**** BEGIN LOGGING AT Wed Aug 19 02:59:56 2009 Aug 19 05:30:53 good morning Aug 19 08:09:03 sob...cgit,tinderbox and bugzilla passed away... Aug 19 08:09:21 "OE services" Aug 19 08:10:02 ant_work: ? Aug 19 08:10:16 apache down? Aug 19 08:10:35 server ueberlastet? Aug 19 08:12:20 at least ping is answered Aug 19 08:17:34 ah, now it's all on melo.openembedded.org Aug 19 08:20:04 works here.. Aug 19 08:21:17 so..Pacific 1 - Atlantic 0 Aug 19 08:28:08 morning Aug 19 08:49:49 zecke: slow as hell, but cgit reappeared for a while...now has gone again Aug 19 09:04:12 has someone tried to optimize the linux boot process? What can I do for this subject? some ideas? Aug 19 09:12:24 sx: hey Aug 19 09:12:56 sx: check the elinux.org wiki? use init=/my-start.sh to just start what you need Aug 19 09:16:32 zecke, hey, I've read some documents of TI for optimizations on kernel and some ideas for rootfs. They use initng instead of sysvinit, but I don't know how to change it on my rootfs or how to specify it before compile rootfs Aug 19 09:17:47 sx: well, there is no easy way in OE to replace the sysvinit... if you want to change that, it will be a lot of work and would be appreciated Aug 19 09:18:23 sx: if you want your system to boot fast right now, just start what you need... maybe have a small shell script Aug 19 09:21:04 zecke, yep, now it boots in 40 seconds now. I've removed some packages "kernel* update-modules update-rc.d qt-embedded-assistant libqtassistantcliente4 qt-embedded-demos qt-embedded-designer libqtdesignercomponentse4 libqtdesignere4 libqthelpe4 qt-embedded-linguist openssh-sftp-server task-base-bluetooth task-base-wifi task-base-smbfs update-alternatives-cworth wireless-tools wpa-supplicant wpa-supplicant-passphrase bluez4 blueprobe" Aug 19 09:21:19 using opkg remove Aug 19 09:22:25 just I make a system that can play simple videos with sound and contains qt4 support. Aug 19 09:23:47 sx: what apps needs to be started on startup? do you need udev? sshd? Aug 19 09:25:02 sx: e.g. for a simple Qt app... e.g a static /dev is doing well, and the only thing I would starts is myapp -qws Aug 19 09:25:02 sshd - no, but udev yes, because I use my own drivers thru /dev directory Aug 19 09:25:34 sx: and device numbers for this are not static? Aug 19 09:25:43 zecke, yep, it is enough for me. I have mplayer and ffmpeg also Aug 19 09:26:04 zecke, device numbers are fixed Aug 19 09:27:18 sx: i would go with init=/my-init on the kernel command line and just init the stuff I need (and then maybe later call the normal init) Aug 19 09:29:37 zecke, I've searched for init=/my-start.sh but couldn't find nothing. I've no experiances on init process, could you recommend a manual? Aug 19 09:30:04 sx: *sigh* Aug 19 09:30:32 sx: normally the kernel tries to launch /sbin/init... and some others... with init=FOO you can say the kernel should start FOO Aug 19 09:30:40 sx: the content of FOO is really up to you Aug 19 09:33:35 zecke, I understood, thanks. But first I have to learn what is happening in sbin/init, but it is binary file :) Aug 19 09:33:51 I'd wanted a manual for that Aug 19 09:37:47 sx: google for sysvinit? Aug 19 09:38:19 sx: there is not much magic... it goes single init, then its target init level and is running the scripts... Aug 19 09:41:05 zecke, I think it just runs the scripts in etc/rcX.d with a priority specified in the names of the scripts S10.. S20.. right? Aug 19 09:41:30 sx: right Aug 19 09:41:55 then if I remove some scripts or edit them, I can do the same thing, right? Aug 19 09:43:33 in my rootfs in inittab file it is selected runlevel 5, that means on boot up it runs the scripts in rc5.d dir Aug 19 09:43:48 zecke, am I correct? Aug 19 09:43:55 sx: might be :) Aug 19 09:44:53 and exactly it runs the scripts in the rcS.d dir also Aug 19 09:45:37 in rc5.d I have 4 scripts, dbus, syslog, usb-gadget, rmnologin only Aug 19 09:46:44 I thing, the important ones are in rcS.d dir Aug 19 09:55:25 03Steve Sakoman  07org.openembedded.dev * r820d71b058 10openembedded.git/ (conf/checksums.ini recipes/atk/atk_1.27.90.bb): atk: add 1.27.90 Aug 19 09:55:32 03Steve Sakoman  07org.openembedded.dev * rad48abd781 10openembedded.git/recipes/iozone3/iozone3_263.bb: iozone3: fix GNU_HASH related build failure Aug 19 09:55:40 03Steve Sakoman  07org.openembedded.dev * rd53e913c2a 10openembedded.git/recipes/gnome/gnome-desktop.inc: gnome-desktop: fix path to python binary in the gnome-about script Aug 19 09:56:41 zecke, if I remove udev or sysfs, alsa(oss) doesn't work Aug 19 09:56:59 sx: if you believe that... Aug 19 09:57:42 zecke, what you mean? Aug 19 09:57:58 sx: it is a free world, you are allowed to believe whatever you want Aug 19 09:58:32 sx: the truth is, linux systems booted and worked before udev, and even now you can replace udev with mdev of busybox, or just not use it and go back to mknod to statically create device nodes Aug 19 10:00:43 zecke, it'll improve the boot process only with 5 seconds... Aug 19 10:01:04 sx: the point I try to make is... if you want to have a fast boot... don't waste your time with an init system, don't waste your time creating device nodes you don't need, don't waste your time chmoding device nodes you don't need Aug 19 10:01:17 zecke, why does not work alsa when I don't use sysfs for you? Aug 19 10:02:12 zecke, you are right, it's absolutely true Aug 19 10:02:25 sx: strace will give you your answer? I don't know how you define "does not work alsa"... if you execute an application and it fails to play music, then use strace to see which files it opens Aug 19 10:05:23 zecke, so thanks Aug 19 10:05:31 haha Aug 19 10:06:36 zecke, if I remove the udev, it is enough to mknod to statically create device nodes, right? no need to init something? Aug 19 10:07:19 sx: right, you don't need to mount /dev with tmpfs and could create these nodes once and statically Aug 19 10:08:18 zecke, thank you again, I can thing on that Aug 19 10:21:26 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rfa85c7299c 10openembedded.git/recipes/at91bootstrap/at91bootstrap.inc.old: Also commit, one file forgotten in earlier commit Aug 19 10:21:26 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r503c46f523 10openembedded.git/recipes/linux/linux-2.6.30.2/at91sam9263dfc/defconfig: Update at91sam9263 linux config Aug 19 10:21:27 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r0f8f1e5a3c 10openembedded.git/recipes/alsa/ (alsa-state.inc alsa-state_atmel.bb): Make atmel specific alsa configuration/missing files Aug 19 10:21:27 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rd7ae94b221 10openembedded.git/recipes/alsa/ (4 files in 2 dirs): Make atmel specific alsa configuration Aug 19 10:21:30 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r1be899045c 10openembedded.git/recipes/u-boot/ (u-boot.inc u-boot_2009.08-rc2.bb): Change do_deploy to do_install for u-boot, to get the binaries Aug 19 10:21:33 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r0f84439a21 10openembedded.git/recipes/at91bootstrap/at91bootstrap-2.13-rc1/update: The update script to update the fs, after the first boot, should not be in AT91bootstrap Aug 19 10:21:39 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r78d0a25906 10openembedded.git/recipes/at91bootstrap/ (14 files in 4 dirs): useful files provided with at91bootstrap/do proper install Aug 19 10:21:42 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rb39070e522 10openembedded.git/recipes/at91bootstrap/ (at91bootstrap.inc at91bootstrap_2.13-rc1.bb): Allow custom defconfigs in at91bootstrap Aug 19 10:21:45 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r90408d793b 10openembedded.git/recipes/gcalctool/ (3 files in 2 dirs): Add ac HOST_PKG_CONFIG to use host tools instead of target tools Aug 19 10:21:50 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r1ded99b665 10openembedded.git/recipes/xserver-kdrive-common/xserver-kdrive-common/Xserver: Remove -rgba switches, if Xorg is used Aug 19 10:21:53 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rd696835e24 10openembedded.git/ (3 files in 2 dirs): Aug 19 10:21:58 Create an atmel version of x11-gpe-image Aug 19 10:22:00 Select xserver-kdrive-1300 Aug 19 10:22:02 conf/machine/include/AT91.inc defies a number Aug 19 10:22:04 of extra packages, for multimedia. Aug 19 10:22:06 They are removed, and are now available in Aug 19 10:22:08 images/x11-gpe-image_atmel.bb Aug 19 10:22:10 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r7b01097e24 10openembedded.git/recipes/xtscal/xtscal_0.6.3.bb: Explain which X extensions are available, if we do not find XCALIBRATE - update bb file to include Aug 19 10:22:14 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r939cdd34fb 10openembedded.git/recipes/xorg-xserver/ (4 files in 2 dirs): Add xserver-kdrive-1.6.1 (not fully tested) Aug 19 10:22:19 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rd736e444fb 10openembedded.git/recipes/Multimedia/Multimedia-1.0/ (My Mustang.jpg My-Mustang-2.jpg): Add some pictures for testing purposes Aug 19 10:22:26 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * ra8fa0fb1d7 10openembedded.git/recipes/atmel-fixes/ (3 files in 2 dirs): Add script to update fs after first boots, user.jffs2 needs contents though Aug 19 10:22:31 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rf46dd4116d 10openembedded.git/recipes/xorg-xserver/xserver-xorg-1.6.1/ (3 files): add some patches, maybe not useful with new xcalibrate approach Aug 19 10:22:34 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r5997b32f98 10openembedded.git/scripts/make-local.conf.sh: Allow to create a single board conf directory Aug 19 10:22:37 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * re503ee55dc 10openembedded.git/conf/machine/at91sam9g45ek.conf: Add board config for sam9g45ekes Aug 19 10:22:40 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r9168e7c714 10openembedded.git/ (2 files in 2 dirs): Add support for at91sam9g45ekes Aug 19 10:22:47 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rdc81547f9e 10openembedded.git/recipes/at91bootstrap/ (6 files in 2 dirs): Update AT91SAM9G45 RomCode_Replacement to a working version... Aug 19 10:22:50 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * r8fa08985cd 10openembedded.git/recipes/samba/samba-3.3.0/at91sam9263dfc/smb.conf: Add a more useful smb.conf Aug 19 10:22:53 03Ulf Samuelsson  07ulf/linux-2.6.30.2 * rcf35b7954a 10openembedded.git/recipes/xtscal/xtscal-0.6.3/xtscal-0.6.3-verbose-XInitExt-error.patch: Explain which X extensions are available, if we do not find XCALIBRATE Aug 19 11:24:07 msmith_: ping Aug 19 11:37:49 zecke: hi Aug 19 11:40:16 msmith_: according to the mailinglist you would qualify for commit access now Aug 19 11:41:38 msmith_: you will have to send your key somewhere, and iirc you should read the commit policy again as some of your commit messages had the wrong format Aug 19 11:41:47 hello all Aug 19 11:42:47 sweet, i'll take another look.. do you remember which ones had the wrong format? Aug 19 11:43:07 i know sometimes the first line is a little long... Aug 19 11:43:18 msmith_: no, I might be mistaken your patches with someone elses... *sorry* Aug 19 11:43:28 no worries. Aug 19 11:43:30 msmith_: e.g. "file/class: MSG" Aug 19 11:43:34 hi zecke Aug 19 11:44:46 who do keys go to these days? cbrake? Aug 19 11:48:23 msmith_: cbrake, I can do them too, the wiki says mickey :} Aug 19 11:49:23 ok i'll mail you some base64, thanks :) Aug 19 11:49:35 If you post them on the list it won't matter, they're public keys anyway :) Aug 19 11:50:01 cbrake: ping Aug 19 11:50:09 lol Aug 19 11:52:44 gitweb.openembedded.org / cgit.oe.org down? Aug 19 11:52:54 is git still working? Aug 19 11:54:07 seems to be Aug 19 11:56:19 cgit works here (asia) Aug 19 11:57:50 morning Aug 19 11:58:12 zecke: are you using macosx? Aug 19 12:00:38 hrw: once a month Aug 19 12:01:36 zecke: can you suggest any usable terminal program? default one is terrible (default config) Aug 19 12:03:21 uf.. after playing with prefs I got alt key working at least... Aug 19 12:04:07 hrw: I'm using the default one :) Aug 19 12:04:18 hrw: you can build konsole if you want to Aug 19 12:05:18 zecke: this is not mine Mac so I prefer to not build software here. Aug 19 12:05:44 zecke: how to teach default one to send PgUp/PgDn keys? Aug 19 12:06:12 hrw: on my notebook keyboard it like fn + shift + up Aug 19 12:07:15 msmith_: please try to clone now Aug 19 12:07:17 ok.. Aug 19 12:09:58 hmm, permission denied (publickey). Let me go make sure i've got the url right Aug 19 12:10:25 msmith_: git@git.openembedded.org Aug 19 12:10:27 msmith_: check git urls Aug 19 12:10:54 Yeah, that did it. Aug 19 12:11:25 i was trying username msmith, because i can't read. :P Aug 19 12:11:33 cloning now. Aug 19 12:11:40 thanks! Aug 19 12:23:26 03Rolf Leggewie  07org.openembedded.dev * rc251327d68 10openembedded.git/conf/distro/minimal.conf: minimal.conf: build opie 1.2.4 Aug 19 12:23:27 03Rolf Leggewie  07org.openembedded.dev * rad0ba47801 10openembedded.git/classes/base.bbclass: (log message trimmed) Aug 19 12:23:27 base.bbclass: introduce OE source mirror network as last resort for do_fetch Aug 19 12:23:27 Updating recipes and checksums.ini for vanishing upstream hosts or simple Aug 19 12:23:29 restructuring can quickly become a chore for OE devs if taken seriously. Aug 19 12:23:31 The OE user experience suffers severely when at any given point in time Aug 19 12:23:33 4 to 5 recipes on average fail do_fetch for a simple image. Aug 19 12:23:35 The solution is for OE to provide its own caching of upstream sources. Aug 19 12:28:08 Yay!!!!!!!!!!! Aug 19 12:28:17 the source mirror is finally done Aug 19 12:28:29 Laibsch: Thanks :) helps me out quite a bit :) Aug 19 12:28:42 great Aug 19 12:28:49 let me know if there are sources missing Aug 19 12:29:06 I will be monitoring the apache logs, as well. Aug 19 12:29:24 Right now, the sources aren't yet fully complete. Aug 19 12:33:14 Laibsch: how you populated mirror? Aug 19 12:35:14 find . -name *.bb -exec bitbake -c fetch {} \; Aug 19 12:35:21 something like that Aug 19 12:35:36 zecke is working on revitalising oetest for that purpose Aug 19 12:35:47 ko Aug 19 12:35:54 so that hg/git/etc. can be skipped Aug 19 12:48:44 msmith_: thank you for your reaction on the mailinglist about the public ssh key, but I didn't managed to get it working with your code yet (not because it's wrong, but because I'm newbie and not having experience with python) Aug 19 12:48:58 msmith_: thank you for your reaction on the mailinglist about the public ssh key, but I didn't managed to get it working with your code yet (not because it's wrong, but because I'm newbie and not having experience with python) Aug 19 12:49:13 sorry for doubleposting Aug 19 12:49:52 msmith_: should I just add the python part you send at the end of the recipe (like this http://pastebin.com/d7c830fb3)? Aug 19 12:52:13 Laibsch: I will try to write script which will use conf/checksums.ini file to populate mirror fro DL_DIR contents Aug 19 12:52:20 s/fro/from Aug 19 12:52:31 that would be cool Aug 19 12:52:44 and should not be too hard to write Aug 19 12:53:00 it is a matter of composing few scripts into one and adding some glue Aug 19 12:53:33 Are you thus suggesting to exclude sources not mentioned in checksums.ini? Aug 19 12:53:41 Right now, I just export DL_DIR Aug 19 12:53:42 exactl Aug 19 12:54:07 Why not just DL_DIR? Aug 19 12:54:07 Laibsch: my DL_DIR contains propertiary archives so cannot be shared Aug 19 12:54:17 OK Aug 19 12:54:25 In those circumstances, it's different Aug 19 12:54:57 and I have there nearly 23GB in 12368 files Aug 19 12:55:14 including svn/git/cvs/hg/bzr checkouts Aug 19 12:56:05 hrw: Maybe you want to suggest a cherry-pick to the stable branch? Aug 19 12:56:14 ? Aug 19 12:56:18 I think it would benefit even more from a mirror network Aug 19 12:56:37 Laibsch: anyone can suggest to stable/2009 Aug 19 12:57:07 Laibsch: and I am using macosx now on friend's mac - not friendly environment for me Aug 19 12:57:50 Why does "bitbake -f -c fetch ttf-sazanami" end successfully but doesn't fetch the source? Aug 19 12:58:09 because wget gets something insead of 404? Aug 19 13:00:24 but wouldn't that end up in DL_DIR? and shouldn't that throw a checksum error? Aug 19 13:00:34 I see no "NOTE: fetch $URL", either Aug 19 13:00:40 It's pretty mysterious Aug 19 13:09:05 first version of script is doing just checking which files are to be mirrored Aug 19 13:11:21 takes a lot of time with 7148 entries to check Aug 19 13:13:46 03Michael Smith  07org.openembedded.dev * r1eb3a77d18 10openembedded.git/contrib/source-checker/oe-checksums-sorter.py: Aug 19 13:13:46 oe-checksums-sorter.py: speed up about 4X using a hash Aug 19 13:13:46 OK, this saves a grand total of one second a few times a week, but it Aug 19 13:13:46 was bugging me... Aug 19 13:13:46 Signed-off-by: Michael Smith Aug 19 13:13:58 03Michael Smith  07org.openembedded.dev * r3555703d29 10openembedded.git/contrib/source-checker/ (oe-checksums-sorter.py oe-source-checker.py): (log message trimmed) Aug 19 13:13:58 source-checker: usability tweaks Aug 19 13:13:58 checksum sorter, source checker: Aug 19 13:13:58 * exit(1) on failure; Aug 19 13:13:59 * send errors to stderr; Aug 19 13:14:01 * make executable Aug 19 13:14:03 checksum sorter: Aug 19 13:14:17 Laibsch: http://marcin.juszkiewicz.com.pl/download/ - check oe-source-mirror-build.py Aug 19 13:14:56 cool Aug 19 13:14:57 thanks Aug 19 13:14:58 tsjsieb: Yeah, that's what I did. Your recipe looks good to me so there is something else going on that I don't understand. Aug 19 13:15:03 Laibsch: it should generate list of files which have proper md5 sums and are good to be copied Aug 19 13:15:24 Yes, I'm right now looking at audioplayer which does not have a correct checksum Aug 19 13:15:47 Laibsch: use contrib/source-checker/ script to check your DL_DIR first Aug 19 13:15:54 OK Aug 19 13:16:00 Laibsch: there is a script which will tell you which files are broken Aug 19 13:16:30 it is (c) OpenedHand but was written by me during my OH hours Aug 19 13:16:32 ;D Aug 19 13:17:44 msmith_: thanks :) this is the error I get with my recipe: http://pastebin.com/d646b58a7 Aug 19 13:18:07 hrw: how about hardlinks instead of copying? maybe optionally. Aug 19 13:18:35 Laibsch: that script just prints list of filenames. what you will do with it is your choice Aug 19 13:18:41 I see Aug 19 13:19:20 I think that this is a best way. Mirror building is not something you do daily so... ;D Aug 19 13:19:51 and I have to end with 'I do not know Python' mantra... Aug 19 13:20:02 tsjsieb: ah, ok, so my block of code is giving you trouble. Aug 19 13:20:06 hrw: Can I possibly motivate you to take a look at why a poky x11 image boots fine in qemu but an Angstrom or minimal one does not? I tried my luck, but I'm stuck so far. I'm not much of an X expert and that does not help. Aug 19 13:20:13 tsjsieb: Try indenting it with tabs instead of spaces. (one tab = 8 spaces) Aug 19 13:20:15 not that I know Python but I wrote too much stupid code in it Aug 19 13:20:31 Laibsch: qemuarm one? Aug 19 13:20:33 tsjsieb: It gets tacked onto the end of a python func, populate_packages, in package.bbclass so the indentation needs to match. Aug 19 13:20:42 hrw: yes Aug 19 13:20:52 I tried qemux86 but no real difference Aug 19 13:20:58 Laibsch: share me images/kernels as I do not have such ones at all Aug 19 13:21:04 OK Aug 19 13:21:10 Just a second Aug 19 13:21:29 Laibsch: I will not look at them before tuesday anyway Aug 19 13:21:49 running qemu over 1Mbps connection is not comfortable Aug 19 13:27:13 hi Aug 19 13:29:15 I'm testing an image and want to put "things" in rootfs. I want to rebuild only image without repopulate rootfs but when I execute -c build bitbate do first do_rootfs!! is possible force this ? Aug 19 13:30:48 sushisan: images are build by populating rootfs Aug 19 13:31:58 hrw: yes, but I already have my rootfs, I'm only remake the ext2 image Aug 19 13:32:35 that do not change anything Aug 19 13:35:11 yes because I change something in rootfs Aug 19 13:35:43 sushisan: you can't do that, rootfs construction and image building are tied together. Aug 19 13:35:54 hacking tmp/rootfs by hand and attempting to re-image it will not work. Aug 19 13:37:50 :-( Aug 19 13:38:28 msmith_: I replaced al spaces in the file with tabs now, but now I get this error: http://pastebin.com/d17583382 Aug 19 13:38:38 that way of thinking is one of reasons of cleaning tmp/rootfs after creaion of rootfs images Aug 19 13:40:04 ok.. but my idea is not to use my hacked rootfs how final image . is only for testig (in my case the xorg.conf) Aug 19 13:41:21 if you need to fiddle the rootfs by hand, your best option is to mount the image as a loopback device (mount img.ext2 /path -o loop) and hack it that way. Aug 19 13:43:17 the inability to separate do_rootfs from image building is not just an arbitrary policy thing, there are technical reasons why it is not possible to re-use tmp/rootfs for image rebuilds. Aug 19 13:49:04 * * OE Bug 5298 has been RESOLVED (FIXED) by Aug 19 13:49:06 * * Update resolvconf recipe to 1.45 Aug 19 13:49:08 * * http://bugs.openembedded.net/show_bug.cgi?id=5298 Aug 19 13:50:35 tsjsieb: remove the leading '/' from '/home/root/.ssh' in the call to os.path.join() Aug 19 13:50:47 my bad.. Aug 19 13:51:05 pb__: OK...TXS! Aug 19 13:53:13 Who does the gateway from git to the commit ml? Aug 19 13:53:58 c Aug 19 13:55:00 broonie: was that an answer to my question? Aug 19 13:55:46 no Aug 19 13:56:16 Laibsch: is that no a postcommit hook in the git repo? Aug 19 13:56:42 what's the FILESPATH magic to make it look in other overlays for file:// URI's Aug 19 13:57:52 hi from train Aug 19 13:57:55 msmith_: this is the next error that appears: http://pastebin.com/d209a2245 (I hope I don't take too much of your time) Aug 19 14:00:07 msmith_: and this is how the recipe looks now, I think the indenting is o.k. http://pastebin.com/d1ec01577 Aug 19 14:01:53 XorA: I don't see anything Aug 19 14:02:23 tsjsieb why cant you use postinstall for this task? Aug 19 14:03:40 tsjsieb: the indentation is wrong. Try to the match Aug 19 14:03:42 err Aug 19 14:03:47 try to match the one i sent you Aug 19 14:04:00 dirperm_hack() is a function, the 3 lines below it need to be indented one more tab Aug 19 14:04:23 cbrake: would that be you territory? How the mails to the commit ml are formatted? Aug 19 14:04:42 woglinde: I tried it but it doesn't seem to work for me (see mailinglist: http://thread.gmane.org/gmane.comp.handhelds.openembedded/25886 ) Aug 19 14:06:50 th1_: do you use COLLECTIONS? Aug 19 14:07:15 th1_: http://pastebin.ca/1535112 Aug 19 14:08:03 * * OE Bug 5009 has been marked as DUPLICATE of bug 3122 by Aug 19 14:08:05 * * audioplayer-1.0.tar.bz2: do_fetch failed - wrong checksum Aug 19 14:08:07 * * http://bugs.openembedded.net/show_bug.cgi?id=5009 Aug 19 14:12:49 msmith_, yes I use collections Aug 19 14:12:51 sorry got distracted Aug 19 14:13:28 tjsieb okay didnt read the email before Aug 19 14:14:25 great, msmith_ this is exactly what I was looking for, thanks a million Aug 19 14:14:34 where did you find it from? Aug 19 14:17:06 th1_: from my overlay :) Aug 19 14:17:10 hehe Aug 19 14:17:15 i asked the same q on the mailing list a few months back. Aug 19 14:17:34 it should be on the wikipage describing collections Aug 19 14:18:03 basically with that, plus amend-recipes.inc, i can easily break any oe recipe without modifying the source :P Aug 19 14:18:15 amend-recipes? Aug 19 14:18:41 oh, i feel another link coming on Aug 19 14:18:59 http://article.gmane.org/gmane.comp.handhelds.openembedded/25159 Aug 19 14:19:02 it's for "require/includes"? Aug 19 14:19:05 * * OE Bug 5300 has been created by  Aug 19 14:19:07 * * RFE: some changes to the mails sent to commit ml Aug 19 14:19:09 * * http://bugs.openembedded.org/show_bug.cgi?id=5300 Aug 19 14:19:43 it lets you create an amend.inc in your overlay (e.g. myoverlay/recipes/bash/amend.inc) that gets parsed after the main recipe, so you can override or append to do_install(), etc. Aug 19 14:20:09 ah, brilliant Aug 19 14:22:19 these are very nice, they should be bundled in :) Aug 19 14:28:53 msmith_: the recipe seems to work successful now, thank you very much for your help! Aug 19 14:29:25 turns out I was using BBFILES_COLLECTIONS which is not the same as COLLECTIONS apparently Aug 19 14:29:33 so it didn't work out of the box with the FILESPATH.. Aug 19 14:29:55 now I have both COLLECTIONS and BBFILE_COLLECTIONS, don't know if that'll work Aug 19 14:34:10 doesn't seem to cause harmful radiation Aug 19 14:44:28 th1: probably COLLECTIONS will override BBFILE_COLLECTIONS (or maybe vice versa) Aug 19 14:45:02 bitbake doesn't know about COLLECTIONS only BBCOLLECTIONS apparently Aug 19 14:45:21 but BBCOLLECTIONS supports priorities Aug 19 14:46:21 yes COLLECTIONS overrides BBFILE_COLLECTIONS if it's not set already Aug 19 14:46:52 so it's like a user-friendlier interface to BBCOLLECTIONS+BBPATH Aug 19 14:46:56 I guess I don't need the other one then Aug 19 14:48:24 how can I setup a new build environment, and ask it to reuse packages created by an earlier build rather than rebuild everything? Aug 19 14:48:40 that's my biggest issue right now Aug 19 14:48:57 packaged-staging helps there Aug 19 14:49:00 I want developers to be able to check out my overlay, run the setup script, and bitbake whatever, it will then get the packages from the build machine Aug 19 14:49:08 But there's caveats Aug 19 14:49:22 is there any doc on it anywhere? Aug 19 14:49:27 namely, host binaries are cached but path dependent Aug 19 14:50:14 ok Aug 19 14:50:37 can it be worked around e.g. by only using target staging packages? Aug 19 14:50:48 The default path, iirc, for the cache is ${TMPDIR}/deploy/pstage Aug 19 14:50:59 You'll probably want that to be outside of ${TMPDIR} Aug 19 14:51:02 and yes, target stuff is cached Aug 19 14:51:35 ok what do I need to do to set it up? Aug 19 14:51:47 What distro are you using? Aug 19 14:51:51 angstrom Aug 19 14:51:59 or you mean for build? Aug 19 14:52:01 It's already made the caches for you Aug 19 14:52:13 see tmp/deploy//pstage, or deploy/pstage Aug 19 14:52:17 yes Aug 19 14:52:20 Now, re-using the cache... Aug 19 14:52:21 it's made that Aug 19 14:52:37 ok, mv tmp/pstage out of the way, rm -rf tmp, put tmp/pstage back Aug 19 14:52:41 It should Just Work Aug 19 14:52:50 ah Aug 19 14:52:50 Tartarus: hey... Aug 19 14:52:58 hi ant_work Aug 19 14:52:58 talking about the devil... http://projects.linuxtogo.org/pipermail/openembedded-devel/2009-June/011810.html Aug 19 14:53:09 I can't tell it "look for pstage in /srv/global/pstage" or something? Aug 19 14:53:24 th1_, you can Aug 19 14:53:32 Tartarus: I have to answer to you on the ML, just did some more tests before Aug 19 14:53:43 bout /pstage, read ^^^ Aug 19 14:53:47 th1_ , sec Aug 19 14:53:51 ok Aug 19 14:54:03 about my second issue (misplaced /cross), just moving the binaries solves all Aug 19 14:54:10 in local.conf, set DEPLOY_DIR_PSTAGE to /srv/global/pstage Aug 19 14:54:29 ant_work, btw, my mail is semi down right now, need to go kick grub around a bit on my workstation Aug 19 14:54:55 np, it's just wrong-recreated path somewhere... Aug 19 14:55:10 some nasty sed I suppose Aug 19 15:12:46 hi, I've bitbaken xorg-image with angstrom as distro config,usb0 doesn't come up Aug 19 15:12:54 here what I've tried Aug 19 15:13:14 *#/bin/sh ash in a init script,linked in runlevel5 Aug 19 15:13:21 =>keyboard doesn't seem to work Aug 19 15:13:24 worked with SHR Aug 19 15:13:35 so how can I debug the image without the serial cable? Aug 19 15:14:00 I've no serial cable on my htcdream,and I can't build it here(no tools for that) Aug 19 15:14:12 03Stanislav Brabec  07org.openembedded.dev * r8ccde0a871 10openembedded.git/recipes/gpsd/gpsd_2.38.bb: gpsd: Disabled parallel build as it sometimes fail. Aug 19 15:15:50 03Michael Smith  07org.openembedded.dev * r9a29a0aea3 10openembedded.git/recipes/openssh/openssh_5.2p1.bb: Aug 19 15:15:50 openssh 5.2p1: remove DEFAULT_PREFERENCE = "-1" Aug 19 15:15:50 This version has been working well on x86. Aug 19 15:15:50 Signed-off-by: Michael Smith Aug 19 15:15:51 03Michael Smith  07org.openembedded.dev * re05310f495 10openembedded.git/recipes/openssh/ (openssh.inc openssh_4.6p1.bb openssh_5.2p1.bb): (log message trimmed) Aug 19 15:15:56 openssh.inc: cleanup Aug 19 15:15:58 Because of leading slashes in FILES_*, some files were being installed Aug 19 15:16:00 twice: e.g. /etc/ssh/moduli was in both openssh and openssh-sshd. Now Aug 19 15:16:02 openssh is an empty metapackage, which I think was the plan. Aug 19 15:16:04 Other cleanups: Aug 19 15:16:06 * switch to INC_PR Aug 19 15:18:33 msmith_: congrats! Aug 19 15:21:12 haha yes.. baby's first commit Aug 19 15:22:43 i guess i'll slowly leak my patch queue into the tree over the next couple of days.. the ones that got acks anyway Aug 19 15:23:14 msmith_: yeah, it is so much better for me :) Aug 19 15:24:59 yup. Hey, if I sent in a patch for a non-core recipe, let's say mesa-dri, and it didn't get any feedback.. is it ok to push or should i resend and ask for an ack? Aug 19 15:26:29 mmm...what could I do...log commands to file...that would take a lot of time Aug 19 15:27:40 msmith_: feel free to commit Aug 19 15:27:54 anything non-core does not need review Aug 19 15:28:06 Gnutoo: i don't suppose network console would be an option? :P Aug 19 15:28:10 If there are identifiable maintainers, you should of course contact them Aug 19 15:28:19 cool, thanks. Aug 19 15:28:20 like machine maintainers or recipe maintainers Aug 19 15:28:25 over USB that won't work Aug 19 15:28:40 msmith_: take a look at who touched the file with git log Aug 19 15:28:49 that should give you an idea of who to ping Aug 19 15:29:06 yeah, ok. Aug 19 15:29:57 msmith_: oh, i wanted to tell you.. there's a quirk with amend.inc usage & interaction with the bitbake cache. i'm sure you've noticed, but if not, should point it out. creating a new amend.inc won't invalidate the recipe's cache, so if the amend.inc alters provides/depends/rprovides/rdepends, you have to manually touch the recipe or a config file to force cache invalidation to get it to pick up the amended dep changes Aug 19 15:30:18 i have a fix for it, but its both a fix in the metadata and a fix to bitbake, unfortunately :) Aug 19 15:30:23 Oh yeah, i've definitely run into that :P Aug 19 15:31:53 and netconsole would only print kernel messages Aug 19 15:31:55 what i do to fix it is manually inject every amend.inc that *might* exist at some point into the __depends variable, which is what bitbake uses to figure out what files will invalidate the cache (include/require'd files end up there). unfortunately, if you don't have the fix to bitbake as well, it ends up invalidating the cache every time you run bitbake.. not good Aug 19 15:32:00 I need to run commands Aug 19 15:32:10 such as cat /proc/net/dev Aug 19 15:32:30 I'll cat it with a init script...but that would be rather long to do: Aug 19 15:32:38 I also modified amend-recipes.inc to just look in COLLECTIONS + overrides rather than use FILESPATH. It makes parsing a bit faster because it doesn't need to evaluate FILESPATH when rebuilding the cache. Aug 19 15:33:05 extract the sdcard,mount it ,make the script,boot ,shutdown,extract-sdcard mount it etc... Aug 19 15:33:08 moves the work to the package build phase i guess. Aug 19 15:36:37 I'd really like to rework the way we do all of this path stuff.. bbpath, bbfiles, bbfile_*, filespath.. not sure what would be better than this, but something must be :) Aug 19 15:38:55 Heh yeah it's a bit confusing for sure. Works well though Aug 19 15:52:13 pH5: hey! Aug 19 15:52:21 pH5: how is your phd going? Aug 19 15:57:25 hej zecke! Aug 19 15:57:48 spent a lot of time in front of the pc, lately - time to get back into the lab. Aug 19 15:58:32 just passed T minus one year last month. Aug 19 15:58:55 pH5: did you ever managed to build your transistor? Aug 19 16:03:40 msmith_: yeah, that's true. the only reason it is the way it is now is because bbpath/bbfiles seemed the fastest route to go, very little new development effort involved to implement.. i guess its a good thing that we've managed to progress as well as we have given some of the hackish roots :) now to figure out how to do some refactoring without breaking the world.. Aug 19 16:03:48 hehe Aug 19 16:03:51 zecke: unfortunately, no Aug 19 16:05:00 pH5: that is sad Aug 19 16:05:48 CMakeLists.txt is such a lame filename. i think my mom could think of a better one Aug 19 16:05:50 hrmph Aug 19 16:05:53 msmith: was the fix for madwifi dependencies in your patch queue by any chance? Aug 19 16:06:10 madwifi dependencies? Aug 19 16:06:16 th1_: do you mean the kernel staging one? Aug 19 16:06:23 i don't think i have anything specific to madwifi Aug 19 16:06:36 to be honest i haven't even tried to build it since just after i made that patch... Aug 19 16:07:44 msmith_, no the one from Chris Larson "Let the distro control whether we include hostap or madwifi" Aug 19 16:08:13 I had a broken build and pulled the half-done fix from his branch but it's not upstream yet Aug 19 16:08:47 erm, i never pushed that? dangit. okay, its time for me to go through all of this mvl6 stuff and figure out what needs pushing Aug 19 16:08:58 pH5: oh :( Aug 19 16:09:29 http://patchwork.openembedded.org/patch/884/ Aug 19 16:09:34 but there was a typo in it Aug 19 16:09:45 dbug -> dbus Aug 19 16:09:52 oh yeah.... i totally forgot about that Aug 19 16:09:54 * kergoth shakes head Aug 19 16:11:47 either way it's certainly a good thing for me, so I hope you're going to push it Aug 19 16:13:28 the mvl6 release is happening quite soon, then things should settle down a bit more for me, and i can get back to pushing stuff Aug 19 16:13:41 i really wish git was better at managing cherry picks to long lived branches Aug 19 16:14:44 kergoth: do you know anything about libopt? Aug 19 16:16:08 nope, fraid not Aug 19 16:55:59 Hi! I'm going to port OE to ARM9 Phytec LPC3250 board and now I'm trying to build minimal distro for it. Could anybody tell me, please, where can I find list of packages which are included in this distr? I'm going to make downloading and building tasks parallel. Looked at minimal.conf, but haven't seen much there. Aug 19 16:56:53 the distro doesn't control what recipes/packages are available. Aug 19 16:56:58 the distro makes high level policy decisions Aug 19 16:57:09 look in openembedded/recipes/ to see what you can build Aug 19 16:58:09 mmm...oops it was not ifup but dropbear Aug 19 16:58:39 Hm, looks like I'd better read manual more. I was sure that distr is related with packages list. Thanks, kergoth Aug 19 16:59:32 well, the distro can define things like which init or ssh server will be included by task-base, and both distro and machine can append extra packages to task-base-based images, but they don't control what's available, or even directly what goes into the image Aug 19 16:59:35 Raydan, also look into the images Aug 19 17:03:55 thanks, right now I'm looking at 'Useful target' wiki page Aug 19 17:46:13 hi Aug 19 17:47:48 I'm trying to compile xf86-video-cirrus and alp_driver breaks with: error: xorg-server.h: No such file or directory Aug 19 17:47:53 but I have: Aug 19 17:47:57 tmp/staging/i586-kaeilos-linux/usr/include/xorg/xorg-server.h Aug 19 17:47:59 ---------------- Aug 19 17:48:01 ?? Aug 19 17:48:04 some idea? Aug 19 17:48:05 * * OE Bug 5301 has been created by likewise(AT)gmx.net Aug 19 17:48:07 * * libvorbis configure.in hardcodes -cpu=750 for powerpc Aug 19 17:48:10 * * http://bugs.openembedded.org/show_bug.cgi?id=5301 Aug 19 17:48:33 Well, is it building with -I.../usr/include/xorg ? Aug 19 17:48:40 sushisan: xorg/xorg-server.h != xorg-server.h Aug 19 17:48:44 heh Aug 19 17:48:44 It might be the sources need updating to use Aug 19 17:48:50 or what kergoth said :) Aug 19 17:48:54 :) Aug 19 17:50:02 mmmmmmmm Aug 19 17:50:36 when I made my image I have : ERROR: Multiple .bb files are due to be built which each provide virtual/xserver Aug 19 17:50:46 but local.conf have : PREFERRED_PROVIDER_virtual/xserver = "xserver-xorg" Aug 19 17:51:17 kergoth: maybe have some issue iwth this? Aug 19 18:43:28 bah, I wish you could tell BB a range of threads to have live Aug 19 18:43:44 having 5 do_install or do_package running is different from 5 do_compiles Aug 19 18:43:46 03Leon Woestenberg  07org.openembedded.dev * r9e1ca1d4be 10openembedded.git/recipes/openssl/openssl.inc: openssl: Add linux-gnuspe-powerpc to openssl.inc target list. Aug 19 18:43:56 03Leon Woestenberg  07org.openembedded.dev * r7615b8d9b8 10openembedded.git/conf/distro/include/angstrom-eglibc.inc: Aug 19 18:43:56 angstrom-eglibc: TARGET_OS = linux-gnuspe for specific powerpc machines. Aug 19 18:43:56 Signed-off-by: Leon Woestenberg Aug 19 18:44:20 Tartarus: i think ideally we'd indicate which tasks tend to consume which resources. disk, cpu, network, and indicate how many threads per resource Aug 19 18:44:25 Or some max number of stuffs to have live, and split between passing to make and BB threads to have live Aug 19 18:44:26 Tartarus: I thought there is at most one thread of each type? Aug 19 18:44:29 its what i wanted back when we first started this thing, but never go timplemented Aug 19 18:44:36 afaik, anyway Aug 19 18:44:49 kergoth, maybe that would work too, I dunno Aug 19 18:44:59 but on this box, make -j 8 is sane Aug 19 18:45:04 but -j 8 w/ 1 thread sucks Aug 19 18:45:17 -j6 and 4 or 5 threads is usually a win Aug 19 18:45:20 well, if i know i have a crappy 5400rpm drive on my laptop, i might want one disk thread, 2-4 network threads, and 2 cpu threads, or something Aug 19 18:45:37 (ie the times we have 3 do_compiles going and thus too many make threads is won out by the times it's 5 do_installs) Aug 19 18:45:56 kergoth, yeah, but you can't really break things up that well Aug 19 18:46:07 you could -c fetchall and then be done w/ network, but that's a disk and a little cpu Aug 19 18:46:17 compile is disk and cpu, even if you have -pipe Aug 19 18:46:38 morning kergoth Aug 19 18:46:45 true, maybe you're right and we just need to specify per task Aug 19 18:47:05 X do_compiles, Y do_installs, Z do_fetches, and then set an overall max Aug 19 18:47:07 hey pb_ Aug 19 18:47:28 kergoth, with some prioritizing too Aug 19 18:47:31 trying to keep track of what i have and haven't pushed up from mvl6 to OpenEmbedded and what from OpenEmbedded I've cherry picked into mvl6 is a royal pain in the ass Aug 19 18:47:36 ie if I have 3 compiles going, I want nothing else going Aug 19 18:47:39 long lived branches in git are annoying, cherry pick just isn't smart enough Aug 19 18:49:25 yeah, I don't think I've ever really used an SCM system that handles long-lived branches all that well. you'd think git would be able to do a better job than it seems to, though. Aug 19 18:49:42 Do we have -pipe enabled by default? Aug 19 18:49:47 Well, cherry picking is the problem really. Aug 19 18:50:04 likewise_: I don't think so Aug 19 18:50:34 git-cherry tries to be smart, in that it checks to see if a given patch-id is in both histories, but thats not quite enough, since a diff might have to change in order to be applied in the cherry-pick Aug 19 18:50:40 so you need to store the old patch-id too Aug 19 18:50:44 like in the commit message Aug 19 18:50:48 right Aug 19 18:51:17 pb_: why not? Is it broken in some respect? Aug 19 18:51:37 no, it works fine, it just doesn't buy you much. Aug 19 18:51:47 on most workloads it isn't much of an advantage and it can even be a net loss in some cases. Aug 19 18:51:51 hmm, I didn't know we didn't use -pipe by default.. Aug 19 18:52:14 I guess it depends on how much memory you have really? Aug 19 18:52:19 and how fast or not your disk is Aug 19 18:52:27 It used to make more sense in the past but these days the OS gives you pretty much identical performance without it. Aug 19 18:52:31 Tartarus: in most cases the files in /tmp never hit the disk anyway Aug 19 18:52:46 Especially with tmpfs Aug 19 18:52:59 even if you aren't using tmpfs there, which lots of folks are, they generally just sit in buffer cache rather than getting flushed to the disk itself. Aug 19 18:53:07 ah, true Aug 19 18:53:18 and, if you use make -jN, you get the parallelism benefit as well Aug 19 18:53:39 something to put on my list of stuff to play with at some point Aug 19 18:53:51 so in a lot of cases, the only net effect of -pipe is that you have cc1 and gas both sitting in RAM all the time rather than just one or the other. Aug 19 18:54:02 if you have enough ram then that isn't a big deal, but neither is it a win Aug 19 18:54:57 also, in the days when -pipe was invented, cpp was a separate process. now that it's integrated with cc1, the benefit of -pipe is further eroded. Aug 19 18:55:39 yeah, but it's easy enough to toss in and test on a full build and see if it comes out as anything more than noise Aug 19 18:55:40 still, there probably are some cases when it's a win, it's just not all that easy to predict when that will be true. it'd be interesting to see some benchmarks. Aug 19 18:55:59 right Aug 19 18:56:13 What flag would be global enough to add "-pipe" to, to affect all builds? Aug 19 18:56:29 likewise_: you could add it to ${CC}, I guess. Aug 19 18:56:34 I can report some from-scratch builds from a build server. Aug 19 18:56:40 or the global CFLAGS Aug 19 18:56:49 pb_: thanks. Aug 19 18:56:59 Personally I'd feel safe just doing what the kernel does, they care about this stuff a lot. Aug 19 18:57:18 or even patch the gcc driver to make -pipe the default for your test build Aug 19 18:57:24 erk, 362 commits still listed that i might need to push upstream yet.. this makes me sad. Aug 19 18:57:32 broonie: heh, I never feel safe doing what the kernel does. Aug 19 18:57:32 * kergoth grumbles Aug 19 18:58:41 kergoth: doh Aug 19 18:58:47 kergoth: i wonder what would happen if you tried to rebase your branch to the latest org.openembedded.dev Aug 19 18:58:52 maybe rebase is a little smarter than cherry? Aug 19 18:59:05 good question Aug 19 18:59:28 I'd have to generate some more fake merge commits to manage it though Aug 19 18:59:36 see, this collection doesn't have all of upstream oe, just certain recipes Aug 19 18:59:47 oh, i see. Aug 19 18:59:50 i generated a fake merge commit at the point where i copied most of them, so i can get a slightly more sane list Aug 19 18:59:54 :) Aug 19 18:59:59 heh. mad git-fu Aug 19 19:00:52 One of the things that made me like OE when I was looking into it was that it had been through a few SCM migrations, and the history had mostly survived. Aug 19 19:01:04 what I'd kind of like to do is go over each file in this repo, get a hash of the git object for it, then go to the other branch and see if that object exists, and if so, find the first commit that points at a tree containing this object, and exclude the commits prior to that from my list of commits for that file Aug 19 19:01:35 since i dont particularly care how i got there, as long as the other branch has gotten there Aug 19 19:01:43 hrm yes.. that would get you somewhere.. Aug 19 19:02:09 that way if a set of commits were squashed when they went into the other tree, I'd manage to exclude that whole set of commits.. Aug 19 19:02:09 hmm Aug 19 19:03:01 i've seen quite complicated answers to git questions on stackoverflow... I wonder if someone else in the world is running up against this and has a crazy script for it Aug 19 19:03:30 there've been multiple threads on the git list in the past talking about this Aug 19 19:03:48 http://lists.zerezo.com/git/msg667742.html Aug 19 19:04:16 but, that particular one was a proposal to modify the the git object headers rather than building on top of git with a new porcelain, so the git folks shot it down Aug 19 19:04:20 (can see their point) Aug 19 19:05:28 for now i just have a blacklist file with a list of commits i know exist in both branches, and then a script that does a git log without those.. better than nothing, but still suboptimal for a workflow Aug 19 19:05:30 ah well.. Aug 19 19:07:08 the annoying part of a workflow like this is that you can't just inject extra bits into the commit message to keep track, because that metadata will be removed when something gets pushed upstream :) Aug 19 19:08:53 you need some way to steganographically hide the metadata. So it inserts typos into your commit message, and the xors of the typos combine to give you a local commit id :P Aug 19 19:09:12 then when it gets pushed upstream the data is still there... but not bothering anybody Aug 19 19:09:53 hehe Aug 19 19:10:26 well, if its just one person doing the work, could just store the metadata locally the way I'm doing for the hash blacklist, write wrappers for the git commits that manipulate/use it Aug 19 19:16:47 do any of you guys use packaged_staging for developer builds? Aug 19 19:17:26 what do you mean by developer builds? Aug 19 19:17:54 anyone here a udev expert, by chance? Aug 19 19:17:59 I mean for the developer to checkout a repo and then do a build of the packages he/she is working on but faster because they don't have to build all the build-deps first Aug 19 19:18:06 I know a fair bit of udev Aug 19 19:20:08 oh, we use it in that way, th1 Aug 19 19:20:22 mvl6 lets you build from prebuilt binaries, its just pstage + changes Aug 19 19:20:38 hi all: does anyone know how to continue a oe build after it has failed given you have removed the recipe that was failing? Or how can I leave the recipe in but have oe skip it and continue with the other recipes? Aug 19 19:20:41 we added some things to it, but most of those things were trivial Aug 19 19:20:44 so how do you handle the path problem in pstaging? Aug 19 19:20:55 use a chroot? Aug 19 19:20:56 which path problem? :) Aug 19 19:21:07 with the staging packages having to be in the same path as they were built for Aug 19 19:21:32 we've handled it on a case by case basis, making the native and cross packages relocatable. in some cases using wrapper scripts, in others by overriding paths via env vars (i.e. autotools), etc Aug 19 19:21:55 ok Aug 19 19:22:02 also used $ORIGIN in rpath in BUILD_LDFLAGS Aug 19 19:22:26 sure would be nice if more open source projects were properly relocatable Aug 19 19:22:33 yes Aug 19 19:23:00 I've been thinking of a fakechroot style thing Aug 19 19:23:16 its easier for people who maintain their own collections to make it functional in this way, just because of the limited number of native/cross stuff Aug 19 19:23:30 yes Aug 19 19:23:41 itd be really cool to use OpenEmbedded to produce a rootfs to run OpenEmbedded in, and ship that rootfs as a live cd, chroot environment, vmware player image, whatever Aug 19 19:23:49 ideally I'd prefer not to maintain a private collection other than for the packages that are specific to our project Aug 19 19:24:07 in fact, having to maintain lots of hte stock packages is probably a show stopper for us adopting it in the project Aug 19 19:24:32 that makes life easier, though you dont get the benefits of the speedy as hell parse time due to using a minimal collection instead of upstream Aug 19 19:24:43 doing that has spoiled me, now i get all annoyed building upstream :) Aug 19 19:24:45 true Aug 19 19:25:17 but maintaining a distro is not the main focus of our project :) Aug 19 19:25:23 oh, i should push my newcollection.inc upstream at some point. you run bitbake -c newcollection bash (with the right vars set in local.conf), and it deploys bash and all of its deps into your new collection, including filespath bits Aug 19 19:25:33 this may be a stupid question but would it be possible to run multiple concurrent parse threads when rebuilding the cache? Aug 19 19:25:52 i think richard was wanting to do that, i think its possible and not done right now Aug 19 19:25:55 that sounds very handy Aug 19 19:26:20 so I could do bitbake -c newcollection xenclient-iovm-image :) Aug 19 19:26:41 and I'd go from parsing 7000 files to some 700 Aug 19 19:27:02 yep, exactly Aug 19 19:27:25 I thought about doing it another way, have a script delete all the bb's that don't contribute to the build Aug 19 19:27:47 but tbh the parsing overhead is only a problem when doing one or a few pacakges Aug 19 19:28:27 if touching local.conf didn't invalidate the world, i wouldn't mind so much Aug 19 19:28:28 heh Aug 19 19:28:29 btw for making an image.. is it possible to use a "package repository" as an overlay where it'll pull all the precompiled packages from a URI instead of building them locally? Aug 19 19:29:16 e.g "foobar-1.0.2-r7 is in repo so pull it instead of build it" Aug 19 19:29:28 when building the image Aug 19 19:30:54 good question Aug 19 19:30:59 re Aug 19 19:31:38 that would be extremely useful Aug 19 19:32:05 for developer builds... and for an auto image builder that feeds from a world-builder Aug 19 19:32:35 e.g. every 5 minutes get the list of packages at the worldbuilder, if it's changed then build the images Aug 19 19:33:19 if there's nothing like that already I might work on it after our alpha release when I have time Aug 19 19:34:52 florian: good evening Aug 19 19:35:20 hey pb_ Aug 19 19:35:58 * kergoth git rev-list HEAD..foundation-sync -- conf|while read hash; do if ! grep -q $hash .git/cherry/*; then echo $hash; fi; done|git diff-tree --stdin --pretty=medium|tig Aug 19 19:36:51 git cherry-list foundation-sync ;) Aug 19 19:37:53 kergoth: that newcollection sounds VERY handy, also would let you just package the metadata you need for a given image, that could be handy. Aug 19 19:38:13 i'll see about pushing it this week, there won't be any objection to it Aug 19 19:38:34 I don't see how it could break anything if you don't use it ;) Aug 19 19:39:04 i meant objection from mvista :) policy is push everything unless there's a very specific reason to do otherwise, which there won't be here Aug 19 19:39:34 even bitbake -c newcollection world would be useful as it would get rid of old versions and the like Aug 19 19:40:01 Citrix policy is similar Aug 19 19:41:39 btw do you work with any of the old MIPS UK guys? Aug 19 19:41:50 the only troublesome bit is filespath, since OVERRIDES will vary from build to build. would have to make it slightly smarter for that case, have it grab the actual *dirs* in filespath for each entry thats relative to FILE_DIRNAME Aug 19 19:41:51 I know some of us went to work for mvista Aug 19 19:41:55 Question, is http://pastebin.com/d5aa3e64d (lack of LEAD_SONAME and a few missing libs from the package) anything to worry about (ARMv7 EABI)? Aug 19 19:43:06 kergoth: I did not realise your mvista. Aug 19 19:43:22 DJWillis, it's a bug but it probably won't affect you Aug 19 19:43:45 those should be packaged separately AFAICS Aug 19 19:44:13 DJWillis: yep, I'm on the mvl6 team Aug 19 19:44:32 th1: I did not think it was an issue that was going to hit me but the first time I have noticed it so I figured I would check. Aug 19 19:45:05 any opinion on the add of a new patchtool to patch.bbclass that will apply patches to the tree using git-apply rather than patch or quilt? can be useful, since git patches can carry extra bits, like permissions info Aug 19 19:45:16 we have it locally, just wondering if its worth pushing Aug 19 19:45:58 DJWillis, the memusage lib is a debugging tool, the other one I think is for international domain name<>utf-8 trnaslation Aug 19 19:46:16 unless another package breaks from lacking one of htem I don't think you have to worry Aug 19 19:46:30 kergoth: I have a few people I know looking at mvl6 as an option for 'commericial OE'. Aug 19 19:46:38 glad to hear it Aug 19 19:47:41 if someone whines to me about how hard OE is, I'll certainly steer them to mvl6 Aug 19 19:47:50 kergoth: and on that patchtool question, gets a +1 from me as I had an odd issue with GIT patches and permission changes that it would have prevented. Aug 19 19:47:58 kergoth, any idea when there wil be price info? Aug 19 19:48:03 DJWillis: k, will add it to the list Aug 19 19:49:45 03Stanislav Brabec  07org.openembedded.dev * rcc7cd41c0f 10openembedded.git/recipes/util-linux-ng/util-linux-ng.inc: util-linux-ng: Fixed empty packages. Provide util-linux. Aug 19 19:49:56 th1: thanks for that info, I did wonder what memusage was for. Aug 19 19:50:13 Crofton|work, I'd not whine that it's hard but that the documentation is not quite as good as it could be ;) Aug 19 19:52:02 heh Aug 19 19:53:34 03Stanislav Brabec  07org.openembedded.dev * r90d97f3fb9 10openembedded.git/conf/distro/angstrom-2008.1.conf: angstrom-2008.1.conf: Prefer util-linux-ng over util-linux. Aug 19 19:55:00 03Leon Woestenberg  07org.openembedded.dev * re28e9d0b52 10openembedded.git/recipes/linux/linux-2.6.30/calamari/defconfig: Aug 19 19:55:00 linux-2.6.30: calamari/defconfig CONFIG_WIRELESS_EXT=y for madwifi. Aug 19 19:55:00 Signed-off-by: Leon Woestenberg Aug 19 20:26:30 florian: Would you be so kind to delete my @openembedded.org account? I receive nothing but spam. Aug 19 20:29:16 03Chris Larson  07org.openembedded.dev * raea9acb329 10openembedded.git/ (7 files in 5 dirs): (log message trimmed) Aug 19 20:29:16 Let the distro control whether we include hostap or madwifi. Aug 19 20:29:16 - Add base_ifelse convenience function to base.bbclass. Aug 19 20:29:16 - Replace all conditionals in the recipes relating to hostap and madwifi to Aug 19 20:29:16 look for their name in COMBINED_FEATURES rather than looking for specific buses (pci, pcmcia). Aug 19 20:29:20 - Change the default COMBINED_FEATURES to enable: Aug 19 20:29:22 - madwifi, when: Aug 19 20:33:24 03Chris Larson  07org.openembedded.dev * rac729cee60 10openembedded.git/classes/base.bbclass: Aug 19 20:33:24 base.bbclass: catch bb.MalformedUrl in do_fetch. Aug 19 20:33:24 Signed-off-by: Chris Larson Aug 19 20:50:14 * kergoth makes a commit that checks to see if wget supports --no-check-certificate before using it.. Aug 19 20:56:32 03Stanislav Brabec  07org.openembedded.dev * r3bd6ffb5ba 10openembedded.git/: Merge branch 'org.openembedded.dev' of git@git.openembedded.org:openembedded into org.openembedded.dev Aug 19 20:56:42 03Stanislav Brabec  07org.openembedded.dev * rada6f217e9 10openembedded.git/ (5 files in 2 dirs): swfdec: Fixed dependencies and package splitting. Disabled static. Packaged stable 0.8.2. Aug 19 21:34:58 Laibsch: git ml notifications: /home/git/git-notify Aug 19 21:35:20 Laibsch: I'm not sure where it originally came from, I only tweaked it during the server move Aug 19 21:35:54 kergoth: will this patchtool replace quilt ? Aug 19 21:36:06 it isn't default Aug 19 21:36:14 you enable it in a recipe that wants it Aug 19 21:36:26 we may want ot think about enabling it globally though, quilt is actually a fair bit slower Aug 19 21:36:31 something to think about Aug 19 21:36:34 for the future Aug 19 21:37:00 quilt is nice to develop patches Aug 19 21:37:07 cbrake: thanks. Aug 19 21:37:07 what tool is it ? Aug 19 21:37:17 cbrake: Maybe you can take a look at Aug 19 21:37:19 !oebug 5300 Aug 19 21:37:21 * * Bug 5300, Status: UNCONFIRMED, Created: 2009-08-19 14:05 Aug 19 21:37:22 * * : RFE: some changes to the mails sent to commit ml Aug 19 21:37:23 * * http://bugs.openembedded.org/show_bug.cgi?id=5300 Aug 19 21:37:24 ? Aug 19 21:38:05 yeah, quilt is nice for that. but, we could default to git-apply or patch, and the user opt-in to the use of quilt for the specific recipe they're working on Aug 19 21:38:09 PATCHTOOL_pn-foo = "quilt" Aug 19 21:38:18 then hop into devshell for foo, or whatever Aug 19 21:38:28 * cbrake likes using git instead of patch for applying patches Aug 19 21:38:51 I've been thinking bout using git to track input/output of tasks in general. capture what do_install installed, what do_compile outputted, etc Aug 19 21:38:51 guilt might be a nice fudge Aug 19 21:39:05 along the lines of tools like clearmake, though itd be in a local repo, not upstream Aug 19 21:44:27 03Michael Smith  07org.openembedded.dev * r1d18a75825 10openembedded.git/site/x86_64-linux: Aug 19 21:44:27 site/x86_64-linux: Add some vars from ix86-common Aug 19 21:44:27 ix86-common had a few updates in the past couple of years; bring Aug 19 21:44:27 them to x86_64. Aug 19 21:44:27 Some of these could probably be moved to common-glibc. Aug 19 21:44:29 Signed-off-by: Michael Smith Aug 19 21:44:37 03Michael Smith  07org.openembedded.dev * r29f099ea70 10openembedded.git/ (14 files in 2 dirs): (log message trimmed) Aug 19 21:44:37 site/common-linux: centralize bash configure vars Aug 19 21:44:38 If the other site files are accurate, almost all of the bash vars Aug 19 21:44:40 are universal across Linux. Aug 19 21:44:42 Did this to fix bash on x86_64. Also bump bash PRs to help track Aug 19 21:44:44 breakage later. Aug 19 21:44:46 Signed-off-by: Michael Smith Aug 19 21:52:01 03Michael Smith  07org.openembedded.dev * r9ea3013639 10openembedded.git/ (5 files in 2 dirs): Aug 19 21:52:01 syslinux: upgrade to 3.82 and split into a few packages Aug 19 21:52:01 3.63 is no longer available. Aug 19 21:52:01 Signed-off-by: Michael Smith Aug 19 21:57:19 03Chris Larson  07org.openembedded.dev * re20e2ebb2a 10openembedded.git/classes/package_rpm.bbclass: Aug 19 21:57:19 package_rpm: don't run against native/cross recipes, explicitly Aug 19 21:57:19 In the past, it only avoided adding the rpm-native/fakeroot-native deps if Aug 19 21:57:19 PACKAGES is empty, but unfortunately some native recipes have PACKAGES set to Aug 19 21:57:19 non-empty, so let's explicitly check for native/cross. Aug 19 21:57:23 Signed-off-by: Chris Larson Aug 19 21:57:25 03Chris Larson  07org.openembedded.dev * r08f57c984f 10openembedded.git/conf/collections.inc: (log message trimmed) Aug 19 21:57:28 collections.inc: sync with MVL6. Aug 19 21:57:30 - Gather info on the collections into COLLECTIONSINFO. Aug 19 21:57:32 - Store the collection info for the current file in COLLECTIONINFO. Aug 19 21:57:34 - Add has_collection() and get_collection() utility functions. Aug 19 21:57:36 - Make collection_unpack pass back the collection name. Aug 19 21:57:38 - Just use != rather than symmetric_difference on the set comparision. Aug 19 21:57:40 03Chris Larson  07org.openembedded.dev * r7001e26671 10openembedded.git/recipes/distro-version/distro-version.bb: Aug 19 21:57:43 distro-version: new general recipe, based on angstrom-version. Aug 19 21:57:45 Signed-off-by: Chris Larson Aug 19 21:57:49 03Chris Larson  07org.openembedded.dev * ree09ed852d 10openembedded.git/classes/ (src_distribute.bbclass src_distribute_local.bbclass): (log message trimmed) Aug 19 21:57:56 src_distribute*.bbclass: cleanup, revamp parts, distribute all sources. Aug 19 21:57:58 - Kill the big shell function with the case statement used in _local in Aug 19 21:58:00 favor of a set of bitbake variables, with the appropriate one selected Aug 19 21:58:04 as SRC_DISTRIBUTECOMMAND. Aug 19 21:58:06 - Add and use SRC_DISTRIBUTE_DLONLY to toggle distribution of all sources Aug 19 21:58:08 versus just those downloaded to DL_DIR. This variable expects to be set Aug 19 21:58:10 03Chris Larson  07org.openembedded.dev * r6b86d4060b 10openembedded.git/classes/devshell.bbclass: Aug 19 21:58:13 devshell.bbclass: set a default for XAUTHORITY and export DBUS_SESSION_BUS_ADDRESS. Aug 19 21:58:17 Signed-off-by: Chris Larson Aug 19 21:58:19 03Chris Larson  07org.openembedded.dev * ra6a23f7369 10openembedded.git/classes/src_distribute.bbclass: Aug 19 21:58:22 src_distribute: force do_fetch execution in do_distribute_sources. Aug 19 21:58:26 Works around src_distribute failures with the use of packaged staging. Aug 19 21:58:28 Signed-off-by: Chris Larson Aug 19 21:58:30 03Chris Larson  07org.openembedded.dev * rf9b6fb8272 10openembedded.git/recipes/ncurses/ncurses.inc: Aug 19 21:58:35 ncurses: don't use libtool. Aug 19 21:58:37 Signed-off-by: Chris Larson Aug 19 21:58:39 03Jeremy Puhlman  07org.openembedded.dev * r7ab4382774 10openembedded.git/classes/src_distribute_local.bbclass: Aug 19 21:58:42 src_distribute_local: Change cp -f to cp -fr to account for directories. Aug 19 21:58:46 Signed-off-by: Jeremy Puhlman Aug 19 21:58:48 Signed-off-by: Chris Larson Aug 19 22:00:22 * kergoth continues to go through his backlog of bits to push.. Aug 19 22:03:41 heh, I'll run a clean build in a bit :) Aug 19 22:13:33 03Denys Dmytriyenko  07org.openembedded.dev * r1307fdab98 10openembedded.git/recipes/rt-tests/rt-tests_0.28.bb: rt-tests: fix GNU_HASH QA error Aug 19 22:30:32 hmm Aug 19 22:30:47 kergoth: whoa, lots of commits... Aug 19 22:31:31 heh :) just getting started Aug 19 22:31:41 kergoth: what's COLLECTION(S)INFO for? Aug 19 22:32:34 can be useful to programmatically get to the current collection, and get the unpacked collection path after the fact Aug 19 22:32:50 without that, all you have is bits in bbfiles, which are hard to rip back apart, since the user couldve changed it, and the tarballs in COLLECTIONS Aug 19 22:34:17 I see. I haven't used packed collections extensively yet... Aug 19 22:34:32 also, with has_collection, you can make decisions based on what collections are being used. for MVL6, i can enable hostap/madwifi/wifi only if the collection that has the wireless stuff is available (what MSD the customer buys will affect what collections we give them) Aug 19 22:35:28 not sure how useful the collectioninfo will be to others, but no point not pushing it, adding it doesn't hurt anything, and the resource usage is minimal Aug 19 22:35:35 so, what drives what in your setup - BBPATH drives COLLECTIONS or vice versa? Aug 19 22:37:43 COLLECTIONS is a space separated list of directories or tarballs, in priority order, highest to lowest Aug 19 22:37:58 collections.inc sets BBFILES, BBFILE_* based on that, and also updates bbpath if it's missing any of them Aug 19 22:38:51 in mvl6, all of our collections are tarballs, so we had to do a bit of other stuff, to bootstrap the unpacking of the collections to make available the real bitbake.conf, but thats the jist Aug 19 22:43:49 thanks Aug 19 22:44:50 * kergoth would kind of like to get bitbake to accept an easier setup mechanism like COLLECTIONS directly, rather than all the other vars, but... Aug 19 22:45:10 kergoth: btw, wanted to ask you if you are familiar with AUTOREV Aug 19 22:45:27 nope, that stuff happened after i left, and i haven't gotten caught up on it since coming back, yet Aug 19 22:45:54 I see Aug 19 22:50:28 kergoth: denix: you are using packaged staging...does it work well for you? I still have issues... Aug 19 22:50:49 I sent two msg to oe-dev ML Aug 19 22:52:24 actually there is a third issue...with packages wrongly or rather not staged (one of them uses package_stagefile_shell) Aug 19 22:53:33 i have a fair bit of work to do on the pstage stuff. there are a lot of pstage commits that i haven't had a chance to push yet, need to clean it all up, it'll take a while Aug 19 22:53:57 np, remember RP one from Poki Aug 19 22:54:32 * XorA suspects klibc broke it Aug 19 22:57:57 XorA: strange error on rebuild from pstage was ' * ERROR: Cannot satisfy the following dependencies for kexecboot: Aug 19 22:57:58 * kexec-klibc-static ' Aug 19 22:58:13 in the recipe we have: PACKAGES =+ "kexec-klibc-static kdump-klibc-static" Aug 19 22:58:15 ant__: blody hell, I was just messing with yah :-) Aug 19 22:58:25 and it builds normally, but not from pstage Aug 19 22:58:30 ant__: I didnt know it was actually broke :-D Aug 19 22:59:03 XorA: you guess what? zaurus-updater is the second sucker :) Aug 19 22:59:19 ant__: I haven't played with pstage lately - it worked for me some time back though Aug 19 22:59:43 * XorA has no zaurus :-D Aug 19 22:59:52 the biggest issue I see is the /cross dir is misplaced on rebuild... Aug 19 23:00:16 if I move all under /cross/armv5te/ * then is fine Aug 19 23:01:16 XorA: I'm adding RPROVIDES = "kexec-klibc-static kdump-klibc-static" Aug 19 23:12:42 mickey|zzZZzz: use Twitter... :) Aug 20 00:44:58 03Chris Larson  07org.openembedded.dev * r9f35410c92 10openembedded.git/recipes/opkg/ (opkg.inc opkg/opkg-intercept-cleanup.patch): Aug 20 00:44:58 opkg: explicitly clean up the opkg-intercept dirs rather than rm -rf. Aug 20 00:44:58 Signed-off-by: Chris Larson Aug 20 00:45:08 03Chris Larson  07org.openembedded.dev * r83a6a8745d 10openembedded.git/recipes/opkg/ (6 files): Aug 20 00:45:08 opkg: use INC_PR. Aug 20 00:45:08 Signed-off-by: Chris Larson **** ENDING LOGGING AT Thu Aug 20 02:59:56 2009