**** BEGIN LOGGING AT Tue Aug 25 02:59:56 2009 Aug 25 03:22:18 khem : re http://bugs.openembedded.org/show_bug.cgi?id=5282 my hushmail account expired, i may redo my toolchain overnight Aug 25 03:23:13 i started fabricating the enclosure for the dht-walnut i have been working on, though. i got an older es1373 sound card working, as well as a promise tx4 sata card, so everything is going well. Aug 25 03:23:59 im waiting on a 2gb ide slc ssd dom (hah) that had to be shipped from korea -> montreal -> me Aug 25 03:25:58 m4t, do you have a build environment that works for the dht-walnut? I have two of those boards. I fiddles with them a bit, but did not get them to work. I am pretty sure I know why now, but I'd like to get them to work. Aug 25 03:26:16 yes i do Aug 25 03:26:42 i have a patched u-boot 1.2.0 package, and a heavily patched linux 2.6.24.7 kernel as well Aug 25 03:26:59 i also have a patch to enable the gpio pins Aug 25 03:27:02 If you get that where you can share the setup to build, I'd like to see it. Aug 25 03:27:47 yea ive got a patch, and i can tar up my configs Aug 25 03:27:59 everything was done with gcc 4.4.0 + eglibc Aug 25 03:29:06 What do you use as a build environment? I mean, do you build on a Linux distro? Aug 25 03:29:09 i also have a patch to qemu 0.9.1 that enables a dht-walnut target, helped me test firmware Aug 25 03:29:12 yes , linux Aug 25 03:29:25 What distro and version? Aug 25 03:29:37 ubuntu 9.04 Aug 25 03:30:57 So, you installed the gcc yourself. (.04 comes with gcc 4.3.3. (.10 comes with 4.4.1 Aug 25 03:31:14 oh, im referring to the cross compile toolchain Aug 25 03:31:35 the host cc is stock. Aug 25 03:31:37 Good point. I checked the wrong thing. Aug 25 03:35:55 m4t, what did the dom cost? I have not found anyone who sells slc doms. Aug 25 03:36:27 kind of pricy, like 50$ Aug 25 03:36:57 it had *just* been released, hence the need to have it sent direct from korea Aug 25 03:41:50 http://www.filefactory.com/file/ah30ga5/n/openembedded_walnut_full_patch Aug 25 03:42:06 ive made some changes since then Aug 25 03:42:35 most of the debian etchnhalf security patches are in there too Aug 25 03:43:20 hmm Aug 25 03:43:51 that will give you u-boot 1.2.0 for the walnut Aug 25 03:44:05 i changes a few of the puts() so it identifies the cpu as an ibm rather than an amcc Aug 25 03:44:20 just an up-port from the old walnut u-boot target Aug 25 03:45:08 that diff is from a git checkout at least 1.5mo old btw Aug 25 03:47:23 http://pastebin.com/d67b69b9a is openembedded/conf/machine/dht-walnut.conf Aug 25 03:48:39 http://pastebin.com/da35ea54 is openembedded/conf/distro/angstron-2008.1.conf Aug 25 03:49:32 i dont know if the 2.6.24.7 has pax enabled, if it does, it will break all dynamic binaries Aug 25 03:49:39 :/ Aug 25 03:52:28 Severian : i can also send you an initramfs and a uboot 1.2.0 image, if you just want something to play with for the time being Aug 25 03:54:12 I am interested. I don't know what to do with them, but I hope I have enough info to figure it out. Aug 25 03:54:55 try http://www.elinux.org/flameman/dht-walnut and elinux.org/dht-walnut Aug 25 03:55:05 very good starting point, there is a lot of info there Aug 25 03:55:18 I got the 2 pastebin files. I have not figured out how to get that patch yet. Aug 25 03:55:27 hrm Aug 25 03:55:45 yea i just googled around for a file dump, do you have any other sites where i can post it? ~1.5m Aug 25 03:56:36 I think it may require javascript, which I don't have. If you are ready to do it, I can setup an ftp server for a little while. Aug 25 03:56:59 i could do that, just msg me the info Aug 25 03:58:01 i got u-boot.diff as well, its a separate file actually Aug 25 04:36:55 m4t, what kind of case are you building? Do you have drawings or pictures yet? Aug 25 04:37:10 i'll show you what it's going into, one sec Aug 25 04:40:55 http://slacksite.com/images/StorEdge/multipack_SCSI.jpg http://www.computerliquidationcenter.com/store/catalog/1144_image3.jpg Aug 25 04:41:45 i am ~70% done, if you count software Aug 25 04:42:02 you can work the plastic in that enclosure like wood with a dremel :o Aug 25 04:43:32 That should end up looking very professional. I assume you doing this for yourself. Aug 25 04:43:45 yea, nas-type device Aug 25 04:44:28 with a bit of epoxy putty and a green paint marker, i should have a couple hot swap sata bays Aug 25 04:45:24 i think the walnut itself will be mounted on a piece of lexan, which will slide into a few mirror hangers i found at a hardware store, and match up edge-to-edge with the existing scsi backplane Aug 25 04:45:33 i havent figured out how to secure the pci card yet Aug 25 05:19:26 hmm Aug 25 05:22:02 kergoth: hey Aug 25 05:22:09 hey zecke_ Aug 25 05:27:42 opinion on making bb.event obey an 'events' var flag to control which events a given event handler is run for? might be beneficial, particularly given in trunk bb.msg fires events (ew) Aug 25 05:36:08 http://dl.getdropbox.com/u/112715/Files/inscm-0.tar.bz2 - heh, experimenting Aug 25 07:11:53 good morning Aug 25 07:12:32 good morning Aug 25 07:27:50 good morning Aug 25 07:28:12 how can I specify a directory where i have .ipk packages for opkg ? Aug 25 07:28:41 from command line Aug 25 08:21:01 good morning Aug 25 08:40:42 good morning Aug 25 08:52:01 morning Aug 25 08:57:21 hi hrw Aug 25 09:27:03 florian: good morning Aug 25 09:27:18 zecke_: good evening Aug 25 09:31:23 pb__: hey Aug 25 09:47:25 03Roger Monk  07org.openembedded.dev * rf1b3ee8407 10openembedded.git/recipes/gnome/gnome-games_2.24.0.bb: gnome-games: fix previous bad patch -I/SDL error Aug 25 10:02:12 argh... Aug 25 10:02:31 I installed debian on vortex86sx... Aug 25 10:04:12 lack of fpu kills.. Aug 25 10:09:10 mckoan: did you ever got Debian kernel booted on it? Aug 25 10:09:21 mckoan: 486 Debian kernel Aug 25 10:10:20 03Thomas Zimmermann  07shr/import * rf7a4e3cf80 10openembedded.git/recipes/opensync/ (libsyncml_0.5.4.bb wbxml2_0.10.7.bb): Updated libsyncml and wbxml2 reciepe to current versions Aug 25 10:10:31 03Sebastian Krzyszkowiak  07shr/import * r544e533407 10openembedded.git/conf/distro/include/preferred-shr-versions.inc: preferred-shr-versions: bump frozen-bubble from 1.0.0 to 2.2.0 Aug 25 10:17:08 hrw: I do actually sell it with Debian on it Aug 25 10:17:46 hrw: with a customized kernel of course, due to missing MMU Aug 25 10:18:12 -> lunch Aug 25 10:23:10 03Koen Kooi  07org.openembedded.dev * r1548843c0e 10openembedded.git/recipes/octave/ (octave.inc octave_3.0.3.bb): octave: fix QA issues Aug 25 10:35:47 hello Aug 25 10:35:54 while building on oe Aug 25 10:36:01 i'm getting the following error Aug 25 10:36:10 c ompiler cannot find executables Aug 25 10:36:20 | autoreconf: configure.ac: not using Libtool | autoreconf: running: /net/wintermute/vol1/home/anish/oebbtmp/staging/i686-linux/usr/bin/autoconf --include=/net/wintermute/vol1/home/anish/oebbtmp/work/i686-linux/xtrans-native-1_1.2.3-r0/xtrans-1.2.3/ --include=/net/wintermute/vol1/home/anish/oebbtmp/staging/i686-linux/usr/share/aclocal-1.10 --include=/net/wintermute/vol1/home/anish/oebbtmp/staging/i686-linux/usr/share/aclocal --force -- Aug 25 10:36:35 can sumone suggest something Aug 25 10:36:58 anish: yes, do your homework :) Aug 25 10:37:50 anish: autoconf points you to config.log, it is a text file, it can be viewed with a text editor Aug 25 10:38:09 pb__: master, when do you plan to hold OEDEM in cambridge? Aug 25 10:39:34 can it be due to toolchain error? Aug 25 10:40:11 anish: How hard is it to look at config.log? Aug 25 10:40:43 i'm just looking Aug 25 10:41:08 zecke_: I think mickey proposed some dates in november before. maybe the weekend of the 14th but I'll check. Aug 25 10:41:39 pb__: no need to, I can search mail too :) Aug 25 10:42:28 zecke_: heh, too late. seems it was the 7th in fact. Aug 25 10:42:34 it was 7th Aug 25 10:54:12 pb__: did you manage to test the glibc change? Aug 25 10:54:48 zecke_: sorry, not yet. I think I did manage to build an image successfully last night, though, so I am all set to test it now. Aug 25 10:55:07 every time I try I keep getting sidetracked by unrelated failures. :-} Aug 25 11:00:21 03Holger Hans Peter Freyther  07org.openembedded.dev * racdd06bbf0 10openembedded.git/ (4 files in 2 dirs): Aug 25 11:00:21 libmusicbrainz: Add version 3.0.2 and fix link error Aug 25 11:00:21 With gcc 4.3 stdc++ symbols leak into the library. One way to Aug 25 11:00:21 solve it is to use a custom linker script, another one is to Aug 25 11:00:21 just point gcc to the c++ library. The second option is used Aug 25 11:00:23 here right now. Aug 25 11:07:57 argh.. Aug 25 11:08:08 who played with openssh packages recently... Aug 25 11:08:25 hrw: check the mailinglist archive Aug 25 11:08:55 hrw: what is your problem with them? Aug 25 11:09:01 found reason Aug 25 11:09:07 someone broke upgrade path Aug 25 11:09:31 msmith_: it was you... openssh-keygen replaced ssh-keygen but not conflicts etc Aug 25 11:31:00 hi RP Aug 25 11:42:09 03Marcin Juszkiewicz  07org.openembedded.dev * r8d516d8ef6 10openembedded.git/recipes/openssh/openssh.inc: (log message trimmed) Aug 25 11:42:09 openssh: make openssh-keygen conflict with ssh-keygen (old name of it) Aug 25 11:42:09 Should fix situation after e05310f4958ba6c73488d398325670704401ee70 commit. Aug 25 11:42:09 Before: Aug 25 11:42:09 * Package openssh-keygen wants to install file /usr/bin/ssh-keygen Aug 25 11:42:11 But that file is already provided by package * ssh-keygen Aug 25 11:42:13 After: Aug 25 11:46:01 uf. vortex booted Aug 25 11:51:28 hrw: thanks, I didn't know how to handle the package rename Aug 25 11:51:54 So with the CONFLICTS, if someone has ssh-keygen installed and tries to install openssh-keygen, it'll give them an error message pointing them to uninstall ssh-keygen? Aug 25 11:52:02 is there anyway to say.. openssh-keygen replaces ssh-keygen? Aug 25 11:52:11 morning all Aug 25 11:52:43 mickey|patio: hey Aug 25 11:52:44 RP: hey! Aug 25 11:53:35 msmith_: replacing was always tricky with .pkg Aug 25 11:53:37 morning folks Aug 25 11:53:42 yo Z. Aug 25 11:53:55 mickey|patio, RP, msmith: hi Aug 25 11:54:28 hi mickey|patio Aug 25 11:55:53 Cx486SLC roxx Aug 25 11:56:02 does oe_runmake definition depends on some settings? I can bet I saw it running "make -e", but now (Angstrom+oe stable) i see just "make"... Aug 25 11:56:09 RP: Any chance you could take a look at the wm831x LED/backlight driver for 2.6.32? Aug 25 11:56:30 broonie: right, I need to sort out those trees Aug 25 11:57:41 jest: git grep EXTRA_OEMAKE Aug 25 11:57:49 RP: Sorry to nag but those are the only two wm831x drivers I need to chase into mainline. Aug 25 11:58:19 broonie: Yes, no problem. I will see what I can do Aug 25 11:58:56 hi mickey|patio Aug 25 11:59:01 hey pb_ Aug 25 12:00:22 msmith_: "Replaces: ssh-keygen" Aug 25 12:00:45 (and "Provides:", "Conflicts:", generally) Aug 25 12:15:30 ... lib/nof/libgcc_s.so: undefi Aug 25 12:15:32 ned reference to `fmsub' Aug 25 12:15:37 :( Aug 25 12:16:00 what can I do? Aug 25 12:17:36 RP: I bother you 'cause you're the author, sorry ^_^: packaged-staging fixes for OE-dev : http://pastebin.com/m4df6908 Aug 25 12:18:15 plus the interesting improvements in Poky oneday, please Aug 25 12:18:28 in OE too Aug 25 12:19:36 ant_work: You mean you want that submitted to OE? Aug 25 12:20:15 the fixes,yes, please. Then, there was a patch for speeding auto-staged things iirc Aug 25 12:21:23 one issue came deleting all but /pstage, the other was a mis-location of compilers after rebuild from pstage Aug 25 12:22:02 pb_: master? i have one unrelated question Aug 25 12:26:04 zecke_: I am at your service Aug 25 12:26:55 morning everyone Aug 25 12:27:38 pb_: great, I have some weird problems with ssh on a new machine. Job Control (CTRL+C, CTRL+Z, even closing a ssh connection) is not working properly... have you seen that before? Aug 25 12:29:14 zecke_: that does sound weird. what exactly happens? Aug 25 12:29:27 I don't recall ever having seen anything like that before. Aug 25 12:29:51 pb_: Well, I see ^C printed on the console.. and I'm running a bash Aug 25 12:30:11 pb_: honestly speaking the only thing I could think of is changing the keymap... besides that I have no idea of what could be wrong Aug 25 12:31:25 zecke_: if it says ^C then I guess the keymap is correct. is this over ssh, or at the local console? Aug 25 12:31:29 what does "stty -a" say? Aug 25 12:32:31 presumably the same bash binary works ok on other machines, right? Aug 25 12:33:35 pb_: it is a desktop :) Aug 25 12:33:41 pb_: good idea about stty Aug 25 12:35:52 re Aug 25 12:36:20 pb_: -iutf8 vs. utf8 Aug 25 12:37:40 pb_: no difference... Aug 25 12:38:23 hm. in that case I think I am out of ideas for the moment. sorry. Aug 25 12:44:59 pb_: *sigh* :) Aug 25 12:49:25 zecke: days ago I was lost in vi for similar reasons Aug 25 12:49:40 ant_work: how did you resolve it? Aug 25 12:50:00 I fear the issue is still there (Openembedded console-image) :/ Aug 25 12:50:16 I extracted the SD and edited on cesktop :D Aug 25 12:50:53 ant_work: oh, for me it is ubuntu karmic Aug 25 12:54:04 i know when i built bash for x86_64 i had to add some lines to the OE site file to get it to realize job control works Aug 25 12:54:25 but that's just because it was cross compiling... presumably ubuntu wouldn't cross compile :) Aug 25 13:09:35 pb_: more questions... do you know when gdb 6.9 is due? Aug 25 13:09:37 03Holger Hans Peter Freyther  07org.openembedded.dev * r344230ab19 10openembedded.git/ (conf/checksums.ini recipes/oprofile/oprofile_0.9.5.bb): Aug 25 13:09:37 oprofile-0.9.5: Add oprofile version 0.9.5 Aug 25 13:09:37 We can drop the other two patches as they have been applied. Aug 25 13:09:39 03Holger Hans Peter Freyther  07org.openembedded.dev * r2ec03e65dd 10openembedded.git/recipes/gdb/ (3 files): Aug 25 13:09:40 gdb: Make 6.8 packages match regarding patches and license Aug 25 13:09:42 6.8 was the first release to use GPLv3+, make sure the license Aug 25 13:09:44 is set correctly. Also make sure the karmic compile fix is applied Aug 25 13:09:46 for everyone. Aug 25 13:10:55 zecke_: allegedly 7.0 is due in september, I don't think there will be a 6.9 release Aug 25 13:11:50 pb_: okay, how do you know these plans? Aug 25 13:12:23 zecke_: see the gdb mailing list, http://sourceware.org/ml/gdb/2009-08/msg00186.html for example Aug 25 13:12:39 pb_: okay, I just looked at the schedule on the website :) Aug 25 13:13:33 * zecke_ has to get his alloca removal in Aug 25 13:15:29 what's that? Aug 25 13:16:11 pb_: In QtWebKit we use a switch to put each function into a section to make ld do a better job to remove unused... Aug 25 13:16:30 pb_: but this makes the elf parsing code in gdb use too much memory in alloca leading to a segfault :) Aug 25 13:17:28 (automake Q) how can i inject additional dependencies for one of the standard targets (e.g. dist) ? Aug 25 13:18:54 zecke_: ah, heh Aug 25 13:19:13 mickey|patio: if it's just an additional dependency, you should be able to write it as usual: Aug 25 13:19:18 dist: NEW-TARGET Aug 25 13:19:49 pb_: ah, so this is usually consecutive unless you add a 'command' section right below? Aug 25 13:20:38 ~curse NIH way of vendors working with kernel Aug 25 13:20:40 May you be reincarnated as a Windows XP administrator, NIH way of vendors working with kernel ! Aug 25 13:20:47 yeah, all the dependencies are cumulative. Aug 25 13:20:58 pb_, mickey|patio, zecke_ what do we need to do to get OEDEM on track for Cambridge? Aug 25 13:21:35 Crofton|work: a definite date and place for the invitation Aug 25 13:21:59 :) Aug 25 13:22:13 exactly Aug 25 13:22:25 mickey@opal:/local/pkg/fso/fso-abyss$ LANG=C make dist Aug 25 13:22:25 make: *** No rule to make target `src/consts.c', needed by `distdir'. Stop. Aug 25 13:22:28 hmm Aug 25 13:22:36 are we all happy with november 7-8? Aug 25 13:22:49 btw... what do we need to get permission to use OE logo on businesscards? Aug 25 13:22:53 pb_: I am Aug 25 13:22:59 * florian too Aug 25 13:23:16 hrw: we need to talk about that Aug 25 13:23:24 hrw: that would be a good point for the agenda Aug 25 13:23:25 yes, nov. 7-8 is the best compromise Aug 25 13:23:26 mickey|patio: DISTFILES? Aug 25 13:23:38 zecke_: good idea. trying... Aug 25 13:23:40 I will have to see if someone pays me a flight to europe Aug 25 13:23:40 florian: oedem is after elc ;( Aug 25 13:24:09 hrw: can you imagine someone complaining? ;) Aug 25 13:24:23 hmm, no Aug 25 13:24:32 DISTFILES is no target Aug 25 13:24:47 i guess we just need to require make all before make dist Aug 25 13:25:27 that is a bit odd. presumably you would expect to have a rule for generating src/consts.c, right? Aug 25 13:25:30 mickey|patio: wait Aug 25 13:25:51 what files is it generated from? if they aren't in distfiles then you would lose in that way. Aug 25 13:26:09 mickey|patio: http://repo.or.cz/w/memprof.git?a=blob;f=Makefile.am;h=53e0f1e029331a6dd13a511765924c553a676d35;hb=d8005f48542d8280e84063fa99f15f9e1a1c099a Aug 25 13:26:12 pb_: vala :) Aug 25 13:26:17 it must be vala Aug 25 13:26:21 shipping them with EXTRA_DISTFILES Aug 25 13:26:22 iirc, automake does a two-step dist to ensure that everything is consistent: it makes dist, unpacks the resulting file, and then makes dist again in that clean tree. Aug 25 13:26:45 hm, that should be fine then. weird. Aug 25 13:26:53 oh, dist-hook! Aug 25 13:27:25 anyway, apropos OEDEM, looks like we are all agreed. nov 7 also has the benefit that you can go and watch the fireworks display in the evening. Aug 25 13:27:40 sounds good Aug 25 13:28:12 hmm, darn Aug 25 13:28:20 (dist) Aug 25 13:28:26 still no go, even with dist-hook Aug 25 13:28:28 at least, I think it'll be on the 7th. it's usually the saturday closest to nov 5, when we (perversely) celebrate our government not being blown up. Aug 25 13:28:34 hehe Aug 25 13:29:17 mickey|patio: hmm... are you sure your Makefile got regenerated? Aug 25 13:29:49 grr, now udev-124-r1 refuses to unpack for some reason. Aug 25 13:29:51 * pb_ stabs it Aug 25 13:29:54 pb_: oh more stuff for you... gtk_widget_realize... how would you force that recursively? gtk_container_for_all? Aug 25 13:30:01 pb_: oh, broken pipe? Aug 25 13:30:03 yes, 100% sure Aug 25 13:30:07 zecke_: yeah, broken pipe Aug 25 13:30:16 zecke_: I thought gtk_widget_realize was recursive anyway Aug 25 13:30:17 pb_: udev 124 unpacking is pita Aug 25 13:30:18 pb_: new GNU tar + python == saddness Aug 25 13:30:21 pb_: it is not Aug 25 13:30:39 pb_: (realize) the first show of the widget is still slow :) Aug 25 13:31:04 zecke_: oh, right, yeah. it recurses upwards but not downwards. Aug 25 13:31:21 argh, wait Aug 25 13:31:21 i.e. realizing a widget realizes its parent, but not its children Aug 25 13:31:48 so, if you happen to have a pointer to the most deeply nested widget, you could just realise that one and... voila. Aug 25 13:31:54 otherwise, yeah, I guess gtk_container_for_all. Aug 25 13:32:31 zecke_: is there a solution for the broken pipe, other than casting udev into a fiery pit of doom? Aug 25 13:33:02 pb_: yes, khem ray should push his patch... python should SIG_DEFL the sigpipe Aug 25 13:33:20 pb_: (realize)... right... Aug 25 13:33:27 ah right, yeah, I remember that patch now Aug 25 13:33:51 that was the one where he failed to give appropriate credit to some dude on irc :-} Aug 25 13:34:01 yeah... ego problems *sight* :} Aug 25 13:35:59 mickey|patio, zecke_: so, I'll send an invitation for oedem at my office and warn people that they need to register in advance rather than just showing up. if we get an unmanageable number of responses then we will have to look at moving to another venue nearby (e.g. th1_ office, which is on the other side of town). Aug 25 13:36:02 question I have: which kernel subsystem would you use for 3D accelerometer? Aug 25 13:36:17 input, I guess Aug 25 13:36:31 on board here vendor used hwmon Aug 25 13:36:56 pb_: the official invitation as an e.V. general assembly has to be sent by the board, but yes, it would be good to send an informal invitation to gather an idea about the participation numbers. Aug 25 13:37:18 ant_work: I just don't really have the time :( Aug 25 13:37:24 mickey|patio: ah, I thought you were doing the e.V. general assembly separately now. Aug 25 13:37:31 ant_work: I'll see if I can sort something out, maybe make a branch for testing Aug 25 13:37:46 pb_: yes, and no. we're doing a small one to fix the voting situation Aug 25 13:37:55 pb_: but next OEDEM was planned as the first 'big' one Aug 25 13:39:31 ah right, I see Aug 25 13:43:31 * XorA is tempted to turn up at the EGM just to get some .de beer! Aug 25 13:44:24 XorA: heh, I don't think that the entire german beer output will be reserved for the egm. Aug 25 13:44:39 England for Guy Fawkes day will be great Aug 25 13:44:53 Crofton|work: fireworks are always great :-) Aug 25 13:45:41 Crofton|work: we wont mind if you want to renenact it :-) Aug 25 13:47:16 heh Aug 25 13:47:44 I had to explain to an American that if they same thing happened today, in the US, Guy Fawkes would have become a national hero .... Aug 25 13:55:14 Crofton|work: so how are the US guys doing with hating scotland at the moment, they must be a bit torn in the middle? Aug 25 13:56:16 no one seems excited Aug 25 13:56:54 everyone is trying to figure out to pretend everyone in the US already has healthcare Aug 25 13:56:55 Crofton|work: so all made up by the media then Aug 25 13:58:14 probably Aug 25 14:20:46 ugh Aug 25 14:20:49 * kergoth grumbles Aug 25 14:27:30 zecke_: heh, just re-running bitbake seems to have made the udev problem go away Aug 25 14:27:51 * pb_ goes to rinse his mouth out with carbolic soap after mentioning that word Aug 25 14:28:56 whew, totally didn't think about that.. when using a recipe inside of an scm checkout of a source tree, and generating PV from git info, had to mark .git/HEAD as a dependency with bb.parse.mark_dependency to ensure the cache got invalidated when i checkout a different version of the source Aug 25 14:29:49 good grief, why does console-image build flex?! Aug 25 14:30:04 kergoth: heh Aug 25 14:30:19 pb_, stock answer, check the depends graph :) Aug 25 14:31:10 rxvt recipe has a funny bug. it downloads from cvs into a file xxxx-todaysdate-xxx.tar.gz .. and the next day you try to build it again, and it looks for "xxxx-todaysdate-xxxx" instead of "xxxx-yesterdaysdate-xxxx" and thus fails to build.. Aug 25 14:34:46 Crofton: apparently libpam depends on flex. I find that fairly hard to believe. Aug 25 14:35:19 maybe it uses lib flex as aprt of its parsing? Aug 25 14:36:42 yeah, I guess that could be true. Aug 25 14:37:07 I'm not quite sure how pam ended up in there in the first place. I guess I chose my image rather poorly. Aug 25 14:37:11 it doesn't seem crazy Aug 25 14:37:14 :) Aug 25 14:38:27 hrmph, downside to inscm operation, things get.. interesting, sharing a ${S} between both native and non-native recipes Aug 25 14:38:30 * kergoth grumbles Aug 25 14:43:42 gah. my head hurts. wonder if there are any other ways to improve the active development use case(s) Aug 25 14:45:00 th1_: cvs fetcher fetch today code Aug 25 14:48:24 Crofton|work: hm, it turns out that flex doesn't actually ship any shared libraries, so that can't be the explanation Aug 25 14:49:08 the flex recipe just builds a single binary package, "flex" (plus -dbg/dev/doc noise of course) and flex.ipk just contains the single file /usr/bin/flex. Aug 25 14:49:35 when glibc gets done packaging I guess I'll find out whether anything actually depends on it. Aug 25 14:49:44 doesn't flex also stage libfl.a? Aug 25 14:50:35 ah, possibly. Aug 25 14:50:47 static libraries are a bit too retro, I didn't think of that. :-} Aug 25 14:50:50 :) Aug 25 14:51:32 yeah, you're right, there's libfl.a in staging. I guess some of pam must be linking with that. Aug 25 14:52:40 newer flexes do support building a shared libfl.so; I suppose it'd make sense to update. Aug 25 14:53:02 ah, nice Aug 25 14:53:17 though, for my purposes, throwing pam into the nearest abyss is an easier solution :-} Aug 25 14:54:17 has anyone thought about changing the shlibdeps, etc code to check if a recipe that emits the runtime packages we need exists in DEPENDS or task dependencies, to check for deps coming in due to "optional" stuff? Aug 25 14:54:45 if a dep on libfoo suddenly shows up and we don't dep on foo, but can use it if it exists, thats a problem in build determinism Aug 25 14:54:52 hmm Aug 25 14:54:55 no, but that's quite a neat idea. I did something similar for the autofoo .m4 files a while back. Aug 25 14:55:05 yeah, thats what made me think about it :) Aug 25 14:55:13 heh Aug 25 14:55:31 we've gotten bit by a lot of these sorts of bugs for mvl6 Aug 25 14:55:37 since we use the pstage packages Aug 25 14:55:48 suddenly you can get a pstage binary that links against libfoo when foo hasn't been built Aug 25 14:56:51 yeah, I can see that would be a pain. Aug 25 14:57:25 should be easy enough to have shlibdeps detect that, I think. Aug 25 14:58:00 would think so. i'd think pkgdata would have the necessary info to map back to the recipe name from the runtime package, and scanning DEPENDS + task depends is easy enough Aug 25 14:58:14 once you've found the binary package names for each shared library, you'd then just need to map those back to recipes and then check to see if they feature in DEPENDS, etc. Aug 25 14:58:20 right, exactly Aug 25 15:31:54 hrmph Aug 25 15:31:56 kergoth: idea is nice Aug 25 15:32:09 k its on the ever growing list.. Aug 25 15:32:35 hey cminyard Aug 25 15:32:48 kergoth: share a list? Aug 25 15:33:26 will clean it up.. right now I've just been adding items to an outline, its just a giant pile of random ideas Aug 25 15:34:08 * kergoth wonders why our event handlers use getName() instead of isinstance() Aug 25 15:39:58 okay, so, does anyone have any opinions on whether this sort of thing is worth pursuing / polishing? http://dl.getdropbox.com/u/112715/Files/inscm-1.tar.bz2 Aug 25 15:46:43 kergoth, here Aug 25 15:47:12 didn't need anything, just hey :) Aug 25 15:47:22 ok, hello :) Aug 25 16:07:38 hmm, we should make it easier / cleaner to leverage B vs S Aug 25 16:08:29 enable it in kernel.bbclass with EXTRA_OEMAKE = "'O=${B}'" or somesuch.. Aug 25 16:08:48 add an autotools_builddir to do the ${B} in the task dirs flags, remove that from base.bbclass Aug 25 16:08:58 set B in bitbake.conf to a sane value != S Aug 25 16:11:03 * kergoth adds to the list Aug 25 16:22:08 anyone familiar with the webkit recipe Aug 25 16:24:04 specifically why is the do_compile_prepend() necessary and what does it do Aug 25 16:24:08 you know what i was mulling over last night.. bbclasses as python eggs. require > version 1.0 of autotools, or whatever... Aug 25 16:24:18 * kergoth isn't, sorry Aug 25 16:29:31 th1_: dftables thing? Aug 25 16:30:03 kergoth: I forgot that ${B} exists Aug 25 16:30:12 hehe, many do, cause we never use it Aug 25 16:30:19 but itd be nice to do so for projects that support it Aug 25 16:30:25 hwyes Aug 25 16:30:29 hrw, yes, even Aug 25 16:31:19 th1_: simple - we build it with BUILD_CC/CXX to have it built for host not for target. it is used during build but normally it is built with TARGET_CC/CXX so you are not able to run it Aug 25 16:31:29 th1_: we do that for many things Aug 25 16:33:24 ok Aug 25 16:33:31 makes sense Aug 25 16:33:33 getting from home to oedem is simple. time to search for booking... Aug 25 16:33:39 I will try to build latest svn Aug 25 16:33:58 hrw: http://dl.getdropbox.com/u/112715/Files/OpenEmbedded.html.zip Aug 25 16:34:09 exported my current outline (like i said, a mess) as html Aug 25 16:34:28 some of that are ideas I've already scrapped, even.. Aug 25 16:34:29 heh Aug 25 16:37:49 kergoth: which tool you use for it? Aug 25 16:37:59 omnioutliner on my mac :) Aug 25 16:39:26 OE_LT_RPATH_ALLOW sounds like rpath handling in libtool or sth like that Aug 25 16:39:40 yeah thats what i figured too, never got around to checking exactly Aug 25 16:43:38 kergoth: idea for -dbg packages was to keep one per recipe Aug 25 16:44:26 hrw: do you recall why? i have image.bbclass modifications which automatically emit a debugging image that corresponds to the image you built, for doing easier debugging with gdb + gdbserver. doing that is a pain if the dbg info is always global. Aug 25 16:45:18 I think that it was for simplicity of manual working Aug 25 16:45:29 hmm Aug 25 16:45:32 kergoth: console-image-dbg idea sounds nice for me Aug 25 16:46:11 hmm, will have to give it some more thought Aug 25 16:46:52 kergoth: but with current -dbg packages you can still use gdb+gdbserver. just gdb will get larger rootfs Aug 25 16:47:26 how would you tell the gdb on the build machine where the debug info is, without a populated rootfs that has the .debug files? Aug 25 16:47:38 i guess if you use nfsroot Aug 25 16:47:56 no, I created debug rootfs by hand use of opkg Aug 25 16:48:06 ahh, yeah, thatd work Aug 25 16:49:03 so your code to create image-dbg is nice addon (if possible to disable/enable it) Aug 25 16:49:07 i'll think about it some more and see about pushing the -dbg image bits upstream. right now its not optional, every image emits both dbg and non :) Aug 25 16:49:07 03Jeremy Lainé  07org.openembedded.dev * r728dd52472 10openembedded.git/recipes/linux/ (3 files in 2 dirs): linux-2.6.29: update lm73 driver for boc01 Aug 25 16:49:08 * kergoth nods Aug 25 16:49:15 will add a flag Aug 25 16:49:22 and console-image-dev? Aug 25 16:49:33 thats a good idea, easy enough to add Aug 25 16:49:45 i was thinking -doc would be useful too for some folks. set a var to get the docs in your images Aug 25 16:50:00 but for that we need to enable depchains for -doc just in tasks, to get task-base-doc, etc Aug 25 16:50:05 started looking into it last week Aug 25 16:50:30 when you are at it then maybe make it generic? Aug 25 16:50:47 -dbg, -dev, -doc, -locale-LANG1 -locale-LANG2 Aug 25 16:50:54 yeah thats what i was just thinking about.. some way to produce image/rootfs variants Aug 25 16:50:56 hmm Aug 25 16:51:51 it will allow us to kill some code for locales Aug 25 16:52:47 shouldnt be too much work. some callback mechanism to inject new variants via python functions that manipulate some vars in the metadata for each one before executing the do_rootfs, and a global config variable with a list of variants to build Aug 25 16:54:03 nfsroot generation sounds interesting but it would require very good testing to not do too much under sudo... Aug 25 16:54:41 yeah, good point, dangerous Aug 25 16:58:06 kergoth: by disk images you mean hdd image with partition table + rootfs in partition? Aug 25 16:58:14 yeah Aug 25 16:58:21 i think poky has some of that already Aug 25 16:58:26 haven'tt had a chance to look Aug 25 16:58:47 bootimg.bbclass? Aug 25 17:07:29 kergoth: "Possible to rebuild a package / invalidate its stamps when a dependent lib's ABI changes?" is not handled by 'whitelist' scheduler? Aug 25 17:07:37 ah.. ABI.. right Aug 25 17:08:06 whitelist is any build/rebuild, i wanted just whent he abi changes Aug 25 17:08:21 and i have that implemented internally in mvl6... you don't want to know how ugly the implementation was, but it works Aug 25 17:09:12 kergoth: you do not want to see my 32KB long recipe used to build image Aug 25 17:09:17 yikes Aug 25 17:10:13 and it uses image rootfs to generate package which will be used to build image Aug 25 17:10:46 erk, fyi, when setting PV based on execution of an external tool, make sure the external tool's output doesn't change from one task to the next Aug 25 17:10:47 hehe Aug 25 17:11:00 * kergoth just had do_setscene get a stamp with a different PV than the other stamps... bitbake got very confused Aug 25 17:14:12 cleanall task is very useful when doing builds inside source trees Aug 25 17:16:39 * hrw back to playing blindly with board Aug 25 17:20:54 wonder if a variable to toggle build/packaging of kernel modules from the kernel recipe would be useful. could toggle all the modules in the defconfig, but if a person just doesn't care about any of the stuff that isn't builtin, such a toggle would be a quick way to speed up the build process.. Aug 25 17:20:55 hmm Aug 25 17:22:39 i kind of like how simple and easy android is. get repo, run 2 repo commands, run make. and in there its easy to do active development in the source trees of the components.. Aug 25 17:23:26 heh Aug 25 17:23:30 That's all I'll say :) Aug 25 17:23:33 heh :) Aug 25 17:24:04 i think the inscm stuff I'm playing with is a step in the right direction. i still think a more project based bitbake command would be good. hmm Aug 25 17:25:16 why do we have a static copy of the webkit makefiles and config etc. instead of using them from svn?? Aug 25 17:25:33 th1_: because svn do not allow to fetch one file Aug 25 17:25:34 bye Aug 25 17:26:08 ok bye Aug 25 17:26:14 but it makes it hard to upgrade :/ Aug 25 17:27:36 hmm.... could add ${P} to OVERRIDES in this recipe, then do conditional bits for different versions, so a single recipe would be useful no matter what version i have checked out.. hmm Aug 25 17:29:40 I think the way webkit is done with static copies of the files from top level is very weird Aug 25 17:30:32 sounds awfully hackish Aug 25 17:30:55 in fact it is possible to get the version: http://trac.webkit.org/browser/trunk/configure.ac?rev=47748&format=raw Aug 25 17:30:55 i mean, if it came down to it, you could override do_fetch and svn cat the remote path or something, if you really only had to get those specific files.. Aug 25 17:31:05 yeah Aug 25 17:31:08 pretty sure svn cat http://foo works Aug 25 17:32:19 lool Aug 25 17:32:27 check in fileS: "update-webkit.sh" Aug 25 17:32:47 now if ever there was uglyness... Aug 25 17:33:24 haha Aug 25 17:33:35 and why wasn't that just done in do_fetch, already? Aug 25 17:33:40 * kergoth shakes head Aug 25 17:33:46 pb_: my virtual/lib build fine yesterday with micro/eglibc/qemuarm Aug 25 17:34:06 yeah do_fetch would work Aug 25 17:34:32 I don't have time to fix it properly now though. have to have a working X+Midori+NetworkManager on Friday Aug 25 17:35:30 i love how many upstreams are using git nowadays. makes integration so easy Aug 25 17:39:37 I have an oe snapshot thats a bit behind which fails to build gdb on a platform as it encounters uninitialized var warnings turned to errors - current oe uses the exact same compiler flags and compiler version and doesn't encounter these - any idea why or what could have changed that would affect this? I can't bring my snapshot up to date at this point in time and am looking for a surgical fix Aug 25 17:41:21 have you tried to copy the recipe from head? Aug 25 17:41:55 ya, the recipe has not changed. I've diffed the gdb workdir too and the code shows no diff - gotta be something in the toolchain Aug 25 17:42:46 only thing I see that changed is my snapshot uses autoconf 2.61 and the current oe uses autoconf 2.63 - not sure if/how that affects anything Aug 25 17:45:52 not sure Aug 25 17:48:43 at this point I wouldn't mind simply disabling the package until I can resync with oe-git but 'DEFAULT_PREFERENCE_pn-gdb = "-1"' does not seem to disable gdb from the image Aug 25 17:49:09 defalut preference is just used to control defalut provider selection among multiple providers of a thing Aug 25 17:49:13 it doesn't control whether something is built Aug 25 17:49:38 you could do ASSUME_PROVIDED += "gdb", which would make it not be built, but if your image wants the binary package, it'd fail there Aug 25 17:50:44 hmm... so I guess I would need some special per-machine python handling to remove the package from the list when creating the image Aug 25 17:51:34 most likely.. mangle PACKAGE_INSTALL/IMAGE_INSTALL or whatever Aug 25 17:51:35 * kergoth shrugs Aug 25 17:51:37 or find out why the image wants gdb and edit the corresponding recipe Aug 25 17:51:41 maybe I'll try to patch the gdb recipe and include '-Wno-uninitialized' as I think thats all thats needed to build gdb (again not sure why that has changed in current OE though) Aug 25 17:52:04 tharvey, you can use devshell to experiment like that Aug 25 17:52:23 devshell failed for gdb - not even sure if devshell works for anything anymore? Aug 25 17:52:29 you know what i want.. i want bitbake to realize that if a task's value is the empty string, that task doesn't need to be run, can be considered to be complete, don't fire off a build of that task at all Aug 25 17:52:33 works for everything I've tried it for Aug 25 17:52:45 then a recipe could easily say, i dont need to fetch, do_fetch () {} and you wouldn't even see it run a fetch for it Aug 25 17:53:55 03Sebastian Krzyszkowiak  07shr/import * r3a1f16c74c 10openembedded.git/recipes/navit/ (navit.inc navit_svn.bb): navit: don't need -dev package to run anymore Aug 25 17:53:56 03Martin Jansa  07shr/import * r2260fcc988 10openembedded.git/recipes/navit/ (4 files in 2 dirs): Fix navit library path. navit.xml was renamed to navit_shipped.xml in svn rev 2492. Aug 25 17:53:56 03Sebastian Krzyszkowiak  07shr/import * r9c3c6893d5 10openembedded.git/recipes/navit/files/navit.desktop: navit: fix QA issue in desktop file Aug 25 17:53:57 03Michael  07shr/import * rbcf6b2b5c1 10openembedded.git/recipes/gtk-webcore/ (files/ua-iphone.patch midori_0.1.2.bb): midori: add iPhone useragend string. Closes #565 Aug 25 17:55:39 tharvey: if you use gnome-terminal, you may want 6b86d4060be13734848bcda7bc52760765baeafe Aug 25 18:04:17 kergoth, thanks - that was the issue Aug 25 18:04:27 with devshell Aug 25 18:07:24 np Aug 25 18:12:07 kergoth, devshell working now? Aug 25 18:12:36 should be, if not open bugs :) Aug 25 18:16:03 i wonder how much work it would be to determine when a recipe is requiring that the classes it inherits be done at that position in the file, rather than prior to the recipe's metadata Aug 25 18:16:07 hmm Aug 25 18:16:15 hi Aug 25 18:16:44 im trying to build the meta-toolchian package for the minimal distribution and an armv4t board. rootfs builds fine but the meta toolchian always fails with Aug 25 18:16:47 | /home/oe/mxm-7114/tmp/work/i686-armv4t-sdk-oe-linux-gnueabi/gcc-cross-sdk-4.2.4-r5/gcc-4.2.4/gcc/tsystem.h:103:20: error: string.h: No such file or directory Aug 25 18:17:14 and ither headers missing like errno.h time.h stdio.h .... anyone got an idea? Aug 25 18:20:01 minimal doesn't work for meta-toolchain as the gcc sdk packages don't like /include Aug 25 18:20:32 but when i remember correctly it worked some time ago Aug 25 18:20:36 didn't it? Aug 25 18:20:52 probably not, no Aug 25 18:21:04 iirc minimal has always been /include not /usr/include Aug 25 18:27:19 really? micro is like that, but I didn't think minimal was. Aug 25 18:27:46 pb_ i think so, too Aug 25 18:28:43 ok, gcc-cross-sdk package tries to get the includes from ${HOST] staging dir instead of TARGET Aug 25 18:29:16 it looks in staging/i686-linux/usr/include instead of arm-oe-linux-gnueabi/usr/include ... strange Aug 25 18:42:21 looks like SDK_CFLAGS is set wrong Aug 25 18:42:42 export SDK_CFLAGS="-isystem/home/oe/mxm-7114/tmp/staging/i686-linux/usr/include -isystem/home/oe/mxm-7114/tmp/staging/i686-linux-sdk/usr/include -fexpensive-optimizations -frename-registers -fomit-frame-pointer -O2" Aug 25 18:42:54 03Sebastian Krzyszkowiak  07shr/import * rafab50e479 10openembedded.git/conf/distro/include/shr-autorev-unstable.inc: shr-autorev-unstable: bump navit to 2520 Aug 25 18:43:00 03Sebastian Krzyszkowiak  07shr/import * r4df36da9b5 10openembedded.git/recipes/navit/navit_svn.bb: navit: remove .so.0 patch for navit.xml, not needed anymore Aug 25 18:43:02 shouldnt it be arm-oe-linux-gnueabi instead of i686-linux Aug 25 18:52:56 * kergoth thinks about what itd take to put together an uber-minimal rootfs builder that doesn't use packaging at all, using his inscm project which only has busybox & linux-2.6 Aug 25 19:12:50 haha Aug 25 19:21:34 ok somethings really wrong with the gcc-cross-sdk package Aug 25 19:21:54 damn why does it include the i686-linux dir.... Aug 25 19:22:16 yes, there's an assumption about stuff being /usr/include somewhere in there Aug 25 19:22:20 haven't had the time to go root it out Aug 25 19:23:14 03Evan Doiron  07org.openembedded.dev * r40cc71035c 10openembedded.git/ (conf/checksums.ini recipes/rsync/rsync_3.0.6.bb): Aug 25 19:23:14 rsync: add v3.0.6 Aug 25 19:23:14 Signed-off-by: Evan Doiron Aug 25 19:23:14 Signed-off-by: Michael Smith Aug 25 19:23:14 Acked-by: Khem Raj Aug 25 19:23:19 hmmmm Aug 25 19:23:33 Taratus: no .. the problem is it includes the wrong arch Aug 25 19:23:44 int includest the native stating dir instead of the target one Aug 25 19:24:15 host staging dir never contains glibc header files because thers no such package Aug 25 19:28:19 pb_, thanks for stepping up to the "crease" on OEDEM Aug 25 19:29:44 hm Aug 25 19:29:54 let me see if there is an easy jet flight Aug 25 19:31:43 * kergoth really needs to get some cash together to get to one of these conferences where OE folks are gathering, or talk MV into sending him Aug 25 19:32:03 woglinde: germanwings are probably your best option for flying to stansted Aug 25 19:32:15 pb hm which date it was again? Aug 25 19:32:22 easyjet only fly to luton, which is a little bit further away Aug 25 19:32:23 hi pb btw. Aug 25 19:32:25 woglinde: nov 7 Aug 25 19:32:28 hi woglinde :-) Aug 25 19:33:06 does everyone use mdev with uclibc builds these days? Aug 25 19:33:11 udev does not seem to want to build Aug 25 19:33:25 re Aug 25 19:33:47 jo hrw Aug 25 19:33:55 hm I could fly with hrw Aug 25 19:34:01 if he comes Aug 25 19:34:24 woglinde: to oedem? Aug 25 19:34:29 hrw yes Aug 25 19:34:45 hm no flyes on nov 7 Aug 25 19:34:46 woglinde: ryanair is other option Aug 25 19:34:58 woglinde: go on 6th and back on 9th Aug 25 19:35:10 hrw I have to work Aug 25 19:35:16 kergoth: yah, you can get flights quite cheaply at the moment. I just checked DFW->LON on aa.com and it quoted me about $350 for november, which isn't a bad price. Aug 25 19:35:52 you might get something a bit cheaper if you tell it that you live in the us, they seem to apply a special tax to non-americans for some reason. Aug 25 19:35:52 huh, not bad indeed. maybe i should start budgeting a certain amount every month into a travel account for conferences Aug 25 19:35:53 heh Aug 25 19:35:56 heh Aug 25 19:36:24 hrw and I have more hotel costs Aug 25 19:36:36 woglinde: I know Aug 25 19:36:56 woglinde: but otherwise you have to fly very early morning to not miss half of even Aug 25 19:36:57 t Aug 25 19:36:59 kergoth: I guess at that price it shouldn't be too hard a sell to your boss at MV. Aug 25 19:37:13 woglinde: and I need to add 2-3h for szczecin<>berlin trip Aug 25 19:37:42 perhaps.. but MV just cut everyone in the company's pay by 8%, they're being awfully strict on "unnecessary" expenses Aug 25 19:37:44 hrw yes Aug 25 19:37:49 well, not just cut, but a few months back Aug 25 19:37:54 stupid economy Aug 25 19:37:57 kergoth: doh, I see Aug 25 19:38:00 hrw or you come earlier and sleep here Aug 25 19:38:05 +day Aug 25 19:38:24 schoenefeld is only 20 minutes from here Aug 25 19:39:04 Crofton: did you ever figure out the udev/ppoll issue? Aug 25 19:39:16 let me check Aug 25 19:39:20 I think that must have been someone else ... Aug 25 19:39:47 * Kero ponders... what would be on an OEDEM agenda; and can I get my customer to pay, i.e. dropping provided debian from the board and install not-quite-provided OE instead... Aug 25 19:39:52 Crofton: http://www.beagleboard.org/irclogs/index.php?date=2009-05-21 Aug 25 19:39:56 cbrake: iirc, the ppoll thing just needed to be fixed in the kernel. I don't think there was any other solution. Aug 25 19:39:57 that was clear back in May Aug 25 19:40:10 i love how fast do_clean / rm -rf tmp is when doing builds inside source trees. no unpacking Aug 25 19:40:16 hmm Aug 25 19:40:35 ...or just come because of openmoko ;) Aug 25 19:40:39 heh Aug 25 19:40:46 that was a long time ago, I forget Aug 25 19:40:48 Kero: yes, good point, we should formulate an agenda. Aug 25 19:40:51 gcc-cross-sdk in stable is missing zecke-no-host-includes.patch, that was the reason that the build of meta toolchain failed. the gcc dir of -dev and -stable has a huge drift! Aug 25 19:41:40 pb_: how are you, been a while :) iPAQ is gathering dust, but my openmoko finally does duty as a daily phone Aug 25 19:41:40 woglinde: 09:35 sxf ryanair -> 10:35 stn and back 18:45 -> 21:30. germanwings do not fly on 7th Aug 25 19:42:17 yeah found the same Aug 25 19:42:23 77.34 Aug 25 19:42:42 woglinde: add some time for airports + train STN<>Cambridge Aug 25 19:42:52 if you land 1035 at STN you would probably reach the venue around 1200 Aug 25 19:43:08 ~12:00 at oedem and go from there at 15:00-16:00 Aug 25 19:43:14 so, not a disaster, but you might consider flying in the previous night instead. Aug 25 19:43:22 do you think we will have any time for hacking before/after? Aug 25 19:43:41 hms Aug 25 19:43:45 damn opodo Aug 25 19:43:48 woglinde: when you end work? what time? Aug 25 19:44:04 woglinde: 16:25 and 21:55 are flights on friday Aug 25 19:44:07 hrw hm I can fly friday no prob Aug 25 19:44:19 but thursday would be to early Aug 25 19:44:30 woglinde: friday for me is fine Aug 25 19:44:45 hm Aug 25 19:44:46 hrw: did you have a chance to have a look at the qemu issue? Aug 25 19:44:54 Laibsch: nope, sorry Aug 25 19:44:56 btw. I was today at schoenefeld with tobi Aug 25 19:45:02 Laibsch: I have 3 projects at same time Aug 25 19:45:02 OK Aug 25 19:45:03 looking at the planes Aug 25 19:45:24 Crofton: on the friday/monday, you mean? probably not if it's at my office; we would need the space on those days. Aug 25 19:45:26 hrw but we would need some hotel Aug 25 19:45:44 might be able to find a hacking space somewhere else though. Aug 25 19:46:11 woglinde: double room should not be so expensive Aug 25 19:48:36 have a nice rest of day Aug 25 19:48:40 bye hrw Aug 25 19:48:55 hm let us finalyze us this this week Aug 25 19:49:07 have to ask my wife first Aug 25 19:55:53 pb_, it seems like there are companies near there that would be interested .... Aug 25 19:56:32 jo ant Aug 25 19:56:50 hey woglinde Aug 25 19:56:52 woglinde: there are plenty of hotels in the area, it should be easy to find something. Aug 25 19:57:36 Crofton: yeah, th1 mentioned that he might have space in his office as well. Aug 25 19:57:55 pb I have no idea Aug 25 19:58:07 never was in london or uk jet Aug 25 19:58:58 ah, I see. Aug 25 19:59:06 what type of hotel do you want? I can probably find something for you. Aug 25 19:59:14 pb hehe Aug 25 19:59:24 I will come with hrw Aug 25 19:59:32 so let us discuss this tomorrow Aug 25 19:59:37 righto Aug 25 20:01:06 ok Aug 25 20:01:18 I have an embaressing question Aug 25 20:01:35 how do I get a screen shot in "windows" Aug 25 20:01:54 crtl print Aug 25 20:02:04 then open ms paint Aug 25 20:02:04 where does the fiole go? Aug 25 20:02:07 crtl v Aug 25 20:02:10 ah Aug 25 20:05:44 pb_, let me know how the signups look, if I need to try and get our office as much notice as possible would help Aug 25 20:08:26 ugh, trying to figure out all the ways to extract version info from scms for inscm is going to make my head explode Aug 25 20:08:44 http://www.flickr.com/photos/32615155@N00/3857143522/ Aug 25 20:08:48 like, what if they're using a subtree merge for the individual components.. hrmph Aug 25 20:11:06 Crofton: neat, how do you like the intronix? Aug 25 20:11:14 well Aug 25 20:11:19 without it I would be lost Aug 25 20:11:27 it is annoying due to the windows factor Aug 25 20:11:31 nod Aug 25 20:11:37 but I already needed windows for the fpga sw Aug 25 20:11:48 crfotn hm this is bad Aug 25 20:12:07 getting the altera download pod and the intronix working at the same time was a pain Aug 25 20:12:29 Crofton: maybe http://www.totalphase.com/ will make a logic analyzer some day Aug 25 20:12:44 Crofton: I have a USB analyzer which works well on Linux Aug 25 20:12:55 cool Aug 25 20:13:01 or drivers for the intronix Aug 25 20:16:55 so did Russell shut down the ARM maillists or not? I'm still getting mail .... Aug 25 20:17:19 cbrake transfers Aug 25 20:17:38 you should have automaticly susbscribed to the new list Aug 25 20:17:39 s Aug 25 20:18:05 woglinde: I'm still getting mail from lists.arm.linux.org.uk Aug 25 20:18:54 okay good nite Aug 25 20:19:04 * * OE Bug 5304 has been created by chrisfriedt(AT)gmail.com Aug 25 20:19:06 * * cups-1.2.12 fails, No GNU_HASH in the elf binary, libcups.so.2 libcupsimage.so.2 Aug 25 20:19:08 * * http://bugs.openembedded.org/show_bug.cgi?id=5304 Aug 25 20:29:46 re Aug 25 20:52:23 anyone know makedevs well? Aug 25 20:52:24 hmm Aug 25 20:55:29 kergoth: what is question? Aug 25 20:55:53 can the device table format handle globs or recursion or anything, for doing permissions fixups outside of /dev? Aug 25 20:56:13 ah.. device tables.. Aug 25 20:56:33 sorry, don't know :) Aug 25 20:56:57 np Aug 25 20:57:04 but you can try to add more column with globs :) Aug 25 20:57:09 heh :) Aug 25 20:57:23 hmm Aug 25 20:57:37 i wonder what itd take to take the stored output of a fakeroot execution and use it to replay some/all of those changes under sudo Aug 25 21:19:53 denix0: I'm guessing you wrote the ruby 1.8.5 recipe Aug 25 21:21:04 I have a 1.8.6 update, with recipe for gtk binding, working on rubybgems right now. Aug 25 21:54:37 Crofton: Maybe you want to take a look at the FreeDNS service from namecheap.com (I have one domain with them). They offer DNS services for free even for those domains not registered by them. You could then untie oe.org from your personal account and provide redundancy in managing DNS Aug 25 21:55:02 http://www.namecheap.com/freedns/free-manage-dns.aspx Aug 25 21:55:09 Kero: wrong guess :) I'm not familiar with ruby Aug 25 21:56:28 denix: you are not denis at denix ?, thought I was making a decent guess there... Aug 25 21:58:24 Laibsch, Crofton: I use mydomain.com for free dns... Aug 25 21:59:07 remind me toward the end of the year :) Aug 25 21:59:19 Kero: that's me, all right. but I have nothing to do with ruby recipes :) Aug 25 21:59:40 Kero: I guess you looked at my commit renaming packages/ into recipes/ Aug 25 22:00:00 Kero: try "git log --follow" to see the complete history Aug 25 22:00:07 denix: hey! we have russian ML now on googlegroups Aug 25 22:00:38 denix: hm you are right. I had not expected something like that ;) Aug 25 22:01:00 Jay7: sign me up! :) Aug 25 22:01:22 Jay7: what's the name? Aug 25 22:01:36 ah, got it Aug 25 22:01:57 Kero: fyi, --follow only works for git log . if you want to see a path, you can specify both before and after locations. git log -- packages/foo recipes/foo Aug 25 22:02:02 s/path/directory/ Aug 25 22:02:29 denix: http://openembedded.ru/ Aug 25 22:02:36 there is link Aug 25 22:02:55 norguthar is owner Aug 25 22:03:10 btw, he is translating OE's manual to russian Aug 25 22:03:21 look at docs.openembedded.ru Aug 25 22:05:46 hmm, is there any way to subscribe to a googlegroups ML fron a non-google/gmail account? Aug 25 22:06:12 denix: should be Aug 25 22:06:54 hrw|gone: you wrote the ruby 1.8.5 recipe, Jan 2007 :) Aug 25 22:11:08 denix: ask slapin to subscribe when you see him Aug 25 22:11:21 Jay7: ah, can only do it over the email, not on the web... need to send an email to +subscribe@googlegroups.com Aug 25 22:12:08 denix: hm.. I'm sure that I've send some letters in googlegroups from other domain.. Aug 25 22:12:29 ah, well Aug 25 22:14:49 all: I am somewhat how a noob here so please forgive me if I am asking dumb questions... Does anyone know how to change the toolchange for a angstrom build? Also has anyone been able to get GFX SDK up and running? Aug 25 22:15:52 what toolchain? what GFX? Aug 25 22:17:09 I want to use arm2008q3-72 (as far as the GFX, I should be able to get this working it I can build angstrom with this tool chain) Aug 25 22:20:14 * Kero is doing something weird: unpkacing v1.3.0 to create v.1.3.5, but now the do_* steps look in the package-1.3.5 directory, which does not exist (yet). Aug 25 22:20:38 should probably rename the dir before doing those steps... Aug 25 22:22:56 nevertheless, I need to know the name of that dir by all nice preset capital DIR variables... Aug 25 22:26:26 jeaneus: you'd need to use external-toolchain - there were some discussions lately here and on the mailing list Aug 25 22:26:41 too easy: ${S} Aug 25 22:28:07 denix: I tried setting an external-toolchain within my local.conf, but I am not sure how to set this. I also try to adjust the one in angstrom.2008.1.conf, but no luck Aug 25 22:28:08 jeaneus: or you can check existing projects using CSL as external-toolchain for some pointers, like arago-project.org (shameful plug :) Aug 25 22:28:47 denix: I just started to review that local.conf Aug 25 22:28:53 jeaneus: bah, you are in TI... so, why not Arago? Aug 25 22:29:24 jeaneus: but not everything from OE/Angstrom can be built with CSL... Aug 25 22:29:35 as of GFX, I think you meant SGX... Aug 25 22:31:14 denix:correct Aug 25 22:51:04 denix: thanks for the input.... Aug 25 22:56:57 jeaneus: no problem Aug 25 23:15:25 hmm, whats a single name for "configure, compile, install", that *isnt* "build" Aug 25 23:17:49 installing-autotools-enabled-package-from-sources? :) Aug 25 23:18:44 erk, i think i broke bitbake Aug 25 23:19:06 I'm working on builds where the recipe is inside a local scm checkout of some source tree, and it builds in that source tree Aug 25 23:19:11 but the shit hits the fan with native vs non-native Aug 25 23:19:14 since they share a ${S} Aug 25 23:19:21 configure/compile/install can interleave Aug 25 23:19:27 with BB_NUMBER_THREADS > 1 Aug 25 23:19:50 so i was thinking about trying toc hange the tasks around a bit Aug 25 23:20:08 * Crofton|work questions kergoth's sanity Aug 25 23:20:41 Crofton|work: http://dl.getdropbox.com/u/112715/Files/inscm-1.tar.bz2 .. i think its a worthwhile endeavor, if a pain in the ass. mind taking a look and giving an opinion? Aug 25 23:21:17 maybe after I mow the lawn some .... Aug 25 23:21:46 hehe, k Aug 25 23:23:20 ERROR: No method available to unpack /home/kergoth/code/inscm/components/quilt.lock (unsupported file type?) ... err. Aug 25 23:23:23 what the fuck Aug 25 23:24:39 where the hell is that coming from Aug 25 23:26:39 ooh, i think i may have worked around this interleaving issue Aug 25 23:27:11 its kind of twisted, though Aug 25 23:27:24 i sift through the metadata in an anonymous python function to basically implement "deltask" Aug 25 23:27:32 and transitions bits over to a different task.. Aug 25 23:27:37 all to avoid messing with base.bbclass Aug 25 23:28:26 * denix concurs with Crofton about kergoth's sanity... :) Aug 25 23:29:01 had to sift through and change any task depends and deps flags to point at the new task instead of any of the independent tasks, in order to merge these .. hmm Aug 25 23:29:04 * kergoth thinks Aug 25 23:29:32 the problem is, a lockfile is insufficient. could still get interleaving tasks touching ${S} Aug 25 23:29:37 * kergoth sighs Aug 25 23:29:52 i really like this style of project, i think app/kernel devs would love it, but jesus Aug 25 23:50:46 hmm Aug 25 23:50:53 i want an equivalent to 'world' which doesn't obey EXCLUDE_FROM_WORLD Aug 25 23:54:16 grr Aug 25 23:54:38 would be nice if exec_task checked stamp existence before it ran, but i guess the cooker is responsible for that Aug 25 23:54:39 hrm Aug 25 23:55:49 * kergoth shakes head, need to step back and rethink this Aug 26 00:03:14 * kergoth 's head explodes Aug 26 00:22:13 http://kergoth.pastey.net/123377 Aug 26 00:22:20 yikes. Aug 26 00:31:28 * kergoth is starting to think he's a crazy bastard Aug 26 00:37:11 hmm, this stuff would fit together well with that private staging stuff i was playing with Aug 26 00:46:20 !!!! that is nuts Aug 26 00:46:34 kergoth: what's the case where you need to build both blah and blah-native inscm? Aug 26 00:49:54 well, as an example, i have the OpenEmbedded shasum app sources in a git repository, and the native and target recipes both reside there. the classes they inherit set things up so that the builds occurs in that source tree Aug 26 00:50:50 if both the native and target recipe are going to be built, there's a danger of the tasks that touch ${S} interleaving Aug 26 00:50:56 yeah. Aug 26 00:51:02 so you could, say, theoretically anyway, have foo-native configure, then foo compile, bam Aug 26 00:52:05 * kergoth isn't yet completely convinced that this is the best avenue to pursue, but its the best he's managed to come up with such far to satisfy the app/kernel developer use cases better Aug 26 00:52:10 I guess you could take the easy way out and forbid native + inscm... since the dev cycle for non-native is longer, that's the one you want to shorten the most Aug 26 00:52:46 How do you clean up between non-native & native? Aug 26 00:52:47 yeah, that's true. or could split them. clone one copy of shasum.git with target, one with native, recipe wise Aug 26 00:53:20 thats a good point, right now it assumes you can re-run configure/compile/install/stage and have it take the new settings. might want to run do_clean, or some subset of it, prior to configure.. Aug 26 00:53:21 yeah but then you aren't really building in the scm... although it might be predictable enough that people could get used to it Aug 26 00:53:22 * kergoth shakes head Aug 26 00:53:38 (the cloning two copies, i mean) Aug 26 00:54:01 I guess the complication is your recipe is actually in the SCM, right? Aug 26 00:54:05 the annoying part there is, if it really is maintained in git or something, and you source control the recipes, you'd have to have multiple branches, one with each incarnation of the recipe, or something Aug 26 00:54:49 right now it is, but it doesn't necessarily have to be that way. i was also thinking that you could create, say, linux-external.bb, make linux-external inherit inscm, but point ${S} to an external location of a checkout/clone, and keep linux-external in a metadata repository Aug 26 00:55:16 that'd separate the recipe and sources, and still have the same dillemma with recipe variants like native/cross/sdk, though :) Aug 26 00:55:32 But that right there seems like the most common use case to me Aug 26 00:55:44 the package scm has a makefile.. the OE scm has the recipe Aug 26 00:55:56 given the way OpenEmbedded works today, -external recipes would be the best way to go, if they want to interact with us Aug 26 00:55:59 then it would sort of make sense that each recipe within a directory needs to have its own checkout directory Aug 26 00:56:07 but i think the other way is an interesting approach for people making their own collections/overlays Aug 26 00:56:21 its more like what folks like debian/rh do, like what android does Aug 26 00:56:29 Yeah it is interesting. Aug 26 00:56:30 regardless, the same classes will work for both Aug 26 00:56:43 So is the idea that I would add a source checkout to my COLLECTIONS? Aug 26 00:56:49 exactly Aug 26 00:56:59 and the .bb's in the checkout would inherit inscm, and it would Just Work Aug 26 00:57:03 yup Aug 26 00:57:08 and they could make changes, rebuild real fast Aug 26 00:57:10 low overhead Aug 26 00:57:20 yeah. Aug 26 00:57:27 i like that a lot. Aug 26 00:57:44 and of course, you can combine the two/three ways of doing things Aug 26 00:57:55 so they could use stock OpenEmbedded + some inscms for their apps or kernel Aug 26 00:58:29 i can't think of many cases where i'd need to build both -native and non-native in this way. but it does feel a bit hacky that it wouldn't work. Aug 26 00:58:57 well, if you look at what richard did for his virtualnative branch for poky, native can be automatic, no need to create a new recipe Aug 26 00:59:03 just set a variable to get a -native available for a given recipe Aug 26 00:59:20 in which case there's no real way to avoid it, if you happen to be using inscm for something that has -native Aug 26 00:59:27 i dunno how likely that is, but what the heck Aug 26 01:00:09 http://dl.getdropbox.com/u/112715/Files/inscm-1.tar.bz2 is the prototype, before i had to add more intricacies, but it gives you the idea of one possible project area using entirely inscm components Aug 26 01:00:11 It's almost like you need a way to specify a set of packages that need to build serially, but don't depend on each other Aug 26 01:00:19 yeah, pretty much Aug 26 01:00:34 these two cant build at the same time. a recipe-wide lockfile or something.. Aug 26 01:01:02 actually, you're right, that wouldnt be enough.. can't let the tasks interweave, it needs to build the -entire- recipe, then move on Aug 26 01:01:23 kind of defeats all the fanciness richard did with the taskdata/runqueue, at least for a small piece of it Aug 26 01:02:28 it's like every recipe could try to acquire a lock by default... ${PN}.lock, say... but for inscm stuff you'd have to pick some name for hte lockfile shared between blah and blah-native Aug 26 01:02:36 but i guess that's sort of what you did on your pastebin :) Aug 26 01:02:58 yeah, sort of. i tried to wrap up everything that touched ${S} into a single task, to make that portion of it serial Aug 26 01:03:05 the packaging stuff is free to run whenever, its in WORKDIR, not S Aug 26 01:03:49 ah well, this is disgusting, but it works. guess i should try adding upstream OpenEmbedded to my collections and make sure nothing explodes Aug 26 01:06:00 yeah actually if it's just a special case for blah and blah-native inscm, it seems pretty sane Aug 26 01:06:15 inherit aiiieee :P Aug 26 01:06:19 hehe Aug 26 01:06:44 i dislike that this 'build' task is named do_populate_staging, since it does more than that, but there are cross-recipe deps on do_populate_staging executions, so i need to provide one to satisfy that Aug 26 01:16:33 what's 'inscm'? Aug 26 01:17:12 prototype implementation for doing builds within source trees. no fetch/unpack/etc, just check out your project from your scm of choice, with the recipe in the repository, and add that to your collections/bbfiles, and go Aug 26 01:17:43 do_clean just does a make clean + some metadata, no re-unpacking/patching for a rebuild Aug 26 01:17:56 ah. very cool Aug 26 01:18:23 the goal is to improve functionality for people doing active development on applications & kernels, speed up the build/test process Aug 26 01:18:43 right now OpenEmbedded is awesome for build/distro folk, but for devs its pretty weak Aug 26 01:19:21 sadly, getting it workign is a royal pain in the ass Aug 26 01:19:21 its kind of annoying to try to fix a bug and/or add features to an existing application that has an existing recipe Aug 26 01:19:29 yeah, exactly Aug 26 01:19:53 so this would fix that, so long as you have it in an SCM Aug 26 01:20:29 yep. you could keep the actual recipe in a regular OE repository if you wanted, but the key is that the entire current source tree is available locally, and it builds there Aug 26 01:20:48 the scm specific bits are some bits i added to do things like extract PV from the git repository info Aug 26 01:20:52 but those are optional Aug 26 01:20:59 and we need to add that stuff for other scms Aug 26 01:21:08 well, i do, if i plan on getting this thing upstream :) Aug 26 01:34:26 hmm Aug 26 01:35:10 hey zecke Aug 26 01:39:17 woo, works awesome together with more traditional stuff from the OpenEmbedded repo, and packaging didn't get broken... hmm Aug 26 01:39:27 guess now i clean it up, rename it, and work on more version extraction crap Aug 26 01:39:48 also add more convenience functions for do_clean, and maybe add a var to control which of them are run instead of overriding do_clean Aug 26 01:39:51 hmm Aug 26 01:52:44 * kergoth ponders Aug 26 01:59:59 kergoth: morning Aug 26 02:00:24 zecke, gm Aug 26 02:02:52 hmm, i wonder if preferences interact properly with BBCLASSEXTEND Aug 26 02:02:57 anyone tested this? Aug 26 02:03:19 if quilt.bb has native in BBCLASSEXTEND, and there's a real actual quilt-native.bb in another lower priority collection, it seems to be preferring the latter Aug 26 02:03:26 which doesn't seem correct to me Aug 26 02:03:27 hmm Aug 26 02:06:26 hmm, i wonder how to manage do_clean here Aug 26 02:07:33 maybe its best if i move the stamps into the source tree, but then... hrm Aug 26 02:07:51 maybe i should move WORKDIR back into the source tree.. i was going to avoid that, but.. hrmph Aug 26 02:23:53 * kergoth wonders what the best interface for interacting with the version extraction routines would be Aug 26 02:24:24 any opinions on that? Aug 26 02:24:25 hmm Aug 26 02:28:24 oh.. i know, i'll just make variables like SVNREV, SVNPATH, GITDESC, GITREV, etc, and let the user construct their PV with them Aug 26 02:29:40 yeah i like that... i don't know if you need to bother with branch names since especially in git they tend to have unwelcome characters Aug 26 02:31:59 wonder if inscm could be a BBCLASSEXTEND class.. hmm Aug 26 02:32:49 guess it could, but if you used that instead of manually doing it, you'd be stuck with the default version.. guess that works Aug 26 02:34:14 what's a better name for this, insource? insrctree? externalsourcetree? inscm isn't really appropriate, since you don't *have* to use scm, a local tree would work as well, just with no magic version extraction Aug 26 02:41:36 i like insrctree Aug 26 02:41:44 or insourcetree if you like vowels Aug 26 02:46:49 the bb python module could really use a Popen() wrapper that sets up the env for the child process to contain all exported vars from the metadata Aug 26 02:47:03 could replace a lot of hacky spots in the classes **** ENDING LOGGING AT Wed Aug 26 02:59:57 2009