**** BEGIN LOGGING AT Thu Nov 18 02:59:58 2010 Nov 18 06:42:11 hi guys, i am trying to setup the oe build environment on my fedora 12 Nov 18 06:42:56 i didn't understand the instructions given about the ccache thing Nov 18 06:43:09 can anyone please explain to me what to do ? Nov 18 06:43:17 i have just done yum install bitbake Nov 18 06:43:37 version 1.8.10 Nov 18 07:17:30 03Sebastian Krzyszkowiak  07master * r6638e662d8 10openembedded.git/recipes/freesmartphone/opimd-utils_git.bb: Nov 18 07:17:30 opimd-utils: update SRCREV Nov 18 07:17:30 Signed-off-by: Martin Jansa Nov 18 07:17:40 03Martin Jansa  07master * r96179e1d31 10openembedded.git/recipes/linux/ (4 files in 3 dirs): Nov 18 07:17:40 linux(-openmoko)-2.6.34: update openmoko.patch with gta01 support Nov 18 07:17:40 Signed-off-by: Martin Jansa Nov 18 07:17:42 03Martin Jansa  07master * rb31d65ccc1 10openembedded.git/ (5 files in 3 dirs): Nov 18 07:17:42 linux-openmoko: remove old version based on 2.6.29-rc3, switch gta01 to 2.6.34 Nov 18 07:17:42 Signed-off-by: Martin Jansa Nov 18 07:25:06 PHP 5.2.13 for the last month fails to compile. Nov 18 07:25:29 It can't find libcrypto (OpenSSL), later mysql libs, another libs, etc.. Nov 18 07:25:47 All those libs are staged into native and target Nov 18 07:26:06 Ideas why it can't find libs? Nov 18 07:32:32 Btw Nov 18 07:32:47 org.oe.dev or org.oe.stable should be better used? Nov 18 07:41:45 what is the meaning of a build environment? Nov 18 07:41:58 what am i supposed to use oe exactly for? Nov 18 07:42:35 is it just like the qmake environment for qt? Nov 18 07:44:24 build environment: OS with some required libs and setups Nov 18 07:44:41 OE - Open EMBEDDED, EMBEDDED devices. Nov 18 07:48:28 good morning Nov 18 07:54:32 HI all while building the OE for am3517-evm i am getting error in qemu , please help me to fix that error . Nov 18 08:23:52 03Martin Jansa  07master * r20ca75903a 10openembedded.git/recipes/xorg-lib/ (9 files in 2 dirs): Nov 18 08:23:52 pixman: add version 0.21.2 Nov 18 08:23:52 Signed-off-by: Martin Jansa Nov 18 08:23:55 03Martin Jansa  07master * r9f01acb255 10openembedded.git/ (47 files in 6 dirs): Nov 18 08:23:55 xorg: new versions for 2010-11-17 Nov 18 08:23:55 Signed-off-by: Martin Jansa Nov 18 08:23:57 03Martin Jansa  07master * r7ba0adcf4e 10openembedded.git/recipes/ (35 files in 3 dirs): Nov 18 08:23:57 xorg: remove recipes for versions not pinned in any distro config Nov 18 08:23:57 Signed-off-by: Martin Jansa Nov 18 08:24:32 * davidlt back Nov 18 08:31:23 Any ideas for PHP? Nov 18 08:32:58 davidlt: is OE an OS? Nov 18 08:33:42 I am using Gumstix, which is pretty much heavily based on OE Nov 18 08:35:33 http://old.nabble.com/PHP-Fails-to-build-td30110142.html Nov 18 08:37:41 davidlt: my idea of OE is that it simply provides the uclibc, and other libraries fineturned for embedded applications to compile your programs for embedded systems, Nov 18 08:38:02 davidlt: how is OE an OS? is it equivalent to an OS in terms of ANDROID? Nov 18 08:39:09 ksinkar: OE is not an OS Nov 18 08:39:51 it's metadata about different versions of and dependencies between software packages, and a set of tools for building and deploying said packages in a way that respects the dependences Nov 18 08:40:17 mrj10: thank you Nov 18 08:40:23 you can use it to build the kernel and userspace utilities and applications for any OS you want Nov 18 08:40:44 it is especially well set up for Linux, but there is nothing fundamental preventing you from extending it to work for something else Nov 18 08:41:06 mrj10: thank you very much for that enlightening explanation Nov 18 08:41:14 don't get me wrong, i think it would be a ton of work to try something else Nov 18 08:41:35 np Nov 18 08:42:28 in the git folder that i just downloaded it has also mentioned in /sites avr32? Nov 18 08:44:57 what board do you have? Nov 18 08:45:21 support is generally added on a per-board basis. board configuration files can be found in conf/machine/ Nov 18 08:45:47 i have LPC2468 and OMAPL138, i want to add support for that Nov 18 08:45:58 it looks like the two avr32 boards supported out of the box are at32stk1000 and atngw100 Nov 18 08:46:34 btw the avr32 comment was an exclamation that even avr 32 was supported Nov 18 08:46:45 oh gotcha Nov 18 08:46:59 sorry, misinterpreted. Nov 18 08:47:14 so the LPC2468 is a little ARM7 board? Nov 18 08:47:20 i have hawkboard (omapl138)and olimex e2468(lpc2468) board Nov 18 08:47:30 now both have different requirements Nov 18 08:48:01 the omap board has MMU cause of ARM 9 while lpc board is without mmu coz of arm 7 Nov 18 08:48:12 yeah Nov 18 08:48:48 OE has recipes to build uclibc, but i'm not sure about uclinux Nov 18 08:49:00 hawkboard is supported out of the box Nov 18 08:49:05 see conf/machine/hawkboard.conf Nov 18 08:50:18 this is pretty cool Nov 18 08:50:47 do the angstrom guys use OE for their server as the backend? Nov 18 08:53:18 i'm not sure *exactly* what you're asking, but i believe the answer is likely yes Nov 18 08:53:40 angstrom and OE are tightly linked, Koen Kooi contributes heavily to both (and is in this channel often) Nov 18 08:53:49 uclinux may be more of a project Nov 18 08:53:50 morning Nov 18 08:54:23 but somebody tried to get it going for blackfin; there are still remnants of that in the OE repository if you want to take a peek. i believe they could build a minimal version of angstrom Nov 18 08:54:25 hrw: morning Nov 18 08:55:03 he announced his intentions here http://blackfin.uclinux.org/gf/project/toolchain/forum/?_forum_action=ForumMessageBrowse&thread_id=586&action=ForumBrowse , you might touch base with him if you have uclinux-specific questions Nov 18 08:55:38 and others on the channel are generally helpful with creating and debugging recipes, so long as you read through the user manual before asking http://docs.openembedded.org/usermanual/usermanual.html Nov 18 08:56:44 mrj10: this was THE document that I was looking for Nov 18 08:58:19 :) glad to help Nov 18 08:58:24 i must be off to bed, but best of luck Nov 18 08:58:56 gm Nov 18 08:59:31 hi effem Nov 18 09:09:35 klajdsfkljakldfjakljdsfkljaldsf Nov 18 09:09:37 ads Nov 18 09:09:38 f Nov 18 09:10:46 hehe Nov 18 09:10:49 keyboard works. Nov 18 09:10:56 little kids Nov 18 09:10:57 or cat Nov 18 09:16:06 Tartarus, ericben|away : http://www.kaeilos.com/debian-lenny-and-git-native Nov 18 09:20:15 morning Nov 18 09:34:49 hi jineld Nov 18 09:37:06 hi jin Nov 18 09:37:15 hi blindman Nov 18 09:43:08 hi woglinde :) Nov 18 09:58:23 gm Nov 18 09:59:58 hi likewise Nov 18 10:00:55 hi likewise Nov 18 10:06:09 03Martin Jansa  07master * rc3119c5f43 10openembedded.git/recipes/linux/ (3 files in 3 dirs): Nov 18 10:06:09 linux-2.6.34: small update of om-gta0* defconfig Nov 18 10:06:09 Signed-off-by: Martin Jansa Nov 18 10:32:51 org.oe.dev vs. org.oe.stable? Any comments? Nov 18 10:33:33 .dev Nov 18 10:33:46 stable/2009 is 1.5y old Nov 18 10:35:08 what is release cycle? 1.5. is a long time... Nov 18 10:39:48 davidlt: random Nov 18 10:47:14 If I want to recompile stuff on bibake, but I want clean build (no garbage from older builds). Can i do it somehow in fast way? Nov 18 10:47:29 I don't want to remove the whole tmp dir and wait 8 hours just to test. Nov 18 10:48:00 Is it possible only delete sysroot and bitbake would recompile few packages and reconstruct sysroot? Nov 18 10:51:02 mickey|office: good morning Nov 18 10:59:56 pb_: good mornign! Nov 18 11:00:31 davidlt: AFAIK would be a mess Nov 18 11:01:03 Why? Nov 18 11:01:56 You know the order how pacakges should be staged, you have already build all packages. That's like a snapshots. So Bitbake should be always able to build sysroot Nov 18 11:14:28 Someone knows how I can boot an uImage via tftp in U-Boot? I can load the file with the tftp command, but bootm just says "Wrong Image Format for bootm command" -- Am I missing something? Nov 18 11:18:43 Jordinar: you're OT, and is hard to answer without knowing details Nov 18 12:06:08 03Martin Jansa  07master * r2fa21974c6 10openembedded.git/recipes/python/python-elementary_svn.bb: Nov 18 12:06:08 python-elementary: add missing python-evas to DEPENDS Nov 18 12:06:08 Signed-off-by: Martin Jansa Nov 18 12:10:29 re Nov 18 12:22:17 hi woglinde Nov 18 12:22:29 hi pb Nov 18 12:36:49 ls Nov 18 12:37:05 rm Nov 18 12:37:50 hm, where did that come from? better do rm -rf (01:29:38 PM) woglinde: rm Nov 18 12:39:47 hi mwester Nov 18 13:08:36 anyone having an address of the angstrom src mirror handy? Nov 18 13:09:10 php released a new version, old version is moved but also fails the md5sum, would like to compare Nov 18 13:11:15 jo mickeyl Nov 18 13:23:59 hi ant Nov 18 13:25:43 * XorA prods OE for being broken at the moment Nov 18 13:26:35 hm? Nov 18 13:26:40 procps wont compile Nov 18 13:26:42 how? Nov 18 13:26:51 which is essential for all images Nov 18 13:26:57 good morning Nov 18 13:27:47 why? Nov 18 13:27:53 http://tinderbox.openembedded.net/packages/procps/ Nov 18 13:27:59 didnt see any commits toward procps Nov 18 13:28:22 dont know, my guess is it relies on host stuff somewhere Nov 18 13:28:30 ? Nov 18 13:29:38 tinderbox would help Nov 18 13:29:58 well tinderbox shows success, so I am assuming something is different on my host Nov 18 13:30:41 * Tartarus goes trolling for acks for help2man- Nov 18 13:31:16 look in the #marsattacks channel :-) Nov 18 13:31:36 * XorA wonders how many other OE developers are using Fedorsa 14 Nov 18 13:32:38 * mwester is about to try Nov 18 13:32:51 hi kgilmer Nov 18 13:33:03 64 bit or 32? Nov 18 13:33:22 64bit Nov 18 13:33:35 XorA: I'm so sorry :) Nov 18 13:33:53 could be worse, could be an ubuntu release :-) Nov 18 13:34:07 :p Nov 18 13:34:37 I need more disk to get my 64-bit install "OE-capable"... but it's in the works. Nov 18 13:47:39 Anyone built evtest recently? Nov 18 13:47:43 Got a checksum mismatch overnight Nov 18 13:48:05 http://old.nabble.com/latest-console-image-fails-to-build---procps-ps-td30234659.html Nov 18 13:48:11 not just me having procps issue Nov 18 13:49:50 BB_VERSION = "1.8.18" & METADATA_BRANCH = "overo" Nov 18 13:49:59 doesn't match Nov 18 13:50:04 our tests Nov 18 13:50:37 Tartarus: just cleaned evtest, removed the source, and bitbaked ok here Nov 18 13:51:16 local problem then, thanks Nov 18 13:52:35 xora -> METADATA_BRANCH = "overo" Nov 18 13:52:38 no help Nov 18 13:52:41 sorry Nov 18 13:52:50 overo is cooking too much on its own Nov 18 13:53:53 XorA: the deps seems ok (DEPENDS = "ncurses" like Gentoo) Nov 18 13:54:53 morning Nov 18 13:55:27 hi kergoth slept well? Nov 18 13:56:14 quite Nov 18 13:56:39 lucky :) Nov 18 14:00:50 Tartarus: heh, the little one keeping you guys up late, i assume :) Nov 18 14:01:11 last night was actually pretty good Nov 18 14:01:14 1130 -> 0500 Nov 18 14:01:29 and it wasn't a "FEED ME NOW!!!" right at 5 either Nov 18 14:01:42 Tartarus: how old? Have a 1 year old here. Sleeps 8:30-6:30 about every night ;) Nov 18 14:02:24 foerste haha I have 4 old here sleeps from 10.00pm to 7.30 Nov 18 14:02:36 foerster, she's almost a month old now Nov 18 14:02:45 Then my son is 2 Nov 18 14:02:50 oh wow, a fresh one! Nov 18 14:02:53 and is up around 0600 Nov 18 14:03:35 we're really hoping that she takes after how he used to be, he used to sleep 7p->8a or better in his crib Nov 18 14:03:43 in his real bed? no, he's up when he wakes up the first time Nov 18 14:04:27 03Graeme Gregory  07org.openembedded.dev * raf4bfb02ca 10openembedded.git/recipes/procps/procps_3.2.8.bb: Nov 18 14:04:27 procps_3.2.8.bb : kludge linker ordering Nov 18 14:04:27 Was facing a linker ordering issue on Fedora 14 which prevented this recipe building. Adding the libproc as last thing to get linked "fixes" this issue and Nov 18 14:04:27 allows build to complete. Nov 18 14:05:00 xora hehe Nov 18 14:11:42 bleeding edge Nov 18 14:16:05 hi woglinde Nov 18 14:25:55 php's autoconf stuff is a nightmare for cross compiling. Lots of AC_TRY_RUN everywhere, effectively ends up disabling lost of stuff. Nov 18 14:26:46 i think we should seriously consider patching autoconf to disable the 'when crosscompiling' argument to AC_TRY_RUN/AC_RUN_IFELSE Nov 18 14:26:56 better to supply the results via the cache than just pick an arbitrary result Nov 18 14:27:04 just error out instead Nov 18 14:27:19 php makes it even worse -- they unset ac_cv* when checking things! Nov 18 14:27:27 ieehks Nov 18 14:27:31 oww Nov 18 14:27:38 yea, making my life hell right now Nov 18 14:27:39 bad developers, bad, no cookie for you Nov 18 14:27:40 php not only suckz at programming Nov 18 14:28:16 but i do have php5.3.3 with mysqlnd (and dynamic extensions) for gearman finally working Nov 18 14:29:08 but it required some CFLAGS hacks in my recipe to re-enable dlopen support Nov 18 14:31:51 03Michael Smith  07master * re06468eb13 10openembedded.git/recipes/udev/udev_141.bb: Nov 18 14:31:51 udev 141: fix postinst check for /lib/udev/rules.d symlink Nov 18 14:31:51 [ -d foo ] succeeds even if foo is a symlink to a directory. Nov 18 14:31:51 Signed-off-by: Michael Smith Nov 18 14:31:51 Acked-by: Tom Rini Nov 18 14:39:58 Hey guys! I added in a --chroot function --build and --clean so the user can specify more of what to do Nov 18 14:40:04 to my script Nov 18 14:40:13 script? Nov 18 14:40:21 http://neo-technical.wikispaces.com/openembedded Nov 18 14:40:45 http://neo-technical.wikispaces.com/file/view/initial.sh Nov 18 14:41:29 it lets the user build openembedded in an easy to use chroot environment so the user doesnt need to touch his/her system Nov 18 14:43:25 normaly users who build oe dont need initial.sh Nov 18 14:44:32 interesting Nov 18 14:44:44 So you setup an archlinux chroot and then ensure that has the needed host deps? Nov 18 14:45:12 or just extract an archlinux chroot? Nov 18 14:45:37 its actually an arch linux live cd mounted in a folder, by unsquashing the filesystem Nov 18 14:46:35 But do you then ensure that the live cd has the needed host deps with another script? Nov 18 14:46:53 Or are we just lucky and arch linux has them all by default? Nov 18 14:46:59 no chroot.sh covers that Nov 18 14:47:06 initial.sh runs chroot.sh Nov 18 14:48:02 it would be very ugly keeping it all in a single file, nasty sed lines. Nov 18 14:48:35 There we go, interesting concept Nov 18 14:48:58 Rather than per host distro "you need x" type things just grab a small 'normal' linux dist and update i Nov 18 14:48:58 t Nov 18 14:49:30 If you're in a chroot, why does this mess with networking? Nov 18 14:49:46 Or rather, bringing up networking Nov 18 14:49:52 arch linux needs to download packages Nov 18 14:49:53 should just be able to steal the main resolv.conf Nov 18 14:50:34 Right, but the host already has functional networking Nov 18 14:50:41 well the problem is that with DHCP and static IP in arch linux, it has to be in the rc.conf file Nov 18 14:50:45 And inside the chroot you don't need to redo networking fully Nov 18 14:50:54 You just need to have /etc/resolv.conf have valid info Nov 18 14:51:02 (and /proc and /sys mounted) Nov 18 14:51:17 hmm ill test and see if that works Nov 18 14:51:46 Yeah, regular old chroots don't provide a giant separation Nov 18 14:52:54 dont forgot to umount them Nov 18 14:54:51 i havent actually built a full image with this script since my CPU is awfully slow Nov 18 14:55:34 so unmounting without having a finished build would be guess work Nov 18 14:56:52 what is the suggested package to manage an USB-wifi device in OE ? Nov 18 14:57:15 madwifi-ng (broken)? or maybe iw ? Nov 18 14:57:15 oh wow that does work! Thanks Tartarus Nov 18 14:57:40 damnit. anyone used python's multiprocessing module's process pool before? Nov 18 14:57:51 i can create the processes fine manually, but when i use the pool, it hangs waiting for results Nov 18 15:00:47 kergoth dont use it Nov 18 15:00:58 why? Nov 18 15:01:00 its using real new processes Nov 18 15:01:03 no threads Nov 18 15:01:05 sorry X keeps crashing. Nov 18 15:01:12 this sometimes suckz really Nov 18 15:01:15 ? Nov 18 15:01:19 of course it uses processes Nov 18 15:01:22 threads are useless in python Nov 18 15:01:26 global interpreter lock Nov 18 15:01:42 for my case it sucked Nov 18 15:01:55 well, it just cut the parsing time in half Nov 18 15:01:56 a process didnt get closed when parent was closed Nov 18 15:02:00 so i'd say its an improvement Nov 18 15:02:08 and a file was writen until killing Nov 18 15:02:15 user error, you can tell it when the processes should die Nov 18 15:02:24 just pool.close() to stop submission of tasks, and they'll exit on their own Nov 18 15:02:25 no didnt worked Nov 18 15:02:36 hm Nov 18 15:02:56 at least, thats what the docs claim ;) Nov 18 15:03:11 problem is when pool.close() isnt called Nov 18 15:04:09 maybee for oe it might works Nov 18 15:04:23 for my dbus app it somehow didnt Nov 18 15:06:13 you have to run close() or terminate() for the processes to exit Nov 18 15:06:17 you can then call join() to wait for them to die Nov 18 15:06:21 see the docs Nov 18 15:06:38 but it seems broken for me anyway, at least i can't get imap to work Nov 18 15:06:40 map works fine Nov 18 15:06:42 hmm Nov 18 15:16:02 * kergoth gives up Nov 18 15:26:38 someone remembers what command did "pull --rebase" in monotone? Nov 18 15:28:56 03Tom Rini  07org.openembedded.dev * re01784b748 10openembedded.git/ (4 files in 3 dirs): (log message trimmed) Nov 18 15:28:56 help2man: Add native recipe, drop check for being provided. Nov 18 15:28:56 With a recent change to auto* to not build the man pages for Nov 18 15:28:56 native recipes, we can make build progress without help2man Nov 18 15:28:56 being there already. This adds help2man to local.conf.sample Nov 18 15:28:56 as an example of something that is assumed to be provided, Nov 18 15:28:57 drops the sanity.bbclass check and adds a help2man-native recipe. Nov 18 15:29:07 03Tom Rini  07org.openembedded.dev * rc63403cb9c 10openembedded.git/recipes/glib-2.0/ (11 files in 2 dirs): Nov 18 15:29:07 glib-2.0: Fix glib-mkenums to use /usr/bin/env perl Nov 18 15:29:07 Signed-off-by: Tom Rini Nov 18 15:29:09 03Tom Rini  07org.openembedded.dev * r8da17586c5 10openembedded.git/recipes/autoconf/autoconf.inc: Nov 18 15:29:09 autoconf.inc: Use 'which' to find m4 Nov 18 15:29:09 We need to make sure that if m4-native is built, it's used here Nov 18 15:29:09 as there are hosts that otherwise have too old of an m4. When Nov 18 15:29:09 @M4@ is replaced here it's not used in conjunction with #! so this Nov 18 15:29:09 is safe. Nov 18 15:29:10 Signed-off-by: Tom Rini Nov 18 15:32:46 hrw: iirc, you needed to "mtn merge" explicitly after pulling Nov 18 15:35:04 oh hell Nov 18 15:35:10 hrm Nov 18 15:36:01 thx Nov 18 15:52:39 03Martin Jansa  07master * r2631a8e9d9 10openembedded.git/conf/machine/om-gta01.conf: Nov 18 15:52:39 om-gta01: update list of recommended modules and autoload Nov 18 15:52:39 Signed-off-by: Martin Jansa Nov 18 16:09:07 why is help2man trying help4man directory? :) Nov 18 16:09:36 ? Nov 18 16:09:38 what'd I break? :( Nov 18 16:09:56 arg Nov 18 16:09:59 SRC_URI = "${GNU_MIRROR}/help4man/help2man-${PV}.tar.gz" Nov 18 16:10:40 use "${GNU_MIRROR}/${BPN}/${BPN}-${PV} instead Nov 18 16:10:50 should I push fix? Nov 18 16:11:02 doing now Nov 18 16:11:36 ok Nov 18 16:12:04 03Tom Rini  07org.openembedded.dev * r3a0c366031 10openembedded.git/recipes/help2man/help2man_1.37.1.bb: Nov 18 16:12:04 help2man: Use ${BPN} in SRC_URI Nov 18 16:12:04 Martin Jansa's idea (after spotting my typo). Nov 18 16:12:04 Signed-off-by: Tom Rini Nov 18 16:12:38 thx Nov 18 16:12:43 which package provides iwconfig? Nov 18 16:12:51 wireless-tools Nov 18 16:12:58 woglinde: thx Nov 18 16:13:13 mckoan when you have a kernel >= 2.6.34 better use iw Nov 18 16:14:04 wirless-api will die soon completly in the kernel Nov 18 16:17:04 how do you make it so a SRC_URI that calls out a file that has no versioning in its filename gets stored in DL_DIR with version info? Nov 18 16:17:10 woglinde: 2.6.34 Nov 18 16:21:20 tharvey hm Nov 18 16:22:27 have this issue for a recipe that pulls in wireless firmware that has version info in the path, but not hte file name Nov 18 16:23:29 http://www.kernel.org/pub/linux/kernel/people/chr/carl9170/fw/${PV}/carl9170-1.fw - for the various versions only 'carl9170-1.fw' is in my DL_DIR which collides with other versions Nov 18 16:24:33 hello all .. Nov 18 16:24:42 new to oe .. Nov 18 16:25:15 do any one got cross compilation successfull for openjdk for arm Nov 18 16:32:21 vlrk master? Nov 18 16:32:57 woglinde: ..! Nov 18 16:33:25 vlrk master-branch Nov 18 16:33:49 I am one of the openjdk maintainers in oe Nov 18 16:33:56 but didnt test it for a while Nov 18 16:44:36 I don't remember how to avoid checksum control for a package Nov 18 16:45:36 http://paste.ubuntu.com/533912/ Nov 18 16:48:07 I replaced the number in the recipe :-D Nov 18 17:02:43 have a nice rest of the day Nov 18 17:18:43 hi, Nov 18 17:18:53 was something decided for the release? Nov 18 17:19:05 can we still push to org.openembedded.dev or master Nov 18 17:19:07 ? Nov 18 17:19:28 mckoan|away, can you investigate why the checksum changed Nov 18 17:19:32 master isn't being frozen, it's being branched off Nov 18 17:19:47 so feel free to develop as usual, of course attempting to avoid breakage as always Nov 18 17:21:52 dangit, where was that wiki page wiwth the tsc minutes again Nov 18 17:21:57 * kergoth_ kicks wiki.openembedded.org Nov 18 17:29:48 wtf Nov 18 17:29:49 DatabaseError: database disk image is malformed Nov 18 17:30:17 that happened using parallel parsing, when trying to interact with persist_data -- very strange, given sqlite3 does locking Nov 18 17:30:23 i don't see how it could have gotten corrupted Nov 18 17:30:25 * kergoth_ ponders Nov 18 17:30:31 i wonder.. Nov 18 17:30:45 kergoth: http://wiki.openembedded.net/index.php/TSCDecisions Nov 18 17:30:54 pb_: thanks. we should link there from the TSC page :) Nov 18 17:31:01 kergoth: genius! Nov 18 17:31:11 :P Nov 18 17:38:08 kergoth_, ok thanks a lot Nov 18 17:45:46 03Graeme Gregory  07org.openembedded.dev * r3a3e91b7cc 10openembedded.git/conf/machine/ (omapzoom2.conf omapzoom36x.conf): Nov 18 17:45:46 omapzoom2/3.conf : update the serial port config Nov 18 17:45:46 The update of the kernel enabled the omap serial driver so the console has Nov 18 17:45:46 moved port number as 8250 driver now only handls the two uarts on debug board. Nov 18 17:45:56 03Graeme Gregory  07org.openembedded.dev * rca8c7ab1a5 10openembedded.git/MAINTAINERS: MAINTAINERS : now Im active again relist myself as zoom2/3 maintainer Nov 18 17:45:58 03Graeme Gregory  07org.openembedded.dev * r4b2096d6df 10openembedded.git/recipes/sysvinit/sysvinit_2.86.bb: sysvinit_2.86.bb : bump PR for the omapzoom2/3 changes in serial port Nov 18 17:45:59 03Graeme Gregory  07org.openembedded.dev * ra5f336891b 10openembedded.git/recipes/linux/ (4 files in 4 dirs): Nov 18 17:45:59 linux-omap-zoomsync : update the kernel to latest release from TI. Nov 18 17:45:59 Unified the config as both machines are from outside identical and same kernel Nov 18 17:45:59 should work on both. Nov 18 17:48:05 how do you make it so a SRC_URI that calls out a file that has no version in its 'filename' gets stored in DL_DIR with version info? example: http://www.kernel.org/pub/linux/kernel/people/chr/carl9170/fw/1.9.1/carl9170-1.fw gets stored as $DL_DIR/carl9170-1.fw which collides with other versions of it Nov 18 17:48:52 there *should* be a destination filename parameter Nov 18 17:48:55 but i dont' know if there actually is Nov 18 17:49:31 param? on the end of the SRC_URI element you mean? Nov 18 17:50:42 yeah, ;foo=bar Nov 18 17:51:08 hmmm... where would I dig for that? couldn't find documentation on fetcher params Nov 18 17:51:40 the code? ;) Nov 18 17:52:00 tried that... got lost in the 'modularity' of bitbake ;P Nov 18 17:52:20 bitbake/lib/bb/fetch/wget.py and bitbake/lib/bb/fetch/__init__.py Nov 18 17:52:20 found wget fetcher but not where params got delt with Nov 18 17:59:27 hm.. I've read that NCQ + cfq is worst case Nov 18 17:59:48 so NCQ + noop should be used Nov 18 18:00:04 I think I'll benchmark this Nov 18 18:01:36 seems this would only be an issue for wget fetcher and I don't see any support in there for renaming/versioning the object of the fetch - surprised this isn't a more common issue Nov 18 18:01:51 tharvey: there's an email on one of the lists to add it Nov 18 18:02:01 i think its still in my inbox waiting for me to look at it, actually.. Nov 18 18:02:13 heh... a bitbake patch I presume? Nov 18 18:02:41 yeah, probably has to be, unless you talk do_fetch into renaming it after the fact Nov 18 18:02:46 which you could do as a temporary hack Nov 18 18:03:01 ya, thats what I was thinking of doing Nov 18 18:11:06 03Dan Kruchinin  07org.openembedded.dev * r94990e7d1f 10openembedded.git/recipes/perl/ (perl-rdepends_5.10.1.inc perl_5.10.1.bb): Nov 18 18:11:06 recipes/perl: add another missing runtime dependency Nov 18 18:11:06 * Storable uses Carp::Heavy Nov 18 18:11:06 Signed-off-by: Dan Kruchinin Nov 18 18:11:06 Signed-off-by: Roman I Khimov Nov 18 18:11:18 03Chris Larson  07master * rb3d5432cff 10bitbake.git/lib/bb/persist_data.py: Nov 18 18:11:18 persist_data: handle locked db for SELECT Nov 18 18:11:18 Parallel processes interacting with the persist_data db can quite easily Nov 18 18:11:18 explode without this. Nov 18 18:11:18 Signed-off-by: Chris Larson Nov 18 18:11:28 03Chris Larson  07master * r0c8e0f3191 10bitbake.git/lib/bb/ (cache.py cooker.py taskdata.py): Nov 18 18:11:28 Pre-explode rundeps/runrecs in CacheData Nov 18 18:11:28 Signed-off-by: Chris Larson Nov 18 18:11:29 03Chris Larson  07master * r7171019b11 10bitbake.git/lib/bb/cache.py: Nov 18 18:11:29 cache: make use of defaultdict Nov 18 18:11:29 Signed-off-by: Chris Larson Nov 18 18:11:30 03Chris Larson  07master * r3a776ef947 10bitbake.git/lib/bb/cache.py: Nov 18 18:11:30 cache: use set for clean, checked Nov 18 18:11:31 Signed-off-by: Chris Larson Nov 18 18:12:19 kergoth_: how is multithreaded parsing? Nov 18 18:12:31 is there anything to test already? :) Nov 18 18:15:17 Jay7, https://github.com/kergoth/bitbake/compare/master...parallel-parsing-prep Nov 18 18:16:30 Jay7: Mentor has two working implementations now. Nov 18 18:16:35 one a coworker did, one i did Nov 18 18:16:39 heh Nov 18 18:17:10 Jay7: https://github.com/kergoth/bitbake/tree/parallel-parsing - clone that and check out that branch and see how it behaves, if you would Nov 18 18:22:01 wow Nov 18 18:22:51 03Chris Larson  07master * r67984ba0ac 10bitbake.git/lib/bb/cache.py: Nov 18 18:22:51 cache: Add debug msg for a nonexistant dep file Nov 18 18:22:51 If a recipe depends on a file, and that file is out of date, we show a Nov 18 18:22:51 message, but if that file was removed, we do not, until now. Nov 18 18:22:51 Signed-off-by: Chris Larson Nov 18 18:23:01 it's amazing Nov 18 18:23:01 03Chris Larson  07master * r06420ff839 10bitbake.git/lib/bb/cache.py: Nov 18 18:23:01 cache: pyflakes/pep8/pylint cleanup Nov 18 18:23:01 Signed-off-by: Chris Larson Nov 18 18:23:02 03Chris Larson  07master * re084430446 10bitbake.git/lib/bb/cache.py: Nov 18 18:23:02 cache: use new style classes Nov 18 18:23:02 Signed-off-by: Chris Larson Nov 18 18:23:07 but broken a bit :) Nov 18 18:23:16 how so? Nov 18 18:23:21 one moment I'll recheck Nov 18 18:23:24 k Nov 18 18:23:29 if there's an error, i can fix it, just need the error :) Nov 18 18:23:32 not hitting any here right now Nov 18 18:26:04 03Tom Rini  07org.openembedded.dev * r487b3d0c42 10openembedded.git/recipes/binutils/binutils-cross-sdk_2.19.51.bb: Nov 18 18:26:04 binutils-cross-sdk: Add version 2.19.51 (for micro) Nov 18 18:26:04 Signed-off-by: Tom Rini Nov 18 18:30:06 kergoth_: I've a lot of different messages :) Nov 18 18:30:25 Jay7: pastebin please Nov 18 18:30:33 http://pastebin.ca/1995646 Nov 18 18:30:42 this looks like some race Nov 18 18:30:49 wtf Nov 18 18:30:54 hmm Nov 18 18:30:55 yeah it does Nov 18 18:30:56 strange Nov 18 18:30:57 clean tmpdir Nov 18 18:31:10 something about persist_data isn't parallel safe Nov 18 18:31:19 * kergoth_ reviews persist_data code Nov 18 18:31:21 I hit it twice per about 5-6 runs Nov 18 18:31:23 k Nov 18 18:31:49 it could be that its sharing a single database connection between the threads, inherited due to the fork Nov 18 18:31:53 hmm Nov 18 18:31:58 6 cores, 4 BB threads, -j2 Nov 18 18:32:15 it automatically parses with 2.5 * # of cores by default Nov 18 18:32:17 it is still doing something Nov 18 18:32:21 you can specify with BB_NUMBER_PARSE_THREADS Nov 18 18:32:38 I'll break it by ^C Nov 18 18:32:54 hrmm Nov 18 18:33:05 might be not handling exceptions in the child processes well too, need to review that Nov 18 18:33:41 well this time w/o races Nov 18 18:34:17 http://pastebin.ca/1995649 Nov 18 18:35:48 but parsing 7212 recipes in 40 secons is amazing for me now :) Nov 18 18:35:51 hehe Nov 18 18:36:09 only cuts my parse time in half on the 2 core box, but even that is quite lovely Nov 18 18:36:25 task_deps one is just weird. Nov 18 18:36:26 hmm Nov 18 18:36:56 and it loads all cores for 100% Nov 18 18:36:59 nice Nov 18 18:37:28 the default # of parse threads is likely okay for some, but you should tweak for optimal performance particularly if you have fast disks Nov 18 18:37:32 since its all IO bound Nov 18 18:37:56 imho parse threads number == cores number is good default Nov 18 18:38:13 at least for now Nov 18 18:38:18 from my tests, that doesn't perform nearly as well Nov 18 18:38:21 but we can certainly switch Nov 18 18:38:30 <3 multiprocessing.cpu_count() :) Nov 18 18:38:33 may be for 24 cores it will hit into disks io bw limit :) Nov 18 18:38:36 hehe Nov 18 18:40:03 hmmm Nov 18 18:41:45 If a package needs to install into the gstreamer directory on the rootfs, does the package need to explicitly define install -d ${D}/usr/lib/gstreamer-0.10? Nov 18 18:42:21 well, it's not going to magically create all the dirs you're installing into for you Nov 18 18:42:30 it depends, is the asnwer Nov 18 18:42:37 if you're running make install, it likely handles that already Nov 18 18:42:47 if you're manually installing in do_install, yes, you have to create directories before you put things in them Nov 18 18:44:37 oh ok. I was hoping through the use of DEPENDS or similar I could assume the directory had already been created and also looking to make the recipe independent of gstreamer version -- i.e. not need to hardcode the -0.10 Nov 18 18:45:35 well, the thing is, installation is separate from packaging. Nov 18 18:45:50 at do_install time we're installing files into a fresh root where they'll then go into an ipk or rpm or deb Nov 18 18:46:03 the dir won't exist until the packages of your deps are actually installed Nov 18 18:46:19 do_install doesn't directly install into a rootfs Nov 18 18:46:22 does that make it clearer? Nov 18 18:48:00 bb master takes 03:10 to parse Nov 18 18:49:36 somewhat. Looking at the .ipk file, assuming I used the install -D above, ./usr/lib/gstreamer-0.10/. I was just trying to avoid replicating the same install line in several different package recipes and also looking for orthogonality. Nov 18 18:52:17 Jay7: pull and try now if you would Nov 18 18:52:28 Jay7: this isn't a known fix, its a theoretical one, spotted a potential issue Nov 18 18:52:49 for the task_deps issue, that is Nov 18 18:52:53 still working on the race Nov 18 18:52:56 re Nov 18 18:53:12 hey woglinde Nov 18 18:53:22 lol Nov 18 18:53:35 seems someone from qtcreator team lives in my house Nov 18 18:54:01 my wife wondered why the only our house in the street is pixeld at streetview Nov 18 18:54:38 so we googled his name Nov 18 18:54:42 very intressting Nov 18 18:54:47 hah Nov 18 18:55:22 Jay7: okay, pull again, theoretical workaround for the race Nov 18 18:55:33 (i think its the sharing of the connection object between the processes which is hte problem) Nov 18 18:55:47 Jay7: let me know if you manage to see it break again :) Nov 18 18:55:52 in the meantime going to try to find a cleaner method Nov 18 18:55:55 ok, pulled Nov 18 18:56:03 wait a moment ;) Nov 18 18:56:06 :) Nov 18 18:56:17 khem, ping Nov 18 18:56:20 sounds great, 40 seconds Nov 18 18:56:27 hm maybee I will ask him the next time I am seeing him Nov 18 18:56:29 UnboundLocalError: local variable 'appends' referenced before assignment Nov 18 18:56:29 blindvt: yes Nov 18 18:56:35 kergoth_: ^^ Nov 18 18:56:37 Jay7: details if you would :) Nov 18 18:56:46 line number, perhaps? Nov 18 18:56:52 * kergoth_ tries to find it Nov 18 18:57:02 does release-2010.12 sounds ok name for release branch :) Nov 18 18:57:03 http://pastebin.ca/1995668 Nov 18 18:57:11 khem, did you see the sparc quadmath patch? Wasn't gcc about to add libquadmath itself ATM? Nov 18 18:57:12 khem yes Nov 18 18:57:19 khem: sounds good to me Nov 18 18:57:22 khem: yes Nov 18 18:57:35 and hi! Nov 18 18:57:39 all! Nov 18 18:57:52 Jay7: pull & try again :) Nov 18 18:58:02 try and error Nov 18 18:58:03 hihi Nov 18 18:58:06 hehe Nov 18 18:58:28 likewise: hey Nov 18 18:58:36 ok Nov 18 18:59:31 blindvt: yes libquadmath is proposed for gcc Nov 18 18:59:32 http://pastebin.ca/1995671 Nov 18 18:59:43 damnit Nov 18 18:59:46 * kergoth_ glares at bitbake Nov 18 18:59:53 hehe :) Nov 18 19:01:03 okay, thats an easy enough fix, i think Nov 18 19:01:04 hmm Nov 18 19:01:37 clearly i tested more from scratch builds than i did from cache builds :) Nov 18 19:01:56 khem what was what uclibc needs for libcgroup? Nov 18 19:02:07 maybe blindvt has some sparetime Nov 18 19:02:09 still not sure if this is the cleanest way, but i think its cleaner than the obvious alternative Nov 18 19:06:40 hmmm Nov 18 19:09:47 hi, little question. I have an old ipaq h2200 and my kernel is 2.6.21. I have problems with the udev-151 package and i read somewhere the udev-compat141 is made for this problem. But how can i select that one instead of udev-151 when i run bitbake task-base? Nov 18 19:11:39 Jay7: now its inexplicably hanging :| Nov 18 19:11:43 but only when going from the cache Nov 18 19:11:44 hrm Nov 18 19:12:06 oh, i think i see Nov 18 19:12:33 cazze: PREFERRED_VERSION_udev_local = "141" Nov 18 19:12:50 khem thx Nov 18 19:14:53 cazze: or you can add PREFERRED_PROVIDER_udev-compat_local = "udev-compat141" Nov 18 19:14:59 as well Nov 18 19:15:17 ok Nov 18 19:15:44 that the wrong way Nov 18 19:15:52 grrr Nov 18 19:16:16 add h2220 to the machine overides in udev_151.bb Nov 18 19:16:58 XorA: but everytime i sync it will be lost :-( Nov 18 19:17:11 cazze: learn about branches Nov 18 19:17:29 cazze: you can check it in, then git pull --rebase to keep it on top of upstream Nov 18 19:17:43 cazze: and even better submit the patch for inclusion and gain fame Nov 18 19:18:04 XorA: héhé, i have a little head and it is already much to take on 2 days of playing with oe :-) Nov 18 19:18:33 XorA: but as soon as there is place again i'll will try it :-) Nov 18 19:18:49 cazze: start at line RPROVIDES_udev_append = " udev-compat-wrapper" and how to do it should be obvious Nov 18 19:23:03 ERROR: Fetch failed: Unable to fetch URL http://cairographics.org/releases/pixman-0.21.2.tar.gz;name=archive from any source. Nov 18 19:23:07 can someone fix it Nov 18 19:23:44 I'll check, but today I downloaded it OK again Nov 18 19:24:01 I mean on different host than where I prepared that patch Nov 18 19:24:25 03Andreas Oberritter  07master * rc2c8fe4c56 10openembedded.git/classes/pkgconfig.bbclass: (log message trimmed) Nov 18 19:24:25 pkgconfig.bbclass: don't blindly install every .pc file Nov 18 19:24:25 * Instead of installing the file, print a log message. Later, Nov 18 19:24:25 this function could be removed. Nov 18 19:24:25 * .pc files must be installed explicitly to avoid causing Nov 18 19:24:25 build problems. E.g. libsoup-2.4-gnome.pc must not be Nov 18 19:24:26 installed if libsoup-2.4 is built without gnome support, Nov 18 19:24:35 03Khem Raj  07master * ree2644c927 10openembedded.git/recipes/opencvs/opencvs_cvs.bb: Nov 18 19:24:35 opencvs_cvs.bb: Re-arrange SRC_URI so to avoid a CVS fetcher bug where it extracts the end of SRC_URI as path Nov 18 19:24:35 Signed-off-by: Khem Raj Nov 18 19:24:37 khem: and the file is there on that URL :/ Nov 18 19:26:01 JaMa: yes Nov 18 19:27:36 khem: so is there a problem or not? :) Nov 18 19:30:18 Jay7: wow, this sped up the already cached version too, i think. pull and try again Nov 18 19:30:47 JaMa: lemme see Nov 18 19:31:09 JaMa: seems this time it worked Nov 18 19:31:22 JaMa: probably a blip in my connection or somewhere in interner Nov 18 19:31:22 kergoth_: I'm checking now only clean versions Nov 18 19:31:42 wow Nov 18 19:31:50 kergoth_: that fix for opencvs_cvs.bb is interesting we should actually fix bitbake for that Nov 18 19:31:56 it progresses to preparing runqueue Nov 18 19:32:46 Jay7: hehe. progress! Nov 18 19:34:21 kergoth_: http://pastebin.ca/1995703 Nov 18 19:34:35 but it starts the build Nov 18 19:35:24 hmm, okay, i think i just need to make absolutely certain that the parsing threads are dead before the build begins, so it doesn't spot those children going away and think they're tasks Nov 18 19:36:01 they're supposed to exit on their own when we're done handling the files, though, maybe its just a timing thing, taking too long to go away on their own, i should probably explicitly join them Nov 18 19:38:11 hm.. I've started bitbake -p and it is hanging after parsing.. Nov 18 19:38:19 hmm, i thought i fixed that Nov 18 19:38:21 grr Nov 18 19:38:23 it's doing something but show nothing Nov 18 19:38:35 clearly my concurrency skills are rusty :P Nov 18 19:38:50 you have any :) Nov 18 19:38:54 I have almost none :) Nov 18 19:39:22 mostly kernel level stuff, i dont' have mjuch experience with threading or even multiple process work, other than the basics Nov 18 19:40:03 it's still doing something Nov 18 19:40:08 so I'll interrupt Nov 18 19:40:11 k Nov 18 19:40:19 okay, lets try this Nov 18 19:40:37 http://pastebin.ca/1995709 Nov 18 19:40:42 interrupt bt Nov 18 19:40:57 right, okay, i think this should do it Nov 18 19:40:59 pull and try yet again Nov 18 19:41:12 * kergoth_ kicks off more test parsing Nov 18 19:42:15 * kergoth_ 's going to have to do some code cleanup after the last of these issues goes away :P Nov 18 19:42:58 yikes, ^C during the parse is really ugly now Nov 18 19:43:05 need to catch that exception in the child i think Nov 18 19:43:15 oh well, one nstep at a time Nov 18 19:43:31 khem, you around? Nov 18 19:43:56 kergoth_: now it is done ok Nov 18 19:44:11 whew, getting there :) Nov 18 19:44:12 cached too Nov 18 19:44:23 I'll test multiple times ;) Nov 18 19:44:26 you notice that all cached is faster now too? Nov 18 19:44:32 i think thats like 10 seconds instead of 25 here Nov 18 19:44:46 yes Nov 18 19:44:58 one thing you might not have noticed: the progress bar isn't being updated for already cached items, it doesn't show it until its starting the parsing of actual recipes Nov 18 19:45:03 which could be good or bad, i'm not sure Nov 18 19:45:39 might be a good thing, because every one of those events has to get send to the UI to display it, it does it for every single file normally Nov 18 19:45:50 it is starting from 95% :) Nov 18 19:45:54 yeah, hehe Nov 18 19:46:03 the stats are for the total recipes Nov 18 19:46:16 ooh, i'm actually building Nov 18 19:46:17 no errors Nov 18 19:46:19 * kergoth_ does a little dance Nov 18 19:47:52 well, 20 cached parses was fine Nov 18 19:48:00 let's try 20 clean parses :) Nov 18 19:48:16 cool Nov 18 19:48:25 now to fix ^C behavior while parsing Nov 18 19:52:12 Parsing of 7212 .bb files complete (6780 cached, 432 parsed). 438 targets, 19 skipped, 2 masked, 0 errors. Nov 18 19:52:17 this is for cached Nov 18 19:52:22 Parsing of 7212 .bb files complete (0 cached, 7212 parsed). 7794 targets, 329 skipped, 2 masked, 0 errors. Nov 18 19:52:25 and for clean Nov 18 19:52:46 why it reparse some targets? Nov 18 19:52:55 there are recipes that can't be cached Nov 18 19:52:56 is it normal behaviour? Nov 18 19:52:58 e.g. SRCREV usage Nov 18 19:53:01 ah, ok then Nov 18 19:53:03 can't know PV unless it gets it each time Nov 18 19:53:10 so this is expected, and what it did beforee Nov 18 19:54:17 just have not noted before Nov 18 19:54:20 * kergoth_ nods Nov 18 19:54:41 well.. seems bitbake -p is working ok Nov 18 19:54:43 hehe Nov 18 19:54:48 need to decide if i should update parsing progress for caced items Nov 18 19:54:58 ah.. hehe.. Nov 18 19:55:04 cached, blah Nov 18 19:55:06 no, ok Nov 18 19:55:19 * Jay7 was afraid by long pause :) Nov 18 19:55:49 right, exactly, hanging or appearing to hang is not good Nov 18 19:55:50 about 10 parses is ok Nov 18 19:55:58 k Nov 18 19:56:14 kergoth right crtl-c was a problem too with multiprocess Nov 18 19:56:17 in my casew Nov 18 19:56:28 http://www.bryceboe.com/2010/08/26/python-multiprocessing-and-keyboardinterrupt/ may be useful Nov 18 19:56:31 trying it now Nov 18 19:57:04 crtl-c has serval problems with multithread and using more than one system with a mainthread Nov 18 19:57:26 * woglinde spend lot of time at mastherthesis with it Nov 18 20:01:45 hmm... Nov 18 20:01:57 seems I got hang again Nov 18 20:02:11 hmm, k Nov 18 20:02:23 16 run of 20 :) Nov 18 20:02:58 yes, it hangs Nov 18 20:03:35 interrupt bt: http://pastebin.ca/1995726 Nov 18 20:06:17 got it again Nov 18 20:07:00 http://pastebin.ca/1995732 Nov 18 20:08:26 let's build something now Nov 18 20:17:57 03Philip Balister  07org.openembedded.dev * r8b8142421e 10openembedded.git/recipes/tasks/task-sdk-gnuradio-native.bb: task-sdk-gnuradio-native : Add pkgconfig-dev. Nov 18 20:18:08 03Philip Balister  07org.openembedded.dev * ra528a922da 10openembedded.git/recipes/dhcp/ (dhcp4.inc dhcp_4.1.1-P1.bb): dhcp : Create directory for state storage. Convert to INC_PR. Nov 18 20:18:21 damnit, getting a broken pipe with the queues on interrupt now Nov 18 20:18:22 hmmm Nov 18 20:19:19 hm.. Nov 18 20:19:45 noop is making system unusable Nov 18 20:19:50 by the way I gave on crtl+c Nov 18 20:19:51 up Nov 18 20:19:52 now I have deadline Nov 18 20:19:55 signal handling in bitbake is such a mess Nov 18 20:19:56 and use telnet Nov 18 20:20:03 cfq was not good too Nov 18 20:20:04 used Nov 18 20:20:16 Jay7: takking your suggestion and backing off the default to cpu_count() Nov 18 20:20:32 kergoth_: for parse it is good Nov 18 20:20:39 at least for me here Nov 18 20:20:49 it just lagged me a little :) Nov 18 20:20:53 all cores are loaded for 100% Nov 18 20:20:54 could do * 2 instead of * 2.5 Nov 18 20:21:31 and my WD Black is fast enough :) Nov 18 20:21:46 heh :) Nov 18 20:21:47 btw, good idea to test it on WD Green Nov 18 20:22:06 grr, keyboardinterrupt handling still blows in bitbake :| Nov 18 20:22:34 and I think we should port OE to freebsd or even better to solaris Nov 18 20:23:05 because linux with all interactive/realtime patches is unusable under high load Nov 18 20:23:19 well.. may be not all patches Nov 18 20:23:23 but cfq is fake Nov 18 20:23:35 Jay7: try new "sched: auto" patch Nov 18 20:23:45 I have it now and it really makes quite a difference Nov 18 20:23:46 freebsd is behaving better Nov 18 20:23:49 Jay7: http://www.phoronix.com/scan.php?page=article&item=linux_2637_video&num=1 Nov 18 20:23:50 :) Nov 18 20:24:08 Total: 1261 processes, 3681 lwps, load averages: 1165.53, 1269.34, 1460.86 Nov 18 20:24:14 that about solaris :) Nov 18 20:24:14 d'oh Nov 18 20:24:25 it's still pretty usable Nov 18 20:24:39 but signals are coming slowly ;) Nov 18 20:24:48 (that is not my solaris) :) Nov 18 20:24:53 ynezz hehe right Nov 18 20:24:59 I wanted to test the patch too Nov 18 20:25:03 hi woglinde Nov 18 20:25:11 Jay7: try pulling again, ^C should be slightly less screwy now, and the progress bar is updated for cached items Nov 18 20:28:24 kergoth_: started 20 clean parses again Nov 18 20:28:39 k Nov 18 20:30:09 hehe.. after first parse files are cached by vm Nov 18 20:30:31 hehe Nov 18 20:30:34 nice Nov 18 20:30:57 may be good idea to parse with tmpdir on tmpfs Nov 18 20:31:12 it doesn't write the cache until the end, but yeah, you're right Nov 18 20:31:32 hmm, my from cache parse isn't as speedy as it was Nov 18 20:31:34 * kergoth_ checks top Nov 18 20:31:34 and good to test it on some 8-cores machine Nov 18 20:32:12 damnit, why is it back to 25 seconds Nov 18 20:32:18 hm Nov 18 20:32:19 hehe Nov 18 20:35:57 i wonder if the progress bits slowed down the fully cached version Nov 18 20:36:03 better usability, but slower Nov 18 20:36:14 progress bars are for nerds :p Nov 18 20:36:40 kergoth_: wait a bit, I'll test on cached too Nov 18 20:36:42 what's annoying is the server/ui thing, we can't just update the progress display, we have to send an update through the queue to get to the ui Nov 18 20:36:52 hrm Nov 18 20:37:26 kergoth_: update by timer with fake values ;) Nov 18 20:37:32 right in UI :)) Nov 18 20:38:12 hehe Nov 18 20:38:29 it's tempting to just implement a throttle Nov 18 20:38:36 update every N recipes Nov 18 20:39:15 my builds are run bb redirected to tee Nov 18 20:39:21 so, no progress anyway Nov 18 20:39:21 autobuilders would like throttles Nov 18 20:39:25 btw Nov 18 20:39:30 if anyone else wants to try this, clone git://github.com/kergoth/bitbake and check out the parallel-parsing branch Nov 18 20:39:35 kergoth_: why not add cmdline option? Nov 18 20:39:43 Jay7: for # of parsing threads? Nov 18 20:39:44 --progress Nov 18 20:39:48 oh, progress, right Nov 18 20:39:57 or --no-progress :) Nov 18 20:40:04 it already can operate without a progress display, of sorts, when stdout isn't a tty Nov 18 20:40:12 but i'm not sure if its smart enough to stop the progress events Nov 18 20:40:28 its a good idea Nov 18 20:40:44 or --fast :) Nov 18 20:40:57 another idea would be to disable all progress in favor of a "Parsing....done" display automatically when >N recipes were cached Nov 18 20:41:05 since we knkow the parse time will be more sane Nov 18 20:41:34 Parsing [........] done Nov 18 20:41:49 that + a throttle so the . is every N% would be nice Nov 18 20:41:50 and draw a dot per 10-100 recipes Nov 18 20:41:52 yeah Nov 18 20:42:06 i think we need a better way to handle progress, we don't do a very good job of it right now Nov 18 20:42:08 it would work with redirected output too Nov 18 20:42:09 in general Nov 18 20:42:27 have you seen the deal where people use iterators for progress? Nov 18 20:42:34 or may be Parsing... [10%] ... [20%] ... Nov 18 20:42:41 you construct an iterator that automatically updates progress as you move through it Nov 18 20:42:44 pass it a callback or whatever Nov 18 20:42:46 very cute Nov 18 20:42:56 for i in foo: ... do real work ... Nov 18 20:42:58 hm.. Nov 18 20:43:00 and it just shwos progress Nov 18 20:43:03 shows Nov 18 20:43:08 who is listening for events - UI or server? Nov 18 20:43:23 guh, don't get me started :) Nov 18 20:43:34 the UI sends commands to the server, and the server sends events to the UI, for display Nov 18 20:43:37 both directions Nov 18 20:43:50 right now they're the same actual physical process, so its just a queue, a list or something Nov 18 20:43:56 we do have an xml/rpc version though Nov 18 20:44:00 which uses it in both directions Nov 18 20:44:08 the commands are few and far between, no big deal Nov 18 20:44:11 so you can start drawing a dot per 10 seconds and stop by event from server :) Nov 18 20:44:12 but events can be a bit much Nov 18 20:44:18 hehe Nov 18 20:44:25 no updates at all :) Nov 18 20:44:33 but still some progress bar Nov 18 20:44:41 Jay7: http://code.activestate.com/recipes/576986-progress-bar-for-console-programs-as-iterator/ Nov 18 20:45:02 hmm, not the best sample Nov 18 20:45:07 http://pypi.python.org/pypi/progressbar/2.3-dev Nov 18 20:45:12 there, check out the examples on progressbar there Nov 18 20:45:30 note particularly the injection of widgets into the text display :) Nov 18 20:45:33 even rate, etc Nov 18 20:45:37 pretty nifty Nov 18 20:45:54 kergoth_: well, clean parses was fine Nov 18 20:45:57 no hangs Nov 18 20:45:57 cool Nov 18 20:45:59 thats a start Nov 18 20:46:03 not cached Nov 18 20:46:04 still could have bugs lurking somewhere Nov 18 20:46:15 * kergoth_ kicks off a build Nov 18 20:46:32 yes, it slower than before Nov 18 20:46:51 will have to investigate the progress situation as a separate task Nov 18 20:46:53 i think Nov 18 20:46:58 ok, np Nov 18 20:47:01 since that existed both before and after this :) Nov 18 20:47:07 we just didn't realize it Nov 18 20:47:24 imho, you can commit current work to bb master Nov 18 20:47:48 going to do a bit more review / cleanup first, make sure i really like thistion Nov 18 20:47:50 or wait one day, I'll fire testbuilder up this night Nov 18 20:47:51 er Nov 18 20:47:57 this implementation Nov 18 20:48:10 erk, i think my build had too many threads Nov 18 20:48:14 laptop was not happy with me Nov 18 20:48:52 we actually have two mostly working implementations of this, my coworker did a different one Nov 18 20:49:10 but it didn't use multiprocessing, the child processes ended up running the cooker too, led to problems and confusion Nov 18 20:50:02 * kergoth_ reviews all the changes he had to make to fix these bugs to make sure the code didn't get any worse Nov 18 20:53:52 clean parses are fine too Nov 18 20:53:58 just slower Nov 18 20:55:29 should be about the same as without parallel parsing, since it doesn't spawn off workers for those, it just adds the data from the cache directly into its cachedata Nov 18 20:56:31 one issue i see is that the progress event sends along all the statistics, skipped, etc Nov 18 20:56:39 unnecessary, that's only needed for the summary at the end of hte parse Nov 18 20:56:47 progress only needs current / total Nov 18 20:58:39 kergoth: parse in 1.10 is 2m52s, parse with parallel is 1m15s Nov 18 20:58:49 much faster Nov 18 20:58:50 :) Nov 18 20:58:59 glad its working for you Nov 18 20:59:08 foerster: how much cores you have? Nov 18 20:59:17 * kergoth_ should test it on a faster machine, been using his slow work laptop Nov 18 20:59:38 4 core, Xeon i7 Nov 18 21:01:03 developers should be handed slowest machines like I have then you think about making thinks faster :) Nov 18 21:01:15 hehe Nov 18 21:01:27 ok release-2010.12 branch is created Nov 18 21:01:31 :) boss sent me 2 of these beasts, plus a laptop of similar power Nov 18 21:01:49 and testers should have fast machines to catch races ;) Nov 18 21:01:51 foerster: which company do u work for seems to be nice boss and rich company Nov 18 21:01:53 and 4 21" monitors Nov 18 21:01:56 i bought a decent upgrade for my home machine the other day, phenom ii x2 3.5ghz and an ssd, and the damn processor was DOA, might get it back from newegg in another 9 days :| Nov 18 21:02:04 small company called LayerZero Power Systems Nov 18 21:02:08 i work from home too, so it's great Nov 18 21:02:13 kergoth_: why phenom Nov 18 21:02:16 why not i7 Nov 18 21:02:31 wallet :P Nov 18 21:02:38 kergoth_: heh ok :) Nov 18 21:02:45 this was the price of an i5, not an i7 :) Nov 18 21:02:51 imho for compilation they are almost equal Nov 18 21:03:05 * khem has never had more than 50 dollars in pocket since getting married Nov 18 21:03:17 more cores is better than more GHz Nov 18 21:03:21 * kergoth_ had enough trouble talking himself into paying this much to upgrade Nov 18 21:03:26 kergoth_: and performance of i5 too :) Nov 18 21:03:49 the x6 2.8ghz was the same price as the x4 3.5ghz phenom, went with the latter, might have been better to go with the former instead though :\ Nov 18 21:04:03 I go to micro center often these days confused to put a system together Nov 18 21:04:03 khem: my home machine right now is a core 2 duo E4500 Nov 18 21:04:07 almost anything is a good upgrade Nov 18 21:04:10 hehe Nov 18 21:04:13 kergoth_: haha yes Nov 18 21:04:24 I have p4 Nov 18 21:04:56 I might go for an i7 Nov 18 21:05:05 like 720 or 750 Nov 18 21:05:11 nice Nov 18 21:05:32 opteron x12.. :) Nov 18 21:05:34 I am waiting for black friday Nov 18 21:05:42 we make decent money, my fiance and i, but we have a lot to save for too. a kid in the next few years, emergency fund, down payment on a house.. Nov 18 21:05:45 try to budget Nov 18 21:05:50 heh Nov 18 21:06:19 kergoth_: sure. my argument is I have not bought a computer since 2006 really Nov 18 21:06:27 nice Nov 18 21:06:41 i think mine is around there too for the most part Nov 18 21:06:41 last one was MBP with Intel Core2 w/o duo Nov 18 21:06:48 btw, I'm still looking for some remote OE/linux/freebsd-dependent work ;) Nov 18 21:06:51 the doors wouldn't stay on the damn case :) Nov 18 21:06:56 well I did some upgrades like memory and harddisk Nov 18 21:07:26 its too easy to get sucked into the upgrade madness, have to make myself hold back and only even research hardware every few years Nov 18 21:07:32 Jay7: how well versed are you with freebsd Nov 18 21:07:40 Jay7: that is internals Nov 18 21:08:14 khem: about 10 years as sysadmin, ~5 as ISP sysadmin Nov 18 21:08:14 kergoth_: yeah I think every 4-5 years is a good time for new mother board Nov 18 21:08:25 Jay7: btw, there's still definite room for improvement with this branch, its slowed a bit due to the fact that it hands off the tasks one by one, rather than in chunks Nov 18 21:08:28 khem: agreed Nov 18 21:08:30 Jay7: cool. Nov 18 21:08:46 kergoth_: the one I am looking at has SATA3 Nov 18 21:08:58 khem: but still w/o commit bit ;) Nov 18 21:09:00 and I am looking at WD HD with SATA3 Nov 18 21:09:07 but I know core member :) Nov 18 21:09:09 Jay7: ok Nov 18 21:09:33 yeah FreeBSD is lot more about old style of committer and bla bla Nov 18 21:09:46 lot of folks hark about that here Nov 18 21:09:50 khem: current SATA 3 HDD are fake almost.. Nov 18 21:09:53 i picked up a 120gb ssd, will supplement with a couple of wd greens in raid 0 most likely, i should really pick up a good size black at some point Nov 18 21:10:08 I say get some distributed SCM like git,hg and do your stuff you dont need to be a committer Nov 18 21:10:10 they are still in SATA2 speed Nov 18 21:10:30 kergoth_: SSD will be wear out if you do builds on it wont it Nov 18 21:11:01 yeah thats a danger, no idea in real terms what the lifetime is Nov 18 21:11:08 could always stick to srctree builds on it :) Nov 18 21:11:36 kergoth_: about improvements - I'm always ready to test it ;) Nov 18 21:12:05 Jay7: btw release branch is created. Shortly we will send notice to ml but you can start pounding with your builds on it Nov 18 21:12:08 now that it seems solid i'll have to give it some thought Nov 18 21:12:19 and because this is lxc container I can give shell here Nov 18 21:12:22 multiprocessing.Pool automates the chunking, but it seems busted Nov 18 21:12:24 anyone know what 'Qt for Embedded Linux data directory has incorrect permissions: /tmp/qtembedded-0' means? perms look correct to me Nov 18 21:12:28 Jay7: btw. did efikamx images build for you ? Nov 18 21:12:31 at least i can't get it to work Nov 18 21:12:42 khem: yes, last build was successfull Nov 18 21:12:43 kergoth_: SSD's will wear out faster than HD Nov 18 21:12:57 building right now again to test new bitbake Nov 18 21:13:07 well, its flash, thats not surprising, can only erase a block so many times Nov 18 21:13:50 I'll buy second WD Black and create RAID0 to use under OE TMPDIR Nov 18 21:17:35 Jay7: let me know if you run into any more hangs - i worry there might be issues remaining there Nov 18 21:27:06 kergoth_: ok Nov 18 21:27:41 khem: I'll run testbuilder on release branch then Nov 18 21:28:06 ant__: hello, I hope I'll find bug in kexecboot today Nov 18 21:28:08 khem: I have an issue building with uclibc-git Nov 18 21:28:20 s/I'll find/I've found/ Nov 18 21:28:21 hey Jay7 Nov 18 21:28:35 khem: http://tinderbox.openembedded.net/public/logs/task/10346025.txt Nov 18 21:28:40 at least I've patch :) Nov 18 21:28:47 ^^ is fin vs. glibc / eglibc Nov 18 21:33:22 khem: it looks like ffmpeg has same problem Nov 18 21:33:30 http://www.mail-archive.com/uclibc@uclibc.org/msg04949.html Nov 18 21:34:10 Subject: [PATCH] enable log2f and exp2f Nov 18 21:35:09 Attached patch enable those function to allow ffmpeg/x264/... to build Nov 18 21:35:19 ant__: hmmm Nov 18 21:35:42 january stuff Nov 18 21:36:34 shit, just hit a hang with parallel parsing Nov 18 21:36:36 what went wrong this time.. Nov 18 21:36:38 hrm Nov 18 21:36:48 * kergoth_ ponders Nov 18 21:37:16 kergoth_: deadlock ? Nov 18 21:37:29 most likely, have to ensure the queues are empty at the time of the joins Nov 18 21:37:32 parallel parsing? sweet! Nov 18 21:37:42 not sure if there was an exception raised or if it was exiting cleanly, though Nov 18 21:37:55 kergoth_: I'll recheck on tmpfs Nov 18 21:37:55 denix0: git://github.com/kergoth/bitbake, parallel-parsing branch, feel free to try it out Nov 18 21:37:58 Jay7: k Nov 18 21:38:09 kergoth_: was it clean parse? Nov 18 21:38:09 kergoth_: thanks! Nov 18 21:38:16 Jay7: yeah Nov 18 21:38:19 ok Nov 18 21:38:44 khem: seems Applied to master and 0_9_30 branches. Nov 18 21:38:54 does anyone else wonder why bitbake doesn't abort when there's a parse error, right away? Nov 18 21:39:06 it's going to exit out anyway, it won't build with any parse errors Nov 18 21:39:11 so what exactly is the point of parsing everything up front? Nov 18 21:39:17 ant__: We need to select it in .config Nov 18 21:39:17 might as well exit now and let me fix it Nov 18 21:39:42 cool bitbake reports METADATA_BRANCH = "release-2010.12" Nov 18 21:39:49 fantabulous Nov 18 21:40:05 hi, is that normal: Nov 18 21:40:11 pkg-config --variable=libdir libccgnu2 => /usr/lib Nov 18 21:40:15 that's in devshell Nov 18 21:41:29 khem: I see #define HAVE_LOG2F Nov 18 21:41:31 pkg-config --variable=libdir libccrtp1 does the same Nov 18 21:41:35 /usr/lib Nov 18 21:41:37 which results Nov 18 21:42:14 in that: LIBS='-L/usr/lib -pthread -lccgnu2 -ldl -lrt -L/usr/lib -pthread -lccrtp1 -lccgnu2 -ldl -lrt -lcrypto -lpthread ' Nov 18 21:42:23 khem: checking for return log2f(2); in math.h... yes Nov 18 21:46:57 ant__: hmmm can you write a small test case which uses log2f Nov 18 21:47:01 and see if it lnks Nov 18 21:52:23 khem: I was inspecting libm.a Nov 18 21:52:35 and I see Nov 18 21:52:37 log2f.o: Nov 18 21:52:37 00000000 N $ Nov 18 21:53:07 same for exp2f Nov 18 21:54:44 ant__: dump libm.so Nov 18 21:55:00 with readelf -aW and grep for the symbol Nov 18 22:00:30 khem: I don't understand...http://pastebin.ca/1995834 Nov 18 22:01:40 hi kergoth_ , concerning what you told me yesterday for MULTIMACH_ARCH & amend.inc, I don't see the link with base_deps(d) in base.bbclass. May you please explain me this ? Nov 18 22:03:53 kergoth_: got it Nov 18 22:04:18 it hangs again Nov 18 22:08:05 kergoth_: interrupt bt http://pastebin.ca/1995840 Nov 18 22:08:15 this time one ^C was not enough Nov 18 22:11:53 omg.. Nov 18 22:12:23 pgrep python | wc -l -> 43 Nov 18 22:12:55 ericben: what did you want to change in the MULTIMACH_ARCH/amend.inc patch? Nov 18 22:14:10 denix0: here is what kergoth_ told me yesterday : the concept is good, but I think the logic of how MULTIMACH_ARCH is set is slightly more complex than that. Your patch won't catch the case where PACKAGE_ARCH_ is set to MACHINE_ARCH. See lines 58 through 89 in base.bbclass. We should likely shift that either out of an anonymous python function, or move that logic into lib/oe/ or a def'd python function, so we can call it from amend Nov 18 22:14:28 denix0: and I don't see the link with these lines Nov 18 22:15:29 denix0: else the goal is to put lines 358 to 389 of base.bbclass in a function and use this function both in base.bbclass& amend.bbclass Nov 18 22:15:48 in order to avoid duplicated code Nov 18 22:15:48 hmm, interesting. let me check the base class... Nov 18 22:16:55 ericben, hi Nov 18 22:18:49 hi gnutoo I've not forgotten you for the iMX31's kernel ... I hope to find some time for this soon Nov 18 22:19:06 ok nice Nov 18 22:19:13 kergoth_: got it again on parsing angstrom-2008.1/tosa Nov 18 22:19:15 ericben, I've found the response to my question btw Nov 18 22:19:25 I was looking at the wrong kernel Nov 18 22:19:27 for audio ? Nov 18 22:19:32 I was looking at 2.6.27 Nov 18 22:19:41 I should have looked at 2.6.30 Nov 18 22:19:42 for serial Nov 18 22:19:57 ah ok so I don't remember what was the question :-D Nov 18 22:20:08 kergoth_: there was 11 python processes, after ^C two left Nov 18 22:20:09 basically there were 2 serial systems Nov 18 22:20:13 I wondered which one to port Nov 18 22:20:36 2.6.30 has only one Nov 18 22:20:41 so I'll use that one Nov 18 22:20:50 gnutoo: 2 serial system for the iMX ? Nov 18 22:20:55 hm... now next parse started and there is 17 processes Nov 18 22:21:01 at least 2 different CONFIG_SERIAL Nov 18 22:21:05 or something like that Nov 18 22:21:08 let me check again Nov 18 22:21:23 in Freescale's kernel yes, there are 2 drivers, in kernel.org only one Nov 18 22:21:31 ok Nov 18 22:21:36 it was freescale indeed Nov 18 22:22:16 but now in 2.6.37-rc, i.MX31's support is not bad so it will be quite easy to port it Nov 18 22:22:54 CONFIG_SERIAL_8250 and CONFIG_SERIAL_MXC Nov 18 22:23:01 wow Nov 18 22:23:02 good idea Nov 18 22:23:08 and wifi is easy Nov 18 22:23:17 but need to be converted to the new runtime api Nov 18 22:23:20 ok 8250 is for the external UART (16C550 type) Nov 18 22:23:21 for the power Nov 18 22:23:29 ok Nov 18 22:23:30 MXC is for the i.MX's UART Nov 18 22:23:45 so unless you have an external uart on the bus, you don't need the 8250 Nov 18 22:23:55 only CONFIG_SERIAL_MXC appear in their 2.6.30 port if I remember well Nov 18 22:24:35 well.. 16 python processes is normal on parsing.. Nov 18 22:24:51 so here's the question that is left: Nov 18 22:24:59 do I need only serial? Nov 18 22:25:06 for bringup? Nov 18 22:25:17 or do I need something else like gpio etc...? Nov 18 22:25:25 or board fixup Nov 18 22:25:28 or something like that? Nov 18 22:26:19 kergoth_: btw, BB_SCHEDULER is speed here Nov 18 22:26:31 and rm_work inherited Nov 18 22:26:48 JFYI Nov 18 22:27:35 hang again Nov 18 22:27:37 gnutoo: that depends on the hardware. But in you case I think to start only declaring a serial port would be the first step to get kernel log which will allow you to enable (and debug) other devices Nov 18 22:27:48 4 python processes Nov 18 22:28:12 ericben, ok thanks a lot Nov 18 22:28:44 ant__: that says that log2 is there but no log2f Nov 18 22:28:44 I'll try linux-next Nov 18 22:28:50 kergoth_: ps output: http://pastebin.ca/1995856 Nov 18 22:29:04 * khem fixed opie-tasks Nov 18 22:29:45 khem: after first ^C only first two processes left Nov 18 22:29:50 khem: -notes? Nov 18 22:30:41 first message was to kergoth_ :) Nov 18 22:32:09 and second process (pid 7922) is alive after all parses was done Nov 18 22:32:11 denix0: did you check base.bbclaass ? Nov 18 22:32:43 ericben: sorry, have to leave for now, be back in ~30 minutes Nov 18 22:32:51 khem: right :/ Nov 18 22:33:11 that's why I don't understand... Nov 18 22:33:37 denix0: ok have to sleep now, will see that tomorrow, you can drop me a personal message and I'll check this tomorrow morning Nov 18 22:33:41 bye Nov 18 22:35:16 Jay7: yes notes Nov 18 22:35:19 khem: more strangeness... Nov 18 22:35:21 Jay7: I am pushing a fix Nov 18 22:35:40 khem: please push it to release branch too Nov 18 22:35:47 03Khem Raj  07master * r6dd8fa67b3 10openembedded.git/recipes/opie-notes/ (2 files in 2 dirs): Nov 18 22:35:48 opie-notes_1.2.4.bb: Fix build errors encountered with gcc 4.5 Nov 18 22:35:48 * This should get us a step closer to compiling opie-image with gcc 4.5 Nov 18 22:35:48 Signed-off-by: Khem Raj Nov 18 22:36:04 Jay7: I will push it to release branch once you test it :) Nov 18 22:36:08 I need more acks Nov 18 22:36:16 ah.. hm.. Nov 18 22:36:28 Jay7: so follow the steps I suggested to checkout release branch Nov 18 22:36:33 and then cherry pick this fix Nov 18 22:36:42 seems I should add multi-branches support to testbuilder :) Nov 18 22:37:05 ok Nov 18 22:37:10 Jay7: you can simply do git cherry-pick 6dd8fa67b3858353765921a2f648459f7170d653 Nov 18 22:37:33 that will apply this commit to your locally checked out branch based on release. Nov 18 22:37:38 then if it works Nov 18 22:37:53 send me a message and I will pull it into release branch Nov 18 22:37:53 hmm Nov 18 22:38:10 So, virtual:native:xcb-proto needs a dep on python-native Nov 18 22:39:32 Tartarus: for some python bindings ? Nov 18 22:39:46 or modules Nov 18 22:39:53 | checking for xmllint... /usr/bin/xmllint Nov 18 22:39:53 | checking for a Python interpreter with version >= 2.5... none Nov 18 22:39:53 | configure: error: no suitable Python interpreter found Nov 18 22:40:01 I guess it could use it from target python Nov 18 22:40:04 oh I see Nov 18 22:40:51 it seems it does need python but if it needs it at build time to execute then obviously we are left with choice to use python-native Nov 18 22:41:42 having things on lvm is very useful Nov 18 22:41:55 * Jay7 have done snapshot of data under testbuilder Nov 18 22:41:56 looking at C++ errors that gcc 4.5 is finding it seems older gcc versions were nuts when it came to c++ Nov 18 22:42:37 ant__: in some quiteness I will see the log2f issue Nov 18 22:42:56 I see smthg Nov 18 22:43:01 everyone who tests the release branch gets a free one bug fix from me :) Nov 18 22:43:02 moment Nov 18 22:43:42 khem, what if i test it 10 times, do i get 10 free bug fixes ? :P Nov 18 22:43:54 khem, per person or per combo? :) Nov 18 22:43:56 heh 10 combinations yes Nov 18 22:44:24 And, aww, missed the testing tag Nov 18 22:44:33 03Tom Rini  07org.openembedded.dev * r0610b3f1a8 10openembedded.git/recipes/xcb/xcb-proto.inc: Nov 18 22:44:33 xcb-proto.inc: Add python-native to DEPENDS Nov 18 22:44:33 Can't build without python >= 2.5 available. Nov 18 22:44:33 Signed-off-by: Tom Rini Nov 18 22:44:34 khem: http://pastebin.ca/1995865 Nov 18 22:44:34 Tartarus: per person for sure. Per combination I could try Nov 18 22:44:35 khem: Is that the tag you're going to do the release stuff from? Nov 18 22:44:39 grr Nov 18 22:44:45 say 5 combos = 1 more bug fix Nov 18 22:44:46 what am i missing here.. Nov 18 22:44:50 khem: I've got a lotta cpu cycles I'm working on throwing at things :) Nov 18 22:44:59 hmm Nov 18 22:45:06 Tartarus, my understanding is he will update the branch with fixes, the release at the end Nov 18 22:45:12 Tartarus: cool pount is 24/7 with different combos Nov 18 22:45:26 yes we will cherry pick fixes into the branch Nov 18 22:45:36 Crofton, yeah, but i wanted to switch my stuff over to the testing branch today :) Nov 18 22:45:38 the tag is a starting point on master Nov 18 22:45:48 Tartarus: switch to release branch Nov 18 22:45:58 yes its live Nov 18 22:46:02 khem: pull that bugfix in then please :) Nov 18 22:46:13 Tartarus: opie-image one ? Nov 18 22:46:13 half the hosts are rhel5 so no python 2.5 Nov 18 22:46:20 the xcb one i did now Nov 18 22:46:31 * khem kicks fetchmail Nov 18 22:46:53 and woop, getting some actual complete builds now i think :) Nov 18 22:47:13 khem: sry, look at this (3 *libc) http://pastebin.ca/1995872 Nov 18 22:48:08 ant__: ok I see that its missing from libm.so Nov 18 22:48:18 I think we dont have STUBS enabled in uclibc Nov 18 22:48:27 or it may be a real bug Nov 18 22:48:31 khem: So... Nov 18 22:48:40 Give me some combos you want build tested, that aren't atm Nov 18 22:48:51 Or that you were testing you don't have the cycles for :) Nov 18 22:49:16 Tartarus: what machines do you have at home ? Nov 18 22:49:19 khem: once built I'll do a size-comparison Nov 18 22:49:28 of console-image Nov 18 22:49:38 take those as machines and use minimal/minimal-uclibc and x11-image/native-sdk-image Nov 18 22:49:59 Tartarus: I think we need some x86 and ppc testing Nov 18 22:50:01 khem: I can build test tons Nov 18 22:50:04 and boot test little Nov 18 22:50:12 I am sure many people will test arm and its variants Nov 18 22:50:25 Tartarus: lets start with build tests Nov 18 22:50:27 khem: So give me machine/distro/build-targets you want built Nov 18 22:50:27 we should think about qemu-based testing automation Nov 18 22:50:33 And I can queue 'em up and let you know Nov 18 22:50:33 then we can move into booting mode Nov 18 22:50:40 Tartarus: I plan to do lot of qemu's Nov 18 22:50:53 other than efikamx Nov 18 22:50:57 I can add testing pattern to testbuilder Nov 18 22:51:00 right Nov 18 22:51:00 which is my real work horse Nov 18 22:51:04 khem: atm glibc and eglibc image-sizes are almost equal (glibc one is slightly smaller) Nov 18 22:51:08 but we should have it :) Nov 18 22:51:10 So atm I've got p2020ds and mpc8315e-rdb building a few things Nov 18 22:51:27 Tartarus: ok. Nov 18 22:51:28 And I've got tons of cycles to burn Nov 18 22:51:35 So.. wishlist? Nov 18 22:51:56 Tartarus: we should provide queueing interface ;) Nov 18 22:52:06 Tartarus: use minimal/minimal-uclibc for x11-image/console-image and native-sdk-image Nov 18 22:52:08 I have free cpu time here too Nov 18 22:52:17 that will make it like 12 combos for you Nov 18 22:52:19 Jay7: Once i get a chance to run this past some mgmt folks, I'm going to ask the -members list for ideas Nov 18 22:52:49 oh I also need some folks to test angstrom-2010.x Nov 18 22:52:50 khem: Well, uclibc isn't booting on ppc atm, recall? :) Nov 18 22:52:57 because it uses libtool 2.4 Nov 18 22:53:08 hm.. angstrom-2010.. Nov 18 22:53:10 Tartarus: ok I owe that fix to you :) Nov 18 22:53:15 may be I'll get it Nov 18 22:53:20 but not tonight Nov 18 22:53:20 tinderbox is tinderbox.oe.org right? Nov 18 22:53:27 yes Nov 18 22:53:34 no Nov 18 22:53:35 .net :) Nov 18 22:53:37 Crofton yes oe in expanded form Nov 18 22:53:38 khem, I am firing off angstrom 2010 now Nov 18 22:53:46 Crofton: cool. Nov 18 22:53:48 reporting to tinderbox Nov 18 22:54:02 I have some stuff in an overlay, but I care about the results Nov 18 22:54:03 Crofton: cool and do your most interesting image. Nov 18 22:54:13 then if you have more cycles do some other images too Nov 18 22:54:15 usrp-embedded-gnome-image Nov 18 22:54:23 heh Nov 18 22:54:26 cool. No one other than you will do that Nov 18 22:54:28 so good Nov 18 22:54:30 brain cycles keeping is alos a probelm Nov 18 22:54:36 lol Nov 18 22:54:39 trying to apporve hw today Nov 18 22:55:16 hm.. Nov 18 22:55:26 grg: I guess you will help with qemu-mipsel ? Nov 18 22:55:33 so I will not consider that Nov 18 22:55:33 khem, indeed Nov 18 22:55:48 may be good idea to provide distro/image/machine cube with builders' names and results on intersection :) Nov 18 22:55:57 builders cube :) Nov 18 22:56:11 Jay7: yes infact we could use some sort of spread sheet Nov 18 22:56:14 may be spliced by distro Nov 18 22:56:15 i'd like to get micro/micro-image working Nov 18 22:56:20 *sliced Nov 18 22:56:22 grg even better Nov 18 22:56:35 micro has this flattened out tree structure Nov 18 22:56:38 but i don't know anything about the packaging of eglibc... and it fails Nov 18 22:56:50 (see my post from yesterday) Nov 18 22:56:56 grg: ok we might have to fix it Nov 18 22:57:01 grg: I remember Nov 18 22:57:19 so its complaining about a .so.x coming from a -dev package right ? Nov 18 22:57:25 yes Nov 18 22:57:26 eglibc was built fine on minimal/efikamx/console-image about hour ago Nov 18 22:57:36 khem: how is it eglibc images are not smaller than glibc ones? Nov 18 22:57:39 i couldnt see why a libc6-dev package was being installed Nov 18 22:57:48 ant__: eglibc is 2.12 Nov 18 22:57:56 and glibc prolly is 2.9 Nov 18 22:58:02 so eglibc has more code per say Nov 18 22:58:07 may be more syscalls etc. Nov 18 22:58:30 and I think we use -ggdb3 Nov 18 22:58:37 which could disable some opts Nov 18 22:58:56 but there is a nasty problem with gcc-4.5 that ggdb3 hides Nov 18 22:58:59 in ld.so Nov 18 22:59:00 yes, glibc is 2.9 Nov 18 22:59:10 ant__: compare with eglibc-2.9 Nov 18 22:59:19 but still surprised Nov 18 22:59:33 pls ty to fix uclibc ;) Nov 18 22:59:33 hi all, does anyone know how to resolve this error with a python library?: ERROR: QA Issue with omniorbpy-dev: No GNU_HASH in the elf binary: '/home/mekoniak/oe/build/tmp/work/armv7a-angstrom-linux-gnueabi/omniorbpy-3.4-r0/packages-split/omniorbpy-dev/usr/lib/python2.6/site-packages/_omniConnMgmtmodule.so.3.4' Nov 18 22:59:42 yeah Nov 18 22:59:44 hang on Nov 18 22:59:56 juggling here Nov 18 23:00:07 stupid packages not respecting LDFLAGS Nov 18 23:00:38 ant__: we should, oh btw. uclibc git might be our best bet Nov 18 23:00:40 to test on Nov 18 23:00:51 yes, uclibc-git Nov 18 23:00:58 sorry for the imprecision Nov 18 23:01:07 ant__: np Nov 18 23:01:23 ka6sox: you pinged me earlier Nov 18 23:02:13 anyone remind me the rough from of the GNU Hash fix so i can find an example commit Nov 18 23:03:13 Crofton: look at wrong LDFLAGS Nov 18 23:03:19 Crofton: saw this that you commented on: http://old.nabble.com/No-GNU_HASH-found-in-elf-binary-td23072960.html Nov 18 23:03:20 well.. testbuilder is fired Nov 18 23:03:48 angstrom-2008.1 + minimal for collie, akita, tosa, efikamx, ben-nanonote Nov 18 23:03:57 console/x11/opie-images Nov 18 23:03:57 figured it didnt work in my case because of it being python and not C Nov 18 23:04:31 mekoniak: .so isn't python at all :p Nov 18 23:04:32 progress should be available on oestats server :) Nov 18 23:04:53 mekoniak, did you try this? Nov 18 23:04:55 http://cgit.openembedded.org/cgit.cgi?url=openembedded/commit/&id=ac6ff675b74d1fb5d21d51f0718dbc6159564b52 Nov 18 23:05:01 and people check that :) Nov 18 23:05:12 I figured not since its a lib, but I have zero experience with Python :-) Nov 18 23:05:16 gotta run Nov 18 23:05:42 yeah, I did try that Nov 18 23:06:08 OK, added console and x11 image to nas-server and meta-toolchain, was already doing native-sdk-image Nov 18 23:06:21 khem, when do you need garnet again? Nov 18 23:07:05 covering e300c3 and e500v2 targets Nov 18 23:07:09 (ppc) Nov 18 23:07:53 mekoniak, and you cleaned after you made the change? Nov 18 23:08:01 pastebin the entire log Nov 18 23:08:32 ka6sox: oh I was about to pound on it Nov 18 23:08:42 I forced a re-configure... should be sufficient, right? Nov 18 23:08:43 ka6sox: is it needed to be offline Nov 18 23:11:43 Jay7: this is my vicious loop of various combos Nov 18 23:11:45 http://pastebin.com/5ibDHnCE Nov 18 23:12:09 there is no log for do_package. Most recent log is qa_package: http://pastebin.com/1HvPvb91 Nov 18 23:12:18 khem: OK, read your mail now, can I cherry-pick 0610b3f1a82e8f07cad894fb9ad4e075fd9d68fa into release-2010.12? Nov 18 23:12:46 khem, go ahead please...let me know when its done :D Nov 18 23:13:51 ka6sox: it may be days Nov 18 23:13:58 anyway I think we should have full list of supported distros/machines/images somewhere on wiki Nov 18 23:14:04 ka6sox: is it something quick you need then I can wait few hours Nov 18 23:14:24 Jay7: whatever we test will be listed in release notes Nov 18 23:14:26 this allow people to see what combinations are not tested yet Nov 18 23:15:48 Tartarus: sure go ahead Nov 18 23:17:18 khem go ahead...I'll wait. Nov 18 23:18:27 well... Nov 18 23:18:30 * Jay7 -> sleep Nov 18 23:20:37 ka6sox: the machine needs a reboot due to openssl security patch Nov 18 23:21:44 kk...re boot? Nov 18 23:24:16 hrm Nov 18 23:24:58 ka6sox: yes now plz Nov 18 23:25:09 ka6sox: I have applied all the upgrades Nov 18 23:25:23 its running a rm task but no worries kill it Nov 18 23:25:29 I will restart it after reboot Nov 18 23:25:39 kk Nov 18 23:27:41 ugh...forgot how bad satellite connections are with ssh sessions. Nov 18 23:28:18 ka6sox: hrmm so Nov 18 23:28:26 khem, should be coming back up. Nov 18 23:28:38 it garnet waiting for physical touch :) Nov 18 23:29:29 ah its up Nov 18 23:29:30 nice Nov 18 23:48:28 ericben|away, just serial doesn't work Nov 18 23:48:42 as I guessed I needed other stuff too such as memory map Nov 18 23:52:40 khem: ping Nov 18 23:52:45 Bah! Nov 18 23:52:50 * Tartarus hates branches Nov 18 23:53:30 03Tom Rini  07org.openembedded.dev * rb794b9d9be 10openembedded.git/recipes/tasks/task-sdk-bare.bb: Nov 18 23:53:30 task-sdk-bare: Set LIBC_PKGS for uclibcgnuspe Nov 18 23:53:30 Signed-off-by: Tom Rini Nov 18 23:53:44 khem: I was going to ask about ^^^ but put it into 2010.12 first by accident Nov 18 23:53:45 ok? Nov 18 23:55:23 Tartarus: saw that Nov 18 23:56:08 its fine I think the change it sane Nov 19 00:00:34 'nite Nov 19 00:17:50 JaMa: I need your help to identify the cairo issue Nov 19 00:17:58 can you tell me what image should I build Nov 19 00:18:07 and what version of cairo shows the problem Nov 19 00:18:16 MACHINE DISTRO and IMAGE Nov 19 00:18:21 and cairo version Nov 19 00:26:08 a new cairo issue? again? Nov 19 00:26:59 khem: uclibc+x11 image seems to fail due to tls stuff Nov 19 00:27:05 pixman, known? Nov 19 00:35:18 Tartarus: not known Nov 19 00:35:28 jama bumped pxman versions Nov 19 00:35:55 Tartarus: let my builder fail Nov 19 00:35:59 and I will see Nov 19 00:36:09 Tartarus: btw. is it with uclibc-git Nov 19 00:36:19 git should use nptl Nov 19 00:42:37 khem the tracker package needs hall built Nov 19 00:43:06 do you mind I add that and commit to the release branch Nov 19 00:45:05 khem, unmodified sane-toolchain Nov 19 00:45:09 so no, 0.9.31 Nov 19 00:50:15 Sigh, ppc+uclibc is in pretty poor shape atm Nov 19 00:50:26 is -lnsl something you're allowed w/ uclibc/ Nov 19 00:50:39 | powerpc-oe-linux-uclibcspe-gcc -mcpu=8548 -mspe=yes -mabi=spe -mhard-float -mfloat-gprs=double -L/var/hudson/workspace/sb_openembedded_autobuilder/BBPKGS/native-sdk-image/DISTRO/minimal-uclibc/MACHINE/p2020ds/label/sb-u804/tmp/sysroots/ppce500v2-oe-linux-uclibcspe/usr/lib -Wl,-rpath-link,/var/hudson/workspace/sb_openembedded_autobuilder/BBPKGS/native-sdk-image/DISTRO/minimal-uclibc/MACHINE/p2020ds/label/sb-u804/tmp/sysroots/ppce500v2-oe-linux-uclibcspe/usr/lib Nov 19 00:50:39 | /var/hudson/workspace/sb_openembedded_autobuilder/BBPKGS/native-sdk-image/DISTRO/minimal-uclibc/MACHINE/p2020ds/label/sb-u804/tmp/sysroots/i686-linux/usr/ppce500v2/lib/gcc/powerpc-oe-linux-uclibcspe/4.5.2/../../../../powerpc-oe-linux-uclibcspe/bin/ld: cannot find -lnsl Nov 19 00:50:48 oh my, that was a tad longer of a line than I thought Nov 19 00:50:51 ~lart Tartarus Nov 19 00:50:51 * ibot cuts off Tartarus's head with a halberd that could have been a little bit sharper Nov 19 00:57:47 going to leave a mark Nov 19 01:02:00 Crofton: sure go ahead Nov 19 01:02:28 Tartarus: use uclibc-git Nov 19 01:02:40 we might suggest not to use older versions with this release Nov 19 01:02:45 ok Nov 19 01:02:59 I'll push to release and cherry pick to dev Nov 19 01:03:08 something like that, when the build stops ... Nov 19 01:14:15 puleaudio do_install failed Nov 19 01:15:33 I have a package that builds a version.c file based on the output of "git --describe". Is there a reasonable way to make that work with an OpenEmbedded git:-based SRC_URI? Nov 19 01:16:56 (Because OE's build directory has the files but gets rid of the .git directory, that metadata is gone by the time the compile stage runs.) Nov 19 01:25:14 khem: This seutp isn't easy to drop stuff into, so if you want the change, cherry pick into the branch and I'll kick things :) Nov 19 01:25:54 khem: I see you didn't have time for the qemu boards last tag, do you have time this time? If not I can easily toss them in Nov 19 01:29:35 khem, also, should we do something like http://wiki.openembedded.org/index.php/Testing but for the release? Nov 19 01:37:39 anyone know how to get "import platform" to work on python? Nov 19 01:37:52 fails to load on my build Nov 19 01:57:03 So, I managed to get what I want by borrowing some code from gitpkgv.bbclass to build a string that I can use as the GIT_DIR environment variable. Nov 19 01:57:40 It would be handy if OE did that for me, based on having a git:// SRC_URI :) Nov 19 01:58:16 that was quick, i was about to explain that fetching/unpacking will be reworked a bit when the new fetch module is created Nov 19 01:59:39 Jay7: okay, i can't reproduce the hang anymore, with the latest commit. pull and see if it goes away :) Nov 19 02:06:59 03Philip Balister  07org.openembedded.dev * rdd5c67b727 10openembedded.git/recipes/tracker/tracker_0.6.95.bb: tracker : Add DEPENDS on hal. Nov 19 02:09:18 ok Nov 19 02:09:29 this is weird, something in a build fork bombs me Nov 19 02:09:30 http://pastebin.com/MMfjKUDp Nov 19 02:10:36 stupid screen/gterm is truncating **** ENDING LOGGING AT Fri Nov 19 02:59:58 2010