? Jun 12 12:58:33 curious as to what release 037a833ed05a86d01ea27a2c32043b86c549be1b was shipped in Jun 12 12:58:42 rburton: not latest sync, but it's a clone Jun 12 12:58:56 $ git describe 037a833ed05a86d01ea27a2c32043b86c549be1b Jun 12 12:58:56 v3.5-rc5-79-g037a833 Jun 12 12:59:08 thanks Jun 12 12:59:17 so thats probably why my debian with ancient 3.2 doesn't work Jun 12 12:59:54 * rburton boots his fedora18 machine, turns out it has a use after all Jun 12 13:02:17 woo works Jun 12 13:02:32 walters: i don't suppose you know if f18 has a touchscreen calibrator thingy? Jun 12 13:02:48 no idea sorry Jun 12 13:02:59 np, i'll dig out xinput-calibrator Jun 12 13:03:16 oh, no point, i can't actually connect both video and usb at once on this machine Jun 12 13:03:21 ffs, cursed. Jun 12 13:17:13 ok, i'm lost in kernel configuration again. how can i easily turn on a kernel config option in linux-yocto? Jun 12 13:17:21 literally just one option i want to be a "m" Jun 12 13:19:00 zeddii: ^ Jun 12 13:22:21 put it into "foo.cfg" Jun 12 13:22:26 and add foo.cfg to your SRC_URI. Jun 12 13:29:48 Hello. Jun 12 13:30:34 seebs: fsarbu talked to you a couple of days ago about that strange issue with udev-hwdb Jun 12 13:31:06 we basicaly managed to find the cause of this issue: having dash as sh Jun 12 13:32:29 Is there anybody who can help me understand why if a dependency is updated, the packages depending in it are not recompiled? Jun 12 13:32:38 on* Jun 12 13:33:21 Bug: icu updated, harfbuzz not recompiled (still has references to icu 50). Pango fail as it links with harfbuzz. Jun 12 13:33:33 fails* Jun 12 13:50:42 zeddii: easier than i thought :) Jun 12 14:47:49 how can I obtain the rootfs as a cpio.gz archive ? Jun 12 14:48:50 hmmm Jun 12 14:48:53 I know that Jun 12 14:49:40 IMAGE_FSTYPES ?= "cpio ext2 ext2.gz.u-boot ext2.gz" Jun 12 14:49:53 maybe cpio.gz? Jun 12 14:52:59 and this should be adde din my "meta-custom/conf/machine/qemux86.conf" file Jun 12 14:53:00 ? Jun 12 14:53:37 try local.conf Jun 12 14:53:53 IMAGE_FSTYPES = "cpio cpio.gz" Jun 12 14:54:21 thank you. will try this now Jun 12 14:59:58 agherzan: It probably did get recompiled however it probably didn't correctly relink Jun 12 15:00:19 agherzan: This is why we made the S != B changes so try and correctly rebuild software Jun 12 15:00:52 agherzan: offhand I don't know if harfbuzz supports that or whether you'll have S != B enabled (depends on which distro you chose) Jun 12 15:01:23 RP indeed. It was recompiled. But it was a no-op log. Jun 12 15:03:27 RP I know I can recompile it cleaning it first. But this is not an option as I'm using a build server which should support incremental builds... Jun 12 15:06:27 agherzan: which DISTRO are you using? Does it use the S != B code Jun 12 15:17:51 RP nope. O have a custom one. But why isn't this fixed in the package itself? Jun 12 15:17:58 I* Jun 12 15:19:29 agherzan: Its a problem with autotools :/ Jun 12 15:19:48 agherzan: add the include to turn on b!=s, it's lovely Jun 12 15:21:39 rburton RP got it. Will try Jun 12 15:21:41 Thanks Jun 12 15:22:06 agherzan: I figured out that much and then added http://git.yoctoproject.org/cgit.cgi/poky/tree/meta/conf/distro/include/seperatebuilddir.inc Jun 12 15:24:15 RP yes. Harfbuzz is there. Jun 12 15:24:20 Thanks a lot Jun 12 15:25:06 agherzan: the gotcha is that you'll need to do a build from clean, autotools doesn't like doing a B!=S build in a previously configured directory :( Jun 12 15:25:25 Aaaaaaaa Jun 12 15:25:39 Need to clean the server right? Jun 12 15:25:55 sadly, yes. Jun 12 15:26:12 only for the packages that use B!=S, but that's quite a long list :( Jun 12 15:26:19 Shoot. But ok. As this will result in more stable incremental builds. Jun 12 15:27:53 Now, the obvious question. Why would this fail in a b=s and not in a b! =s Jun 12 15:27:58 ? Jun 12 15:28:38 because with b!=s, we can blow away the build directory when rebuilding. otherwise, .la files can survive a reconfigure and old link information persists Jun 12 15:28:42 evil evil evil libtool Jun 12 15:29:03 (need to bring up the patches to kill .la files in $libdir again) Jun 12 15:56:10 otavio, I'm available to look at outgoing mail now. I'll check the logs for the information you sent earlier. Jun 12 16:21:11 anyone have any tips for flushing out race conditions in a corpus of recipes? (e.g. they don't call out their dependencies correctly, so they succeed or fail depending on the timing of populate_sysroot tasks) **** ENDING LOGGING AT Wed Jun 12 16:48:41 2013 **** BEGIN LOGGING AT Wed Jun 12 16:50:19 2013 Jun 12 16:50:25 i'd say inspect the source, read the readme, read the configure.{in,ac}, see what they actually look for and under what circumstances. **** ENDING LOGGING AT Wed Jun 12 16:52:05 2013 **** BEGIN LOGGING AT Wed Jun 12 16:54:22 2013 Jun 12 17:54:01 I'm looking for a Directfb GUI or demo Jun 12 17:54:12 anyone has any starting point? Jun 12 18:09:42 Anyone have any thoughts on my creating a more general framework for automatic dep stuff as a part of the work on Bug 4451? E.g. hooks for writing and reading the paths without duplicating the pattern from shlibs code, or something? Jun 12 18:09:42 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=4451 enhancement, Medium, 1.5, kergoth, IN PROGRESS IMPLEMENTATION , Feature request: automatic rdepends based on python dependencies Jun 12 18:09:46 * kergoth ponders Jun 12 18:20:11 while fetching gcc-cross-initial, i was expecting bitbake to find it in a shared sstate-cache and therefore just retrieve it from cache. it doesn't, and bitbake-diffsigs reports differences with all the patch files in the SRC_URI. The specific message is like "Dependency on checksum of file ${TOPDIR}/meta/recipes-devtools/gcc/gcc-4.7/avoid-oob-array-access.patch was [added|removed]" Jun 12 18:20:32 should this be the case? Jun 12 18:20:36 sounds like the patches themselves changed then. it checksums file:// file contents Jun 12 18:21:00 i'll check that Jun 12 18:26:29 hi all Jun 12 18:28:12 got a question: why 64-bit libedit.so is not in build sysroot of yocto? I turned on multilib, and after I upgrade to dylan, only 32-bit libedit is in sysroot Jun 12 18:28:33 kergoth: a diff shows that they are identical Jun 12 18:30:03 this may not be caused by libedit itself, is there any place to say what would be in sysroot, bot both 64-bit and 32-bit? Jun 12 18:33:47 j8: no idea then, afaik it should be going by the file contents, and shouldn't add and remove the dependency unless the checksum changes Jun 12 18:33:50 * kergoth shrugs Jun 12 18:36:29 i did not mention that the ${TOPDIR} varies between the two users Jun 12 18:38:34 that *shouldn't* be a factor, since it should only be including the URI, not the localpath, in the checksum, but if it is, that'd definitely be a bug Jun 12 18:45:26 Pretty quiet in here today for some reason Jun 12 18:47:21 * kergoth ponders Jun 12 18:49:32 nobody knows anything about my question/ Jun 12 18:49:34 ??? Jun 12 18:53:25 your question is unclear. 'build sysroot', which sysroot? do you mean teh sysroot for the build machine, native, not target? and what do you mean by 'turned on multilib', just included the multilib .conf/.inc, or configured it also? and how did you configure it? Jun 12 18:54:33 Unrelated to that, my thoughts on possibly cleaning up the automatic dependency handling: https://gist.github.com/kergoth/5768039 Jun 12 18:54:59 build sysroot means on build host (not target), multilib is turned on like this in local.conf: require conf/multilib.conf MULTILIBS = "multilib:lib32" DEFAULTTUNE_virtclass-multilib-lib32 = "x86" Jun 12 18:55:29 i got link error on "-ledit", therefore I found there is no 64-bit libedit in build/tmp/sysroot/xxx Jun 12 18:55:31 afaik multilib is used for target, and isn't used for native at all Jun 12 18:55:41 could be wrong, but that's my understanding Jun 12 18:56:24 correct, but you need to compile all of those target on build machine, right? Jun 12 18:57:02 actually my question is pretty simple, what controls what in sysroot? Jun 12 18:57:10 kergoth: no, sorry; I mean that my goal is to identify which recipes have dependency races so I can force my team members to fix them. my current approach is to use a very high BB_NUMBER_THREADS and clean-build repeatedly, on the hope that the parallelism increases the probability of such races being triggered. Jun 12 18:57:34 but that's proving inadequate Jun 12 18:58:18 evanp: oh, I see. well, you might be interested in a couple of testing scripts I created for trying to identify such cases. not perfect, but better than nothing. hold a moment Jun 12 18:58:56 evanp: https://github.com/kergoth/oe-test-scripts#readme Jun 12 18:59:03 evanp: specifically greedy-deps-tests may be of interest to you Jun 12 18:59:17 kergoth: thanks very much Jun 12 18:59:30 evanp: it builds a specified target, then cleans and rebuilds everything piecemeal with that populated tmpdir, using buidhistory to compare the packages, watching for modified runtime dependencies. Jun 12 18:59:47 evanp: the example output there is from an actual bug i spotted and fixed using it Jun 12 19:01:43 hmm, might be able to improve this Jun 12 19:01:48 haven't looked at it in some time Jun 12 19:03:22 evanp: if you do end up trying it, let me know how it goes Jun 12 19:03:35 kergoth: ok, will do Jun 12 19:04:55 afaik i'm the only one using these at the moment. trying to identify such problems is rather nontrivial :) Jun 12 19:06:24 we used to have quite a lot of reloc bugs, but i think those are all fixed now. it wouldn't surprise me if some greedy dep bugs still exist though Jun 12 19:50:09 * kergoth ponders Jun 12 20:31:19 rburton: when you have time, can you look at top 10 patches in http://git.openembedded.org/openembedded-core-contrib/log/?h=jansa/mesa before I finish them and send them for review? Jun 12 22:05:55 what the.. NameError: global name 'd' is not defined in expanding a flag Jun 12 22:06:32 * kergoth scratches head Jun 12 22:22:03 holy spamming YoctoAutoBuilder Jun 12 22:22:12 heh Jun 12 22:22:15 that's a lot of failures Jun 12 22:22:25 ah, well the green was nice while it lasted... Jun 12 22:23:17 I guess the buildtools-tarball wasn't reinstated Jun 12 22:51:12 sgw_, ping Jun 12 22:56:17 exit Jun 12 23:01:33 ant_home: what's up? Jun 12 23:09:13 hi Jun 12 23:09:44 I'm embarassed to insist but maybe this pov will convince you Jun 12 23:09:50 http://imgbin.org/index.php?page=image&id=14117 Jun 12 23:11:45 sgw_, I'd finally point out that image_types.bbclass has this reference to cramfs-native that currently isn't provided Jun 12 23:13:13 it's not about introducing (yet)another fs, is only about image generation Jun 12 23:13:27 ant_home: your already adding other layers, why not add meta-oe? Jun 12 23:14:09 to have xserver-nodm-init overriden? Jun 12 23:14:14 no thanks Jun 13 01:06:47 * kergoth ponders Jun 13 01:13:40 WIP: https://gist.github.com/kergoth/5770431 Jun 13 02:26:03 WARNING: Automatic python depends for libnewt-python-python: set(['string', 'types']) Jun 13 02:26:03 WARNING: libnewt-python-python depends on string(python-stringold) Jun 13 02:26:08 WARNING: libnewt-python-python depends on types(python-core) Jun 13 02:26:09 woot. Jun 13 02:27:22 now to add scanning of python scripts in the bin directories, and the version specific dep on python-core based on the schebangs **** ENDING LOGGING AT Thu Jun 13 02:59:59 2013