**** BEGIN LOGGING AT Tue Nov 16 02:59:58 2010 Nov 16 03:03:38 does anyone know why I should have the dhcp bb file create /var/lib/dhcp for it to put runtime files in? Nov 16 03:11:57 khem: yes, -O2 rootfs works on ppc Nov 16 03:15:00 what are the thought about making /var a tmpfs? Nov 16 03:19:21 It can be useful, yeah :) Nov 16 03:22:03 dhcp expects to see /var/lib/dhcp Nov 16 03:22:15 I can create it and put it in the package Nov 16 03:22:34 which is ok, since Angstrom looks like /var is in a root Nov 16 03:22:49 /var/volatile is on a tmpfs Nov 16 03:23:35 if I put it in a script then it could be created even if /var was tmpfs Nov 16 03:23:59 but, it looks like it keeps lease info there, which it might want on the next boot Nov 16 03:25:40 Moving all of /var to tmpfs requires additional help, yes Nov 16 03:25:56 ok Nov 16 03:26:00 Some sort of support package for it, perhaps Nov 16 03:26:07 yeah Nov 16 03:26:18 I think I'll make the path at install Nov 16 03:26:29 Personally, I'd love to see that since it just might make squashfs based rootfs work :) Nov 16 03:26:31 and package with the client and go from there Nov 16 03:27:01 but I think there is lots of work fixing things that assumed existence of directories at boot time Nov 16 03:27:29 and right now, I want an annoying message to go away :) Nov 16 03:27:54 WWPD Nov 16 03:30:05 poky looks to have the smae problem Nov 16 03:42:29 kgilmer, gm Nov 16 03:46:59 hi crofton :) Nov 16 03:47:59 I just heard the Chilean ambassador to Japan talk about how important IPv6 is to the future. strange. Nov 16 03:50:57 heh Nov 16 05:38:40 Jay7: thats right ltrace is not ported to sh it should be disabled for sh3/sh4 Nov 16 06:51:47 gm Nov 16 06:55:17 mawnin Nov 16 08:02:16 what do you guyz think of pokylinux ? Nov 16 08:08:33 03Martin Jansa  07master * rae1ed0e4ed 10openembedded.git/recipes/openmoko-3rdparty/mokosuite2_git.bb: mokosuite2: bump SRCREV Nov 16 08:52:19 03Martin Jansa  07master * rfaa743ab65 10openembedded.git/recipes/openmoko-3rdparty/mokowm_git.bb: mokowm: bump SRCREV Nov 16 09:15:13 dj-death: I developed both OE and Poky Nov 16 12:03:15 http://marcin.juszkiewicz.com.pl/2010/11/16/pandaboard-my-story/ Nov 16 12:04:12 Hi Nov 16 12:04:36 i want to submit a new board patch to the oe Nov 16 12:05:02 nani__: git format-patch + git send-email to OEML Nov 16 12:05:15 ok Nov 16 12:06:28 but i have a linux-omap-psp kernel patch can i include this patch also while patching modified reciepes Nov 16 12:06:34 yes Nov 16 12:06:41 for my new board Nov 16 12:07:04 what is mean by machine id to a new board Nov 16 12:07:57 Is there any process of registering the new board in OE Nov 16 12:09:23 no Nov 16 12:09:33 new board is new machine definition Nov 16 12:09:42 adding such is done by sending patches Nov 16 12:09:57 ya can i directly send the machine.conf file patch Nov 16 12:10:40 ok . Thank u Nov 16 12:12:45 for including my new board i have modified some files can i take the patch of all files at a time or i have to take individual patch for every file i have modified Nov 16 12:17:50 nani__: combine it to some logical groups Nov 16 12:17:52 e.g. per package Nov 16 12:59:05 can i submit psp kernel patches and u-boot patches along with this recipe patches Nov 16 12:59:25 to oe mailing list Nov 16 13:05:57 nani__: split'em Nov 16 13:06:08 separate patches for packages (recipes) Nov 16 13:06:41 then just send to ML Nov 16 13:06:52 people will show your mistakes ;) Nov 16 13:43:46 03Andreas Oberritter  07org.openembedded.dev * r330b0377a0 10openembedded.git/classes/lib_package.bbclass: Nov 16 13:43:46 lib_package.bbclass: fix packaging of static libs, inherit binconfig Nov 16 13:43:46 * Package ${libdir}/*.a and /lib/*.a into ${PN}-static instead of ${PN}-dev Nov 16 13:43:46 * Although it packaged ${bindir}/*-config into FILES_${PN}-dev, it did Nov 16 13:43:46 not include the required mangling applied by binconfig.bbclass. Nov 16 13:43:46 Signed-off-by: Andreas Oberritter Nov 16 13:43:47 Acked-by: Tom Rini Nov 16 13:57:34 anyone ever used/build uclibc++ ? build fails for me Nov 16 14:33:44 hrw: what's your preference Nov 16 14:34:08 dj-death: now I do not use any of them. Nov 16 14:34:23 dj-death: which one to use is a hard question. Nov 16 14:34:53 dj-death: it depends is your hardware supported by any of those, if yes then how good it is supported Nov 16 14:35:04 for omap3 I would choose OE Nov 16 14:35:18 for example Nov 16 14:35:27 ok Nov 16 14:36:26 since the board we want to build for has no support in both oe and poky, I can't really choose with this criteria Nov 16 14:37:08 does oe has more developers than poky ? Nov 16 14:38:16 yes Nov 16 14:39:25 so it seems a better choice Nov 16 14:40:25 I don't really see why the 2 projects haven't merge at some point Nov 16 14:41:33 when looking at the trees it seems that the biggest difference is the way recipes are stored ;) Nov 16 14:42:12 that's one part of the difference, yes. Nov 16 14:42:25 see the mailing list threads, richard purdie has discussed why poky exists before Nov 16 15:15:14 mk, just about done with the 2to3 on master, but of course it can't be tested due to all of the python2 in the metadata Nov 16 15:15:17 * kergoth_ rolls eyes Nov 16 15:16:56 2to3? Nov 16 15:17:40 converts python 2 to 3 Nov 16 15:17:50 can't do everything, but covers most of it Nov 16 15:17:54 its a start, anyway Nov 16 15:18:05 figured we should get a branch for it going Nov 16 15:18:15 ah ok Nov 16 15:26:15 03Koen Kooi  07org.openembedded.dev * r503f015b08 10openembedded.git/recipes/x-load/ (2 files in 2 dirs): x-load git: add patch for beagleboard to fix u-boot in fat32 partitions Nov 16 15:35:26 he...BB people still battling with x-load / u-boot :p Nov 16 15:35:50 guh, i hate the way bb.cache works Nov 16 15:35:51 * kergoth_ thinks Nov 16 16:33:25 hi! i read through the documentation for openembedded and bitbake, but still, it never talks about what sets you can install. how do I install lxde or xfce? it only tells me how to install x11 in http://wiki.openembedded.org/index.php/Getting_started#Start_building Nov 16 16:34:13 also, it doesnt say what is included in x11-image. do i get a browser? do i get a text editor? or do I only get the base X11 environment like twm or something? Nov 16 16:34:59 i think more documentation should be written about "installing packages via bitbake" which includes a nice big list Nov 16 16:35:59 at least some more examples should be shown. like installing gnome or something, and at least a way to find out for yourself what packages you can install Nov 16 16:37:28 I am a developer (for other stuff) and would love to start getting my feet wet with openembedded, but first i need to get a simple understanding of installing packages. i know how to write my own bb files and such, thats all simple stuff, i write gentoo ebuilds and PKGBUILDS for arch linux all the time Nov 16 16:37:34 thank you in advance! Nov 16 16:39:13 i know in here it lists all the packages individually http://cgit.openembedded.org/cgit.cgi/openembedded/tree/recipes but no groups :( Nov 16 16:39:23 x11-image for example Nov 16 16:39:33 x11-image is in recipes/images Nov 16 16:39:46 usually coupble of task-something which are recipes/tasks Nov 16 16:39:50 couple Nov 16 16:40:19 oh sweet thanks! can you install multiple "images" at once? Nov 16 16:40:44 no but think about tasks as groups Nov 16 16:40:53 so you instal multiple tasks in one image Nov 16 16:41:05 so you cant do "bitbake xfce46-image x11-image" Nov 16 16:41:18 yes you can build multiple images Nov 16 16:41:44 and also install packages from multiple images on target device (from generated feed) Nov 16 16:41:59 but you flash your target with 1 images (the closest to your needs) Nov 16 16:42:47 so if you run "bitbake xfce46-image x11-image" for example, it will generate two files not just a single one? like they wont be merged Nov 16 16:44:11 NTU: did you wish to have xfce inside x11-image? Nov 16 16:44:44 i want an xfce environment with X (which includes xf86-input-evdev) and all the required drivers Nov 16 16:44:47 or you wish to know a way to install xfce on some target device? Nov 16 16:45:00 xfce46-image will have Xorg too :) Nov 16 16:45:14 oh! good they dont split it up Nov 16 16:45:40 because xfce by itself is generally useless if it doesnt include xsetroot and all that stuff Nov 16 16:45:45 yes, just use xfce*-image Nov 16 16:46:05 you can look into recipes/images/xfce* Nov 16 16:46:10 NTU: it will be clear, if you read that recipe Nov 16 16:47:07 yup! Nov 16 16:47:15 i should have done that first :P Nov 16 16:50:08 if you choose DISTRO = "angstrom-201x" something along those lines in build/conf/local.conf should you install the angstrom*-image instead? Nov 16 16:50:58 im on a really slow netbook (1.6 GHz i believe, intel atom, single core) and rather not have to build a whole new image all over again which is why i rather not trial and error Nov 16 16:51:34 my 4 core Phenom II X4 955 is currently... broken. :S Nov 16 17:06:37 brb Nov 16 17:19:17 gm Nov 16 17:21:06 * kergoth_ ponders Nov 16 18:08:06 has anyone got a working kernel config for the ipaq h5450?? Nov 16 18:22:33 Any recommendations for removing package contents from sysroots? Nov 16 18:24:18 more context? Nov 16 18:24:42 some, such as meta-toolchain have hooks to them remove things Nov 16 18:24:57 others, such as the build sysroot, we don't have a good reason (yet?) to Nov 16 18:27:08 Trying to just get a clean environ without removing tmp and starting from scratch. Nov 16 18:27:36 So you want to make use of packaged staging Nov 16 18:27:43 lemme see, the default dir is.. Nov 16 18:27:45 pstage is automatic Nov 16 18:27:48 just bitbake -c clean it Nov 16 18:27:52 clean uatomatically uninstalls from sysroot Nov 16 18:28:11 if you want to really start from scratch entirely, you should be removing tmpdir Nov 16 18:28:14 ah, that's another way to parse that sentance, possibly more correctly :) Nov 16 18:30:23 ok thanks. I've found that bitbake -c clean sometimes(ever?) actually removes content from sysroots. Nov 16 18:30:36 then you've found something that isn't the case Nov 16 18:30:44 at least not using current oe, which uses packaged staging Nov 16 18:32:54 ack. must be driver error Nov 16 18:33:15 or you were just using an old oe tree, perhaps Nov 16 18:33:29 note that -c clean operates only against the recipe you tell it to, not its dependencies Nov 16 18:34:55 yeah that latter bit has always bothered me. I guess something can be done with white list but this is set by my distro (Arago) and didn't want to mess with it. Nov 16 18:35:43 -c clean is always an explicit operation, and is almost always only done for debugging Nov 16 18:40:46 oh.. master freeze announced Nov 16 18:41:20 is it actually physically frozen on the server? if not, its not going to work Nov 16 18:41:25 with a hook, that is Nov 16 18:41:39 I think it is just announced Nov 16 18:41:48 dman Nov 16 18:41:56 I have a bugfix :) Nov 16 18:42:18 for this one, I think social engineering is OK Nov 16 18:42:22 i don't think a freeze is going to work, unless either all commits are blocked, or only commits from a person designated as the one to pull fixes Nov 16 18:42:30 s/only commits/only commits allowed/ Nov 16 18:42:37 * kergoth_ shrugs Nov 16 18:42:37 it is more like feature freeze Nov 16 18:42:54 bugfixes should be commited Nov 16 18:43:10 khem: am I right? Nov 16 18:43:33 Jay7: yes feature freeze for sure Nov 16 18:43:44 Jay7: even bug fixes should be on ml Nov 16 18:43:56 just to make sure Nov 16 18:44:09 * Jay7 is thinking about not-so-good version of kexecboot in tree :\ Nov 16 18:44:45 khem: btw, how to disable ltrace in x11-image for sh4? Nov 16 18:45:20 or may be should I build other image? Nov 16 18:45:42 03Koen Kooi  07org.openembedded.dev * rab49fda5f7 10openembedded.git/conf/distro/angstrom-2010.x.conf: angstrom next: switch to eglibc 2.12 to align with angstrom-yocto Nov 16 18:45:45 Jay7: it can be done with COMPATIBLE_TARGET_SYS = (?!sh[34]) Nov 16 18:45:48 I need web-interface for my builder :) Nov 16 18:46:07 khem: may be good idea to commit this fix? Nov 16 18:46:27 Jay7: but I am not sure if image will complain because what this will remove it from parse tree Nov 16 18:46:33 havent tried it Nov 16 18:46:38 can you test it out Nov 16 18:46:40 well, let's try Nov 16 18:49:21 hm Nov 16 18:49:34 it have LTRACE_ARCH=sh in do_compile() Nov 16 18:49:52 khem, still trying to learn the GPL licensing - you say you read iftop's license and it should be GPLv2+ - why the +? The COPYING file sys GNU GENERAL PUBLIC LICENSE Version 2, June 1991 Nov 16 18:51:08 tharvey: it also says "at your option any later version" Nov 16 18:51:14 thats why + Nov 16 18:51:27 in theory you can say its gplv3 Nov 16 18:51:34 ah... I didn't understand that statement you made earlier Nov 16 18:51:36 ERROR: Nothing RPROVIDES 'ltrace' (but /storage/oe/openembedded/recipes/tasks/task-cli-tools.bb RDEPENDS on or otherwise requires it) Nov 16 18:51:47 Jay7: yes thats what I thought Nov 16 18:51:54 ok so if that verbiage is there you 'must' say GPLv2+ - ok thx Nov 16 18:52:01 resubmitting Nov 16 18:52:03 Jay7: ok let me give you another fix on top of this one Nov 16 18:52:36 if I submit a v2 or v3 patch am I supposed to state what I changed from previous versions and how? Nov 16 19:01:11 Jay7: http://pastebin.com/JvHVAbeP Nov 16 19:01:20 Jay7: try this one on top of that fix Nov 16 19:01:42 tharvey: usually yes Nov 16 19:03:59 khem: build continued Nov 16 19:05:55 http://pastebin.com/Pvvd5RFA Nov 16 19:05:57 full diff Nov 16 19:08:09 khem: Want me to post or just push doing sane-toolchain -O2 for ppc? Nov 16 19:10:48 Tartarus: I think we should post it Nov 16 19:11:07 the change should be done in sane-toolchain-eglibc I think Nov 16 19:11:21 Jay7: ok Nov 16 19:11:57 yeah Nov 16 19:12:02 testing uclibc now quick too Nov 16 19:12:17 Tartarus: btw did you try with newer kernels too for ppc ? Nov 16 19:12:33 hi Nov 16 19:12:41 khem, not past .35, no Nov 16 19:12:47 ok Nov 16 19:12:51 thats good. Nov 16 19:13:06 I think in next two weeks we should settle on various versions Nov 16 19:13:14 for qemu and other recipes Nov 16 19:13:19 and make the defaults works Nov 16 19:21:10 yeah Nov 16 19:24:48 khem: build is failed again in other place Nov 16 19:26:01 | * resolve_conffiles: Existing conffile /var/tmp/oe/rootfs/x11-image/etc/device_table is different from the conffile in the new package. The new conffile will be placed at /var/tmp/oe/rootfs/x11-image/etc/device_table-opkg. Nov 16 19:26:02 | * opkg_install_cmd: Cannot install package locale-base-fr-fr. Nov 16 19:26:04 NOTE: package x11-image-1.0-r0: task do_rootfs: Failed Nov 16 19:26:15 sh@t.. tinderbox doesn't catch this :( Nov 16 19:27:03 about locale - it can be my local.conf Nov 16 19:27:28 I have this there: GLIBC_GENERATE_LOCALES = "en_US.UTF-8 en_GB.UTF-8 de_DE.UTF-8 ru_RU.UTF-8" Nov 16 19:28:06 but image should respect this as I understand Nov 16 19:31:54 Jay7: thats could be something else Nov 16 19:46:17 is there a mirror of the handhelds.org web site available? Nov 16 19:46:41 where is get-send-email ... Nov 16 19:46:49 trying install on Fedora Nov 16 19:47:18 Crofton: something about perl+git Nov 16 19:47:34 iirc, it was written on perl Nov 16 19:47:40 yeah Nov 16 19:48:18 yes, I remember right Nov 16 19:48:58 Crofton: yum install git-email Nov 16 19:49:45 and also install perl-Authen-SASL Nov 16 19:49:52 to use gmail Nov 16 19:53:57 heh Nov 16 19:54:09 I was searaching for git-send-email :) Nov 16 19:54:15 found it Nov 16 20:18:44 hm, what I am doing wrong here? I have in my local.conf: Nov 16 20:18:44 PREFERRED_PROVIDER_libmysqlclient_local = "mysql5" Nov 16 20:18:57 but still I get: NOTE: multiple providers are available for runtime libmysqlclient (mysql5, mysql) Nov 16 20:19:08 (distro = minimal) Nov 16 20:21:40 anyone a hint ? Nov 16 20:25:42 effem try -vvv Nov 16 20:25:52 to see whats goin on Nov 16 20:25:57 and good nite Nov 16 20:26:03 nite Nov 16 20:27:28 ok, I am practiving with git gend-mail :) Nov 16 20:32:20 this is what -vvv -D -D -D gives (trimmed): http://www.pastebin.ca/1993778 Nov 16 20:33:18 03Martin Jansa  07org.openembedded.dev * r124f6bb67d 10openembedded.git/recipes/uhd/uhd.inc: Nov 16 20:33:18 udh.inc: use RDEPENDS_${PN} Nov 16 20:33:18 Signed-off-by: Philip Balister Nov 16 20:33:41 wondering whether my pinning is wrong as the log says things like: NOTE: checking PREFERRED_PROVIDER_mysql5 Nov 16 20:33:54 instead of PREFERRED_PROVIDER_libmysqlclient Nov 16 20:35:59 * Crofton wondrs why that did not pickup th acks from pw Nov 16 20:36:44 strange, you used the script ? Nov 16 20:36:51 yeah Nov 16 20:37:03 and I could see the acks Nov 16 20:37:11 no worries Nov 16 20:37:19 hm, for me it always worked Nov 16 21:25:16 I get this error *** Error: Package name contains illegal characters, (other than [a-z0-9.+-]) Nov 16 21:25:23 kernel-2.6.31.12-ER1 Nov 16 21:25:28 is the package name Nov 16 21:25:35 is it illegal to have big letters Nov 16 21:25:36 its the uppercase letters in the revision Nov 16 21:25:39 yep Nov 16 21:25:46 hmm Nov 16 21:25:55 why is that Nov 16 21:26:22 pretty sure its the package manager complaining, not us, no? Nov 16 21:26:45 khem, ok, uclibc needs -O2 too Nov 16 21:27:01 yeah, it is Nov 16 21:27:17 (package manager) Nov 16 21:27:33 I wish bitbake would do the griping in that case, from when I hit that a few years back Nov 16 21:28:21 Tartarus: yes right Nov 16 21:29:00 do_package_write_ipk failed Nov 16 21:29:05 so I think yes its opkg Nov 16 21:29:46 khem: so the bug, is it with libgcc.so or libgcc_s.so ? Nov 16 21:29:50 writing up a comment Nov 16 21:33:26 is Michael Smith here? Nov 16 21:35:44 Tartarus: its with libgcc.a Nov 16 21:35:51 ok, thanks Nov 16 21:36:08 # gcc 4.5 has a problem with 'leakage' into libgcc.a with -Os Nov 16 21:36:12 good enough? Nov 16 21:36:26 Tartarus: yes. Nov 16 21:36:27 fine Nov 16 21:37:01 k, I'll post this patch in a little bit, doing a build to confirm now Nov 16 21:40:12 03Andreas Oberritter  07org.openembedded.dev * ref0de0741b 10openembedded.git/recipes/libfribidi/ (files/libtool-update.patch libfribidi_0.10.4.bb): (log message trimmed) Nov 16 21:40:12 libfribidi-0.10.4: update recipe, fix packaging Nov 16 21:40:12 * added LICENSE Nov 16 21:40:12 * removed configure patch Nov 16 21:40:12 * use lib_package and binconfig to package all installed files Nov 16 21:40:12 * convert to new style staging Nov 16 21:40:12 * oe-stylize Nov 16 21:40:19 03Andreas Oberritter  07org.openembedded.dev * rcc397a6c47 10openembedded.git/recipes/libungif/libungif_4.1.3.bb: (log message trimmed) Nov 16 21:40:19 libungif-4.1.3: update recipe, fix packaging Nov 16 21:40:19 * Use lib_package to move static libraries to -static Nov 16 21:40:19 and *.la and *.so symlinks to -dev. Nov 16 21:40:19 * Converted to new style staging. Nov 16 21:40:19 * Ran oe-stylize.py. Nov 16 21:40:20 Signed-off-by: Andreas Oberritter Nov 16 21:50:43 khem: OK with new recipes or want to hold off? Nov 16 21:52:36 I would think new stuff is fine Nov 16 21:53:55 Tartarus: I guess new recipes are ok Nov 16 21:54:04 but need to be reviewed Nov 16 21:54:39 Tartarus: I will be more cautious of changes to classes and common metadata Nov 16 21:55:36 k, so will you ack/push tharvey's nuttcp recipe since you nak'd v1/v2? :) Nov 16 21:55:39 and would prefer no changes to toolchain Nov 16 21:55:42 unless needed Nov 16 21:55:49 yes will do Nov 16 21:56:03 actually I will go thru the patchwork in the evening Nov 16 22:00:51 qt4-embedded takes forever on this poor system Nov 16 22:05:08 khem: i feel your pain, qt4 for me means a couple hours of compiling + 1.5 hours linking Nov 16 22:15:52 khem, so, um, uclibc crashed anyhow Nov 16 22:17:08 And it's all -O2 not -Os, from run.do_compiles Nov 16 22:17:09 anyone experience with uclibc++ ? this one does not compile for me Nov 16 22:26:40 Using meta-toolchain to create an SDK results in a nice environment-setup script. Anyone know of examples where the cmake.bbclass was used to generate toolchain.make files for an SDK package? Nov 16 22:33:38 mrj 1.5 hrs of linking ?????????? Nov 16 22:33:42 thats insane Nov 16 22:34:01 hmmm probably I will have to look at what linker is smoking Nov 16 22:34:18 Tartarus: hmmm Nov 16 22:34:47 eFfeM: uclibc++ is kind of stale in OE Nov 16 22:35:06 eFfeM: you may wanna look at aboriginal linux Nov 16 22:35:19 and see if there are some patches you can steal Nov 16 22:35:27 Tartarus: where does it die Nov 16 22:37:12 khem, actually I got patches for a more recent version, but was not too sure about them, so thought I.d build the current recipe first Nov 16 22:37:12 hmm Nov 16 22:37:15 0.1.9 Nov 16 22:37:23 the new one is 0.2.2 Nov 16 22:37:34 I added a task, what do I need to do to make it some bitbake can run it Nov 16 22:37:36 like Nov 16 22:37:39 it is on my todo list Nov 16 22:37:43 khem: same place, oddly Nov 16 22:37:45 khem, is there a maintainer ? Nov 16 22:37:48 bitbake -c copy_to_target uhd-srctree Nov 16 22:38:10 denix, do you know why INITSCRIPT_PARAMS for gstreamer-ti gstti-init is 'start 30 5 2 . stop 40 0 1 6 .' and not something more like 'defaults 30' (curious why runlevel 3,4 are not supported) Nov 16 22:38:27 qemuarm + uclibc (minimal-uclibc for dist) is fine Nov 16 22:46:20 kergoth_: do you still feel responsible for uclibc++ ? git blame gives your name for most of the lines :-) Nov 16 22:46:54 nope, haven't touched it Nov 16 22:50:25 eFfeM: if you can add a working recipe I would be interested Nov 16 22:50:40 Tartarus: hmmm, is it with nptl ? Nov 16 22:51:00 Tartarus: check if it works with uclibc git recipe Nov 16 22:51:03 that uses nptl Nov 16 22:51:11 Tartarus: thats the last I checked against Nov 16 22:53:54 khem, what I have is supposed to be working, but i have to review the patches and maybe improve them, will try to find some time for it this week Nov 16 22:54:11 eFfeM: sooner the better Nov 16 22:54:19 argh, damnit gvim Nov 16 22:54:21 * kergoth_ glares Nov 16 22:54:46 kergoth_: that g infront of vim is problem I guess Nov 16 22:55:30 khem in that case I'll try to work on it tomorrow, (about bedtime now), I think I still owe you one (or 2 or 3 or more :-) ) Nov 16 22:55:38 its the gtk ui specifically Nov 16 22:55:42 will sent it to ML for review Nov 16 22:55:46 it doesn't seem to be supporting multiple fonts in guifont correctly Nov 16 22:55:49 eFfeM: excellant Nov 16 22:55:52 its supposed to be able to fall back if a font doesn't eixst Nov 16 22:55:58 but it doesn't do that, it just displays wrong instead Nov 16 22:56:17 * khem uses vim Nov 16 22:56:27 or kate Nov 16 22:56:54 i usually do, but its annoying being limited by the terminal, colorscheme wise Nov 16 22:57:13 hmm you can use schemes with terminal too Nov 16 22:57:23 yep, but you're limited to the number of colors the terminal supports Nov 16 22:57:28 khem: i just used default in sane-toolchain Nov 16 22:57:29 8, 16, 88, or 256, depending Nov 16 22:57:30 it should support 256 colors now a days Nov 16 22:57:39 Tartarus: hmmm I see Nov 16 22:57:51 last i checked, gnome-terminal was still 16. course i don't use gnome-terminal, but still :) Nov 16 22:57:54 Tartarus: which would be 0.9.31 I guess Nov 16 22:57:54 i use urxvt, 88 color Nov 16 22:58:11 hmm Nov 16 22:58:12 kergoth_: whats you version of gnome-terminal Nov 16 22:58:19 anyone have a vim 7.3 package for ubuntu handy? Nov 16 22:58:19 mine here has 256 colors Nov 16 22:58:27 yeah 0.9.31 Nov 16 22:58:36 7.3 has the 'conceal' feature Nov 16 22:58:37 which is handy Nov 16 22:58:39 Tartarus: may be we should up that to git Nov 16 22:59:00 because for release git would work better than any releases Nov 16 22:59:02 http://hamberg.no/erlend/files/cute_python.png Nov 16 22:59:04 check that out Nov 16 22:59:23 kergoth_: thats looks bs Nov 16 22:59:43 my neighbor is greek he can understand that prolly Nov 16 22:59:47 I disagree. having a real lambda is quite nice Nov 16 23:00:00 and it doesn't change the code, only how it looks -- unless you move the cursor onto that line Nov 16 23:00:11 uses conceal to fold away text inside the line Nov 16 23:00:19 its like folding but not line-wise Nov 16 23:00:21 fantastic Nov 16 23:00:28 there are other cool uses Nov 16 23:00:34 it can show ansi color codes in actual ansi color Nov 16 23:01:17 seriously I will be confused more by that Nov 16 23:01:25 but thats me Nov 16 23:01:43 i might disable all but the ones that are obvious. i'm sure a mathematician would love it Nov 16 23:01:50 e.g. set membership Nov 16 23:01:56 yeah Nov 16 23:02:09 what do people think about experimenting with layers after the release Nov 16 23:02:42 somehow split the recipes into multiple layers and keep a core which is aligned to poky Nov 16 23:02:47 * kergoth_ is all for it, but is distracted by bitbake development at the moment Nov 16 23:03:03 * kergoth_ is knee deep in trying to refactor so the parallel parsing support is less ugly Nov 16 23:03:10 I may be able to do heavy lifting depending upon time but no promises Nov 16 23:05:37 khem: So, just try out the current uclibc_git.bb? Nov 16 23:07:26 Tartarus: yes Nov 16 23:07:44 got the P_V magic handy? If not I'll recall what it is Nov 16 23:07:52 Tartarus: should be selectable with PREFERRED_VERSION_uclibc = "git" Nov 16 23:08:00 ah, k Nov 16 23:08:00 and PREFERRED_VERSION_uclibc-initial = "git" Nov 16 23:08:14 PREFERRED_UCLIBC_VERSION = "git" :) Nov 16 23:08:27 I dont think we have that in sane-toolchain Nov 16 23:08:28 but add it Nov 16 23:08:29 its ok Nov 16 23:08:30 * Tartarus waits for this timed test to finish Nov 16 23:09:01 hmm Nov 16 23:09:37 i wish we had more python folk messing with bitbake itself, i could use opinions on things :| Nov 16 23:15:30 is namedtuple a sane way to avoid creating a method or constructor with a shitload of methods? i'm thinking its faster than creating a config object manually, for cases where functions have too many args Nov 16 23:15:35 * kergoth_ ponders Nov 16 23:17:25 kergoth_: arent named tuples a python3 thing Nov 16 23:17:31 they were added in 2.6 Nov 16 23:18:27 from perf point they are as efficient as normal tuples I guess Nov 16 23:18:50 should be, i'd think, yeah Nov 16 23:18:54 03Frans Meulenbroeks  07org.openembedded.dev * r28d660f862 10openembedded.git/recipes/tgt/tgt_1.0.9.bb: Nov 16 23:18:54 tgt: removed 1.0.9; Nov 16 23:18:54 apparently this was forgotten when moving to 1.0.10 Nov 16 23:18:54 Signed-off-by: Frans Meulenbroeks Nov 16 23:20:09 kergoth_: I guess it makes code look better and readable too Nov 16 23:20:40 okay, bb.cache is a seriously leaky abstraction. it tries to make it transparent whether you're using the cache or not, but its *never* transparent, you always have to be aware, the cached data is used for an entirely different purpose than the regular data, and if we want a full recipe parse cache, we can implement it in bb.parse and make it truly transparent Nov 16 23:20:40 calling it a day, cya :-) Nov 16 23:20:44 with my python-fu I dont see any downsides Nov 16 23:21:34 I'm *trying* to decouple some of our modules some, they're *really* tightly bound at the moment -- god help you if you try to write a python script other than bin/bitbake that uses the modules to do stuff Nov 16 23:21:47 heh Nov 16 23:22:48 yes moving caching to parse is good as you see its broken anyway Nov 16 23:24:06 one of my coworkers did a functional implementation of parallel parsing, but its pretty ugly. each subprocess forks off of the main one and continues using its copy of the cooker, running its state machine to do its thing, collecting statistics and results in members of the class, then that has to somehow make it back to the main process, and the event queues have to be cleared, or the different forks of the main process will all try to send t Nov 16 23:24:40 i'm trying to make it so we can just have the subprocess take a filename as input and pass back an object with the data we need for the cache and runqueue generation, and thats it Nov 16 23:24:50 more functional, less procedural, less confusing Nov 16 23:24:53 but its a twisty maze in there Nov 16 23:24:58 so i'm getting turned around :) Nov 16 23:25:12 filename = recipe ? Nov 16 23:25:17 yeah Nov 16 23:25:26 makes sense Nov 16 23:26:07 okay, thanks, needed to make sure i wasn't completely insane here Nov 16 23:26:08 heh Nov 16 23:29:21 kergoth_: yeah I think its better to launch smaller threads to generate a piece of cache which than another consolidation thread keeps adding into the final cache as the little threads do the work may be per recipe Nov 16 23:30:12 * kergoth_ nods Nov 16 23:30:31 it also gets quite confusing when the same code is run in many processes, its hard to keep straight exactly which context you're in when going through the code Nov 16 23:30:43 end up having conditionals for is this the child or the parent in the methods and crap Nov 16 23:31:30 yeah unfortunately multi-threading is something human brains can not do when it gets more than 2 threads Nov 16 23:31:33 :) Nov 16 23:31:46 so I agree Nov 16 23:44:03 ok, now to make git send-email actually post to the list :) Nov 16 23:45:05 OK Nov 16 23:45:09 Anyone around that has built in a tmpfs? Nov 16 23:45:57 I tried not long ago but gave up after a few attempts Nov 16 23:46:11 tmpfs running out of space? Nov 16 23:46:27 like way before you would expect? Nov 16 23:46:32 yes, and it seems like you can't have deploy, pstage in diff tree's Nov 16 23:46:49 the whole of deploy, no Nov 16 23:46:52 just the whole of pstage, yes Nov 16 23:47:14 so you have pstage moved outside of tmp? Nov 16 23:47:19 yeah Nov 16 23:47:24 but not deploy? Nov 16 23:47:27 right Nov 16 23:48:05 ah, hmm, wait, maybe i know my tmpfs problem Nov 16 23:48:08 or at least something ot check Nov 16 23:48:15 my pstage is 5.1G - my deploy is 2.1G Nov 16 23:48:36 I would love to see you post something to wiki or maillist about how you set yours up with all the do's and don'ts Nov 16 23:49:05 # Keep pstage out of TMPDIR Nov 16 23:49:10 PSTAGE_DIR = "/scratch/OE-upstream/pstage.${LIBC}.${DISTRO}" Nov 16 23:49:12 is all it needs Nov 16 23:50:29 how big/inodes are you making your tmpfs? Nov 16 23:50:47 default, which i think is way too few Nov 16 23:51:02 I do recall running out of inodes when I used default Nov 16 23:51:48 Well, i'll know when I error out this time Nov 16 23:52:00 passed in a more normal-fs nr_inodes Nov 16 23:52:50 why is it that DEPLOY_DIR must be within TMPDIR? Nov 16 23:54:11 It's not Nov 16 23:54:12 Er Nov 16 23:54:13 sorry Nov 16 23:54:30 images are fine Nov 16 23:54:36 But kernel stuff is what wrecks it Nov 16 23:55:40 such as ipk? Nov 16 23:57:06 DEPLOY_DIR_IPK is the bulk for me - 1.5G of my 2.8G DEPLOY_DIR Nov 16 23:57:19 no, it's kernel do_deploys Nov 16 23:57:22 iirc Nov 17 00:05:08 Tartarus: ant__ was building in tmpfs some time ago with separated pstage dir Nov 17 00:05:49 k Nov 17 00:05:55 it must just be an inode thing then Nov 17 00:06:16 kergoth_: keep me informed about multithreaded parser progress please :) Nov 17 00:06:27 will do Nov 17 00:06:47 khem: well, what about our build? Nov 17 00:07:03 any one use dreambox ? Nov 17 00:07:41 I have no idea about device_table and locale-base-fr-fr bugs.. Nov 17 00:10:06 tried to clean glibc and rebuild Nov 17 00:10:08 w/o success Nov 17 00:10:24 any idea how to use gmail for smtp, but not have your gmail address show up in the email? Nov 17 00:11:00 ah.. seems it is using eglibc Nov 17 00:11:26 Crofton, does gmail handle the domain or no? Nov 17 00:11:31 if no, I think you can't Nov 17 00:11:45 ok Nov 17 00:12:05 Crofton: use envelope-from? Nov 17 00:12:24 isn't working Nov 17 00:12:33 heh.. seems gmail is too smart Nov 17 00:12:55 yeah Nov 17 00:12:56 did you have your address attached to gmail into gmail account settings? Nov 17 00:13:22 i.e. as address from which you can send mail via web-interface Nov 17 00:13:49 morning Nov 17 00:14:52 Jay7, trying that now Nov 17 00:15:19 as other option register separate account specially for this :) Nov 17 00:15:41 ah Nov 17 00:15:49 I think this worked, thanks JaMa|Off Nov 17 00:15:51 btw, gmail can use 'plussed' addresses iirc Nov 17 00:15:52 Jay7, Nov 17 00:16:14 yourmail+something@gmail.com Nov 17 00:17:00 I've even heard that it can mark mail arriving to such address with 'something' label Nov 17 00:17:11 but have not checked this Nov 17 00:17:33 its not automatic, you have to create filters for it Nov 17 00:17:38 but its doable easily enough Nov 17 00:19:36 adding the address I am subscribed to the OE list from to my gmail account worked Nov 17 00:19:59 cool Nov 17 00:20:18 Crofton: can you write some short instructions on wiki? Nov 17 00:20:33 about using git-send-email Nov 17 00:21:18 I copied from the wiki Nov 17 00:21:25 I suppose I could add some notes Nov 17 00:21:31 ah.. good :) Nov 17 00:22:58 wow Nov 17 00:23:10 khem: build was finished Nov 17 00:23:48 I've rebuild eglibc w/o GLIBC_GENERATE_LOCALES Nov 17 00:24:17 but requiring fr_FR locale is wrong imho Nov 17 00:24:29 Jay7: hey Nov 17 00:24:37 classes/image.bbclass:IMAGE_LINGUAS ?= "de-de fr-fr en-gb" Nov 17 00:24:39 haha Nov 17 00:24:45 that's a really lame default Nov 17 00:24:51 also, arbitrary Nov 17 00:24:57 forgotten imho Nov 17 00:25:12 * kergoth_ 's had that overridden in his site.conf for so long he probably forgot about it Nov 17 00:25:17 I'm deleting some old tmpdirs..the eglibc was fine *withy* locales generation Nov 17 00:25:34 damnit, i broke the bitbake cache somehow.. hrm Nov 17 00:25:52 s@i+...can't get used to this new kb Nov 17 00:25:58 well.. so, I should note that GLIBC_GENERATE_LOCALES should be consistent with IMAGE_LINGUAS then Nov 17 00:26:48 btw, lot of images have it overriden Nov 17 00:27:11 well, ok Nov 17 00:27:31 khem: build successed anyway Nov 17 00:27:40 so you can push that changes Nov 17 00:29:15 kergoth_: I noticed some/most images are setting IMAGE_BASENAME.. Nov 17 00:29:22 they should be, yes Nov 17 00:29:29 well, exporting Nov 17 00:29:43 oh, that's silly Nov 17 00:30:14 is it even whitelisted? Nov 17 00:30:26 not sure what you mean Nov 17 00:30:27 ah ha..first grep match.. Nov 17 00:30:37 ./classes/image.bbclass:# "export IMAGE_BASENAME" not supported at this time Nov 17 00:30:52 that just means its doing IMAGE_BASENAME[export] = "1" Nov 17 00:30:58 due to old bitbake not supporting the export directive without = Nov 17 00:31:05 k Nov 17 00:31:10 doesn't explain why its being exported at all :) Nov 17 00:31:18 still it seems to me unnecessary Nov 17 00:31:21 yep Nov 17 00:31:22 i agree Nov 17 00:31:23 right Nov 17 00:39:28 seems I should setup git-send-email too.. Nov 17 00:48:05 * kergoth_ grumbles Nov 17 00:51:28 late here...good night Nov 17 01:02:31 khem: I've posted patch to ML Nov 17 01:02:31 khem, ping? Nov 17 01:09:07 Any ideas? Nov 17 01:09:08 http://tinderbox.openembedded.org/public/logs/task/10211209.txt Nov 17 01:09:15 tracker dies looking for hal Nov 17 01:14:53 Crofton: what is in config.log? Nov 17 01:15:09 not sure Nov 17 01:15:22 I am trying to bitbake hal and restarting is my guess Nov 17 01:18:41 * Jay7 have started testbuilder for minimal distro and goes sleep Nov 17 01:19:27 gn Nov 17 01:20:13 Jay7, are you still using that spreadsheet? Nov 17 01:20:41 Crofton: btw, have you seen your mail in your gmail.com mailbox? Nov 17 01:20:57 which email? Nov 17 01:21:10 which was sent by git-send-email Nov 17 01:21:12 the confirmation email for the new address goes to that ddress Nov 17 01:21:31 ka6sox-work: I've not updated it Nov 17 01:21:33 I thnk it stopped going there Nov 17 01:21:56 ka6sox-work: have nothing new to show.. Nov 17 01:22:29 ok tracker is compiling now Nov 17 01:22:30 I've sent my patch to ML by git-send-email but I don't see it in inbox Nov 17 01:22:39 looks like it need some form of depends on hal Nov 17 01:22:49 * Crofton uses a multi core builder Nov 17 01:23:57 Jay7, thanks. Nov 17 01:24:06 multi-core only means it fails faster. Nov 17 01:24:55 :)) Nov 17 01:25:15 ka6sox-work: btw, I'm using 4/2 threads now Nov 17 01:25:35 Jay7: BB/make or make/BB ? Nov 17 01:25:39 bb/make Nov 17 01:25:55 tried tuning it up higher? Nov 17 01:26:04 seems more bb threads will not produce better time.. Nov 17 01:26:04 * Tartarus is starting to question his rule of thumb there Nov 17 01:26:09 Jay7, kewl! that should help to keep it out of iowait. Nov 17 01:26:17 HDD bw limit reached Nov 17 01:26:38 but I should analyze vmstat output.. Nov 17 01:26:55 right..thats why multi-core and the rule of J#= 2X number of Cores FAILS. Nov 17 01:27:19 I had been doing -j#*1.5, threads=N Nov 17 01:27:35 above 2cores and -j 4 seems to hit a wall. Nov 17 01:27:50 ideal case is bb threads == No of cores and -j2 :) Nov 17 01:28:11 it's sure true that very few things can make use of larger -j numbers Nov 17 01:28:20 gcc and *glibc and that's about it Nov 17 01:28:43 I've tought today about distributed building system Nov 17 01:29:38 Jay7, I'm thinking about 2 cores on each of 4 machines with -j4 and icecc Nov 17 01:29:40 nothing useful.. Nov 17 01:30:06 but autoconfig and other things will slow that down. Nov 17 01:30:12 yeah Nov 17 01:30:22 ka6sox-work: 4 cores have almost the same price imho Nov 17 01:30:27 That's one of the bottlenecks to making icecc a win over single big boxe Nov 17 01:30:29 s Nov 17 01:30:30 +- 20$ :) Nov 17 01:30:53 Jay7, its mostly about I/O BW...not cores Nov 17 01:31:00 Phenom II X4 is very cheap now Nov 17 01:31:15 'now'? heh, I don't wanna know how cheap they are now Nov 17 01:31:19 see, I'm using 4/2 :) Nov 17 01:31:25 * Tartarus got one when they were kinda new and would have sworn it was cheap then Nov 17 01:31:29 really...its *all* about I/O BW when building Nov 17 01:31:40 model name : AMD Phenom(tm) II X4 940 Processor Nov 17 01:31:44 is what i've got local Nov 17 01:32:05 Tartarus: I've awaited for X6 and got it now :) Nov 17 01:32:08 I have 2ea 2378's Nov 17 01:32:16 so 8 cores. Nov 17 01:32:33 there are 12-cored opterons already.. Nov 17 01:33:33 I should sleep! Nov 17 01:33:45 nn Jay7 Nov 17 01:33:48 sleep well. Nov 17 01:43:05 khem: Still fails, still -O2 not -Os **** ENDING LOGGING AT Wed Nov 17 02:59:57 2010