**** BEGIN LOGGING AT Thu Dec 16 02:59:57 2010 Dec 16 05:11:16 03Chris Larson  07master * r1f5d49ce64 10bitbake.git/lib/bb/cooker.py: Dec 16 05:11:16 cooker: don't swallow recipe parse errors Dec 16 05:11:16 Signed-off-by: Chris Larson Dec 16 05:41:29 grg: what part of the OE manual do i need to look at to try to build my module? Dec 16 05:42:10 gandhijee, it shouldn't be dissimilar from other recipes Dec 16 05:42:54 i tried using ARCH= and that magic, but i get errors Dec 16 05:43:23 i haven't done it myself, but i'm sure there are other examples in the oe tree Dec 16 05:43:48 ok Dec 16 05:44:12 gandhijee, have a look at madwifi Dec 16 05:47:26 gandhijee, and grepping for KERNEL_PATH will show a few more Dec 16 05:47:45 thanks Dec 16 05:49:17 gandhijee, there doesn't appear to be much consistency in how this is done for kernel modules... if you're feeling ambitious, maybe you could write a .bbclass for kernel modules and convert some of these to it? Dec 16 05:49:39 if i figure out how to do it i will Dec 16 05:49:44 :) Dec 16 05:49:59 i am about to rip out the tree from oe right now to try to compile the module Dec 16 05:52:19 * grg sees module.bbclass and feels foolish Dec 16 05:57:20 grg: i missed it too =P Dec 16 07:59:25 does proc have to return a string? Dec 16 07:59:37 i mean, does a proc entry have to return a string Dec 16 08:04:05 03Martin Jansa  07master * re01fb362ee 10openembedded.git/ (13 files in 2 dirs): Dec 16 08:04:05 efl,e-base: instruct fetcher to keep scmdata with new scmdata=keep param Dec 16 08:04:05 * bitbake master was excluting .svn dir while creating snapshot tarballs Dec 16 08:04:05 * read Dec 16 08:04:05 http://www.mail-archive.com/bitbake-dev@lists.berlios.de/msg00880.html Dec 16 08:04:06 for details Dec 16 08:04:06 Signed-off-by: Martin Jansa Dec 16 08:47:54 good morning Dec 16 08:57:29 03Sylvain Paré  07master * re0235ba322 10openembedded.git/recipes/gwaterpas/ (4 files in 2 dirs): Dec 16 08:57:30 gwaterpas: first recipe for this app in its 0.3 version Dec 16 08:57:30 * see http://wiki.openmoko.org/wiki/Gwaterpas Dec 16 08:57:30 * thanks to playya for his help! Dec 16 08:57:30 Signed-off-by: Sylvain Paré Dec 16 08:57:30 Signed-off-by: Martin Jansa Dec 16 08:57:40 03Martin Jansa  07master * rf8dacd8b68 10openembedded.git/recipes/tasks/task-shr-feed.bb: Dec 16 08:57:40 task-shr-feed: add gwaterpas, comment out few broken (mpd, shr-launcher, ventura, maxima) Dec 16 08:57:40 Signed-off-by: Martin Jansa Dec 16 08:58:41 03Koen Kooi  07org.openembedded.dev * r6ee127712f 10openembedded.git/recipes/gstreamer/ (gstreamer/query-buffers.patch gstreamer_0.10.31.bb): Dec 16 08:58:41 gstreamer 0.10.31: add patch for query buffer api Dec 16 08:58:41 Signed-off-by: Koen Kooi Dec 16 08:58:41 03Koen Kooi  07org.openembedded.dev * r2e15fbd75c 10openembedded.git/recipes/ti/libdce_git.bb: Dec 16 08:58:41 libdce: bump SRCREV for h264 fixes Dec 16 08:58:42 Signed-off-by: Koen Kooi Dec 16 08:58:43 03Koen Kooi  07org.openembedded.dev * r0bccf4e7ce 10openembedded.git/recipes/ti/gst-ducati_git.bb: Dec 16 08:58:43 gst-ducati: add WIP recipe Dec 16 08:58:44 Signed-off-by: Koen Kooi Dec 16 09:00:03 gm Dec 16 09:02:11 Jay7: have you tried to patch what I suggested for gettext Dec 16 09:03:13 hey, how do i turn an interger into a string inside the kernel? Dec 16 09:05:56 gandhijee: itoa() Dec 16 09:06:49 eFfeM_work: hi Dec 16 09:11:06 yeah itoa is a normal c function Dec 16 09:38:08 can someone with latest gcc-4.5 try to build kexecboot-klibc? here it segfaults now arm-oe-linux-gn[10171]: segfault at 0 ip 00002ba56cc076e0 sp 00007fffdd784718 error 4 in libc-2.12.1.so[2ba56cb9c000+163000] Dec 16 09:39:30 ah it's not gcc-4.5 but arm-oe-linux-gnueabi-klcc Dec 16 09:39:54 jama armv4? Dec 16 09:40:23 armv5te (spitz) Dec 16 09:41:20 hm that might be a strange problem Dec 16 09:41:33 ant had some segfaults too, where I had not Dec 16 09:41:39 hm or was it dietlibc Dec 16 09:41:41 hm hm Dec 16 09:41:47 it was building ok about a week ago Dec 16 09:41:53 but thats was some time ago Dec 16 09:42:04 and rebuilding klcc-cross_1.5.20.bb didn't help Dec 16 09:42:05 since than only the samba stuff was pachted Dec 16 09:42:31 that would be tragedy when one is fixed and the other is broken Dec 16 09:43:42 if I run it manualy in terminal then it builds ok Dec 16 09:44:01 maybe it's related to changed subprocess handling in bitbake? Dec 16 09:47:19 morning Dec 16 09:47:54 eFfeM_work: I've tried proposed change and it did not help Dec 16 09:48:16 I don't know from where libtool get -L/usr/lib while relinking Dec 16 09:48:24 and why it decide to relink on install Dec 16 09:50:32 hi jay7 Dec 16 09:50:49 jay7 urgs so another libtool error Dec 16 09:51:13 woglinde: I got it on gettext Dec 16 09:51:21 so it may be gettext error Dec 16 09:51:55 dont thinkso Dec 16 09:52:14 gettext maintainer knows enough about libtool and autotools Dec 16 10:09:29 03Martin Jansa  07master * reba0ae0805 10openembedded.git/recipes/linux/ (4 files in 3 dirs): Dec 16 10:09:29 linux(-kexecboot)_git: bump SRCREV for 2.6.37-rc6 Dec 16 10:09:29 Signed-off-by: Martin Jansa Dec 16 10:12:02 spitz with 2.6.37? Dec 16 10:13:26 it should work except power management Dec 16 10:13:41 offline charging is broken from .24 Dec 16 10:13:55 uh Dec 16 10:14:00 fix it Dec 16 10:14:02 *g* Dec 16 10:14:38 heh.. I'm not a kernel guru :) Dec 16 10:15:14 Jay7: trying to upgrade kexecboot rev, to check that broken gui again on spitz.. Dec 16 10:15:31 JaMa|Wrk: It will happens, I'm sure :) Dec 16 10:15:32 will let you know (hopefully tonight Dec 16 10:15:56 Jay7: I know.. only change is that I've rebuilt it from scratch Dec 16 10:16:28 I need some time w/o anybody around to sit down and code some things from my queue.. Dec 16 10:17:37 http://alexthunder.livejournal.com/309815.html Dec 16 10:17:49 good description of programmer's work :) Dec 16 10:27:02 hm.. kexecboot-klibc_git build failing in interesting manner Dec 16 10:27:15 klcc segfaults on cfgparser Dec 16 10:27:43 that's the same I got today (see 10:37:08 Dec 16 10:28:45 :( Dec 16 10:29:49 JaMa|Wrk: but there is angstrom-2008.1 with gcc-4.3 iirc Dec 16 10:32:59 qemux86 is still broken Dec 16 10:33:50 Jay7: i thought the -L/usr/lib came from those two lines Dec 16 10:34:17 eFfeM_work: I've tried to rebuild libtool and gettext with commenting out similar lines in ltmain.* Dec 16 10:34:29 this didn't help me Dec 16 10:34:38 but may be I've missed something Dec 16 10:35:05 good morning Dec 16 10:35:10 Jay7: I think the gettext ltmain.sh is generated but (dirty trick) you could patch it in do_install_prepend or so Dec 16 10:35:11 hi eric Dec 16 10:35:19 did anyone here received apandaboard after ordering it on digikey ? Dec 16 10:35:34 eFfeM_work: I've patched it by hands after bitbake -c patch Dec 16 10:36:17 Jay7: ah ok, maybe you should do that by hand after do_configure (or at least check afterwards if your patchis not overwritten) Dec 16 10:36:25 ah, may be.. Dec 16 10:36:25 * eFfeM_work is not a libtool expert either Dec 16 10:36:59 mine was supposed to ship 11/21, then 12/12 (a sunday) then 12/19 (an other sunday) and now 6/1 (which is not a sunday so maybe someone can ship it ;-) Dec 16 10:38:16 Jay7: btw make sure to patch the version in i686* or x64* (the native one) and not the target one Dec 16 10:38:40 did you mean libtool or gettext now? Dec 16 10:39:57 Jay7: bitbake master? Dec 16 10:40:04 JaMa|Wrk: yes Dec 16 10:41:10 Jay7: maybe your problem is the same kind of problem I have with LANG : if now we get variable from the environment like LANG it means we can also get other variables which can affect the configure/compile process. Dec 16 10:41:12 hi obi Dec 16 10:41:26 ericben: hm.. good idea Dec 16 10:41:35 I'll try to run with env -i Dec 16 10:41:37 hi Dec 16 10:42:32 Jay7: try bitbake at 4d081a0ed759bd526ab01849d650bd9e8d80ddd1 Dec 16 10:42:51 Jay7: this is the commit before the one which starts to create trouble for LANG Dec 16 10:51:02 hm.. Dec 16 10:51:22 bitbake hangs right after printing Build Configuration Dec 16 10:53:07 wow.. Dec 16 10:53:21 after ^C it seems to started doing something Dec 16 10:53:34 according to ps output Dec 16 10:53:40 but still nothing on screen Dec 16 10:56:33 Jay7 I meant libtool (was afk) Dec 16 10:56:38 hi boi Dec 16 10:58:42 ah... seems bb was waiting for oestats answer.. Dec 16 11:02:05 o_O Dec 16 11:02:17 JaMa|Wrk: clean build may succeed Dec 16 11:02:34 I've cleaned klibc, kexecboot and kexec-tools-klibc Dec 16 11:02:45 and seems it will be compiled ok Dec 16 11:03:27 but, I've rolled bitbake back Dec 16 11:03:37 to revision ericben suggested Dec 16 11:03:50 well.. let's try with HEAD Dec 16 11:04:05 make WITH_HEAD=yes :) Dec 16 11:07:51 hehe.. Dec 16 11:07:59 bb HEAD is failing the same manner Dec 16 11:11:57 so bitbake is guilty somehow.. Dec 16 11:17:37 Hi Dec 16 11:19:13 using release-2010.12, i try to build x11-image for i586 Dec 16 11:19:34 but i errors Dec 16 11:19:51 NOTE: Parsing finished. 0 cached, 6909 parsed, 301 skipped, 0 masked. Dec 16 11:19:51 ERROR: Parsing errors found, exiting... Dec 16 11:20:17 full log here: http://pastebin.com/raw.php?i=zPbEMkLw Dec 16 11:20:33 * Jay7 bisecting bb to find what commit break klibc Dec 16 11:25:50 my config http://pastebin.com/raw.php?i=81XXjeAH Dec 16 11:29:06 machine "ibm-a21m" : http://pastebin.com/Wy8Mf5aD Dec 16 11:30:35 Jay7: it works ok when you run "run.do_compile.$$" manually in terminal Dec 16 11:30:47 Jay7: so it will be something with subprocesses Dec 16 11:30:58 JaMa|Wrk: it works ok when you run run_oemake in devshell too :) Dec 16 11:31:18 using official "debian bitbake 1.8.18-2" Dec 16 11:39:10 with 4262c26d36d1c1b6801ac422716a227c1f6b9589 I get ERROR: Task 79 (/storage/oe/testbuilder/openembedded/recipes/images/initramfs-kexecboot-klibc-image.bb, do_package_update_index_ipk) failed with exit code '1' Dec 16 11:42:28 so, I'll mark this as good anyway because kexec-tools and kexecboot was compiled Dec 16 12:02:54 20929afdd87e5124891f121c3640aa9efa368c2c Dec 16 12:03:11 this is :) Dec 16 12:04:25 but not sure how.. Dec 16 12:07:11 kergoth_: bitbake rev 20929afdd87e5124891f121c3640aa9efa368c2c break building of kexec-tools-klibc Dec 16 12:08:26 args Dec 16 12:09:09 before it compiles, this and after - not Dec 16 12:09:56 should go now Dec 16 12:10:05 * Jay7 -> afk Dec 16 12:15:52 03Martin Jansa  07master * r4dd35de374 10openembedded.git/recipes/phoneme/phoneme-advanced.inc: Dec 16 12:15:52 phoneme-advanced: force ARM mode, otherwise build fails on armv4t Dec 16 12:15:52 Signed-off-by: Martin Jansa Dec 16 13:26:21 khem: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46883 is fixed. Dec 16 13:27:42 khem: I think we shouldn't have patches targeted for gcc 4.6 in what we call gcc 4.5. gcc 4.5 should be gcc 4.5 + bug fixes and we coudl have a gcc head or gcc-next for people who want bleeding edge patches in their toolchain Dec 16 13:29:56 ericben: by 'patches targeted for gcc 4.6' you mean 'linaro patches'? Dec 16 13:30:48 ericben: and I guess this one will be merged to 4.5 branch as soon as 4.5.2 is released and branch open again Dec 16 13:30:57 hrw: not specialy Linaro's patches, whatever patch provides a new feature expected for inclusion in next release Dec 16 13:31:17 ericben: even if this gives you faster/smaller code? Dec 16 13:32:07 I think the default toolchain for many distro should not be a test bed for gcc. Of course new patches improve thing somewhere but they also break things on other parts/archs Dec 16 13:32:34 ericben: if you find regression then report it Dec 16 13:33:12 hrw: that's what was done. In the present case : gcc-4.5 breaks some compilations for armv4 and armv5. Unless I'm mistaken, Linaro targets armv7 so maybe we should reduce the Linaro patches to armv7. Dec 16 13:33:38 linaro targets armv7 but with 'make no harm' to previous ones Dec 16 13:34:01 hrw: sure but how much is that tested on arch < armv7 ? Dec 16 13:34:41 ericben: if there is regression due to linaro patch then please report it at https://bugs.launchpad.net/gcc-linaro Dec 16 13:34:57 hrw: when I checkout OE to build an image I don't expect that the default compiler is broken for such common arch than armv4 or armv5. In that case something is wrong somewhere Dec 16 13:35:25 hrw: the problem was reported by Khem on Linaro's lists Dec 16 13:35:49 and was fixed today by a Codesourcery person Dec 16 13:36:41 and that CS person is working at Linaro too Dec 16 13:37:30 hrw: if I understand well, it was not directly a Linaro's patch problem. It was caused by a trunk patch which was backported in gcc 4.5 and labeled gcc-4.5-linaro-r99379.patch Dec 16 13:37:57 the problkem is not to criticize Linaro's work, I know we will all benefit from this work Dec 16 13:37:58 it was problem due to linaro patch Dec 16 13:38:52 so next step is to add http://gcc.gnu.org/ml/gcc-patches/2010-12/txt00064.txt to gcc-4.5 recipe Dec 16 13:39:13 hrw : I'm actually on this :). The root problem is : is OE's default toolchain intended to be a moving target ? Dec 16 13:39:49 gcc-4.5 is default for at least distro minimal, angstrom 2010.x Dec 16 13:40:09 for 2010 we want to be able to fiddle with toolchain Dec 16 13:40:18 for 2008, it used stable toolchain :) Dec 16 13:41:23 ericben: it is also default in ubuntu 11.04 (with linaro patchset) Dec 16 13:41:45 ubuntu does not care about the variety of machines OE does Dec 16 13:42:13 I know Dec 16 13:51:05 hrw: ok, that's the risk with having bleeding edge software after all, my only concern is should be have a default stable toolchain or do we accept a bleeding edge toolchain which the possibility to have broken targets/packages from one day to the other because of new patches added to gcc. Dec 16 13:54:00 ericben: thats the problem of .dev tree Dec 16 13:54:26 how can I download only the kernel used by linux-kexecboot_2.6.21-hh20.bb ? Dec 16 13:54:31 with stable core you will have toolchain updates only after they got tested to work for targets Dec 16 13:54:42 mckoan: bitbake -b linux-kexecboot_2.6.21-hh20.bb -cfetch? Dec 16 13:54:47 hrw: thx Dec 16 13:54:53 mckoan: you have to play with old ipaqs? Dec 16 13:55:01 hrw: moreover, gcc-linaro-4.5-2010.12-0 is based on svn167157 and our is on 167449 so we are even in advance which may explain some bug Dec 16 13:55:06 hrw: I need a driver TSC2200 Dec 16 13:55:51 I wrote it on 2004 for linux-2.4 and now I'd like not to reinvent the wheel Dec 16 13:56:05 ericben: but as 4.5 branch gets only bugfixes, I agree with khem's gcc SRCREV bumps to get those fixes Dec 16 13:57:10 JaMa|Wrk: yes (that's waht I also did on my gcc-4.5.2 patch), the problem is when we add other patches which are not in gcc's 4.5 branch Dec 16 13:58:12 ericben: 4.5 branch is now closed because of 4.5.2 release.. as I said before.. some of them will probably get there as soon as 4.5.2 is out Dec 16 13:58:57 it's frozen since 8th Dec Dec 16 13:59:01 mckoan: ah Dec 16 13:59:08 JaMa|Wrk: ok Dec 16 14:00:39 hrw: by "stable core" do you mean something like Yocto ? Dec 16 14:05:28 03Stefan Schmidt  07master * r795c18b0eb 10openembedded.git/recipes/matchbox-keyboard/ (5 files in 3 dirs): matchbox-keyboard-inputmethod: Add support for bug20 layout Dec 16 14:05:35 03Stefan Schmidt  07master * r7106d1eae3 10openembedded.git/recipes/initscripts/ (initscripts-1.0/bug20/bootmisc.sh initscripts_1.0.bb): initscripts-1.0: Add special bootmisc for bug20 to set the GPS serial port correctly. Dec 16 14:05:36 03Stefan Schmidt  07master * r6b5c27f05d 10openembedded.git/recipes/udev-rules-buglabs/ (7 files in 2 dirs): udev-rules-buglabs: Add recipe for Bug 2.0 specific udev rules Dec 16 14:05:37 03Stefan Schmidt  07master * r75306a9922 10openembedded.git/recipes/bug-icon-theme/mb-bug-icon-theme.bb: bug-icon-theme: Add buglabs icon theme for matchbox Dec 16 14:05:47 03Stefan Schmidt  07master * rb8ce0dd695 10openembedded.git/recipes/bug-desktop-icons/ (8 files in 2 dirs): bug-desktop-icons: Add recipes for some bug desktop icons. Dec 16 14:05:48 03Stefan Schmidt  07master * r7570e05a25 10openembedded.git/recipes/bug-logrotate/ (bug-logrotate_0.1.bb files/bug-rules): bug-logrotate: Add logrotate config file for bug Dec 16 14:07:26 ericben: yocto is good example Dec 16 14:07:35 l1nux: have you made any progress with your parse errors problem? Dec 16 14:09:01 * Tartarus hopes to fix one of his failing builds now, perhaps and waits on CIA Dec 16 14:09:01 03Tom Rini  07org.openembedded.dev * r8118cb14a0 10openembedded.git/recipes/dri/libdrm.inc: Dec 16 14:09:01 libdrm.inc: Add cairo to DEPENDS. Dec 16 14:09:01 Fixes: Dec 16 14:09:01 | make[3]: Entering directory `/.../tmp/work/i686-oe-linux/libdrm-1_2.4.22-r5.0/libdrm-2.4.22/tests/modetest' Dec 16 14:09:01 | CC modetest.o Dec 16 14:09:01 | modetest.c:59:19: fatal error: cairo.h: No such file or directory Dec 16 14:09:02 Signed-off-by: Tom Rini Dec 16 14:09:11 stefan_schmidt: wow, why separate recipe for each icon? not one with few sub packages? Dec 16 14:10:12 hrw: ok I expect to have time to look at Yocto in 2 weeks. Dec 16 14:10:37 JaMa|Wrk: Mostly because it was what we already had in the overlay. And I found it clean enough during review to get it upstream. Dec 16 14:10:49 JaMa|Wrk: Maybe I should redo it. Will think about it. Dec 16 14:11:53 stefan_schmidt: well with that really fast parsing in bitbake master it's not big issue.. I was just surprised to notice it Dec 16 14:12:25 JaMa|Wrk: Some stuff needs to get sorted into a different dir anyway sometime. Dec 16 14:12:45 JaMa|Wrk: A lot of the stuff is only interesting for the bug images. Dec 16 14:12:59 JaMa|Wrk: With a better dir layout we would avoid parsing for all others then. :) Dec 16 14:13:20 JaMa|Wrk: Will work on this when we redo our layout for work towards yocto Dec 16 14:14:02 yes, I plan to do the same with shr stuff Dec 16 14:27:06 foerster no :( Dec 16 14:27:27 l1nux: I had a similar problem in the past I think. Do you happen to have OVERRIDES specified in any of your conf files? Dec 16 14:27:42 There was a change a month or so ago in OE that altered that usage. Dec 16 14:27:54 and I remember having similar problems. Dec 16 14:29:55 "OVERRIDES specified in any of your conf files ?" ".bb" files ? Dec 16 14:30:13 .conf Dec 16 14:30:28 yes minimal.conf Dec 16 14:30:53 http://cgit.openembedded.org/cgit.cgi/openembedded/commit/?id=5b485324c2ac637e9fc6d40753ee64fd6907db69 Dec 16 14:31:32 i removed 2 lignes: Dec 16 14:31:35 # MACHINE_CLASS ?= "" Dec 16 14:31:35 # MACHINE_OVERRIDES += "${MACHINE_CLASS}" Dec 16 14:31:45 I think you want those lines. Dec 16 14:32:07 At least, the problem I had.. I had OVERRIDES set in a local distro mydistro.conf file Dec 16 14:32:17 and it needed to be switched to the lines like you removed Dec 16 14:33:45 is that the standard minimal.conf that's in oe/conf/distro, or a modified one in an overlay? Dec 16 14:34:32 is standard minimal.conf 2 lignes modified Dec 16 14:35:32 what happens if you put those lines back in? Dec 16 14:35:50 can you pastebin that file? Dec 16 14:37:03 yes i try later, actualy i build latest openwrt Dec 16 14:37:51 JaMa|Wrk: please take a look on the final gitpkgv version. Dec 16 14:38:09 otavio: did you send it to ML? Dec 16 14:38:18 JaMa|Wrk: I think gitver is a different use-case and I have not enough skills to properly merge them Dec 16 14:38:18 ah now it arrived :) Dec 16 14:38:21 JaMa|Wrk: did it Dec 16 14:38:40 foerster, oe need toolchain for build i586 (in i586 host) ? Dec 16 14:39:22 it'll try to build its own, unless you specify ASSUME_PROVIDED I believe Dec 16 14:39:35 JaMa|Wrk: I kept it producing mostly same output (I cut at 7 chars the rev since it is clear enough and easy to read) Dec 16 14:39:36 otavio: did you read those 2 pastebins I've sent you yesterday? Dec 16 14:39:40 but I've never done that, so others may be able to provide further info Dec 16 14:39:46 JaMa|Wrk: I did not Dec 16 14:39:54 shame on you then :) Dec 16 14:40:14 grr Dec 16 14:40:15 23:06:52 < JaMa|Zzz> otavio: http://pastebin.com/S3CDUjWD or even http://pastebin.com/uV4cpA0Y Dec 16 14:40:17 JaMa|Wrk: what was that? Dec 16 14:40:30 otavio: you should keep the format consistent Dec 16 14:40:58 that's why my version has "0.0-%s-g%s" for GITVERTAG and "%s+%s" otherwise Dec 16 14:41:41 otavio: and from my basic programming classes I still don't like exceptions for normal program flow Dec 16 14:42:30 JaMa|Wrk: I see Dec 16 14:42:41 JaMa|Wrk: doing it right now; except from it, it looks OK? Dec 16 14:43:17 otavio: and please s/PV = "gitr${SRCPV}"/PV = "1.0+gitr${SRCPV}"/g + example for PKGV = "1.0+gitr${GITPKGV}" Dec 16 14:44:40 s/${GITPKGVTAG}should/${GITPKGVTAG} should/g Dec 16 14:45:31 and maybe even real example what to expect in those vars, to be clear for everybody :) Dec 16 14:46:45 JaMa|Wrk: http://pastebin.com/C47qGdBA <= this? Dec 16 14:46:50 * otavio at lunch, brb Dec 16 14:50:58 otavio: I think you have to move "commits = bb.fetch.runfetchcmd" after that os.chdir() Dec 16 14:52:42 03Chris Larson  07master * rd527ca4415 10bitbake.git/lib/bb/ui/knotty.py: Dec 16 14:52:42 knotty: exit with 1 if we see a critical log message Dec 16 14:52:42 Signed-off-by: Chris Larson Dec 16 14:52:45 03Chris Larson  07master * ref97bd82a3 10bitbake.git/lib/bb/runqueue.py: Dec 16 14:52:45 runqueue: fix remnant rq/rqdata reference from poky sync Dec 16 14:52:45 Signed-off-by: Chris Larson Dec 16 14:53:31 otavio: and please also add how the tags need to look like for sortable GITPKGVTAG (ie v1.0, v1.1, v2.0, but not with tag like tested_2010-11-04 beween those) Dec 16 14:54:03 Jay7: what error re: kexec-tools-klibc? Dec 16 14:54:37 kergoth_: segfault Dec 16 14:54:46 ? Dec 16 14:54:49 03Chris Larson  07master * r0ea334430d 10bitbake.git/lib/bb/build.py: Dec 16 14:54:49 Revert "build: kill stdout in python functions" Dec 16 14:54:49 We have a better solution for this on the separate-ui-and-server branch, so Dec 16 14:54:49 this version is no longer necessary. Dec 16 14:54:49 This reverts commit 20929afdd87e5124891f121c3640aa9efa368c2c. Dec 16 14:54:49 Signed-off-by: Chris Larson Dec 16 14:55:05 nothing bitbake can do will cause python to segfault, if it does, its a bug in your python.. Dec 16 14:55:06 kergoth_: arm-oe-linux-gnueabi-klcc segfault Dec 16 14:55:25 sounds like a klcc bug Dec 16 14:55:43 segfault is *never* correct behavior, regardless of the root cause Dec 16 14:55:50 kergoth_: but with different bitbake version or when run.do_compile.$$ sourced to terminal it works ok Dec 16 14:56:09 kergoth_: so I guess it's something in env which makes it segfault Dec 16 14:56:32 try reading what i just said again Dec 16 14:58:00 regardless, i just reverted the commit you say causes it -- the separate-ui-and-server branch kills stdout/stderr in the entire bitbake server instead. Dec 16 14:58:09 kergoth_: I agree it's not right reaction to segfault, but any hint what was changed in bitbake 20929afdd87e5124891f121c3640aa9efa368c2c that it reveals this klcc bug? Dec 16 14:58:35 it changes stdout/stderr to go to /dev/null. Dec 16 14:58:43 so if the compiler breaks when that happens, i don't know what to say about that Dec 16 15:00:51 * JaMa|Wrk trying it with that revert Dec 16 15:01:28 the underlying issue that resulted in that commit is that we can't use non-standard bitbake UIs, because the python tasks spew crap all over the console Dec 16 15:01:37 when they should be using bb.msg to output things cleanly Dec 16 15:09:58 * kergoth_ thinks about adding unit tests to bb.process/bb.build Dec 16 15:10:18 (given the number of bugs that cropped up, its clearly in need of it) Dec 16 15:11:18 damnit Dec 16 15:11:23 we need to get the server ui split merged asap Dec 16 15:12:01 lets see, there was one issue remaining with that -- the possibility of the UI exiting without the server doing so as well Dec 16 15:12:19 just typed that , you beat me to it. Dec 16 15:12:23 :) Dec 16 15:12:29 right, we need some failsafe shutdown Dec 16 15:12:30 took me a second to remember what the issue was Dec 16 15:13:07 i also have goggle with right click -> pastebin of error messages, and it puts the url in your clipboard :) Dec 16 15:13:10 code needs cleaned up though Dec 16 15:13:13 the ui execution is wrapped in a finally that does a server.stop/server.join, maybe we just need a server.terminate() if the join times out or so there Dec 16 15:13:23 ah Dec 16 15:13:26 yes, I was thinking Dec 16 15:13:36 server.stop just exits the loop in the server Dec 16 15:13:39 that won't help with a SIGKILL/SIGTERM, we'd still need a signal handler Dec 16 15:13:58 we should probably do a cooker shutdown, and wait til there are no items left in the idle functions Dec 16 15:14:22 ah. hmm. Dec 16 15:14:38 rather, cooker stop (that's the one that says stop now, I think) Dec 16 15:14:41 we probably want a cooker stop, actually Dec 16 15:14:42 right Dec 16 15:14:58 that way, idle funcs will only run probably one more time to do the cleanup Dec 16 15:15:52 we can pass a pid of the UI, and server can check it periodically. Dec 16 15:16:02 then later, if in daemon mode, ignore the pid checking Dec 16 15:16:46 hmm, ugly, but i can't think of any other way to handle someone killing the UI forcibly in a way we can't catch Dec 16 15:16:53 right Dec 16 15:17:52 can have a flag, tied_to_ui = true, in the server for now. If tied_to_ui: check ui_pid Dec 16 15:18:28 ericben: hmm, I think its ok to have some bug fixes on top and we will always need them Dec 16 15:18:30 too bad we can't say "if no one is left on the other end of ui_channel, quit" Dec 16 15:18:33 gm all Dec 16 15:18:59 kergoth_: Right. I don't think the multiprocessing stuff does that, does it? Dec 16 15:19:11 should just use a damn socket Dec 16 15:20:12 hmm Dec 16 15:20:19 ericben: we will get some regressions which we should fix. If you take a release you can also get problems. We are just trying to make better performing toolchain as default for every user so they get a nicer experience. Dec 16 15:20:29 recv() Dec 16 15:20:30 Return an object sent from the other end of the connection using send(). Raises EOFError if there is nothing left to receive and the other end was closed. Dec 16 15:20:45 i wonder if that happens if it was closed forcibly Dec 16 15:21:20 03Koen Kooi  07org.openembedded.dev * r52e3070906 10openembedded.git/recipes/gstreamer/ (4 files in 2 dirs): Dec 16 15:21:20 gstreamer 0.10.31: add 2 more patches (crop and pad_alloc) Dec 16 15:21:20 0 Dec 16 15:21:20 Signed-off-by: Koen Kooi Dec 16 15:29:35 foerster: oops -- if we just terminate if it takes too long to join, it hoses us if the UI wanted to shut down slowly via a cooker shutdown rather than stop Dec 16 15:29:40 its happily finishing a task.. BOOM, DEAD Dec 16 15:29:42 hehe Dec 16 15:30:06 nice :) Dec 16 15:30:25 * kergoth_ guesses letting the server force itself to shutdown is the better approach :) Dec 16 15:32:46 khem: hi. OK the drawback with this policy is that users can be afraid to use OE because of the risk to have an unstable result. Dec 16 15:41:33 foerster: https://github.com/kergoth/bitbake/commit/2e8c1763 at least works for the case where the user runs 'kill ' Dec 16 15:41:51 hrm Dec 16 15:42:38 erm, i meant terminate, not shutdown Dec 16 15:42:40 * kergoth_ rolls syees Dec 16 15:42:44 :) Dec 16 15:44:38 giving the server the ui pid is more full proof, as itd handle KILL, though :) Dec 16 15:45:10 right, otherwise, we'll have no idea. Of course, do people really KILL bitbake? Dec 16 15:45:31 that'd only be last resort if it was not responding, I'd think. Dec 16 15:45:32 i'd hope not, more likely they'd ^C the UI or close it with the X Dec 16 15:45:36 right Dec 16 15:45:41 even TERm is much more likely Dec 16 15:45:44 KILL is really last resort Dec 16 15:45:59 in which case, its on them :) Dec 16 15:46:10 generally, things cannot be expected to shutdown nicely with kill, so yes, it's on them. Dec 16 15:46:41 plus, I do ps aux|grep bitbake|awk '{print $2;}'|xargs kill or something stupid Dec 16 15:46:43 maybe this is good enough, in that case. though we still should get the server to more cleanly shut down the cooker regardless Dec 16 15:46:44 to kill em all Dec 16 15:47:22 server.stop should call cooker.stop Dec 16 15:47:35 since we may get to server.stop in bin/bitbake if the UI threw an exception Dec 16 15:47:46 so cooker.stop was never called at that point Dec 16 15:47:54 right Dec 16 15:48:23 call cooker.stop, and we have to wait for idle funcs to be done i think Dec 16 15:48:24 we don't want to do it in server.stop() itself though, of course, since its called by the UI, would likely want to check on the cooker when the main loop is broken? Dec 16 15:48:30 right Dec 16 15:48:37 stop sets the event flag Dec 16 15:48:59 so the main loop of server needs to do a bit more Dec 16 15:49:30 it needs to call cooker.stop, and then run the idle funcs to let the cooker actually do the shutdown Dec 16 15:49:57 * kergoth_ nods Dec 16 15:50:34 i think that will clean up many of the variou issues we've seen Dec 16 15:50:55 that is, if the ui bails uncleanly, the server will still do a nice shutdown Dec 16 15:51:00 yep Dec 16 15:51:22 jumping back and forth btw python and c++ is a joy :) Dec 16 15:51:25 hehe Dec 16 15:51:33 if( Dec 16 15:51:36 shit, no ( Dec 16 15:52:07 use more snippets/templates? if ;) Dec 16 15:52:19 actually, i should Dec 16 15:52:30 * kergoth_ has snipMate installed in his vim, but always forgets to use it Dec 16 15:52:40 i've been using netbeans for a few months Dec 16 15:52:42 not too bad Dec 16 15:52:50 a bit heavy - good thing I have 4GB ram Dec 16 15:53:22 i love the *idea* of a good IDE, but none seem to be up to par, particularly for portable options. i haven't spent m uch time with netbeans though, should play with it more Dec 16 15:53:31 has git status coloring in the file listing, which is cool Dec 16 15:53:37 right click git status, diff Dec 16 15:53:54 cute Dec 16 15:53:55 it's not perfect, but it's the best of the options I've tried Dec 16 15:54:22 foerster: I use ksh and emacs; it works perfect =-) Dec 16 15:54:35 visual studio actually seems pretty nice from the little time i've spent with it, too bad its not portable, and the only python for it is IronPython Dec 16 15:54:40 * kergoth_ = vim junkie Dec 16 15:55:18 with netbeans, I didn't need to do much of anything to get php, python, c/c++ etc all supported Dec 16 15:55:19 just works Dec 16 15:55:30 Biggest problem I've had with IDEs is whenever I try and use one again I still need to spend too much time making basic stuff work Dec 16 15:55:54 I spent far too much time trying to make Eclipse+Maven2 behave a few weeks back Dec 16 15:55:57 i'm *invested* in vim. i've spent so much time tweaking it that its horrifying considering anything else Dec 16 15:56:06 i only use it for editing, not actual building/etc, although it's gdb support is nice Dec 16 15:56:12 And it still didn't do the one thing that would have been real nice, autocomplete java class stuff, since it didn't figure out my layout Dec 16 15:56:17 JaMa|Off: http://pastebin.com/NWZM61Wt <= finished and tested Dec 16 15:56:46 * kergoth_ still isn't very happy with his python completion in vim though :( using omnicompletion, but it's not very smart Dec 16 15:56:53 JaMa|Off: IMO we shouldn't comment about tagging style since it can in fact " work " in case it is sortable but not in a standard version Dec 16 15:57:03 JaMa|Off: so release-1, release-2 work Dec 16 15:57:16 kergoth_: ^ please take a look on this too. Dec 16 15:59:01 looks reasonable to me Dec 16 16:39:14 Hi .. Dec 16 16:39:36 Please help me i am strucking up with fallowing error Dec 16 16:39:41 checking for arm-angstrom-linux-gnueabi-gcc... arm-angstrom-linux-gnueabi-gcc -march=armv7-a -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp -mthumb-interwork -mno-thumb | checking whether the C compiler works... no | configure: error: in `/home/mistral/stuff/openembedded/tmp/work/armv7a-angstrom-linux-gnueabi/module-init-tools-3.4-r6/module-init-tools-3.4': | configure: error: C compiler cannot create executables | See `config.log' for mor Dec 16 16:44:35 nani_: what did you find when you saw config.log? Dec 16 16:48:49 It is a huge file of some 1000 lines Dec 16 16:49:16 Yes Dec 16 16:49:25 i hate config.log Dec 16 16:49:25 But if you look for "error: C compiler cannot create" Dec 16 16:49:30 you'll get pretty close to the error Dec 16 16:49:32 why cant' the actual error be at the bottom of the file? Dec 16 16:49:35 * kergoth_ shakes head Dec 16 16:51:29 does anyone know if python can raise an exception other than SyntaxError when doing a compile()? Dec 16 16:51:29 It is telling about PATH settings of LD flags and CFLAGS Dec 16 16:51:39 nani_: search for the error, as Tartarus says Dec 16 16:56:34 ah, right, it can raise eitehr SyntaxError of TypeError -- i love it when the docs actually have the info i'm looking for Dec 16 16:56:49 * kergoth_ thinks more docs need to cover what exceptions may be raised Dec 16 16:58:18 hey, got a filename too long exception from linux-omap-psp_2.6.32.bb. (trying to compile angstrom/beagleboard kernel) . Can I tell the script to ignore the md5 file that is too long ? :> Dec 16 17:05:38 03Koen Kooi  07org.openembedded.dev * r4926294e52 10openembedded.git/recipes/lmbench/lmbench_3.0-a9.bb: Dec 16 17:05:38 lmbench: package scripts that lmbench-run needs Dec 16 17:05:38 Signed-off-by: Koen Kooi Dec 16 17:13:54 * kergoth_ keeps adding more and more to https://github.com/kergoth/bitbake/wiki/TODO Dec 16 17:16:47 kergoth_: trying to keep yourself busy, eh? :) Dec 16 17:16:53 apparently :) Dec 16 17:18:01 kergoth_: concerning the prolem I met with perl because of LANG of the system : did you understood what can cause this ? Dec 16 17:18:14 it bugs me the way better_exec handles execution of python from our recipes -- the way it separates the "file" thats in the compiled code from the "realfile" and all, so i started experimenting with https://gist.github.com/743677 -- correctly adjusts the line numbers of the code object and even a SyntaxError exception Dec 16 17:18:25 03Michael 'Mickey' Lauer  07org.openembedded.dev * rc78db7ab87 10openembedded.git/ (2 files in 2 dirs): linux-n900: enable nat, connection tracking, etc. to get tethering Dec 16 17:18:46 ericben: that was fixed yesterday, see commit bebdb55 in bitbake master Dec 16 17:19:10 can you git add -i a chunk of a chunk? :) Dec 16 17:19:14 yes Dec 16 17:19:17 awesome Dec 16 17:19:20 hit 'e' to edit the hunk Dec 16 17:19:21 * foerster tries Dec 16 17:19:21 kergoth_: ok thanks I'm launching a build to test it Dec 16 17:19:30 you can directly remove bits and it won't stage the bits you removed Dec 16 17:19:34 but they'll stay as local changes Dec 16 17:19:37 works really well Dec 16 17:19:41 fantastic Dec 16 17:19:48 so 's' when you can, otherwise 'e' Dec 16 17:19:48 * foerster is beginning to love git Dec 16 17:19:52 is my rule of thumb Dec 16 17:20:26 ericben: was a thinko, somehow i changed the old LC_ALL=C to LANG=C when redoing bb.build/bb.process :P Dec 16 17:20:36 unintentionally Dec 16 17:24:01 kergoth_: https://github.com/foerster/bitbake/commits/separate-ui-and-server Dec 16 17:24:14 goggle right click on log -> pastebin, and puts it in your clipboard Dec 16 17:24:27 nice Dec 16 17:24:29 not cleaned up yet, but proof of concept Dec 16 17:24:46 i show the whole log as a child of the taskfailed line Dec 16 17:24:56 then you can right click that and pastebin it Dec 16 17:25:09 i threw two test commits for the server shutdown stuff on my version of that branch, seems to function, needs a bit more testing Dec 16 17:25:29 cool. my branch is rebased on master as of I think this morning. Dec 16 17:25:49 we should separate the goggle, etc bits from the main server/ui split code, so the latter can be merged, and you can keep the rest out until they're no longer WIP ;) Dec 16 17:26:11 we can do that pretty easily. Dec 16 17:26:24 goggle is pretty functional now, I like it. Dec 16 17:26:24 yeah, the commits are pretty well split, just need a -i Dec 16 17:26:32 yep Dec 16 17:27:00 you already rebased the other ones, so any last fixes to the server parts can be moved around and extracted Dec 16 17:28:46 i'll do that sometime today, its past time to get the split into master Dec 16 17:29:09 i was building with master earlier and hit ^C and got tracebacks and was sad Dec 16 17:29:10 ;) Dec 16 17:29:16 lol Dec 16 17:29:35 people keep coming in here talking about their 1.8 not working Dec 16 17:29:41 i'd just to master just for parallel parsing! Dec 16 17:29:47 s/just/jump/ Dec 16 17:30:25 i'm getting a bit concerned with the poky sync stuff, some of it has definite performance impact, even without the sstate metadata. debating releasing before the rest of that goes in, and again after Dec 16 17:31:01 they probably don't notice since their metadata is so small in comparison Dec 16 17:31:10 heh, probably Dec 16 17:31:18 adding the checksumming bits doubled the up front parse time Dec 16 17:31:29 woah Dec 16 17:31:31 anyone who's been using parallel parsing will see their ETA go from 1 minute to 2 Dec 16 17:31:33 which hurts Dec 16 17:31:40 son of a.. Dec 16 17:31:41 yea Dec 16 17:31:46 can it be disabled? Dec 16 17:31:48 :) Dec 16 17:32:01 --no-checksum Dec 16 17:32:27 i thought about that, that's the only way we could get the codebases merged Dec 16 17:34:40 what are they doing with checksum? Dec 16 17:35:25 it's for the sstate stuff. bitbake integrated packaged-staging that uses metadata checksumming to determine if the package is valid, basically Dec 16 17:35:54 hm. I have no idea what any of that is :) Dec 16 17:35:59 lets you add flags to the metadata for bits to run before/after a task, and you can add bits that tell bitbake 'this is already done, skip me and anything that leads up to me' Dec 16 17:36:11 oh, got it Dec 16 17:36:16 it's basically cached binaries. the ability to use previous build output for a recipe instead of building it from scratch Dec 16 17:36:33 by using checksums, we ensure that touching a variable that's used will invalidate the cached results Dec 16 17:36:56 i wrote most of that code, it uses pysh to parse shell and ast to parse python to track down nearly all variable references Dec 16 17:36:59 nightmare Dec 16 17:37:12 sounds like a joy Dec 16 17:39:32 03Koen Kooi  07org.openembedded.dev * r556261d7ff 10openembedded.git/recipes/lmbench/lmbench_3.0-a9.bb: Dec 16 17:39:32 lmbench: install 'lmbench' script into /usr/bin so 'lmbench-run' actually works Dec 16 17:39:32 Signed-off-by: Koen Kooi Dec 16 17:41:09 03Chris Larson  07master * r1448875530 10bitbake.git/lib/bb/cooker.py: Dec 16 17:41:09 cooker: fix remnant rq/rqdata issue from poky sync Dec 16 17:41:09 Signed-off-by: Chris Larson Dec 16 17:45:41 here we go Dec 16 17:46:04 03Bob Foerster  07master * r65b615c6df 10bitbake.git/ (6 files in 4 dirs): Dec 16 17:46:04 Run the server and UI in separate processes Dec 16 17:46:04 This uses the python multiprocessing module, both to spawn the server process Dec 16 17:46:04 and for communication between the processes. Dec 16 17:46:04 Signed-off-by: Bob Foerster Dec 16 17:46:05 Signed-off-by: Chris Larson Dec 16 17:46:16 03Chris Larson  07master * r416dfe211f 10bitbake.git/lib/bb/ (server/process.py utils.py): Dec 16 17:46:16 server: kill stdin/stdout/stderr Dec 16 17:46:16 This ensures that nothing run from the server can touch the console, in Dec 16 17:46:16 particular event handlers and python tasks, both of which can use bb.msg or Dec 16 17:46:16 the bitbake loggers to send output to the UI in a correct fashion instead. Dec 16 17:46:16 Signed-off-by: Chris Larson Dec 16 17:46:26 03Bob Foerster  07master * rbdd7813d8e 10bitbake.git/lib/bb/ (cache.py cooker.py event.py ui/knotty.py): Dec 16 17:46:26 Show the user progress when loading the cache Dec 16 17:46:26 Signed-off-by: Bob Foerster Dec 16 17:46:26 Signed-off-by: Chris Larson Dec 16 17:46:28 03Bob Foerster  07master * r588d339f2b 10bitbake.git/ (4 files in 3 dirs): Dec 16 17:46:28 server: fix interrupt handling for process Dec 16 17:46:28 SIGINT is now blocked within the server context, thus allowing the UI to fully Dec 16 17:46:28 handle all user interaction. There is no longer a need to check for Dec 16 17:46:29 KeyboardInterrupt Exceptions anywhere within the server context. Dec 16 17:46:29 Signed-off-by: Bob Foerster Dec 16 17:46:30 Signed-off-by: Chris Larson Dec 16 17:46:30 03Chris Larson  07master * r3f85132039 10bitbake.git/ (bin/bitbake lib/bb/server/process.py): Dec 16 17:46:31 Forcibly shut down the server if the UI is terminated Dec 16 17:46:31 Signed-off-by: Chris Larson Dec 16 17:46:39 nice Dec 16 17:46:47 03Chris Larson  07master * r20eb69cfc6 10bitbake.git/lib/bb/server/process.py: Dec 16 17:46:47 server: ensure the cooker is stopped when shutting down Dec 16 17:46:47 Signed-off-by: Chris Larson Dec 16 17:46:55 everything is in but the changes to the UIs now Dec 16 17:46:59 awesome Dec 16 17:47:17 if you're happy with the current UI stuff, kill the WIPs in the commits and the like and we can get those merged too Dec 16 17:47:32 i'll need to clean some stuff up first Dec 16 17:47:36 considering they were all broken before, anything that unbreaks them is an improvement, no reason to hold off ;) Dec 16 17:47:39 sure, np Dec 16 17:47:41 right Dec 16 17:47:44 that's what I was thinking Dec 16 17:47:57 they're not complete yet, but they at least function to some extent :) Dec 16 17:48:00 yep Dec 16 17:48:42 the separate-ui-and-server in my github is now your ui commits rebased against the newly merged master Dec 16 17:48:47 if you want to avoid an extra -i Dec 16 17:48:55 * kergoth_ kicks off some more test builds Dec 16 17:48:57 thanks, that'll be easier Dec 16 17:52:06 * kergoth_ works on the poky sync some more Dec 16 17:53:44 hmmm, interesting, just got a CRITICAL: message when a task failed -- the UI should have printed that with our log formatter, which should have use FATAL Dec 16 17:56:00 pb_: could you remind me what the current TSC meeting schedule is, so i can update my calendar? :) Dec 16 17:56:04 * kergoth_ keeps forgetting about it Dec 16 18:00:52 khem: ping Dec 16 18:08:18 evening Dec 16 18:11:10 well.. now kexec-tools-klibc have double segfault :) Dec 16 18:11:52 seems klcc should be revised to work w/o std{in,out,err} Dec 16 18:12:29 Jay7: what function is running those? Dec 16 18:12:51 shell functions should be fine, its only python tasks/functions that get affected by killing stdin/stdout/stderr Dec 16 18:13:09 http://tinderbox.openembedded.net/builds/111893/ Dec 16 18:13:12 may be all you need to do is change it to use oe.process instead of os.system Dec 16 18:13:31 this build would be fine with older bb Dec 16 18:13:50 yes, you said that 3 or 4 times already Dec 16 18:13:51 I don't complain against bitbake Dec 16 18:13:52 i think i got it Dec 16 18:14:04 just trying to find how to resolve this Dec 16 18:14:16 imho, you should be informed at least :) Dec 16 18:14:39 hmm, that makes no sense. Dec 16 18:14:51 do_compile is a shell task, shell task stdout goes to a log file Dec 16 18:15:47 I should eat something then will return to this Dec 16 18:16:07 * kergoth_ 'll try reproducing locally Dec 16 18:27:06 hmm, wonder if something akin to dict's setdefault would be useful in DataSmart objects for the metadata Dec 16 18:27:18 though, less of an issue using oe.types Dec 16 18:34:27 hmmmm Dec 16 18:38:51 kergoth_: I confirm your LANG fix solves the perl problem I met Dec 16 18:38:58 whew Dec 16 18:38:59 good Dec 16 18:39:03 sorry about that Dec 16 18:39:22 kergoth_: now I know this server is configured in french ;-) Dec 16 18:39:29 let me know if you hit any other issues with master, need to make sure its solid Dec 16 18:40:43 I will keep you updated in case of any problem related to bitbake Dec 16 18:41:08 thanks Dec 16 18:41:28 * kergoth_ should probably check on the state of 1.8 with OE Dec 16 18:42:06 oh, that reminds me, i was going to write a script that displays what commits can be applied to current cleanly, similar to the red/green indicators on github's fork queue Dec 16 18:42:20 might be handy for little fixes to maintanance branches Dec 16 18:42:26 Tartarus: pong :) Dec 16 18:42:53 khem: Later today I'm going to RFC sane-toolchain related stuff so that external toolchains work there (ie DISTRO=minimal) Dec 16 18:42:55 fine in theory? Dec 16 18:43:44 Tartarus: whats needed Dec 16 18:43:49 kergoth_: 1.8 seems to create problems actually from message read here yesterday. 1.10 seems OK and the person which had problems with 1.8.18 also had problems with 1.10.1 Dec 16 18:44:03 so in fact 1.10.0 worked for him Dec 16 18:44:05 hmm, will do some test builds with both and see how it goes Dec 16 18:44:07 do u want to accomodate external toolchain versions into sane toolchain ? Dec 16 18:44:13 thats odd, 1.10.1 only has a couple minor bugfixes Dec 16 18:44:35 kergoth_: all that from memory as I was on other things at the same time Dec 16 18:44:38 * kergoth_ nods Dec 16 18:44:41 thanks for the info Dec 16 18:45:03 khem: making use of TOOLCHAIN_TYPE Dec 16 18:45:10 khem: only angstrom/shr work atm Dec 16 18:45:10 ericben: people ok Dec 16 18:45:22 khem: what ? Dec 16 18:46:27 heh half cooked message Dec 16 18:46:48 jesus, 1.8 is PAINFUL Dec 16 18:46:51 (slow) Dec 16 18:47:11 anyone still running 1.8, stop that Dec 16 18:47:37 :p Dec 16 18:48:07 also, anyone in here using 1.10, please test with master, need to get it solid for a release Dec 16 18:48:11 please :) Dec 16 18:54:11 ericben: that feeling of unstability Dec 16 18:54:28 do u think its paramount because we find couple of bugs Dec 16 18:54:36 in gcc Dec 16 18:55:08 I have worked with gcc closely in OE and also in general and seen lot more issues with FSF releases Dec 16 18:55:58 I think its normal issues what we are seeing. Dec 16 18:56:32 Tartarus: OK that will be fine Dec 16 18:57:11 khem: the main question is : do we want to default a compiler which is evolving on a weekly basis ? Dec 16 18:59:29 I'm not against testing bleeding edge software but I think having a source tree which by default can show regressions from one week to the other because of toolchain changes is not a good thing Dec 16 19:11:58 * kergoth_ works on 1.8 builds with oe Dec 16 19:20:16 03Chris Larson  07master * r4b36be445a 10bitbake.git/lib/bb/utils.py: Dec 16 19:20:16 utils: drop unnecessary debug msgs from mkdirhier Dec 16 19:20:16 Signed-off-by: Chris Larson Dec 16 19:26:30 * kergoth_ sighs Dec 16 19:31:01 hi, I'm crosscompiling R to arm arch (adding a recipe for OpenEmbedded), and when I run autotools, I've got everything to compile fine, but after the compilation stage, I don't know why the Make tries to run R, and of course it fails the build as the binary is for arm and I'm on an x86 PC. Dec 16 19:32:09 any hints on where should I look to disable this execution? I suppouse in the configure.ac and Makefile.am, but apart from that I'd like some more specific hint. Dec 16 19:32:17 right.. that issue isn't an oe issue, its a 'how do i crosscompile R' issue. ask them, or read the makefiles and figure out why its doing it Dec 16 19:33:13 kergoth_, ok, already asked in #r, but there's few movement there :) Dec 16 19:33:41 its hard to say why it's running it. if its happening during the compile/install, not the configure, then it wont be in configure.ac, look at the Makefile.am/Makefile.in files Dec 16 19:34:06 pespin: there should be a bb recipe already for R Dec 16 19:35:00 after all koen got 1000 currency units for that feat or was it even 2k? Dec 16 19:35:19 scmdata=keep is cute -- I think I can get rid of my kludge in my recipe now which went and found the git dir Dec 16 19:35:43 * kergoth_ really looks forward to the bb.fetch rewrite Dec 16 19:35:55 dm8tbr, only r-native Dec 16 19:36:37 pespin: http://farm5.static.flickr.com/4117/4827023258_ca07c8f9d6_o.png - this is a beagle-board screenshot. I'm pretty sure he cross-compiled it Dec 16 19:36:57 the recipe isn't in openembedded/recipes/r/ at least.. Dec 16 19:37:18 dm8tbr, http://cgit.openembedded.org/cgit.cgi/openembedded/tree/recipes/r Dec 16 19:37:35 yeah, I have the binary and evyerhing needed to run it I think Dec 16 19:37:51 so I could run it in the arm board Dec 16 19:38:09 but as the make is failing, OE can't build the package Dec 16 19:38:56 pespin: ah on the ML he said something on his desktop was broken and he built directly Dec 16 19:42:25 03Chris Larson  07master * raf81d52e48 10openembedded.git/classes/base.bbclass: Dec 16 19:42:25 base: resurrect messages for bitbake 1.8 Dec 16 19:42:25 Signed-off-by: Chris Larson Dec 16 19:42:44 dm8tbr, here's how my recipe looks like atm: http://paste.pocoo.org/show/305731/ Dec 16 19:43:14 btw, I have to add that final #sed to some precompile stage Dec 16 19:43:22 where should I put it? Dec 16 19:43:39 some kind of do_compile_prepend() ? Dec 16 19:44:27 pespin: I just jumped in because I remembered koen's stunt. better talk to him directly :) Dec 16 19:45:13 ericben: ok. sometimes we find bugs in compilers too it does not change to frequently in general Dec 16 19:46:01 ericben: Take a release of gcc and you will spend time stabilizing that too Dec 16 19:46:23 given the amount of devs we have in OE it will be very hard if no one tests it Dec 16 19:46:50 and we will never have quality that good. Dec 16 19:47:30 ericben: while from release standpoint I think its too much to expect from a .dev branch I think Dec 16 19:48:06 compilers need space to develop as well and if we dont do that on .dev where else can it be Dec 16 19:48:07 bitbake 1.8 is building away happily on oe master Dec 16 19:48:15 it was building before too, just not as verbosely Dec 16 19:50:23 foerster is ok Dec 16 19:50:31 l1nux: good Dec 16 19:50:39 but i have other problem :( Dec 16 19:50:48 Missing or unbuildable dependency chain was: ['x11-image', 'xf86-video-radeon'] Dec 16 19:51:30 is needed by ibm A21m Dec 16 19:51:51 any solution ? Dec 16 19:52:53 who specified video-radeon? At a quick glance, I only see video-radeonhd in the recipes (note the hd at the end) Dec 16 19:55:38 l1nux: your machine config probably needs hd at the end of radeon Dec 16 19:55:45 yeah Dec 16 19:55:50 i try it :) Dec 16 19:56:38 MACHINE_DRI_MODULES = "radeonhd" ? Dec 16 19:57:07 XSERVER xf86-video-radeonhd Dec 16 19:57:19 right ? Dec 16 19:58:07 the xf86 line for sure, don't know on the other one. I don't use radeonhd, and don't know what MACHINE_DRI_MODULES does. Dec 16 19:58:24 ok Dec 16 19:58:24 the XSERVER xf86 line is the error you received, so that's all I know. Dec 16 20:01:05 khem: I understand your point of view. Dec 16 20:02:06 ericben: and similarily I understand yours Dec 16 20:02:21 ericben: but look we are making gcc better Dec 16 20:03:11 ericben: eventually these arm improvements will go upstream so instead of hitting them in 4.6 its better to be proactive Dec 16 20:04:02 ericben: any issues you find in GCC we have I will be ready to help in fixing/finding that Dec 16 20:04:32 khem: I understand this and have seen your reactivity ! Dec 16 20:09:48 khem: that's always the problem to chose between bleeding edge's improvements and a sort of stability with older versions Dec 16 20:21:22 ericben: Yeah, I think I always try to make sure that gcc modification w.r.t code changes are well tested and are only bugfixes Dec 16 20:21:34 ericben: with Linaro I took a step further I understand Dec 16 20:21:55 but then to make OE attractive to lot of new cortex users its a must Dec 16 20:22:20 We would agree that there are lot of users who have beagleboards now a days Dec 16 20:23:58 ericben: ironically, as I speak I am about to push a change which fixes the samba issue and upgrade to 4.5.2 :) Dec 16 20:25:04 ericben: hopefully it wont scare you off Dec 16 20:25:07 l1nux: the error says xf86-video-radeon and you say xf86-video-radeonHD? Dec 16 20:25:17 khem: :-D Dec 16 20:25:45 here at work if something breaks due to gcc then hell breaks on us Dec 16 20:26:12 I'm not scared. I've spent 2 days working on the maintenance of an fltk/nanox dev so I had time to think to other things ;-) Dec 16 20:26:13 devs keep injecting bugs into software but thats ok I think its unfair for compiler Dec 16 20:26:15 heh Dec 16 20:26:50 ericben: ok so you will try it out :) Dec 16 20:27:24 JaMa|Off i have modified my machine conf from xf86-video-radeon to xf86-video-radeonhd Dec 16 20:27:56 khem: I always try master, else I wouldn't meet these problems :-) Dec 16 20:28:13 and work now Dec 16 20:28:28 l1nux: ah sorry I haven't read whole backlog before reply Dec 16 20:28:29 thanks to foerster :) Dec 16 20:28:37 ericben: cool thx Dec 16 20:31:04 JaMa: Do u have any pain points for gcc 4.5 ? Dec 16 20:31:35 03Michael 'Mickey' Lauer  07org.openembedded.dev * r54c9371f6a 10openembedded.git/recipes/linux/linux-nokia900/defconfig: linux-n900: adjust defconfig as per last commit Dec 16 20:33:30 khem: currently only that u-boot Dec 16 20:33:56 khem: I've upgraded to u-boot master, but still debuging it (doesn't boot even when build with old toolchain sofar) Dec 16 20:34:13 Hi Dec 16 20:34:30 Please help me to fix this error Dec 16 20:34:44 configure: error: C compiler cannot create executables | See `config.log' for more details. | ERROR: Function do_configure failed NOTE: package libtool-cross-2.2.6b-r27.1: task do_configure: Failed ERROR: TaskFailed event exception, aborting Dec 16 20:35:15 I searched it but unable to find the exact solution Dec 16 20:35:36 how to fix thhis Dec 16 20:37:49 khem: did you notice my msg about weak assignment in sane-toolchain for LIBTOOL_HAS_SYSROOT ? Dec 16 20:41:51 JaMa: older toolchain you mean gcc-4.5 without linaro ? Dec 16 20:42:06 JaMa: IOW vanilla gcc 4.5 Dec 16 20:43:10 khem: no last working is gcc-4.4.4 Dec 16 20:43:45 khem: and for those tests I was using even older with gcc-4.2.1 (packaged by openmoko back then) Dec 16 20:45:52 err hmmm Dec 16 20:45:55 if you need my help Dec 16 20:45:59 then let me know Dec 16 20:47:09 khem: I can give you workdirs working and failing, but I'll try a bit more to make latest u-boot work Dec 16 20:48:03 that will be usefull in future and if the problem with gcc-4.5 is still there, then there should be more people complaining Dec 16 20:48:25 JaMa: ok. Dec 16 20:48:54 yeah I would say try to get is reproduced on latest uboot Dec 16 20:48:59 then we have a better case Dec 16 20:58:44 JaMa: I think we need to figure where it really dies Dec 16 20:58:52 otherwise workdirs wont make much use to me Dec 16 21:00:11 I know, today I've found some new documentation about debuging relocation issues in u-boot so hopefully it will help me with that upgrade Dec 16 21:16:35 JaMa: hello Dec 16 21:16:50 JaMa: did you have a look into the last revision of the class? Dec 16 21:16:55 JaMa: I'd like to push it Dec 16 21:18:52 hello Dec 16 21:19:16 otavio: yes looks ok Dec 16 21:19:36 otavio: wrt that tagging example.. Dec 16 21:21:19 JaMa: ?? Dec 16 21:22:00 otavio: I didn't want to show that only v1.0 is working ok.. but that you cannot use it if there are completely unrelated tags between Dec 16 21:22:00 otavio: ie in your example release-1, release-test, release-2 wont work Dec 16 21:22:01 sortable GITVERTAG Dec 16 21:22:21 if I undestand git describe right it show latest tag whatever it is, right? Dec 16 21:23:31 otavio: so for example in OE repo we have nice sortable tested_2010-.. tags, but release-2010.12-branchpoint between will break sortable GITVERTAG between tested_2010-11-04, release-2010.12-branchpoint, tested_2010-11-12 Dec 16 21:25:35 otavio: so you should be pretty sure that source git repo is using and will be using consistent AND sortable tag names, otherwise you'll ship something to target which cannot be upgraded without bumping PE Dec 16 21:29:27 JaMa: sure I fully agree however it is not "class bussiness" to handle it Dec 16 21:30:00 yes.. but warning in GITVERTAG example can save few damned souls :) Dec 16 21:30:35 JaMa: please propose a text about it then. I have doubt how to describe it Dec 16 21:31:25 gitver lets you provide a hook to mangle the describe output, perhaps you need something like that, or a way to inject a tag filter or something Dec 16 21:31:29 ok, I'll send it as reply to your patch, maybe some native speaker will comment or make it easier to understand Dec 16 21:32:12 maybe define expected tag format and warn/reject everything else Dec 16 21:38:48 kergoth_: I agree with that but let's do that later. Dec 16 21:39:07 kergoth_: I want to not diverge from oe and I need to use it Dec 16 21:39:19 kergoth_: in fact I have prepared a patch for freerdp to use it Dec 16 21:58:41 JaMa: I have delegated some patched in patchwork to you Dec 16 22:15:25 03Khem Raj  07master * rbd1714a382 10openembedded.git/recipes/gcc/ (4 files in 3 dirs): Dec 16 22:15:25 gcc-4.5: Update to 4.5.2 release Dec 16 22:15:25 * port the patch for PR 46883 which fixed samba build issue Dec 16 22:15:25 * Drop PR 42172 reverting patch as its not required anymore Dec 16 22:15:25 * Refresh patches to apply cleanly on 4.5.2 Dec 16 22:15:25 Signed-off-by: Khem Raj Dec 16 22:15:34 03Khem Raj  07master * r3010d62670 10openembedded.git/recipes/eglibc/ (eglibc_2.12.bb eglibc_svn.bb): Dec 16 22:15:34 eglibc_2.12.bb/eglibc_svn.bb: Grab more bug fixes move to svn tip Dec 16 22:15:34 Signed-off-by: Khem Raj Dec 16 22:50:31 03Tom Rini  07master * r3f93cb690b 10openembedded.git/recipes/db/ (db5.inc db_5.0.21.bb): Dec 16 22:50:31 db5: Switch to using our libtool Dec 16 22:50:31 Signed-off-by: Tom Rini Dec 16 23:11:32 hi grg Dec 16 23:11:40 khem, mornin' Dec 16 23:11:47 aussies are not doing good this ashes Dec 16 23:11:55 no. we suck. badly Dec 16 23:12:21 (england still suck, we just suck worse :)) Dec 16 23:12:26 haha Dec 16 23:12:27 grg: btw, once you have asked my if I see any problem with making -force-reinstall just sequence of remove and install Dec 16 23:12:58 JaMa|Off, that sounds vaguely familiar. I think i did that. Dec 16 23:12:59 grg: not sure how that special case for reinstall handled it before.. but -force-reinstall libc6 is very bad idea now Dec 16 23:13:04 grg: I have some important fixes done for gcc and eglibc so give it a whirl Dec 16 23:13:22 khem, sure Dec 16 23:13:26 JaMa|Off: I will test your x11 patches Dec 16 23:13:35 grg: same for ie -force-reinstall opkg Dec 16 23:14:00 JaMa|Off: if I build an x11-image using angstrom or minimal it should pick the new versions right provided I cherry pick your patch Dec 16 23:14:06 khem: great thanks Dec 16 23:14:16 khem: yes, right Dec 16 23:14:46 or even with qt4-x11-base-image too Dec 16 23:14:47 right Dec 16 23:15:22 I think we should update wiki news section more often Dec 16 23:15:44 like when we get major/key component updates like x11, QT, gcc etc. Dec 16 23:16:19 khem: what about that weak assignment with LIBTOOL_HAS_SYSROOT in sane-toolchain.inc? Dec 16 23:17:01 JaMa|Off: What happens ? it does not get picked Dec 16 23:17:09 oh I see Dec 16 23:17:14 it should be conditional Dec 16 23:17:18 khem: I don't build qt4-x11-base-image here but yes it should build and work ok with X11R7.6 (that's practically what you get without any include) Dec 16 23:17:28 right Dec 16 23:17:53 khem: it is conditional (?=), but bitbake.conf sets it to default "no" Dec 16 23:18:05 hrmmm Dec 16 23:18:20 JaMa|Off: right I missed that Dec 16 23:18:29 khem: and sane-toolchain.inc has ?= "yes" and it's included always later from ${DISTRO}.conf Dec 16 23:18:32 khem: +1 no updating news section Dec 16 23:18:32 can you post a patch if you have it in kitty Dec 16 23:18:38 'we are alive' Dec 16 23:18:39 yeah I got it Dec 16 23:18:44 JaMa|Off, here is the change, in case you want to propose a different behaviour: http://code.google.com/p/opkg/source/detail?r=538 Dec 16 23:18:56 Jay7: I will ask for write access to wiki front page Dec 16 23:19:17 khem: just make the assignment "=" in sane-toolchain? or do you know something better? Dec 16 23:19:27 hm.. s/no/to/ in my phrase about news Dec 16 23:19:51 JaMa|Off: yes hard assignment Dec 16 23:19:56 but try it out Dec 16 23:20:24 khem: changing default in bitbake.conf to "yes" would force ie angstrom and other distros with older libtool to override it in their config.. Dec 16 23:20:27 ok will try Dec 16 23:20:48 I dont want to change bitbake.conf Dec 16 23:20:57 untill we make 2.4 preferred as default Dec 16 23:21:19 I've resolved it by this: http://git.openembedded.org/cgit.cgi/openembedded/diff/conf/distro/shr.conf?id=8e2783a27920cb01d1a094eeb014d31b7df1d274 Dec 16 23:21:45 * Jay7 -> sleep Dec 16 23:22:01 JaMa|Off: yeah seems ok to me propogate that to sane-toolchain.inc Dec 16 23:22:04 with comment please Dec 16 23:22:51 yup, will test tonight with toolchain upgrades and post tomorrow Dec 16 23:22:58 thanks and good night Dec 16 23:23:24 khem: can you look on qemux86 btw? Dec 16 23:23:33 hmmm Dec 16 23:23:38 whats wrong Dec 16 23:23:41 momen Dec 16 23:23:48 Jay7: use 12.5 Dec 16 23:23:54 http://tinderbox.openembedded.net/builds/111776/ Dec 16 23:23:55 .13 has issues I know Dec 16 23:24:10 it is with cross-localedef Dec 16 23:24:34 ah yes Dec 16 23:24:36 thats fixable Dec 16 23:24:49 I've raised this here http://thread.gmane.org/gmane.comp.handhelds.openembedded/40566 Dec 16 23:25:02 seems subject line was very uninformative :) Dec 16 23:25:39 heh Dec 16 23:25:49 Jay7: what DISTRO/MACHINE Dec 16 23:26:40 khem: qemux86 with any of {anstrom-2008.1, angstrom-2010.x, minimal) Dec 16 23:26:56 if you mean "unknown arch:i686 for locale_arch_options" Dec 16 23:26:59 really hmm Dec 16 23:27:02 bad Dec 16 23:27:04 me Dec 16 23:27:23 I think this should be easy to fix :) Dec 16 23:30:32 now I'm sleeping :) Dec 17 00:03:36 khem: still around? Dec 17 00:06:02 yes Dec 17 00:06:07 24/7 Dec 17 00:06:19 and I am buying a phone which can run IRC Dec 17 00:06:31 heh Dec 17 00:06:36 but terminal should have 256 color suppport Dec 17 00:06:38 going to post the sane-toolchain.inc stuff in a minute Dec 17 00:06:41 sure Dec 17 00:11:55 there Dec 17 00:30:47 well, should be, hmmm Dec 17 00:31:02 I sent them to myself via git send-email first, so i know that box can mail, in general Dec 17 00:31:39 * Tartarus mails to gmail Dec 17 00:32:13 gmail got 'em, wth.. Dec 17 00:34:17 better be sending from the address that's subscribed to the list ;) Dec 17 00:38:17 heh, yes, it's alleging to come from there Dec 17 00:40:32 Anyone around with admin privs? Dec 17 00:43:16 what happened Dec 17 00:56:17 patches aren't showing up Dec 17 01:00:00 Jay7: not getting a segfault building kexec-tools-klibc here. what distro/machine? Dec 17 02:13:23 http://kevinw.github.com/2010/12/15/this-is-your-brain-on-vim/ Dec 17 02:49:45 If there is any interest, I have written some recipes for TPM: http://git.alwaysinnovating.com/cgit.cgi/ai.openembedded.dev/tree/recipes/aios/tpm **** ENDING LOGGING AT Fri Dec 17 02:59:59 2010