**** BEGIN LOGGING AT Fri Sep 10 02:59:57 2010 Sep 10 03:45:11 did someone have a doc how to generate an image for efikamx Sep 10 03:45:26 if possible with e17 as windows manager Sep 10 03:53:56 I've been having a devil of a time building Openembedded successfully. I'm working with a Beagleboard C2 and I've done this before, but lately I can't seem to get anywhere. Should I expect an Angstrom 2008.1 build from org.openembedded.dev to build and result in working images right now? Sep 10 03:55:22 if it doesn't build, open a bug Sep 10 03:55:30 failing to build is never intended Sep 10 03:57:47 Even on the .dev branch? I want to make sure that I'm not being *more* of an idiot than usual, first. Sep 10 03:59:39 yes. Sep 10 04:00:10 OK. Thank you. I'll triple check things and file a bug if I can't resolve it. Sep 10 04:00:12 .dev is actively developed, so things *will* break from time to time, but it's never intended, never expected, and should always be reported Sep 10 04:00:48 I'm building on Ubuntu 10.04 at this point... no additional expected weirdness there? Sep 10 04:02:34 not really, no. lots of us use it Sep 10 04:03:04 The only other "different" thing I'm doing is running with INHERIT = "rm_work"? Sep 10 04:03:19 lots of people do that too, including me Sep 10 04:03:27 and khem, and .. Sep 10 04:03:29 :) Sep 10 04:04:42 Oh? 8> Sep 10 04:07:04 I'll try it without the rm_work, and thanks again for the help. Sep 10 04:10:08 todd_: http://wiki.openembedded.net/index.php/Testing Sep 10 04:10:14 todd_: see the first two entries Sep 10 04:10:49 can always try building from the tag indicated as last successful build there Sep 10 04:11:23 Great. Thanks. I was ignorant of that particular resource for knowing "what should be working". Sep 10 04:13:01 it's quite recent Sep 10 04:13:20 many attempts have been made to improve QA, this one seems quite promising Sep 10 04:15:23 Yes. It is a nice looking page and easily understood. Good mix of hosts, etc. Cool! Sep 10 06:02:39 kergoth: Tartarus thanks I found the problem. It was missing entry in insane.bbclass Sep 10 06:16:29 gm Sep 10 06:26:36 03Frans Meulenbroeks  07org.openembedded.dev * r0cbb6ba712 10openembedded.git/recipes/pulseaudio/pulseaudio.inc: Sep 10 06:26:37 pulseaudio.inc: set INC_PR back to 10, was actually dropped to 0 in a previous commit Sep 10 06:26:37 Signed-off-by: Frans Meulenbroeks Sep 10 06:36:13 03Koen Kooi  07org.openembedded.dev * r61879a6709 10openembedded.git/recipes/openjdk/openjdk-6-common.inc: openjdk-6: fix including things like 'openjdk-6-jdk' into tasks and images Sep 10 06:49:21 03Koen Kooi  07org.openembedded.dev * re2715ccb68 10openembedded.git/recipes/opencv/openscenegraph_2.9.7.bb: openscenegraph: use gcc atomic builtins to make cmake happy Sep 10 07:21:55 good morning Sep 10 07:39:20 morning Sep 10 07:42:46 hi hrw Sep 10 07:51:47 03Koen Kooi  07org.openembedded.dev * r5a7d77d927 10openembedded.git/ (3 files in 3 dirs): Sep 10 07:51:47 linux-omap-psp 2.6.32: OMAP: DSS2: OMAPFB: swap front and back porches for both hsync and vsync Sep 10 07:51:47 Framebuffer's left and right margins are relative to the active pixel Sep 10 07:51:47 area. Front and back porches are relative to the sync area. Sep 10 07:51:47 Left margin was wrongly assigned to front porch (and right to back), Sep 10 07:51:47 this patch fixes it. Sep 10 07:52:28 #/msg nickserv identify blablabla Sep 10 07:54:25 likewise: you might want to pick a different irc password :-) Sep 10 07:54:42 I know, I suck :-) Sep 10 07:54:56 in pidgin you can configure autoidentify Sep 10 07:55:10 made that mistake myself also once Sep 10 07:58:09 likewise: btw, gratz on your SPE proggie Sep 10 08:00:24 morning Sep 10 08:00:44 eFfeM_work: tnx, performance sucked though. Sep 10 08:01:05 that's probably the reason why Freescale will drop the DSP instructions in their next-gen PPC. Sep 10 08:01:13 and why no examples can be found on Internet Sep 10 08:01:54 likewise: ah, misread the log, thought you achieved a 60% gain, not 60% of the original Sep 10 08:02:42 Why must autoconf suck so greatly. Sep 10 08:02:51 mroning Sep 10 08:03:12 nar__: that is because it wants to stay on par with libtool Sep 10 08:05:18 heh! Sep 10 08:05:55 nar__: I call it autohell Sep 10 08:06:22 amen Sep 10 08:07:07 I've got a double-whammy... hdf5 compiles a 'test program' in addition to the autoconf stuff, that ... generates compiler behavior information! Sep 10 08:07:38 Which to me says, somebody said 'we should use autoconf' and then assigned the task to someone who's never used it. Sep 10 08:08:23 Question: should we have recipes like glibc-native? I think I bumped into an issue where the regular recipe builds but the native one does not, apparently for the native package I pick up a host include file that is not in my sysroots and on target comes from glibc/eglibc (btw it is crypt.h) Sep 10 08:09:04 I think if you want a true canadian cross-compiler you need to build glibc to bootstrap your compiler against Sep 10 08:09:29 If you're building your toolchain with OE Sep 10 08:10:10 But it does seem goofy doesn't it. Sep 10 08:12:22 will check a build log for it later on Sep 10 09:18:52 Ugh. automake, autoconf, etc always seem to install incorrectly for me.. I always get unquoted string errors. Sep 10 09:21:13 seems newer automake versions are more picky wrt unquoted strings Sep 10 09:37:57 eFfeM_work: yes, I reached additional 60% performance (factor 1.6 improvement), but my aim was 400% improvement. Sep 10 09:38:20 morning Jin^eLD (and all) Sep 10 09:39:08 ah ok, thought it would do the same instruction in parallel for two sets of data, in that case 60% seems about right to me Sep 10 09:39:14 hey likewise Sep 10 09:39:17 unless teh instructions are more powerful Sep 10 09:39:29 vi Sep 10 09:46:05 mickeyl: good morning Sep 10 09:58:52 mahlzeit Sep 10 09:58:56 oops wrong window Sep 10 09:59:07 good plan though Sep 10 09:59:08 enjoy Sep 10 09:59:51 :) thx Sep 10 10:13:13 Hi! Is it possible to fetch sources from git repository via HTTP? Tried to set SRC_URI="http://;protocol=git", but OE fetches html file instead of sources Sep 10 10:15:28 gim: try something like here SRC_URI = "git://github.com/nytowl/BT-Configure.git;protocol=http" Sep 10 10:25:38 Jin^eLD: thanks, it worked :) But now OE fetch fails with "fatal: Not a valid object name 1". Also it is suspicious for me that it prints "From http:///project" instead of "From http:///project.git", though in recipe SRC_URI contain ".git" suffix Sep 10 10:31:37 gim, best try an example from another recipe and use that as a starter Sep 10 10:42:40 object name 1 means the revision should be specified (this is the hexadecimal commit id) - it defaults to 1 but this is invalid for GIT Sep 10 10:42:44 gim: ^^^ Sep 10 10:43:28 gim: See recipes/linux/linux-powerpc-fsl_git.bb where I recently did that. Sep 10 10:44:42 gim: http://cgit.openembedded.org/cgit.cgi/openembedded/diff/recipes/linux/linux-powerpc-fsl_2.6.35.bb?id=72fed19769b30515c18636acf30426b5d834fcaf Sep 10 10:49:35 wb Sep 10 10:49:44 oops wrong window once again :) dammit Sep 10 11:09:53 Jin^eLD: the first thing I did this morning was posting my freenode password, so you are forgiven :) Sep 10 11:11:21 likewise: it is in the log so it is there for all to recover Sep 10 11:11:39 likewise: pity to loose such a strong password that way ;-) Sep 10 11:11:40 03Chase Maupin  07org.openembedded.dev * r46bcffb8d7 10openembedded.git/recipes/psplash/ (3 files in 2 dirs): (log message trimmed) Sep 10 11:11:40 psplash-ti: update configurability patch Sep 10 11:11:40 * Update the configurability patch to work with revision 422 of Sep 10 11:11:40 the psplash sources. This patch causes the psplash background Sep 10 11:11:40 to be black which is the color we want to use. Sep 10 11:11:40 * Bump the PR Sep 10 11:11:40 Signed-off-by: Chase Maupin Sep 10 11:11:42 03Andreas Mueller  07org.openembedded.dev * r3e84f62cc7 10openembedded.git/recipes/clutter/ (clutter/workaround-broken-mipmaps.patch clutter_1.2.12.bb): Sep 10 11:11:42 clutter: the font display problem is worked around. Sep 10 11:11:42 * see http://lists.clutter-project.org/pipermail/clutter-devel-list/2010-September/000071.html Sep 10 11:11:43 for more details Sep 10 11:11:43 Signed-off-by: Andreas Mueller Sep 10 11:11:44 Signed-off-by: Koen Kooi Sep 10 11:11:49 03Andreas Mueller  07org.openembedded.dev * re51ed5feba 10openembedded.git/recipes/clutter/clutter-box2d_0.10.0_git.bb: Sep 10 11:11:49 clutter-box2d: renamed clutter-box2d_0.10.0_git.bb -> clutter-box2d_git.bb Sep 10 11:11:49 Signed-off-by: Andreas Mueller Sep 10 11:11:49 Signed-off-by: Koen Kooi Sep 10 11:12:17 Jin^eLD: I think, "fatal: Not a valid object name 1" appears when you are pulling from git and there is no S = "${WORKDIR}/git" line Sep 10 11:26:50 khem: hi, gettext-native builds OK, but now there's a problem building guile-native, it fails on configure stage due to undefined AM_GNU_GETTEXT macro :\ Sep 10 11:36:34 is there some minimal content in a .bbappend file to be identifiable as .bbappend? Sep 10 11:36:35 I get Sep 10 11:36:37 ERROR: /home/frans/workspace/gtd/tv.axon.openembedded.2010/recipes/urjtag/urjtag_git.bbappend is not a BitBake file while parsing /home/frans/workspace/gtd/tv.axon.openembedded.2010/recipes/urjtag/urjtag_git.bbappend Sep 10 11:36:59 and the content: Sep 10 11:37:00 SRC_URI += "file://parts.patch;striplevel=2 " Sep 10 11:37:08 plus two lines of comment starting with # Sep 10 11:53:30 nevermind got it, accidently picked up an older bb Sep 10 11:55:48 eFfeM_work: I have changed the password to something of similar strength :-) Sep 10 11:56:56 ah, you have an algorithm! guess you'll append one, so now it will be blahblahblahblah, if so it indicates it happened before Sep 10 12:16:06 ka6sox-work: build slugos-image for le and be with only a minor problem (strip error in kernel) :-) Sep 10 12:36:48 has anyone had trouble building firefox: http://tinderbox.openembedded.net/packages/743500/ Sep 10 13:00:26 cbrake, I hear sakoman has Sep 10 13:13:10 cbrake: ff build will fail if your build machine has the same version of ff installed Sep 10 13:13:55 sakoman: wow, that is special Sep 10 13:14:02 cbrake: the falure looks like this: http://build.sakoman.com/public/logs/task/24149.txt Sep 10 13:14:18 cbrake: indeed, quite annoying Sep 10 13:14:29 sakoman: yup, same thing here Sep 10 13:14:59 cbrake: my "solution" was to upgrade my host to ff 3.6.9 Sep 10 13:15:17 sakoman: heh, ok Sep 10 13:16:05 cbrake: I didn't feel like (nor have the time) to go wading through the ff build system Sep 10 13:16:58 cbrake: this has happened in the past too -- I think there is something broken in cross builds w/ ff Sep 10 13:17:54 sakoman: yeah, I hear that Sep 10 13:48:26 ericben|away: are you sure the virtual/egl patch works? Sep 10 13:48:39 I'm doing bitbake world for calamari / minimal and I get http://www.pastie.org/1150120 Sep 10 13:49:04 this is with dev head, you see that at the top the patches from koen from earlier today are pulled in, so yours should also be there i guess Sep 10 13:50:17 argh, -ETMTOSO Sep 10 14:04:16 morning Sep 10 14:05:40 hi kergoth_ Sep 10 14:32:53 thoughts on using something like http://gist.github.com/573056 instead of our silly "urldata" thing? Sep 10 14:34:12 if it works then why not Sep 10 14:35:22 I was also thinking about doing a different sort of a fetcher class Sep 10 14:35:30 one that you pass in all the options for, no direct metadata usage Sep 10 14:35:41 and one which fetches directly into WORKDIR, using DL_DIR as a *cache* Sep 10 14:35:45 not a primary destination Sep 10 14:36:01 this way things like git can just clone from DL_DIR into WORKDIR instead of using the usual unpack method Sep 10 14:48:45 Hrm. I have a goofy idea. For autoconf stuff that tries to execute programs.. I bet you can just wrap that command in something that scp's the binary to a staging machine and reads back the output. Sep 10 14:49:02 Clunky as hell but seems like it'd reduce the churn of porting poorly constructed autoconf scripts Sep 10 14:49:19 Which I've been in the middle of for a week now Sep 10 14:49:26 does someone in OE did multilib builds? Sep 10 14:49:33 ppc32/64 for example Sep 10 14:49:38 ! Sep 10 14:49:50 nar: would be easier to do what scratchbox does Sep 10 14:50:22 unfamiliar with scratchbox. I'm googlin' Sep 10 14:50:23 it uses binfmt_misc to automatically do what you discussed without wrapper scripts -- any attempt to run an ARM (for example) binary would automaitcally be done remotely Sep 10 14:50:31 sexy Sep 10 14:50:34 either to qemu or real hardware Sep 10 14:51:12 hmm Sep 10 14:52:00 dealing with crosscompile issues in autoconf scripts isn't *that* difficult, its a specific set of steps, most of the time. someone should probably document the most common ones somewhere though :) Sep 10 14:52:32 what i'd love to do is use the binfmt_misc method to offload the execution as a way to prepopulate the cached test results for our site files, for new platforms Sep 10 14:52:44 but continue to use the site method for most builds. best of both worlds Sep 10 14:52:50 The package I'm stuck on right now has autoconf, then builds a special binary that produces a couple pages of compiler settings that it then ingests as well Sep 10 14:52:58 So it's a little tricker than the usual autoconf stuff. Sep 10 14:53:19 Trying to do it 'right' and not just hack and slash on their configure.in Sep 10 14:53:43 Hmm, so thats an interesting idea kergoth Sep 10 14:53:59 There's a big tarball of autoconf macros someone distributes. Sep 10 14:54:03 i assume those settings are specific to the hardware -- i'd add a mechanism to only run the binary if a file with the settings doesn't already exist Sep 10 14:54:24 then populate that file yourself based on previous runs Sep 10 14:54:34 I bet you could iterate over them and whatever applicable tests run, cache the results such that CONFIG_SITE would pick them up automatically Sep 10 14:54:55 Thats such a good idea I'm going to say it's my idea. Sep 10 14:55:00 We'll call it 'nar's technique' Sep 10 14:55:09 nar: there's something in contrib in oe, someone had started to gather the macros we need to run to populate the site files for a new platform Sep 10 14:55:10 (I kid!) Sep 10 14:55:17 but it's nowhere near complete Sep 10 14:55:19 hmm Sep 10 14:55:49 http://ac-archive.sourceforge.net/ Sep 10 14:55:54 see contrib/site-conf/ Sep 10 14:55:56 Thats the package I'm thinking of Sep 10 14:56:24 the autoconf macro archive doesn't include the macros many packages include and ship themselves Sep 10 14:56:35 its just a good resource if you're needing a macro to check for something when working on an app, really Sep 10 14:56:44 yep Sep 10 14:56:49 I gotcha Sep 10 14:57:23 * kergoth_ has a clone of it somewhere from savannah, there's some good stuff in there - i particularly like the macros that generate relative paths from one location to another, e.g. bindir to libdir, so you could use that to assist in making an app relocatable Sep 10 14:58:05 hrm Sep 10 14:58:41 nar: have you watched the video or read the slides of landley's firmware linux presentation, the one where he goes over the issues inherent to crosscompilation? Sep 10 14:58:50 I have not Sep 10 14:58:57 You send me to the google once again Sep 10 14:59:00 it's apt, and accurate, as far as i can tell. a useful overview Sep 10 14:59:09 http://free-electrons.com/blog/elc-2010-videos/ Sep 10 14:59:25 http://elinux.org/images/4/44/Developing_for_non_x86_targets_using_QEMU.pdf and http://free-electrons.com/pub/video/2010/elc/elc2010-landley-qemu-crosscompiling.ogv Sep 10 14:59:39 he has pretty strong opinions, but there's still useful information there whether you agree with his conclusions or not Sep 10 14:59:47 i think we can all agree crosscompilation is a big pain in the ass Sep 10 14:59:54 just differ on how we deal with it Sep 10 15:00:17 For me it's fine until autoconf/libtool come into the picture. Sep 10 15:00:31 that's pretty common :) Sep 10 15:00:45 For something that's supposed to help distributing software... it kind of does the opposite. Sep 10 15:02:19 At this point tho I've done pretty good. Got a heap of self-ported stuff in our OE tree Sep 10 15:02:27 nice Sep 10 15:02:28 <3 OE Sep 10 15:02:52 If I was any good with git I'd try contributing. heh. Sep 10 15:03:02 there's a lot i'd like to see done differently, so i'll just say we suck less than most of the alternatives :) Sep 10 15:03:12 HEH Sep 10 15:03:14 check out the Pro Git book online, its free and really quite good Sep 10 15:03:33 When our chip vendor suggested we use 'buildroot' I laughed in his face before I could stop myself. Sep 10 15:03:40 hehe Sep 10 15:03:48 buildroot is great if you're doing bringup and the like Sep 10 15:03:52 but its not a distro Sep 10 15:03:57 long term maintenance is not so fun Sep 10 15:04:25 I still need to document my single image thing Sep 10 15:05:02 I made a nifty thing where you take a kernel+initramfs, concatenate it with an ext2 filesystem into a single image -- you get the sizes and offsets and compile that into a u-boot script that you stick on the head Sep 10 15:05:31 From u-boot you run a single command and it loads the whole thing in one swoop Sep 10 15:05:45 Meaning you can then distribute a single file to customers for upgrades, which, for our app is kinda the way to go Sep 10 15:05:53 nice Sep 10 15:06:22 I know too much about u-boot now. Sep 10 15:07:40 heh, i used to - someday i'll start keeping a engineering journal or something so i stop forgetting everything i used to know.. Sep 10 15:08:20 heh! I know what you mean. If I had to go back and do board bring up on our last chip again I'd be lost without my wiki pages Sep 10 15:08:29 'We get to start over from scratch!' Sep 10 15:27:16 http://marcin.juszkiewicz.com.pl/2010/09/10/ubuntu-cross-compilers-part-2/ Sep 10 15:30:54 hrw, nice :) Sep 10 15:33:51 kergoth_: I have to admit that lot of work was possible due to openembedded code ;D Sep 10 15:46:06 hmmmm. Sep 10 15:46:18 anyone have any ideas on how i can put together a reliable unit test for bitbake -k? Sep 10 15:46:24 hmm.. perhaps... Sep 10 15:47:08 How so, like, verifying each step? Sep 10 15:47:13 i'm thinking a new task which depends on A and B, and both A and B inherit a test class which make their fetches fail Sep 10 15:47:14 Or just logging it? Sep 10 15:47:20 -k specifically. Sep 10 15:47:26 -k makes it continue rather than halting at the first error Sep 10 15:47:30 yeah Sep 10 15:47:35 -k is evil. :D Sep 10 15:47:45 this means i need bitbake running at least two different operations, and one has to fail and i have to see that it attempts the second Sep 10 15:48:01 ohh, to verify that -k works? Sep 10 15:48:16 imo -k should be default in master eventually. unless you're debugging, its likely that you want bitbake to go as far as it can before the build requires your attention.. Sep 10 15:48:19 yeah Sep 10 15:48:22 master has a bug with -k Sep 10 15:48:26 sometimes it stops at a failure Sep 10 15:48:27 Could have two recipes fight over a file lock? Sep 10 15:48:29 but i don't know when or why Sep 10 15:48:42 it *seems* related to an exception raised in a task Sep 10 15:48:50 particularly fetch md5 failure seems to cause it to just halt Sep 10 15:49:06 but i can't fix it until i can reliably and predictably reproduce it Sep 10 15:49:21 Hmm. I'd have to disagree. I don't think -k is a reasonable production setup Sep 10 15:49:32 why? Sep 10 15:50:33 It introduces variability in duration, and quality. If something dies @ 5 minutes in but I don't find out for an hour I'm twiddling my thumbs when I could have been fixing the issue Sep 10 15:50:55 And long term you wouldn't have any builds fail Sep 10 15:51:01 what do you mean? Sep 10 15:51:03 But, I thin you're prolly looking agt this from a completely different angle than me Sep 10 15:51:05 -k doesn't make it stop failing Sep 10 15:51:14 it just collects the errors seen and gives them to you at the end Sep 10 15:51:37 hmm Sep 10 15:51:45 I run all my builds through colorize Sep 10 15:51:46 heh Sep 10 15:51:59 I must be doing it wrong Sep 10 15:53:30 lets say you have a task that wants Foo and Bar, and those two require the crosscompiler. it builds the crosscompiler successfully, but Foo fails to compile. with -k, it will go ahead and try to finish up Bar before halting, getting you closer to your end goal before halting the process and telling you that Foo failed with this problem Sep 10 15:55:07 if you're sitting there waiting on the build, or are specifically testing Foo, it makes sense to fail right away so you can fix it and move on Sep 10 15:55:23 but if you're not babysitting it, i doubt you want it to stop and let your box idle when it could be getting closer to a successful build Sep 10 15:57:02 but it depends on what you're trying to accomplish, its hard to pick a default for something like that Sep 10 15:57:08 the current behavior is based on gmake, which has a similar option Sep 10 16:12:39 yah, I hear ya Sep 10 16:19:05 anyone happen to know where I'd find a package providing /usr/sbin/iw ? Sep 10 16:19:56 i guess I could grep -w Sep 10 16:20:32 jconnolly: recipes/iw/iw_*.bb? Sep 10 16:20:35 just a guess :-} Sep 10 16:20:54 ah Sep 10 16:20:56 I'm using stable Sep 10 16:21:08 ah, then you may be out of luck Sep 10 16:21:08 i'll try to merge it in from dev locally, see what happens. thanks pb_ Sep 10 16:21:15 I guess it wouldn't be hard to backport from .dev Sep 10 16:21:22 yeah. that's what I'll try Sep 10 16:21:27 righto, cool Sep 10 16:27:36 hi Sep 10 16:44:49 03Frans Meulenbroeks  07org.openembedded.dev * rb985e56c97 10openembedded.git/recipes/perl/libxml-libxml-perl_1.70.bb: Sep 10 16:44:49 libxml-libxml-perl_1.70.bb: fix native build Sep 10 16:44:49 Signed-off-by: Frans Meulenbroeks Sep 10 16:53:10 03Frans Meulenbroeks  07org.openembedded.dev * rd8ee8e6623 10openembedded.git/recipes/perl/libsoap-lite-perl_0.710.08.bb: Sep 10 16:53:10 libsoap-lite-perl: updated to 0.712 Sep 10 16:53:10 Signed-off-by: Frans Meulenbroeks Sep 10 16:59:46 03Frans Meulenbroeks  07org.openembedded.dev * rb3c0b957fa 10openembedded.git/recipes/perl/ (2 files): Sep 10 16:59:46 libxml-namespacesupport-perl: updated to 1.11 Sep 10 16:59:46 Signed-off-by: Frans Meulenbroeks Sep 10 17:22:16 hello all Sep 10 17:22:36 I would stick to #angstrom, but that room is full of crickets... Sep 10 17:23:12 http://pastebin.ca/1937399 <-- Failing to build module-init-tools-cross Sep 10 17:23:26 error seems to imply that glibc failed to build Sep 10 17:23:43 brokencodes, I answered in #angstrom :) Sep 10 17:23:54 thanks Sep 10 17:24:11 install glibc-static on the build machine Sep 10 17:24:27 awesome Sep 10 17:28:52 this -k bug is so weird. Sep 10 17:32:17 ah ha Sep 10 17:35:41 kergoth_: my bitbake -k died after 13000 tasks Sep 10 17:35:46 kergoth_: is it something same Sep 10 17:35:52 you are seeing Sep 10 17:35:58 both master and 1.10 have a known -k bug Sep 10 17:36:04 just nailed it down, testing a fix now Sep 10 17:36:13 oh whats the behaviour Sep 10 17:36:15 er, i know master does, not sure about 1.10 Sep 10 17:36:23 may be its the one I ran into Sep 10 17:36:34 if a task exists with bb.fatal() or sys.exit(), the build stops there instead of continuing the way it should Sep 10 17:39:02 03Frans Meulenbroeks  07org.openembedded.dev * re9c03d628e 10openembedded.git/recipes/perl/libtask-weaken-perl_1.03.bb: Sep 10 17:39:02 libtask-weaken-perl_1.03.bb: new recipe Sep 10 17:39:02 Signed-off-by: Frans Meulenbroeks Sep 10 17:39:03 03Frans Meulenbroeks  07org.openembedded.dev * rae19546edc 10openembedded.git/recipes/perl/libclass-inspector-perl_1.24.bb: Sep 10 17:39:03 libclass-inspector-perl_1.24.bb: new recipe Sep 10 17:39:03 Signed-off-by: Frans Meulenbroeks Sep 10 17:40:57 kergoth_: I think it could be the one I ran into Sep 10 17:41:00 not sure Sep 10 17:41:02 though Sep 10 17:42:46 NOTE: Sending SIGTERM to remaining 2 tasks Sep 10 17:42:49 thats what I see Sep 10 17:42:51 in the end Sep 10 17:44:04 here's a question Sep 10 17:44:09 how many failed tasks does it show in the count? Sep 10 17:44:14 in the status at the end Sep 10 17:54:35 khem, found the cause of the 'with 256' seen with the failed tasks all the damn time Sep 10 17:54:45 khem, incorrect usage of the exitcode returned by waitpid Sep 10 17:54:49 its shifted up by 8 bits Sep 10 17:54:55 so 256 is an exit code of 1 Sep 10 17:57:12 hi khem, you got mail :-) Sep 10 18:30:53 hmm, i think i see how to fix the log message ordering issue with the 'none' server Sep 10 18:39:22 khem, ping Sep 10 18:53:35 Crofton_|work: hello Sep 10 18:53:57 hey Sep 10 18:54:25 I need to try a small test case building something with the atmiic intrinsics on the omap3 Sep 10 18:54:48 basically, a build fails on the omap3 using gcc there Sep 10 18:55:00 hrm Sep 10 18:55:01 but succeeds in OE, with gcc-cross Sep 10 18:55:41 cross is much faster, but native on the omap3 is helpful for some work Sep 10 18:55:42 something i did results in not saving/using the log iles for python task failures.. it works in 1.0.. course, i actually rather like it that way, but i think it should be more explicit.. Sep 10 18:55:44 * kergoth_ bisects Sep 10 19:06:38 03Koen Kooi  07org.openembedded.dev * rdf10473931 10openembedded.git/recipes/angstrom/exquisite-theme-angstrom.bb: exquisite-theme-angstrom: another workaround for do_unpack breakage Sep 10 19:07:46 uh? what the heck is koen doing Sep 10 19:07:57 "file://angstrom/" works fine Sep 10 19:08:55 if it didn't, it should be reported as a bug and fixed, not "another workaround" Sep 10 19:10:27 likewise: thanks, today my tooth hurts Sep 10 19:10:40 likewise: but may be on sunday I will do something Sep 10 19:10:48 my brain has stopped working Sep 10 19:11:05 all WBC are deployed in the jaw Sep 10 19:11:13 and its calling for more help :) Sep 10 19:11:29 Crofton_|work: ok Sep 10 19:11:47 Crofton_|work: so you want to know if there is any small testcase Sep 10 19:11:56 I need to make one Sep 10 19:12:01 ok Sep 10 19:12:08 03Florian Boor  07org.openembedded.dev * r968f736473 10openembedded.git/conf/machine/mini6410.conf: mini6410.conf: Initial support for S3C6410 based Mini6410 board. Sep 10 19:12:09 03Florian Boor  07org.openembedded.dev * r41019c38e9 10openembedded.git/recipes/linux/ (4 files in 2 dirs): linux: Add 2.6.28 patches and config for Mini6410. Sep 10 19:12:12 instrisics should work Sep 10 19:12:16 with gcc 4.5 and 4.4 Sep 10 19:12:19 yeah Sep 10 19:12:35 and woglinde posted a patch for 4.3 as well Sep 10 19:12:44 re Sep 10 19:12:45 4.3 worked Sep 10 19:12:51 another ugly kernel patch for oe Sep 10 19:12:52 yeah it should I believe Sep 10 19:13:12 florian: ugliness is neverending Sep 10 19:13:16 I'm not sure what happened when we switched to gcc-4.5.x Sep 10 19:13:39 Crofton_|work: 4.5 should have them too Sep 10 19:14:06 if you tell me which intrinsic you are interested in I can find ore Sep 10 19:14:08 more Sep 10 19:14:20 I know, but I have a funny build error, just can't reproduce it eaily Sep 10 19:14:28 khem: yeah... and vendor kernel patches are attracting it ;) Sep 10 19:14:57 khem: good luck with the tooth(fairy) Sep 10 19:15:11 I'll try and get back to it in a bit Sep 10 19:15:20 need to get some NEON intrinsics going atm Sep 10 19:15:35 likewise: yeah my wisdom tooth it is there is no place to grow Sep 10 19:15:47 I guess before its born I must kill it Sep 10 19:20:18 damnit, there's another little bug in bitbake to fix.. Sep 10 19:20:22 * kergoth_ mutters Sep 10 19:20:41 the untracked delvar exception handler appears to be catching KeyboardInterrupt Sep 10 19:26:17 kergoth_: hey, sorry for the joke... :) no offence meant Sep 10 19:48:20 heh we are passing -j4 to glibc build thats not gonna parallelize it Sep 10 19:48:28 now I see why it takes a long time Sep 10 19:48:42 usually eglibc is the last one to complete Sep 10 19:48:53 passing -j4? Sep 10 19:48:55 03Chris Larson  07org.openembedded.dev * r88cb518488 10openembedded.git/classes/base.bbclass: Sep 10 19:48:55 Fix unpack issue with bitbake 1.10 Sep 10 19:48:55 Signed-off-by: Chris Larson Sep 10 19:48:57 03Chris Larson  07org.openembedded.dev * r15c1b7982f 10openembedded.git/recipes/ (3 files in 3 dirs): Sep 10 19:48:57 Revert now unnecessary do_unpack bug workarounds Sep 10 19:48:57 Signed-off-by: Chris Larson Sep 10 19:49:58 khem, okay, -k *should* behave again Sep 10 19:50:05 kergoth_: nice Sep 10 19:50:06 can't repro a failure anymore at least Sep 10 19:50:12 I will know Sep 10 19:50:17 in next run Sep 10 19:50:20 cool Sep 10 19:50:30 if it goes beyond 13000 task :) Sep 10 19:50:46 i just realized that the conversion to logging resulted in no log files for python tasks Sep 10 19:50:57 the worker processes don't print the messages anymore, they go to the server/ui Sep 10 19:51:06 so the worker stdout is empty, if you use bb.note/etc Sep 10 19:51:17 i guess thats okay, so the log file only has stdout *other than* bb.msg stuff Sep 10 19:51:21 which means no duplicated output Sep 10 19:51:48 can always add the bb.msg bits to the log file via adding a new logging streamhandler for workers Sep 10 19:52:49 kergoth_: how can I use the PARALLELISM info myself Sep 10 19:52:56 what do you mean? Sep 10 19:53:04 let me explain Sep 10 19:53:10 some apps dont use -j Sep 10 19:53:20 so I want to know the number Sep 10 19:53:27 and contruct that for the app Sep 10 19:53:40 like say the app used PARALLELISM=n Sep 10 19:53:45 then I want to pick n Sep 10 19:53:51 and pass that manually Sep 10 19:53:58 to oe_runmake Sep 10 19:54:03 PARALLELISM := "${@'${PARALLEL_MAKE}'.replace('-j', '')}" Sep 10 19:54:07 PARALLEL_MAKE = "" Sep 10 19:54:08 in that order Sep 10 19:54:42 PARALLELMFLAGS=-j Sep 10 19:54:46 is what I need to pass Sep 10 19:54:54 hi there Sep 10 19:54:56 okay, just use PARALLEL_MAKE directly then Sep 10 19:54:59 it contains -jwahtever Sep 10 19:55:12 just use := to expand it before you override it to empty to disable it, if you do so Sep 10 19:55:19 er, that wasn't very clear Sep 10 19:55:32 if hte recipe empties PARALLEL_MAKE, make sure you use := when referencing the old value before it does so Sep 10 19:55:34 that's better Sep 10 19:55:37 heheh Sep 10 19:55:56 it seems that there's a problem compiling guile-native, here's log: http://pastebin.com/VLWkwgeW Sep 10 19:56:13 is there autotools guru to help me? Sep 10 19:57:10 kergoth_: I need to empty it Sep 10 19:57:25 but before I empty it I need to transfer its content Sep 10 19:57:28 right, so PARALLELMFLAGS := "${PARALLEL_MAKE}" Sep 10 19:57:31 then PARALLEL_MAKE = "" Sep 10 19:57:47 := makes it expand it immediately rather than lazily Sep 10 19:58:03 OE_EXTRA_RUNMAKE is what I can sue to pass extra stuff to make right Sep 10 19:58:07 (I've experienced gettext-native compile problemsand fixed it with this patch: http://thread.gmane.org/gmane.comp.handhelds.openembedded/36800 , but it doesn't look like it causes guile-native build problem :\) Sep 10 19:58:19 EXTRA_OEMAKE :) Sep 10 19:58:29 huh Sep 10 19:58:41 * kergoth_ just realizes that bitbake is using NOTE: for task failure .. Sep 10 19:58:44 EXTRA_OEMAKE Sep 10 19:58:46 NOTE: package bison-native-2.3-r8.0: task do_fetch: Failed Sep 10 19:58:52 that's an ERROR, no? :P Sep 10 19:58:57 kergoth_: yes good eye Sep 10 20:05:39 hi ericben|away Sep 10 20:05:42 ah he's away Sep 10 20:06:15 hmm, thats odd, i thought KeyboardInterrupt didn't inherit from Exception.. Sep 10 20:06:19 hmm Sep 10 20:06:30 it doesn't.. Sep 10 20:06:33 * kergoth_ scratches head Sep 10 20:07:27 i know what i should do.. i'm sick of the task events deciding on the formatting of the message displayed Sep 10 20:45:55 kergoth_: http://pastebin.com/zeijSg2P Sep 10 20:46:12 while doing bitbake -c clean libgles-omap3 Sep 10 20:46:24 using rm_wrok and packaged stageing Sep 10 20:46:46 libgles-omap3 did not build btw. so I was trying to rebuild it and clean it before that Sep 10 20:47:28 bug in packaged-staging.bbclass Sep 10 20:48:30 03Chris Larson  07org.openembedded.dev * r8607ea2929 10openembedded.git/classes/packaged-staging.bbclass: Sep 10 20:48:30 packaged-staging: add missing debug level Sep 10 20:48:30 Signed-off-by: Chris Larson Sep 10 20:54:31 kergoth_: ok works now Sep 10 20:54:36 cool Sep 10 20:56:50 should add an argument to bitbake to specify a logging config file Sep 10 21:04:02 hmm, kind of nice changing the formatting to remove NOTE:. it's really rather unnecessary Sep 10 21:04:10 bitbake libgles-omap3 it fetched it and exited Sep 10 21:04:11 odd Sep 10 21:04:19 hmm, weird Sep 10 21:04:52 second time worked Sep 10 21:05:28 very weird. Sep 10 21:15:01 damint, change to FuncFailed arguments is breaking things -- i forgot oe uses it directly.. it really shouldnt' Sep 10 22:04:53 http://www.angstrom-distribution.org/unstable/sources/OMAP35x_Graphics_SDK_setuplinux_3_01_00_06.bin <---- file not found, making base-image Sep 10 22:05:02 Sorry wrong room Sep 10 22:29:58 found it here... http://software-dl.ti.com/dsps/dsps_public_sw/sdo_sb/targetcontent//gfxsdk/3_01_00_06/3.01.00.06/OMAP35x_Graphics_SDK_setuplinux_3_01_00_06.bin Sep 11 00:35:30 hey Sep 11 00:35:48 is it possible to set ${prefix} from /usr to / for a single package? Sep 11 00:38:10 you just want to move things in do_install, most likely Sep 11 00:38:23 ie mv ${D}${bindir}/foo ${D}${base_bindir}/foo Sep 11 00:38:32 if you really want to do that, its a oneliner. prefix = "" Sep 11 00:38:59 or /, i can never remember which of the pieces has the slash.. Sep 11 00:39:00 heh Sep 11 00:40:43 And really, most packages don't want to do this :) Sep 11 00:40:45 no, i need to call configure (actually the idiotic cmake can't install outside the ${prefix} Sep 11 00:40:47 There are cases where it's needed Sep 11 00:41:00 so i can't get anything outside /usr Sep 11 00:41:24 * kergoth rolls eyes at cmake Sep 11 00:41:31 i've changed my package to match, but it's an ugly hack Sep 11 00:41:57 anyway, it's where i am now so... Sep 11 00:42:16 well my surprise was the attitude on #cmake: why do you wanna install stuff in /etc ? Sep 11 00:42:24 i mean geez - why not? Sep 11 00:43:03 anyway, the lesson i got is that no program should install configuration files or such; hope this workaround works :( Sep 11 00:44:06 so i can actually install stuff in /etc Sep 11 00:44:45 i so miss autotools right now... **** ENDING LOGGING AT Sat Sep 11 02:59:57 2010