**** BEGIN LOGGING AT Tue Mar 20 02:59:57 2007 Mar 20 03:45:43 I'm trying to compile opie-recorder with the .dev branch and get this message: cvs [update aborted]: cannot rename file .new.language. to language.png: File exists Mar 20 03:45:58 Anyone know how to fix that? Mar 20 03:53:10 n/m, figured it out. Mar 20 04:06:45 SirManiac: you wewre right. It was the spacing Mar 20 04:07:23 goot to hear Mar 20 07:22:16 good morning! Mar 20 07:25:05 so what is currently a kinda "bootstrap" image to be build ? Mar 20 07:27:57 can i create a image out of task-base ? like bitbake base-image ? Mar 20 07:35:31 oh never mind Mar 20 07:36:00 i thought there is no bootstrap-image anymore , but it still exists Mar 20 07:49:33 Where can I specify the machine byte order? Mar 20 08:00:52 morning all Mar 20 08:00:56 hi Mar 20 08:01:44 hillct: make a different bb.file to handle the sound files. something like sounds-openpbx.bb Mar 20 08:21:05 morning all Mar 20 09:15:18 morning... Mar 20 09:15:40 do any of of guys use meta-toolchain? Mar 20 09:15:46 good morning all Mar 20 09:16:39 I don't quite understand the directory structure of the filan tarball it creates Mar 20 09:16:56 there are several include directories... Mar 20 09:17:01 /usr/local/i586/oe/i586-sts-linux/include/ Mar 20 09:17:12 which seems to include most of the headers Mar 20 09:17:18 tkp: you'd have to ask florian Mar 20 09:17:40 but then there is also /usr/local/i586/oe/include/ which contains a couple more (including libiberty.h from binutils) Mar 20 09:18:07 koen: he about on irc ever? Mar 20 09:18:07 morning all Mar 20 09:18:15 ~seen florian Mar 20 09:18:17 hey RP Mar 20 09:18:54 A package I'm writing needs bfd.h, which appears to come from binutils. Mar 20 09:18:54 florian was last seen on IRC in channel #gpe, 12h 15m 56s ago, saying: 'hi mmp'. Mar 20 09:19:30 I was going to alter the binutils-cross-sdk do_install() to also install bfd.h Mar 20 09:19:39 (it currently only does libiberty.h for some reason) Mar 20 09:20:09 hi Mar 20 09:20:31 but now I see in my toolchain (which was built from meta-toolchain) what libiberty.h has been installed into this strange other location... and I'm not sure why Mar 20 09:20:31 koen: does Debian/ARM guys finally got armel as official port or it is still unofficial one? Mar 20 09:20:59 hrw: they need to rebuild it on their underpowered arm machines Mar 20 09:21:07 hrw: so I guess it will take some months Mar 20 09:21:11 koen: thx Mar 20 09:23:30 so quick to ubuntu armel :-) Mar 20 09:24:46 * XorA considers buying a n2100 again as his machine died last night Mar 20 09:25:09 hmmm Mar 20 09:25:23 bitbake 1.9.0 decides to hold back messages at random it seems Mar 20 09:26:06 drat Mar 20 09:26:20 don't build both 'dates' and 'openmoko-dates' Mar 20 09:26:31 openmoko-dates now Depends: on dates Mar 20 09:30:06 koen: "holds back messages"? Mar 20 09:32:36 -c rebuild just says "rebuild started" "rebuild ended" and nothing in between Mar 20 09:33:00 and parsing seemed to be stuck at 20% till it suddenly completed Mar 20 09:34:32 koen: not good. When did that start happening? Mar 20 09:34:49 this morning after 'svn up' Mar 20 09:36:56 It does it here too :/ Mar 20 09:42:44 morning all Mar 20 09:45:54 03rpurdie * r799 10bitbake/lib/bb/build.py: build.py: Use a try/finally to make sure logs get redirected after function execution even if an exception occurs Mar 20 09:46:00 koen: That might fix it... Mar 20 09:46:05 hi ade|desk Mar 20 09:46:10 RP: thanks! Mar 20 09:46:50 koen: Its nice to have found that as I think that might have been causing some head scratching in the UI code (where all the changes make a bug like that much harder to find!) Mar 20 09:54:24 tkp: florian went online :) Mar 20 09:55:07 hi Mar 20 09:55:20 florian: hi.. Mar 20 09:55:28 I've been told that you use meta-toolchain?? Mar 20 09:55:47 I have a couple of questions about the layout of the sdk that gets created Mar 20 09:56:29 there seems to be more than one lib dir Mar 20 09:56:42 /usr/local/i586/oe/lib/ & /usr/local/i586/oe/i586-sts-linux/lib/ Mar 20 09:57:18 the first one seems to contain only specific versions of a couple of so's... whilst the second one contains most of the rest Mar 20 09:58:25 I currently only have the second one in my include dirs Mar 20 09:58:36 should I be including both of them? Mar 20 09:59:35 it's the same with include dirs too... there is /usr/local/i586/oe/include/ & /usr/local/i586/oe/i586-sts-linux/include/ Mar 20 10:00:05 the first one only contains a couple of headers, the second contains all the rest Mar 20 10:00:16 why have the headers been split up like this? Mar 20 10:01:27 it seems to be only libiberty.h and ansidecl.h from binutils in /usr/local/i586/oe/include/ Mar 20 10:02:51 I am in ned of bfd.h in my toolchain include dir (which also comes from binutils), but am unsure of where to install it to... the base_libdir or libdir Mar 20 10:03:46 I'm thinking libdir, since that is the only one in my search path... but perhaps my search path is wrong, and should be including the files from base_libdir too? Mar 20 10:11:03 one sec, phone call Mar 20 10:23:28 re Mar 20 10:24:01 florian: wb Mar 20 10:24:19 tkp: indeed, but i don't really know the reason to split this stuff. which compiler and distribution did you select? Mar 20 10:25:04 florian: this is my own distro... but it's gcc 3.3.4 Mar 20 10:25:25 binutils 2.15.94.0.1 Mar 20 10:26:19 anyone know who wrote the binutils-cross-sdk packages? Mar 20 10:26:37 tkp: I suggest switching to 2.16 if you plan to use on-device toolchain Mar 20 10:26:59 hrw: I don't want an on device toolchain Mar 20 10:27:15 but some software I'm compiling needs bfd.h, from binutils Mar 20 10:28:40 tkp: the ehader files should go to some include dir Mar 20 10:28:48 or... would using bfd.h require me to have binutils on the target device? Mar 20 10:29:16 florian: right... they do, but not to the same one all all the other toolchain header files Mar 20 10:29:49 the sdk tarball seems to have two different include dirs Mar 20 10:31:56 tkp: and binutils is the only bit with the headers in that different dir? Mar 20 10:32:29 no.. there are a couple from gcc too... Mar 20 10:32:49 if you look at the gcc recipie, they have gone out of theirt way to move these headers and libs Mar 20 10:32:55 their* Mar 20 10:33:20 gcc-package-sdk.inc Mar 20 10:35:21 tkp: gcc-package* seems to be arcane and buggy Mar 20 10:36:52 so, you think it is wrong that these libs/headers should live in /usr/local/i586/oe/lib rather than /usr/local/i586/oe/i586-sts-linux/lib/ ? Mar 20 10:38:11 /usr/local/i586/oe/i586-sts-linux/lib/ looks like the right place Mar 20 10:38:36 that would allow you to install different i586 toolchains if you want Mar 20 10:38:48 koen: thats what I would have thought... at least, thats where everything else installs it's libs too Mar 20 10:38:51 right Mar 20 10:39:26 so, the bug is with gcc-package-sdk, and binutils-cross-skd-* Mar 20 10:39:33 I haven't gone so far with a new sdk... i still fight against gcc-cross-sdk Mar 20 10:43:33 anyone experienced a GLIBC problem with "query_module"? Mar 20 10:44:06 In my glibc_2.3.5+cvsSOMETHING I read in tmp/work/.../build-arm-linux/misc/stub-syscalls.c this line: "weak_alias (_no_syscall, query_module)" Mar 20 10:44:14 I just can't understand why someone has gone to so much trouble to move these libs out of /usr/local/i586/oe/i586-sts-linux/lib/ and into /usr/local/i586/oe/lib/ in the first place Mar 20 10:44:28 tkp: I can't either Mar 20 10:44:29 and later compilation of modutils barks that calling "query_module" will always fail Mar 20 10:48:20 koen : if we have a machine without fpu, shouldn't the -msoft-float by used to build all packages ? Mar 20 10:48:42 Ifaistos: depends Mar 20 10:48:56 koen : on what ? Mar 20 10:49:01 Ifaistos: doing softfloat in kernel or userspace is a distro decision Mar 20 10:50:35 koen : i was under the impression that once defined it is used thought out, but just noticed it dod not Mar 20 10:51:00 Ifaistos: TARGET_FPU="soft" should trigger userspace softfloat Mar 20 10:51:46 koen : well it does not :) Mar 20 10:56:43 koen : i am building with soft-float enabled and this is for example how speex compiles -> http://rafb.net/p/amM6wM97.html Mar 20 10:57:07 Ifaistos: speex could be resetting CFLAGS Mar 20 10:58:08 koen : its not only speex :( Mar 20 10:59:05 koen : i was thinking it was added to TARGET_CC_ARCH Mar 20 10:59:48 Ifaistos: yes, but that doesn't help if the app resets CFLAGS Mar 20 11:00:59 koen : i don't think i does. -mcpu=405 which is set by the machine config shows up Mar 20 11:01:19 weird Mar 20 11:01:42 Hmm, seems that arm-linux/include/asm/unistd.h doesn't have the sys_query_module define in it, i'll try and patch linux-libc-headers 2.6.18 (my box runs ancient kernel 2.4.21, which uses this syscall) Mar 20 11:03:13 koen : this is my local.conf ->http://rafb.net/p/fYxjfL97.html Mar 20 11:07:33 Hi - i am trying to use OE behind a http proxy => works fine for wget, but what do I do about builds trying to get something from a cvs repository? Mar 20 11:08:19 oops i see it's in the FAQs already, sorry Mar 20 11:10:00 well i found this here: http://www.openembedded.org/search/node/FETCH_COMMAND but not a really helpful entry...maybe the author (koen?) could improve it a bit? Mar 20 11:10:38 * koen should stop importing stuff to new system Mar 20 11:10:52 everyone thinks I wrote OE and the wiki in one day ;) Mar 20 11:12:31 koen, yes :-) Mar 20 11:13:30 kOEn: nick suggest all Mar 20 11:25:58 morning Mar 20 11:28:56 03koen 07org.oe.dev * rda190699... 10/ (1 packages/cairo/cairo_1.4.2.bb): cairo: add 1.4.2 (latest stable) + surface-cache patch Mar 20 11:38:53 hrw: moin! Mar 20 11:41:24 hi schurig Mar 20 12:06:14 Hi! Mar 20 12:06:47 RP: did kernel-image cleanup went online yet? I don't see commits from you Mar 20 12:15:43 bbl Mar 20 12:24:05 hrw: would it be OK to use linux-libc-headers 2.6.18 even on target where one has an ancient (2.4.21) kernel? Mar 20 12:24:36 hey mickeyl Mar 20 12:24:47 morning Mar 20 12:24:49 hrw: if yes, then I'd like to commit a patch (via SRC_URI_append_mnci) that adds sys_query_module and sys_create_module, otherwise depmod from modutils won't work Mar 20 12:24:55 mickeyl: moin? Mar 20 12:25:01 mickeyl: ahem, moin! Mar 20 12:25:02 ~utg Mar 20 12:25:08 ~ugt Mar 20 12:25:10 methinks ugt is Universial Greeting Time. Created in #mipslinux, it is a rule that states that whenever somebody enters an IRC channel it is always morning, and it is always late when the person leaves. The local time of any other people in the channel, including the greeter, is irrelevant. http://www.total-knowledge.com/~ilya/mips/ugt.html Mar 20 12:25:12 :) Mar 20 12:25:18 schurig: l-l-h does not have to much kernel Mar 20 12:25:20 mickeyl: the ! was a typo Mar 20 12:25:35 schurig: for OZ we used 2.6.11 l-l-h and 2.4.18/2.6.x kernels Mar 20 12:25:46 mickeyl: yeah, I'm so much outside OZ now that I'd like to ask before committing Mar 20 12:25:58 hi mickeyl Mar 20 12:26:02 hey hrw Mar 20 12:26:36 schurig: yes, it all depends what minimal version you pass to glibc Mar 20 12:26:40 mickeyl: with glibc2.3.5+cvsBLAH and linux-libc-headers 2.6.18 there was no query_module in the libc, therefore modutils didn't work Mar 20 12:27:25 schurig: hrw is the man. Since working on OpenMoko I didn't have time to look at anything else :/ Mar 20 12:27:29 zecke: "pass a minimal version to glibc"? How do I do this? Mar 20 12:28:35 schurig: ./configure --help Mar 20 12:28:45 schurig: I don't remember the name of the option Mar 20 12:29:37 hrw: I see, l-i-h 2.6.11.1 still includes the line "#define __NR_query_module (__NR_SYSCALL_BASE+167)", whereas l-i-h 2.6.18 doesn't anymore Mar 20 12:30:44 schurig: I would add it in OE then. Mar 20 12:31:01 hrw: you sure? Mar 20 12:31:10 hrw: there must be a reason why it got removed Mar 20 12:31:48 koen: the reason is that in 2.6.xx, it's done very different (modutils vs. module-init-tools) Mar 20 12:32:39 koen: I believe that they kept the syscall in asm-*/unistd.h for some time until someone did a janitoring job Mar 20 12:33:05 I think the better solution is to use llh 2.6.11 for 2.4.ancient kernels Mar 20 12:33:19 instead of re-adding cruft to llh 2.6.recent Mar 20 12:34:13 koen: i agree, if 2.6.11 works for me, I'll use that Mar 20 12:34:44 schurig: if .11 doesn't work, feel free to patch .18 Mar 20 12:34:46 good point koen Mar 20 12:35:13 * schurig does the "make mrproper; make nfsroot; make coffee" dance Mar 20 12:43:49 koen: is there a reason there is no gcc-cross-initial_4.1.2.bb ? Mar 20 12:44:31 schurig: haven't had time to test it yet Mar 20 12:45:04 koen: okay, so I'll test gcc-cross-4.1.1 with my filesystem now :-) Mar 20 12:45:18 koen: oops, s/4.1.1/4.1.2/ Mar 20 12:48:11 heh Mar 20 12:48:24 someone mailed to the oe bugzilla ml Mar 20 12:54:35 what is the link in http://www.openembedded.org/wiki/ to get the list of recently changed pages? Mar 20 13:17:34 schurig: sadly there isn't :( Mar 20 13:20:40 psokolovsky: I need to merge that since no objections were made Mar 20 13:20:56 RP: yes, please do! ;-) Mar 20 13:21:34 RP: all sorts of log files now start taking ~18 GB :( Mar 20 13:21:58 RP: I can't reproduce it 100%, but it happens with half the packages Mar 20 13:22:19 (I suspect a bitbake bug, but it could be a screwed build) Mar 20 13:22:26 koen: There are new log files but they shouldn't take 18GB Mar 20 13:23:14 let me see what's inside the files Mar 20 13:23:50 koen: Also, some of the messages you're used to seeing with bitbake will now go to log files with trunk. We need this change for the UI(s) so we'll just have to iron out any bugs... Mar 20 13:24:24 I suspect tinderclient will appreicate the log files too... Mar 20 13:24:44 RP: I get a lot of '|' in the log files Mar 20 13:25:02 koen: Can you pastebin an example? Mar 20 13:25:44 koen: or an example package? Mar 20 13:26:03 koen@bitbake:/data/build/koen/OE/build/tmp/angstrom/work/c7x0-angstrom-linux-gnueabi/angstrom-x11-image-1.0-r5/temp$ du -hs *4.0K ipkg.conf5.5G log.do_rootfs.16773 Mar 20 13:26:20 and an excerpt: http://rafb.net/p/LDJrpk82.html Mar 20 13:27:14 koen: Its the output from wget Mar 20 13:28:35 or it looks a bit like it anyway. We need to work out what command is generating that in do_rootfs and clean up the output Mar 20 13:31:17 koen: maybe wget doesn't realize that it doesn't run on a tty and produces animated download graphics, e.g. a rotating bar or a progress bar Mar 20 13:39:35 schurig: could be Mar 20 13:39:48 schurig: could be bitbake prefixing messages with '|' Mar 20 13:39:56 you can use -q with wget not to output any messages Mar 20 13:40:16 koen: I think we need a proxy, but with some special use cases Mar 20 13:41:05 koen: e.g. if you don't have a connection it is better to show an old page before you show none :) Mar 20 14:01:22 Ifaistos: I was able ot get it all working properly in the single file. I do have a separate bb file for openpbx.org-perl which is a contrib package included in the openpbx distribution, as for sounds, they're places in openpbx.org-sounds as I intended, and pulled from cache as expected. Mar 20 14:04:28 * koen|away tries with wget -q Mar 20 14:11:06 Ifaistos: my reasoning being that if you want to use openpbx, you need the sounds. They aren't an optional build in that respect, unless (and this is why they aren't included in the main package) you intend to record all your own voice prompts for even the basic features such as conferencing and voicemail. Some people will record a few of their own prompts but it's important to have a base set of them. Mar 20 14:14:49 can I set a PREFERRED_VERSION on the bitbake command line along with the package name? Mar 20 14:15:50 RP: do you know something about the Linux stack usage? Mar 20 14:16:01 RP: e.g. a task does it have a kernel and a userspace stack Mar 20 14:16:15 RP: or is the usersspace stack used when entering the kernel? Mar 20 14:19:36 RP: it's definately bitbake/OE doing all the '||||||||' Mar 20 14:21:02 RP: http://ewi546.ewi.utwente.nl/tmp/huge.log Mar 20 14:21:12 a 6MB excerpt Mar 20 14:23:36 koen|away: nice log Mar 20 14:25:48 zecke: every task has it's own kernel- and userspace stack, yes Mar 20 14:25:57 zecke: I'm not sure about threads or tasklets Mar 20 14:26:27 schurig: one more question, don't run away Mar 20 14:26:38 * schurig stays put Mar 20 14:26:39 schurig: where is the stack inside the task_struct? Mar 20 14:26:50 schurig: is it within thread_info? Mar 20 14:27:09 schurig: I see that the kernel is able to map from stack to process (the gloryious current macro) Mar 20 14:29:03 zecke: http://kerneltrap.org/node/6399 Mar 20 14:29:49 zecke: it said there "kernel stack space of a process", and I assumed that this is a user process Mar 20 14:32:28 schurig: thanks. now I need to find the sysenter code Mar 20 14:33:28 zecke: I was afk, I see you have an answer now :) Mar 20 14:33:51 koen|away: ok, we'll have to find where that is coming from... Mar 20 14:34:11 zecke: for ARM? it's in arch/arm/kernel/entry-*.S Mar 20 14:36:00 any implementation is good enough :) Mar 20 14:38:17 koen|away: Do you see that in all logfiles or just rootfs? Mar 20 14:38:38 several logfiles grow to insane sizes Mar 20 14:38:54 let me see about keymaps Mar 20 14:39:10 koen|away: Only upon task failure? Mar 20 14:39:59 hard to say Mar 20 14:40:07 it fails eventually due to space issues Mar 20 14:40:24 There is that :) Mar 20 14:40:55 koen|away: Try turning off BBINCLUDELOGS Mar 20 14:41:26 In fact I'm 100% sure its BBINCLUDELOGS... Mar 20 14:41:39 * koen tries Mar 20 14:42:22 I think I'm just going to remove that code since we won't want it in the final UI code and I don't want to spend time trying to make it work since it will get removed anyway... Mar 20 14:43:05 fine by me Mar 20 14:43:18 if it fixes the problem i'm seeing :) Mar 20 14:44:20 It does mean bitbake won't quote logiles when it fails... Mar 20 14:46:06 03rpurdie * r800 10bitbake/lib/bb/build.py: build.py: Remove BBINCLUDELOGS for now since its breaking things. It will return in the tty UI handler eventually Mar 20 14:49:44 RP: BBINCLUDELOG = "no" still fills up log.do_rootfs Mar 20 14:50:03 koen: I think you'd have to remove the variable entirely but try svn up Mar 20 14:53:26 03koen 07org.oe.dev * r06f7b5ed... 10/ (1 packages/linux/linux-gta01_2.6.20.bb): linux-gta01: remove bogus CMD_CON that breaks the gsm modem Mar 20 15:05:15 03koen 07org.oe.dev * r7b0d8ae8... 10/ (1 packages/gsm/files/gsmd packages/gsm/libgsmd_svn.bb): libgsmd: hack around bug where the kernel console is pointed to the gsm modem Mar 20 15:09:22 RP: svn up seems to have solved it! Mar 20 15:09:24 RP: thanks! Mar 20 15:10:34 koen|away: good, thanks for testing. That was quite nasty :) Mar 20 15:34:08 hey, i am trying to build a distro for the simputer (http://simputer.org), i tried the getting started guide and have now built core-utils based on linux-2.6.18. Now i see the target has a bunch of ipkg files, how can you install this onto the device? Mar 20 15:42:34 hi all Mar 20 15:44:12 zecke: do you have some time for me? Mar 20 15:45:29 shres: ipkg install Mar 20 15:49:23 mr_nice:if you don't mind a delay on answers shoot Mar 20 15:49:32 mr_nice: I'm sure other people can help you as wel Mar 20 15:59:28 zecke: no, delay is not important. the question self is about the simpad-batterty driver. I have tried to change it to fit to the new ucb changes. but I get a null pointer. I don't understand how the simpad_battery struct works. here is the changed code (http://pastebin.ca/403693) and here is the original (http://pastebin.ca/403694). Mar 20 16:03:58 hi, all! Mar 20 16:04:19 is it officially ok to use bitbake-1.8 with OE.dev Angstrom distro? Mar 20 16:05:31 mr_nnice: my driver was unfinished but let me check Mar 20 16:07:15 mr_nice: search the old driver from Juergen Mar 20 16:07:41 zecke: from 2.4? Mar 20 16:07:50 mr_nice: yes, and show me the backtrace :) Mar 20 16:08:05 mr_nice: all the calculations will be taken from 2.4 driver Mar 20 16:08:26 zecke: ok. Mar 20 16:08:44 mr_nice: you will implement a "class driver" for the ucb and this driver will register to the battery interface Mar 20 16:09:08 so the skeleton is present, but hooking things together might need work, actually I think this is the class I stopped working on :} Mar 20 16:09:36 mr_nice: show me the backtrace and I might remember where I stopped Mar 20 16:10:06 Hello Mar 20 16:10:16 I thought of a nice feature to add Mar 20 16:10:25 in bitbake Mar 20 16:10:27 koen, i want to build a directory structure i can copy across to the machine, i know this sounds stupid but deploy/i686/ doesnt have ipkg binary in it. I will have to do a ipkg -d /target/dir * in my deploy/arm i suppose Mar 20 16:10:58 shres: tmp/staging/i686-linux/bin should have an ipkg-cl in it Mar 20 16:11:58 mr_nice: looking at the code I would say the call to ->ucb is wrong Mar 20 16:11:59 shres: also look at OE/classes/image.bbclass Mar 20 16:11:59 What if bitbake, after finishing a job, emitted a log of all failed packages, and why they failed (basically the 10 last lines from the output for each package)? Mar 20 16:12:20 Building "world" would be made somewhat easier that way Mar 20 16:12:38 mr_nice: no it is not. the battery_class is inside the simpad_battery_class Mar 20 16:12:56 B_Lizzard: It already does that as it goes along, its just a data collection issue Mar 20 16:13:11 Aye Mar 20 16:13:12 slapin_nb: you may use bitbake 1.8.0 with angstrom Mar 20 16:14:25 RP: what's the main diff between bb 1.6 and 1.8? Is somewhere a docu? Or just "svn log"? Mar 20 16:14:25 It would just be an issue of getting a failed package's name, and adding the last 10 lines from the output. Mar 20 16:14:28 zecke: http://pastebin.ca/403726 Mar 20 16:14:52 And adding that to a logfile whenever such a failed package occured Mar 20 16:15:05 I have two .bb that will not compile, so i added a mask to not use them Mar 20 16:15:06 BBMASK = "/home/zero/oe/org.openembedded.dev/packages/file/file-native_4.20*.bb /home/zero/oe/org.openembedded.dev/packages/gcc/gcc-cross-initial_4.1.1.bb" Mar 20 16:15:16 mr_nice: that is easy?! Mar 20 16:15:27 but it doesn't work, can someone say why I'm dumb? I can mask one ofr the other, but not both Mar 20 16:15:37 s/ofr/or Mar 20 16:16:28 schurig: cat Changelog and/or what was on the mailing list about the multithreading a while back Mar 20 16:17:00 schurig: Basically, the core was rewritten to compute the dependency chain in advance and allow multiple tasks to run at once Mar 20 16:18:18 RP: any news on the 'rebuild dependant tasks' bit? Mar 20 16:18:19 mr_nice: which battery patch did you use? Mar 20 16:19:26 koen|away: No, I just keep forgetting to look at it. I will write that down this time... Mar 20 16:19:51 zecke: I was using your driver and it was broken because of changes on the ucb1x00 stuff. Then I was searching on the way the keyboard driver was registering to ucb and I tried to adapt that to your driver Mar 20 16:20:31 RP: no problem, the indt guys asked for it so to ease their testing Mar 20 16:20:35 zecke: http://pastebin.ca/403694 is the original Mar 20 16:20:41 mr_nice: the adaption must be right, but I can't find hwmon/battery/battery.c in mainline, so what additional patch do you use? I found a copy at hh.org Mar 20 16:21:06 koen: The code in trunk should be able to support it easily enough Mar 20 16:21:18 mr_nice: BUG: at drivers/hwmon/battery/battery.c:129 battery_class_register() Mar 20 16:21:18 schurig: some infos about what's new can also be found in http://linuxtogo.org/~mickeyl/tools/FOSDEM2007.pdf Mar 20 16:21:19 RP: that's why I asked :) Mar 20 16:21:34 mr_nice: going to 129 leads to a line like Mar 20 16:21:40 mr_nice: WARN_ON(!bat->name); Mar 20 16:22:29 mr_nice: http://handhelds.org/cgi-bin/cvsweb.cgi/linux/kernel26/include/linux/battery.h?rev=1.11&content-type=text/x-cvsweb-markup Mar 20 16:22:41 zecke: mom Mar 20 16:22:58 mr_nice: you at least need to set name with some string Mar 20 16:23:07 mr_nice: static struct simpad_battery_t simpad_battery = { .battery = { Mar 20 16:23:14 .name = "simpad-battery" Mar 20 16:23:24 mr_nice: or better check the other drivers to have a common naming Mar 20 16:24:02 mickey|lunch: nice presentation! Mar 20 16:24:14 thanks. unfortunately i couldn't talk about it :/ Mar 20 16:24:28 let me steal some stuff for CELF :) Mar 20 16:24:29 we should get this into the next OOO newsletter Mar 20 16:24:32 sure, go ahead Mar 20 16:24:36 but credit me! :) Mar 20 16:24:38 s/me/us/ Mar 20 16:24:54 koen: credit mickeyl as "Chris Larson and others" Mar 20 16:25:04 heh Mar 20 16:25:04 yeah Mar 20 16:25:12 it's always like that :) Mar 20 16:25:20 :} Mar 20 16:26:03 mr_nice: you also need to set class_dev to something and optionally id Mar 20 16:27:11 mickey|lunch: do you have the 'source' of the pdf available? Mar 20 16:27:22 zecke: this is the batery.c file. I can't remember from which version of the hh.org kernel it was taken from. Mar 20 16:27:23 koen: if you can handle *.ppt Mar 20 16:27:30 mr_nice: http://handhelds.org/cgi-bin/cvsweb.cgi/linux/kernel26/arch/arm/mach-pxa/hx4700/hx4700_battery.c?rev=1.7&content-type=text/x-cvsweb-markup as an example Mar 20 16:27:45 mr_nice: you we will need to set name and this backtrace should go away Mar 20 16:27:45 i can probably export it to .txt Mar 20 16:28:15 mr_nice: any progress on submitting the patches to rmk? can one congratulate already? Mar 20 16:28:36 koen: .rtf underway Mar 20 16:28:41 mickey|lunch: thanks! Mar 20 16:28:58 you can add ProGear into webpads ;D Mar 20 16:29:23 zecke: ok I will try it. Sorry I didn't hand it in atm :/ I will do it this weekend! Mar 20 16:29:47 zecke: thx. I will try it out now. Mar 20 16:29:58 is there something I can add to my .bb file to keep generated binaries from being stripped? Mar 20 16:30:00 mr_nice: name is mandantory, id is optional Mar 20 16:30:30 mr_nice: I'm glad you spend your energy on the simpad :) Mar 20 16:31:13 zecke: I love my simpad :). hopefully you will get yours repaired :D Mar 20 16:31:31 I see that unstripped binaries are kept in .debug folders, but if I could keep that step from happening for my development builds that would help Mar 20 16:32:23 mr_nice: finding it is a issue :) Mar 20 16:32:24 mickey|lunch: page 16 has typo Mar 20 16:32:41 xuumbi: .debug folders should end up in -dbg packages Mar 20 16:33:04 I'm trying to set up a cvs checkout for a package, but bitbake keeps saying the version is not available. Is there a good way to debug the cvs checkout? Mar 20 16:33:04 xuumbi: and -dbg packages should not be installed into a rootfs by default Mar 20 16:34:45 mickey|lunch: and bugsquashing are first weekend not last Mar 20 16:35:07 oops :D Mar 20 16:35:17 that's why I am always alone Mar 20 16:35:19 mickey|lunch: very nice presentation Mar 20 16:35:26 thanks :) Mar 20 16:35:33 a pity we didn't get a slot for that Mar 20 16:36:07 mickey|lunch: other time Mar 20 16:36:30 "Introduction to the Gnome Mobile and Embedded Initiative" Mar 20 16:36:33 mickey|lunch: what do you think about creating 'OE History' page in wiki? how it starts, who joined, which team etc Mar 20 16:36:39 seems like they are finally going to launch it at ELC Mar 20 16:37:54 hrw: sounds good in principle Mar 20 16:38:52 A history page would be good. I was thinking bitbake should have something about its history too... Mar 20 16:39:15 "What was changed, when and why" Mar 20 16:39:46 yeah Mar 20 16:39:59 let's do it asap... as long as we can remember the reasons Mar 20 16:40:24 http://www.openembedded.org/wiki/History will be ok? Mar 20 16:40:51 sure Mar 20 16:42:50 mickey|lunch: There were some emails I wrote that should cover the bits of bitbake I've had a part in. I can't really say what happened before 1.4 though... Mar 20 16:43:17 when i can breathe again, i try to remember earlier stuff Mar 20 16:43:46 mickey|lunch: I keep hoping to be able to breathe but it doesn't work out like that :/ Mar 20 16:50:42 mickey|lunch: http://www.openembedded.org/history Mar 20 16:52:10 good! Mar 20 16:52:40 the veterans can fill in the early years Mar 20 16:52:55 on Ania's laptop I have some more entries in other presentation Mar 20 16:56:38 So, does anyone know how to debug the CVS file fetcher? -DDD doesn't provide anything useful. My SRC_URI is "cvs://anonymous@cvs.savannah.gnu.org/sources/classpath;module=classpath" and I can check out that module manually, but bitbake says my version is not available. I'd like to check out the module with todays date, 20070320, so I set SRCDATE to that. I read http://bitbake.berlios.de/manual/ch03s03.html, but I still don't know what I'm doing exactly. Mar 20 16:59:32 I would tweak cvs fetcher in bitbake to add some debugs Mar 20 17:00:08 zecke: YUHU!! it works many thx! Mar 20 17:01:25 zecke: do you remeber if the driver from Juergen included the values for T-Sinus pads? Mar 20 17:03:37 anybody remembers which kernel/patch contain PXA capture device driver? Mar 20 17:03:50 pxa capture? Mar 20 17:03:55 like the camera interface? Mar 20 17:04:02 does anyone know what is the overhead of a package when installed ? i mean the added space need it store list etc etc Mar 20 17:05:31 I am trying to figure out the cut-off point of creating a package, to see what is the minimum package size that makes sence Mar 20 17:06:07 i.e putting a 100 bytes file in package mean 1K in package overhead Mar 20 17:06:20 in the flash of the target Mar 20 17:06:36 koen, yes Mar 20 17:07:22 I remember this pxa_capture.c floating around but I'd like to know where the latest version is Mar 20 17:07:25 slapin_nb: ask on #openezx, they updated the driver to 2.6.16 iirc Mar 20 17:11:15 cu Mar 20 17:22:20 mr_nice|scummvm: no, I don't remember Mar 20 17:23:10 any autofoo experts around? Mar 20 17:23:22 how's the proper deal with foo-config files? Mar 20 17:23:32 i.e. checking for existance and taking --cflags and --libs into account Mar 20 17:23:36 <--- qmake weenie Mar 20 17:23:51 mickeyl: not at all, fix the upstream to use pkgconfig :) Mar 20 17:23:57 ah no Mar 20 17:24:05 i don't want to do that for so many projects :) Mar 20 17:24:44 mickeyl: check libsdl-ttf, -image I have fixed these tests there Mar 20 17:24:58 ok Mar 20 17:34:42 heh, nm Mar 20 17:34:53 the library in question already has pkgconfig support Mar 20 17:36:46 mickeyl: PGK_CHECK_MODULES :) Mar 20 17:36:55 ya Mar 20 17:36:58 i know at least about that Mar 20 17:37:00 PKG even Mar 20 17:37:00 and used it Mar 20 17:37:10 all autoconf stuff in OM2007 is mine Mar 20 17:37:14 :D Mar 20 17:37:29 mickeyl: the error message in chordmaster is bogus Mar 20 17:37:38 mickeyl: the default one is good enough :) Mar 20 17:37:47 mickeyl: then use [] around litlerals :) Mar 20 17:38:01 AC_PREQ(2.53) gets AC_PREQ([2.53]) Mar 20 17:38:16 mickeyl: why do you need the maintainer mode? :) Mar 20 17:38:43 mickeyl: BTW: I have a small diff that adds () around the *_get_type() calls Mar 20 17:39:18 no idea actually Mar 20 17:39:29 lots of that was taken from ohand autotools usage :) Mar 20 17:39:43 let the patches roll in :) Mar 20 17:40:07 probably at 19:30 or 20h I need to finish chapter four first... :} Mar 20 17:40:55 cool Mar 20 17:45:10 need some help with dynamic packaging Mar 20 17:45:12 do_split_packages(d, asterisk_libdir, '^app(.*)\.so$', 'asterisk-app%s', \ Mar 20 17:45:12 'Asterisk application for %s', extra_depends='') Mar 20 17:45:25 this creates the packages fine.... but Mar 20 17:45:40 instead of using arch is using machine Mar 20 17:45:53 Packaged contents of asterisk-res-musiconhold into /proj/oplinux-0.2/tmp/libc/deploy/ipk/dht-walnut/asterisk-res-musiconhold_1.2.16-r1_dht-walnut.ipk Mar 20 17:46:35 is this correct ? Mar 20 17:46:52 if asterisk is machine specific, it might be Mar 20 17:47:08 koen: no its not Mar 20 17:48:19 koen : opsis-asterisk_1.2.16-r0_ppc405.ipk Mar 20 17:49:44 koen : should the created packages be named exactly like .bb ? now the bb is opsis-asterisk and the packages are asterisk- Mar 20 17:51:27 Ifaistos: dunno Mar 20 17:53:26 03koen 07org.oe.dev * r159e0203... 10/ (1 packages/tzdata/tzdata_2007c.bb): tzdata: also rconflict with timezones Mar 20 18:10:54 mickey|sports: e17 is switching to pkgconfig :) Mar 20 18:11:14 Hi Mar 20 18:11:45 is there a way to run an openembedded generated system from qemu ? Mar 20 18:12:20 i've been trying with qemu-system-arm and fails with "unimplemented cp15 register read" Mar 20 18:19:41 RP: How hard is it to add a driver to the leds subsystem, where controlling the leds are done by writing to a pic/avr connected to an uart? Mar 20 18:20:00 Do you always try in the real hardware? no emulation? Mar 20 18:20:15 saladino: I do. Mar 20 18:20:38 saladino: http://folks.o-hand.com/richard/qemu Mar 20 18:21:51 humm Mar 20 18:22:01 seems that it can't be run as is Mar 20 18:22:13 some modifications are needed, i'll take a look Mar 20 18:22:14 thx Mar 20 18:22:32 NAiL : Why don't you do it from userspace ? its much much easier Mar 20 18:23:11 Ifaistos: yeah, it probably is. Either way I can't do it :-P Mar 20 18:25:10 NAiL : Does the led "controller" have a full protocol or you just send it a byte where its bit defines the state of a led ? Mar 20 18:26:41 send a byte. the pic then decides what actions to take. Mar 20 18:30:05 NAiL : its easy then, i'll post a couple of scripts you can use Mar 20 18:31:40 Ifaistos: the uart is two-way. It also handles the buttons. Mar 20 18:32:32 NAiL : do you poll it or it sends a byte when someone presses a button ? Mar 20 18:33:00 I haven't completely deciphered the code, but I think it sends a byte Mar 20 18:33:47 NAiL : hmmm that complicates stuff a bit Mar 20 18:34:33 NAiL : anyway have to go for dinner now Mar 20 18:43:09 hi Mar 20 19:02:25 hmm.. when I have a uboot multi-file image, where first image is my kenrel and the 2nd one is the ramfs, how can I point the kernel to the ramfs? Mar 20 19:02:32 somehow I fail to figure that out Mar 20 19:06:11 Jin^eLD, hopefully, the file is loaded in one of your /dev/mtdblockN Mar 20 19:06:52 HopsNBarley: aha.. so I have to pass a boot arg like root=/dev/mydev to the kernel? Mar 20 19:07:20 i forget the syntax, but i think it's initramfs=/dev/mtdblockX (or initrd, depends on the type of RD) Mar 20 19:07:52 I guess what I have to figure out is where the image got loaded to Mar 20 19:07:59 can I find this out from uboot? Mar 20 19:08:18 scratch that - iniramfs is located within the kernel image itself, so it would be initrd. sorry Mar 20 19:08:43 mhm Mar 20 19:08:47 well Mar 20 19:09:03 I followed some howto that said that I could separately add a cpio.gz archive that would then be used as an initrd file Mar 20 19:09:04 how are the images getting down? via tftp? Mar 20 19:09:20 and overload the kernel's ramfs Mar 20 19:09:39 the kernel's mtd map better agree with where you are putting things (-; Mar 20 19:09:48 yes, I packged the kernel and the cpio.gz archive which is supposed ot be the ramfs into one uimage file, and that is being fetched via tftp Mar 20 19:10:28 did the cpio get put in the kernel build? If so, that's an initramfs, and kernel should automatically use it. Mar 20 19:10:39 no, it did not Mar 20 19:10:59 that's what I actually am trying to do, I do not want to touch the kernel build, but I want to add the ramfs afterwards via uboot Mar 20 19:11:13 what's the difference in behavior between RRECOMMENDS and RSUGGESTS ? Mar 20 19:11:47 HopsNBarley: found information about how to Initialize initramfs from an External File Mar 20 19:12:02 ah, okay. so the initrd will reside at a particular location. the kernel has a map of the flash which is compiled it. it has to agree with where the flash is loaded. Mar 20 19:12:26 the map will also give you the mtd partition info, so you'll know which N to use for /dev/mtdblockN Mar 20 19:12:36 where do I find the map? Mar 20 19:13:17 down in kernel/drivers/mtd/(something) some times physmap.c, sometimes other places. depends on the patch/kernel. Mar 20 19:13:26 uh.. Mar 20 19:13:34 not helping? Mar 20 19:14:17 not sure yet :) Mar 20 19:15:56 yes, that is true Mar 20 19:15:58 oops wrong window Mar 20 19:16:08 HopsNBarley: no, the sources with the maps are not really helping.. Mar 20 19:16:25 so the 2nd image, that is inside a uimage, will be loaded automatically to this /dev/mtd thing? Mar 20 19:16:35 do I undersatnd that correctly? Mar 20 19:16:55 I thought I'd have to tell uboot about it or something.. Mar 20 19:17:04 when I have two separate images I am supposed to give both addresses to bootm Mar 20 19:17:16 but I could not find any info on how to boot with one multiimage Mar 20 19:20:11 Jin^eLD: not sure if you figured this out yet, but you pass an initramfs to the kernel using the initrd mechanism. Mar 20 19:20:13 the mtd is linear flash space. that's why the map is important: it says where each partition starts and stops. however, they can overlap. Mar 20 19:20:45 that's how one flash file can update both kernel partition and rootfs partition. Mar 20 19:21:03 but the padding between them and the flash map in the kernel must all be perfectly aligned. Mar 20 19:24:18 cbrake: yes that is how I understood it, but right now I am failing to figure out how to tell it where the initrd is located - it is somewhere inside my multi-file uimage Mar 20 19:25:08 Jin^eLD: hmm, hopefully u-boot loads into RAM at some location, and then the kernel needs to know this RAM address Mar 20 19:25:37 cbrake: I guess so :) the question is - how to figure out this address? Mar 20 19:25:55 uboot only tells me Mar 20 19:25:56 Image 0: 1496774 Bytes = 1.4 MB Mar 20 19:25:56 Image 1: 7625112 Bytes = 7.3 MB Mar 20 19:26:27 Jin^eLD: that is neat -- I did not know uImage could contain multiple images. Mar 20 19:26:45 I discovered that today :) Mar 20 19:27:13 http://www.denx.de/wiki/view/DULG/CombiningKernelAndRamdisk Mar 20 19:27:22 you can produce that with mkimage Mar 20 19:28:44 I also figured out that if you have two uimages, you can tell bootm where the kenrel and the ramdisk are Mar 20 19:28:58 there it is easy - since you load both images yourself you know where they are going Mar 20 19:29:06 but with this multithing.. damn Mar 20 19:29:40 Jin^eLD: I guess start groking the uboot code. I don't see any clues on the web site. Mar 20 19:29:48 me neither.. Mar 20 19:45:26 03nail 07org.oe.dev * rf5036628... 10/ (1 conf/distro/foonas.conf): foonas: Update config, moving more stuff away from foonas-image.bb Mar 20 19:45:31 03nail 07org.oe.dev * r93e745b0... 10/ (1 conf/machine/n2100.conf conf/machine/turbostation.conf): turbostation,n2100: Update machine definitions Mar 20 19:45:36 03nail 07org.oe.dev * ra43f5ce9... 10/ (5 files in 2 dirs): n2100, turbostation, linkstation: Move more stuff away from foonas-image to machine configs Mar 20 19:45:42 03nail 07org.oe.dev * r67907ca0... 10/ (5 files in 2 dirs): n2100, turbostation, linkstation: Move machine-stuff out of classes Mar 20 19:45:46 03nail 07org.oe.dev * r0ec41d71... 10/ (1 packages/images/foonas-image.bb): foonas-image: Move preprocessing away Mar 20 19:56:22 Good evening. Mar 20 20:07:36 hey sirfred Mar 20 20:33:45 re Mar 20 20:51:49 koen, when are you traveling to san jose? Mar 20 20:53:21 RP: is any of the o-hand Gtk freaks online? Mar 20 21:05:46 03justinp 07org.oe.dev * r6cb06ec6... 10/ (3 files in 2 dirs): angstrom-e: add e-image and tasks for angstrom Mar 20 21:38:58 03koen 07org.oe.dev * r35b8bd70... 10/ (3 files in 2 dirs): disapproval of revision '6cb06ec6fefb58a8630562ccfdd7e9b156c05f3d' Mar 20 21:42:25 hmm what did koen remove... Mar 20 21:43:08 zecke|moko: I think it was JustinP's e-image stuff Mar 20 21:44:07 hmm Mar 20 21:44:33 zecke|moko: it just made the ML Mar 20 21:45:32 sounds reasonable :} Mar 20 21:47:32 zecke|moko: so is enlightenment now considdered depricated? Mar 20 21:47:55 hvontres|poodle: hehe Mar 20 21:48:08 hvontres|poodle: I don't know koen's reasoning but gpe-bootsplash is deprecated Mar 20 21:48:16 (according to koen's mail) Mar 20 21:50:43 zecke|moko: ok..I thought I didn't see it in there to begin with..but it's in task-angstrom-e.bb Mar 20 21:55:06 hi zecke|moko Mar 20 21:57:00 pb_: hey Mar 20 21:57:16 pb_: I lost my soul, http://page.mi.fu-berlin.de/~freyther/FeedReader.png Mar 20 21:58:00 pb_: I'm totally g_* and already stumble across bugzilla.gnome.org and have a hard copy of all bug reports near my pillow Mar 20 21:58:05 pb_: but how are you? Mar 20 23:14:34 zecke: crumbs, have you fallen victim to some kind of mind control? Mar 20 23:15:17 pb_: I stopped being stoned, this could have a small influence :) Mar 20 23:15:27 zecke: I am fairly well, I guess. been spending a lot of time redecorating bits of my house and that kind of thing. Mar 20 23:15:29 zecke: heh Mar 20 23:15:39 pb_: At least I want to write the same app for Qtopia, so I'm not completely lost yet Mar 20 23:15:45 :-} Mar 20 23:16:05 ;) Mar 20 23:16:18 whoops Mar 20 23:16:34 I wish xchat wouldn't put the close button there. Mar 20 23:16:41 hehe, I thought your wife had taken over the control :) Mar 20 23:16:50 heh Mar 20 23:16:55 no, she has gone to bed :-} Mar 20 23:18:06 Here everyone smaller than me is sleeping as well :-) Mar 20 23:18:12 hehe Mar 21 00:25:04 NAiL: The LED subsystem itself doesn't care. If you can write a driver, you can connect it to the subsystem, even if you have to use a wait queue Mar 21 00:25:23 ah Mar 21 00:53:34 * RP -> sleep **** BEGIN LOGGING AT Wed Mar 21 02:24:18 2007 **** ENDING LOGGING AT Wed Mar 21 02:59:56 2007