**** BEGIN LOGGING AT Thu Aug 28 02:59:56 2008 Aug 28 05:05:38 how do I get linphone_1.6.0 to configure/compile? Aug 28 05:06:10 I've tried defining SPEEX_LIBS and SPEEXDSP_LIBS without success Aug 28 05:07:56 errors: http://fastpaste.net/146/ Aug 28 05:17:29 ?? => http://bugs.openembedded.net/attachment.cgi?id=9704 Aug 28 05:18:05 * * OE Bug 4527 has been created by denis(AT)denix.org Aug 28 05:18:07 * * omap3evm.conf - allow local.conf to override PREFERRED_PROVIDER_virtual/kernel Aug 28 05:18:09 * * http://bugs.openembedded.net/show_bug.cgi?id=4527 Aug 28 05:33:05 * * OE Bug 4528 has been created by hans(AT)eds.org Aug 28 05:33:07 * * moko-gtk-engine fails to build when dash is sh Aug 28 05:33:09 * * http://bugs.openembedded.net/show_bug.cgi?id=4528 Aug 28 07:11:31 morning Aug 28 07:17:53 gm Aug 28 08:06:02 bonjour Aug 28 09:05:24 woglinde: ping Aug 28 09:05:53 where was that .stable viewmtn? Aug 28 09:07:12 np, found http://amethyst.openembedded.net/oe/viewmtn/viewmtn.py/ Aug 28 09:08:58 well, sort of...Internal Server Error Aug 28 09:11:25 jupp Aug 28 09:11:28 sorry Aug 28 09:11:35 I would you tell this yesterday Aug 28 09:11:43 but somehow forgot Aug 28 09:13:03 there is avanti_ having problems with glibc_2.3.2+cvs20040726.bb in stable Aug 28 09:13:22 I see there have been recent changes in .dev Aug 28 09:13:30 *g* Aug 28 09:13:32 2004 Aug 28 09:13:36 is 4 years old Aug 28 09:13:50 and rather I am concentraded on uclibc Aug 28 09:14:07 I can not stand the pain compilig glibc 2 times about 3 hours Aug 28 09:14:17 when uclibc finishes under 3 minutes Aug 28 09:15:20 hi woglinde Aug 28 09:15:24 hi pb Aug 28 09:15:46 hi pb_ shorttail Aug 28 09:15:50 heh Aug 28 09:15:51 hi ant_work Aug 28 09:15:55 pb hm I integrated your linxuthreads code for arm in uclibc, was fairly easy, but I have to test if it really runs Aug 28 09:16:28 woglinde: I guess it should do, linuxthreads is pretty simple Aug 28 09:16:37 we will see Aug 28 09:16:49 otherwis I had no chance to compile webkit under uclibc Aug 28 09:17:21 woglinde: how's the 'locale' status in uclibc? Aug 28 09:17:30 ant dont know Aug 28 09:17:40 I dont need locale really Aug 28 09:17:47 you should know about German .... Aug 28 09:17:50 hi genesis Aug 28 09:17:50 he he Aug 28 09:18:21 ant libiconv and gettext is compiling fine Aug 28 09:18:45 you can try it out Aug 28 09:18:47 *g* Aug 28 09:19:20 if you have time Aug 28 09:19:29 set ANGSTROM_MODE= "uclibc" in local.conf Aug 28 09:20:14 pb I wonder why the arm stuff was oversight at the uclibc linuxthread import Aug 28 09:20:25 pb maybee while it was moved to ports Aug 28 09:21:25 yeah, dunno Aug 28 09:21:36 woglinde: I did in the past some test-builds with angstrom / uclibc and never had compile issues Aug 28 09:21:37 linuxthreads has supported arm for years Aug 28 09:21:50 I have to check the latest version now Aug 28 09:22:18 me and Jay are working at one busibox image, all against uclibc Aug 28 09:22:23 pb only two were missing Aug 28 09:22:30 files Aug 28 09:23:10 ant khem pushed nptl support in Aug 28 09:23:14 but only for eabno Aug 28 09:23:17 ups ebai Aug 28 09:23:32 args Aug 28 09:23:34 eabi Aug 28 09:23:52 we are on EABI (Zaurus) Aug 28 09:24:56 I'd hope it is already new-posix-threads-library enabled :D Aug 28 09:25:14 ant nope Aug 28 09:25:23 ah, really? Aug 28 09:25:37 linuxthreads.old is default Aug 28 09:25:42 I got nptl with glibc-2.5.1 on Gentoo IIRC Aug 28 09:25:47 3 yars ago... Aug 28 09:25:55 in uclibc Aug 28 09:26:06 oops...sorry..I have to read carefully Aug 28 09:26:20 nice improvement Aug 28 09:26:22 thx Aug 28 09:27:27 hm Aug 28 09:27:31 BTW, on X86 2cpu server I've never seen the 30% performance boost promised by nptl Aug 28 09:27:38 we should disable examples for qt stuff Aug 28 09:27:52 ant intel cpu? Aug 28 09:27:57 old PIII Aug 28 09:28:00 aeh pentium4? Aug 28 09:28:05 *g* Aug 28 09:28:05 older Aug 28 09:28:14 hm Aug 28 09:28:19 p3 should be okay Aug 28 09:28:55 the kernel is monolithic, no preempt Aug 28 09:32:00 ant_work: the performance improvement for nptl is pretty much dependent on your workload. Aug 28 09:32:28 I don't think there are any situations where nptl should actually be slower on x86, but there are certainly many cases where it would be hard to measure much difference compared to linuxthreads Aug 28 09:33:06 never done real benchs Aug 28 09:33:41 woglinde: btw you know what is pulling glibc_2.3.2 ? Aug 28 09:33:50 and DISTRO = "generic" :=) Aug 28 09:34:16 "generic" needs some love Aug 28 09:35:50 just program like real men and dont use threads :-) Aug 28 09:36:14 yah, good point Aug 28 09:40:51 # equery hasuse nptl Aug 28 09:40:59 [I--] [ ] sys-libs/libstdc++-v3-3.3.6 (5) Aug 28 09:41:28 only this compat lib has nptl in Gentoo's useflags Aug 28 09:57:03 ant in dev generic is now minimal Aug 28 09:57:22 xora pfff Aug 28 09:59:57 woglinde: ok, he was using sta(b)le Aug 28 10:14:39 florian_: good morning Aug 28 10:15:35 good morning Aug 28 11:14:58 morning Aug 28 11:15:00 ~curse rpm Aug 28 11:15:01 May the fleas of a thousand camels infest your most sensitive regions, rpm ! Aug 28 11:15:50 gm Aug 28 11:23:08 hrw: these rpm appreciations...are for RP? Aug 28 11:32:12 why? Aug 28 11:46:27 any reason not to push http://bugs.openembedded.net/attachment.cgi?id=9951 Aug 28 11:48:29 whib bug it is? Aug 28 11:48:38 I can push it Aug 28 11:49:26 I can push it, I just wanted to make sure no one hadd a "philosophy" issue :) Aug 28 11:49:38 http://bugs.openembedded.net/show_bug.cgi?id=4527 Aug 28 11:49:56 Crofton: for me it is ok Aug 28 11:50:01 great Aug 28 11:50:18 I don't like that the kernel has a dependency on local config Aug 28 11:50:28 but, it is important for people to try out things Aug 28 11:50:50 and I understand how hard it is to get one sane kernel for a dev board Aug 28 11:52:16 heh.. I am forgetting more and more from monotone ;( Aug 28 11:52:40 heh Aug 28 11:52:47 I am pushing now Aug 28 11:53:09 when we switch, you will be my git manual :) Aug 28 11:57:00 ;d Aug 28 11:59:44 I commited, with note warnng about chaos if you forget you overrode the kernel locally Aug 28 12:00:06 * * OE Bug 4527 has been RESOLVED (FIXED) by philip(AT)balister.org Aug 28 12:00:08 * * omap3evm.conf - allow local.conf to override PREFERRED_PROVIDER_virtual/kernel Aug 28 12:00:10 * * http://bugs.openembedded.net/show_bug.cgi?id=4527 Aug 28 12:05:01 when you boot a machine programmed with Angstrom, the serial terminal displays Anstrom in ascii characters before the log-in prompt.. what file/package sets that up? Aug 28 12:10:56 /etc/motd Aug 28 12:11:05 or /etc/issue(.net) Aug 28 12:13:50 hrw: thank you. Aug 28 12:20:04 re Aug 28 12:20:11 i am trying to do a "make slugosle-package" buildrun ... but somewhere down the line during patching of madwifi a patch fails and aborts the build ... how can i prevent this specific package from being build and keep the rest of the package build going ? Aug 28 12:20:36 make? Aug 28 12:21:00 woglinde: it's a slugos generic Makefile that essentially does a bb distro Aug 28 12:21:06 nearly all devs using bitbake Aug 28 12:21:52 yes and channel topic says This is not a distro or machine support channel Aug 28 12:22:30 woglinde: i realise that ... i am just having a hard time trying to figure out what the hell i am supposed to set in local.conf to prevent this specific package from even building Aug 28 12:22:37 and what I heard slugos hasnt many support in oe these time Aug 28 12:22:42 s Aug 28 12:22:59 local.conf is not the root Aug 28 12:23:06 the root lies in task stuff Aug 28 12:23:34 never mind i guess ... i'll just have to figure it out on my own Aug 28 12:24:04 hm Aug 28 12:24:08 if he cant wait Aug 28 12:27:12 ant still there? Aug 28 12:28:09 re Aug 28 12:29:12 ant how I mad kexecboot kernel? Aug 28 12:29:14 make Aug 28 12:29:38 just bitbake linux-kexecboot (for supported machines) Aug 28 12:30:08 ah hm Aug 28 12:30:14 (it pulls initramfs-kexecboot and kexecboot) Aug 28 12:30:20 where this ends up? Aug 28 12:30:23 in deploy? Aug 28 12:30:28 in a big kernel in /deploy Aug 28 12:30:32 yep Aug 28 12:30:55 for collie and poodle is even better packaged IIRC Aug 28 12:30:59 hms Aug 28 12:31:04 look at 4520 and 4523 Aug 28 12:31:08 Nothing PROVIDES 'linux-kexecboot' Aug 28 12:32:03 hm seems it is not commited yet? Aug 28 12:32:07 right? Aug 28 12:32:23 ni, it is Aug 28 12:33:43 is in /packages/kexecboot Aug 28 12:33:52 but only 2.6.26 atm Aug 28 12:33:54 hm Aug 28 12:34:04 what I have to do for the simpad Aug 28 12:34:18 probably initramfs-bootmenu-image Aug 28 12:34:35 (pfalcon work, uclibc) Aug 28 12:34:37 simpad isnt far from the collie Aug 28 12:34:51 this last one is too big for zaurus Aug 28 12:36:15 but if your kernel supports kexec and initramfs...thesing's kexecboot could work Aug 28 12:36:23 don't think is kernel specific Aug 28 12:36:37 yes Aug 28 12:36:38 perhaps teh issue mtd != mtdblock Aug 28 12:36:55 but what I have to change in kernelrecipe for this Aug 28 12:37:19 or have I only change kexecboot Aug 28 12:37:26 try COMPATIBLE_MACHINE Aug 28 12:37:53 in linux-kexecboot.inc Aug 28 12:38:13 yes found it too Aug 28 12:38:51 remember the defconfig must be set for initramfs Aug 28 12:39:02 and kexec Aug 28 12:39:14 ah okay Aug 28 12:39:52 jupp it is Aug 28 12:40:35 urgs perl-native is needed Aug 28 12:42:05 woglinde: hmm Aug 28 12:42:07 ? Aug 28 12:42:26 CONFIG_INITRAMFS_SOURCE="initramfs.cpio.gz" Aug 28 12:42:35 sorry... Aug 28 12:42:36 http://tinderbox.openembedded.net/packages/linux-kexecboot/ Aug 28 12:43:19 hm Aug 28 12:43:22 seems Aug 28 12:43:33 it pulls in linux-2.6.26 Aug 28 12:44:12 hvontres is proposing addition of 2.6.24 (oebug 4523) Aug 28 12:44:35 yes Aug 28 12:44:37 (patch attached to 4520 and 4523 btw) Aug 28 12:45:06 we are waiting for thesing last check and commit Aug 28 12:45:37 but if you could extend it to other machines, good Aug 28 12:48:22 hm Aug 28 12:48:30 its to much work Aug 28 12:52:33 oh, shade Aug 28 12:54:57 maybe later Aug 28 12:59:28 hail zecke_ Aug 28 13:39:31 does anyone have experience with using OE in a commercial environment? I want to make sure we're taking care of GPL compliance correctly. I just want to see what other people's solutions are Aug 28 13:39:40 bitbake makes this easy, which is nice Aug 28 13:41:43 frikker: read src_distribute classes Aug 28 13:42:06 hrw: thanks for the heads up. will do. Aug 28 13:42:27 hrw: is that a file? or where is it? Aug 28 13:46:05 * * OE Bug 4529 has been created by mdarland(AT)pager.net Aug 28 13:46:07 * * libnl_1.1 compile fails on 2.6.26 kernel Aug 28 13:46:09 * * http://bugs.openembedded.net/show_bug.cgi?id=4529 Aug 28 13:54:03 OE/classes/src_distribute* Aug 28 13:54:46 frikker: also read http://pokylinux.org/doc/poky-handbook.html#ref-classes-src-distribute Aug 28 14:04:27 hrw: a technical question about linux-rp: we have one defconfig.machine each version of /linux/linux-rp-2.6.2x and other defconfig.machine in /kexecboot/linux-kexecboot-2.6.2x Aug 28 14:05:01 can this be optimized? using patches against stock (original kernel) defconfig? Aug 28 14:10:02 Intel acquires OpenedHand Aug 28 14:10:16 now this is official: http://o-hand.com/ Aug 28 14:11:16 hrw: now you are working for intel? Aug 28 14:11:19 how did it sound?...just let me know...do you stay or do you go? Aug 28 14:11:26 (clash ?) Aug 28 14:12:25 zecke_: papers time for me Aug 28 14:25:59 hrw: thanks for that poky link. freaking awesome. Aug 28 14:45:26 ehmm, what's this 'conman: connection manager' from moblin ? Aug 28 14:46:32 ant_work: Handles network connections like wifi, wired lan, bluetooth, etc. Aug 28 14:46:45 ant_work: The mobiln version from NetworkManager Aug 28 14:46:51 it is the phoenix ? Aug 28 14:46:54 :=) Aug 28 14:47:16 nah, but perhaps better for small systems Aug 28 14:47:30 Needs to be evaluated Aug 28 14:50:51 and about connman-gnome? (The GTK+-based user interfaces for an applet and a properties dialog) Aug 28 15:29:46 re Aug 28 15:38:27 re Aug 28 15:52:43 03  07org.openembedded.dev * r83a88adec4 10OE.dev/packages/libnl/ (8 files in 2 dirs): libnl add 1.1, remove outdated ones Aug 28 15:52:43 03  07org.openembedded.dev * r88e31644d2 10OE.dev/ (3 files in 3 dirs): add python-connexion, a versatile (network) connectivity library Aug 28 15:52:45 03  07org.openembedded.dev * r9bd826f0fa 10OE.dev/: Aug 28 15:52:45 merge of '8f5711b8ee04e00c48f5997987d6ab4600e5dc19' Aug 28 15:52:47 and 'b04f946bcf310267d7557532bc06f235d79f8226' Aug 28 15:52:49 03  07org.openembedded.dev * rf7ed81b527 10OE.dev/packages/classpath/ (18 files in 5 dirs): (log message trimmed) Aug 28 15:52:52 classpath: fix config.rpath, autotools and ecj -d Aug 28 15:52:54 * add ecj_java_dir.patch for issuse when ecj is not cable to Aug 28 15:52:56 to create directory-structure with parameter -d Aug 28 15:52:58 * add autotools.patch for m4 stuff Aug 28 15:53:00 * copy config.rpath always from /usr/share/gettext, so no crude hacks are Aug 28 15:53:02 needed anymore Aug 28 15:53:06 03  07org.openembedded.dev * rca9d7399cc 10OE.dev/: Aug 28 15:53:08 merge of '1d6a9becb57e689050e4b9b1fb6482e04e3e5111' Aug 28 15:53:10 and 'dd34c96ee9d492fd08a5d56c4405aee69f2740ee' Aug 28 15:53:14 03  07org.openembedded.dev * r6a87013692 10OE.dev/packages/images/fso-image.bb: fso-image: remove bogus compatible machine *again* Aug 28 15:53:17 03  07org.openembedded.dev * rd13402ae40 10OE.dev/packages/ppp/ (ppp-2.4.3/ppp-2.4.3-mppe-mppc-1.1.patch ppp_2.4.3.bb): Aug 28 15:53:22 packages/ppp/ppp_2.4.3.bb : Add mppe support Aug 28 15:53:24 Bump PR Aug 28 15:53:26 03  07org.openembedded.dev * rfa5755c45e 10OE.dev/packages/opie-console/ (4 files in 2 dirs): Aug 28 15:53:29 opie-console: fix errors thrown with gcc-4.3 Aug 28 15:53:31 * serval headers used function arguments with the same name, Aug 28 15:53:35 this is no good coding style and gcc-4.3 is yelling Aug 28 15:53:37 * add double_name.patch Aug 28 15:53:39 * bump PR Aug 28 15:53:41 03  07org.openembedded.dev * rc50aed4b2e 10OE.dev/packages/tangogps/tangogps_0.9.2.bb: tangogps: Updated to version 0.9.3. Aug 28 15:53:44 03  07org.openembedded.dev * r6fdf4f0f31 10OE.dev/packages/mozilla/files/powerpc/jsautocfg.h: mozilla: fix powerpc builds Aug 28 15:53:47 03  07org.openembedded.dev * r49b2ac1b11 10OE.dev/packages/ppp/ppp_2.4.3.bb: ppp: fix parse error Aug 28 15:53:51 CIA-24: welcome back Aug 28 15:53:56 03  07org.openembedded.dev * r0f331e9307 10OE.dev/site/common-glibc: site common-glibc: add ac_cv_fread_reads_directories Aug 28 15:53:59 03  07org.openembedded.dev * r1e595f610e 10OE.dev/packages/gstreamer/gst-ffmpeg_0.10.4.bb: gst-ffmpeg: 0.10.4 is working great, make it the default Aug 28 15:54:06 03  07org.openembedded.dev * r3f9b068bec 10OE.dev/packages/ (3 files in 2 dirs): cairo, xservers: bump PR due to pixman abi change for arm Aug 28 15:54:11 03  07org.openembedded.dev * r6cdb826968 10OE.dev/classes/scratchbox-compat.bbclass: scratcbox-compat.bbclass: New bbclass. Aug 28 15:54:14 03  07org.openembedded.dev * r89dfe4666e 10OE.dev/: Aug 28 15:54:16 merge of '045e00ff4301aa05fddab174a3ac5bf9c260343b' Aug 28 15:54:18 and '90204f8950d2ea5d6cc182ba36902a2e41bdfc7e' Aug 28 15:54:24 03  07org.openembedded.dev * rfd5f5bbf28 10OE.dev/packages/intltool/ (intltool-native_0.40.3.bb intltool_0.40.3.bb): intltool: update to 0.40.3 Aug 28 15:54:27 03  07org.openembedded.dev * r8b3e988c02 10OE.dev/packages/intltool/intltool_0.40.3.bb: intltool 0.40.3: fix shoddy m4 macro Aug 28 15:54:32 03  07org.openembedded.dev * r063a9909c3 10OE.dev/packages/sugar/sugar-base_0.82.0.bb: sugar-base: add more python rdepends Aug 28 15:54:37 03  07org.openembedded.dev * r67f8023272 10OE.dev/: Aug 28 15:54:39 merge of '0e9b0714e6ab764eaa6afb85fbd5d15b41175604' Aug 28 15:54:41 and '45a9c4a970e79c5b8343a4a4ff110589bcc19d2f' Aug 28 15:54:43 03  07org.openembedded.dev * r26cb1a5679 10OE.dev/packages/gnome/ (files/acinclude.m4 libwnck_2.23.6.bb): libwnck: update to 2.23.6 Aug 28 15:54:48 03  07org.openembedded.dev * rd0eef6f485 10OE.dev/packages/intltool/intltool-native_0.40.3.bb: intltool-native 0.40.3: replace patch with sed Aug 28 15:54:51 03  07org.openembedded.dev * r30d81a4040 10OE.dev/packages/sugar/sugar-base_0.82.0.bb: sugar-base: it now gets to the first run wizard with these extra depends Aug 28 15:54:56 03  07org.openembedded.dev * r1e27f3ef18 10OE.dev/: Aug 28 15:54:58 merge of '72708eace9dcdb70c5700231b4f92541bb90b6ac' Aug 28 15:55:04 and '91a0ad611226219a56b014a2e89ec3d70647d2eb' Aug 28 15:55:06 03  07org.openembedded.dev * re187493ff7 10OE.dev/: Aug 28 15:55:08 merge of '4dc9e841adc6770da91963ff705d5101d2ff3417' Aug 28 15:55:10 (77 lines omitted) Aug 28 15:55:12 03  07master * rdc218f50a4 10OE.dev/: Aug 28 15:55:14 merge of '3cb2978c67d7c62746b9e01b6002b3798f3418e1' Aug 28 15:55:18 and '81a0838330d2a072edda6c27b5e8c63f40172553' Aug 28 15:55:20 03  07master * rd6b76c18a2 10OE.dev/packages/gnome/gnome-python-desktop_2.23.0.bb: gnome-python-desktop: also build wnck and rsvg modules Aug 28 15:55:23 03  07master * r71c4981463 10OE.dev/packages/gnome/hippo-canvas_0.3.0.bb: hippo canvas: add 0.3.0 Aug 28 15:55:28 03  07master * r063a9909c3 10OE.dev/packages/sugar/sugar-base_0.82.0.bb: sugar-base: add more python rdepends Aug 28 15:55:37 03  07master * r30d81a4040 10OE.dev/packages/sugar/sugar-base_0.82.0.bb: sugar-base: it now gets to the first run wizard with these extra depends Aug 28 15:55:40 03  07master * r1e27f3ef18 10OE.dev/: Aug 28 15:55:42 merge of '72708eace9dcdb70c5700231b4f92541bb90b6ac' Aug 28 15:55:46 and '91a0ad611226219a56b014a2e89ec3d70647d2eb' Aug 28 15:55:48 03  07master * re187493ff7 10OE.dev/: Aug 28 15:55:50 merge of '4dc9e841adc6770da91963ff705d5101d2ff3417' Aug 28 15:55:52 and 'fd1d401ab19d2e2f42e5f93edbf8609e4e3012de' Aug 28 15:55:54 03  07master * r5d7762ecd9 10OE.dev/packages/libgdbus/ (.mtn2git_empty libgdbus_git.bb): libgdbus: Add recipe from OM git. Dependency of connman. Aug 28 15:55:57 03  07master * rd3a1d0fb2a 10OE.dev/conf/distro/include/sane-srcrevs.inc: sane-srcrevs.inc: Sane revs for libgdbus and connman. From OM git. Aug 28 15:56:04 03  07master * r9ca0a0876f 10OE.dev/packages/connman/ (5 files in 2 dirs): connman: Connection manager from moblin. Recipe from OM git. Aug 28 15:56:07 03  07master * r9dfb253c95 10OE.dev/packages/avahi/mango-lassi_git.bb: mango-lassi: fix install Aug 28 15:56:14 03  07master * r3bfcb4e7aa 10OE.dev/: Aug 28 15:56:18 merge of '93275fa73551aabe0b186dde99fd16be158043ab' Aug 28 15:56:20 and 'eed8dc866370e81bc5849ef5b2ddfeb8e87d4f1c' Aug 28 15:56:22 03  07master * r3969386af3 10OE.dev/packages/gstreamer/gst-meta-base_0.10.bb: gst-meta-base 0.10 add wav demuxer to meta-audio Aug 28 15:56:25 03  07master * r6d06d0ac23 10OE.dev/conf/machine/omap3evm.conf: Aug 28 15:56:27 omap3evm.conf : Allow override of kernel. See bug #4527. Aug 28 15:56:29 Be very careful with this, if you forget the override Aug 28 15:56:33 is in local.conf confusion will result :) Aug 28 15:56:35 03  07master * rb62e46df61 10OE.dev/packages/linux/ (7 files in 3 dirs): linux-2.6.26.bb: package linux 2.6.26 for mpc8313e-rdb Aug 28 15:56:38 (16 lines omitted) Aug 28 15:56:45 ui Aug 28 16:01:43 morning all Aug 28 16:02:01 thesing: morning :) Aug 28 16:03:43 thesing: when you get a chance, could you take a look at the new patches for #4523 ? I only added defconfigs for pxa models, since collie seems to be happy with 2.6.23 Aug 28 16:03:51 err 2.6.26 Aug 28 16:04:24 sure Aug 28 16:09:37 hi thesing Aug 28 16:14:58 he thesing Aug 28 16:15:40 dd and dd_resuce are cool tools Aug 28 16:17:10 hvontres|work: you forgot to bump PR of linux-kexecboot_2.6.26 (I did it) Aug 28 16:17:37 hm I would be glad if the patch for 2.6.25 would be integrated Aug 28 16:17:42 so I can test for the simpad Aug 28 16:18:17 hvontres|work: we disable sizecheck for all zaurus kernels with this patch. I think we should only disable it for zaurus machines that are known to work with kexecboot (and sharp bootloader) Aug 28 16:18:36 woglinde: which patch for 2.6.25? Aug 28 16:18:36 thesing: please have look at the progress of 4520 Aug 28 16:26:34 thesing: Ok. so just disable for poodle and collie for now. I geuess maintainers for the other machines will need to fix the conf files once they are happy Aug 28 16:27:24 thesing: the C code is ok: here bootlog http://www.rafb.net/p/4gONec91.html Aug 28 16:27:45 yue see the mtdparts are passed Aug 28 16:28:38 ant_work: why is MAX_CMDLINE defined? one can just use COMMAND_LINE_SIZE Aug 28 16:28:50 Jai7: ^^? Aug 28 16:28:51 thesing: any ideas on how to make zaurus_make_installkit work better? Adding a seperate check for each machine seems kludgy to me Aug 28 16:29:50 thesing: I think max_cmdline 1024 is sane Aug 28 16:30:32 that may be, but kexecboot includes a header from kernel where COMMAND_LINE_SIZE is defined. So this should be used. Aug 28 16:31:13 I think it was recently bumped to 1024...before was smaller Aug 28 16:31:34 (recently for me can be years ago :-) Aug 28 16:32:07 wuff Aug 28 16:32:49 thesing: I just have quick search by kernel but found no good define :) Aug 28 16:33:03 thesing: let it be COMMAND_LINE_SIZE Aug 28 16:33:12 there is no special meaning Aug 28 16:49:11 hvontres|work: did you test you current version? Aug 28 16:51:08 hvontres|work: the move of IMAGE_DEVICE_TABLES to linux-kexecboot.inc does not seem to work here. Aug 28 16:53:58 bbl Aug 28 16:59:04 thesing not yet. Aug 28 16:59:48 thesing: any idea what happens if you have the device tables in a regualr image? Aug 28 17:00:23 * hvontres|work slaps head Aug 28 17:00:44 hvontres|work: you have more static entries in /dev but since we use udev it doesn't hurt (other than a few bytes of space) Aug 28 17:02:43 Actually if the devices table is used then you have devices available before udev starts -- depending on the machine/distro that might be a very good thing (it certainly is for SlugOS) Aug 28 17:04:05 did anyone of you attempt a hal upgrade to 0.5.11? Aug 28 17:04:30 hvontres|work: I like the idea of only adding the additional stuff if we need it. I tried moving it to initramfs-kexeboot-image but that doesn't work either. Please fix that. Aug 28 17:05:22 thesing: ok. Aug 28 17:07:10 thesing: one other thing I noticed is that you put the keyboard at /dev/event0 instead of /dev/input/event0. was that intentional? Aug 28 17:09:07 hvontres|work: if I added /dev/input/event0 to the device table the node wasn't created. If it works for you feel free to change it. Aug 28 17:09:22 hvontres|work: I added a comment in bugtracker Aug 28 17:10:58 thesing: thanks. I'll fix it when I get home then Aug 28 17:11:18 hvontres|work: thanks. Aug 28 17:12:12 bye Aug 28 17:12:45 hvontres|work: and for the installkit stuff: I'm not sure if we should drop it. Or if we should introduce a new machine feature and check for it. Aug 28 17:25:55 thesing: we have real problem with installkit now.. our main reason is u-boot :) Aug 28 17:26:17 updater.sh is not compatible with u-boot.. Aug 28 17:26:45 so we should disable it or create new updater :) Aug 28 17:37:12 Jay7: if you come up with a solution to change this I will look at it (maybe sth. like an USE_UBOOT or BOOTLOADER=uboot in machine conf and then test for that) Aug 28 17:38:21 KERNEL_TYPE=uImage :) Aug 28 17:38:43 we cant use updater with uImage Aug 28 17:38:58 so it is enough (imho) Aug 28 17:40:31 zecke pasted "Emergency Numbers" at http://paste.lisp.org/display/66024 Aug 28 17:40:40 *KERNEL_IMAGETYPE Aug 28 17:42:11 Jay7: yes should be enough. you could also make other stuff conditional to that (like the linux-kexecboot.inc in bug 4520) Aug 28 17:42:51 thesing: now I have fix zaurus-2.6.inc agaings hardcoded zImage Aug 28 17:43:20 but I dont know what to deal with updater yet.. Aug 28 17:44:12 installing u-boot first time is not user-friendly :) Aug 28 17:45:46 * Jay7 is away for ~1 hour Aug 28 17:45:48 Jay7: More user-friendly than Redboot :-P Aug 28 17:46:04 NAiL: I can't compare :) Aug 28 17:46:15 because I have no redboot device :) Aug 28 17:52:20 Jay7: Trust me, it sucks. ;) Aug 28 18:00:13 how do I change the patchlevel that patch.bbclass uses to patch? looks like there is perhaps a variable strippath or something but its not clear to me how to use that Aug 28 18:03:14 ah... pnum= Aug 28 18:21:25 what does something like 'do_patch[dirs] = "${WORKDIR}"' do exactly? does the 'dirs' key contain a list of dirs to perform the task on? Aug 28 18:23:01 tharvey: my guess: it tell do_patch from which dir the patch command should be run. Aug 28 18:26:57 hi mr_nice Aug 28 18:27:56 hi mr_nice how goes the git stuff? Aug 28 18:30:02 woglinde: hi Aug 28 18:30:09 thesing: hi Aug 28 18:30:39 thesing, ah yes - and more... exec_func_shell function of bitbake/lib/bb/build.py - states that the 'dirs varflag' contains a list of dirs that need to be created prior to execution and the last item in the list is where it will chdir/cd to Aug 28 18:30:51 thesing: for u-boot quite well. I did not add stuff to the linux tree at the moment. Aug 28 18:32:25 mr_nice: how is the u-boot status? what hardware does work and what doesn't? Aug 28 18:32:43 mr_nice hm uboot what you are doing? Aug 28 18:35:10 thesing: mq200, serial both flash chips are working. resume is still not working but slims wanted to help me thre a bit (he did the redboot port). pcmcia is not working but jffs2. So you can use it to flash and boot angstrom but not to resume from suspend Aug 28 18:35:38 he fine Aug 28 18:35:51 pcmica yes with network Aug 28 18:35:55 would be nice Aug 28 18:36:13 then netbooting and nfsroot Aug 28 18:36:28 no more testing with flashing or serial download Aug 28 18:36:37 yes that woud be nice :) Aug 28 18:42:29 woglinde: i think redboot is able to boot from network with a ne2k card. But I am not familiar with it Aug 28 18:43:28 woglinde: slims told me that the only important thin is to plug in the card before you boot the simpad Aug 28 18:43:50 mr_nice yes it is Aug 28 18:43:50 slim made it work Aug 28 18:45:50 mr_nice: why do you put so much energy into u-boot? Redboot seems fine for me. Aug 28 18:47:27 thesing: hm, it is just for fun. and some stuff do not work with redboot at the moment Aug 28 18:48:21 thesing: I uploaded the source which slims gave me to a git repro at linuxtogo. Aug 28 18:49:27 thesing: adding network support to the hh.org loader could be the easiest way Aug 28 18:49:43 * Jay7 is back Aug 28 18:49:58 mr_nice: I'm not sure if its worth the work. Aug 28 18:50:24 * thesing thinks it would be better to improve kernel stuff or userspace. Aug 28 18:53:10 thesing: you are right with that opinion simpad has two well working bootloaders and there is no need for another one. it is more for fun. I am also trying to get lirc working on simpad Aug 28 18:54:19 userspace stuff would be more important than kernel stuff Aug 28 18:54:39 lirc via serial port adapter? Aug 28 18:54:56 no simpad has a irda port :) Aug 28 18:55:41 a lot of people using the simpad with wince as some cind of remote controll Aug 28 18:56:45 a quite bulky one. Aug 28 18:56:52 I don't know why they are using the simpad for that. I like my "normal" remote controll much more Aug 28 18:57:25 thesing: yes. It is some kind of clone for a real expansive remote controll Aug 28 18:58:12 thesing: how does your gps stuff? Aug 28 18:58:42 thesing: I was considering to give navit a try Aug 28 18:59:46 mr_nice: I found a side where you can grab google-maps maps for use with gpsdrive. But other than that nothing. Aug 28 19:00:21 mr_nice: I forbid myself coding for stuff not related to my diploma thesis. Aug 28 19:00:50 thesing: that is a good idea! what is your diploma thesis about? Aug 28 19:01:39 buying things in webshops with minimal (or "optimal") traffic usage Aug 28 19:06:31 realworld traffic (e.g. trucks)? Aug 28 19:08:25 mr_nice: no network traffic Aug 28 19:14:09 thesing: from client side? Aug 28 19:14:27 or from side of webshop owners? :) Aug 28 19:14:41 Jay7: yes Aug 28 19:14:58 Jay7: from client and network as whole. Aug 28 19:15:19 thesing: seems interesting :) Aug 28 20:12:30 re Aug 28 21:00:10 re Aug 28 21:01:04 thesing: wb Aug 28 21:01:12 we are patching right now... Aug 28 21:01:20 stay tuned :-) Aug 28 21:04:53 if I have a recipie named aaa_2.4, which provides Python, and bbb_2.5 wich also provides Python, bitbakes automatically selects the last one? Aug 28 21:05:57 RealSmasho: not sure. But you can choose one by setting PREFERRED_PROVIDER_python = "aaa" Aug 28 21:07:10 thanks thesing Aug 28 21:09:19 but Im really trying to understand Python recipies.. Aug 28 21:14:21 night all Aug 28 21:50:45 re Aug 28 21:51:37 Hm. Anybody know if the XIP functionality in kernel 2.6.25 will work on a SL-C3200 (spitz)? Aug 28 21:51:53 (feel free to tell me to buzz off if this is the horribly wrong channel for this kind of question) Aug 28 21:57:07 hrw|gone: ok, now I know what you meant with 'exciting' :) Aug 28 21:57:27 * florian reading http://www.linuxdevices.com/news/NS7559902579.html?kc=rss Aug 28 21:58:38 texel`: interesting question Aug 28 22:01:15 texel`: is XIP working with NAND and NOR? what flash does the SL-C3200 have? Aug 28 22:01:29 texel`: install u-boot and compile an image with -x switch. Let un know Aug 28 22:39:00 zecke_: it looks like * NOR flash supports execute in place (XIP). NAND flash must first be shadowed in RAM before being executable. Aug 28 23:02:15 ant_: Bah. The last uboot required me to completely reflash the NAND in the 3200 I have. Aug 28 23:02:21 I'd rather not go the uboot way. O.o Aug 28 23:02:26 If that's at all possible. Aug 28 23:02:57 zecke_: I have no idea if it's supported, which is exactly why I'm asking the question -- was hoping other devs would have tried this one before. =op Aug 28 23:03:24 zecke_: or rather, I don't know if it'll work with the Zaurus' NAND flash. Aug 28 23:03:35 zecke_: the in-configure help only talks about NOR flash. Aug 28 23:05:12 At the moment, I have a debian armel distro on the Z, which I was preparing to compile a vanilla 2.6.25 kernel with. Hence the question. =o) Aug 28 23:09:48 ant_: I know ;) Aug 28 23:10:06 I guessed...;-) Aug 28 23:11:21 So guessing from ant_'s statement, the XIP support probably won't work with the Z, since it's NAND. Aug 28 23:11:40 but with kexecboot you don't really need it... Aug 28 23:11:54 you'll have all the ram again Aug 28 23:11:57 Hence the use of uboot? Aug 28 23:12:05 if the kernel is too big Aug 28 23:12:13 and for custom rootfs Aug 28 23:12:45 allowing to boot different distros from different media Aug 28 23:13:44 (kernel is in /boot) Aug 28 23:16:44 Hm. Aug 28 23:16:52 The kernel is in /boot thing is very tempting. Aug 28 23:17:11 But... grr... don't. want. to semi-brick. the Z again. O.o Aug 28 23:17:17 ehmm...need some documentation... Aug 28 23:17:36 is not yet 100% in oe Aug 28 23:17:48 work in progress Aug 28 23:18:15 look at bugs 4520 and 4523 Aug 28 23:18:30 and then bitbake linux-kexecboot Aug 28 23:21:58 u-boot is optional atm Aug 28 23:27:08 Hm. So then what is kexecboot, then? Aug 28 23:27:57 Just a custom kernel with kexec built in that provides a menu of options, and lets you start a different kernel via kexec? Aug 28 23:28:01 oi oi... :) Aug 28 23:28:08 quick q Aug 28 23:28:16 how do i force a file to be installed after others with opkg? Aug 28 23:28:17 texel`: exactly Aug 28 23:28:18 ie Aug 28 23:28:23 i have a .bb with a list of files Aug 28 23:28:36 i want 1 files installed very last - after others have been installed Aug 28 23:28:37 Hm. Interesting. Aug 28 23:28:47 I'll take a look. Thanks for the pointer. Aug 28 23:28:53 raster: same package? why do you want? Aug 28 23:28:58 you're welcome :-) Aug 28 23:29:09 raster: sounds you want to have a post_inst hook and execute stuff after everything has been installed? Aug 28 23:29:14 zecke_: .deskltop file is installed before icon .png Aug 28 23:29:21 .desktop file Aug 28 23:29:25 problem is Aug 28 23:29:36 raster: I can guess :) Aug 28 23:29:37 a .desktop file install/uninstall triggers a "refresh app list" Aug 28 23:30:02 as such its just a race condition no matter what if .desktop is installed before any files it needs (eg icons or executable) Aug 28 23:30:05 * * OE Bug 4530 has been created by likewise(AT)gmx.net Aug 28 23:30:07 * * ffmpeg Makefile calls install with -s, which results in host strip being called Aug 28 23:30:08 so it should always get put in last Aug 28 23:30:08 :) Aug 28 23:30:10 * * http://bugs.openembedded.net/show_bug.cgi?id=4530 Aug 28 23:30:43 raster: I think it is pretty random and depends on the way we use tar to pack the files Aug 28 23:30:52 yeah Aug 28 23:31:01 i was hoping that just re-orderign the file list makes that happen Aug 28 23:31:02 it doesntr Aug 28 23:31:03 :( Aug 28 23:31:07 doesnt Aug 28 23:31:12 raster: and then probably on opkg how we copy the files to the destination (or if we copy to a tmp dir and then do find . to copy them) Aug 28 23:31:26 yeah Aug 28 23:31:29 that will make a difference Aug 28 23:31:42 install into tmpdir then rename() to destination is always safest Aug 28 23:31:51 but even then - it's a race cond Aug 28 23:32:02 raster: drop tick a mail on how opkg is installing files :) Aug 28 23:32:46 there must be a way to do this Aug 28 23:32:57 raster: If we use opendir on the tmpdir we depend on the order of the FS to list the files :) Aug 28 23:33:01 as such XDG/FDO dont have any way to "signal apps" when a .desktop is installed Aug 28 23:33:05 so its a matter of polling Aug 28 23:33:12 or fnotify/inotify Aug 28 23:33:28 efreet uses inotify to have the kernel tell me when something happens to the .desktops Aug 28 23:33:45 so once a .desktop appears Aug 28 23:33:54 the whole chain of "ooh look - new app!" starts Aug 28 23:33:56 raster: my computation yields to ask tick if it is possible to copy .desktop files the last Aug 28 23:34:12 the problem is opkg can take many seconds betwene installing a .desktop and possibly installing the .png for it too Aug 28 23:34:26 he knows Aug 28 23:34:31 i described this in the ticket Aug 28 23:35:00 i just am wondering if i can have a test at fixing it here so i know that in principle it is possible Aug 28 23:35:27 raster: don't use that word (ticket), it makes me angry :) Aug 28 23:35:43 raster: a test would be to hack on opkg-build-package Aug 28 23:36:17 hahaha Aug 28 23:36:18 dude Aug 28 23:36:21 stop doing all tickets Aug 28 23:36:25 just do tickets thatare for "you" Aug 28 23:36:39 zecke_: too many customer problems at work? =op Aug 28 23:38:48 raster: yeah, I'm searching for a new job... if we can't even hire a bugmaster... Aug 28 23:39:15 raster: opkg-build is what I'm looking at Aug 28 23:39:45 zecke_: i can more than understand why its painful Aug 28 23:40:03 zecke_: also the fact that om's "design" doesnt let you necessarily address user issues Aug 28 23:40:18 as user wantx X in the default setup - design wants Y Aug 28 23:40:21 too bad for user Aug 28 23:40:28 raster: hehe, I'm not even on that level... Aug 28 23:40:55 raster: I'm more well, we can not even support multiple keyboards... it is not even bad faith... give us time to create the technical means Aug 28 23:41:08 raster: you could hack opkg-build (a shell script) Aug 28 23:41:17 raster? long time no see Aug 28 23:41:34 raster: to add .desktop files at the end Aug 28 23:41:54 zecke_: i wonder if hacking opkg-build is the right way Aug 28 23:42:02 i was more looking for an existign mechanism.. if there is one Aug 28 23:42:05 kergoth: yo yo Aug 28 23:42:24 raster: line 237, but I think we need to handle this on the opkg side to have some kind of unpack/install order Aug 28 23:42:37 raster: no there is not Aug 28 23:42:52 aha Aug 28 23:42:53 ok Aug 28 23:42:55 so hacking time Aug 28 23:43:16 well i COULD have used a postinst that mv's the .desktop out of the dir to /tmp abd back in again Aug 28 23:43:17 for example Aug 28 23:43:20 that'd re-trigger it Aug 28 23:43:22 but thats ugly Aug 28 23:44:10 raster: touch would be enough, but then it would flash twice per package :) Aug 28 23:44:40 zecke_: hmm ok. yup. just tar up dir contents. opendir will just do the rest. Aug 28 23:44:48 zecke_: yeah. its still bad Aug 28 23:49:45 raster: .desktop files where at the end? test with the front? But I assume that opendir on the device will damange anything you try :) Aug 28 23:50:05 raster: we could patch jffs2 to list files in a sane manner ;) Aug 28 23:50:38 zecke_: nah - it's not jffs2 Aug 28 23:50:48 its that the installer untars the .desktop into applications/ Aug 28 23:51:06 before the icon .png is untarred into icons/ Aug 28 23:51:15 so the .desktp exists before the icon .png that it referrs to does Aug 28 23:51:19 thus the icon is blank Aug 28 23:51:46 so its really more a matter of just forcing a file install order Aug 28 23:54:56 hmm Aug 28 23:55:05 i'll leave it in tick's corner with opkg for now Aug 28 23:55:08 i have other fish to fry Aug 28 23:56:12 no quick and dirty option here as it'll need some special "exceptions" list of files to install given a particular ordering (eg a "post-insta files" list or such Aug 28 23:57:21 * florian votes for fixing the application that reads the desktop files Aug 29 00:04:34 florian: you can't Aug 29 00:04:42 as there is no fix Aug 29 00:04:47 its just "change" Aug 29 00:04:58 i explained who/how Aug 29 00:05:52 raster: keep it simple... just delay loading the desktop file Aug 29 00:06:32 already wait 1 second Aug 29 00:06:35 so i make it 2 Aug 29 00:06:42 a package that takes > 2 breaksd Aug 29 00:06:43 i make it 5 Aug 29 00:06:45 packages > 5 break Aug 29 00:06:47 make it 20 Aug 29 00:06:52 one that takes > 20 breaks Aug 29 00:06:54 it doesnt fix it Aug 29 00:07:20 you just play cat and mouse with "how slow is your flash/writes today/ how big is the package?" games Aug 29 00:07:24 its no fix Aug 29 00:07:35 fix is to install the "trigger" last Aug 29 00:07:46 as xdg/fso provide no other trigger mechanism as part of the standard Aug 29 00:08:01 raster: right, that would be a bad idea. think about a better mechanism. Aug 29 00:08:05 the ONLY trigger avaoilable is detecting a change in the file tree. Aug 29 00:08:19 now u can go invent some new trigger mechanism Aug 29 00:08:30 and u still have to go modify all packages to use that trigger Aug 29 00:08:30 that's overkill Aug 29 00:08:50 raster: Make better assumption but keep it simple. Aug 29 00:08:54 so given the work neded to come up with a new trigger... mayaswell just do it the simpel way - install the trigger last so all it depends on is already there Aug 29 00:09:16 offer me a solution that 1. is standards compliant, 2. sane to do, 3. isnt just cat ns mouse in playing with race times Aug 29 00:10:43 the .desktop says "use icon x.png" Aug 29 00:10:50 x.png inst installed yet Aug 29 00:10:59 raster: Read the desktop file as soon it appears. If it references another file you fail to read then retry later. Aug 29 00:11:02 that fact is you have to cache results of an ixon search due to complexity Aug 29 00:11:05 icon search Aug 29 00:11:12 aaah u cant do that Aug 29 00:11:25 I guess it is safe to assume the file will show up some time. Aug 29 00:11:26 why? there are tonnes of broken .desktops aroudn that reference icons that dont - and never will exist Aug 29 00:11:38 so u sti and always poll looking for an icon that will never come ot be Aug 29 00:11:43 so it's not safe Aug 29 00:11:49 also searching for an icon is expensive Aug 29 00:11:54 you have to hunt a lot of directories Aug 29 00:12:01 try multiple extensions etc. etc. Aug 29 00:12:14 multiply this by 100's of applications and it gets nasty Aug 29 00:12:33 thus u need to cache a result once found - be that a "it doesnt exist" or a "i found it here" Aug 29 00:12:45 Ok but then its safe to give up at some point. Aug 29 00:13:02 again - race condition Aug 29 00:13:09 so how long do u wait to give up? Aug 29 00:13:49 in the end u are just creating ugly hacks around what fundamentally is a broken install order issue - avoiding fixing the real problem Aug 29 00:14:28 raster: Sometimes you need to make things simple - just ry after ten seconds, one minute, five minutes. Or if you are cool then wait till the installer application process dissapears :) Aug 29 00:14:50 u cant do that Aug 29 00:14:57 as u dont know what is installing it Aug 29 00:15:24 and ten seconds, 1 minute, 10, whatever doesnt change it Aug 29 00:15:29 its a hack around the original problem Aug 29 00:15:33 not a fix Aug 29 00:15:40 raster: yes, but a vaild one Aug 29 00:16:04 u cant check for the installer process Aug 29 00:16:16 othjerwise u tie the code to a particular packaging system Aug 29 00:16:25 and now u are polling the proc table for it Aug 29 00:16:27 raster: on the other hand... on what do you want to rely if you don't know what is installing the packages? THen you can't rely on the installer doing it in the right order. Aug 29 00:16:30 and doing fuzzy matches on process names Aug 29 00:16:33 thats just not sane Aug 29 00:16:52 the problem is the installer installs files Aug 29 00:17:02 and installs a file that references others thaat dont exist yet Aug 29 00:17:09 and these are key important files Aug 29 00:17:20 so fixing the installer to install int he right order fixes it simply for everything Aug 29 00:17:30 illume/efreet/e isnt just for oe Aug 29 00:17:34 it runs on everything Aug 29 00:17:50 .this problem has never happened on my desktop Aug 29 00:17:58 that i can remember Aug 29 00:18:04 i suspect debian has an install order Aug 29 00:18:42 raster: No... the installer does not necessary know how files reference each other. The reference could be even in a binary or whatever. Aug 29 00:19:12 indeed Aug 29 00:19:23 so u have a .desktop saying "u can execute this binary" before the binary exists Aug 29 00:19:25 same with icon Aug 29 00:19:41 so now ujust have different bug reports Aug 29 00:19:50 "i install3ed app x - but i try and run it and it doesnt work" Aug 29 00:19:54 but if u want 1 minute it will Aug 29 00:20:08 as it takes a minute between installing .desktop and the executable Aug 29 00:20:17 It would be better if there was a specified mechanism for the installer signalling that some package is ready for use which triggers the update. Aug 29 00:20:18 OE is broken again ? Aug 29 00:20:30 florian: there is a mechanism. Aug 29 00:20:33 install /desktop last Aug 29 00:20:36 .desktop Aug 29 00:20:37 :) Aug 29 00:20:51 that6 should work with all existing software with no breakages Aug 29 00:20:56 (that look at .desktop files) Aug 29 00:21:14 it is by far the cleanest, simplest and most robust fix Aug 29 00:21:23 raster: right, but this solves it for desktop files (== a known type) only. A generic mechanism would be much more safe. Aug 29 00:21:28 and its the correct fix as u install dependencies for the .desktop before u install the .desktop Aug 29 00:21:46 well the generic mechanism is package dependencies Aug 29 00:21:58 ie u install my-app and it depends on myapp-core Aug 29 00:22:07 do it like Microsoft, and just have opkg prompt to reboot. :p Aug 29 00:22:14 I agree, that would solve the problem on moast devices for most packages currently. Aug 29 00:22:19 myapp ONLY installs a .desktop - it suckes in myapp-core whcih has all the executables, datafiles, icons etc. Aug 29 00:22:30 but that is truly ugly Aug 29 00:22:39 yeah Aug 29 00:22:48 i haven't really been paying attention to the conversation, but isn't there a way to force a refresh in the ui from a postinst in some fashion? Aug 29 00:22:55 now we have a huge blowout in package numbers Aug 29 00:22:57 and confusion Aug 29 00:23:13 kergoth: no. as designers dont want it :) Aug 29 00:23:26 kergoth: and evebn so - thats a workaround. the bug reports will still come in Aug 29 00:23:39 u want a solution where it "just works" Aug 29 00:23:41 no workarounds Aug 29 00:23:48 no race conditions Aug 29 00:24:16 florian: btw. "dependencies" are just a way to get the package manager to generate an install order for you Aug 29 00:24:17 :) Aug 29 00:24:41 so as such an install order for files within a package is really a viable solution as its a small enough set. Aug 29 00:25:20 * florian agrees Aug 29 00:25:29 i just want a clean solution Aug 29 00:25:30 not hacks Aug 29 00:25:45 you are right - i can go and add hacks to rtry and wait and retry and wait Aug 29 00:25:56 but in the end u end up with bgu reports for "it takes a minute to update my icon" Aug 29 00:26:15 or some packages just are so big to install your idea of waiting long enough is not long enough Aug 29 00:26:26 or some other thing that hen breaks the race condition again Aug 29 00:26:46 florian: hi :) is this guy some new blood to OE ? Aug 29 00:28:35 hi Marex Aug 29 00:29:35 raster: I really see the requirement... but its really hard since you need to be aware of all your dependencies. Aug 29 00:29:50 yeah - i know Aug 29 00:30:04 but - for packages that install .desktops... it wouldnt be that hard Aug 29 00:30:08 just remember the Aug 29 00:30:54 FILES_LAST_${PN} += "${datadir}/applications/*.desktop" Aug 29 00:30:56 instead of Aug 29 00:31:00 FILES_${PN} += "${datadir}/applications/*" Aug 29 00:31:10 so u can have a FILES_FIRST, FILES_LAST Aug 29 00:31:19 where order is preserved as per the file list Aug 29 00:31:30 and otherwise normal files get installed in "any order" as currently Aug 29 00:31:34 doesnt break anything Aug 29 00:31:40 is a new feature to use Aug 29 00:31:45 lets packages transition to use it Aug 29 00:31:51 and is simple Aug 29 00:31:52 :) Aug 29 00:32:45 yes, that's somethign we should do... for oe we can hide that in a class anyway Aug 29 00:33:45 yup Aug 29 00:35:21 hrm... time to get some sleep. I fear the little ones won't sleep longer than 8am :-/ Aug 29 00:35:34 eek! Aug 29 00:36:27 yeah... Aug 29 00:36:48 good night Aug 29 00:37:41 nite Aug 29 02:28:35 03  07org.openembedded.dev * r7a9fac56ca 10OE.dev/packages/gcc/ (7 files in 2 dirs): Aug 29 02:28:35 Refresh patches based on 4.3.1 for 4.3.2 Aug 29 02:28:35 All it needed was regenerating configure scripts. Aug 29 02:28:35 -Khem Aug 29 02:28:41 03  07master * r7a9fac56ca 10OE.dev/packages/gcc/ (7 files in 2 dirs): Aug 29 02:28:41 Refresh patches based on 4.3.1 for 4.3.2 Aug 29 02:28:42 All it needed was regenerating configure scripts. Aug 29 02:28:44 -Khem Aug 29 02:52:38 khem: thanks **** ENDING LOGGING AT Fri Aug 29 02:59:58 2008