**** BEGIN LOGGING AT Sat Aug 29 02:59:56 2009 Aug 29 03:12:07 m4t: nice Aug 29 03:12:09 does it boot Aug 29 03:13:49 will try it in a bit Aug 29 03:13:56 need to do an initramfs for now Aug 29 03:14:04 ive been waiting on this slc dom for nearly a month :/ Aug 29 03:14:10 thought it'd be here today Aug 29 03:14:57 kergoth: once you do the braindump you will have one in me :) Aug 29 03:15:22 heh :) Aug 29 03:16:05 maybe i should do that before i think about how to rewrite the ast / semantic model for our recipes, then delegate ;) Aug 29 03:16:35 smart Aug 29 03:16:37 sure Aug 29 03:16:44 I will have like 10 days Aug 29 03:16:55 mandatory PTO starting long weekend Aug 29 03:17:01 ah Aug 29 03:25:53 zecke has started to improve the parser stuff by introducing a rudamentary ast... but things still aren't ideal, in my opinion. everything is flat, and files data flows into one another directly when an include happens, etc... I'd like to see something of a semantic model... an object model which reflects how it'll be used. hand back Tasks and Variables and stuff.. The cooker and other bitbake interfaces shouldn't need or want to see the Aug 29 03:25:54 implementation details of our datastore... Aug 29 03:27:26 kergoth: is there a wiki with all these proposed changes listed somewhere ? Aug 29 03:27:48 i don't think so. my personal list is at http://dl.getdropbox.com/u/112715/Documents/OpenEmbedded%20Tasks.html/index.html Aug 29 03:28:03 there are some things i haven't added to the list yet, though Aug 29 03:28:11 like the parser stuff Aug 29 03:28:26 wow. nice long list Aug 29 03:28:29 just what I wanted Aug 29 03:28:59 some of that is probably pretty trivial to implement, if something looks like a good change to make, feel free to dive in... :) its not just bitbake, but general OpenEmbedded stuff, even some specific recipe issues Aug 29 03:29:26 for the past couple years I've just gathered random notes about OpenEmbedded in an outline, and yesterday i extracted actionable tasks from it into that list you're looking at Aug 29 03:30:12 (-->dive in:) I hope to. Aug 29 03:30:31 how do you make that sort of a document ? The checkboxes and nesting is cool Aug 29 03:30:51 that's mostly specific things, i have a lot of more fuzzy things floating around my head and haven't really gotten that specific yet.. thats in the old document still Aug 29 03:31:03 it's exported HTML from OmniOutliner Pro on my MacBook Aug 29 03:31:19 * kergoth likes that you can click the arrows on the web page to collapse sections :) Aug 29 03:32:21 :) indeed. Aug 29 03:32:49 * rkirti bookmarks the page and prepares to go to sleep - 9 am here :-/ Aug 29 03:33:09 khem: do you think a sort of... "This is what happens when you run bitbake foo" would be useful? i.e. has the cooker module do this, which has the taskdata module collect this data, and gets this from the cache, and ... as a way of seeing the common execution flow amongst the python modules? Aug 29 03:33:18 * kergoth isn't good with diagrams :) Aug 29 03:35:47 kergoth: yeah general flow annotated with source code Aug 29 03:43:13 okay, sounds good. that's what i found most confusing diving back into it after my absence.. I'm going okay.. there are like 42 modules in this package, what the fuck do they all do, and which use which.. Aug 29 03:49:59 oh no Aug 29 03:50:00 Kernel panic - not syncing: junk in compressed archive Aug 29 03:50:06 * m4t wonder wtf Aug 29 03:51:56 i gave defconfig a path to an untar'd minimal-image Aug 29 03:52:11 owned by root, with opkg/lock chmod o+r Aug 29 03:55:45 can i give it a .cpio to CONFIG_INITRAMFS_SOURCE= Aug 29 03:55:46 ? Aug 29 03:56:12 i read through the actual helper program in the kernel that deals with those, but i wasnt able to figure out what it expects Aug 29 03:58:30 gen_init_cpio is the programused Aug 29 04:16:15 Kernel panic - not syncing: broken padding Aug 29 04:16:51 after giving defconfig a direct path to the cpio Aug 29 04:18:19 could mkimage be borked? Aug 29 04:18:27 * m4t is testing the ram module Aug 29 04:20:46 m4t: are you using uboot Aug 29 04:21:09 yea Aug 29 04:21:21 maybe its because i tried to embed minimal-image, rather than initramfs Aug 29 04:22:35 i havent messed with u-boot since i almost bricked my board with a typo Aug 29 04:22:44 but it has worked without issue so far Aug 29 04:24:33 if it fails again with an initramfs image, i will try the old eglibc version that was known to work Aug 29 04:27:00 m4t which version Aug 29 04:27:42 of eglibc? i think its 2.9 Aug 29 04:28:27 i already have an initramfs directory in tmp.old/deploy/eglibc/images/dht-walnut Aug 29 04:32:28 no i meant uboot Aug 29 04:33:16 oh 1.2.0 Aug 29 04:39:32 Kernel + compressed initramfs + uncompressed initramfs must fit in memory at same time Aug 29 04:39:44 can you make sure thats the case Aug 29 04:41:22 it is, i have 128mb of ram on the board Aug 29 04:41:37 i actually just booted a uimage a couple days old Aug 29 04:41:51 Linux dht-walnut 2.6.24.7 #1 Sun Aug 23 14:44:11 EDT 2009 ppc unknow Aug 29 04:44:26 ## Booting image at 00400000 ... Aug 29 04:44:26 Bad Header Checksum Aug 29 04:45:26 oh oops, i just tried to bootm a u-boot img Aug 29 04:54:33 hrm new kernel/toolchain + old initramfs boots Aug 29 05:10:55 o hey khem it seems to work :| Aug 29 05:11:04 i un-cpio'd minimal image Aug 29 05:15:54 udev hangs, i can ^C and get udhcpc to grab an ip, but i guess then the console device is not available Aug 29 06:03:49 heh, an 18m initramfs/ works, but a 14m shows 'junk in compressed archive' Aug 29 06:05:00 maybe its reading perms wrong Aug 29 06:39:34 03Sebastian Krzyszkowiak  07shr/import * rbf222e1f18 10openembedded.git/recipes/freesmartphone/opimd-utils_git.bb: opimd-utils: provide all pixmaps in package Aug 29 06:39:35 03Sebastian Krzyszkowiak  07shr/import * r90bce97e56 10openembedded.git/recipes/tasks/task-shr-feed.bb: task-shr-feed: add irssi Aug 29 06:39:35 03Sebastian Krzyszkowiak  07shr/import * r9414e7de2a 10openembedded.git/recipes/tasks/task-shr-feed.bb: task-shr-feed: add xournal and evince Aug 29 06:39:36 03Sebastian Krzyszkowiak  07shr/import * r552fcc8394 10openembedded.git/recipes/tasks/task-shr-feed.bb: task-shr-feed: add cu Aug 29 06:39:39 03Sebastian Krzyszkowiak  07shr/import * r7d157baa55 10openembedded.git/recipes/tasks/task-shr-feed.bb: task-shr-feed: add ipython Aug 29 06:39:42 03Thomas Zimmermann  07shr/import * raf7bf0bd79 10openembedded.git/recipes/aceofpenguins/aceofpenguins-launcher_0.3.bb: Fixes error in aceofpenguins-launcher Aug 29 06:39:45 03Frederik Sdun  07shr/import * re7b8359070 10openembedded.git/recipes/mokomaze/ (3 files in 2 dirs): (log message trimmed) Aug 29 06:39:48 Now the patches for mokomaze Aug 29 06:39:50 * Frederik Sdun [22.08.2009 15:34]: Aug 29 06:39:52 > Sorry. hit the y key Aug 29 06:39:54 > -- Aug 29 06:39:56 > IRC: playya @ Freenode, Gimpnet Aug 29 06:39:58 > xmpp: playya@draugr.de Aug 29 06:40:04 03Sebastian Krzyszkowiak  07shr/import * r64b9179512 10openembedded.git/recipes/tasks/task-shr-feed.bb: task-shr-feed: add phoneme-advanced-foundation Aug 29 06:40:07 03Thomas Zimmermann  07shr/import * r1e8b362d3e 10openembedded.git/recipes/openmoko-3rdparty/om-neon_svn.bb: Correcting theme path within neon Aug 29 06:49:20 i think what might be happening is that the initramfs is too large, and it is overwriting the still-uncompressing data in the uImage as it extracts Aug 29 06:49:45 http://meld.mvista.com/group_discussion.aspx?discussionid=96c1e4ec4db34b6baf396917f4e1f79c Aug 29 07:01:50 khem : micro image works fine w/uclibc and 4.1 as initramfs Aug 29 07:03:09 micro-base-image Aug 29 07:03:19 im trying micro-image now, to hopefully see if udev works Aug 29 07:39:38 m4t: try micro with gcc 4.4.1 Aug 29 09:42:51 hi Aug 29 09:55:53 hi woglinde Aug 29 09:57:26 he pb Aug 29 09:58:08 hm Aug 29 09:58:23 what are koen and brian doing with the ti dvsdk stuff Aug 29 09:58:35 now the duplicated same packages Aug 29 10:42:26 need a spapshot of openembedded Aug 29 10:42:58 i'm on dialup Aug 29 10:44:19 i mean snapshot Aug 29 10:47:13 snapshot is 100mb too Aug 29 11:05:38 03Benjamin Schieder  07shr/import * rd8f64d0cf1 10openembedded.git/shr-unstable/openembedded/recipes/brickshooter/ (brickshooter.desktop brickshooter_0.04.bb readyforfr.patch): (log message trimmed) Aug 29 11:05:38 New Package: brickshooter Aug 29 11:05:38 Hi all. Aug 29 11:05:38 I've just created a first working version of brickshooter for SHR. Aug 29 11:05:39 You can get a binary version from opkg.org: Aug 29 11:05:41 http://www.opkg.org/package_270.html Aug 29 11:05:43 I've attached a patch against shr-unstable/openembedded, please bear Aug 29 12:47:21 how to use gcc or g++ on angstrom console-image? Aug 29 12:48:15 or should i cross-compile on host machine rather than on beagleboard? Aug 29 12:49:05 ???? Aug 29 12:49:16 why you want compile native? Aug 29 12:49:24 you cannt compile on console image Aug 29 12:49:28 there is no compiler in Aug 29 12:49:31 i dunno its looks easier Aug 29 12:49:42 for what? Aug 29 12:49:45 hello world? Aug 29 12:49:46 maybe Aug 29 12:50:06 why not for other stuff? Aug 29 12:51:24 then i should make .bb recipes and include soft in that way? Aug 29 12:51:24 space and time consumption Aug 29 12:51:37 and the pain to setup the software your self Aug 29 12:51:48 if you are building the packages Aug 29 12:52:00 you can compile your soft through bitbake too Aug 29 12:55:24 why then when i did instructions from http://docs.openembedded.org/usermanual/html/recipes_examples.html#recipes_helloworld_example Aug 29 12:56:01 when i try to run /usr/sbin/myhelloworld i got error: -sh: myhelloworld: Permission denied Aug 29 12:56:34 whats ls -la usr/sbin/myhelloworld Aug 29 12:57:00 ans where is hello world example mention the compiling on the device? Aug 29 12:57:53 8 -rw-r--r-- 1 root root 6479 Aug 27 2009 myhelloworld Aug 29 12:58:55 you know the permissions on unix? Aug 29 12:59:29 not perfect, but i guess u are talking about chmod -x myhelloworld Aug 29 12:59:55 right Aug 29 13:00:15 and thats wyh the example has do_install Aug 29 13:00:21 with install -m 0755 -d ${D}${bindir} ${D}${docdir}/myhelloworld Aug 29 13:00:42 and i cant be changed by doing chmod -x ? Aug 29 13:00:52 ????? Aug 29 13:01:01 on the device you are the ruler Aug 29 13:01:08 you can what you want Aug 29 13:01:33 you can even mount a partition readonly and noexecutable Aug 29 13:01:40 yes, but when i did chmod -x myhelloworld its didnt make any change at permissions on myhelloworld Aug 29 13:01:59 maybee you try chmod 755 Aug 29 13:02:12 or chmod a+x Aug 29 13:03:11 chmod 775 worked and now i can run myhelloworld Aug 29 13:04:54 yeah it works since 30 years Aug 29 13:04:58 so no suprise Aug 29 13:05:54 thanks for help, im using linux for about week Aug 29 13:06:21 ah okay Aug 29 13:07:41 one more querstion, when i build a image with my new soft (like myhelloworld) Aug 29 13:08:38 i must to copy whole new rootfs or i cant do it with other way? Aug 29 13:11:11 you can copy over the ipk Aug 29 13:11:19 and use opkg/ipkg install bla.ipk Aug 29 13:12:10 where bla.ipk should be? Aug 29 13:13:31 i mean where oe creates the .ipk file Aug 29 13:14:43 ok, i found it in deploy/glibc/ipk/ Aug 29 13:16:45 hm seems your learning curve is very stall at the moment Aug 29 13:18:23 hm why u think like that? Aug 29 13:21:00 linux since one week now oe and bitbake Aug 29 13:21:22 that are not trival stuff Aug 29 13:23:46 i think same, for now i got lack of knowledge about developing under linux Aug 29 13:25:13 but i dont see any problems with obtaining it Aug 29 13:26:21 surikatga did you only windows before? Aug 29 13:28:23 woglinde: windows and dos on 486 ;p Aug 29 13:28:29 why u asking? Aug 29 13:29:34 ah okay dos Aug 29 13:29:43 so you are familiar with commandline Aug 29 13:30:09 because its so diffrent Aug 29 13:46:59 re Aug 29 13:49:23 hi rkirti Aug 29 13:49:30 hellow woglinde Aug 29 13:49:36 err...hello Aug 29 14:55:45 03Sebastian Krzyszkowiak  07shr/import * r60d6be0692 10openembedded.git/recipes/mokomaze/mokomaze-0.5.5/fsoraw.patch: mokomaze: fix building Aug 29 15:15:16 re Aug 29 15:54:34 where/how do I see how OE replaces the variables in the Makefile/configure.ac with the OE ones ? Aug 29 16:01:40 ???? Aug 29 16:19:06 huhu ph5 Aug 29 16:21:45 wogm Aug 29 16:22:32 err that was my son Aug 29 16:23:36 khem hihi Aug 29 16:24:10 he has learned to pull the chair back and climb on it Aug 29 16:24:18 fine Aug 29 16:25:14 hej woglinde Aug 29 16:29:07 pH5: hi long time Aug 29 16:29:47 hi khem Aug 29 16:34:04 I remember you being quite active when I was new Aug 29 16:45:31 jo gnutoo Aug 29 16:45:38 woglinde, hi Aug 29 17:21:07 khem: I'm looking into the ppoll/uclibc issue now Aug 29 17:21:46 khem: the beagleboard omap kernel already has the ppoll patch applied, so if uclibc is looking at the kernel headers, I'm not sure why it did not compile ppoll support Aug 29 17:21:54 (still digging) Aug 29 17:22:56 cbrake: we dont use same sources for real kernel and linux-libc-headers Aug 29 17:23:21 cbrake: I had the patch locally here but deleted it mistakenly Aug 29 17:24:47 khem: so we probably don't want to patch linux-libc-headers globally? Aug 29 17:25:15 khem: as some machine kernel's won't have ppoll support? Aug 29 17:25:26 cbrake: that correct. Aug 29 17:25:34 Thats why I was still pondering on it Aug 29 17:25:59 probably apply the patch conditionally only for machines which will use kernel with ppoll Aug 29 17:26:11 khem: hmm, seems like a rather strange problem Aug 29 17:26:35 khem: I wonder if no one but me is building udev with uclibc? Aug 29 17:26:41 yeah ideally we should generate libc-headers from the same kernel sources these days Aug 29 17:26:46 khem: and you I guess :-) Aug 29 17:26:51 cbrake: I was building it too Aug 29 17:27:01 and had exact same problem Aug 29 17:27:07 I patched 2.6.29 headers Aug 29 17:31:23 khem: so uclibc is a "machine" package, so there is no problem patching specific for a machine Aug 29 17:32:38 this will be linux-libc-headers package actually Aug 29 17:32:43 which is not general Aug 29 17:32:56 khem: glibc does not seem to be that way, rather armv7a, etc Aug 29 17:32:56 I mean which is general Aug 29 17:33:08 cbrake: glibc has emulation Aug 29 17:33:21 so it does not matter if syscalls exist or not Aug 29 17:33:27 khem: IC so that is why udev works with glibc -- I wondered about that Aug 29 17:33:59 yep Aug 29 17:34:44 khem: hmm, build/angstrom-2008.1/tmp/deploy/glibc/ipk/armv7a/linux-libc-headers (so this is not machine specific?) Aug 29 17:34:59 uclibc solely depends on kernel headers if syscall is not wired it will not configure it Aug 29 17:35:56 cbrake: I think we can do the patch like SRC_URI_beagleboard += "ppoll patch" Aug 29 17:38:45 khem: it still seems to me that patching a recipe for a machine seems that the resulting package should be split out by machine, but mayby I'm missing something Aug 29 17:43:55 hmm from distribution pov seems like that Aug 29 17:44:20 is it possible to get udev going without ppoll/pselect Aug 29 17:44:44 I don't know Aug 29 17:45:25 I would say that would be safest Aug 29 17:47:52 khem: I should probably actually be using mdev ... Aug 29 17:48:00 03Anton Olkhovik  07shr/import * ref485f5f92 10openembedded.git/recipes/mokomaze/mokomaze_0.5.5.bb: Mokomaze: fix package runtime depends Aug 29 17:50:21 hi, can i get support about oxnas? Aug 29 17:50:39 whats oxnaxs? Aug 29 17:51:17 western digital mybook world edition, and i have see a patch in git repo Aug 29 17:55:20 limon: what do you need Aug 29 17:56:08 i'm trying get eabi support, for switch arm to armel debian Aug 29 17:56:23 and i'm stuck at udev need 2.6.18, but my kernel 2.6.14 Aug 29 17:56:42 i see patch for 2.6.24, is it working? Aug 29 17:57:21 debian? why do you ask at #oe? Aug 29 17:57:37 booxter: patch in oe repo Aug 29 17:57:46 2.6.14 is full of security holes Aug 29 18:02:31 cbrake: yeah I think udev also should check for support before using i Aug 29 18:02:33 it Aug 29 18:13:06 I don't think it's possible to make modern udev or hal work without ppoll. Aug 29 18:13:52 the glibc emulation is pretty worthless, it just translates one kind of failure into another. Aug 29 18:14:15 pb lol Aug 29 18:15:25 but yeah, personally I use mdev anyway. Aug 29 18:23:20 pb hm right want to look at it Aug 29 18:24:07 pb did you build some easy support for mdev now? Aug 29 18:25:52 on in 1.15 we have ifplugd yeah Aug 29 18:26:01 yeah ppoll/pselect unless wired in kernel dont make much sense Aug 29 18:26:51 how special, minimal image with uclibc is now 6.3MB :-( Aug 29 18:27:04 cbrake hm Aug 29 18:27:17 gettext is occupying to much space Aug 29 18:27:31 but didnt find the time to cleanup my gettext stuff Aug 29 18:27:59 micro-image is 1.7MB Aug 29 18:28:14 hm Aug 29 18:28:16 a little more tolerable, but a far cry from my 1MB uclibc images built some time ago Aug 29 18:28:29 cbrake???? Aug 29 18:28:51 I think you mix something Aug 29 18:29:00 woglinde: uclibc minimal-image built with 1-year old OE snapshot were about 1MB Aug 29 18:29:04 nope Aug 29 18:29:08 never Aug 29 18:29:47 woglinde: which gcc version ? Aug 29 18:30:01 khem??? Aug 29 18:30:08 woglinde: http://pastebin.ca/1547414 Aug 29 18:30:20 cbrake but would be intressting whats got gibber Aug 29 18:30:23 args bigger Aug 29 18:30:50 question was to cbrake Aug 29 18:31:00 image size also depends on compiler Aug 29 18:31:11 and uclibc features included Aug 29 18:31:51 khem: looks like 4.3.3 Aug 29 18:32:12 khem: fairly recent OE, angstrom, PXA270 Aug 29 18:32:39 cbrake: I think you should untar both rootfs and compare big tickers Aug 29 18:32:49 like uclibc itslef Aug 29 18:33:04 or some other largish files Aug 29 18:35:31 cbrake: I just observed that u-boot .bss size overgrew with gcc 4.4.1 Aug 29 18:35:35 on ppc Aug 29 18:35:45 khem: woglinde: http://pastebin.ca/1547420 Aug 29 18:35:53 kernel image is an obvious one Aug 29 18:36:01 hi all Aug 29 18:36:08 gettext as well Aug 29 18:37:22 morning florian Aug 29 18:37:28 I guess micro-image is more what I want these days for a simple inittramfs Aug 29 18:39:20 cbrake: gettext libiconv are mising in one of them Aug 29 18:40:06 khem: nod, and those obviously are not needed for a simple initramfs Aug 29 18:40:38 for my custom image, some of that crud is getting pulled back in, so I guess I get to track down the dependencies Aug 29 18:40:44 otherwise sizes are same almost Aug 29 18:41:09 cbrake: micro-image isn't actually that micro. try micro-base-image. Aug 29 18:42:05 iirc, micro-image has udev and a bunch of other stuff that isn't really compatible with low footprint Aug 29 18:42:43 pb___: cool, its building mdev :-) Aug 29 18:43:44 pb___: 1.6MB for micro-base-image vs 1.7MB for micro-image Aug 29 18:44:48 comparison: http://pastebin.ca/1547429 Aug 29 18:52:28 cbrake: hm, not really the difference I was expecting. are you using DISTRO=micro? Aug 29 18:53:16 pb___: no, angstrom Aug 29 18:53:25 pb___: I'll try micro Aug 29 18:53:34 oh, angstrom. that would account for some of the extra space although I would still have expected the difference between micro-image and micro-base-image to be more. Aug 29 18:54:49 ah, I remember now, I added dropbear to micro-base-image. I guess that was probably not such a good idea for your situation, you might want to take it out again. Aug 29 18:57:07 hi florian Aug 29 18:59:11 hm Aug 29 18:59:19 kernel is included in one Aug 29 18:59:41 has anybody had luck with building pulseaudio - tinderbox reports successful builds Aug 29 19:00:06 rkirti whats you rproblem? Aug 29 19:00:18 I tought I fixed all building bugs Aug 29 19:01:37 woglinde: sorry, thats an unfinished line that wasnt meant to be entered - I realised midway I need a git pull :-/ Aug 29 19:01:50 hi, there is a file command equivalent in angstrom console-image? Aug 29 19:01:53 woglinde: though the fixes affect only the newer version I guess Aug 29 19:02:18 woglinde: since tinderbox shows only >=0.9.15 successful Aug 29 19:02:31 * khem cherry picks darwin host fixes Aug 29 19:05:40 what type of file is /sys/class/leds/beagleboard::usr0/trigger in angstrom console-image? Aug 29 19:09:52 hmmm we should also check for wget package in sanity.bbclass Aug 29 19:15:32 morning kergoth Aug 29 19:15:50 hi kergoth Aug 29 19:16:46 hey Aug 29 19:17:06 surikatga: a text file, I would guess Aug 29 19:19:00 yes but when i "cat triggers" i got: opt1 opt2 [opt3] Aug 29 19:19:11 also when echo opt1 > triggers Aug 29 19:19:30 from cat i will get [opt1] opt2 opt3 Aug 29 19:19:45 and want to know how work with this type of file under c Aug 29 19:21:41 well, much the same as any other file. you can open it with fopen(), write to it with fputs(), and so on. Aug 29 19:23:13 thanks Aug 29 19:27:45 pb___: but can u explain me why "echo "[opt1] opt2 opt3" > trigger" wont work when "echo opt1 > trigger" works pretty well Aug 29 19:28:05 its not normal for text file, right? Aug 29 19:42:39 rkirti yes I only fixed the newest Aug 29 19:44:30 hmmm python on mac returns a different string for md5sum and it seems bitbake does not parse it correctly Aug 29 19:45:02 looking at verify_md5sum Aug 29 19:46:34 e.g. md5 cooker.log returns MD5 (cooker.log) = 856e897f59b8a42a09403469b45439bf Aug 29 19:46:54 md5sum cooker.log returns ab3742bae533785da188a1939f1d1149 cooker.log Aug 29 20:24:31 if I use python 2.6 from ports q Aug 29 20:26:34 any thoughts on filesystems for OE builds? My ext3 build partition with 400GB of data seems to be getting a bit slow ... Aug 29 20:29:45 I have ext4 Aug 29 20:31:30 this is interesting -- adding mtd-utils to micro-image pulls in a bunch of extra stuff: http://pastebin.ca/1547566 Aug 29 20:32:22 that is unfortunate as I like to use mtd-utils for system update functionality in an initramfs Aug 29 20:32:24 hi Aug 29 20:36:22 holy shit Aug 29 20:36:36 there exist springpython Aug 29 20:37:25 any body familiar with usb port programming Aug 29 20:37:46 cbrake hm whats this? Aug 29 20:37:48 /armv5te/libgettextsrc_0.17-r2.3_armv5te.ipk Aug 29 20:41:42 woglinde: strange, I'm not sure where that came from Aug 29 20:41:58 woglinde: do you know what causes the gettext package to be pulled in? Aug 29 20:42:21 some locale package Aug 29 20:42:25 What causes an error like this --> http://pastebin.com/d7881b29b ? Aug 29 20:42:48 cbrake hm try opkg remove libgettextlib Aug 29 20:43:09 woglinde: I can run this on a target as I only have 2MB of space for this FS :-( Aug 29 20:44:03 cbrake qmeu? Aug 29 20:44:05 cbrake: mtd-utils need libuuid lzo and zlib Aug 29 20:44:18 rkirti hm Aug 29 20:44:49 These critical programs are missing or too old: as ld gcc Aug 29 20:45:09 arm-angstrom-linux-gnueabi-gcc: No such file or directory Aug 29 20:45:39 rkirti: woglinde this is because gcc-initial is needed Aug 29 20:45:50 for glibc-initial Aug 29 20:45:59 and somehow gcc is removed Aug 29 20:46:22 toolchain build has a strict sequence Aug 29 20:47:01 I would have preferred a feature in bitbake where I could say if gcc-cross is built and installed dont bother to even consider the intermediate packages Aug 29 20:47:11 unfortunately this is not done Aug 29 20:47:43 rkirti: best is clean gcc-cross gcc-cross-intermediate and gcc-cross-initial Aug 29 20:47:54 and glibc-intial Aug 29 20:47:56 khem: thats just what I did Aug 29 20:48:06 then and also glibc Aug 29 20:49:07 khem: I had avgue error with glibc that said my compiler didnt support -E and hence I cleaned the above stuff you mentioned . I forgot gcc-cross-initial though :-/ Aug 29 20:49:11 will check out again Aug 29 20:49:21 a/avgye/vague Aug 29 20:49:34 rkirti yeah Aug 29 20:49:55 actually there are same binaries referred by the packages generated by these recipes Aug 29 20:50:08 and if you clean one the rest gets corrupted Aug 29 20:50:24 but bitbake thinks they are ok and does not rebuild them Aug 29 20:50:44 thats why it skipped gcc-cross-initial and went to build glibc-initial Aug 29 20:51:14 I see Aug 29 20:51:43 if there was a way to flag the conflicts in bitbake then it would not happen Aug 29 20:52:28 and I think once glibc and gcc-cross are built intermidiate packages are not needed Aug 29 20:52:37 they are only needed to bootstrap the toolchan Aug 29 20:52:47 s/toolchan/toolchain/ Aug 29 20:53:54 "flagging the conflicts" ? Are there other such issues too..I mean is this a generic problem ? Aug 29 20:55:27 this is generic Aug 29 20:55:35 and very specific to toolchain Aug 29 20:55:48 I was thinking of adding capability to handle it Aug 29 20:55:51 rpm has it Aug 29 20:56:16 with rpm you can say this package is obsoleted by y Aug 29 20:56:32 and if you install y then x gets removed automatically Aug 29 20:56:34 jo ant Aug 29 20:57:00 khem: ah ok Aug 29 20:57:13 anybody using mac for OE build here ? Aug 29 20:57:35 khem hm zecke did some times ago Aug 29 20:57:36 no matter what I try bitbake can not find md5sum binary Aug 29 20:57:44 but then installing linux on his macs Aug 29 20:57:52 and the binary is there in my path Aug 29 21:02:46 hm springpython rockz Aug 29 21:02:55 http://springpython.webfactional.com/1.0.0/reference/html/remoting.html Aug 29 21:05:08 wuahah Aug 29 21:06:04 lol Aug 29 21:06:17 we should rewrite bitbake with spring stuff Aug 29 21:07:20 khem , any ideas about the bogus initramfs image? do you think it ends up being too large and is overwritten as it uncompresses? Aug 29 21:07:45 that would be my first guess Aug 29 21:07:57 m4t: you said something else worked ok with gcc 4.4.1 Aug 29 21:08:06 so I would not suspect toolchain Aug 29 21:08:33 yea, i got the micro and micro-base images working Aug 29 21:08:40 as well as the eglibc initramfs/ directory Aug 29 21:08:49 which was a base-image i think Aug 29 21:09:00 right Aug 29 21:09:09 so I think somehow it grows big Aug 29 21:09:18 I suspect bss size growth Aug 29 21:09:38 .bss, okay Aug 29 21:11:27 why would .bss cause the initramfs to be too big? Aug 29 21:12:21 the actual uImages are around 6.6mb Aug 29 21:12:38 ~5mb of that is the initramfs Aug 29 21:12:39 okay, m4t Aug 29 21:15:08 cbrake: what's the build failure you're seeing with micro-uclibc? Aug 29 21:16:20 pb___: http://pastebin.ca/1547622 Aug 29 21:19:16 cbrake: are you trying to use eabi, or the old abi? Aug 29 21:19:57 pb___: I assume eabi, but I'm not sure as all I did was configure for micro-uclibc distro and hit go Aug 29 21:20:02 it sounds as though uclibc is configuring itself for old abi (and the target triple of cm-x270-oe-linux-uclibc would sort of support that) yet the kernel thinks you are on eabi. Aug 29 21:20:26 yeah, it ought to be eabi by default Aug 29 21:20:53 yeah thats the problem I think Aug 29 21:21:25 have a look in sane-toolchain magic which sets this triplet magically Aug 29 21:23:48 cbrake: TARGET_OS setting in micro.conf is problematic Aug 29 21:23:52 prolly here Aug 29 21:25:18 cbrake: ah, heh. micro-uclibc currently only selects eabi for armeb, not for regular arm Aug 29 21:26:01 try setting TARGET_OS_arm = "linux-uclibcgnueabi" in local.conf and see if the situation improves for you. Aug 29 21:26:02 pb___: locally I removed it here and since it uses sane-toolchain includes I let it rely upon that Aug 29 21:27:46 khem: so just remove the TARGET_OS settings all together? Aug 29 21:28:15 in micro.conf yes Aug 29 21:28:38 I wonder if micro and micro-uclibc distros could be combined like they are in Angstrom? Aug 29 21:28:42 but that said it should work with OABI too so somehow uclibc is broken Aug 29 21:29:14 cbrake: right now if you use DISTRO=micro LIBC=uclibc its same Aug 29 21:29:29 I have intended to remove -uclibc versions Aug 29 21:29:43 ok, that sounds good Aug 29 21:30:47 cbrake: personally I would prefer to keep them separate. uclibc and glibc binaries are not interoperable, and it seems like a reasonable thing to require the distro to pick one and stick to it. Aug 29 21:31:01 I would be sort of tempted just to ditch the glibc version of micro altogether, it seems like a bit of an oxymoron. Aug 29 21:32:01 pb___: yah that would be fine too Aug 29 21:32:31 but, if folks do want to use glibc with micro-type policies for whatever reason, it would seem to make sense that it should be a separate DISTRO. Aug 29 21:34:32 cbrake: I'm reading your rants abous (initramfs) size Aug 29 21:34:46 the only solution appears to be klibc Aug 29 21:35:38 ant__: why do you say that? Aug 29 21:35:59 we had 1,2mb size-limiy for kernel+initramfs Aug 29 21:36:11 we still have (Zaurus) Aug 29 21:36:17 fixing this problem with mtd-utils hauling in gettext and suchlike doesn't seem like it should require klibc Aug 29 21:36:24 we tried uclibc ...no way Aug 29 21:36:45 I tried to bake a minimal-image-with-mts timne ago Aug 29 21:36:54 with uclibc was around 2.6 mb Aug 29 21:37:01 (1 yrs ago) Aug 29 21:37:11 *with mtd Aug 29 21:37:41 ant__: did u use USE_NLS=no Aug 29 21:37:48 ah, if your goal is 1.2MB for kernel _plus_ initramfs (which means, I guess, about 200k for the initramfs itself), you might be right that klibc would be required. Aug 29 21:38:17 khem: I tried with allposible patches iirc (no-gettext) Aug 29 21:38:35 uclibc itself will not fit into 200K Aug 29 21:38:35 it didn't sound like cbrake's requirements were quite that stringent; he seemed to be talking about 1MB for the initramfs and I think that should be comfortably achievable with uclibc. Aug 29 21:38:55 I would think so Aug 29 21:39:07 1mb is luxury :) Aug 29 21:39:09 statically linked busybox might be an option Aug 29 21:39:14 if you have like 1M Aug 29 21:39:59 from a 2008-09 snapshot of OE, I'm able to build 2.2MB initramfs with mtd-utils, and some other things Aug 29 21:40:13 maybe I was a little unreasonable suggesting 1MB Aug 29 21:40:40 those sizes are before compression, right? Aug 29 21:41:19 after, I was talking about nand size Aug 29 21:41:32 pb___: the sizes in the installed-package-sizes.txt lots are likely before compression, but I really don't know where they come from Aug 29 21:41:40 1MB compressed size should be easily doable; my nslu2 images are coming out at about 800k compressed at the moment and I think they still have a certain amount of bloat in them. Aug 29 21:42:00 pb___: the image sizes I'm throwing around are cpio.gz compressed images Aug 29 21:42:03 1MB uncompressed would be a bit more of a challenge but I imagine it still ought to be achievable. Aug 29 21:42:25 cbrake: hm, right. they do seem awfully high then. Aug 29 21:42:47 looking at the images that I happen to have lying around here, the last .tar.gz image I built for nslu2 is 660970 bytes. Aug 29 21:43:08 I don't imagine the tar-vs-cpio thing makes a lot of difference. Aug 29 21:43:17 Angstrom~cpio.gz│ 89141│Aug 29 03:28│ Aug 29 21:43:20 :) Aug 29 21:43:37 I guess I can try one with mtd-utils in and see what happens Aug 29 21:44:22 pb___: that would be interesting Aug 29 21:44:26 pb___: I added some bloat because I had to add my nand-read code in userspace (klibc) Aug 29 21:44:45 I think this coul be just a couple of lines in kernel-space Aug 29 21:45:29 btw..I missed the move (years ago) from mtd.h to mtd-user.h Aug 29 21:45:35 I did try building an ipk for mtd-utils and it didn't seem to have any obvious dependency on gettext, so I suspect you might have been suffering from some angstrom-specific effect there. Aug 29 21:45:54 :/ Aug 29 21:46:56 hmm, should micro-uclibc be building eglibc-initial? Aug 29 21:47:30 no but it has a dependency Aug 29 21:47:39 incorrect one on glibc Aug 29 21:47:48 cbrake: I ended up coding my routines for reading the nand...I really don't depend from mtd-utils anymore (well, I still need flash_eraseall & co.) Aug 29 21:48:22 ant__: nod, I use flash_eraseall and nandwrite Aug 29 21:48:30 well, the code is 'stolen' from mtd-utils anyway ;) Aug 29 21:48:37 PREFERRED_PROVIDER_virtual/libintl Aug 29 21:48:48 ant__: would probably be nice to have a nice compact library for implementing system update mechanisms Aug 29 21:49:14 cbrake: no, I don't think it should be building any *glibc bits. you should just have uclibc-initial and uclibc. Aug 29 21:49:34 * cbrake looks again ... Aug 29 21:49:39 fwiw, my mtd-utils.ipk has: Aug 29 21:49:40 Depends: libc6 (>= 2.6.1), libgcc1 (>= 4.2.4), libz1 (>= 1.2.3), liblzo1 (>= 1.08), libuuid1 (>= 2.16) Aug 29 21:49:46 do you have gettext in there as well? Aug 29 21:50:02 oh, wait, that's the wrong one. that must be a glibc build. Aug 29 21:50:19 cbrake: punt this line PREFERRED_PROVIDER_virtual/libintl = "glibc" Aug 29 21:50:28 from micro.conf Aug 29 21:50:29 pb___: I built some months ago, with a patch c/o woglinde nuking gettext Aug 29 21:50:38 re kergoth Aug 29 21:50:54 gah: http://pastebin.ca/1547648 Aug 29 21:51:02 ant__ hm really nuking? Aug 29 21:51:10 lol Aug 29 21:51:28 looks like I still have some work to get libc stuff set up correctly for ARM EABI Aug 29 21:51:33 cbrake: doh, suck Aug 29 21:51:45 it's a bit sad that uclibc doesn't feature there at all Aug 29 21:52:02 heh Aug 29 21:53:03 but yeah, it does sound like the configuration bits are somewhat broken at the moment. Aug 29 21:53:11 cbrake: hmm how does your local.conf look like Aug 29 21:53:53 khem: http://pastebin.ca/1547653 Aug 29 21:54:43 ehm..do you guys suggest to try eglibc? Aug 29 21:54:57 (for normal images) Aug 29 21:55:38 afaik it can be configures uclibc-alike Aug 29 21:55:42 eglibc worked without issues for me Aug 29 21:55:54 size? Aug 29 21:55:56 4.4.0/2.9 Aug 29 21:56:10 real savings? Aug 29 21:56:27 cbrake: for now try DISTRO=micro and LIBC=uclibc in your local.conf Aug 29 21:56:58 1.5Mlibc-2.10.1.so Aug 29 21:57:10 cbrake: micro-uclibc is broken because it includes distro/micro.conf before defining LIBC='uclibc' Aug 29 21:57:19 khem: setting TARGET_OS_arm = "linux-uclibcgnueabi" seems to have fixed the multiple providers are available for virtual/libc issue Aug 29 21:57:49 m4t: you are using gcc 4.4.1/eglibc 2.10 Aug 29 21:58:10 khem: will "DISTRO=micro and LIBC=uclibc in your local.conf" have the same effect with no TARGET_OS settings? Aug 29 21:58:24 cbrake yes Aug 29 21:58:31 khem: ok Aug 29 21:58:37 trying ... Aug 29 21:59:13 then probably we should remove micro-uclibc.conf and as pb___ suggested use uclibc as defualt system C lib on micro itself Aug 29 21:59:40 and if someone wants to use eglibc or glibc he can override LIBC in his local conf Aug 29 22:00:15 khem: sounds good to me -- seems like we should get some of this stuff cleaned up Aug 29 22:00:27 yes Aug 29 22:01:12 I dare to say the (override of the ) choice of which libc to use is appropriate in local.conf Aug 29 22:01:48 ant__: is that an assertion ? Aug 29 22:01:58 after rethink...yes Aug 29 22:02:42 I just reworded your sentence in fact :) Aug 29 22:02:53 I think I would prefer just to see micro.conf and micro-uclibc.conf fixed so that they each work correctly. There doesn't seem to be any real need to delete either of them, and I'm not sure that suddenly flipping micro itself from glibc to uclibc would be very well received by its users. Aug 29 22:03:36 pb__ thats easy then micro-uclibc.conf will look like this Aug 29 22:04:17 LIBC = "uclibc" require conf/distro/micro.conf Aug 29 22:04:24 as of now Aug 29 22:04:31 yes, possibly Aug 29 22:04:40 khem: seems to be working here Aug 29 22:05:22 khem yea, that eglibc 2.10.1 came from 4.4.0 Aug 29 22:05:27 cbrake: yeah I expected it to Aug 29 22:05:33 dinner bell, its been fun, thanks for all the help Aug 29 22:05:39 vs. 404K libuClibc-0.9.30.1.so Aug 29 22:05:52 that is soft float, it might be smaller without that Aug 29 22:06:26 m4t: eglibc is trying to hit middle ground it probably will never become as small as uclibc Aug 29 22:06:43 yea Aug 29 22:07:37 ubuntu 9.04 w/out soft float: 1.3M /lib/libc-2.9.so Aug 29 22:07:56 different arch, too Aug 29 22:08:07 m4t: debian is already using eglibc in experimental Aug 29 22:08:14 i read that Aug 29 22:08:15 so future ubuntu will have eglibc too :) Aug 29 22:08:21 yup probably Aug 29 22:08:36 * khem back to do some mac stuff Aug 29 22:43:36 ~lart epiphany devs...killall -9 epiphany practically on each javascript... Aug 29 22:44:26 oh, yes, CIA is on vacation :) Aug 29 23:24:48 gm Aug 29 23:32:41 hi likewise Aug 30 00:09:56 likewise_: what do you mean by 'fix' -g support? i used it multiple times in my internal training presentation on debugging in mvl6, at MV, with success. just used a shell snippet to generate all the -I's from a grep of bitbake -s, to trim down the output Aug 30 00:15:26 kergoth: it has been broken numerous times, let me check what it does for me now Aug 30 00:17:28 kergoth: ok, the single character per node bug seems gone, much better Aug 30 00:18:21 springpython rockz Aug 30 00:20:38 okay good nite Aug 30 00:33:26 kergoth: indeed, the bugs seem fixed, I should have checked before I wrote my email. I once added regex support to -I, and used native and cross in them mostly. Aug 30 00:38:46 good nite **** ENDING LOGGING AT Sun Aug 30 02:59:57 2009