**** BEGIN LOGGING AT Tue Nov 30 02:59:57 2010 Nov 30 03:10:36 khem, gcc-cross-sdk still fails for me with micro-uclibc: http://tinderbox.openembedded.org/packages/1149123/ Nov 30 04:00:13 hey guys, i did a bitbake -c clean and then a bitbake minimal-install, but i can't seem to get the kernel modules and kernel to go in to my images directory Nov 30 04:00:18 what step am i missing? Nov 30 04:31:46 hi good mornging Nov 30 04:31:48 my ${OE_HOME}/tmp folder doesn't have a staging folder why is it so ? Nov 30 04:32:35 because it was renamed. see the sysroots dir instead Nov 30 04:37:40 ok thanks Nov 30 04:39:14 nm Nov 30 07:22:45 gm Nov 30 07:26:53 03Martin Jansa  07master * r61e98efe0b 10openembedded.git/recipes/shr/ (e-wm-theme-illume-neo_git.bb elementary-theme-neo_git.bb): Nov 30 07:26:53 e-wm-theme-illume-neo,elementary-theme-neo: bump SRCREV Nov 30 07:26:53 Signed-off-by: Martin Jansa Nov 30 07:27:04 03Martin Jansa  07master * r434d85dca9 10openembedded.git/recipes/evopedia/evopedia_git.bb: Nov 30 07:27:04 evopedia: bump SRCREV Nov 30 07:27:04 Signed-off-by: Martin Jansa Nov 30 07:45:22 Has anyone been able to compile the wvdial, wvstreams, ppp packages to use the USB modem on the beagleboard ? Nov 30 07:48:27 screwgoth: wvdial is broken on ARM i've heard Nov 30 07:48:34 screwgoth: just use chat and ppp directly Nov 30 07:48:50 dm8tbr: Really ? Ok... Nov 30 07:49:08 dm8tbr: I am not aware about chat .. Is it a package similar to wvdial ? Nov 30 07:49:56 it's on a lower level, less fancy shmancy Nov 30 07:50:30 everybody used ppp+chat before such things like wvdial or god forbid UI software came along ;) Nov 30 07:50:58 so I'm sure using the search engine of your least distrust you should be able to find some instructions for what you plan to do Nov 30 07:52:05 yup.... I'm already there.... thanks for the heads-up.... Nov 30 07:52:26 I'll also dig out and see if wvdial is really broken for ARM ..... thanks for the tips Nov 30 07:55:21 it sure as hell didn't want to build for me back then and I remember koen saying it is broken too Nov 30 08:00:03 i have a mips target board with me , i want to build a nano for it , what should be my distribution for that ..? Nov 30 08:00:58 i am currently using DISTRIBUTION="qemumips" in local.conf Nov 30 08:01:03 will this work ..? Nov 30 08:10:25 good morning Nov 30 08:11:28 Is the platform module not available for python in OE? Can't find anything that provides it. Nov 30 08:38:05 * Jay7 is races-catcher... Nov 30 08:39:16 http://tinderbox.openembedded.net/public/logs/task/11999070.txt Nov 30 08:42:44 03Frans Meulenbroeks  07master * r9d3d91887f 10openembedded.git/recipes/php/php_5.2.13.bb: Nov 30 08:42:44 php: removed --enable-memory-limit configure argument; does not exist any moe Nov 30 08:42:44 Signed-off-by: Frans Meulenbroeks Nov 30 08:42:52 03Frans Meulenbroeks  07master * rb0d9ce644c 10openembedded.git/recipes/php/php_5.2.13.bb: Nov 30 08:42:52 php: reinstated --without-iconv, this seems to generate configure issues Nov 30 08:42:52 issues are wrt native paths being used. Nov 30 08:42:52 e.g. get that fork is not found Nov 30 08:42:52 adding --with-iconv-dir did not help Nov 30 09:08:54 we need antispam plugins in wiki Nov 30 09:09:32 we got few hundred spam accounts created Nov 30 09:09:40 removing them one by one is waste of time Nov 30 09:10:13 same with removing spam pages Nov 30 09:13:11 hrw: I thought it was already done Nov 30 09:14:54 then I hope that someone know how to use it Nov 30 09:23:24 can we block the ip they are from ? Nov 30 09:23:40 or we need to have confirmation or captcha or so when registering Nov 30 09:23:48 * eFfeM is not a wiki expert Nov 30 09:30:36 captcha + disable anonymous edit (if not done already) Nov 30 09:32:22 Jay7: they do user edits Nov 30 09:32:40 it is not 'how to prevent from spam' Nov 30 09:32:50 morning Nov 30 09:32:51 right now we can disable new user registration Nov 30 09:32:53 we need 'how to remove spam in one run' Nov 30 09:33:27 such as "select that 200 users and block them" + "remove all edits of those users" Nov 30 09:33:46 Jay7: who will register real users? Nov 30 09:33:49 well.. not right now, but when wiki admin will be found and awaken :) Nov 30 09:34:05 what works well on wordpress is the speed check, humans just cant type as fast as bots Nov 30 09:34:34 XorA: it will be fixed soon ;) Nov 30 09:35:46 hrw: I mean quick temporary solution Nov 30 09:36:39 well.. anyway wiki (and server under wiki) admin is needed Nov 30 09:37:09 XorA: in wordpress I use NoSpamNX Nov 30 09:37:29 hrw: cant remember what I use, but it works well Nov 30 09:58:55 kergoth: progressbar should be installed with setup.py, right? seems it's not and even ply dependency is not checked in master HEAD Nov 30 10:06:38 kergoth: ah that progressbar.py in bitbake code is bundled version of http://pypi.python.org/pypi/progressbar/2.2 ? can I use system version? Nov 30 10:19:49 we need http://www.mediawiki.org/wiki/Extension:Nuke installed Nov 30 10:23:40 hrw why? Nov 30 10:24:09 hrw: cunning plan to take over world Nov 30 10:24:41 XorA, I wonder :D Nov 30 10:25:01 we need its companion though...Pave Nov 30 10:25:01 ka6sox: mass remove spam Nov 30 10:25:26 okay...Let me get it installed when I wake up. Nov 30 10:25:30 ka6sox: I can take care of oe wiki spam but not by removing user by user, page by page Nov 30 10:25:32 2am here now. Nov 30 10:25:48 ka6sox: I am surprised that you are still awake Nov 30 10:26:11 was working on getting the temp sensor working in the kernel for the Pre2 Nov 30 10:26:29 since any day I suspect we will see 1.6ghz phones. Nov 30 10:27:52 we will Nov 30 10:28:05 and eventually phone will keep us warm in winter Nov 30 10:28:07 I've already seen the patch. Nov 30 10:28:25 who knows..maybe even tommorrow. Nov 30 10:28:32 ka6sox: pre2 has 3730/1GHz Nov 30 10:28:33 ? Nov 30 10:28:43 3630/1Ghz Nov 30 10:29:09 yes.. 1.6GHz OC users... Nov 30 10:29:31 n900 has 3430/600MHz and people even clocks to 1.0-1.1GHz Nov 30 10:30:12 hrw, I've seen one Pre going 1.4Ghz (but for how long I don't know!) Nov 30 10:31:30 I still have a problem with selection of next phone after n900 Nov 30 10:32:05 ok. bbsoon Nov 30 10:43:32 usr/klibc/syscalls/SYSCALLS.i:20: Could not parse input: "static inline void crash_setup_regs(struct pt_regs *newregs," Nov 30 10:43:45 hehe.. seems this is klibc issue on mips.. Nov 30 10:48:58 morning bluelightning Nov 30 10:52:09 hi ka6sox Nov 30 10:58:58 well.. kexec patch seems broke klibc Nov 30 10:59:38 this one: klibc_kexecsyscall.patch Nov 30 11:25:04 oh..well, this one was around since initial klibc commit Nov 30 11:25:44 possibly the patching was done in an arm-egocentric way ;) Nov 30 11:26:04 something is grabbing crash_setup_regs() definition from asm/kexec.h and store it in SYSCALLS.i Nov 30 11:26:13 then parser fails on this Nov 30 11:26:32 I suspect that something is cpp Nov 30 11:27:55 yes, it is Nov 30 11:28:39 strange, why this not happens on arm?.. Nov 30 11:32:40 qemu compiling w/o parallel make is so looong Nov 30 11:45:04 my angstrom-build uses busybox. is there a bitbake command to find out which recipe gives me that? Nov 30 11:47:45 gives you what? you mean recipe version? Nov 30 11:48:01 if you don't want to use busybox, just look at conf/distro Nov 30 11:48:14 it's not defined in recipe, it's distro prefference Nov 30 11:49:19 ynezz: ok. I'd like to use it, but I need to figure out why I don't get the ulimit command Nov 30 11:56:13 eek: remote: aborting due to possible repository corruption on the remote side. Nov 30 11:56:26 second time this went ok... Nov 30 11:56:36 gm btw Nov 30 12:03:57 tasslehoff: it will be probably some config option Nov 30 12:04:10 morning likewise Nov 30 12:11:27 ynezz: yep Nov 30 12:14:33 likewise, what repo? Nov 30 12:16:43 is there something similar to "bitbake virtual/kernel -c menuconfig" for busybox? Nov 30 12:42:06 tasslehoff: what do you want from it? Nov 30 12:42:15 menuconfig exists only for kernels Nov 30 12:43:57 hm.. Nov 30 12:44:17 ant_work: klibc builds fine w/o #include Nov 30 12:44:30 and I've found why it was ok on arm Nov 30 12:44:50 Jay7: just trying to figure out why ulimit is missing. Nov 30 12:44:50 tosa's 2.6.24 have no that function definition Nov 30 12:45:09 tasslehoff: check busybox recipes Nov 30 12:45:19 Jay7: busybox itself has a "make menuconfig" where you can configure what you want supported. Nov 30 12:45:21 I bet for configure options :) Nov 30 12:45:37 Jay7: yeah Nov 30 12:52:51 but busybox seems to have defconfig, like the kernel does, so maybe menuconfig'ing from bitbake could've been done Nov 30 12:54:07 Hi, I'm trying to bitbake angstrom-gnome-image.bb and I'm getting an error about kissing dependancies for various gtk-themes Nov 30 12:54:30 tasslehoff: just modify do_make to make menuconfig Nov 30 12:54:48 Out of these I found that only gtk-engine-clearlooks has a recipes ... there are no recipes for gtk-engine-thinince ,etc ...how is that possible ? Nov 30 12:54:54 tasslehoff: or run.do_compile and change oe_runmake line to make menuconfig one, or something like that Nov 30 12:57:26 screwgoth: because no-one wrote them I guess Nov 30 12:57:41 ynezz: that could work. Nov 30 12:57:50 XorA: Ok ... but then why are they past of one of the task recipes ? Nov 30 12:58:26 screwgoth: see if someone removed them without checking tasks? Nov 30 12:58:43 I've build that image about 1-2 weeks ago and it was fine Nov 30 12:59:21 Jay7: so the patch is arm-only nowadays? Nov 30 12:59:39 ant_work: I suspect this patch will broke anything a bit later :) Nov 30 13:00:21 ant_work: I'm trying to build with fixed patch version (w/o include line Nov 30 13:00:43 pls check with latest klibc-git and purged includes Nov 30 13:00:49 tosa and ben-nanonote was fine Nov 30 13:01:01 ok Nov 30 13:01:02 XorA: Hmm ... maybe I am mistaken ... I think the gtk-engines recipe should build all the other gtk-engines ... Thanks for the inputs Nov 30 13:01:13 Hi Everyone! I am trying to setup OE on fedora 10.. after seting up OE git using oebb.sh (following wiki), i am trying to exectute bitbake nano which gives me error on ::: FATAL: coreutils-native-8.5: ftp://ftp.gnu.org/gnu/coreutils/coreutils-8.5.tar.gz cannot check archive integrity ERROR: Error in executing python function in: /home/rathi/OE/sources/openembedded/recipes/coreutils/coreutils-native_8.5.bb Nov 30 13:01:20 How to fix this? Nov 30 13:01:50 install md5sum Nov 30 13:01:52 ant_work: qemux86 is fine too Nov 30 13:02:05 Thank you XorA, OE rocks! Nov 30 13:02:07 I think Nov 30 13:02:35 been a long while since I used F10 :-D Nov 30 13:02:47 ant_work: what is better to do now? write to ML or just push one-line change? :) Nov 30 13:03:25 have you built a 2.2x kernel without the patcgh? Nov 30 13:03:36 2.62x Nov 30 13:03:54 if yes (tosa) please remove the patch Nov 30 13:03:55 2.6.24 for tosa, 2.6.35 for qemux86, 2.6.36 for nanonote Nov 30 13:04:08 ok then the patch is obsolete/dangerous Nov 30 13:04:11 patch is needed Nov 30 13:04:20 because of kexec_load Nov 30 13:04:29 ah, wait Nov 30 13:04:31 @ XorA: Sorry buddy, md5sum util is already present Nov 30 13:04:37 but include before is not needed seems Nov 30 13:04:57 navaltiger: maybe your python is too old? Nov 30 13:05:00 ant_work: check recipes/klibc/klibc-1.5.20/klibc_kexecsyscall.patch Nov 30 13:05:10 navaltiger: just re-read the error message Nov 30 13:05:27 Jay7: I see, the conflicting header Nov 30 13:05:37 ant_work: yes Nov 30 13:05:52 I'll check again with tosa and 2.6.36 Nov 30 13:06:06 if it will be ok, I'll commit to master Nov 30 13:06:43 seems we (kexecboot) are only users of klibc + that kexec patch Nov 30 13:07:08 well, yes Nov 30 13:07:09 kexec-tools-klibc-2.0.2/syscall.patch Nov 30 13:07:09 NOTE: package coreutils-native-8.5-r5.0: task do_fetch: Started NOTE: The checksums for '/home/navaltiger/OE/sources/downloads/coreutils-8.5.tar.gz' did not match. Expected MD5: 'c1ffe586d001e87d66cd80c4536ee823' and Got: '1a9a22a2bae302c710520d42c6c730f5' Expected SHA256: 'dd243e2ac7d0ed203b170ecfb8299643779e7501fec2b45ae1cea8a450fa8ca0' and Got: '8abaffdfbd8552664101721ee63975296ef7a9b3dce4721a56316baec00693b8' NOTE: Your checksum Nov 30 13:07:23 Jay7: try to remove both :p Nov 30 13:07:52 Do i really need Psyco JIT Compiler??? Nov 30 13:08:00 navaltiger: no you dont Nov 30 13:08:15 navaltiger: try to remove /home/navaltiger/OE/sources/downloads/coreutils-8.5.tar.gz and restart build Nov 30 13:08:29 navaltiger: ah that error makes more sense, something corrupted your download on the way Nov 30 13:08:36 Jay7: Let me try. Nov 30 13:08:45 Jay7: last time we checked there were issues with syscalls numbering iirc Nov 30 13:08:47 ant_work: this may break kexec-tools then :) Nov 30 13:08:53 but I'll check Nov 30 13:09:08 fwiw x86 doesn't need any syscall hack Nov 30 13:09:10 btw, I've not tried to build kexec-tools after klibc change.. Nov 30 13:09:46 I can only imagine thesing was obliged to hack it years ago... Nov 30 13:10:10 but then arm has been merged in libc Nov 30 13:10:16 *klibc Nov 30 13:20:33 ant_work: that #include should break arm too on 2.6.36 Nov 30 13:21:50 hm...no, I built it with Nov 30 13:27:36 kergoth: I'm not sure but seems that tasks of one recipe can be running in different bb threads Nov 30 13:28:14 so I see now that do_sizecheck is finished after do_uboot_mkimage started Nov 30 13:28:48 imho, all recipe's tasks should run on same bb thread Nov 30 13:28:58 this may slowdown build but exclude races Nov 30 13:30:16 ant_work: hm.. yes, was built fine.. Nov 30 13:32:09 ah.. crash_setup_regs() is #ifdef'ed there Nov 30 13:53:07 ant_work: kexec-tools fails to build w/o that patch on klibc Nov 30 13:54:07 what's failing? kdump ? Nov 30 13:54:18 hm.. can't say now.. Nov 30 13:54:49 http://tinderbox.openembedded.net/public/logs/task/12015176.txt Nov 30 13:54:58 tinderbox is great sometimes Nov 30 13:55:02 kexec.c Nov 30 13:55:13 kexec.o even Nov 30 13:58:24 | purgatory/purgatory.c:2:20: fatal error: limits.h: No such file or directory Nov 30 13:58:26 :( Nov 30 13:58:29 mipsel Nov 30 14:00:20 Jay, I can't test now, but it looks like now __NR_kexec_load is defined Nov 30 14:00:33 it wasn't when first recipes where committed (for arm) Nov 30 14:00:38 thus the patch/hack Nov 30 14:01:00 ant_work: will you available today's evening? Nov 30 14:01:05 yes Nov 30 14:01:08 see http://git.kernel.org/?p=utils/kernel/kexec/kexec-tools.git;a=blob;f=kexec/kexec-syscall.h Nov 30 14:01:22 ok, we can try to cleanup this then Nov 30 14:03:39 gmorning Nov 30 14:11:35 hm.. there is no purgatory/arch/mips/include dir Nov 30 14:11:40 so, no limits.h.. Nov 30 14:11:43 wtf?.. Nov 30 14:16:01 "Remove purgatory/arch/mipsel/include/limits.h as it is not needed." Nov 30 14:16:10 * Jay7 wonders Nov 30 14:19:09 * Jay7 requesting help of someone with klibc knowledge Nov 30 14:39:55 ka6sox: ("likewise, what repo?") That was when I did a "git clone git@git.openembedded.org:openembedded openembedded.git" Nov 30 14:40:08 ka6sox: the second time, I got no error. Nov 30 14:58:33 Jay7: check kexec-tools-klibc-2.0.2/headers.patch Nov 30 14:58:40 and extend it ;) Nov 30 15:12:53 ant_work: may be just add -Isysroots/.../include ? Nov 30 15:13:03 not sure this is good idea but.. Nov 30 15:13:38 sysroots/mipsel-oe-linux/usr/include/ have limits.h Nov 30 15:13:57 afk Nov 30 15:15:35 be careful, we use nostdinc Nov 30 15:16:14 check purgatory CFLAGS in 'other.patch' Nov 30 16:16:54 with the rate fixed, should i resurrect the progress bar's ETA in bitbake master? i rather like it to have some ballpark figure of how much time i have to wait, do i have time to go get coffee, whatever Nov 30 16:17:35 morning kergoth Nov 30 16:19:08 kergoth_: I like the progress bar. It's usually pretty good - only time when it is off is when you have a partially parsed/cached recipe set. Nov 30 16:19:46 but, to me at least, that's an oddball case, so who cares. Nov 30 16:21:21 kergoth_: sorry, I meant "I like the ETA" Nov 30 16:21:29 i fixed that, though. i just made it so it only considers the # of recipes to be parsed, and the current parsed recipe Nov 30 16:21:35 which stabilizes the rate Nov 30 16:21:47 ah, cool. Didn't see that fix. Nov 30 16:21:58 thats because i havent pushed it yet, wanted to make sure its something people find worthwhile Nov 30 16:22:01 :) Nov 30 16:22:22 ah. Well, I like the ETA. Nov 30 16:22:30 k, thanks for the input Nov 30 16:22:34 hey pb__ Nov 30 16:22:59 kergoth: ah that progressbar.py in bitbake code is bundled version of http://pypi.python.org/pypi/progressbar/2.2 ? can I use system version? Nov 30 16:23:30 JaMa: you're welcome to, yes, but it wont unless you remove bitbake's local copy, or install bitbake with setup.py Nov 30 16:23:45 if you install bitbake, it won't bring along its included copies of its dependencies (PLY, progressbar) Nov 30 16:24:52 I'm installing it with gentoo ebuild (with setup.py) and it's without included copies.. that's why I was asking.. Nov 30 16:24:59 ah :) Nov 30 16:25:05 yes, that should work fine Nov 30 16:25:59 ok, btw I was trying to use bitbake master + your patches from python3 branch without much luck here (it just quits after loading multiprocess module) is it "known" or should it be fine already? Nov 30 16:26:03 if you look at the setup.py, you can see what we depend on Nov 30 16:26:04 requires = ["ply", "progressbar"], Nov 30 16:26:15 i haven't done much testing on the python3 branch yet, could very well be more work to be done Nov 30 16:26:21 work in progress Nov 30 16:26:21 * Jay7 is back Nov 30 16:26:33 * kergoth_ doesn't know 3 very well yet Nov 30 16:26:36 kergoth_: and calling setup should fail if those depends are not found? Nov 30 16:26:57 maybe? Nov 30 16:27:03 * kergoth_ isn't a distutils expert either ;) Nov 30 16:27:10 here it passed ok without ply and progressbar in system Nov 30 16:27:38 kergoth_: I've noted that tasks of same recipe can be running on different bb threads Nov 30 16:27:43 is it true? Nov 30 16:27:49 I'll push xmame stuff Nov 30 16:27:56 every task runs in its own process, and has for years now Nov 30 16:28:10 ah.. Nov 30 16:28:25 03Denis 'GNUtoo' Carikli  07org.openembedded.dev * rd420ff258a 10openembedded.git/recipes/xmame/xmame_0.100.bb: (log message trimmed) Nov 30 16:28:25 xmame: fix runtime endianess issue and bump PR Nov 30 16:28:25 Without that fix we have(at runtime): Nov 30 16:28:25 xmame.SDL Nov 30 16:28:25 error: compiled byte ordering doesn't match machine byte ordering. Nov 30 16:28:29 well.. I misunderstand then Nov 30 16:28:46 each task is spawned in its own process, and each task process reparses its recipe at that time Nov 30 16:28:58 independent of the up front parsing process, which is *only* used to construct the runqueue Nov 30 16:29:24 this was all RP's code while i was away, apparently the impact of reparsing is low, and reduces memory usage a great deal Nov 30 16:31:20 I just looking for source of races Nov 30 16:32:03 I've seen today e.g. that kernel size check task was done in parallel with u-boot mkimage task Nov 30 16:32:18 bitbake schedules tasks to run based on their dependencies. if there is no dependency, then there' no guarantee on execution order. Nov 30 16:34:23 http://savannah.gnu.org/ is down because of passwords leak Nov 30 16:34:34 heard about that, sucks Nov 30 16:34:42 hmm Nov 30 16:34:57 03Chris Larson  07master * r618480f773 10bitbake.git/lib/bb/ (cooker.py event.py ui/knotty.py): Nov 30 16:34:57 cooker: no cached in progressbar and add ETA Nov 30 16:34:57 Rather than updating the progress bar based on the recipe being processed Nov 30 16:34:57 (whether cached or parsed), consider only parsed recipes. This reduces the Nov 30 16:34:57 instability in progress rate introduced by the cached entries, and allows the Nov 30 16:37:49 foerster: fyi, a version of the cache load progress bar which uses the actual position in the file works fine, though currently doesn't give the # of loaded elements in the completion event -- not sure if this is cleaner than storing the number in the file or less clean Nov 30 16:38:27 hi valhalla Nov 30 16:38:38 kergoth_: I don't think the number of loaded elements is too important. Nov 30 16:39:22 that can be added anyway, len(self.depends_cache) after the load Nov 30 16:39:24 if we need it Nov 30 16:39:47 i'm wondering if using st_size + file.tell() is cleaner than num_cached, or less clean, though :) Nov 30 16:39:47 kergoth_: btw, I almost have the triple ctrl-c fixed on my version. Gotta track down one more weird issue. Nov 30 16:39:50 * kergoth_ scratches head Nov 30 16:39:54 ah, nicely done Nov 30 16:40:13 And, i have some oddity happening. Nov 30 16:40:32 I did a fresh build - it hung on building php-native and the task was in state T in ps (stopped). Nov 30 16:40:45 I killed it, tried building php again, and it hung again. Nov 30 16:41:18 i was thinking -- there's really no reason we couldn't use threads instead of processes for the parsing, if the server uses a queue for events -- they're all blocking on I/O, so the global interpreter lock likely isn't a concern. don't know if its worth messing witho or not Nov 30 16:41:23 foerster: huh, that's really strange Nov 30 16:41:39 what causes stopped, is that SIGSTOP/SIGCONT stopped? Nov 30 16:41:53 hmm Nov 30 16:41:55 I think so Nov 30 16:42:03 i don't see what would do that Nov 30 16:42:19 although, looking back, I had configure errors in my php recipe Nov 30 16:42:28 when i switched back to master, i got the errors and I fixed it Nov 30 16:42:37 so maybe something odd with the exceptions or something Nov 30 16:42:45 * foerster shrugs Nov 30 16:43:11 will have to reproduce and figure out when I get some time. I have real day-job work to get done before playing around more. Nov 30 16:46:15 kergoth_: I don't know if there's any advantage to just using threads for parsing. The processes seem to work nicely, so probably no reason to change it. Nov 30 16:46:38 yeah, agreed, if it isn't broken, don't fix it :) Nov 30 16:46:49 plenty of better uses for our time Nov 30 16:47:22 yea, like getting rid of the jacked up runqueue<-->event fire crap Nov 30 16:52:22 hmm, interesting discrepency(sp?) - len(self.depends_cache) after loading gives 7060 entries, yet the stats in the cooker shows 6485 cached -- even skipped doesn't account for the difference, must be inadvertently saving something into the cache which the cooker doesn't end up using Nov 30 16:52:26 * kergoth_ investigates Nov 30 16:53:24 kergoth_: got php build hung again. http://pastie.org/1335716 Nov 30 16:53:44 weird, eh? Nov 30 16:54:06 very Nov 30 16:54:38 happens every time with php with the process server bb Nov 30 16:55:58 I noticed that discrepancy on cache too: here I get 7102 from cache, but end of parsing says 6520. Nov 30 18:30:23 hi Nov 30 18:30:49 hey ericben Nov 30 18:30:55 * khem needs some tea Nov 30 18:30:59 hi Khem, has Gragham Gower comit right ? Nov 30 18:31:11 Graham not Gragham Nov 30 18:32:12 ericben: no, I want to propose him to have commit access actually Nov 30 18:32:17 thats on my todo Nov 30 18:32:24 yes I was about to do the email also ;-) Nov 30 18:32:24 bluelightning: howdy! around? Nov 30 18:32:30 lemme send the message now that I remember Nov 30 18:32:50 * khem delays the tea pot Nov 30 18:33:07 go he unbreak micro distro and seem to be the nearly the only one to test it :-) Nov 30 18:33:54 foerster: my separate-ui-and-master on github has a merge with current master that uses the .st_size/.tell method for the cache progress, and another minor tweak (adds a newline after the cache progress bar) Nov 30 18:33:55 yeah and I will be releaved because most of the time I test and commit his patches Nov 30 18:34:21 * kergoth_ grumbles, flight this afternoon, really hates flying nowadays Nov 30 18:34:49 especially when patchwork is not working Nov 30 18:36:51 kergoth_: ok, cool. I'll check it out when I get a chance. Nov 30 18:37:12 it's trivial, but might save a headache when you decide to merge with master Nov 30 18:38:02 kergoth_: enhanced pat downs :) Nov 30 18:38:06 kergoth_: so here's a question - do you like bbappend or amend? I have amend.inc working here (somehow, through magic). Is either preferred? Nov 30 18:38:08 khem: guh Nov 30 18:38:53 foerster: that's a tough question, really. I like that bbappend is supported by bitbake directly, and performs better, but bbappend is less flexible, since its bound to filename Nov 30 18:39:50 for example, I have approx 10 recipes that I use amend on, typically just to override the base config that's installed or similar. Nov 30 18:39:58 what is wrong with patchwork? Nov 30 18:40:45 hm. I guess I'll stick with amend.inc for now. I don't quite get how it works - it used some voodoo with FILESPATHBASE in my site.conf Nov 30 18:40:47 note that for config files that are accessed via file://, you can avoid the use of amend.inc entirely and just drop the file into the place where you would be putting amend.inc, and adjust FILESPATHBASE to include all your collections Nov 30 18:40:59 amend.inc is found int he same way that file:// files are found Nov 30 18:41:13 via FILESPATH, which is constructed from FILESPATHBASE + FILESPATHPKG + OVERRIDES, in that order Nov 30 18:41:42 and who is responsible for patchwork? Nov 30 18:41:56 wasn't patchwork being moved to a different server or something? Nov 30 18:42:09 yeah ka6sox has been working on it Nov 30 18:42:18 infact its being setup afresh I think Nov 30 18:42:20 having broken PW right before release is not a good idea.. Nov 30 18:42:53 * khem hails mutt Nov 30 18:42:57 dont need pw anymore Nov 30 18:43:08 khem: btw, I need some assistance on klibc/kexec-tools-klibc Nov 30 18:43:10 kergoth_: thanks. I think in most of my cases then, the amend.inc isn't necessary. Nov 30 18:43:37 foerster: yeah, most likely. i find bbappend/amend.inc most useful for adding patches, etc Nov 30 18:43:44 or a quick fix that isn't ready for prime time yet Nov 30 18:44:26 Jay7: ok Nov 30 18:45:12 hmm, still a KeyboardInterrupt in bb.command, bb.data, and bb.shell, wonder if we still need those Nov 30 18:45:22 khem: I've fixed klibc build on mips but now have problems with kexec-tools-klibc Nov 30 18:45:41 limit.h is not found Nov 30 18:45:50 Dudes, to active debug symbols I need uncomment Nov 30 18:45:50 DEBUG_BUILD = "1" Nov 30 18:45:57 INHIBIT_PACKAGE_STRIP = "1" Nov 30 18:45:59 right? Nov 30 18:46:17 Just use it and run a bitbake -v package Nov 30 18:46:22 khem: can we point kexec-tools-klibc to look into sysroots' include dir for limits.h? Nov 30 18:46:31 fidencio[AWAY]: no, things are built with debugging by default and the debug bits go into the -dbg packages. just install those if you need to debug Nov 30 18:46:34 is it good idea? Nov 30 18:46:46 fidencio[AWAY]: PACKAGE_STRIP = "no" Nov 30 18:47:15 fidencio[AWAY]: iirc or as kergoth_ said, you can install -dbg instead Nov 30 18:48:08 but PACKAGE_STRIP = "no" worked a bit better ie while debuging Xorg (loading lot's of libs/drivers itself) Nov 30 18:48:09 kergoth_: which is the simple away to do it in an image? Nov 30 18:48:37 change all packages that compose the image and append the "-dbg" to the packages? Nov 30 18:48:42 try IMAGE_FEATURES += "dbg" Nov 30 18:48:44 in local.conf Nov 30 18:48:50 should do it for any image you build, in theory Nov 30 18:49:42 kergoth_: JaMa: tryint it. many thanks :-) Nov 30 18:49:51 np, let me know how it goes Nov 30 18:49:57 afaik not many people are utilizing IMAGE_FEATURES yet Nov 30 18:50:08 fidencio[AWAY]: there are still those patches related to EFL in patchwork where were some changes requested, do you plan to send PATCHv2 for them? Nov 30 18:50:20 khem: here is log: http://tinderbox.openembedded.net/packages/1152380/ Nov 30 18:50:53 for arm we are placing stdint.h and limits.h by patches Nov 30 18:51:12 upstream have removed 'em for all arch'es Nov 30 18:52:48 JaMa: oh yeah. Nov 30 18:53:20 JaMa: but I'm doing a work that is craping myself. :-\. Finishing it I resent that patches Nov 30 18:54:17 fidencio[AWAY]: ok.. np.. Nov 30 18:55:10 fidencio[AWAY]: GNUtoo|laptop and pespin were interested in enjoy recipes so merging those lightmediascanner patches would be nice, but I'll wait :) Nov 30 18:55:38 basically it's almost finished Nov 30 18:55:54 I'm just waiting for pespin to give me the full RDEPENDS Nov 30 18:55:56 for flac Nov 30 18:56:14 I should also --enable and --disable stuff for lightmediascanner Nov 30 18:56:26 fidencio[AWAY], ^^^ Nov 30 18:56:59 JaMa: are you the maintainer of EFL packages? Which machine are you using? Nov 30 18:57:24 it would be rather nice if we could do a world build without -k -- maybe it should not fail if something is unbuildable, but still fail if something actually fails to build, if you don't do -k Nov 30 18:57:25 EFL is segfaulting with svn HEAD in a babbage machine. Nov 30 18:58:19 kergoth_: build fails! Nov 30 18:58:33 should also profile the runqueue generation for world, it takes an awfully long time Nov 30 18:59:30 kergoth_: http://pastebin.com/9BdgG6eT Nov 30 18:59:55 bah, so there's a recipe there that doesn't create a -dbg package Nov 30 18:59:55 fidencio[AWAY]: sort of, using it on om-gta02,n900,spitz Nov 30 18:59:57 so it breaks Nov 30 19:00:29 hrm Nov 30 19:01:18 ideally itd be nice if bitbake was smart about RRECOMMENDS, not dying if a recommended thing isn't available or buildable, since those things are only recommended, not required Nov 30 19:01:27 fidencio[AWAY]: EFL_SRCREV in OE head works for me and ie svnr55077 built today also works Nov 30 19:01:53 JaMa, it works for me too apart the top-bar bug Nov 30 19:02:24 fidencio[AWAY], did you try to clear .e/ or is it a fresh image and it doesn't apply? Nov 30 19:02:51 JaMa: damn ... I'm using 55071 and I can't choose a profile without a segfault Nov 30 19:02:59 GNUtoo|laptop: yeap. .e and .efreet Nov 30 19:03:10 ok Nov 30 19:03:43 fidencio[AWAY]: previous rev I had 55033 also working fine Nov 30 19:04:17 GNUtoo|laptop: JaMa: Can you start E using valgrind, choose Illume profile and send me a log? please, please, please ;-) Nov 30 19:07:15 * JaMa checking if we already have valgrind 3.6.0 to try it on n900 Nov 30 19:10:43 hmm we have but still with COMPATIBLE_HOST = "^(i.86|x86_64).*-linux" Nov 30 19:11:09 valgrind is not available for arm Nov 30 19:11:10 JaMa: hmm. Ok. nevermind, then :-) Nov 30 19:12:57 khem: 3.6.0 should support armv7-a Nov 30 19:13:16 khem: http://valgrind.org/docs/manual/dist.news.html Nov 30 19:18:17 JaMa: does n900 has cortex in it ? Nov 30 19:18:41 yes Nov 30 19:21:47 hmm seems even ppc is supported too Nov 30 19:22:14 so probably we should make the expression to exxlude say sh,mips,nios Nov 30 19:31:56 Jay7: ok so what was the klibc issue again Nov 30 19:32:38 khem: http://tinderbox.openembedded.net/packages/1152380/ Nov 30 19:32:57 I've fixed klibc already Nov 30 19:33:13 now kexec-tools-klibc should be fixed Nov 30 19:33:43 we have headers.patch that provides stdint.h and limits.h for arm Nov 30 19:34:02 upstream have no such files Nov 30 19:34:49 ok, Nov 30 19:34:58 where is klibc fix Nov 30 19:36:04 http://pastebin.ca/2006988 Nov 30 19:36:14 not commited yet Nov 30 19:37:10 ok Nov 30 19:38:00 khem: about stdint.h and limits.h Nov 30 19:38:13 Jay7: when you build kexec-tools-klibc does it build eglibc too ? Nov 30 19:38:14 what is better way to deal with this? Nov 30 19:38:40 actually if we build a normal toolchain then its ok to suck it from sysroot Nov 30 19:39:00 well.. I'll try clean build Nov 30 19:39:12 please do Nov 30 19:39:26 and if it have these headers in sysroot I think we can use them Nov 30 19:41:43 quick test (clean kexec-tools-klibc and eglibc, then build kexec-tools-klibc) tried to build eglibc too Nov 30 19:41:53 ok Nov 30 19:41:56 I thought so Nov 30 19:42:02 because it needs a toolchai Nov 30 19:42:27 Jay7: I am seeing linking failure with klibc Nov 30 19:42:33 for minimal-uclibc Nov 30 19:42:40 is that what you fixed Nov 30 19:42:58 I've fixed linking of klibc Nov 30 19:43:13 http://pastebin.com/CzzJSPRc Nov 30 19:43:27 yes, it is Nov 30 19:43:32 thats my error I am getting for MACHINE=qemumips DISTRO=minimal-uclibc Nov 30 19:43:35 ok Nov 30 19:43:55 please try with diff applied Nov 30 19:43:56 you have my ack Nov 30 19:44:57 hey, where do i get the zeromq recipe for bitbake? i don't seem to hav et Nov 30 19:46:57 omg.. now eglibc fails to configure.. Nov 30 19:47:09 | configure: error: C preprocessor "mipsel-oe-linux-gcc -E" fails sanity check Nov 30 19:47:15 yeah Nov 30 19:47:29 you have to do this Nov 30 19:48:19 /var/tmp/oe/minimal/mipsel/sysroots/x86_64-linux/usr/mipsel/lib/gcc/mipsel-oe-linux/4.5.2/include-fixed/limits.h:169:61: fatal error Nov 30 19:48:19 bitbake -f -c populate_sysroot eglibc-initial Nov 30 19:48:20 : limits.h: No such file or directory Nov 30 19:48:22 yeah.. Nov 30 19:48:44 if you clean libc basically you cripple the compiler Nov 30 19:48:58 and now using the same compiler to recompile eglibc wont work Nov 30 19:49:06 you have to go thru the bootstrap Nov 30 19:49:10 well, ok Nov 30 19:49:50 foerster: hmm, just hit a waitpid hang with separate-ui-and-server after interrupting, i'll assume that's related to the work you're doing to fix up the shutdown, but wanted to note it in case you hadn't seen it. Nov 30 19:51:22 kergoth_: ok. I don't recall seeing that, but making note of it. You interrupted during parsing, or run? Nov 30 19:51:46 run Nov 30 19:52:02 this is with my merged branch, though i doubt that's a factor Nov 30 19:52:07 doubt it. Nov 30 19:52:45 There seems to be some ordering issues (certainly with the parsing shutdown). Probably others. Nov 30 19:52:57 Did you do a single, or multiple ctrl-c Nov 30 19:52:58 ? Nov 30 19:53:14 multiple. had a fetch hang on me Nov 30 19:53:28 er, that is Nov 30 19:53:39 i hit ^C multiple times, i didn't pay attention to who caught it / whether it was the cooker triple interrupt thing Nov 30 19:53:47 i assume so Nov 30 19:53:56 * kergoth_ just hits it a crapload of times out of habit now.. Nov 30 19:54:17 ok. The UI will catch all of them, if you do it more than 3 times, the it's uncaught in bin/bitbake Nov 30 19:54:27 *UI will catch 3, not all Nov 30 19:55:03 probably need another try/except around server shutdown Nov 30 19:55:10 ah, hmm Nov 30 19:55:18 so you don't ctrl-c it to death :) Nov 30 19:56:14 did the fetch hang due to known reasons, or b/c the server got jacked up (like when I try to build php)? Nov 30 19:56:15 i wonder Nov 30 19:56:24 do the task executing owrker processes ever catch ^C? Nov 30 19:56:32 doesn't SIGINT go to the whole process group? Nov 30 19:56:34 * kergoth_ ponders Nov 30 19:56:37 i block it Nov 30 19:56:46 ProcessServer::run Nov 30 19:56:49 blocks SIGINT Nov 30 19:56:55 should apply to all children Nov 30 19:57:00 ah, good Nov 30 19:57:05 that's a relief Nov 30 19:57:09 that really does save a lot of headaches Nov 30 19:57:23 yes. otherwise, every time I hit ctrl-c it was in the damn parser/etc Nov 30 19:57:27 * kergoth_ nods Nov 30 19:57:30 took me a day to figure out how to block that Nov 30 19:57:31 could be anywhere Nov 30 19:58:06 at first, I was creating a daemon, detaching from the controlling tty, etc. Then I realized I could just block it. Nov 30 19:58:33 hehe. i probably wouldn't have thought about that, unless the signal() i have in the parsing workers reminded me Nov 30 19:58:44 khem: I've found ./mipsel-oe-linux/lib/klibc/include/limits.h in sysroots Nov 30 19:58:53 maybe its time to re-read advanced programming in the unix environment Nov 30 19:58:56 its been a while Nov 30 19:59:00 ah, speaking of that Nov 30 19:59:04 just got a new book Nov 30 19:59:10 this is looking better to use instead *glibc's one Nov 30 19:59:17 "the linux programming interface" Nov 30 19:59:20 looks very good Nov 30 19:59:29 it's the modern day APUE Nov 30 19:59:31 huh, nice, that sounds like a version of the one i just mentioned, but linux specific Nov 30 19:59:34 right, nice Nov 30 19:59:37 will have to check that out Nov 30 19:59:50 $60 on amazon (normal list is $100 I think) Nov 30 19:59:57 Jay7: ok then you have to add that path to compiler include paths Nov 30 20:00:09 Jay7: may be CFLAGS += "-I ..." Nov 30 20:00:21 Jay7: btw. the klibc patch seems to fix the issue Nov 30 20:00:24 so please commit Nov 30 20:00:34 I'll try to remove arm headers.patch and use this for all arch'es Nov 30 20:01:09 kergoth_: worker signals are already removed. Nov 30 20:01:31 cool Nov 30 20:01:53 i have no idea what's stopping php Nov 30 20:02:14 the process tree gets pretty deep there, but I doubt that's a problem. Nov 30 20:02:31 the only time i recall seeing processes stopped/continued was during a suspend/resume Nov 30 20:02:37 which isn't helpful here Nov 30 20:02:39 heh Nov 30 20:02:53 nope. I'm at a loss currently. Nov 30 20:03:05 I was thinking some obscure ulimit kind of thing Nov 30 20:03:30 everything is mostly unlimited under /proc/pid/limits Nov 30 20:05:56 process that trying to read from tty when it is in background can be stopped Nov 30 20:06:50 *stopped*, not sleeping? Nov 30 20:06:57 stopped Nov 30 20:06:58 usually, afaik, waiting on IO is a sleep state Nov 30 20:07:02 weird Nov 30 20:07:06 but this is right for shell Nov 30 20:07:14 shell job control does this Nov 30 20:07:17 oh, so the user can resume it? Nov 30 20:07:19 with fg? Nov 30 20:07:26 huh Nov 30 20:07:29 I'm not sure this is applicable to daemons Nov 30 20:07:43 * kergoth_ thinks it may be time to check susv3 about job control Nov 30 20:08:41 Background processes which attempt to read from (write to when stty tostop is in effect) the terminal are Nov 30 20:08:42 sent a SIGTTIN (SIGTTOU) signal by the kernel's terminal driver, which, unless caught, suspends the process. Nov 30 20:08:51 (from man bash /JOB CONTROL) Nov 30 20:08:53 interesting Nov 30 20:09:07 Jay7: here is patch for kexectools http://pastebin.com/ZXH7rWtM Nov 30 20:09:17 Jay7: this should fix the problems Nov 30 20:09:36 khem: I'll start testbuilds Nov 30 20:09:59 well, ideally none of our task executing processes should be able to touch stdin/stdout/stderr at all, shell or python -- it should be captured into the log and that's it, imo -- python processes can still output to the user with bb.msg.* via the worker->server->ui pipe(s) Nov 30 20:10:13 Jay7: ok, Nov 30 20:10:17 I tried catching SIGHUP, SIGTSTP, SIGTTIN, SIGTTOU to print them, but no luck. Maybe I'll have to do it in the runqueue instead. Nov 30 20:10:56 kergoth_: right. I resurrected the ncurses ui - but fetchers and others walk all over stdout. Nov 30 20:11:01 * kergoth_ nods Nov 30 20:11:38 that's one possible way around this, if it doesn't have any open file descriptors which are ttys, it clearly can't be bitten by the behavior Jay7 is describing Nov 30 20:12:25 check fd's 1,2,3 Nov 30 20:12:38 runqueue maps stdin to /dev/null Nov 30 20:16:47 interesting Nov 30 20:16:57 it always does it at the same spot in the php building Nov 30 20:17:19 http://www.linuxfordevices.com/c/a/News/Electrolux-ProFusion-Infinity-IKitchen/ Nov 30 20:17:30 isnt profusion using OE ? Nov 30 20:17:42 * khem thinks etrunko must know Nov 30 20:21:38 it has an app for recipes :) cool ideas Nov 30 20:22:48 gm Nov 30 20:23:30 khem: yeah, we're using, but not on this refrigerator Nov 30 20:23:48 03Yuri Bushmelev  07master * rf355bb4254 10openembedded.git/recipes/klibc/klibc-1.5.20/klibc_kexecsyscall.patch: Nov 30 20:23:48 klibc: fix linking for mipsel Nov 30 20:23:48 Disable inclusion of . This header have function (at least Nov 30 20:23:48 on mipsel) which breaks klibc's syscalls parser and lead to klibc Nov 30 20:23:48 linking failure. Nov 30 20:24:36 we started to use OE trying to replace ltib Nov 30 20:24:44 etrunko: ok Nov 30 20:25:13 for sure our next product will be using OE ;) Nov 30 20:25:22 etrunko: word Nov 30 20:25:23 thats good sign Nov 30 20:25:40 We should link OE based device released to our news section as well Nov 30 20:25:58 etrunko: what else besides systemd, ofono, bluez, connman and efl you guys need? :) Nov 30 20:26:01 khem: yeah, we will let you know Nov 30 20:26:10 I used ltib in a past life - enjoying OE much more! Nov 30 20:26:40 foerster: hell yeah! Nov 30 20:27:04 stefan_: money!? Nov 30 20:27:23 etrunko: pfff, you have refrigerators Nov 30 20:27:26 khem, thanks for taking the patches Nov 30 20:27:37 stefan_: heh Nov 30 20:28:03 stefan_: yeah. And we can sell icecreams. Nov 30 20:28:07 anyone by any chance any hints how to get php working iwth iconv? it seems to check for libiconv but with eglibc iconv() is part of eglibc Nov 30 20:28:21 * eFfeM feels stuck Nov 30 20:28:31 fidencio[AWAY]: profit! Nov 30 20:28:41 fidencio[AWAY], etrunko: I hope k-s is buying one for each workplace. For fast development of course Nov 30 20:28:53 eFfeM: is it failing, or just not finding something you need? Nov 30 20:28:57 * stefan_ hears php and runs away Nov 30 20:29:04 eFfeM: check the configure file and see how its checking Nov 30 20:29:11 the library names are same Nov 30 20:29:22 php runtime checks everything, and unsets cached ac values Nov 30 20:29:25 weather it comes from eglibc or independently Nov 30 20:29:47 foerster: runtime ? Nov 30 20:29:48 hm.. Nov 30 20:29:52 what do u mean Nov 30 20:30:02 khem it is searching for libiconv but that does not seem to exist, the iconv function is integrated in eglibc Nov 30 20:30:05 does it do some stuff while building Nov 30 20:30:07 AC_TRY_RUN Nov 30 20:30:11 peeked into configure but with little success Nov 30 20:30:19 time to sleep here Nov 30 20:30:20 night all Nov 30 20:30:22 khem: should I bump PR on klibc? Nov 30 20:30:33 I've forget to do this Nov 30 20:30:35 foerster: yeah those AC_TRY_RUN should be cached Nov 30 20:30:39 Jay7: yes Nov 30 20:30:52 ok Nov 30 20:30:59 khem: just a short update. All builds on the release branch before the last update went fine Nov 30 20:31:00 khem: but php's m4 file does a lot of unset ac_* when checking Nov 30 20:31:06 thats's why I reverted the php --with-iconv patch Nov 30 20:31:08 khem: will re-test tomorrow Nov 30 20:31:27 night Nov 30 20:32:16 btw has someone tried a clean build starting with an empty downloads dir ? Nov 30 20:32:23 for release? Nov 30 20:32:30 khem: other Q: you have included 'bits32' dir into CFLAGS Nov 30 20:32:51 what is reason to include bits32 but not include bits64? Nov 30 20:33:00 there are mips64 around :) Nov 30 20:33:23 (but I've not tested any build for mips64) Nov 30 20:34:27 eFfeM: it seems php prefers libiconv over glibc provided iconv() for some bugs Nov 30 20:34:51 Jay7: yes it should be made more configurable I know Nov 30 20:34:59 but klibc does not support mips64 yet Nov 30 20:35:11 well, ok for now then Nov 30 20:35:11 we will do it when it comes to that Nov 30 20:35:19 I'll note this in commit message Nov 30 20:35:24 yes Nov 30 20:41:50 khem, is there a way to get iconv from glibc even when using eglibc? I see uclibc has PREFERRED_PROVIDER_virtual/libiconv ?= "libiconv", can I also pin that in a recipe or only at distro level Nov 30 21:06:20 03Chris Larson  07master * r2be5169a01 10bitbake.git/lib/bb/cooker.py: Nov 30 21:06:20 cooker: don't choke on <100 recipes to parse Nov 30 21:06:20 Signed-off-by: Chris Larson Nov 30 21:07:20 integer divide by zero in progressbar = bad :P Nov 30 21:07:34 hah, yea. Nov 30 21:10:51 hmm, should probably confirm it doesn't explode horribly if toparse is zero too, in the *completely* fully cached case (which doesn't apply to OE) Nov 30 21:11:31 it should just not fire the parsing events in that case Nov 30 21:11:38 is bb used anywhere other than oe? Nov 30 21:12:16 poky, some other oe based things -- its theoretically possible to build something with oe's classes but none of oe's recipes, and thereby have no skipped / SRCREV recipes which can't be cached Nov 30 21:12:27 ah Nov 30 21:16:28 or even a subset of oe's recipes, really.. should really comb through the recipes that use autorev and see if thats really what we want them doing Nov 30 21:23:28 khem, got php building with iconv now, but not 100% happy with it, I had to change the DEPENDS on virtual/libiconv into a DEPENDS on libiconv (distro minimal), how do you feel about that ? Nov 30 21:39:20 meh Nov 30 21:39:53 kergoth_: ping Nov 30 21:39:58 morphis: pong Nov 30 21:40:22 kergoth_: could you please the for-upstream branch of https://github.com/playya/tslib into upstream master? Nov 30 21:40:38 s/could you please/could you please merge/ Nov 30 21:42:11 khem: mipsel (nanonote) and armv5 (tosa) build of kexec-tools-klibc was fine Nov 30 21:42:19 but qemux86 fail Nov 30 21:42:35 but because of other reason Nov 30 21:43:13 http://tinderbox.openembedded.net/public/logs/task/12039234.txt Nov 30 21:43:18 I'm investigating Nov 30 21:43:31 morphis: can you submit a pull request on github for it, so i don't forget? don't have time right now, but don't want to lose track of it -- alternatively, email Nov 30 21:44:36 kergoth_: ok, will tell playya about it as I don't own the repository Nov 30 21:45:18 but will you a mail too Nov 30 21:45:27 k, thanks Nov 30 21:46:26 no problem Nov 30 21:48:26 so I am off Nov 30 21:48:26 gn8 Nov 30 21:51:19 strange.. elf.h is present at least in two places.. Nov 30 22:01:33 eFfeM: providing libiconv separate from glibc is a distro decision Nov 30 22:02:17 seems kexec-tools-klibc is never worked on i686 Nov 30 22:02:27 *never build Nov 30 22:04:08 khem, I understand but minimal uses eglibc and eglibc does not have a libiconv (and minimal does not force it to use libiconv, and neither do some other distros) Nov 30 22:04:58 eFfeM: how angstrom-2010.x behave in this case? Nov 30 22:05:04 it is using eglibc iirc too Nov 30 22:07:36 eFfeM: could you post the config.log somewhere for php Nov 30 22:07:57 eFfeM: I think packaging libiconv along with eglibc/glibc can be a headache Nov 30 22:09:57 khem: I'll give up on i686 Nov 30 22:10:48 it is require more patches Nov 30 22:11:32 e.g. remove/replace 'basename' because klibc doesn't have it Nov 30 22:11:45 Jay7: make sure that the x86 failure is not caused by the patches we are commiting Nov 30 22:11:54 yes, I'll Nov 30 22:11:57 k Nov 30 22:11:59 thats fine Nov 30 22:12:06 as long as we are not regressing its fine Nov 30 22:12:14 Jay7: angstrom blacklists libiconv: Nov 30 22:12:15 ../../conf/distro/include/angstrom-eglibc.inc:ANGSTROM_BLACKLIST_pn-libiconv = "the glibc builtin iconv replacement is used" Nov 30 22:12:30 khem, will do, one sec Nov 30 22:13:42 eFfeM: yes if you choose glibc/eglibc then it provides virtual/libiconv Nov 30 22:13:56 so that makes sure that libiconv is not picked from elsewhere Nov 30 22:15:11 khem, ah ok, understood, issue is that eglibc does not provide a libiconv Nov 30 22:15:42 khem here is the config.log with libiconv in deps (without virtual) http://filebin.ca/mtfjaw/config.log Nov 30 22:15:53 will make the virtual one right away Nov 30 22:17:02 btw in ph_4.2.13.bb I only removed --disable-iconv and added -with-iconv --with-iconv-dir=${STAGING_LIBDIR}/.. Nov 30 22:17:26 mind the .. at the end, guess there is a better macro for it, but the same construct was also used elsewhere Nov 30 22:17:43 removed --without-iconv, not --disable-iconv Nov 30 22:21:03 khem, finally got sick of handling my patches? Nov 30 22:21:08 :P Nov 30 22:22:42 eFfeM: it still finds the one from eglibc Nov 30 22:22:48 grg: heh sort of. Nov 30 22:23:25 but I look at it as I have one more guy to delegate stuff to :) Nov 30 22:23:31 khem, what I gave was the one with libiconv present in depends, now did a bitbake -cclean php libiconv; bitbake libiconv Nov 30 22:23:39 oops that is bitbake php Nov 30 22:23:51 it is running configure now Nov 30 22:24:07 eFfeM:bitbake -c clean should remove traces of libiconv from staging Nov 30 22:24:34 grg: I am sure your help is needed in OE so buckle up Nov 30 22:24:41 khem, yup did that Nov 30 22:24:53 btw need to go Zzzz in 15 mins or so Nov 30 22:25:14 khem, but what will happen to my social life... :) Nov 30 22:28:15 grg: this is community are'nt participating in communities social Nov 30 22:28:34 yeah Nov 30 22:28:38 at least its not facebook Nov 30 22:29:42 people here have true identities :) Nov 30 22:29:44 grg you probably have the most social life of all of your friends! which of them can tell they hang out every day with people from europe and USA Nov 30 22:29:59 khem bitbake -cclean doe snot remove ../../../tmp_minimal/sysroots/armv5te-oe-linux-gnueabi/usr/lib/libiconv.so* Nov 30 22:30:08 how about headers Nov 30 22:30:30 so, micro-uclibc/console-image builds now Nov 30 22:30:54 working on native-sdk-image Nov 30 22:31:32 khem, symlinks remain (libiconv.so and libiconv.so.2, they link to libiconv.so.2.5.0 which is gone Nov 30 22:31:41 grg: hmm I think micro should be enough with console-image isnt it Nov 30 22:31:49 well nothing wrong in making a SDK so go ahead Nov 30 22:31:49 guess this is an issue in staging, manually nuked the symlinks, now retrying Nov 30 22:32:02 eFfeM: thats fine Nov 30 22:32:08 how about iconv.h Nov 30 22:33:01 well.. qemux86 build of kexec-tools-klibc fails on same place with master Nov 30 22:33:06 I'll commit then Nov 30 22:35:09 hm.. Nov 30 22:35:19 khem, hm it configures fine Nov 30 22:35:24 ok Nov 30 22:35:30 so problem is elsewhere Nov 30 22:35:46 btw. you should just use --enable-iconv with out any path Nov 30 22:35:47 compile faisl iconv.h no such file or dir Nov 30 22:36:00 ok now that could be issue Nov 30 22:36:02 it used to give problems before Nov 30 22:36:08 make sure the eglibc-dev has this header Nov 30 22:36:25 I need help with klibc PR bump.. I'm not experienced with that INC_PR stuff.. Nov 30 22:37:10 ant__: hey Nov 30 22:37:20 kehm when I tried with only --enable-iconv I got a configure error on a clean build that fork was not found, and I saw also that it added -I/usr/include and -L/usr/lib to the gcc command when testing things Nov 30 22:37:27 s/kehm/khem/ Nov 30 22:37:27 ant__: where should I bump klibc's PR to do it right? :) Nov 30 22:37:28 hello Nov 30 22:37:58 ant__: khem was helped to fix kexec-tools-klibc too Nov 30 22:38:06 so it builds ok now Nov 30 22:38:08 great Nov 30 22:38:21 we should repackage it a bit like debian Nov 30 22:39:41 please bump INC_PR in klibc.inc Nov 30 22:40:20 khem will try tomorrow evening, too tired now, will also kick off a build from scratch on the autobuilder tomorrow Nov 30 22:40:37 but I feel I tried most of the things Nov 30 22:41:17 eFfeM: good no worries Nov 30 22:41:25 btw. checksums for php native seems wrong for me Nov 30 22:41:32 eFfeM: good night Nov 30 22:43:36 khem, here is a log that had the faulty configure: http://tinderbox.openembedded.net/public/logs/task/11475121.txt Nov 30 22:43:48 still here for a short while, the Mrs is not home from work yet Nov 30 22:43:54 but guess it won't take long Nov 30 22:44:21 i'm unaware of a change for php native, but the location has changed Nov 30 22:44:48 here's my tinderbox php log: http://tinderbox.openembedded.net/packages/php/?page=2 Nov 30 22:45:39 actually should have given http://tinderbox.openembedded.net/packages/php/ but the one I mention above from nov 28 is the one before undid the with-iconv change Nov 30 22:49:40 well.. tosa and nanonote builds was fine Nov 30 22:49:43 pushing Nov 30 22:50:05 Jay7: ok Nov 30 22:51:55 03Yuri Bushmelev  07master * r1a7484525f 10openembedded.git/recipes/klibc/klibc.inc: klibc: bump PR to force rebuild after mipsel fix Nov 30 22:51:58 03Yuri Bushmelev  07master * rfa740ca249 10openembedded.git/recipes/kexec-tools/ (2 files in 2 dirs): Nov 30 22:51:58 kexec-tools-klibc: switch to using klibc's staged includes Nov 30 22:51:58 kexec-tools-klibc is switched to use klibc's includes instead of having Nov 30 22:51:58 patch with own. This will fix building for mipsel as well. Nov 30 22:51:58 NOTE: now kexec-tools-klibc is just using 'klibc/include[/bits32]' dirs Nov 30 22:54:56 hi all Nov 30 22:55:44 I'm trying to build qt4-x11-base-image, but it fails on conflicting device_table and fails to install task-qt4-x11-base Nov 30 22:55:53 does anyone knows how to solve ? Nov 30 22:56:01 alx_: paste that part of log Nov 30 22:56:06 there should be other error Nov 30 22:57:34 Jay7, http://pastebin.com/UzBGyfMK Nov 30 22:57:46 this is last part of the log Nov 30 22:58:13 main error is below about libqtxmlpatterns4 Nov 30 23:00:21 hmmm this is a real issue I think Nov 30 23:00:27 ericben: around ? Nov 30 23:00:46 ericben: task-qt4-x11.bb has libqtxmlpatterns4 but no one seems to provide it Nov 30 23:00:47 khem: not far from bed but still here :-) Nov 30 23:01:05 * khem confiscates ericben quilt Nov 30 23:01:09 lol Nov 30 23:01:27 khem: strange I build it here before subiting the patch I check Nov 30 23:01:54 now I know what quilt means ;-) Nov 30 23:01:55 ericben: I think you might have forgotton to git add/push somethinhg Nov 30 23:02:11 heh Nov 30 23:02:17 I know those errors Nov 30 23:04:02 well, guys, I lost you Nov 30 23:04:46 alx_: dont worry ericben is firing up the engine Nov 30 23:05:55 khem, lol well I've to solve a lot more after this :) Nov 30 23:06:03 strangeI have a libqtxmlpatterns4_4.7.1-r21.0.9 ipk Nov 30 23:06:28 ericben: check the logs Nov 30 23:06:32 alx_: which distro are you using ? Nov 30 23:06:32 who provides it Nov 30 23:06:48 it's qt4-x11-free Nov 30 23:06:59 ike the other qt libs Nov 30 23:07:16 angstrom-2008.1 Nov 30 23:07:22 should go to 2010 ? Nov 30 23:07:30 ericben: will it work with mimimal? Nov 30 23:07:33 alx_: okso qt 4.63 which may not have this Nov 30 23:07:41 khem: it should Nov 30 23:07:45 ok Nov 30 23:07:51 target machine is qemux86 (dont' ask me why!!) :D Nov 30 23:07:53 alx_: yeah use 2010 or minimal Nov 30 23:08:04 khem: I'll launch a build with minimal on my build server now Nov 30 23:08:15 ericben: I will too. Nov 30 23:08:21 angstro-2010.x ? Nov 30 23:08:24 as I want to boot my efikamx into it Nov 30 23:08:26 angstrom-2010.x ? Nov 30 23:08:40 or minimal ... Nov 30 23:08:45 ok should go with minimal Nov 30 23:08:46 alx_: I initialy tested with angstrom 2010.x Nov 30 23:08:49 ericben: so he other qt is the image right ? Nov 30 23:08:57 hah Nov 30 23:09:03 qt4-x11-free Nov 30 23:09:05 I meant Nov 30 23:09:30 qtxmlpattern seems only present in qt 4.7.x Nov 30 23:09:49 oh.. efika.. Nov 30 23:09:54 what do you mean by "so he other qt is the image right ?" ? Nov 30 23:09:58 I'm just sad about having another Nov 30 23:10:00 * Jay7 need to test kexec on efikamx.. Nov 30 23:10:02 night compiling :( Nov 30 23:10:42 ericben: I was typing when my konsole was frozen Nov 30 23:10:54 khem: is there an easy way to select some packages in the task depending on qt's version ? Nov 30 23:10:57 ericben: anyway so I should bitbake qt4-x11-free right Nov 30 23:11:09 this will build qt4 for x11 Nov 30 23:11:21 ericben: yeah thats what I need Nov 30 23:11:26 and it will have some stuff right ? Nov 30 23:11:32 if you want the image bitbake qt4-x11-demo-image Nov 30 23:11:52 like browser etc ? Nov 30 23:11:53 this builds the image with qtdemo launched at boot Nov 30 23:12:23 in qtdemo there are s ome webkit demo Nov 30 23:14:01 DEPENDS = "foo (>=x.y.z)" Nov 30 23:14:24 ok, while I recompile the world together Nov 30 23:14:29 ok thanks I'll try it Nov 30 23:14:35 ericben: I tried that I want to surf real internet Nov 30 23:14:38 may I ask some support for two issues I have ? Nov 30 23:14:46 alx_: one night may not be enough for the world ;-) Nov 30 23:15:06 ericben, true, I hope so, because tomorrow morning I've a demo Nov 30 23:15:15 and another important demo on friday Nov 30 23:15:19 khem: well that depends of the meaning of "real internet" :-D Nov 30 23:15:32 heh I mean I want to play videos Nov 30 23:15:40 HD ones Nov 30 23:15:46 and it should use QT Nov 30 23:16:23 well, on the i.MX515 this won't work without the hardware codecs Nov 30 23:17:16 nite all, (still late) Nov 30 23:17:16 and I don't think it even can play full hd video, only "hd ready" one I think Nov 30 23:17:26 bye eFfeM Nov 30 23:18:59 ericben, do you know about any python-qt recipe I may inspire from ? Nov 30 23:20:02 ericben: hardware codecs from FreeScale ? Nov 30 23:20:10 alx_: there is python-pyqt but I'm not sure it builds against qt 4.7.1 Nov 30 23:20:58 grg: perl and sqlite patches are not yet in master Nov 30 23:21:39 I'm still there with minimal / uclibc (I knoe there is minimal-uclibc building fine ;) Nov 30 23:21:40 ericben: if I have a blue ray DVD player and a computer/laptop which has a blueray DVD drive are there any differences that bluray DVD player offers over laptop ? Nov 30 23:22:02 ant__: minimal/uclibc is no combo dont use it Nov 30 23:22:08 khem: yes : http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=i.MX515&nodeId=0162468rH31143ZrDR633B&fpsp=1&tab=Design_Tools_Tab# at the end there at openmax & codoecs Nov 30 23:22:11 yep, I saw Nov 30 23:22:15 ant__, i don't think any of my patches from yesterday have been pushed Nov 30 23:22:20 ant__: you should simply use minima-uclibc distro Nov 30 23:22:23 still intriguing..not so far it seems Nov 30 23:22:28 grg: no not yet Nov 30 23:22:30 I will Nov 30 23:22:33 already built minimal-uclibc Nov 30 23:22:33 tonight Nov 30 23:22:37 looks fine Nov 30 23:23:13 khem: I don't know anything about blueray except it's expensive :) Nov 30 23:24:11 ericben: heh I was wondering from video/audio quality pov Nov 30 23:24:33 ericben: are these codecs like kernel modules ? Nov 30 23:25:11 they want me to be registered in order to download that crap Nov 30 23:26:33 khem: iirc, #efika people says that mplayer with freescale kernel can play 720p Nov 30 23:26:44 khem: you can ask there Nov 30 23:28:09 khem: no these are some libs but they need the freescale kernel for the display part Nov 30 23:28:45 Jay7: with the hardware codecs or only in software ? Nov 30 23:29:02 ericben: with hw codecs I think Nov 30 23:29:16 ok I confirm (but was not using mplayer) Nov 30 23:29:52 in software something similar to omapfbplay should be possible as it's a cortex a8 with neon like the omap3530 Nov 30 23:29:52 khem: also gentoo and debian ports on efikamx are progressing Nov 30 23:29:59 so.. we can grab something from there Nov 30 23:31:47 khem: I've built several angstrom 2010.x imagesfor amrv5 target on a Fedora 14 (with make 3.82) so the release should work not bad on Fedora Nov 30 23:34:07 ok build launched time to sleep, bye ! Nov 30 23:41:21 ericben|away: gn Nov 30 23:41:27 khem: about last moment hot-fixes, those mentioned by me and ynezz refers to angstrom Nov 30 23:41:58 x264/mesa/e2fsprogs Nov 30 23:42:08 ok, tell me whats needed Nov 30 23:42:28 did it work better with 0.9.30.3 Nov 30 23:42:46 oh, moved straight to git Nov 30 23:43:16 well angstrom does not use git version Nov 30 23:43:35 with git version things will be better for sure Nov 30 23:44:02 http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-November/027347.html Nov 30 23:44:09 ant__: if you test the release branch please update the Testing table with your combinations Nov 30 23:44:18 so should everyone Nov 30 23:44:39 ok, but I'm on master Nov 30 23:45:00 I have some queue... Nov 30 23:47:09 khem: about e2fsprogs there is a pending update c/o blindvt Nov 30 23:47:18 in Patchwork Nov 30 23:47:55 ant__: well two patches will be taken care if angstrom bumps to 0.9.30.3 Nov 30 23:48:09 no way before release :/ Nov 30 23:48:09 the e2fsprogs is a nasty one Nov 30 23:48:19 if you disable tls then git recipe will suffer Nov 30 23:48:40 so best way is to have a DISTRO_FEATURE to indicate if a DISTRO supports TLS or not Nov 30 23:48:51 sure Nov 30 23:48:53 then we could use that feature here Nov 30 23:49:06 but that will be intrusive Nov 30 23:49:21 and I dont want to have something like that at this stage of release Nov 30 23:49:53 I will meet koen tomorrow so I will discuss bumping the uclibc version Nov 30 23:49:56 that should be doable Nov 30 23:50:00 0.9.30.3 needs newer binutils iirc. not upgradable at this point Nov 30 23:50:07 ah I see Nov 30 23:50:10 yes it does Nov 30 23:50:13 for arm Nov 30 23:50:16 :// Nov 30 23:50:45 so for uclibc lovers use minimal-uclibc or micro-uclibc DISTRO Nov 30 23:51:04 I guess thats fair enough Nov 30 23:51:39 angstrom does not have to support uclibc its distro decision Nov 30 23:52:03 FWIW angstrom-2008.1 console, opie, x11, x11-gpe images build fine Nov 30 23:52:04 with binutils "2.20.1" and uclibc "git" like minimal distro but  with gcc-4. Nov 30 23:52:21 gcc-4.3.3 Nov 30 23:52:53 yes they do. Convince the maintainers :) I dont maintain angstrom Nov 30 23:53:01 no point Nov 30 23:53:19 I was just round-testing Nov 30 23:53:31 I do have say in minimal and micro Nov 30 23:53:46 now I only lack minimal-uclibc for my image-size shootout ;) Nov 30 23:54:05 minimal + uclibc ;) Nov 30 23:55:26 * khem needs tea Nov 30 23:56:00 * ant__ needs sleep Nov 30 23:56:05 good night Dec 01 00:20:25 * Jay7 -> sleep Dec 01 01:09:18 hi all. very new to OE. used getting started page to set it up. i've used a beagleboard config file (making angstrom) and then tried bitbake base-image - got a no BBPATH variable set error - i put it into the config file and retried - still same error - can anyone help me troubleshoot please? angstrom channel is dead Dec 01 01:17:08 andyzammy, BBPATH should be set in your shell's environment Dec 01 01:19:10 the oe instructions suggest it can be put into the config file? i've found examples that do just that? Dec 01 01:27:06 andyzammy, i invoke bitbake from a script. like this: http://pastebin.com/02hr9CvZ Dec 01 01:27:55 im pretty sure the instructions say the BBPATH is in the shell environment. bitbake/oe wont find your local.conf without it Dec 01 01:33:45 ah, i believe i'm getting mixed up between BBFILES and BBPATH.. i'm still confused a lil though - i actually have OE in 2 places, installed with different methods - the 2nd one was done with a setup scripts taken from gitorious.. for THAT one, it actually runs but fails with tonnes of EOL while scanning for string literal errors.. Dec 01 01:34:33 i think i'm getting to the problem now.. the 1st one was installed in root, and running as sudo (which is why it's failing to find BBPATH)... but Dec 01 01:35:29 looking at my own path i still can't see where it gets BBPATH from.. Dec 01 01:36:39 would OE_HOME be an equivelent to BBPATH? Dec 01 01:37:34 no Dec 01 01:38:23 BBPATH points to things that can be "include"d by bitbake, starting with the bitbake.conf Dec 01 01:44:42 andyzammy, check that what distro/machine is written in local.conf is valid and existent in OE (got same error and I had a typo in DISTRO) Dec 01 01:45:00 okay, for my 1st installation the problem is solved - add BBPATH to .bashrc for root, but for this 2nd installation, how is it working without a BBPATH? i've checked my $PATH and scanned .bashrc but no mention of BBPATH Dec 01 01:49:08 andyzammy, don't run bitbake as root. it wont work Dec 01 01:49:36 also, you can echo $BBPATH to find out what its set to Dec 01 01:50:36 openembedded may accidentally clobber stuff in your host filesystem if you run things as root... i think there was something added to make it bail out if you try to run as root Dec 01 01:50:51 weired - i get a blank line Dec 01 01:50:56 for BBPATH Dec 01 01:51:23 gotchya, will remove and forget about first installation then Dec 01 01:52:26 2nd one looks like is now nicely building base-image now i've fixed local.conf file, thanks for making me check it alx_.. but still baffling me how it's running at all without BBPATH Dec 01 02:41:27 the base-image build failed: http://pastebin.com/sQzMUjj9 by looks of things something cp didn't like? Dec 01 02:45:08 andyzammy, indeed. You're using busybox on your host? what distro is this? Dec 01 02:46:04 it's on here (ubuntu) but not being used.. should i remove it? Dec 01 02:46:58 andyzammy, you'll probably run into other problems because its not something that is tested. Dec 01 02:47:23 * grg wonders if anyone does OE on OE builds... Dec 01 02:47:39 OE should be self hostable, in principle Dec 01 02:48:38 i don't follow you.. was it using the cp within busybox? Dec 01 02:50:04 khem ping? Dec 01 02:50:22 andyzammy, yes Dec 01 02:50:42 hence your error message. the busybox cp didn't like the long option to cp Dec 01 02:51:34 ah ofc - it's because it comes 1st in the alphabetical order within the /bin folder.... so the fix is just to uninstall it right? Dec 01 02:53:03 andyzammy, i guess. I'm completely unfamiliar with ubuntu so couldn't say categorically Dec 01 02:53:18 andyzammy, i would hope you don't hose your system doing that Dec 01 02:54:57 no, its a swiss tool i used once to create a simple bootable OS-on-a-memstick.. didn't come with the distro.. i did think it weired how OE was fishing around busybox but it didn't even click that it was using it's cp + prob all it's other programs lol Dec 01 02:55:02 thx for the help :) **** ENDING LOGGING AT Wed Dec 01 02:59:57 2010