**** BEGIN LOGGING AT Tue Sep 02 02:59:56 2008 Sep 02 03:01:47 cool Sep 02 03:03:57 i'm trying to rebuild the base-image, but i get this error in log.do_rootfs.* Sep 02 03:03:59 http://rafb.net/p/3B2KB652.html Sep 02 05:06:44 hi, i have oe running, but how can i do dualboot? Sep 02 06:43:45 morning Sep 02 07:35:21 bonjour Sep 02 08:33:02 03mickeyl 07org.oe.dev * r86ded5a8... 10/ (1 packages/linux/linux-openmoko_2.6.22.5.bb): remove linux-openmoko 2.6.22.5 Sep 02 08:33:07 03koen 07org.oe.dev * r5b03bc6b... 10/ (1 packages/ffmpeg/ffmpeg.inc packages/ffmpeg/ffmpeg_git.bb): ffmpeg: move improved staging method from _git to .inc Sep 02 08:33:12 03koen 07org.oe.dev * rf0caa9e3... 10/ (1 contrib/angstrom/build-feeds.sh): angstrom feed builder: more tweaks Sep 02 09:07:41 good morning Sep 02 09:20:52 anyone fancy giving LTG a kick, www.angstrom-distribution.org seems to be no longer talking HTTP Sep 02 09:25:55 * dcordes pokes LTG wiith a stick Sep 02 09:28:03 * florian uses ssh for this purpose Sep 02 09:32:57 http is overrated... Sep 02 09:33:22 hehe Sep 02 09:33:48 mickeyl: but you didnt write the rsync fetcher for bitbake yet Sep 02 09:34:57 morning Sep 02 09:35:04 yo hrw Sep 02 09:37:10 hrw: youre an intel employee, no? Sep 02 09:44:12 Hello - I'm a newbie to open embedded, I've successfully built 'console-base' for my OMAP3x EVM board but have an issue re kernel configuration - can anyone help? Sep 02 09:46:54 rjnbarrass: maybe, but the kernel configuration itself should not be reelated to oe Sep 02 09:48:49 rjnbarrass: Maybe explain what you intend to do Sep 02 09:49:58 hi - yeah... I currently have a development environment set up with the development board configured to boot via the network (tftpboot for kernel, and nfs share for root)... Sep 02 09:50:20 the newly built kernel boot and runs, but then panics because it can't find the root. Sep 02 09:51:38 rjnbarrass: does it even try to mount the filesystem via nfs? Sep 02 09:52:06 check what you are passing to the kernel as ROOT= Sep 02 09:52:25 (bootargs are configured for nfs). (root=/dev/nfs nfsroot=/export/b1,nolock) Sep 02 09:53:23 does your nfs server show any connection attempts? Sep 02 09:53:24 I don't think so (based on experience of the 'shipped' kernel from Texas) Sep 02 09:54:21 as I say - this was was working with the prebuilt kernel from TI... so I don't think there is an issue with the config of the server/nfs export) -- I can ping the dev kit from the server Sep 02 09:55:12 and your newly compiled kernel supports nfs? Sep 02 09:57:24 that is my query... this has been built with an Angstrom configuration for the omap3evm machine under open embedded (bitbake) Sep 02 09:57:45 but I'm not sure how I go about tailoring the configuration to allow setting of kernel options... Sep 02 09:59:48 does that make sense? Sep 02 10:00:11 I'm a complete newbie to the OE world :) Sep 02 10:02:12 rjnbarrass: You can check - find out which kernel bb is used and search for its configuration file. Sep 02 10:02:51 now that is a good question... Sep 02 10:03:02 rjnbarrass: Usually it is in packages/linux///defconfig Sep 02 10:03:17 florian: ah... thanks for the pointer :) Sep 02 10:06:15 florian: there looks like loads of versions under the 'linux' directory... the booting kernel reports itself as 2.6.26-omap1 which is a little interesting Sep 02 10:13:22 florian: okay - i'm going to do some digging around - you have given my a hint... thanks! Sep 02 10:14:50 rjnbarrass: yw Sep 02 10:16:14 Esben: Why does this happen with ltt? Marker control : 'connect' 'net_socket_recvmsg' id: dynamic channel: Sep 02 10:16:31 Esben: I have no trace information at all, what am I missing? Sep 02 10:40:47 jpsaman: have you run ltt-armall? Sep 02 10:41:06 I trace with: Sep 02 10:41:06 Esben: yes, that is where the message came from Sep 02 10:41:24 mount /debugfs Sep 02 10:41:24 ltt-armall Sep 02 10:41:24 lttctl -n $2 -d -l /debugfs/ltt -t /local/traces/$2 Sep 02 10:41:24 sleep $1 Sep 02 10:41:24 lttctl -n $2 -R Sep 02 10:42:01 Esben: this is what I do Sep 02 10:42:03 mount -t debugfs -o rw debugfs /mnt/debugfs Sep 02 10:42:03 ltt-armall Sep 02 10:42:03 lttctl -n trace -d -l /mnt/debugfs/ltt -t /media/trace Sep 02 10:50:04 Esben: any ideas?? Sep 02 10:50:23 no, not really Sep 02 10:50:48 I haven't experienced something like that, or at least, I don't remember seeing anything like that. Sep 02 10:51:03 you have mounted your debugfs? Sep 02 10:55:28 yes Sep 02 10:55:42 debugfs on /mnt/debugfs type debugfs (rw) Sep 02 11:04:21 RP: are you here? Sep 02 11:05:39 Esben: kind of Sep 02 11:06:54 I am getting really close with this "really cool OpenEmbedded" feature :-D Sep 02 11:07:10 but I have an outstanding issue with bitbake, that I hope you can help me with. Sep 02 11:07:40 how do I assign a string like "${foo}" to a variable, without having bitbake do variable substitution? Sep 02 11:08:48 Esben: Why do you want to do that? Sep 02 11:09:12 Esben: The data is actually stored unexpanded so you could just not expand the variable when you read it? Sep 02 11:10:16 I need to call a configure script with --prefix=${foobar} Sep 02 11:11:06 or more correctly, with --exec-prefix=${prefix} which makes it possible to give an alternative prefix when doing make install. Sep 02 11:11:17 and DESTDIR is not possible. Sep 02 11:11:35 Esben: why not override the prefix variable in the install task then? Sep 02 11:11:41 I am building for Windows (MinGW), so I need to have compiled in prefix set to C:\foobar Sep 02 11:11:59 ah, I think I see what you're doing Sep 02 11:12:43 I am trying to do this without changing so much in existing classes and recipes, which means I would like to not have to change all the places where ${prefix} and other location variables are used in FILES_* declarations. Sep 02 11:13:28 Why isn't DESTDIR possible on windows? Sep 02 11:13:37 I am cross building on Linux ;-) Sep 02 11:13:57 so I need to install to fx. DESTDIR/OpenEmbedded Sep 02 11:14:04 with prefix set to C:\OpenEmbedded Sep 02 11:14:21 and I don't want to change existing recipes Sep 02 11:14:38 right, tricky and I see your problem :) Sep 02 11:16:21 I'd hope some way of escaping ${} exists but I don't know offhand Sep 02 11:17:13 I hoped also, but haven't been able to do it. Sep 02 11:17:30 Esben: Look up EXTENDPE in bitbake.conf for one particularly ugly hack Sep 02 11:17:37 I can specify $prefix, but that does not seem to have the same meaning to configure as ${prefix} Sep 02 11:17:53 The \x7d in particular... Sep 02 11:18:27 and \x7d being } I guess... Sep 02 11:18:51 that might just work here also :-) Sep 02 12:30:16 gm Sep 02 12:34:29 hi all Sep 02 12:35:19 RP: hrw: while you're trying to convince Intel guys that Poky needs to be hosted on the intel mirrors-ring, insist a bit more and get them convinced them that openembedded metadata needs as well to be mirrored... Sep 02 12:36:18 anybody has knowledge about debugging frameworkd ? I cannot output the log file Sep 02 12:55:26 RP: maybe we could add something like print_curly_envvar() to bitbake to print out strings like ${foo} Sep 02 13:08:14 configure_prefix = "${@['','${prefix\x7d_'][1]}" Sep 02 13:08:31 actually gets expanded to the value of bitbake ${prefix} variable Sep 02 13:08:47 (forget about the trailing _) Sep 02 13:08:50 :( Sep 02 13:09:20 We may need to add some way to handle this :/ Sep 02 13:09:25 and so does configure_prefix = "${@'${prefix\x7d'}" Sep 02 13:09:47 yes, some way to handle this would be definitely nice. Sep 02 13:10:37 I have here a complete meta-toolchain derived canadian cross build for host=windows target=powerpc up and running on my build=x86_64-linux machine Sep 02 13:10:47 gcc and binutils for now. Sep 02 13:10:56 Esben: Very nice :) Sep 02 13:11:12 but I need a way to have gcc work with windows paths... Sep 02 13:12:13 I guess I will go for setting prefix=C:\OpenEmbedded for now, and then modify all necessary FILES_* variables. Sep 02 13:12:19 I understand. I also know how bitbake does variable expansion internally and know its going to be hard to get it to do what you need :/ Sep 02 13:13:53 I will add a ${fake_prefix} variable, which all packages inheriting from canadian-sdk.bbclass must use when refering to linux build paths, and then have ${prefix} be the real installation path (which for all other host arch's than MinGW should be the same). Sep 02 13:15:34 03koen 07org.oe.dev * rcd61a824... 10/ (1 packages/ffmpeg/ffmpeg_git.bb): ffmpeg git: make default for avr32 Sep 02 13:58:45 03koen 07org.oe.dev * r6e8b3cd5... 10/ (5 files in 2 dirs): gnas-minimal: add ffmpeg support Sep 02 14:40:03 I am trying to install perl, and its modules don't install. I was hoping someone could explain how I can get the perl modules to install too. Sep 02 14:53:04 03koen 07org.oe.dev * rc3a62b48... 10/ (3 files in 3 dirs): gnash-minimal cvs: fix missing trunc() definition, https://savannah.gnu.org/bugs/?24188 Sep 02 15:51:01 morning Sep 02 15:59:09 hi kergoth`work Sep 02 16:01:14 yo kergoth`work Sep 02 16:01:16 hi florian Sep 02 16:01:26 * pb_ returns from Paris Sep 02 16:01:49 hi pb_ Sep 02 16:04:41 * Genesis still in paris :( Sep 02 16:21:05 * * OE Bug 4536 has been created by denis(AT)denix.org Sep 02 16:21:08 * * gcc-cross-sdk 4.3.1 missing include-fixed Sep 02 16:21:09 * * http://bugs.openembedded.net/show_bug.cgi?id=4536 Sep 02 16:44:19 morning all Sep 02 16:44:54 thesing: hi! Sep 02 16:46:26 Jay7: do you have your kexec patch in your oe-tree? Sep 02 16:46:36 thesing: yes Sep 02 16:47:27 I just witing to Andrea (ant) to test kexecboot patch before posting full mtn diff Sep 02 16:47:32 *waiting Sep 02 16:47:47 but I you want I can post entry mtn diff Sep 02 16:48:29 Jay7: it wouldn't hurt. Sep 02 16:49:09 thesing: about strcat.. Sep 02 16:50:20 seems like we can use strcat() Sep 02 16:50:58 but not strncat() - this will use very heavy strlen then Sep 02 16:51:25 I dislike using strcpy/strcat because of lacks of bounds checking Sep 02 16:51:31 but in this case... Sep 02 16:51:44 I should check it once more Sep 02 16:53:47 Jay7: could the strlenn function substituted by "s ? strlen(s) : 0 " ? Sep 02 16:54:14 thesing: yes, it can Sep 02 16:54:26 by #define Sep 02 16:55:16 ok, then I'll replace strnenn with such define and str_append with strcat Sep 02 16:55:25 wait for v.3 :) Sep 02 16:59:51 Jay7: would you like do further stuff on kexecboot? I could give you access to the git tree. Sep 02 17:00:21 thesing: may be some time later.. Now I'm not ready yet :) Sep 02 17:00:57 would you like to commit these changes into main tree? Sep 02 17:01:32 Jay7: yes. It would be a shame not to do so. Sep 02 17:18:02 how can I execute an action that removes files from rootfs, before building the filesystem image? Sep 02 17:18:05 I've modified an image by adding perl as part of the IMAGE_INSTALL variable, however none of the perl modules are being installed. I tried to list the modules individuals, but that failed to work properly for me. Does anyone know how to get the modules to install too? Sep 02 17:20:01 Smashmad: there is IMAGE_PREPROCESS_COMMAND (or so) variable. Sep 02 17:26:18 thesing: fixed :) Sep 02 17:26:53 have some testing now and attach then to bugzilla Sep 02 17:28:44 Jay7: in case you want to do some further work on kexecboot: it would be nice if it only started if a key is pressed (or hold) at boot, and it would be nice if it would be directly used as /init in initramfs. Sep 02 17:29:37 thesing: what should be happens when key on boot will not be pressed? Sep 02 17:29:45 thanks thesing, now I found that its better to prevent the installation of the package which install those files. Is there any way of doing that from user.collection? Sep 02 17:29:54 should boot from last time selected device? Sep 02 17:31:51 Jay7: if you find a way to store this information yes. Otherwise from some (at compile time) predefined location (if available). Also it would be nice to have the arguments kexecboot gets from the initramfs atm defined as defaults at compiletime too. Sep 02 17:35:18 thesing: what exactly kexecboot gets from initramfs atm? Sep 02 17:36:28 Jay7: the angle of the framebuffer and the inputdevice. Sep 02 17:36:33 ah Sep 02 17:36:59 Jay7: see initramfs-kexeboot package Sep 02 17:37:40 It's better to decide angle by known models from /proc/cpuinfo e.g. Sep 02 17:38:39 with u-boot we can pass and store some info from/to u-boot env.. Sep 02 17:39:19 with sharp bootloader - only hardcode or try to decide from /proc Sep 02 17:40:28 Jay7: yes thats a good idea. As for uboot as long as it does not depend on uboot. (i.e. should work somehow with sharp loader ) Sep 02 17:40:42 thesing: Jay7: Could we somehow stroe a last used tag in the selected rootfs and use that to detrmine the nexe default choice? Sep 02 17:41:29 hm.. we can try to store some info on selected boot device Sep 02 17:41:51 and then read this while doing devicescan() Sep 02 17:42:01 yes as grub does. maybe on flash. Sep 02 17:43:05 thesing_dinner: well, I was thinking that in order to exetnd the life of the flash we should only use the rootfs the user wants Sep 02 17:43:57 thesing_dinner: btw, rootwait flag can be set selectively while doing devicescan() Sep 02 17:44:23 I'm considering about adding rootwait flag to boot structure Sep 02 17:45:27 Jay7: IIRC, rootwait is supoosed to replce rootdelay=nn in 2.6.23 and above Sep 02 17:45:59 hvontres|work: we already moved to rootwait in kexecboot but now it is on by default Sep 02 17:46:31 this just added to kexec's cmdline Sep 02 18:00:40 ok. kexecboot compiles at least :) Sep 02 18:03:32 hvontres|work: hvontres|work only reading the flash doesn't do it any harm. Sep 02 18:03:39 bbl Sep 02 19:01:30 hi kergoth Sep 02 19:01:38 hey Sep 02 19:14:46 he florian Sep 02 19:14:51 re Sep 02 19:15:31 hi ant Sep 02 19:15:42 hello Sep 02 19:16:28 ant__: hi Sep 02 19:16:44 Jay7: compiling right now v3 Sep 02 19:16:46 ant__: please test latest version (v.3) :) Sep 02 19:16:50 ;-) Sep 02 19:16:56 rocks! :) Sep 02 19:52:06 03koen 07org.oe.dev * ra25e1466... 10/ (1 packages/bluez/bluez-utils_3.33.bb): bluez-utils: reenable hidd Sep 02 20:24:04 re Sep 02 20:26:20 hey all Sep 02 20:29:48 hi thesing Sep 02 21:13:02 03koen 07org.oe.dev * r9c6207c1... 10/ (4 files in 4 dirs): Sep 02 21:13:02 ffmpeg: bump SRCREV Sep 02 21:13:02 gst-ffmpeg, libomxil: bump PR Sep 02 21:13:06 03koen 07org.oe.dev * rfb263edc... 10/ (7 files in 5 dirs): Sep 02 21:13:06 fennec: add skeletor to build fennec, the mobile browser from mozilla Sep 02 21:13:06 mozilla.bbclass: fixup do_configure Sep 02 21:13:19 wow Sep 02 21:13:22 fennec! Sep 02 21:13:40 where is chrome? :) Sep 02 21:13:41 * Jay7 is wanted fennec :) Sep 02 21:13:59 * zecke wants a pon Sep 02 21:14:03 * zecke wants a pony Sep 02 21:14:12 :) Sep 02 21:24:29 I'm having a difficult time tracking down a jffs2 partition corruption when using an ixp4xx machine with jffs2 filesystem - looks like the fs is left in a corrupt state on writes - anyone experience anything like this? I don't really know where to start looking - enabling JFFS2 debug spams like mad Sep 02 21:55:01 thesing: hi, we are smoothing the last issues Sep 02 21:55:47 ant__: I will look at it tomorrow. Sep 02 21:56:46 thx Sep 02 22:02:35 thesing: almost there (-l must be last) http://paste.org.ru/?xgt09v Sep 02 22:02:44 hehe.. Sep 02 22:09:17 RP: I've found a workaround. I am concatenating the "${prefix}" string using shell variables, and then using that variable in the arguments to configure. Sep 02 22:18:57 nite Sep 02 22:22:48 nite all Sep 03 00:13:54 * Jay7 -> sleep **** ENDING LOGGING AT Wed Sep 03 02:59:57 2008