**** BEGIN LOGGING AT Wed Sep 12 02:59:56 2007 Sep 12 03:46:34 * wick7xx is away: sleep/work Sep 12 06:48:05 03koen 07org.oe.dev * r81e53a59... 10/ (1 packages/gpephone/libidn_0.5.19.bb): libidn: remove bogus dep on 'glibc' to fix uclibc builds Sep 12 06:52:33 03koen 07org.oe.dev * r4f5beacb... 10/ (1 packages/openmoko2/neod_svn.bb): neod: s/MACHINE/MACHINE_ARCH/ Sep 12 07:06:43 seems that http://www.openembedded.org/viewmtn is broken Sep 12 07:24:11 you should get something like this Sep 12 07:24:15 root@xilinx-ml403:~$ cat /proc/cpuinfo Sep 12 07:24:15 processor : 0 Sep 12 07:24:15 cpu : Virtex-4 FX Sep 12 07:24:15 clock : 400MHz Sep 12 07:24:15 revision : 20.48 (pvr 2001 1430) Sep 12 07:24:15 bogomips : 797.90 Sep 12 07:24:17 machine : ML403 Reference Design Sep 12 07:24:19 plb bus clock : 100MHz Sep 12 07:24:32 sorry . wrong window :( Sep 12 07:25:26 ps ax Sep 12 07:29:40 hm.. is someone working on avr32 support? Sep 12 07:29:54 roh: read oe-devel Sep 12 07:30:09 ah Sep 12 07:31:16 koen : any idea why ipkg would not be included in angstrom-minimal-image ? Sep 12 07:31:58 steliosk: proably because rootfs_ipk.bbclass changed Sep 12 07:32:35 koen i cannot find any reference to an archive on the website.. have a tip for me? Sep 12 07:32:58 roh : answer is "soon" Sep 12 07:34:56 nice Sep 12 07:35:02 roh: it's on the bloody frontpage Sep 12 07:35:28 i have the small 2x eth, no lcm board around now Sep 12 07:38:16 koen yep.. found it.. the search and the contact link seems broken Sep 12 08:02:06 morning Sep 12 08:02:22 steliosk: no ipkg because this is minimal image Sep 12 08:02:48 hrw : hi ! Sep 12 08:03:11 hrw : but then it makes it difficult to add stuff Sep 12 08:03:43 hrw : minimal-image should not "prevent" this Sep 12 08:03:53 steliosk: the reason for minimal was other a bit - feel free to add ipkg into it Sep 12 08:05:16 hrw : well it should provide the way to add other packages. its very odd to have to install ipkg by hand Sep 12 08:06:16 steliosk: ok. Sep 12 08:06:40 steliosk: I used minimal image in initrd so needed as small as possible one Sep 12 08:08:03 hrw : in the initrd it makes sense not to have ipkg, but i believe it should be another image i.e initrd-image or something similar Sep 12 08:09:44 steliosk: like I said - feel free to add ipkg Sep 12 08:10:19 hrw : should we add an initrd-image as well ? It might come handy in many situations Sep 12 08:10:25 when using --pad, should the jffs2 image size be equal to 0 modulo the pad size ? Sep 12 08:18:22 is it possible to compile the libavcodec package without having to build the whole ffmpeg source ? Sep 12 08:19:38 steliosk: initramfs-images looks like someone tried Sep 12 08:20:25 and please don't do RDEPENDS += "ipkg" or something Sep 12 08:20:37 some people want to use debs :) Sep 12 08:21:06 hrw, RP: how do you solve http://bugs.openembedded.org/show_bug.cgi?id=2961 in poky? Sep 12 08:27:10 let me see Sep 12 08:28:42 my deb rootfs from Poky has /sbin/init -> /sbin/init.sysvinit Sep 12 08:29:59 iirc vivijim had the same problem Sep 12 08:31:20 1569 update-alternatives: Linking /home/hrw/devel/OH/poky/build/tmp-poky/rootfs//bin/ash to busybox 1570 update-alternatives: Linking /home/hrw/devel/OH/poky/build/tmp-poky/rootfs//usr/bin/awk to ../../bin/busybox Sep 12 08:32:10 update-alternatives: Linking /home/hrw/devel/OH/poky/build/tmp-poky/rootfs//sbin/init to /sbin/init.sysvinit Sep 12 08:35:10 03hrw 07org.oe.dev * r858d3619... 10/ (1 packages/settings-daemon/settings-daemon_svn.bb): settings-daemon: switch to SRCREV (from Poky) Sep 12 08:39:38 ~lart daniel@openmoko Sep 12 08:39:38 * ibot installs a bad bootloader on daniel@openmoko and turns daniel@openmoko into a brick Sep 12 08:39:46 ibot: good lart! Sep 12 08:39:47 Good lart!, good! Have a cookie... oh no! The cookie jar is dry! Sep 12 08:41:48 hrw: which update-alternatives-native does that use? Sep 12 08:43:25 cworth Sep 12 08:45:24 weird Sep 12 08:45:32 80 tasks to openmoko-image... Sep 12 09:58:22 psokolovsky_: hi Sep 12 10:02:13 Hi! Sep 12 10:06:45 cbrake_away: I've been able to manually build a jffs2 image that is accepted by nandwrite Sep 12 10:07:05 unfortunately it still doesn't mount: http://rafb.net/p/ZxdRJh39.html Sep 12 10:08:17 03hrw 07org.oe.dev * re02f3ba6... 10/ (1 packages/openmoko2/openmoko-terminal2_svn.bb): openmoko-terminal2: fix DEPENDS/RDEPENDS error Sep 12 10:08:22 03koen 07org.oe.dev * r32a36cb7... 10/ (3 files in 2 dirs): matchbox-keyboard/: fix PVs Sep 12 10:08:28 03koen 07org.oe.dev * rbf8c5e1e... 10/ (1 packages/tasks/task-openmoko.bb): task-openmoko: add openmoko-dates2 Sep 12 10:10:05 koen: thx - just had to add it Sep 12 10:15:47 hrw: 2007.2 works quite nice on my a780 Sep 12 10:16:03 haven't needed swap yet with the 48mb of ram Sep 12 10:20:11 re Sep 12 10:29:08 bleh, C is ugly Sep 12 10:29:53 mickey|sports will have a heart-attach when he sees my patch Sep 12 10:30:51 what patced? Sep 12 10:31:22 [UGLY PATCH] search sysfs for backlight in neod :) Sep 12 10:31:57 trying to get my a780 working as a phone Sep 12 10:32:54 koen: /sys/class/backlight/[first entry]/ one? Sep 12 10:33:19 + while((d = readdir(dp))) { Sep 12 10:33:19 + if (!(strcmp(d->d_name, ".") == 0) && Sep 12 10:33:20 + !(strcmp(d->d_name, "..") == 0)) Sep 12 10:33:20 + asprintf(&bldir, "%s%s", "/sys/class/backlight/" , d->d_name); Sep 12 10:33:21 that one Sep 12 10:33:31 adapted from the minilite code I wrote years ago Sep 12 10:33:44 so now it's a mix of libc and glib functions Sep 12 10:34:40 morning Sep 12 10:34:46 hey mickeyl :) Sep 12 10:35:06 hey florian Sep 12 10:35:13 ~seen zecke Sep 12 10:35:15 zecke was last seen on IRC in channel #openmoko, 40m 33s ago, saying: 'hi'. Sep 12 10:35:19 hi all Sep 12 10:37:18 mickeyl: http://scap.linuxtogo.org/ Sep 12 10:39:34 * koen leaves to attend a graduation Sep 12 10:39:37 later all Sep 12 10:43:17 cu koen Sep 12 10:44:02 koen: nice screenies. guess we're 80% there w/ OM on the EZX platform Sep 12 10:53:05 gconf-dbus_svn.bb doesn Sep 12 10:53:12 doesn't build for me Sep 12 10:53:24 armv5te Sep 12 11:16:50 03mickeyl 07org.oe.dev * rc4e2e121... 10/ (3 files in 2 dirs): uboot-openmoko 1.2.0+git+svn bump frozen version Sep 12 11:35:43 hrw : Do you remember with which commit you removed the ipkg from minimal-image ? Sep 12 11:39:11 no Sep 12 11:54:03 hi all Sep 12 11:54:09 hey likewise Sep 12 11:54:49 gm Sep 12 12:25:08 03mickeyl 07org.oe.dev * r6a29d411... 10/ (3 files in 3 dirs): psplash: convert to SRCREV, update openmoko branding image Sep 12 12:46:12 nud: so you made some progress ... Sep 12 12:46:40 nud: I noticed in my build that the summary images were padded to block size, but the non-summary images were not. Sep 12 12:46:50 nud: not sure why yet ... Sep 12 12:47:11 cbrake: I made it work twaking the options in compulab-x270.conf Sep 12 12:47:23 using --pad instead of --pad=0x20000 worked Sep 12 12:47:30 you might want to try Sep 12 12:47:48 nud: so it must pad it out to the erase block size then Sep 12 12:47:51 but I still can't manage mounting the flash drive once it has been flashed Sep 12 12:47:57 cbrake: yes Sep 12 12:48:00 nud: looking at your log ... Sep 12 12:50:00 nud: are you using the 2.6.22 kernel built with OE? Sep 12 12:50:09 yes I am Sep 12 12:50:11 nud: almost looks like the kernel does not include jffs2 support Sep 12 12:50:32 mmh, it was built using angstrom-minimal-image-with-mtd-tools Sep 12 12:50:49 nud: try this: mount -t jffs2 mtd3 /mnt/disk Sep 12 12:51:18 nud: /dev/mtd3 is actually a char device. Technically, mtd is neither char nor block, so the correct way to specify is just mtdx Sep 12 12:51:23 uh Sep 12 12:51:25 works Sep 12 12:51:38 and the root option would just be root=mtd3 then ? Sep 12 12:51:43 boot option* Sep 12 12:51:50 I think so Sep 12 12:52:00 * nud feelds dumber everyday, thanks to mtd Sep 12 12:52:26 nud: yes, that is correct. See the defconfig for the compulab kernel build in OE Sep 12 12:52:27 * nud tries to reboot on the NAND partition Sep 12 12:54:03 nud: I'll push your fix for compulab-x270.conf Sep 12 12:54:12 wow, looks like it works Sep 12 12:54:17 thanks a lot cbrake Sep 12 12:54:29 nud: great! Sep 12 12:54:33 I hope I'll be able to work most things out myself now :-) Sep 12 12:55:35 cbrake: so now, what I have to do is to make a new custom image with what I want on my disk ? Sep 12 12:56:30 nud: right -- create a new image recipe (http://bec-systems.com/web/content/view/79/9/) Sep 12 12:56:42 cbrake: and how does it work for config files and such ? Sep 12 12:57:12 nud: I use bbcollections for most of my projects, which is essentially an overlay. Sep 12 12:57:19 ah, ok Sep 12 12:57:29 nud: so if I need to modify a package, I copy it into the bbcollections tree and modify it Sep 12 12:57:33 * nud has his own directory in the upstream overlay, currently Sep 12 12:57:44 cbrake: so you config everything through patches ? Sep 12 12:58:10 nud: what type of config are you talking about specifically? Sep 12 12:58:23 cbrake: files in /etc Sep 12 12:58:45 nud: I usually just copy and modify the package that provides the file Sep 12 12:59:02 ok Sep 12 12:59:04 nud: sometimes it is necessary to patch, but most of the time it is not, as these files are directly in OE Sep 12 12:59:13 nud: for example fstab, network config, etc. Sep 12 12:59:41 cbrake: ok Sep 12 13:00:03 I'll stop annoy you now that cm-x270-specific stuff are working ;-) Sep 12 13:00:22 you've been very kind to answer my questions last days Sep 12 13:02:12 nud: no problem -- glad to see another satisfied OE user :-) Sep 12 13:03:18 cbrake: just a last question: what program do you use to interract with your boards through serial port ? Sep 12 13:03:27 nud: minicom Sep 12 13:03:54 should try this one then, cutecom sucks Sep 12 13:04:02 thanks Sep 12 13:06:36 minicom is best Sep 12 13:12:30 cbrake: have you ever used the video input device on the cm-x270 ? Sep 12 13:14:18 nud: no, I have not Sep 12 13:14:43 ok Sep 12 13:14:48 will look by myself then :-) Sep 12 13:14:56 nud: I had discussions with one of my customer though that used it Sep 12 13:15:28 What's a good example of a Makefile that builds from OE, but also stand-alone on the host? Sep 12 13:16:17 likewise: for applications, I use autotools for everything Sep 12 13:16:21 cbrake: ah, do you have some links ? :-) Sep 12 13:16:35 cbrake: also for your own applications? Sep 12 13:16:50 likewise: autotools rocks when it comes to cross-compiling Sep 12 13:17:08 it's not *that* difficult to figure out, but it's a lot of cut'n'paste though Sep 12 13:17:33 nud: no, I don't have much information -- other than it is _very_ difficult to pipe vido at 30fps at resolutions >= VGA through a PXA270 Sep 12 13:17:51 cbrake: well, I was thinking about QCIF Sep 12 13:18:00 that I need to encode using H261 iirc Sep 12 13:18:13 likewise: yes, I use autotools even for 40 line test apps Sep 12 13:18:21 sounds doable to you ? Sep 12 13:18:43 nud: yeah, QCIF should work, but prototype it first! Sep 12 13:18:52 nud: looked at Intel's IPP primitives for XScale? Sep 12 13:19:05 nud: highly optimized codecs Sep 12 13:19:09 cbrake: prototype it ? Sep 12 13:19:33 likewise: yeah I know (I even have the DVD on my desk) but I'm not sure I will have enough time to look at it Sep 12 13:19:40 nud: yes, built up a prototype and test a few things before spending lots of money making a product Sep 12 13:19:43 likewise: debian and OE took a lot of the available time ;-) Sep 12 13:19:57 cbrake: well, I'm using a devkit right now Sep 12 13:21:30 likewise: avoid setting CC/CFLAGS/LDFLAGS in makefile Sep 12 13:21:52 likewise: also build with CC CFLAGS and remember about LDFLAGS during linking Sep 12 13:22:32 hrw: my current approach is to use CFLAGS ?= and CFLAGS += Sep 12 13:24:56 hrw: for example, in one Makefile I have to specify I'm using -std=c99 Sep 12 13:37:00 hi Sep 12 13:37:17 morning Sep 12 13:42:01 anyone can tell me which package provides gtk-webcore-nrcit? gpe-mini-browser won't build without it Sep 12 13:43:17 If EXTRA_OEMAKE is not set, -e is passed to make. If EXTRA_OEMAKE is set, -e is not passed. Why exactly is this? Sep 12 13:43:51 hi Sep 12 13:44:03 Bernardo: look into gtk-webcore/ dir Sep 12 13:44:44 i have a patch which modifies a Makefile.in. however I guess the change will be overwritten because of ./configure which happens afterwards Sep 12 13:45:08 what is the way to handle such a patch then? Sep 12 13:46:21 rschuster: patch the Makefile.am files instead Sep 12 13:46:50 rschuster: or whatever is needed to generate the correct Makefile.in file -- easier said than done ... Sep 12 13:48:18 cbrake: um. you're right. dont know why I did not think of this :) Sep 12 13:49:21 rschuster: are you the author of: https://www.fsfe.org/de/fellows/robertschuster/weblog/(tag)/openmoko Sep 12 13:50:29 cbrake: yes, I am Sep 12 13:50:29 cbrake: cool. did not know that one can put the tags into the URL :) Sep 12 13:50:53 rschuster: can you give me an RSS feed link for that task? we're going to add your blog to the planet then Sep 12 13:51:13 s/task/tag/ Sep 12 13:52:03 rschuster: :-) your java works looks intersting! Is cacao the best option for Java on ARM? Sep 12 13:55:40 Bernardo, hrw: I see this issue too, but it looks like gpe-mini-browser is not consistent with gtk-webcore in some way. Sep 12 13:56:23 cbrake: yep. atm it is Sep 12 13:56:24 I would suggest midori instead Sep 12 13:56:58 cbrake: we have some trouble on kernel 2.4, linuxthreads devices but no issues on modern ones: linux 2.6 + nptl Sep 12 13:57:32 cbrake: best of all. cacao is going to support icedtea (the 100% free variant of openjdk) Sep 12 13:57:53 cbrake: and tarent is willing to sponsor all me OE work :) Sep 12 13:58:00 s/me/my/ Sep 12 13:58:09 hrw: ok, I'll try that. For now I'll see if building osb-nrcit is enough for gpe-mini-browser to build Sep 12 13:58:22 mickeyl: will do so Sep 12 13:58:26 * rschuster -> food Sep 12 13:58:34 bon appetit Sep 12 13:58:38 rschuster: excellent! That is great news. Sep 12 13:58:48 rschuster: oh great Sep 12 13:59:00 * zecke is starving as well... Sep 12 13:59:22 * zecke is traveling to mickeyl as he promised food... Sep 12 13:59:24 Bernardo: It isn't... I just tried Sep 12 13:59:35 heh Sep 12 13:59:43 zecke: We have some pare food and sweets here too ;) Sep 12 14:00:26 florian: youŕe right, here it fails too Sep 12 14:01:38 anyway, I was just trying to build meta-gpe... What I really do need to do is to set up a build env for openmoko, have a possible project that would use it Sep 12 14:03:36 has anyone out there gotten rhythmbox to build? it is complaining about not seeing a higher version of Gstreamer but there is a higher one built. (.8 vs .10) Sep 12 14:04:55 | configure: error: GStreamer not found, or older than version 0.8.1 Sep 12 14:13:05 mickeyl: I've been reading a little on the neo1973, but I don't understand what is the problem with gps? The project I would be starting in a couple of weeks would need it (at least need a assurance that gpsd/libgpsd will work with it) Sep 12 14:13:37 no gps without some binary blob Sep 12 14:13:53 said binary blub works on OM2007.1 Sep 12 14:13:56 but not on OM2007.2 Sep 12 14:14:00 since we went EABI Sep 12 14:14:06 now we need an updated binary blob Sep 12 14:14:24 hi all, i'm looking for help about an annoying problem. Till yesterday I had a completely working OE environment. Yesteday I made a daily update of the org.openembedded.dev branch and after that I can't use OE because whatever I do i get ERROR: Unable to parse conf/bitbake.conf Sep 12 14:14:39 and I assume that it isn't as simple as asking for a recompiled blob from the gps manufacturer... Sep 12 14:14:49 i am sure that my environment vars are OK Sep 12 14:15:13 any kind of advice is welcome Sep 12 14:15:14 :) Sep 12 14:15:55 gpe-mini-browser built fine for me over here Sep 12 14:16:02 but I could set up a build env for OM2007.1, develop for that, deploy the firsts units running OM2007.1, and later update, right? The functionalities I'll need from the neo, for now, will be basic gsm, gps, and mp3 playing Sep 12 14:16:21 Crofton: ping! Sep 12 14:16:27 mickeyl: you forgot to add that that binary blob is not official anyway Sep 12 14:16:35 bernard_: gsm is very flakey up to now Sep 12 14:16:57 hrw: well, i'm sparing everyone the full story... Sep 12 14:17:40 ;D Sep 12 14:17:52 well, I can live with gps + mp3 on a hw that later will have gsm... And which might even be replaced later on by something with wifi and usb host Sep 12 14:18:22 that is, if this project even goes forward Sep 12 14:18:29 gsm daemon and libs on neo are .... ... Sep 12 14:18:54 each application restarts daemon if is unable to connect to it Sep 12 14:19:31 so during system start gsmd is started few times, few times stopped and no warranty that it works at all Sep 12 14:20:52 in the end if winter is in proper direction then all is working. if it not then no dialer or gsm applet or both Sep 12 14:21:25 well, this is nonsense: Sep 12 14:21:26 Parsing .bb files, please wait...svn: Failed to fetch lock information: 400 Bad Request (http://svn.openmoko.org) Sep 12 14:21:26 svn: PROPFIND request failed on '/!svn/bln/2955' Sep 12 14:21:26 svn: PROPFIND of '/!svn/bln/2955': 400 Bad Request (http://svn.openmoko.org) Sep 12 14:21:46 mickeyl: what's up with moko svn? every 2nd parse end up like this Sep 12 14:21:54 ok, that wouldn't be a problem - for now, but I will need something that can "phone home" or receive a message with updated data, that has gps for position and direction, and that has sound output, and the neo seemed perfect for this Sep 12 14:22:04 psokolovsky__: no idea Sep 12 14:22:06 odd Sep 12 14:24:16 mickeyl: and what do you get with "wget 'http://svn.openmoko.org/!svn/bln/295'"? I get pretty much what it says - null response. you have capacity issues with svn? Sep 12 14:26:05 wget can't connect here Sep 12 14:26:11 svn ls works pretty fine Sep 12 14:29:07 can anyone tell me how to enable an early login using angstrom distro ? I would like to have a window up as early as possible. Right now, it is taking over a minute to boot, if not 2. Sep 12 14:29:29 koen: so, what about that sane-srcrevs.inc? I don't see it referenced anywhere. Is user supposed to include it himself? Sep 12 14:30:16 christopher: someone should start up dropbear as soon as possible, yep Sep 12 14:32:03 psokolovsky__: I'm getting svn: REPORT request failed on '/!svn/vcc/default' when I try to update my openmoko svn checkout Sep 12 14:32:17 psokolovsky__: but svn ls works fine??? Sep 12 14:32:29 cbrake: yep, me too. random errors like that. Sep 12 14:32:41 psokolovsky__: I'm going to try a clean checkout ... Sep 12 14:32:42 well, there's big difference between ls and checkout ;-) Sep 12 14:32:50 /wi 59 Sep 12 14:32:53 erm Sep 12 14:33:50 koen, the did you remove the size option from the udev mounting Sep 12 14:33:53 psokolovsky__: checkout seems to be going ... Sep 12 14:35:57 cbrake: hm, nice. can try immediate update too? Sep 12 14:41:27 psokolovsky__: Fetching external item into 'openmoko2/trunk/src/target/OM-2007.2/applications/openmoko-today2/libkoto' Sep 12 14:41:30 svn: REPORT request failed on '/repos/tasks/!svn/vcc/default' Sep 12 14:41:33 svn: REPORT of '/repos/tasks/!svn/vcc/default': 200 OK (http://svn.o-hand.com) Sep 12 14:41:36 psokolovsky__: checkout did not finish Sep 12 14:43:37 woops Sep 12 14:43:40 nonsense Sep 12 14:44:02 psokolovsky__: http://pastebin.ca/693947 what is up with SVN??? Sep 12 14:44:48 cbrake: I have only the same exclamation %) Sep 12 14:47:51 It's a feature. Sep 12 14:51:41 christopher: I've tried building from zero, and nothing... what distro are you using? I'm running kubuntu gutsy Sep 12 14:53:03 bernardo: ok, I am using angstrom Sep 12 14:53:49 is ffmpeg broken for someone else ? Sep 12 14:54:21 psokolovsky_: is there a way to get the graphic things like the bootsplash or other things first vs starting up daemons etc? just wondering what the cleanest way to prioritize is Sep 12 14:56:22 christopher: standard rcN.d prioritization Sep 12 14:57:06 christopher: use psplash? Sep 12 14:58:12 hrw: I haven't used that. I think I had gpe-bootsplash in there. Sep 12 14:58:16 or something like it. Sep 12 14:58:20 psokolovsky__: my svn problems seem to be related to a http proxy I was using -- svn seems to be working fine now Sep 12 14:58:22 steliosk, ping Sep 12 14:58:27 then use psplash ;) Sep 12 14:58:58 psokolovsky__: ok. I'll take a look and see what they're all doing. I'd just like something to come up as quickly as is possible while it is booting. Sep 12 14:59:08 seems to take such a long time on our little ARM9 Sep 12 14:59:42 christopher: which arm9? Sep 12 14:59:45 christopher: I meant your host distro. I'm also building for angstrom Sep 12 14:59:48 at91rm9200 Sep 12 14:59:57 bernardo: sorry, it is sarge-at91 Sep 12 15:00:03 ah.. right.. you based on sarge-at91 Sep 12 15:00:09 hrw: correct Sep 12 15:00:17 hrw: figured you'd remember after I said it Sep 12 15:00:18 :) Sep 12 15:00:24 christopher: do you have schematics available somewhere? Sep 12 15:00:25 christopher: well, doing that is easy. more complex is loby for some reasonable changes to go into some distro. I'd wish someone bother with dropbear startup order in angstrom... Sep 12 15:00:41 psokolovsky__: what is wrong with dropbear? Sep 12 15:01:03 christopher: udev and portmap seem to take a long time to start Sep 12 15:01:03 hrw: it would be nice to start up it as soon as possible Sep 12 15:01:18 hrw: no, I don't. it's very loosely based on the ek board. fairly simple design Sep 12 15:01:24 psokolovsky__: I do not agree Sep 12 15:01:29 cbrake: yeah, but no way around that, right? Sep 12 15:01:40 hrw: why? Sep 12 15:01:54 psokolovsky__: UI as soon as possible is proper - so I would change order of x11 startup rather Sep 12 15:02:27 christopher: I have not worked through it myself, but there is mdev in busybox. Sep 12 15:02:29 hrw: x11 *cannot* start soon, it's X after all ;-F Sep 12 15:02:50 psokolovsky__: X11 needs udev, keymap, dbus - nothing more Sep 12 15:02:56 christopher: udev is composed of shell scripts, which may be one reason it is so slow Sep 12 15:03:11 cbrake: mdev is fscking fast Sep 12 15:03:32 cbrake: ok. yeah, I know buildroot used static devs. not sure of advantages etc. Sep 12 15:03:42 hrw: is that something that angstrom can do? Sep 12 15:03:45 hrw: I'd assume that X is complete user env, and thus needs *everything*. while dropbear needs at most udev from that list. Sep 12 15:03:45 hrw: is it a lot of work to switch from udev to mdev in an OE build? Sep 12 15:04:28 christopher: OE comes by default with static /dev. unless you bother to run udev, that's what you get. Sep 12 15:04:46 cbrake: defconfig for busybox, one var used in task-base and some work later to have hotplug working properly Sep 12 15:05:08 psokolovsky__: under 2.6 /dev/ nodes id are dynamic Sep 12 15:05:29 christopher: everything is a tradeoff. Static devs are a royal pain ... Sep 12 15:05:37 hrw: ??? Sep 12 15:05:55 cbrake: yeah, figured as much. I figured there was a reason. but if mdev is so much quicker, why use udev? Sep 12 15:05:57 psokolovsky__: I mean: not all /dev/ entries are static Sep 12 15:06:25 christopher: mdev is busybox so you have to test which ver is working properly for you Sep 12 15:06:34 hrw: well, noone says that using static /dev is the way to go ;-) Sep 12 15:08:41 hrw: btw, any idea when we switch to automake 1.10? Sep 12 15:09:19 psokolovsky__: there were few recipes which fails with it Sep 12 15:09:50 hrw: well, at least angstrom-x11-image builds w/o probs for me Sep 12 15:14:16 like it should Sep 12 15:44:35 Crofton : hi Sep 12 15:57:18 mickey|bbiab: thanks for applying that patch Sep 12 16:01:26 * chouimat kicks pata drivers Sep 12 16:08:45 steliosk, when is the PPC confreence> Sep 12 16:09:54 Crofton : 24 and 25 Sep 12 16:11:04 Crofton : http://www.power.org/devcon/07/ Sep 12 16:11:30 bye Sep 12 16:11:33 ready for your talk? Sep 12 16:12:37 apparently, I may have an excuse to add MACHINE = "ps3" Sep 12 16:17:36 grrr, how do I get the davinic evm build to stop checking u-boot git ... Sep 12 16:19:40 Crofton|work: remove uboot from machine.conf and the kernel file Sep 12 16:20:16 there was some uboot version stuff in the machine file Sep 12 16:20:33 and I replaced uboot with uboot-utils in linnux.omap1.inc Sep 12 16:20:39 trying again Sep 12 16:21:38 03koen 07org.oe.dev * r77890e02... 10/ (1 packages/gsm/files/default packages/gsm/libgsmd_svn.bb): libgsmd: use mux0 on ezx phones Sep 12 16:21:47 I need to start getting caught up on gumstix also Sep 12 16:22:15 I suspect we should think about making all kernel recipes use uboot-utils for mkimage Sep 12 16:22:23 and only build u-boot for he board on demand? Sep 12 16:22:32 hmm Sep 12 16:22:47 we still need to setup the variable needed to buid lu-boot though Sep 12 16:24:16 Crofton : yeah, draft presentation passed with no comments. will submit final tomorrow Sep 12 16:24:58 koen, my issue on this machine is the I did not have git installed Sep 12 16:25:08 I thought OE installed git now ... Sep 12 16:25:40 git-native is added to DEPENDS when fetching from git Sep 12 16:25:48 hmmm Sep 12 16:26:05 only when using the OE git fetcher Sep 12 16:26:05 this is during parsing ..... Sep 12 16:26:47 morning Sep 12 16:26:47 http://rafb.net/p/PwaJ4376.html Sep 12 16:26:51 gm Sep 12 16:26:55 Crofton|work : check the xilinx kernels as i had to add some python magic to get git-native installed as i am not using the OE fetcher Sep 12 16:27:06 the python on openmoko project is pathetic Sep 12 16:27:23 half their roadmap is *ALREADY IMPLEMENTED* in OE Sep 12 16:27:32 koen, I had to install git Sep 12 16:27:39 now parsing succeeds Sep 12 16:27:40 and zecke has done the biggest part of the other half Sep 12 16:47:36 Crofton: after switching to binutils 2.17 I could build a kernel image for DaVinci with a sane size, but it still will not boot Sep 12 16:47:59 how does it fail? Sep 12 16:48:13 koen, that is because zecke rocks Sep 12 16:48:27 Uncompressing Linux.. (about 1.5 terminal lines of dots).. done, booting the kernel. Sep 12 16:48:30 and that's it Sep 12 16:48:31 so really strange Sep 12 16:48:46 have you enabled low level debug? Sep 12 16:48:55 maybe it is a board number mismatch? Sep 12 16:49:08 hmm no, I do not think that I enabled low level debug.. Sep 12 16:49:18 board number mismatch? where is that set? Sep 12 16:49:43 Crofton, Jin^eLD: http://svn.o-hand.com/view/poky/trunk/meta/packages/linux/linux-rp-2.6.22%2B2.6.23-rc4/binutils-buildid-arm.patch?rev=2726&view=markup Sep 12 16:50:29 koen: ok, that explains the 2.18 problem Sep 12 16:51:39 mickey|bbiab: who do I need to bribe to add mokoscrolledwindow support to evince: http://scap.linuxtogo.org/files/ef0066dc374ba9bf84e797e84a83b578.png ? ;) Sep 12 16:51:44 but well, I worked around that by picking a different binutils version Sep 12 16:53:34 Crofton: so what number where you talking about? Sep 12 16:53:49 there is some board ID number somwhere Sep 12 16:53:55 I do not know where it is Sep 12 16:54:09 I think we should make sure lowlevel debug is enabled first Sep 12 16:54:41 Crofton: which one is it, CONFIG_DEBUG_KERNEL ? Sep 12 16:55:11 not ure, read the helps Sep 12 16:55:26 I don't have a copy of the source unpacked atm Sep 12 16:55:38 Crofton: ok.. but what you mean is in defconfig, or did you mean something board specific? Sep 12 16:55:50 in the defconfig under kernel hacking I think Sep 12 16:56:09 ah ok, then I was at the right place already Sep 12 16:57:05 basically, I think it sets up the serial port to work with printks earlier in the boot process Sep 12 16:57:41 try the printascii() hack Sep 12 16:58:26 koen? Sep 12 16:59:04 "Use the debugging printascii() stuff for your platform in arch/arm/kernel/debug-armv.S (enabled by CONFIG_DEBUG_LL) and drop a printascii() into printk() (kernel/printk.c) just after the vsprintf() call, and monitor the relevant serial port." Sep 12 16:59:43 if you insert aht printascii(); in that file the kernel will print out messages on the serial port for the early boot process Sep 12 17:00:23 you get duplicated messages if the boot succeeds, but if it boots you can remove the printascii Sep 12 17:00:40 oh, ok, that sounds cool Sep 12 17:00:52 I'll try that Sep 12 17:05:37 mallum: ping Sep 12 17:06:45 mallum: http://scap.linuxtogo.org/files/81a63f34a2991acfc649521d17bfd9a5.png Sep 12 17:08:28 there must be a bug or race condition in MBPanelScaledImage Sep 12 17:08:55 only the panel icons that are changed after initial showing appear scaled Sep 12 17:11:19 atk does not build on my amd64 machine :-( Sep 12 17:11:32 mickeyl: hey Sep 12 17:11:37 hi mallum Sep 12 17:11:55 mallum: i have a problem with MBPanelScaledImage. The one who did it is no longer with OH, right? Sep 12 17:12:06 Crofotn: you were right! Error: unrecognized/unsupported machine ID (r1 = 0x00000356). Sep 12 17:12:09 mallum: no jorn did it, he is still with us :) Sep 12 17:12:17 talking to myself again Sep 12 17:12:19 it wants 00000385 Sep 12 17:12:32 mickeyl: best bet is shoot a mail to matchbox ml ? Sep 12 17:12:59 and reparsing takes forever when I make debug changes to *.class files :-( Guess its time to figure out how to to make bitbake not spend gobs of time on the internet during parsing ... Sep 12 17:13:17 or is there a better way to debug class files? Sep 12 17:13:19 mallum: ok, right. will do Sep 12 17:14:54 cbrake: http://lists.linuxtogo.org/pipermail/openembedded-devel/2007-September/002963.html Sep 12 17:16:45 Crofton: from what I understand the id of the kernel does not match the id that is set in uboot? Sep 12 17:17:11 or did I get that wrong? Sep 12 17:17:34 koen: sounds like a plan Sep 12 17:17:39 that is one possibility Sep 12 17:17:46 I am making wild guesses .... Sep 12 17:19:10 hy all Sep 12 17:19:12 mhm Sep 12 17:19:28 I'll ask in #davinci, lets see if anyone had the same problem Sep 12 17:21:37 heh, three Crofton's in #davinci :) Sep 12 17:24:25 tofl Sep 12 17:25:04 considering the low numbers there, probably noone else alive besides you :) Sep 12 17:26:22 moin Sep 12 17:26:46 I got a new toy (Nokia N800) Sep 12 17:28:40 KhemWork: fine. subcribe here https://bugs.maemo.org/show_bug.cgi?id=1584 ;) Sep 12 17:31:04 so I thought your guess was the same as mine or something, 'bout uboot Sep 12 17:31:08 oops wrong channel Sep 12 17:31:08 :) Sep 12 17:37:06 rschuster: is there any doc on how to reflash stuff on it Sep 12 17:45:19 KhemWork: sure they have a wiki at maemo.org Sep 12 17:45:37 KhemWork: if you prefer a free flasher tool. try this: http://nopcode.org/0xFFFF/ Sep 12 17:45:37 rschuster: how about OE ? Sep 12 17:45:51 OE builds 0xffff by default :) Sep 12 17:45:53 KhemWork: that related to the bug report Sep 12 17:47:00 KhemWork: there is no spec for some of the crucial part of the devices (powermanagement) and no source for their low-level stuff (e.g. bad-block management, bootloader and such) Sep 12 17:47:16 hmmm Sep 12 17:48:24 koen: is maemo also supported in OE then ? Sep 12 17:48:35 KhemWork: but the poky guys have actually built a working image Sep 12 17:49:17 KhemWork: maemo wasn't 100% free until a short time Sep 12 17:50:50 KhemWork: and contained lots of modifications to gtk. AFAIK this has changed Sep 12 17:51:38 KhemWork: all of their stuff is autotooled (AFAIK). it would not be hard to get it in. just needs volunteers :) Sep 12 17:52:02 rschuster: you mean poky's Sep 12 17:52:32 KhemWork: ? Sep 12 17:53:31 rschuster: I will be interested in trying poky Sep 12 17:55:16 rschuster: maemo depends on a forked gtk+ Sep 12 17:55:28 koen: is this still the case? Sep 12 17:55:40 rschuster: yes Sep 12 17:55:48 rschuster: e.g. hildon-fm needs it Sep 12 17:55:51 * chouimat is reading qnx source code Sep 12 17:55:59 and almost everything depends on hildon-fm Sep 12 17:56:06 hi koen Sep 12 17:56:12 rschuster: see packages/maemo3 Sep 12 17:56:15 hey chouimat Sep 12 17:56:31 koen: ah yes. maemo 3. ok Sep 12 17:57:02 rschuster: if someone fixes hildon-fm we can build maemo-mapper in OE :) Sep 12 17:57:11 wow this is neat Sep 12 17:57:23 ERROR: No providers of runtime build target ${IMAGE_INSTALL} (for ['/home/balister/oe/ossie_collection/ossie-image.bb']) Sep 12 17:57:30 anyone have a quick naswer? Sep 12 17:57:36 koen: they are working on a 4.0 release which takes gtk2.10 even (there where using a 2.6 derivative) Sep 12 17:57:48 rschuster: a forked gtk 2.10 Sep 12 17:57:54 Crofton|work: the kernel image from your site did not boot either Sep 12 17:58:01 rschuster: you need to read every nokia pressrelease very carefully :( Sep 12 17:58:01 bugger Sep 12 17:58:09 Crofton|work: maybe I really need to update uboot Sep 12 17:58:22 I guess someone nees to ask on the linux-davinci list ... Sep 12 17:58:35 I guess that someone will be me heh Sep 12 17:58:50 do the ppl there sue OE ? Sep 12 17:58:53 aem, use Sep 12 17:58:59 koen: this is really bad Sep 12 17:59:15 Jin^eLD: sue Outlook Express. no problem :) Sep 12 17:59:24 :)) Sep 12 17:59:48 rschuster: every speech ari jaaki has given the past year mentions how every piece of software not under company control is crap, doomed to fail, kills kittens etc Sep 12 18:00:13 rschuster: so I'm not very hopefull quim and carlos will success when ari insists on being such an asshat Sep 12 18:00:16 ya, that's a bit unfortunate and shows a certain lack of professional distance Sep 12 18:00:37 s/success/succeed/ Sep 12 18:00:54 mickeyl: funny how they mention openmoko in that category as well Sep 12 18:00:55 mickeyl: and some slight misunderstanding on how things are supposed to work nicely in the community ... Sep 12 18:01:00 has something changed in image generation lately? Sep 12 18:01:05 guess FIC is a too small company to count as a company ;) Sep 12 18:01:05 03mickeyl 07org.oe.dev * re5a1ea39... 10/ (3 files in 2 dirs): xchat: remove 2.6.1, add 2.8.4 Sep 12 18:01:10 03mickeyl 07org.oe.dev * rb2ce7bfc... 10/ (1 packages/psplash/files/openmoko/psplash-hand-img.h): psplash: fix buglet in openmoko psplash-hand-img.h Sep 12 18:01:16 03koen 07org.oe.dev * r0f79b6f7... 10/ (1 packages/e17/e-wm_0.16.999.041.bb): e-wm: add version that uses e-dbus instead of the deprecated ecore-dbus Sep 12 18:01:34 Crofton|home: yes, rootfs*bbclass has been changed substantially Sep 12 18:01:36 mickeyl: not that everything is bright and shiny there but better than dozens of lib*-osso ;) Sep 12 18:01:40 RP: can you please comment on my mail about typical kernel module build issue - it was cc: to oe list, though thread runs in angstrom-devel Sep 12 18:02:18 rschuster: agreed Sep 12 18:02:29 this is why I want an easy way to extend an image rather than make my own Sep 12 18:02:56 koen: did you see my q about sane-srcrevs? Sep 12 18:03:22 koen: so, what about that sane-srcrevs.inc? I don't see it referenced anywhere. Is user supposed to include it himself? Sep 12 18:04:01 and btw even manual including doesn't have with broken parses due to svn flipping Sep 12 18:04:16 psokolovsky__: see the thread on oe-devel Sep 12 18:04:25 I am not your personal email reader Sep 12 18:06:30 koen: last thing I see in ML is "and add the output to a conf file somewhere." now there's file with those revs in OE, but: 1) it's not actually used; 2) it doesn't work Sep 12 18:06:50 koen: I specifically interested in these 2 points, and they were not discussed in ML Sep 12 18:13:53 !oebug 2978 Sep 12 18:23:44 anyone got an idea why some of the packages fail ? http://rafb.net/p/njGGek57.html Sep 12 18:26:39 RP: Btw, fresh idea: it could be that bitbake SRCREV additions cause DDoS on OpenMoko servers ;-D Sep 12 18:27:58 steliosk: it tells clearly that you have broken gzip archives Sep 12 18:29:38 psokolovsky__ : with older ipkg versions works without problem Sep 12 18:30:03 psokolovsky__ : any idea if ipkg format changed ? Sep 12 18:30:28 it hardly has anything to do with ipkg format Sep 12 18:30:39 then, test for "gzip magic" is broken ;-) Sep 12 18:30:41 steliosk: apache mime problem Sep 12 18:31:25 koen : could be. the repos are on different servers Sep 12 18:43:31 QNX goes opensource http://www.qnx.com/news/pr_2471_1.html Sep 12 18:44:11 RP: another q: are you sure using "Last Changed Rev" is unambiguous enough? Sep 12 19:02:33 mickeyl: Well, it seems following happens: on the first svn access to a new repo, it hangs, and eventually times out with 400, next attempt goes on well. Now tell me, if its your network, or mine? ;-) Sep 12 19:17:17 RP: ping Sep 12 19:39:39 any clues on how to add a kernel module ? Sep 12 20:27:30 re Sep 12 20:27:38 hi flo_lap Sep 12 20:28:07 polyonymous or bit of you ... how's OE rebuild going? ;) Sep 12 20:32:16 any oe hacker ? how to add kernel to rootfs Sep 12 20:44:35 interfaith, isnt it added by default ? ;) Sep 12 20:45:04 ok kernel modules Sep 12 20:45:22 to add a kernel module Sep 12 20:45:30 then rebuild rootfs Sep 12 20:45:50 searching the build code for a clue Sep 12 20:46:16 just change your kernel config in OE I'd say Sep 12 20:46:43 and do some tweaks in machine config file to autoload the module Sep 12 20:46:44 that should do Sep 12 20:46:45 yes i have tried many commands like that Sep 12 20:47:01 mods to machine def or distro def.. no luck Sep 12 20:47:19 MACHINE_EXTRA _RECOMMENDS etc.. etc.. nothing doing Sep 12 20:47:54 org.oe.dev/packages/linux/somewhere/there/defconfig Sep 12 20:47:56 now i will just hack default config to build into the kernel Sep 12 20:48:24 yes thats one way around the problem Sep 12 20:49:34 interfaith: is your problem that they aren't being built or that they aren't being added to the rootfs image? Sep 12 20:49:36 'kernel-module' shows up in many code blocks.. the trick to change it baffles me Sep 12 20:49:46 not added to rootfs Sep 12 20:49:55 they are in 'work' Sep 12 20:50:09 what distro/image are your building? Sep 12 20:50:15 ixp4xxbe Sep 12 20:50:34 need to add ixp4xx-mac.ko and more Sep 12 20:51:49 several kernel network modules needed qmgr mac not in rootfs Sep 12 20:52:17 interfaith: ixp4xxbe is the machine, what is the distro you have in your locol.conf, and what image are you building with your bitbake command? Sep 12 20:52:29 distro angstrom 2001 Sep 12 20:52:53 bitbake bootstrap-image Sep 12 20:53:11 should be angstromxxx ? Sep 12 20:54:06 should be: DISTRO = "angstrom-2007.1" Sep 12 20:54:17 ~seen zecke Sep 12 20:54:18 yes Sep 12 20:54:19 zecke was last seen on IRC in channel #openmoko, 9h 49m 17s ago, saying: 'the places for the disabled most of the time have poweroutlets...'. Sep 12 20:54:25 angstrom-2007 Sep 12 20:55:46 some angstrom image ? Sep 12 20:56:06 angstrom-minimal ? Sep 12 20:56:35 I would use the distro I mentioned above and would use angstrom-minimal-image as the image Sep 12 20:57:12 so the MACHINE_EXTRA will add in the kernel modules ? perhaps Sep 12 20:57:33 yes add MACHINE_ESSENTIAL_EXTRA_RRECOMMENDS += "kernel-modules" to your machine def Sep 12 20:57:55 and it will add all of the modules that are built as part of your kernel config to the rootfs image Sep 12 20:58:05 alright ! sweet Sep 12 20:58:15 or you could enumerate them individually if you like Sep 12 20:58:30 yes .. i've been scouting such code Sep 12 20:58:55 good luck! Sep 12 20:58:59 k Sep 12 20:59:09 thx ! Sep 12 20:59:25 psokolovsky: pong Sep 12 20:59:43 psokolovsky: I saw the mail earlier and have been meaning to reply Sep 12 20:59:54 RP: Hi! Saw above whining about svn flipping? Sep 12 21:00:02 RP: ok, thanks Sep 12 21:00:30 psokolovsky: I saw a question about '"Last Changed Rev" is unambiguous enough' and didn't understand it. I guess I need to read context Sep 12 21:01:38 RP: well, nevermind, it's apparently the best choice which can be made. The matter was that "Last Changed Rev" != repo rev Sep 12 21:02:18 RP: the matter is however that SVN, in particular OM, is pretty flippy, and fails regularly on svn info Sep 12 21:02:44 psokolovsky: Broken proxy? Sep 12 21:03:03 RP: I again apologize for being out of loop, but can you just quickly tell if OM was the main customer behind SRCREV (specificaly "now") revamp? Sep 12 21:03:49 RP: yes, on whose side? I never saw something like that on my side (except for cases which broken proxy is known to be on other side of course) Sep 12 21:04:50 psokolovsky: I did have OM and OH in mind when I wrote srcrev support. Nobody was a customer as such, it was done in my free time Sep 12 21:05:21 psokolovsky: I had a really strong personal dislike of the old "now" support since it was totally broken Sep 12 21:05:58 RP: yep I mean "customer" not literally. someone just wanted that. someone who. ok, I see. Sep 12 21:06:00 psokolovsky: I've never seen this "flipping". It could be anywhere inbetween the servers I guess Sep 12 21:06:50 RP: the problem, new "now" support is not ideal too. some may say it's even regression... ;-) Sep 12 21:07:10 RP: well, cbrake confirmed that he sees something similar Sep 12 21:07:59 RP: well, if you mention OH as party behind, I won't ask next question, even though it would be pretty helpful to understand what would be the right fix to this issue ;-) Sep 12 21:08:11 RP: well, last q, on another matter: Sep 12 21:08:31 psokolovsky: The old "now" support wasn't usable. It is as simple as that ;-) Sep 12 21:08:56 RP: suppose there's .bb's: foo1 & foo2, both PROVIDEing foo. But foo1 has DEFAULT_PREFERENCE=-1. What should bitbake do if PREF_PROV is not defined? Sep 12 21:09:10 psokolovsky: The new code perhaps isn't ideal either, I don't deny that. All I ask is the problems are clearly explained so we can try and address them Sep 12 21:09:45 psokolovsky: Probably pick foo2 Sep 12 21:10:21 RP: Folks already said, and I personally can second that: it's bad (bad for those people, you see) to make initial parsing depend on network connection. Sep 12 21:10:50 psokolovsky: What was the question you're not asking? Sep 12 21:10:59 RP: yeah, I would think too, that foo2 would-should be picked, but foo1 is still being picked Sep 12 21:11:19 psokolovsky: That's with trunk (1.8.9) ? Sep 12 21:12:26 psokolovsky: I will say this again although I'm getting sick of saying it. My personal opinion is that every svn or other srcrev revision should be locked down in OE by default. If you do that the network is not accessed during parsing. Sep 12 21:12:46 RP: The q was: is necessity of all this "now"/"latest SRCREV" matter is due to lack/non-filledness of build engineer role ;-). So lazy programmers don't want to handle the build stuff right way, so go out of their way to request crazy features being added to build system to support their laziness ;-) Sep 12 21:13:26 RP: With 1.8 branch's HEAD, yes. Specifically, it happens for qte-mt-static (DEF_PREF=-1) vs qte-mt Sep 12 21:13:52 RP: Are you sure it is *not* accessed then? Sep 12 21:14:13 psokolovsky: If its locked down it certainly shouldn't. If it did, that would be a bug that needs fixing Sep 12 21:14:41 RP: and can you have a look at current sane-srcrev.inc in OE and tell if that's proper way of locking? Because with it, I *still* get errors due to SVN flipping. Sep 12 21:14:48 psokolovsky: Nobody has shown me such a bug and as far as I know it doesn't. I have tested that at various points Sep 12 21:16:26 psokolovsky: For packages listed in the .inc file (which you're sure is being included)? Sep 12 21:17:17 RP: I asked koen why it is not included by default. But I added "require conf/distro/include/sane-srcdates.inc" to local.conf Sep 12 21:17:56 psokolovsky: Regarding lazy developers, OE's role is one of software development. Being able to have OE build the latest and greatest revision of a package is certainly a desireable feature. the old "now" support simply didn't work requiring lots of manual intervention. I don't think its a case of being lazy and is more about optimised workflow Sep 12 21:18:49 psokolovsky: Can you reproduce this problem on a specific package atm? Sep 12 21:19:07 RP: Then I again only second opinion already sounded that it would rather be special feature requiring enable, then default, hitting all users ;-) Sep 12 21:19:42 psokolovsky: I just implemented the feature, I was not the one who made it hit all users Sep 12 21:20:18 RP: The problem is fliping SVN. it's not 100% reproducible. But it happened in 70% accessing some OM repos. Sep 12 21:21:07 psokolovsky: Tell you way, try adding comments to the network hitting function calls in bitbake's wvn fetcher, see if it really is hitting the network? Sep 12 21:21:14 s/way/what/ Sep 12 21:21:27 RP: I added retry code and found that hitting it second time always resolves issue. So apparently, it's weird network/server issue. And bitbake is not immune to it. Sep 12 21:21:29 and I mean print statements, not comments :/ Sep 12 21:22:09 psokolovsky: I suspect there is a dodgy proxy/load balancer/whatever in front of the omoko server Sep 12 21:22:32 psokolovsky: That is simply outside bitbake's control Sep 12 21:22:48 psokolovsky: It shouldn't be hitting the network in the first place and we should find out why Sep 12 21:23:08 RP: Well, what I really wanted is discuss issue with you once again, to at least understand how all that should work. Sep 12 21:23:27 RP: So, it did hit network. Sep 12 21:25:11 RP: If you would work on that, and want me to reproduce something, I can do it. Sep 12 21:25:37 RP: Because as you might guess, I have "fixed" it in "right way" already ;-) Sep 12 21:26:58 psokolovsky: All I want is a reproducable demonstration of the problem. Prove its in bitbake and I'm interested. If its an OE misconfiguration, all I can do is jump up and down on the OE list though :/ Sep 12 21:28:13 RP: how should I run bitbake? -DD? Sep 12 21:28:17 psokolovsky: The problem is I don't really have enough info to reproduce this quickly. Which package was it, what revision was it trying to pull, what was SRCREV set to etc.? Sep 12 21:28:39 RP: I can provide bitbake output Sep 12 21:29:14 RP: [pfalcon@blackbox bitbake-1.8]$ svn up Sep 12 21:29:14 At revision 974. Sep 12 21:29:36 psokolovsky: Add a 'bb.msg.plain("I hit the network")' to _latest_revision in lib/bb/fetch/svn.py Sep 12 21:30:03 psokolovsky: Then run bitbake and see if you see that message. If so, tell me which packages it shows for... Sep 12 21:30:05 RP: there's already such stuff Sep 12 21:30:27 psokolovsky: ok, as you said, bitbake -DD should show it Sep 12 21:30:40 psokolovsky: I'd forgotten I'd left that there :/ Sep 12 21:34:56 RP: Well-well, my great apologies, that was really stupid %). A typo in require directive. Damn! I'm sorry for wasting your time! Sep 12 21:35:39 and another good reason to have it in by default... Sep 12 21:37:05 psokolovsky: It happens :) Sep 12 21:37:13 RP: it still goes for OH tasks, dates, worth adding... Sep 12 21:37:28 RP: yeah, it's all gets more and more complex ;-I Sep 12 21:37:46 psokolovsky: Right, they're missing from the lockdown file which is more of a policy question :/ Sep 12 21:38:36 psokolovsky: Hopefully people won't need to worrk about this complexity once things settle down, the boat is just rocking a bit atm Sep 12 21:39:31 RP: yeah. At least I'll post about current state to angstrom-devel Sep 12 21:41:39 psokolovsky: I'm tempted to make the default SRCREV error out, that should get people filling in sane-srcrevs ;-) Sep 12 21:42:39 RP: then how to use latest? old good "now" value for it? Sep 12 21:43:18 psokolovsky: set SRCREV = "${AUTOREV}" in local.conf Sep 12 21:44:08 RP: ok. probably good plan in the long run ;-) Sep 12 21:44:15 in fact I might check in SRCREV = "1" as the default :) Sep 12 21:46:42 RP: well, that's too harsh ;-). error out is harsh too, unless devels really bother to add entries to file... Sep 12 21:47:27 * rwhitby scans the backlog and likes the idea of OE srcrevs being locked down by default and having to do something in local.conf to use the untested bleeding edge Sep 12 21:47:58 morning Sep 12 21:48:05 rwhitby: you missed the message on the mailinglist about that? Sep 12 21:48:09 * mwester-treo agrees Sep 12 21:48:58 koen: Whats your opinion of making SRCREV = "1" by default in bitbake.conf and then letting distros decide if they want to play games with network access? Sep 12 21:49:20 RP: I'm all for it Sep 12 21:49:44 RP: I would have made sane-srcrevs default in angstrom, but I didn't in fear of OM people freaking out Sep 12 21:49:47 mickeyl: around? Sep 12 21:50:01 RP: Can you also have a look here: http://blackbox/busyb/oe/build_20070912_234551/Step1-1.html Sep 12 21:50:15 (and was too lazy to make an AUTOREV files for OM stuff) Sep 12 21:50:24 koen: the OM users definitely wouldn't freak out Sep 12 21:50:30 psokolovsky: no since that DNS won't work for me ;-) Sep 12 21:50:42 rwhitby: The OM devs would ;-) Sep 12 21:50:54 RP: it seems strange that it prints header many times. I don't remember such behavior during console builds... Sep 12 21:50:56 RP: yeah, well they can use AUTOREV, right? Sep 12 21:51:05 RP: Sorry, http://linux-h4000.sourceforge.net/busyb/oe/build_20070912_234551/Step1-1.html Sep 12 21:51:42 rwhitby, koen: We need a copy of sane-srcrevs.inc (insane-srcrevs.inc) which is a copy replacing the revisions with AUTOREV Sep 12 21:51:55 * rwhitby thinks it would be good to OM devs to have sane srcrevs for 99% of the packages so they can concentrate on fixing the 1% which are critical to making a reliable phone call with the Neo Sep 12 21:53:11 naive thought :) Sep 12 21:53:32 psokolovsky: I've seen it before and it is strange. I've never worked out why. I think its something to do with the threading Sep 12 21:54:08 RP: yes, so I thought, but that log for BB_THREADS=1... Sep 12 21:54:22 well ok, not critical... Sep 12 21:56:21 psokolovsky: How were you running bitbake? Sep 12 21:56:39 RP: and just in case, here's the log for that that RPROVIDES issue: http://linux-h4000.sourceforge.net/busyb/oe/build_20070912_234551/Step4-1.html Sep 12 21:56:43 psokolovsky: Even with threads=1 it actually has two, the control thread and the worker ;-) Sep 12 21:56:52 RP: well, that log has the command ;-) Sep 12 21:57:24 psokolovsky: right, its late ;-) Sep 12 21:57:26 RP: it's all under busyb teh light CI system Sep 12 21:57:55 arg, no I have a colleague who has decided to rewrite some code in C (from C++) because it is faster or something Sep 12 21:58:11 RP: so, it's python doing os.system() of what it is logged there - output gotten using shell redirect Sep 12 21:58:33 RP: have a sane-srcrevs.inc, which is default in bitbake.conf, an extra insane-srcrevs that sets everything to AUTOREV, and an moko-autorev.inc for you-know-who Sep 12 21:59:14 koen: sane-srcrevs.inc should be a distro include really, but yes Sep 12 22:00:11 03koen 07org.oe.dev * r763b1ed6... 10/ (1 packages/gpicview packages/gpicview/gpicview_0.1.5.bb): gpicview: add 0.1.5 Sep 12 22:00:13 psokolovsky: I don't know why its doing that (multiple copies of the output), I think you'd have to debug it Sep 12 22:00:14 RP: I have no strong opinions on that Sep 12 22:00:27 psokolovsky: It would be interesting to know what trunk did as that handles output a lot differently Sep 12 22:00:45 koen: I will propose changing things like that on the OE list Sep 12 22:00:54 RP: ok, as I told, it's not critical. when I have time and mood, I'd go for it ;-) Sep 12 22:01:15 psokolovsky: It could well be fixed in trunk and if thats the case I'd just ignore it Sep 12 22:01:25 ok, will try Sep 12 22:01:39 psokolovsky: trunk is an attempt to clean up things I know 1.8 does badly in places ;-) Sep 12 22:02:06 Admittedly trunk has issues of its own atm of course and I'd not recommend for daily use quite yet :/ Sep 12 22:02:18 * koen still needs to try trunk Sep 12 22:02:55 koen: Its at the stage where we should start a "things to fix" list and help creating that list is welcome Sep 12 22:03:32 "* Pizza fetcher doesn't work" Sep 12 22:03:37 exit status and not catching all events that should trigger an exit are the two major ones I know of... Sep 12 22:03:38 Crofton: I've been out of town for a couple of days. Did you settle on the "proper: way to fix the modules.dep issue? Sep 12 22:04:06 sakoman: yes, bug in bitbake, hrw solved it Sep 12 22:04:45 excellent! I'll remove my hack. Hate to keep that stuff around for long -- ends up being junk dna Sep 12 22:07:37 koen: It was not a bug in bitbake ;-) Sep 12 22:14:43 RP: so where was the bug? Sep 12 22:14:56 sakoman: Just in the way the metadata was written in OE Sep 12 22:15:13 sakoman: People were expecting bitbake to handle things in a certain way and it didn't Sep 12 22:15:37 ah, well I'm glad it's fixed. That one was driving me crazy! Sep 12 22:16:00 Worked sometimes, didn't work other times Sep 12 22:16:42 sakoman: Have a look at bug #2966 if you're interested in the details Sep 12 22:17:09 OK, thanks Sep 12 22:21:32 ok, SRCREV mail sent Sep 12 22:21:35 * RP -> back shortly Sep 13 00:00:17 03rpurdie 07org.oe.dev * rbdcb6216... 10/ (5 files in 2 dirs): meta-sdk-*: Update/fix to use package_update_index_ipk Sep 13 01:24:34 03mickeyl 07org.oe.dev * rccee86b3... 10/ (3 files in 3 dirs): Sep 13 01:24:34 alsa-state: ship standard asoundrc (simple dmix configuration) that should Sep 13 01:24:34 be a reasonable default for most devices **** ENDING LOGGING AT Thu Sep 13 02:59:57 2007