**** BEGIN LOGGING AT Fri Apr 30 02:59:57 2010 Apr 30 03:00:12 So nothing flash aware, ok, nm Apr 30 03:00:16 our FS guru left us a few months back but I'm sure he'd have suggestions Apr 30 04:37:10 Tartarus: I figured out the problem with gcc for ppc Apr 30 04:37:31 Tartarus: for now the workaround is that dont use -Os while compiling libgcc Apr 30 04:38:19 compiler generates these function calls in prologue with Os and these functions exist in libgcc but they are hidden Apr 30 06:49:57 03Klaus Kurzmann  07org.openembedded.dev * rd2bff322d8 10openembedded.git/recipes/tasks/task-shr-feed.bb: Apr 30 06:49:57 task-shr-feed: add dbus-daemon-proxy Apr 30 06:49:57 Signed-off-by: Klaus Kurzmann Apr 30 07:31:07 morning Apr 30 07:31:33 what would be a smallest gtk based video player available in oe ? Apr 30 08:04:28 good morning Apr 30 08:07:53 hey mckoan Apr 30 08:08:22 03Martin Jansa  07org.openembedded.dev * ra7dce2f6dd 10openembedded.git/ (4 files in 3 dirs): Apr 30 08:08:22 mesa-dri: add 7.8.1 Apr 30 08:08:22 Signed-off-by: Martin Jansa Apr 30 08:08:23 03Martin Jansa  07org.openembedded.dev * ra003dba3b4 10openembedded.git/ (conf/distro/shr.conf recipes/shr/shr-version.bb): Apr 30 08:08:23 shr: add shr-version recipe and drop its generation from ROOTFS_POSTPROCESS_COMMAND Apr 30 08:08:23 Signed-off-by: Martin Jansa Apr 30 08:08:29 03Martin Jansa  07org.openembedded.dev * r8fe3d92f94 10openembedded.git/recipes/images/ (9 files in 9 dirs): Apr 30 08:08:29 shr-image: drop all tweaks Apr 30 08:08:29 * they should really be applied in recipes where they belong Apr 30 08:08:29 * one of reasons why reflashed image can behave different than older Apr 30 08:08:29 after opkg upgrade Apr 30 08:08:29 Signed-off-by: Martin Jansa Apr 30 08:08:29 03Martin Jansa  07org.openembedded.dev * r597586329d 10openembedded.git/conf/distro/include/sane-srcrevs.inc: Apr 30 08:08:30 EFL: bump SRCREV Apr 30 08:08:30 Signed-off-by: Martin Jansa Apr 30 08:08:31 03Martin Jansa  07org.openembedded.dev * r7f934c85b0 10openembedded.git/recipes/shr/ (shr-settings/elementary.sh shr-settings_git.bb): Apr 30 08:08:32 shr-settings: add elementary.sh Apr 30 08:08:32 Signed-off-by: Martin Jansa Apr 30 08:23:22 good morning Apr 30 08:23:28 hi florian Apr 30 08:23:53 hi florian woglinde Apr 30 08:26:15 hi hrw Apr 30 08:28:42 qemu running on ssh x11 forwarding looks fun Apr 30 08:29:08 o.O Apr 30 08:30:03 hrw hm once I made following Apr 30 08:31:43 from windows with cygwin X via ssh -X to an linux machine and from the linuxmachine via rdesktop to another windowsmachine from this windowmachine vi terminalprogram to an ibm ios5 machine Apr 30 08:31:47 hi rp Apr 30 08:32:57 hi woglinde Apr 30 08:33:08 I need to reorder desk and underdesk equipment to connect laptop to 1GbE switch Apr 30 08:33:37 now it is connected to 'devboards switch' which is just 100Mbps Apr 30 08:41:58 arg bitbake vlc pulls in qt ?!? Apr 30 08:42:20 sure Apr 30 08:42:25 vlc has qt frontend Apr 30 08:42:31 ah Apr 30 08:43:04 i see a vlc-gpe -- will that one come without qt ? Apr 30 08:43:14 dont know Apr 30 08:43:20 look into it and it depends Apr 30 08:43:24 k thx .. i test it Apr 30 08:44:05 yep depends look good Apr 30 08:45:45 can you recommend any small gtk based player ? Apr 30 08:46:25 you could try gstreamer Apr 30 08:46:54 which should be independet from graphic widgets Apr 30 08:50:37 rob_w: yes vlc-gpe should be able to live without qt Apr 30 08:50:59 yep .. i just hope it all compiles Apr 30 08:59:32 03Koen Kooi  07org.openembedded.dev * r5671855bfb 10openembedded.git/recipes/memtester/memtester_4.1.3.bb: memtester: add 4.1.3 Apr 30 09:15:58 hm new init system by lennart Apr 30 09:16:30 gm Apr 30 09:36:41 03Martin Jansa  07org.openembedded.dev * recdbd5c326 10openembedded.git/recipes/openvpn/openvpn_2.1.0.bb: Apr 30 09:36:41 openvpn: add recipe for 2.1.0 Apr 30 09:36:41 Signed-off-by: Martin Jansa Apr 30 10:09:19 better? Apr 30 10:09:22 better :) Apr 30 10:11:10 ok, so, I want to change ${SOMEBASE}/linux-omap3-2.6.32-r51/git/arch/arm/mach-omap2/board-overo.c Apr 30 10:11:19 where is the best place to make that change? Apr 30 10:12:16 the instructions I'm following say to copy the whole of overo-oe/org.openembedded/recipes/linux into user.collections, but I'd rather something a little less brute force Apr 30 10:13:20 I'm only bitbaking omap-console-image and would like to override as little as possible... Apr 30 10:13:35 anyone have some sage like advice? Apr 30 11:43:25 03Koen Kooi  07org.openembedded.dev * r0fbb279f86 10openembedded.git/conf/machine/include/omap3.inc: omap3: bump MACHINE_KERNEL_PR Apr 30 11:43:35 03Koen Kooi  07org.openembedded.dev * r8113f70062 10openembedded.git/recipes/u-boot/ (15 files in 2 dirs): u-boot git: update beagleboard patchset Apr 30 11:43:36 03Koen Kooi  07org.openembedded.dev * r78a5b578ca 10openembedded.git/recipes/linux/ (48 files in 2 dirs): linux-omap-psp 2.6.32: update to latest PSP git, rediff patches Apr 30 11:46:36 03Martin Jansa  07org.openembedded.dev * re38c2c1286 10openembedded.git/classes/image.bbclass: Apr 30 11:46:36 image.bbclass: install_linguas don't install empty list Apr 30 11:46:36 * in some combination of IMAGE_LINGUAS installed packages and available translations it's normal to get empty intersection in pending-locale-packages Apr 30 11:46:36 * don't call opkg install in this case, because it will fail Apr 30 12:24:44 khem: Apr 30 13:16:16 what's better place for run-parts? busybox installs it to /bin/run-parts, debianutils to /usr/bin/run-parts, and u-a then fails, I'll move one of them, but not sure which is better Apr 30 13:21:03 03Robert Schuster  07org.openembedded.dev * r6da4ed714f 10openembedded.git/recipes/llvm/ (llvm-native.inc llvm.inc): Apr 30 13:21:03 llvm.inc: Accomodate to changes made to cmake.bbclass which broke the LLVM builds. Apr 30 13:21:03 llvm-native.inc: Dito. Apr 30 13:21:12 03Robert Schuster  07org.openembedded.dev * r6db59c439e 10openembedded.git/recipes/llvm/ (6 files in 2 dirs): Apr 30 13:21:12 llvm.inc: Added path for nm. Apr 30 13:21:12 llvm2.7 2.6+svnr20091007: Removed prerelease version. Apr 30 13:21:12 llvm2.7-native 2.6+svnr20091007: Dito. Apr 30 13:21:12 llvm2.7 2.7: New recipe. Apr 30 13:21:13 llvm2.7-native 2.7: New recipe. Apr 30 13:23:23 03Robert Schuster  07org.openembedded.dev * r90c620ade8 10openembedded.git/recipes/llvm/llvm2.7_2.6+svnr20091007.bb: llvm2.7 2.6+svnr20091007: Removed. Apr 30 13:27:17 JaMa: debian one Apr 30 13:27:40 hrw: ok, thanks Apr 30 13:52:03 03Martin Jansa  07org.openembedded.dev * rf1b326c83b 10openembedded.git/recipes/busybox/ (busybox-1.15.3/run-parts.in.usr-bin.patch busybox_1.15.3.bb): Apr 30 13:52:03 busybox: add patch for installing run-parts to /usr/bin/ instead of /bin Apr 30 13:52:03 * otherwise debianutils install will fail with Apr 30 13:52:03 update-alternatives: Error: cannot register alternative run-parts to /usr/bin/run-parts since it is already registered to /bin/run-parts Apr 30 13:52:04 Signed-off-by: Martin Jansa Apr 30 14:26:39 morning Apr 30 15:25:44 morning kergoth Apr 30 15:48:34 hi. I was wondering if it was possible to stop bitbake from unpacking a tar.bz2 file during its unpacking step. Apr 30 15:51:51 I have an archive file that I include in SRC_URI that I wish to include as-is in my package. At the moment, I'm achieving this by changing the file name but this is counterintuitive to other developers working with the package Apr 30 15:58:08 ashcharles: I don't think that's supported right now. could be added easily enough -- add a url parameter and check for it in do_unpack in base.bbclass. i.e. "nounpack=1" or somesuch Apr 30 16:53:31 thanks kergoth....that worked nicely :) Apr 30 16:53:41 great :) Apr 30 16:54:33 would other people find this useful? If so, do I just send a patch file to the mailing list? Apr 30 17:01:37 ashcharles: definitely send the patch, see how the response is. Apr 30 17:31:37 whats a good BB_NUMBER_THREADS for a quad-core with 3gig Apr 30 17:31:39 8gig Apr 30 17:31:59 Start at 4 threads, -j6 and go from there Apr 30 17:32:00 03Martin Jansa  07org.openembedded.dev * rff17a80304 10openembedded.git/recipes/shr/shr-e-gadgets_git.bb: shr-e-gadgets: bump SRCREV Apr 30 17:32:56 oh PARALLEL_MAKE Apr 30 17:33:31 Yeah, an important one to set too ;) Apr 30 17:33:46 And how big you can grow really depends on the underlying quadcore and such Apr 30 17:34:02 no wonder my cores were idle Apr 30 17:34:10 heh Apr 30 17:52:27 NOTE: generating locale es_NI (UTF-8) Apr 30 17:52:27 mmap: Permission denied Apr 30 17:55:29 is this /proc/sys/vm/mmap_min_addr ? Apr 30 17:55:31 Tartarus: I know the problem about _savegpr_* and _restgpr_* Apr 30 17:55:56 Tartarus: see http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43810 for details Apr 30 17:55:56 khen: Yay Apr 30 17:58:26 Tartarus: I think the minimal solution for us is to compile crtstuff with -O2 Apr 30 17:58:37 Tartarus: I dont think we can change the linking order Apr 30 17:59:21 ok Apr 30 17:59:28 but if we compile crtend.o with -O2 or say without -Os then we wont get accesses to these symbols in crtend.o and we wont have to change the linking order of libgcc Apr 30 18:00:02 alternatively we can also change order of linking but thats more disruptive Apr 30 18:00:08 let me see if I can hatch a patch Apr 30 18:01:13 So hey ashcharles: Do you know what a Signed-off-by line is? Apr 30 18:01:29 Tartarus: SOB? Apr 30 18:01:48 Tartarus: it usually stands for something else... Apr 30 18:01:51 denix0: Yes, that's short hand Apr 30 18:01:58 I'm asking if he knows what it means, rather Apr 30 18:02:04 Good to make sure before having somemone do it Apr 30 18:02:22 ubuntu desktop iso downloaded in 17secs Apr 30 18:02:32 gone are days when it took 3 hours Apr 30 18:03:10 Tartarus: fyi, I'll be fixing some small issue in external-toolchain-csl shortly... Apr 30 18:03:46 denix0: k, what is it? Apr 30 18:04:05 As I think I said before, it worked for me for the CS ppc one too, so long as I got the variables all set right Apr 30 18:04:29 And the siteinfo/etc stuff for dealing with powerpc*linux-gnu Apr 30 18:06:00 hey Tartarus: I guess I should have added a 'Signed-off by: Ash Charles...." to the patch or to the email Apr 30 18:06:14 ashcharles, yeah. Other than that, looked good enough Apr 30 18:06:35 Tartarus: RPROVIDES = "glibc-utils libsegfault glibc-thread-db" - remove glibc-thread-db there, as it prevent installing the actual libthread-db... Apr 30 18:07:09 denix0: Er, that doesn't sound right at first at least Apr 30 18:07:17 What "actual" libthread-db? Apr 30 18:08:09 Tartarus: it is right to attach the patch file or to copy-paste it in the email...I understand there is some issue with some email clients line wrapping things Apr 30 18:08:22 ashcharles, you can just reply to your first email Apr 30 18:09:42 that's the global RPROVIDES for external-toolchain-csl, so all the PACKAGES get Provides: set like that. Then when someone RDEPENDS on libthread-db, the library doesn't get pulled in automatically, as libc and many other already Provide: it... Apr 30 18:10:34 Tartarus, I meant the question in a general way Apr 30 18:10:43 denix0: Ah, that kind of problem Apr 30 18:10:46 OK, makes sense now, thanks Apr 30 18:11:19 ashcharles: in general, email client that doesn't wrap (or lets you insert files directly) and patch desc, S-O-B, patch contents Apr 30 18:13:17 Tartarus: So, I'm not sure global RPROVIDES is really needed there anyway... We end up with packages like "catchsegv" providing glibc-utils, libsegfault, libthread-db1 which is not right :) Apr 30 18:13:35 denix0: Yeah Apr 30 18:13:40 Needs some thought / testing Apr 30 18:14:45 denix0: how can catchsegv provide glibc-utils ? Apr 30 18:14:58 khem: exactly :) Apr 30 18:15:01 is bitbake mad Apr 30 18:16:59 khem: I think the intention was to say entire CSL provides all that, but not the individual packages we generate... Apr 30 18:18:16 Tartarus: should those 3 items instead be in PROVIDES and not RPROVIDES? Apr 30 18:18:24 denix0: look at eglibc recipes Apr 30 18:18:34 how it provided all what glibc provided Apr 30 18:18:53 you should need something similar for external toolchains Apr 30 18:23:32 khem: right, I was thinking to set package-specific RPROVIDES_glibc-utils = "glibc-utils" and so on, but that's not required as we already produce glibc-utils package. Thus it will be picked up by any RDEPENDS = "glibc-utils" just fine... Apr 30 18:25:04 denix0: Yeah, probably PROVIDES not RPROVIDES Apr 30 18:25:09 all in all, the conclusion is that global RPROVIDES for a meta-package is dangerous, as all the subpackages will inherit it... Apr 30 18:25:13 I need to think / test those out sometimes Apr 30 18:27:29 khem: So, e500v2 + gnutls has http://pastebin.com/WcRE6dk0 w/ 4.4.2. Trying 4.4.3 now then 4.5.0 Apr 30 18:31:47 qemu-arm fails with.. Apr 30 18:31:50 mmap(0x8000, 270336, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0) = -1 EACCES (Permission denied) Apr 30 18:35:35 where 3 appears to be the localdef bin Apr 30 18:46:55 this is required: Apr 30 18:46:56 setsebool -P mmap_low_allowed 1 Apr 30 18:47:31 for fedora 12 Apr 30 18:52:16 SELinux is a pita, yeah :) Apr 30 18:53:40 could there be away for bitbake to sanity check that? Apr 30 18:54:00 Yes, probably Apr 30 18:54:04 We used to check for something else Apr 30 18:54:16 git log on classes/sanity.bbclass or classes/qemu.bbclass Apr 30 18:54:29 Would need to redo the logic to check for selinux enabled, and then mmap_low_allowed Apr 30 18:55:13 I thought latest qemu had it fixed, so no low mmap required... Apr 30 18:56:17 i got 0.12.3 Apr 30 18:56:45 it works fine here with 4096 there.. Apr 30 18:57:01 I mean qemu-0.12.3 and 4096 in mmap_min_addr Apr 30 18:57:12 JaMa|GoNe: what distro is that? Apr 30 18:57:16 gentoo Apr 30 18:57:38 but I have report from some other distro that it didn't work there unles mmap_min_addr set to 0 Apr 30 18:57:39 with or w/o SELinux? Apr 30 18:57:56 same git revison, same qemu (provided by OE) Apr 30 18:58:17 here without, the other guy's distro with (probably) Apr 30 18:58:29 I believe gentoo-hardened is the one with SELinux... Apr 30 18:59:45 I don't know what distro he had.. but it failed for him.. then he set it to 0 with sysctl and then locales were built ok (in eglibc) Apr 30 19:00:00 can ask him later.. but now I have to be really gone ;) Apr 30 19:14:44 03Denys Dmytriyenko  07org.openembedded.dev * r5e9f45c577 10openembedded.git/recipes/meta/external-toolchain-csl.bb: Apr 30 19:14:44 external-toolchain-csl: remove unnecessary global RPROVIDES Apr 30 19:14:44 Fixes missing thread-db library needed by gdbserver. Apr 30 19:14:44 Signed-off-by: Denys Dmytriyenko Apr 30 19:16:27 Tartarus: I removed RPROVIDES and tried my builds - nothing seems to be broken and it fixes my immediate problem with missing thread-db lib. Can you please try it on your end? Thanks. Apr 30 19:18:56 denix0: I'll read over the commit Apr 30 20:21:02 khem: 4.5.0 + e500v2 ICEs Apr 30 20:24:09 khem DISTRO=minimal MACHINE=calamari bitbake gcc-cross-initial will do it Apr 30 20:25:46 03Martin Jansa  07org.openembedded.dev * rdacbd2d6e7 10openembedded.git/conf/distro/include/sane-srcrevs.inc: Apr 30 20:25:46 EFL: bump SRCREV a bit more for focus fix Apr 30 20:25:46 Signed-off-by: Martin Jansa Apr 30 20:29:59 denix0: that problem with min_mmap_addr != 0 was on debian unstable without selinux.. Apr 30 20:46:44 heh Apr 30 20:46:51 13:22] khem: 4.5.0 + e500v2 ICEs Apr 30 20:46:52 [13:25] khem DISTRO=minimal MACHINE=calamari bitbake gcc-cross-initial will do it Apr 30 20:47:37 Tartarus: which problem ? Apr 30 20:47:41 is it another one ? Apr 30 20:48:19 another Apr 30 20:48:28 4.4.[23] compile, but gnutls blows up Apr 30 20:48:37 4.5.0 ICEs in gcc building gcc Apr 30 20:48:42 (cross-init) Apr 30 20:49:38 interesting. Apr 30 20:49:41 khem: I tried to build armv4t (om-gta02) and armv5t (spitz) image with gcc-4.5.0 and the result was not so god :/ Apr 30 20:50:13 JaMa: what problem did you see Apr 30 20:50:36 khem: it built fine.. but doesn't boot lots of errors even while booting kernel and when tried in chroot from image built with gcc-4.4.3 is shows alignment errors in almost every binary Apr 30 20:50:45 khem: ie "ls" works Apr 30 20:50:54 khem: but "ls -l" segfaults Apr 30 20:51:03 khem: mmt I'll boot spitz to show you bt Apr 30 20:51:42 JaMa: interesting. I have been using it on qemu did not see these but then qemu is bad at capturing alignment faults Apr 30 20:52:04 Tartarus: Do you have the commandline which causes the segfault Apr 30 20:52:34 scrolled off Apr 30 20:53:03 Tartarus: I just joined the room so no history here Apr 30 20:53:16 if you pasted earlier Apr 30 20:53:18 khem, I mean the ICE itself Apr 30 20:53:22 But... Apr 30 20:53:35 khem: DISTRO=minimal MACHINE=calamari bitbake gcc-cross-initial will do it, with 4.5.0 Apr 30 20:53:53 Tartarus: ok will try to reproduce here Apr 30 20:54:41 Tartarus: because gcc-cross-* uses your local gcc I was interested to know if it was the native gcc on your build box that segfaulted Apr 30 20:55:33 the newly build gcc would only be used for compiling target libraries and there are not many build in gcc-cross-initial phase Apr 30 20:55:45 its only libgcc and startup files Apr 30 20:56:25 JaMa: same setup works well with gcc 4.4.x Apr 30 20:56:31 for you ? Apr 30 20:56:37 khem: ubuntu 9.10 x8664 Apr 30 20:56:45 khem: yes Apr 30 20:57:00 JaMa: I have a osk board which is armv5t Apr 30 20:57:09 I can try it out on real hardware Apr 30 20:57:09 khem: it's image I rebuild quite often for my spitz and om-gta02 Apr 30 20:57:32 khem: waiting to charge spitz a bit (refused to boot now) Apr 30 20:57:55 Tartarus: I have ubuntu 10.04 but gcc on it is same I think Apr 30 20:57:59 4.4.3 Apr 30 20:58:14 may be its 4.4.1 on 9.10 Apr 30 20:58:18 I dont remember Apr 30 20:58:45 I have x86-64 box so should be same as yours Apr 30 20:59:03 * khem 's todo list is growing Apr 30 20:59:29 * JaMa has gentoo with 4.4.3 on x64-64 and no ICE Apr 30 20:59:42 JaMa: and e500v2 target? Apr 30 21:00:18 er, nm Apr 30 21:00:21 Getting stuff confused :) Apr 30 21:03:01 khem: http://paste.pocoo.org/show/208216/ Apr 30 21:03:32 Tartarus: no.. not e500v2 target, just spitz+gta02 Apr 30 21:04:01 JaMa: ok, whats your glibc/eglibc version Apr 30 21:07:08 khem: glibc-2.10.1, basically current oe.dev Angstrom + http://gitorious.org/angstrom/openembedded/commit/171b6167ac571ca3dc6caefa0a6f71bad9040fc7 Apr 30 21:07:36 JaMa: ok and kernel version ? Apr 30 21:07:46 khem: and on gta02 eglibc-2.11 same binutils Apr 30 21:08:15 khem: 2.6.33 Apr 30 21:08:31 JaMa: I have chased a bug in dynamic linker in eglibc which spewed such warnings Apr 30 21:08:36 thats fixed though Apr 30 21:08:44 hmm may be not Apr 30 21:08:53 2.6.33 ok Apr 30 21:09:08 can you try something older Apr 30 21:09:24 may be previous stable version of kernel to verify Apr 30 21:10:06 JaMa: basically you are getting these errors but the programs do work or thats not the case ? Apr 30 21:10:13 they segfault all the time Apr 30 21:10:44 segfaults all the time (when chrooted to that image) Apr 30 21:10:54 and kernel wont boot when booting that image Apr 30 21:11:21 no log here from kernel, because I didn't take serial cable with me Apr 30 21:11:53 JaMa: ok Apr 30 21:12:13 if kernel does not boot thats something new then Apr 30 21:12:24 JaMa: it needs to be investigated Apr 30 21:13:12 hope I can reproduce it with my osk board Apr 30 21:14:04 addresses in the errors you pasted seems like ld.so address to me Apr 30 21:15:07 03Martin Jansa  07org.openembedded.dev * rd99a83922c 10openembedded.git/recipes/shr/shr-e-gadgets_git.bb: Apr 30 21:15:07 shr-e-gadgets: bump SRCREV a bit more Apr 30 21:15:07 Signed-off-by: Martin Jansa Apr 30 21:22:35 RP: ping Apr 30 21:25:27 okay Apr 30 21:26:17 http://gist.github.com/381004 now has 1) regular shell variable reference tracking (${}), 2) python getVar() and expand() reference tracking, 3) python function call tracking specifically for calls to methods in the methodpool (def'd python functions), 4) shell function call tracking between shell functions in the metadata Apr 30 21:26:40 all thats missing now is applying the python analysis with ast to the ${@} snippets as well, and some cleanup.. then integration Apr 30 21:27:04 thoughts? Apr 30 21:28:31 kergoth: anonymous functions ? Apr 30 21:28:53 can use PythonValue for those too, from its perspective its no different, just another block of python code in a string Apr 30 21:28:53 j Apr 30 21:28:53 us Apr 30 21:28:53 t s Apr 30 21:28:57 blah, insane macbook Apr 30 21:29:00 just slightly different integration Apr 30 21:30:52 course, that piece has to be done in bitbake rather than OE, unless I want to poke into __ prefixed variables, but i intend to integrate into bitbake anyway.. Apr 30 21:31:40 ok Apr 30 21:32:11 its that bitbake integration that remains the unknown. do we replace the values in the metadata with these objects at finalize() time directly, or out of band just for ref tracking? if the former, how do we interact with non-string objects already in the metadata? Apr 30 21:35:02 also uncovered some aspects of the datastore behavior that aren't explicit in intent, they aren't documented, but things depend upon the implementation.. like the fact that a ${@} snippet gets ${} expansion applied to the result of the python, not just before the python is eval'd Apr 30 21:35:06 heh Apr 30 21:36:29 right Apr 30 21:36:56 * kergoth thinks that sort of thing really needs to be documented somewhere Apr 30 21:37:48 ${@} is well explained though isnt it Apr 30 21:37:57 or maybe not Apr 30 21:38:01 not this part of it Apr 30 21:38:29 consider this, i doubt people even necessarily realize that ${} expansions inside of python code are expanded before the python is run Apr 30 21:38:38 so bb.data.expand("${somevar}", True) in a task is completely useless Apr 30 21:38:42 '${somevar}' is the same thing Apr 30 21:38:44 s/True/d/ Apr 30 21:39:44 hmm right Apr 30 21:40:31 I've thought about killing that, let the python use getvar/expand themselves to get to it, but shit would break, i'm sure. we have so much shit going on that there are dependencies on implementation details everywhere Apr 30 21:40:39 heh Apr 30 21:41:52 khem: chroot from image with 2.6.26 behaves a bit better, I was able to install gdb at least :) Apr 30 21:42:28 JaMa: ok Apr 30 21:42:50 kergoth: ${@} is bitbake invention or is it std python Apr 30 21:42:57 it's a bitbake thing Apr 30 21:43:03 just indicates that what follows is run in eval() Apr 30 21:43:11 khem: http://paste.pocoo.org/show/208232/ Apr 30 21:44:34 JaMa: is that on a armv5t box Apr 30 21:50:39 khem: yes Apr 30 21:51:04 ok the backtrace is not much useful Apr 30 21:51:18 but it seems its dead somewhere in libc Apr 30 21:51:51 sorry for that.. trying to install at least glibc-dbg Apr 30 21:52:28 backtrace as bad as before.. Apr 30 21:53:51 I'll reflash that gcc-4.5 image on armv4 as there I'm using completely same eglibc/binutils (with gcc-4.4.3 and this gcc-4.5.0) Apr 30 22:01:23 03Stefan Schmidt  07org.openembedded.dev * rb579b5931f 10openembedded.git/recipes/llvm/llvm-native.inc: llvm2.7-native: Add perl-native to the DEPENDS Apr 30 22:27:06 Any folk around that are good with python? could use another set of eyeballs on this code as a sanity check Apr 30 23:02:37 khem: Does minimal-uclibc have a building xwindows image? Apr 30 23:02:46 for the qemu boards Apr 30 23:15:11 val = oe.kergoth.Value("${@d.getVar('foo', True) + ' ${bar}'}", d) Apr 30 23:15:11 assert(str(val) == "value of foo value of bar") Apr 30 23:15:12 assert(set(val.references()) == set(["foo", "bar"])) Apr 30 23:15:16 woot, yet another case handled Apr 30 23:28:57 awesome, now it lets you construct a value directly from components, so we could change "FOO += 'bar'" -> Value([OldValue, "bar"], d), so the previous version still exists, just referenced by the new one Apr 30 23:29:15 * kergoth thinks he's getting close to the functionality of his old prototype Apr 30 23:31:59 * Tartarus thinks he figures out the libfso-glib problem his loop has had Apr 30 23:32:24 can't be PACKAGE_ARCH=all since it has pkgconfig files Apr 30 23:32:26 hmm, i should fire off more loop builds with globally shared cached-binaries Apr 30 23:32:34 * kergoth kicks one off Apr 30 23:33:28 ugh .. Apr 30 23:41:34 03Tom Rini  07org.openembedded.dev * rc669f2444d 10openembedded.git/recipes/freesmartphone/fso-specs_git.bb: Apr 30 23:41:34 fso-specs: Don't use PACKAGE_ARCH=all due to pkgconfig files. Apr 30 23:41:34 Without this, we can't reuse TMPDIR for multiple arches. Apr 30 23:41:34 Signed-off-by: Tom Rini Apr 30 23:41:35 03Tom Rini  07org.openembedded.dev * rb4eb0ccfc8 10openembedded.git/conf/machine/include/tune-ppc603e.inc: Apr 30 23:41:36 tune-ppc603e: Drop -mhard-float Apr 30 23:41:36 -mcpu takes care of this argument for us Apr 30 23:41:36 Signed-off-by: Tom Rini Apr 30 23:41:37 03Tom Rini  07org.openembedded.dev * rff0e53a12c 10openembedded.git/recipes/xorg-lib/libdmx_1.1.0.bb: Apr 30 23:41:37 libdmx: Add fso-specs to DEPENDS Apr 30 23:41:38 Signed-off-by: Tom Rini Apr 30 23:45:36 anybody know whats going on.. http://pastebin.com/RUVYHzYz Apr 30 23:48:21 i dont have any libc in stuff/tmp/staging/x86_64-linux/usr/lib Apr 30 23:50:26 Well, when in doubt, hate ccache May 01 00:02:56 i dont have glibc-static installed May 01 00:07:29 fedora needs glibc-static rpm to build console-image May 01 00:09:34 whee May 01 00:22:12 BB_NUMBER_THREADS = "4" with -j 6 on a quad-core with 8G ram May 01 00:23:51 just dropped by to say congrats on building an awesome system :-D May 01 00:32:15 damnit May 01 00:32:19 pyshyacc fails to parse oe_libinstall May 01 00:36:18 NOTE: References for do_configure: set(['ARM_FP_MODE', 'STAGING_DIR', 'prefix', 'TARGET_VENDOR', 'FILE', 'HOST_OS', 'BASEPKG_HOST_SYS', 'PR', 'EXTRA_AUTORECONF', 'acpaths', 'PV', 'datadir', 'PF', 'PE', 'ARM_FP_PACKAGESUFFIX', 'PN', 'EXTENDPE', 'MULTIMACH_TARGET_SYS', 'BASE_PACKAGE_ARCH', 'S', 'TARGET_ARCH', 'HOST_VENDOR', 'TARGET_OS', 'TMPDIR', 'autotools_do_configure', 'WORKDIR', 'STAGING_DIR_HOST', 'STAGING_DATADIR', 'MULTIMACH_ARCH']) May 01 00:36:18 NOTE: References for do_compile: set(['oefatal', 'die', 'MAKE', 'EXTRA_OEMAKE', 'base_do_compile', 'oenote', 'oe_runmake']) May 01 00:36:21 progress! May 01 00:49:09 packed May 01 00:49:36 I've spent entirely too many hours working the past two weeks May 01 00:49:38 argh, damnit May 01 00:49:43 where to? May 01 00:53:02 damnit, textwrap.dedent() isn't sufficient for this May 01 00:53:11 I really wanted to avoid doing what bitbake does right now for the indentation May 01 00:53:12 hrmph May 01 01:37:37 can anyone tell me why in the dsp-link recipes the variable ${TOOLCHAIN_PATH} is being used, instead of say, ${CROSS_DIR}? May 01 01:40:41 If I had to guess, TI folks using external-toolchain-csl May 01 01:40:46 And a missed item in review May 01 01:45:40 erik006: if you check bitbake.conf, TOOLCHAIN_PATH=CROSS_DIR by default, so what's your problem? :) May 01 01:46:16 and as Tartarus said, it's to be able to use internal and external toolchains May 01 01:57:18 Tartarus: I guess I suspected something like that, thanks for pointing it out though May 01 01:57:40 denix0: I was having issues building the recipes, due to it not finding the arm specific binaries May 01 01:58:31 denix0: for some reason, TOOLCHAIN_PATH was empty, and hence I couldn't build May 01 01:58:51 in my bitbake.conf it does not seemed to be set, which would explain my issues May 01 01:58:56 Tartarus: I guess it's time for RFC to replace CROSS_DIR with TOOLCHAIN_PATH in all recipes... :) May 01 01:59:17 but that's my own fault - so no issues there May 01 01:59:39 erik006: that would explain it :) May 01 02:00:08 I made sure TOOLCHAIN_PATH is set to something sane before using it anywhere May 01 02:01:13 i was using the overo git repository May 01 02:01:35 then checked out the openembedded dev branch to get the latests TI stuff May 01 02:01:44 bad idea, I think May 01 02:02:34 a great learning opportunity though May 01 02:02:38 thanks for your help! May 01 02:06:44 overo should work with latest recipes from ti/ directory in .dev - I remember helping someone with the overo build before... May 01 02:07:34 well, it is likely I was doing it differently I guess May 01 02:08:37 either way, now that I know this, it should be a piece of cake May 01 02:10:02 * denix0 wishes he had your optimism... :) May 01 02:11:26 optimism/lack of experience May 01 02:59:45 whew **** ENDING LOGGING AT Sat May 01 02:59:56 2010