**** BEGIN LOGGING AT Fri Oct 20 02:59:58 2006 Oct 20 03:16:55 where are the -I overrides set in OE? Oct 20 03:35:08 wow. it worked. i think the ppc toolchain is up. Oct 20 04:00:59 RP: btw have you checked out http://wiki.python.org/moin/PythonSpeed/PerformanceTips Oct 20 05:06:41 hey who is giel? Oct 20 05:06:59 why giel as author used (not a proper address) Oct 20 05:07:13 and why does he check-in/add obviously broken and bad stuff? Oct 20 05:08:51 i'm giel, i didn't check in anything, i submitted a .bb to bugzilla for review, it worked for me, and mickey checked it in Oct 20 05:09:06 how/why my name got added during checkin i don't know Oct 20 05:09:50 and why was that a hit-n-run! Oct 20 05:09:51 i hate that Oct 20 06:23:32 morning Oct 20 06:23:37 ~curse koen Oct 20 06:23:44 May you be reincarnated as a Windows XP administrator, koen ! Oct 20 06:30:35 hrw: hi :) I am from jmirc E70 now - so interesting test type from phone qwerty keyboard and look into 3~ -4~ screen. Oct 20 06:30:40 hi hrw Oct 20 06:31:03 * Marex wonders why oe needs ncurses when building bootstrap-image Oct 20 06:31:20 den-ros: ;) Oct 20 06:31:25 Marex: something needs it Oct 20 06:31:40 hrw, dunno what though ;) Oct 20 06:33:06 Marex: bitbake --dry-run --verbose bootstrap-image Oct 20 06:39:29 03hrw 07org.oe.dev * r2754e872... 10/ (4 files in 2 dirs): Oct 20 06:39:29 glibc: re-added 2.4 with DEFAULT_PREFERENCE = "-1" set for ARM Oct 20 06:39:29 - it would be nice if developers will *check* before commiting does Oct 20 06:39:29 changeset break other configurations then their one Oct 20 06:41:06 Marex: found? Oct 20 06:41:23 hrw, nah ... I'll try that later Oct 20 06:41:29 craps ... Oct 20 06:41:35 it screwed again :( Oct 20 06:44:34 NOTE: current path: bootstrap-image (bootstrap-image) -> fakeroot-native (fakeroot-native) -> util-linux (util-linux) -> ncurses (ncurses) -> ncurses-native (ncurses-native) Oct 20 06:44:47 so util-linux need ncurses Oct 20 06:45:04 ah :) Oct 20 06:45:19 | pivot_root.c: In function 'main': Oct 20 06:45:20 | pivot_root.c:27: error: 'SYS_pivot_root' undeclared (first use in this function) Oct 20 06:45:20 | pivot_root.c:27: error: (Each undeclared identifier is reported only once Oct 20 06:45:20 | pivot_root.c:27: error: for each function it appears in.) Oct 20 06:45:20 | make[1]: *** [pivot_root] Error 1 Oct 20 06:45:20 | make[1]: Leaving directory `/stuff/build/tmp/work/armv4t-linux-uclibc/util-linux-2.12r-r8/util-linux-2.12r/mount' Oct 20 06:45:23 | make: *** [all] Error 1 Oct 20 06:45:25 | FATAL: oe_runmake failed Oct 20 06:45:27 dang ... I have to google this :S Oct 20 06:45:36 Marex: PASTEBIN Oct 20 06:45:42 ~pastebin Oct 20 06:45:45 from memory, pastebin is a place to paste your stuff without flooding the channel - try http://pastebin.ca, or http://channels.debian.net/paste Oct 20 06:45:48 oops Oct 20 06:45:49 :x Oct 20 06:45:54 sorry Oct 20 06:49:21 fixed :) Oct 20 06:49:34 replace SYS_pivot_root with __NR_pivot_root Oct 20 06:50:42 Marex: send us patch? Oct 20 06:51:29 hrw, if I knew how exactly to make patch for OE Oct 20 06:51:39 hrw, I can tell you where is the prob Oct 20 06:55:05 util-linux-2.12r/munt/pivot_root.c line 8 Oct 20 06:55:14 replace SYS_pivot_root with __NR_pivot_root Oct 20 06:55:31 but I really dunno if it wont break something Oct 20 06:55:33 Marex: quilt new fix-pivotroot.patch;quilt add util-linux-2.12r/munt/pivot_root.c; edit util-linux-2.12r/munt/pivot_root.c;quilt refresh Oct 20 06:55:42 then send us patches/fix-pivotroot.patch file Oct 20 07:00:23 hello Oct 20 07:00:56 CoreDump|afk: the problem with gnome-vfs-dbus is that we use version for dbus 0.23 and we did not updated it with updating to newer dbus Oct 20 07:01:01 hi Genesis Oct 20 07:01:43 CoreDump|afk: we would need to move to gnome-vfs-dbus_svn with working SRCDATE - probably Florian need to be pestered about it. Oct 20 07:01:46 hrw, can you accept dcc? Oct 20 07:01:49 over irc Oct 20 07:01:49 cu all - time to go work Oct 20 07:01:57 Marex: no - we accept bugtracker Oct 20 07:02:02 ah Oct 20 07:16:16 pb_ are you there? Oct 20 07:26:40 morning Oct 20 07:30:51 morning Oct 20 07:32:29 hrw|work: I was under the impressions gnome-vfs-dbus_svn was the default Oct 20 07:34:03 XorA: not in .oz354x Oct 20 07:34:52 hrw|work: you on dbus 0.9x? Oct 20 07:37:03 actually yes Oct 20 07:37:20 feeds are not completly updated yet Oct 20 07:38:34 03koen 07org.oe.dev * ra72d3fe6... 10/ (1 packages/glibc/glibc_2.4.bb): Oct 20 07:38:34 glibc 2.4: start whitelisting compatible hosts with COMPATIBLE_MACHINE Oct 20 07:38:34 * default_preference is too weak to mask out broken archs Oct 20 07:40:19 hrw|work: make sure dbus-api-change.patch is in oz354x then Oct 20 07:40:28 hrw|work: otherwise svn wont compile Oct 20 07:47:22 hrw|work: and svn20060119 is a working SRCDATE :-) Oct 20 07:51:45 ok Oct 20 08:01:59 good morning all Oct 20 08:03:27 hi koen Oct 20 08:08:10 hey koen Oct 20 08:11:27 Crofton: I will push one small change to u-boot. On my systems it fails on building examples so I removed one example from makefile Oct 20 08:12:07 hm.. it even is in repo.. Oct 20 08:12:39 or no.. it other ver of my patch Oct 20 08:12:59 ~lart modutils Oct 20 08:12:59 * ibot executes killall -HUP modutils Oct 20 08:14:34 * XorA wodners the best weay to fix this mess Oct 20 08:16:30 hi Oct 20 08:18:06 hey nicolasfr Oct 20 08:18:24 koen: have you seen the bug I have submitted about glic2.5 and pthread_kill? Oct 20 08:18:25 nicolasfr: the slowness you were seeing is probably because dbus times out Oct 20 08:18:39 nicolasfr: I've seen it, and I have no idea about that Oct 20 08:19:39 dbus times out? It means I should upgrade dbus? Oct 20 08:20:55 I doubt that would help, but logging in as root seems to fix it Oct 20 08:21:31 03hrw 07org.oe.dev * rcaeaa1c1... 10/ (4 files in 3 dirs): u-boot: do not build helloworld example Oct 20 08:22:58 and about pthread_kill may that be related to using EABI? Oct 20 08:23:06 ~botmail for zecke: 'mtn automate certs ' will show you who committed something when using --author Oct 20 08:23:24 nicolasfr: I suspect we're not installed all headers Oct 20 08:23:50 installing* Oct 20 08:27:26 koen: I have grepped for pthread_kill in the headers in the work dir, but could find any matching .h file, while .so files match ... Oct 20 08:28:08 koen: hum, correction: sygthread.h matches! That may be why Oct 20 08:29:36 koen: thanks, I think I will need to disapprove 497af8d916ea71b5cc3ad879076fb96fd89b3ef9 Oct 20 08:31:37 zecke: or discuss it with mickey|zzZZzz and giel Oct 20 08:32:36 why to disapprove? if someone will maintain it... Oct 20 08:33:06 oe gdb seems to be incapable of running on threaded apps :( Oct 20 08:33:16 morning all Oct 20 08:33:18 http://www.rpsys.net/openzaurus/temp/poky_fixdebianbbclass.patch is my proposal for solving debian.bbclass circular depends Oct 20 08:33:20 hey RP Oct 20 08:33:24 hey Oct 20 08:34:10 | ../src/.libs/libasound.so: undefined reference to `pthread_create' Oct 20 08:34:10 | ../src/.libs/libasound.so: undefined reference to `pthread_mutex_trylock' Oct 20 08:34:10 | ../src/.libs/libasound.so: undefined reference to `pthread_join' Oct 20 08:34:14 applies to poky atm but something in a similar style should work for oe as well. It also cleans up package.bbclass a bit. I should redo it in stages I guess - cleanup, then the actual fixes Oct 20 08:34:21 it just dies with Program terminated with signal SIGTRAP, Trace/breakpoint trap. Oct 20 08:34:23 while compiling alsa lib ... any idea ? :( Oct 20 08:34:42 which (so I have read) indicates that the threadded support is busted Oct 20 08:39:24 anyone seen this before with the oe built gdb? Oct 20 08:41:04 RP: that diff is a bit hard to read since it also reorders methods Oct 20 08:46:22 oh ha mickeyl comitted that crap :} Oct 20 08:46:41 koen: I know. I'll have to redo it. I had to reorder things as I couldn't make any sense of it before doing so :-/ Oct 20 08:47:53 RP: It just takes a bit more coffee to read that diff :) Oct 20 08:48:14 koen: :) Oct 20 08:49:06 There is a lot of code in there screaming for a rewrite Oct 20 08:50:13 OE-ng :) Oct 20 08:50:30 ~hate git Oct 20 08:51:20 RP: would it be possible to 'upgrade' the pkgmaps to something packaged-staging can use? Oct 20 08:52:54 Marex: are you using glibc-2.5? Oct 20 08:53:27 zecke: thanks for calling my efforts crap ;) Oct 20 08:53:44 I'm using uclibc Oct 20 08:53:51 nicolasfr|aw, Oct 20 08:53:55 zecke: i'd be happy to know what's wrong with it so i can give a shot at improving it Oct 20 08:55:07 nicolasfr|aw, I built it with threads, I found that function in libpthread too, but it looks like linked cant find it :S Oct 20 08:55:55 koen: yes, I hope so Oct 20 08:57:42 giel: hey :) Oct 20 09:00:14 ~curse git downloading Oct 20 09:00:16 May the fleas of a thousand camels infest your most sensitive regions, git downloading ! Oct 20 09:03:16 giel: stuff like checking your qmake spec in under qte/qtopia-core Oct 20 09:03:30 giel: and not under qmake. This is obviously broken by design and should never have been comitted Oct 20 09:03:39 QT_ARCH = "arm" Oct 20 09:03:39 QT_ENDIAN = "-little-endian" Oct 20 09:03:43 hello? Oct 20 09:03:58 wtf is that. a) look at the existing qte files which are doing these things just fine Oct 20 09:04:10 and b) check OE how endianes is determined! Oct 20 09:04:16 craptopia-core Oct 20 09:04:45 giel: c) installing the build version of qmake in ${STAGING_BINDIR} overwriting qmake-native Oct 20 09:04:55 WTF this is so bad I'm disapproving it now Oct 20 09:05:32 i'm fine with that Oct 20 09:05:37 http://www.openembedded.org/user-manual&dpage=ch07 <- endianness check Oct 20 09:05:52 i never really meant it to be checked in anyway, i just submitted it to bugzilla to get comments like this Oct 20 09:06:02 giel: actually I don't get why mickeyl pushed it Oct 20 09:06:20 i just needed a working version for my situation, and i thought, i'll show it to others so i can go from there Oct 20 09:06:43 giel: hehe, that is okay. It is not in a quality to be generally used Oct 20 09:06:56 i thought that it was obvious from the comments in the .bb that i don't have a clue what i'm doing for some things Oct 20 09:07:00 giel: did you need any other patches? Oct 20 09:07:39 nope Oct 20 09:07:39 koen: take a look at the tslib plugin for Qtopia Oct 20 09:07:39 zecke: i got a bit confused by the qmake-in-configure vs. external qmake, i wanted to figure that out, but i first needed it just working Oct 20 09:07:39 koen: they do their own linearisation, they have no clue how tslib works :} Oct 20 09:08:30 giel: this is what schocked me :) Oct 20 09:08:33 koen: thanks for the link on endianness Oct 20 09:08:56 zecke: but i certainly understand your disapproval, but hey, i got to start somewhere Oct 20 09:09:34 zecke: i will try to improve the .bb wrt the points you mentioned somewhere next week Oct 20 09:11:49 good morning Oct 20 09:11:49 giel: nice :) Oct 20 09:11:53 hi florian_kc Oct 20 09:23:50 koen: there's indeed a problem with the headers of glibc-2.5. The file "ports/sysdeps/unix/sysv/linux/arm/bits/sigthread.h" which gets staged has not declaration for pthread_kill. While "glibc-2.5/nptl/sysdeps/pthread/bits/sigthread.h" has it. I just copied the file manually and it seems to works fine (compiling mysql ATM). Oct 20 09:25:16 nicolasfr|aw: could you make a patch to the glibc_2.5.bb for that? Oct 20 09:25:22 koen: ok it compiled fine now, that was the problem Oct 20 09:25:33 koen: ok, I can try Oct 20 09:27:44 oops firefox is crashing on angstrom (illegal instruction). no luck :/ Oct 20 09:33:39 giel: ping :) Oct 20 09:33:45 zecke: yes? Oct 20 09:34:49 giel: do you work for phillips? Oct 20 09:34:54 03xora 07org.oe.dev * r3620468c... 10/ (1 packages/linux/linux-smdk2440_2.6.18.bb): linux-smdk2440_2.6.18.bb : add compatible machine Oct 20 09:34:59 zecke: yes Oct 20 09:35:05 -l Oct 20 09:35:07 philips Oct 20 09:35:22 giel: still DVB-H? Oct 20 09:35:39 zecke: mainly Oct 20 09:35:51 zecke: wieso? Oct 20 09:36:17 giel: hehe, I wonder if I should not have bitched about Qtopia :} Oct 20 09:36:32 zecke: oh, that's okay, i can handle criticism Oct 20 09:36:38 zecke: it's the only way to learn :) Oct 20 09:37:09 does 2.6 kernel have capability for different keymaps for differnt keyboards at same time, ie, Zaurus built in and USB Oct 20 09:38:55 koen: http://pastebin.ca/raw/211451 should do the trick Oct 20 09:41:26 * koen tries Oct 20 09:42:21 koen: oops, I copied the wrong file Oct 20 09:42:45 shoudl be nptl/sysdeps/pthread/bits/sigthread.h instead of bits/sigthread.h Oct 20 09:43:38 bits/sigthread.h is the same as the one in arm dir, while nptl/... has pthread_kill declaration Oct 20 09:45:58 http://pastebin.ca/raw/211457 Oct 20 09:46:07 Is oesources.org down? Oct 20 09:46:13 realloc: yes Oct 20 09:46:29 hrw|work: for ever? Oct 20 09:47:45 zecke: could you remind me how bonsai fits into the tinderbox again? Oct 20 09:49:43 XorA: No, not that I'm aware of. Its one of the big limitations of the kernel keymaps Oct 20 09:50:08 koen: hehe Oct 20 09:50:19 koen: |<--- here --->| Oct 20 09:50:29 zecke: I'm going to add poky svn Oct 20 09:50:36 RP: damn Oct 20 09:50:45 koen: wait a second Oct 20 09:51:21 XorA: Another reason the whole things needs rewriting but its in core code nobody dares touch :-/ Oct 20 09:52:07 zecke: http://rafb.net/paste/results/b51e7Z98.html Oct 20 09:52:23 morning folks Oct 20 09:52:26 hey ade|desk Oct 20 09:52:46 XorA: libusb + uinput Oct 20 09:54:40 RP: once upon a time the input guys tried, they were shot down by the x86 guys who just couldnt understand having 2 keyboards Oct 20 09:55:22 koen: an example somewhere? Oct 20 09:55:29 XorA: Hopefully that viewpoint has changed now but it will take effort :-/ Oct 20 09:55:53 XorA: no, but that solution would be a bit backwards Oct 20 09:56:13 realloc: we will create new source mirror(s) Oct 20 09:57:01 Has anyone tried -c fetchall yet? :) Oct 20 09:57:16 koen: how it fits into the tinderbox? Oct 20 09:57:24 koen: BTW: the cmdline looks okay Oct 20 09:58:00 zecke: sideways is another obvious answer to how it fits :) Oct 20 09:58:23 RP: trunk only? Oct 20 09:59:06 koen: the obvious thing is: What changed between this build and the last one Oct 20 09:59:23 hrw|work: I'm afraid so. It uses new freatures in bitbake Oct 20 09:59:27 koen: it can link bugzilla and ci's together Oct 20 09:59:51 koen: but is pysvn working again? Oct 20 09:59:51 ok Oct 20 10:00:12 koen: and at one point it will be able to show all ci's for package ${PN} Oct 20 10:00:22 hrw@bitbake:~/devel/build/dev$ time bitbake -cfetchall gpe-image Oct 20 10:00:26 koen: to the OE tree and the tree where ${PN} is in ;) Oct 20 10:02:22 If we make the fetchers always touch the md5 stamps, we should have enough code for a source mirror apart from the script to run it Oct 20 10:02:47 zecke: pysvn is working again Oct 20 10:05:36 zecke: bonsai is adding a '/' somewhere: http://svn.berlios.de//svnroot/repos/bitbake/ starting at rev:618 Oct 20 10:05:45 and svn://svn.o-hand.com//repos/poky/ Oct 20 10:06:13 koen: don't write /repos Oct 20 10:06:17 koen: write repos :) Oct 20 10:06:22 koen: or change bonsai_update.py Oct 20 10:06:39 morning all Oct 20 10:06:45 hey lrg Oct 20 10:07:41 should I use glibc-2.5 or 2.4? Oct 20 10:08:57 RP: did you see that pangocairo->pangoxft patch on gnome-performance yesterday? Oct 20 10:10:12 zecke: how can I remove a scms from bonsai? Oct 20 10:10:17 zecke: or update it? Oct 20 10:10:42 psql Oct 20 10:11:01 RP: looks like fetchall works Oct 20 10:11:05 Crofton: arm - 2.5 Oct 20 10:11:29 I copied from angstrom last night after the disappering 2.4 .... Oct 20 10:11:30 RP: I still want a different layout for the source/distfiles Oct 20 10:11:36 rebuilt over night :) Oct 20 10:11:47 Crofton: I don't think you won't be able to build 2.4 for arm Oct 20 10:12:04 actually, it was doing some stuff on OSK Oct 20 10:12:25 so I was pretty confused to hear it didn't work Oct 20 10:12:26 zecke: I'm in psql, now what? Oct 20 10:12:42 koen: delete from sometable where id=someid Oct 20 10:13:01 koen: let us coordinate :) Oct 20 10:13:18 koen: what do you want to do? I have changed bonsai_update.py to not add an additional '/' Oct 20 10:13:28 zecke: run bonsai update "Poky SVN" and you'll see what I mean Oct 20 10:14:10 zecke: for some reason it times out, but manually co'ing it works Oct 20 10:14:21 koen: it is not doing co Oct 20 10:14:25 koen: it is doing 'log' Oct 20 10:14:57 koen: but the reason might be simple :) Oct 20 10:17:55 koen: svn log, info all fail Oct 20 10:19:23 "svn log http://svn.o-hand.com/repos/poky/trunk poky" fails, but "svn log http://svn.o-hand.com/repos/poky/trunk" works Oct 20 10:20:45 zecke: "svn log http://svn.o-hand.com/repos/poky/" works as well Oct 20 10:21:12 zecke: a different layout? Oct 20 10:21:26 koen: I did see it, yes Oct 20 10:21:58 hrw|work: excellent :) Oct 20 10:22:15 ~lart gimp Oct 20 10:22:16 * ibot slams gimp against a large cement Tux Oct 20 10:22:34 koen: so http ;) Oct 20 10:22:58 koen: then don't use subversion protocol Oct 20 10:23:13 D'oh Oct 20 10:25:02 koen: I'm writing a bonsai_delete now Oct 20 10:28:09 hey zecke Oct 20 10:28:47 and others Oct 20 10:28:57 koen: Poky is removed :) Oct 20 10:29:10 * zecke wonders who comitted the most to poky Oct 20 10:29:13 RP/zecke: btw have you checked out http://wiki.python.org/moin/PythonSpeed/PerformanceTips Oct 20 10:29:23 zecke: can I add it again? Oct 20 10:29:30 mithro: I had looked at it once :) Oct 20 10:29:36 koen: sure, now use -p "http" please Oct 20 10:29:58 oe@serenity:~/tinderbox-server/tinderbox$ ./bonsai_update.py "Poky SVN"Updating SVN from http://svn.o-hand.com/repos/poky/ starting at rev:0Got 806 possible updates Oct 20 10:30:17 http://www.openembedded.org/bonsai/view/scm/6/ Oct 20 10:30:42 koen: now I need to add the author map Oct 20 10:30:50 koen: I have no idea how to make that go 'fast' Oct 20 10:31:06 zecke: I'd prefer 'accurate' over 'fast' :) Oct 20 10:31:38 mithro: Some good tips there, some I've heard before :) Oct 20 10:31:42 koen: fast as in rapid development Oct 20 10:31:55 RP: yes, its a good start Oct 20 10:32:27 zecke: http://www.openembedded.org/bonsai/query/author/kergoth/ already does the right thing :) Oct 20 10:32:59 koen: yes but richard fails :( Oct 20 10:35:10 I wouldn't worry too much Richard hasn't contributed much of any importance ;) Oct 20 10:36:37 hi lrg Oct 20 10:36:37 hi zecke Oct 20 10:36:47 hey pb_ Oct 20 10:36:49 hi florian_kc Oct 20 10:37:40 hi pb_ Oct 20 10:39:48 mithro: I agree, but it still would be nice to show where all these errors/bugs are from ;) Oct 20 10:40:03 mickey|zzZZzz, you around? Oct 20 10:46:06 hi pb_ Oct 20 10:59:13 zecke: you could implement an amazon style bonsai Oct 20 10:59:28 zecke: "developer that modified this file also modified:" Oct 20 10:59:33 developers* Oct 20 11:01:41 koen: I think that algorithm is patented Oct 20 11:01:58 koen: but I will certainly have to do better statistics Oct 20 11:02:17 koen: Koen comitted to X SCMs, total number of lines added 2 Oct 20 11:11:07 good morning Oct 20 11:11:13 mickeyl: hey :) Oct 20 11:13:36 gm mickeyl Oct 20 11:13:51 any python updates, or do I need to work on something else today? Oct 20 11:13:58 well Oct 20 11:14:05 last night i fixed sdcc (nonnative) Oct 20 11:14:07 took me 1h Oct 20 11:14:11 all these crazy direct pathsh Oct 20 11:14:32 and then I found out, why python-native doesn't honor my --prefix changes Oct 20 11:14:39 heh Oct 20 11:14:44 sounds like progress Oct 20 11:14:46 because autotools supplies --prefix as well Oct 20 11:14:57 i'm now doing a build where --prefix is _really_ changed :) Oct 20 11:15:39 when this works, you're in the game again :) Oct 20 11:15:44 thanks Oct 20 11:16:08 I have a line "prefix=/home/sca" which seems to override autotools Oct 20 11:16:17 yep Oct 20 11:16:51 http://www.digilentinc.com/Products/Detail.cfm?Nav1=Products&Nav2=Programmable&Prod=XUPV2P Oct 20 11:17:10 I need someone to pay me to get OE to build images for the PPC's on this board :) Oct 20 11:20:08 ~seen treke Oct 20 11:20:18 treke was last seen on IRC in channel #oe, 34d 15h 30m 55s ago, saying: 'I wouldn't be up on that if I was working in a morgue :p'. Oct 20 11:20:39 doh 34d Oct 20 11:21:03 anyone know whats up with oesources.org ? Oct 20 11:21:15 ade|desk: orphaned :) Oct 20 11:21:19 gone for good Oct 20 11:21:55 any new mirrors ? Oct 20 11:22:05 hi rp Oct 20 11:22:06 hi mickeyl Oct 20 11:22:13 pb_: master! Oct 20 11:22:44 ade|desk: try ~seen ggilbert Oct 20 11:23:09 ade|desk: greg asked us if he should renew the domain name, we declined that offer Oct 20 11:24:03 why ? Oct 20 11:26:24 ade|desk: it was unmaintained Oct 20 11:26:36 ade|desk: and in the future we would like to have it at sources.openembedded.org Oct 20 11:31:40 cheers pb_ Oct 20 11:31:52 ~lart sd cards Oct 20 11:31:53 * ibot says "boot to the head" and knocks sd cards over Oct 20 11:32:07 mickeyl, any clue when the python stuff will get committed? Oct 20 11:32:37 I have a demo for some potential research sponsor at 11 (our time) :) Oct 20 11:32:47 "My EBN todo list is topped by "bring in test coverage information" which is something Holger Freyther showed me." Oct 20 11:33:04 Crofton: just finished one more test. commit in a couple of seconds Oct 20 11:33:08 thanks Oct 20 11:33:16 Crofton: at least it finds python.h now. the rest is your business ;) Oct 20 11:33:20 hehe Oct 20 11:33:33 last time I rushed to demo the OSK, I wiped u-boot Oct 20 11:34:58 koen: hehe, I have used the mtn gcov stuff Oct 20 11:35:12 koen: and made it work with build dir != src dir and KDEs directory structure Oct 20 11:35:54 03mickeyl 07org.oe.dev * r6c3c55a6... 10/ (4 files in 3 dirs): sdcc: add native-tools patch to make it build Oct 20 11:35:59 03mickeyl 07org.oe.dev * ref3a5a7f... 10/ (16 files in 4 dirs): qemu: sync with poky, remove obsolete ones Oct 20 11:36:07 03mickeyl 07org.oe.dev * rada72513... 10/ (1 packages/python/python-native_2.4.0.bb): Oct 20 11:36:07 python-native 2.4.0 compile with --prefix same as --exec-prefix to fix some rare cases of distutils builds Oct 20 11:36:07 let's see if this breaks anything... Oct 20 11:36:09 03mickeyl 07org.oe.dev * rba012b73... 10/ (1 packages/python/python-native_2.4.0.bb): python-native 2.4.0 this time, the --prefix change is actually effective Oct 20 11:36:14 03mickeyl 07org.oe.dev * r510080de... 10/ (3 files in 2 dirs): add libsdl-native 1.2.11 - this one compiles with gcc4 Oct 20 11:36:19 03mickeyl 07org.oe.dev * r296ff29b... 10/ (4 files in 2 dirs): qemu-native: depends on sdl-native now, add more metadata Oct 20 11:41:46 mickeyl, did usrp actually build? Oct 20 11:42:15 NOTE: package glibc-2.5-r2: task do_populate_staging: started Oct 20 11:42:19 finally Oct 20 11:42:49 koen what about 2.4 sucked on the arm? Oct 20 11:42:57 I have seen 2.4 run on the OSK Oct 20 11:43:06 Crofton: IPC and shm were broken Oct 20 11:43:12 ah Oct 20 11:43:14 Crofton: working on that... Oct 20 11:43:22 ok, didn't get around to testing that Oct 20 11:43:23 it needs boost Oct 20 11:43:27 adding that to DEPENDS Oct 20 11:43:29 ooops Oct 20 11:43:36 I forgot that detail Oct 20 11:43:58 hmmm Oct 20 11:44:11 I wonder if this is a new dependency Oct 20 11:45:26 03mickeyl 07org.oe.dev * r012a3e43... 10/ (1 packages/qemu/qemu-native.inc): qemu-native: install binaries into staging (again) Oct 20 11:45:44 RP: oddly enough, even your dapper qemu binary exhibited the same hang on my machine :/ Oct 20 11:45:51 RP: I'm a bit at a loss now Oct 20 11:47:17 could I ask someone to update eds-dbus_svn.bb in OE and sync with poky? (or should I fill a bug for this?) Oct 20 11:48:06 nicolasfr|aw: and dates, contacts, mail and web :) Oct 20 11:48:17 koen: right!! Oct 20 11:48:32 and also add update-alternatives ;) Oct 20 11:48:43 and usbinit ... Oct 20 11:48:47 nicolasfr|aw: u-a is already in OE Oct 20 11:49:09 oh, was it added recently? Oct 20 11:49:09 nicolasfr|aw: the u-a in poky needs to be checked if it does the right thing Oct 20 11:49:24 ok Oct 20 11:49:29 nicolasfr|aw: no, it has been in ipkg and apt for ages, poky just factores it out :) Oct 20 11:54:44 03mickeyl 07org.oe.dev * r5b49fef6... 10/ (4 files in 2 dirs): qte/qtopia-core: factor out qte_arch function into qte-functions.inc, use this from qtopia-core to get QT_ARCH Oct 20 11:57:20 mickeyl: Does qemu actually lock up? Oct 20 11:57:56 mickeyl: Ctrl+Alt+3 should show a serial terminal and give another way of talking to the image Oct 20 11:58:29 RP: the emulation locks up. qemu still recognizes me giving control to the window (it changes the titlebar label), but even keyboard input is no longer reflected on the screen Oct 20 11:59:26 mickeyl: Its very strange. I've never had it reported or seen it... Oct 20 11:59:32 (before) Oct 20 11:59:56 if(mandriva){sleep(888888888);} Oct 20 11:59:59 RP: nevermind, i'll try on a different machine Oct 20 12:00:37 koen: put there by a redhat engineer fedup of mandriva ripoffs :-) Oct 20 12:01:13 speaking of that Oct 20 12:01:22 * koen applies for OECE Oct 20 12:01:38 OECE? Oct 20 12:01:47 OpenEmbedded Certified Engineer Oct 20 12:02:14 +# prefix = "${STAGING_DIR}/${BUILD_SYS}" Oct 20 12:02:14 +inherit qemu-native.inc Oct 20 12:02:15 -require qemu-native.inc Oct 20 12:02:36 inherit? Oct 20 12:03:02 koen: dont give that mickeyl dude one of those, he keeps checking in strange python stuff Oct 20 12:04:53 XorA: mickeyl pushes strange in general ;) Oct 20 12:05:31 Crofton: configure runs through, it bails out during compile now with Oct 20 12:05:34 linux/compiler.h not found Oct 20 12:05:57 03mickeyl 07org.oe.dev * r70b85757... 10/ (1 packages/sdcc/sdcc-native_2.5.0.bb): sdcc-native: autotools_stage_all doesn't work w/ sdcc, do_stage manually Oct 20 12:06:32 hmmm Oct 20 12:07:12 did you push the boost dependency Oct 20 12:07:50 should appear in a sec. Oct 20 12:07:50 i'm stopping to work on it now, i have a Oct 20 12:07:53 ****load of other things to do :) Oct 20 12:08:18 thats fine Oct 20 12:08:18 thanks for your help! Oct 20 12:08:29 np Oct 20 12:08:29 03mickeyl 07org.oe.dev * r365d8f13... 10/ (1 packages/usrp/usrp_0.12.bb): usrp: depends on boost Oct 20 12:08:35 it's easy to drag me in, if you say python don't work ;) Oct 20 12:08:45 doesn't, even Oct 20 12:10:58 compiler.h is pretty strange .... Oct 20 12:11:06 #define likely(x) __builtin_expect((x),1) Oct 20 12:11:06 #define unlikely(x) __builtin_expect((x),0) Oct 20 12:16:19 http://www.openembedded.org/bonsai/view/scm/6/ - now with a bit of css changed Oct 20 12:16:28 RP: progress! with init=/bin/sh i can get to a login prompt. Oct 20 12:17:04 I wish the usrp guys had some motivation to make things easy for embedded guys .... Oct 20 12:17:04 hrw|work: webmaster! :D Oct 20 12:17:28 RP: can i configure the qemu framebuffer size/resolution? Oct 20 12:18:10 hrw|work: cool Oct 20 12:18:13 mickeyl: just run fbset -xres 640 -yres 480 -n Oct 20 12:18:32 RP: excellent. thanks. Oct 20 12:18:35 hey, i'm in X now Oct 20 12:18:45 guess the computer was just as tired yesterday nite as i was Oct 20 12:18:53 mickeyl: yay. What was the problem? Oct 20 12:19:01 fbset -xres super-secret-xres -yres supersecret-yres -n Oct 20 12:19:31 RP: i have no idea.... today it just works. i didn't change anything Oct 20 12:19:51 RP: now i need to find out how to get the X cursor changing Oct 20 12:20:01 RP: was there some magic involved for that to happen? Oct 20 12:20:08 s/changing/moving/ Oct 20 12:20:25 mickeyl: That all changed recently to use the usb device emulation. You probably need the defconfig from poky Oct 20 12:21:13 RP: i have that one. aah, it works (a bit). the cursor is not supposed to move, it's a stylus right? Oct 20 12:21:28 mickeyl: right :) Oct 20 12:21:39 ok. now let's see whether i can get past that calibration screen# Oct 20 12:21:56 mickeyl: Ideally, the cursor should be turned off so we just have the host X one Oct 20 12:22:08 yeah Oct 20 12:22:25 is there an easy and established way to make use of the pxa270 hw cursor? Oct 20 12:22:31 hmm Oct 20 12:22:42 i can't get the crosshair to move without making drags Oct 20 12:22:51 koen: not really Oct 20 12:22:51 the effect is, the calibration is off then Oct 20 12:23:59 RP: it's calibrated now :) a single click doesn't work here, i need to drag around over every calibration point, but two or three pixels dragging works Oct 20 12:24:04 that is cool Oct 20 12:24:06 really Oct 20 12:24:33 what is the difference between libx11 and diet-x11 Oct 20 12:24:46 Crofton: about 1MB Oct 20 12:24:50 Crofton: missing some rarely used functions Oct 20 12:25:04 utf8, xim, etc. Oct 20 12:25:52 It suggests I define a preffered provider Oct 20 12:26:05 stupid usrp configure can't find sdcc ..... Oct 20 12:26:09 I think diet-x11 is a misnomer Oct 20 12:26:13 heh Oct 20 12:26:16 it should libx11/embedded Oct 20 12:26:26 and they should rename it into xorg-core Oct 20 12:26:30 hmm Oct 20 12:26:38 can't find sdcc in staging .... Oct 20 12:27:55 Crofton: rebuild sdcc-native Oct 20 12:27:57 or Oct 20 12:28:05 just run do_stage again Oct 20 12:28:17 autotools_do_stage doesn't work on sdcc-native which is why i had to revert that Oct 20 12:28:20 (see last commt) Oct 20 12:28:22 commit, even Oct 20 12:28:24 -c populate_staging -f Oct 20 12:28:53 koen: the best trick would be if you could install full libx11 with just a .so to add the missing stuff, instead of a repleace Oct 20 12:29:06 I thought I had rebuilt it ... Oct 20 12:30:06 mickeyl: could you do 'mtn add packages/qemu ; mtn commit packages/qemu -m "qemu: add missing files"' ? Oct 20 12:31:23 *cough* Oct 20 12:31:25 * mickeyl bows Oct 20 12:31:26 sorry Oct 20 12:31:31 :) Oct 20 12:33:33 mickeyl: works now, many thanks Oct 20 12:34:10 np. qemu gave me a lot of headaches, but now it looks ok Oct 20 12:34:28 this will make developing so much easier Oct 20 12:34:33 * mickeyl really excited Oct 20 12:34:44 qemu + iceccd = cheap ARM cluster Oct 20 12:34:49 RP: I'll buy you at least three leffe @ FOSDEM ;) Oct 20 12:35:55 mickeyl: One small thing - in future, if you pull things from poky, can you note it in the commit please. (from poky) or similar is fine Oct 20 12:36:10 RP: i did! "sync with poky" Oct 20 12:36:11 mickeyl: It is very useful and getting ever better with the patches :) Oct 20 12:36:51 mickeyl: Sorry, you did and I was looking at the wrong batch of commits. My applogies! Oct 20 12:36:58 RP: no need to Oct 20 12:37:23 RP: http://www.openembedded.org/bonsai/query/description/poky/ :) Oct 20 12:37:25 credit due to whom it belongs Oct 20 12:37:51 mtn-poky-pluck needed Oct 20 12:38:04 mickeyl, either I am dense, or sdcc-native still isn;t stging Oct 20 12:38:12 XorA: that's what interns are for :) Oct 20 12:38:18 the log fiels for stgae and install are 0 length Oct 20 12:38:21 Crofton: did you pull? Oct 20 12:38:29 yes Oct 20 12:38:29 :) Oct 20 12:38:31 I checked Oct 20 12:38:43 do_stage has a find bin - type ... line Oct 20 12:38:56 mickeyl, did you push? Oct 20 12:38:59 that's correct. can you check whether it works on your system when you do it manually? Oct 20 12:39:09 ok Oct 20 12:39:18 cd inot the src dir and run Oct 20 12:39:23 yep Oct 20 12:39:33 find bin -type f -perm 0755 should list all binaries Oct 20 12:40:13 hmm no files foudn Oct 20 12:40:40 heh 775 on my system Oct 20 12:41:20 man find'ing now Oct 20 12:42:28 ooooh: http://dilbert.com/comics/dilbert/archive/dilbert-20061018.html Oct 20 12:43:27 Crofton: arg, sorry, i didn't think of that. perhaps you can find a better -perm option that only checks the 'x' bit Oct 20 12:43:56 how about find bin -type f -perm -0755 Oct 20 12:44:10 what's the '-' do? Oct 20 12:44:13 even -0555 Oct 20 12:44:19 seems to make the mathcing sloppy Oct 20 12:44:24 cool Oct 20 12:44:26 er matching Oct 20 12:44:28 feel free to change, pleaes Oct 20 12:44:32 will do Oct 20 12:44:35 thanx Oct 20 12:45:44 basically the - checks for all the bits in the mask being set Oct 20 12:45:58 not exact Oct 20 12:46:09 + would check for any of the bits being set Oct 20 12:48:30 RP: i'd like to change the framebuffer resolution right from the beginning, i guess that needs me patching the qemu kernel, right? Oct 20 12:49:14 mickeyl: You might be able to do it from the commandline? Oct 20 12:49:23 oh that'd be cool! Oct 20 12:49:44 mickeyl: If not, I'd like to see patch to do it but yes, worst case patch the kernel Oct 20 12:49:51 mickeyl: Crofton : please check a BSD manual as well :) Oct 20 12:50:02 heh Oct 20 12:50:17 what is BSD? Blue Screen of Death? Oct 20 12:50:19 I was just about to type 'does that work on BSD?' :) Oct 20 12:50:27 Crofton: http://www.freebsd.org/cgi/man.cgi?query=find&apropos=0&sektion=0&manpath=FreeBSD+5.5-stable&format=html Oct 20 12:50:56 it does :) Oct 20 12:51:02 yeah Oct 20 12:51:38 my GNU children Oct 20 12:52:05 It is really hard trying to explain the BSD versus GNU crap to my professor Oct 20 12:52:26 to him it is open source versus the closed source Oct 20 12:52:38 open source politics makes no sense to outsiders Oct 20 12:53:24 Crofton: oh it is not about politics :) Oct 20 12:53:56 politics make no sense to outsiders :) Oct 20 12:54:34 Crofton: it is more about "??!!XX@@!!§§$$$ why doesn't this work on OSX anymore" Oct 20 12:54:48 try explaining american politics to americans .... Oct 20 12:55:08 Crofton: that is easy. "Castro is bad mkaay" Oct 20 12:55:14 hahahahahha Oct 20 12:55:37 zecke, priceless Oct 20 12:55:54 Castro is dying, we will replace him with Chavez Oct 20 12:56:37 Crofton: yeah, hopefully Cuba will not be another Mc Donald parking lot... Oct 20 12:57:24 hopefully south florida won't succed from the union and become part of cuba :) Oct 20 12:57:32 03koen 07org.oe.dev * r8f53c7a7... 10/ (4 files in 3 dirs): conf/: remove references to oesources.org Oct 20 12:57:37 03mickeyl 07org.oe.dev * rec665a22... 10/ (10 files in 2 dirs): qemu: add missing files Oct 20 12:57:38 hehe Oct 20 12:57:42 03koen 07org.oe.dev * r8ce31384... 10/ (1 classes/base.bbclass): base.bbclass: remove oesources.org Oct 20 12:57:42 back to work Oct 20 12:57:47 03koen 07org.oe.dev * rba75ce69... 10/ (1 packages/glibc/glibc_2.5.bb): Oct 20 12:57:47 glibc 2.5: stage correct sigthreads.h Oct 20 12:57:47 * fixes mysql compilation Oct 20 12:57:47 * spotted by 'nicolasfr' in #oe Oct 20 12:57:51 03mickeyl 07org.oe.dev * rf535484a... 10/ (4 files in 3 dirs): tslib 1.0 add configuration file for qemuarm machine - taken from Poky. Oct 20 12:58:03 mithro: any reasoning behind using inherit and not require? Oct 20 12:58:03 ok we have sdcc staging Oct 20 12:58:27 zecke: you mean me or mithro? Oct 20 12:58:32 why do I get preferred verion 2.5 of glibc not availble? Oct 20 12:58:53 mickeyl: you :) Oct 20 12:58:58 zecke: ah Oct 20 12:58:59 zecke: other than being a typo which i fixed in a subsequent commit? ;) Oct 20 12:59:35 * zecke needs to add diff to bonsai Oct 20 13:00:14 pwd Oct 20 13:00:55 hmm, sdcc needs to stage its include files also Oct 20 13:01:48 BTW: the only response to Bonsai was: you need to put DEBUG = False into your settings.py Oct 20 13:02:35 zecke: heh Oct 20 13:02:47 thanks koen for glibc2.5 Oct 20 13:03:36 mickeyl: Ah, the tslib changes will help :) Oct 20 13:03:57 Packaged contents of cu into /data/build/koen/OE/build/tmp/angstrom/deploy/glibc/ipk/cu_1.07-r0_armv5te.ipk Oct 20 13:04:52 mickeyl, what was the issue with sdcc and auto staging? Oct 20 13:05:08 sdcc needs to stage headers and libs for the cross targets Oct 20 13:05:38 RP: :) wondered why your image run xtscal and mine didn't :) Oct 20 13:05:51 RP: answer: TSLIB_TSDEVICE not set => ts.conf 0 bytes :) Oct 20 13:06:18 Crofton: it didn't install anything for me. Because of the rush I didn't look into why exactly Oct 20 13:06:34 I may revisit it :) Oct 20 13:06:36 mickeyl: One of those things you fix and then forget about. My merge scripts would remind me though. I should merge the tslib config for the 770 too... Oct 20 13:06:44 this is a complex install .... Oct 20 13:06:46 RP: ah yeah Oct 20 13:08:02 03koen 07org.oe.dev * rdba1e401... 10/ (1 packages/uucp packages/uucp/uucp_1.07.bb): Oct 20 13:08:02 uucp: add 1.07 Oct 20 13:08:02 * something is funky with autotools, so we manually run libtool and configure Oct 20 13:08:02 goxboxlive: 'bitbake uucp' to get a cu package Oct 20 13:09:02 koen: I can se you have commited it. Thanks a lot. Oct 20 13:09:17 VoodooZ: I got the mirrorball at a florist shop Oct 20 13:09:49 Cool. Oct 20 13:10:07 I'm still looking for the perfect mirror for my omnidirectional camera setup for my robot. Oct 20 13:10:40 Researchers use more complex ($$$) hyperboidal ones though Oct 20 13:12:18 VoodooZ: you could get a bearing as well Oct 20 13:12:29 VoodooZ: altough those would be a bit heavy Oct 20 13:12:35 Spherical has a very low vertical resolution and distorts the image a lot. Oct 20 13:12:43 I've been using a X-mas ornament Oct 20 13:12:50 VoodooZ: that sounds familiar Oct 20 13:12:53 has anyone ever made an internet connection from a Linux computer through a cell phone? I'm thinking this should be possible through bluetooth, but would like some verification from people who have actually done it. Oct 20 13:12:55 The cheap dollar store ones. Oct 20 13:13:02 VoodooZ: calculate one and build the mirror yourself Oct 20 13:13:07 http://robotics.no-ip.org Oct 20 13:13:07 cbrake: many times Oct 20 13:13:13 VoodooZ: so you want to have an almost linear distance function? Oct 20 13:13:37 well, the hyperbolic ones are a min. of 600USD so... Oct 20 13:13:40 cbrake: yep, via ppp over rfcomm Oct 20 13:13:43 I'm just a hobbyist after all. Oct 20 13:15:15 koen: XorA: so can this be done with about any phone? I'm assuming the bluetooth standards cover this type of operation? Oct 20 13:15:20 I just want the robot to have a good 360 view of it's surroundings without too much of a res loss as objects move away. Oct 20 13:16:02 cbrake: as long as phone supports serial over bluetooth and has modem, then as koen says, establish rfcomm connection then just add the gprs dialup stuff to ppp Oct 20 13:16:03 cbrake: yes any phone that hasn't 'motorola' written on it :) Oct 20 13:16:28 cbrake: although java applets with gprs open will prevent dialout on a lot of phones Oct 20 13:17:13 and one other question -- I assume internet connection through a phone is much easier to get working with bluetooth than USB? Oct 20 13:17:25 zecke, well, I don't have the equipment to do such a feat! And too many hobbies and not enough time. Oct 20 13:17:35 cbrake: in my experience a lot of phones dont uspport dialup from usb when they do from bt Oct 20 13:18:14 XorA: koen: thanks for the info! Oct 20 13:18:32 koen, I did try copying the angstrom toolchain setup in the .conf file to the slugos one but it says glibc 2.5 is not available. euh? I'm not expecting it to work but still. It's there. Oct 20 13:20:14 cbrake: alsa with working irda, its pretty much same as bt, you just use ircomm Oct 20 13:20:41 XorA: excellent Oct 20 13:22:54 VoodooZ: I get that as well, but it still builds 2.5 Oct 20 13:23:06 XorA: interesting where stuff like this gets used. I'm looking at connecting a terminal in a tractor cab to the internet via the farmer's phone so that field data can be uploaded near real time. Oct 20 13:23:40 koen, Strange I tried it yesterday with Slugos HEAD and right after that message it started building the old 2.3.x one... Oct 20 13:24:58 cbrake: I once designed a linux box to sit in a ferry to collect polution data from a trailing sensor pod. It could be data dumped over wireless X25 on vodafone Oct 20 13:25:24 XorA: neat Oct 20 13:27:09 cbrake: I used gprs over bt as my primary connection when the local telco couldn't figure out how to get adsl to my home for ~2 months Oct 20 13:28:06 tweaking the ppp scripts is the 'hardest' part Oct 20 13:28:25 koen: wow, nice to have that as a backup. Cell phone companies don't seem quite as progressive in the US. Internet via phone is still pretty expensive here :-\ Oct 20 13:28:45 cbrake: i didn't say it was cheap Oct 20 13:29:09 cbrake: thanks to dial-on-demand my phone bill was €300 :( Oct 20 13:29:20 (3 years ago) Oct 20 13:29:30 koen: nod Oct 20 13:29:43 I've got a flatfee gprs subscribtion now Oct 20 13:29:57 koen: how much is that a month? Oct 20 13:30:10 €17.50 for 1GB/month Oct 20 13:30:57 + 60 hours free voice to other subscribers Oct 20 13:32:59 koen: pretty good -- nothing like there here. Oct 20 13:33:48 "Wireless SpeedStep Power Manager Technology" - intel's name for cpufreq Oct 20 13:34:14 where can i check how to use the icecream.bbclass ? Oct 20 13:34:42 €5 for 100MB/mon or 19.90€/mon (flatrate capped 128kbit/s) are the best deals for gprs/3g here Oct 20 13:36:10 Genesis: depends if the patches from likewise are in Oct 20 13:36:55 i know where are the patch Oct 20 13:37:08 i just need a little help , and i 'll write it on the wiki Oct 20 13:37:20 koen, : "improve it" => link wiki man :) Oct 20 13:37:39 ( http://www.openembedded.org/overview ) Oct 20 13:39:04 Genesis: done :) Oct 20 13:39:25 good :o) Oct 20 13:39:59 documentation is the key :) Oct 20 13:40:12 goxboxlive: http://www.angstrom-distribution.org/repo/?action=details&pnm=cu for the binary Oct 20 13:41:27 zecke, : like you want , so , do you know ? Oct 20 13:45:02 RP: it doesn't look like there is a valid vga setting for 480x640 (which I'm going to need). I'm trying to patch mach-versatile/core.c for now Oct 20 13:45:37 koen, I'm feeling crazy! I did a bb -b glibc_2.5.bb instead just to force it. :) Oct 20 13:45:58 RP: we might think about adding a couple of more panels to mach-versatile and a means to chose them via CMDLINE Oct 20 13:47:28 ggilbert_: any idea when bugs.oe.org will be back? Oct 20 13:49:05 morning Oct 20 13:51:11 this may not look right, but I bet a simple "make install" will stage sdcc-native :) Oct 20 13:54:20 yeah, this is what it was before Oct 20 13:54:29 heh Oct 20 13:54:38 but i wanted to get rid of it Oct 20 13:54:47 seems we can't :/ Oct 20 13:54:58 I think the make files are too smart Oct 20 13:55:20 I seriously doubt most OE users need an 8051 compiler in the tool chain :) Oct 20 13:55:36 I really should have just assumed it was provided :) Oct 20 13:55:47 do_stage() { Oct 20 13:55:47 oe_runmake install Oct 20 13:55:47 } Oct 20 13:55:51 yeah Oct 20 13:55:56 testing now Oct 20 13:55:57 k Oct 20 13:56:25 basically, the usrp hardware has an USB interface with an 8051 Oct 20 13:56:35 Crofton: we can use that to write code for all those chip & pin cards around Oct 20 13:56:36 I need to build the firmware that gets downloaded to it Oct 20 13:59:05 ok, now usrp is barfing at staging Oct 20 13:59:19 more later Oct 20 14:00:18 hrw|work, I am going to push some updates to openomap moving it to glibc2.5 Oct 20 14:05:09 uucp? I thought that died years ago? Oct 20 14:05:47 03crofton 07org.oe.dev * r159a65e4... 10/ (1 conf/distro/openomap.conf): openomap.conf : Move from glibc-2.4 to glibc-2.5. Oct 20 14:05:52 03crofton 07org.oe.dev * rd33422da... 10/ (1 packages/sdcc/sdcc-native_2.5.0.bb): sdcc-native_2.5.0.bb : Fix do_stage. Yes, this really does the right thing. Oct 20 14:07:13 bbiab Oct 20 14:10:12 mickeyl: There is no reason you have to choose predetermined sizes - we could make it support the size direct from the commandline instead ;-) Oct 20 14:11:03 Crofton: ok. I use celinux-test and will set it to 2.3.5 probably Oct 20 14:20:10 zecke: can I edit bonsai templates a bit? Oct 20 14:21:28 ok Oct 20 14:28:59 zecke: 5 files changed, 35 insertions(+), 29 deletions(-) Oct 20 14:33:09 hrw|work: sure Oct 20 14:33:16 hrw|work: just go ahead, we can always svn revert them Oct 20 14:34:29 hrw|work: th is tableheader? Oct 20 14:34:32 Crofton|home: you do know we have an scm, so you don't have to 'save' stuff with '#comment', right? Oct 20 14:34:57 zecke: yes Oct 20 14:36:26 hrw|work: but before koen edited the CSS I was XHTML 1.0 strict ;) Oct 20 14:36:40 so one can write crappy looking and standard coformant sites Oct 20 14:37:07 ;) Oct 20 14:37:10 zecke: I didn't edit the css, I added to OE ones :) Oct 20 14:39:09 zecke, : for the moment , the only things i want from icecc.bbclass is to generate the tarball of the toolchain , do you know how to do that ? Oct 20 14:39:28 Genesis: in that case use create-env from icecream itself? Oct 20 14:39:45 python icecc.bbclass create-env ? Oct 20 14:40:04 hehe no. This is not a supported use case ;) Oct 20 14:40:20 i donno well how to deal with the bbclass Oct 20 14:40:42 Genesis: again use create-env of icecream to create only a tarball Oct 20 14:42:35 koen, I've ended with OE Oct 20 14:43:11 zecke, : i should be a bit silly but i don't understand how to use create-env of icecream Oct 20 14:43:13 sorry, but it's piece of crap ... every second package is screwed and I had to manualy fix them Oct 20 14:43:18 :( Oct 20 14:43:30 it's friday too =) Oct 20 14:43:44 Marex: that's because you don't follow the advice we give you Oct 20 14:44:05 that means getting started Oct 20 14:44:13 no, it wasnt working Oct 20 14:44:16 when I used that Oct 20 14:44:35 we also said "get it working with glibc first" Oct 20 14:44:44 you also didn't do that Oct 20 14:44:51 I wont use glibc Oct 20 14:44:58 I have 16megs of ram Oct 20 14:45:08 do you understand that Oct 20 14:45:11 right Oct 20 14:45:17 back to my original point Oct 20 14:45:18 if I use glibc it'll be slow Oct 20 14:45:25 and useless Oct 20 14:45:33 Marex: that's because you don't follow the advice we give you Oct 20 14:45:40 koen hehe :) Oct 20 14:45:51 why is there uclibc support in OE when it's busted then? Oct 20 14:46:01 cart/horse Oct 20 14:46:17 it works for me Oct 20 14:46:19 uclibc isnt a replacement for glibc, deal with it man Oct 20 14:46:31 but I added a config for my device :) Oct 20 14:46:41 get *something* working first Oct 20 14:47:08 Marex: some software won't even work with uclibc ... Oct 20 14:47:19 zecke: This Page Is Valid XHTML 1.0 Transitional! Oct 20 14:47:32 chouimat, I know that Oct 20 14:47:42 hrw|work: you fixed our css as well? Oct 20 14:47:43 wow Oct 20 14:47:47 but basic rootstrap just has to Oct 20 14:47:54 cuz there are some around Oct 20 14:48:13 hrw|work: but I like borders in tables :} Oct 20 14:48:22 zecke: internal ones? Oct 20 14:48:45 hrw|work: yes, but that doesn't matter. I'm glad you took and take a look Oct 20 14:49:16 zecke, : haaaa you spoke about _ICESCREAM_ not iceccbbclass , sorry i'm bit tired Oct 20 14:49:39 hrw|work: http://www.openembedded.org/oedem Oct 20 14:49:53 hehe Oct 20 14:49:55 excellent Oct 20 14:50:09 might move the old page to oedem06 Oct 20 14:50:13 koen: ;D Oct 20 14:51:21 more beer, less walking Oct 20 14:51:27 hehehe Oct 20 14:51:32 tomas will love that Oct 20 14:51:44 more girls Oct 20 14:51:47 *cough* Oct 20 14:51:50 more dumb girls Oct 20 14:51:55 mickeyl: lol Oct 20 14:52:24 mickeyl: thanks for trying to hook me up, how unselfish :) Oct 20 14:52:33 :D Oct 20 14:52:45 that's what friends are supposed to do, aren't they ;) Oct 20 14:53:02 03rpurdie 07org.oe.dev * r2cb7b3c9... 10/ (1 classes/base.bbclass classes/package.bbclass): base/package.bbclass: Clean up and document things a bit more. Should be no functional changes (from poky) Oct 20 14:53:05 mickeyl: like this: http://www.openembedded.org/oedem ? Oct 20 14:53:06 * mickeyl points to his ring on the finger Oct 20 14:53:17 koen: great Oct 20 14:53:28 brb Oct 20 14:53:52 mickeyl: bound by the one ring to rule them all :) Oct 20 14:53:55 hi my OE build fails on binutils-cross-2.16 ill paste the the errors. http://channels.debian.net/paste/4144 Oct 20 14:54:10 RP: hehe, righti Oct 20 14:56:12 mickeyl: but I hope to not be single next OEDEM, I might be immune as well Oct 20 14:56:20 just updated and tried to "bitbake" something (org.oe.dev) and get: Oct 20 14:56:24 ERROR: Unable to parse conf/bitbake.conf (Could not include required file conf/machine/tune-xscale.conf) Oct 20 14:56:38 BBPATH wrong? Oct 20 14:56:54 mccarthy: florian told me about your problems, I'll convert the _cvs files to svn tonight if I get some time Oct 20 14:57:02 koen, thanks Oct 20 14:57:08 zecke, don't think so Oct 20 14:57:11 zecke: no, missing '/include' inthe require path Oct 20 14:57:47 lindenle: look in /home/lindenle/src/OE/build/tmp/work/binutils-cross-2.16-r3/t Oct 20 14:57:48 emp/log.do_compile.652 Oct 20 14:57:52 lindenle: like it says Oct 20 14:58:18 zecke: actually, this immunity thing is just not working. we are civiliced, so we try to prepend that we are monogamic nowadays, but I'm really sure inside almost every man is still the little voice that recommends "nail everything that moves" :D Oct 20 14:58:18 lindenle: though I suspect lack of texinfo on your system Oct 20 14:58:31 koen, ? Oct 20 14:59:05 03koen 07org.oe.dev * r0bf9a44e... 10/ (1 conf/machine/h3900.conf): h3900.conf: fix include path for tune-xscale Oct 20 14:59:41 mccarthy: that should fix the error you pasted Oct 20 14:59:42 mickeyl: you know that your wife reads this log? Oct 20 14:59:56 time for food Oct 20 15:00:08 zecke: sometimes she does, yes. but it's ok, we talked about these things before :) Oct 20 15:00:14 ~bon appetit Oct 20 15:00:17 bon appetit is, like, smacznego. Guten Appetit. Eet Smakelijk. God Appetitt. Buon Appetito. Buen apetito Bom Apetite. buen apetito Oct 20 15:00:17 XorA: I pated exactly wht was in that log file. it is a conflicting malloc definition. I was just wondering if you had any idea why. Oct 20 15:00:35 http://www.rpsys.net/openzaurus/temp/poky_fixdebianbbclass1.patch is readable now Oct 20 15:01:18 you guys are fast. thanks. Oct 20 15:01:43 lindenle: I cant see that log in that URL, but I have no idea Oct 20 15:02:11 mccarthy: we create errors with the same pace Oct 20 15:02:24 XorA: The log is what is pasted. Oct 20 15:02:37 ERROR: log data follows (/home/lindenle/src/OE/build/tmp/work/binutils-cross-2.1 6-r3/temp/log.do_compile.652) Oct 20 15:02:46 zecke, :) Oct 20 15:02:49 RP: why did we have this conceptual break between PACKAGEFUNCS and bb.build.exec_func("read_subpackage_metadata", d) previously? Oct 20 15:03:42 RP: it looks definitely more clean now Oct 20 15:04:10 mickeyl: It actually happened at the end of populate_packages which is right in the middle of PACKAGEFUNCS Oct 20 15:04:30 lindenle: what branch you using? Oct 20 15:04:37 mickeyl: I just think lots of people editted and added things and didn't fully consider what they were doing Oct 20 15:04:45 RP: *nod* Oct 20 15:04:49 XorA: oz354x Oct 20 15:05:19 mickeyl: poky is slightly different to OE as OE doesn't have emit_pkgdata yet but I will change than, and move it to the end of PACKAGEFUNCS Oct 20 15:05:39 RP: ok, good Oct 20 15:05:48 lindenle: very odd, Im afraid I have no idea Oct 20 15:06:16 mickeyl: I can't guarantee it won't break anything but its something I think we'll just have to try and see Oct 20 15:06:30 RP: i agree. Oct 20 15:07:15 RP: on a slightly related note, i will reserve a timeslot on weekend to work on that ncurses example program we talked about. you'll probably get mail on sunday Oct 20 15:07:22 i can't stand that bogus output any longer ;) Oct 20 15:07:22 zecke: how you like it now? Oct 20 15:08:12 bbl Oct 20 15:08:49 mickey|bbl: probably 30 minutes or so Oct 20 15:08:49 hrw|work: it is not blinking! Oct 20 15:09:06 hrw|work: you might be right, it could be better without grid Oct 20 15:09:13 hrw|work: I leave the freedom to you :) Oct 20 15:12:29 mickey|bbl: sounds good. I'm away next week but might have some time to look at things offline :) Oct 20 15:20:05 zecke: just be carefull when upgrading from girlfriend-cvs.bb to wife_1.0.bb :) Oct 20 15:20:39 hvontres|poodle: wife RREPLACES friends Oct 20 15:21:14 upgrading wife_1.0 to girlfriend_2.0 costs some monies Oct 20 15:21:26 * hvontres|poodle wonders how us geeks ever manage to reproduce...:) Oct 20 15:21:43 hvontres|poodle: when we are drunk a lot of stuff is happening Oct 20 15:22:07 zecke: or we wind up marrying another geek...:) Oct 20 15:22:23 hvontres|poodle: currently I would say, ideally Oct 20 15:22:31 man I need to be more productive again Oct 20 15:25:10 zecke: how do you like it now? Oct 20 15:25:57 can anyone offer me advice n this OZ build problem: http://channels.debian.net/paste/4144 Oct 20 15:28:07 zecke: why cset.description contain

??? Oct 20 15:28:23 usrp built .... Oct 20 15:28:37 how do I get minimal python on the image? Oct 20 15:29:00 hrw|work: do they contain it? could it be a filter like |escape|linebreak? Oct 20 15:29:09 hrw|work: or is it part of the database already? Oct 20 15:29:22 zecke: dont know where do they came from Oct 20 15:29:32 hrw|work: example? Oct 20 15:29:34 zecke: but they are wrong - should not contain

Oct 20 15:29:37 http://www.openembedded.org/bonsai/query/author/rpurdie/?page=1& Oct 20 15:31:22 usrp DEPENDS on several packages Oct 20 15:31:41 I need to install a basic set of python stuff when I install the usrp on the image Oct 20 15:31:58 I imagine this means the usrp bb file needs a RDEPENDS line Oct 20 15:32:07 Crofton: probably Oct 20 15:32:10 python provides several packages, Oct 20 15:32:13 hrw|work: |lnebreaks adds

Oct 20 15:32:24 RDEPENDS = "python" scares me :) Oct 20 15:32:40 will RDEPEnDS = "python-core" work? Oct 20 15:33:00 Crofton: it should, yes Oct 20 15:33:06 zecke: arh Oct 20 15:33:16 hrw|work: http://www.djangoproject.com/documentation/templates/ Oct 20 15:34:10 hrw|work: try linebreaksbr Oct 20 15:34:30 just did Oct 20 15:35:21 zecke: better Oct 20 15:35:23 better? Oct 20 15:35:25 hehe Oct 20 15:35:40 koen: NOTE: package glibc-2.5: completed. Oct 20 15:36:00 I'm a bit afraid to install it though :) Oct 20 15:36:09 zecke: now one of hardcore html validators give 245/255 which is best this site can get Oct 20 15:37:42 hrw|work: which one? Oct 20 15:37:50 hrw|work: I though w3c.org was good enough Oct 20 15:38:09 zecke: http://kwality.polip.com/kwality.cgi?rel=1&rep=1&url=http%3A%2F%2Fwww.openembedded.org%2Fbonsai%2Fquery%2Fauthor%2Frpurdie%40openembedded.org%2F Oct 20 15:38:18 zecke: Polish interface only iirc Oct 20 15:38:29 zecke: look at 'view details' links now Oct 20 15:38:32 is django worth trying? Oct 20 15:38:34 so bonsai is a kwality site? Oct 20 15:38:42 chouimat: definately Oct 20 15:38:58 chouimat: stuff like transactions and performance will be interesting Oct 20 15:39:30 zecke: I'm looking for something to build my website ... currently have php and kolab on this server ... and mt-daapd :) Oct 20 15:39:46 anybody around who can answer a bitbake conf file question? Oct 20 15:41:10 maybe Oct 20 15:42:06 zecke, i'm starting a bring up of a powerpc box. here's a segment from the distro file i'm using: http://pastebin.ca/211931 Oct 20 15:44:35 HopsNBarley: try it :) Oct 20 15:44:43 HopsNBarley: one of the two will work, but not both :) Oct 20 15:45:22 zecke, bitbake picked the glibc-intermediate, and it did build the entire toolchain with just one problem. Oct 20 15:47:03 one of glibc or gcc (forget which) doesn't use the right AR_FOR_TARGET, picking powerpc-linux-ar, when they are names powerpc-603e-linux-gnu-ar. Oct 20 15:47:38 cu Oct 20 15:48:11 i'm curious how this works on x86->arm... i'll check the logs.... Oct 20 15:48:26 HopsNBarley: sounds like glibc :) Oct 20 15:48:39 HopsNBarley: than remove the glibc provider Oct 20 15:50:16 03rpurdie 07org.oe.dev * reeb4aa47... 10/ (7 files in 2 dirs): Oct 20 15:50:16 package.bbclass: Split into two tasks, one which prepares the packages and then Oct 20 15:50:16 package_write which actually generates the packages. The two stage approach Oct 20 15:50:16 allows us to avoid circular dependency issues from classes like debian.bbclass. Oct 20 15:50:16 As the data being emitted into pkgdata/ changed, you need to either wipe tmp or Oct 20 15:50:16 rerun the do_package tasks (wipe the do_package stamps). Everything will Oct 20 15:50:21 repackage anyway due to the new task. Oct 20 15:51:20 zecke, do you mean you think the glibc dependency caused the wrong guess on the ar name? Oct 20 15:51:37 RP: do_installl *and* do_package, otherwise you'll get lots of empty packages Oct 20 15:51:55 HopsNBarley: no, I think glibc has another 'wrong' test which deservers fixing Oct 20 15:52:05 koen: Er, right, yes :} Oct 20 15:52:20 zecke, great, me too. (-; Oct 20 15:52:21 RP: that kept me busy for a while :) Oct 20 15:53:09 zecke, what's current experience with glibc 2.5, is it considered stable? Oct 20 15:53:22 koen: I can imagine :) Oct 20 15:53:47 RP: FILES += 'this', FILES += "that", etc Oct 20 15:54:38 RP: you forgot that add package_deb.bbclass ;) Oct 20 15:54:55 koen: not forgotten but I did note it was missing ;-) Oct 20 15:55:31 hmm, bugs.openembedded.com not coming up for me. Is it working for others? Oct 20 15:55:43 koen: That is waiting for a syncing session. It should be quite safe to add though I guess Oct 20 15:56:05 http://www.carrypad.com/journal/2006/10/pepperpad-3-mini-review_20.html : They should have used OE for their SDK ! Oct 20 15:56:07 cbrake: ggilbert is working on that Oct 20 15:56:11 koen: ok Oct 20 15:56:30 RP: OE is still missing the u-a bits needed for package_deb.bbclass Oct 20 15:56:54 koen: At least the circular depends issue is no longer an issue for u-a :) Oct 20 15:57:01 :) Oct 20 15:57:25 RP: don't forget to send a short note to the list (e.g. forward the commit mail) Oct 20 15:57:48 koen: I replied to the commit with your comment about do_install already Oct 20 15:57:57 zecke, it's actually gcc that is looking for the wrong ar, not glibc. Oct 20 15:58:07 HopsNBarley: okay, fix it :) Oct 20 15:58:30 zecke, i'm working on it, i'm working on it! should i send you a patch??? Oct 20 15:58:54 HopsNBarley: hehe, no but attach one in our lovely bug tracker Oct 20 15:59:13 zecke, will do. Oct 20 16:02:43 HopsNBarley: as we do not call autoreconf on gcc/glibc you will need to patch both the m4 files and the configure script Oct 20 16:03:21 chouimat: django is worth a try, it is really easy to develop with Oct 20 16:03:34 zecke: ok url? Oct 20 16:03:40 chouimat: www.djangoproject.org Oct 20 16:04:50 zecke, okay. however, it may be my configuration: openslug doesn't have this problem. i think i've gone wrong on the TARGET naming somehow. Oct 20 16:05:49 HopsNBarley: then consult an autoconf manual Oct 20 16:06:00 HopsNBarley: it describes the parameter and its parts Oct 20 16:06:07 HopsNBarley: e.g. arch, OS, vendor Oct 20 16:06:36 * chouimat is away: lunch time!! Oct 20 16:09:00 zecke, very few of the conf/distros set the TARGET_VENDOR Oct 20 16:09:35 03rpurdie 07org.oe.dev * r0b255384... 10/ (1 classes/base.bbclass): base.bbclass: Remove some pointless presumably dead code Oct 20 16:09:41 make that conf/machine... Oct 20 16:10:06 HopsNBarley: machines shouldn't set TARGET_VEND0R :) Oct 20 16:10:58 there does seem to be some confusion about what gets set in distros vs. machines. it's hard to tear them apart. Oct 20 16:11:21 actually it's not Oct 20 16:11:36 I agree its really very simple... Oct 20 16:11:37 unless you are used to nslu2's ;) Oct 20 16:11:45 that' Oct 20 16:11:45 HopsNBarley: angstrom does Oct 20 16:11:58 angstrom is the only one that sets it (-; Oct 20 16:12:34 Ask yourself if the problem specific to that machine? Is so its a machine option. If its some kind of policy, its distro Oct 20 16:12:50 anything inherent to the hardware should be in the machine.conf Oct 20 16:13:06 RP: although the tune-* is a corner case Oct 20 16:13:22 i'd *like* to just build openslug for the new board, but that is wedded to the nslu2, so i'm in the process of making openiom. is this the wrong approach? Oct 20 16:13:34 koen: machines should elect which tunes they're capable of supporting, the distro should select it ;-) Oct 20 16:13:38 rp: i've got that right. Oct 20 16:13:50 koen: got that right too! Oct 20 16:14:15 HopsNBarley: zillions of half finished one-off OpenFoo projects are a trait of that so called 'open source' thing Oct 20 16:15:30 koen, in theory is openslug strictly for nslu2? i'm perhaps naively assuming that distro == portable unix distribution, i.e. can run on various h/w. Oct 20 16:15:54 HopsNBarley: try using 'DISTRO=generic' first, and if that turns out to be wrong for your needs check other DISTROs and if those don't fit you as well, build your own based on the experiences with the other ones Oct 20 16:16:13 HopsNBarley: your idea is right, but openslug/nslu2 is a bit ugly Oct 20 16:16:21 koen: uucp buildt successfully Oct 20 16:16:29 goxboxlive: great! Oct 20 16:17:59 HopsNBarley: A well designed distro isn't machine specific. openslug isn't a good example but will hopefully learn :) Oct 20 16:18:54 thanks for clearing that up! after a lot of looking around that was my suspicion. Oct 20 16:19:16 thanks guys! Oct 20 16:19:36 not that openslug is 'bad' it's just too machine specific Oct 20 16:21:06 Yes, don't get me wrong, openslug is just rather focused which is understandable given their interests :) Oct 20 16:24:31 koen: The problem with using the pkgdata functions is they're in python and we need to use them from shell :-/ Oct 20 16:25:16 Hi Oct 20 16:25:25 RP: right :( Oct 20 16:25:34 koen: and the pkgdata functions are per PN, not per P+PR Oct 20 16:26:08 koen: We should try and share the data though... Oct 20 16:27:10 RP: I'm still looking for a clean way to skip straight to packaged staging if the package is present Oct 20 16:27:22 RP: do_fetch_prepend seems a bit to ugly Oct 20 16:29:13 koen: Ideally, we need to add a mechaism to feed something back into bitbake to tell it tasks aren't needed anymore Oct 20 16:29:48 RP: Is there any news on pxa SRAM support? Oct 20 16:30:00 koen: If we have that, you just write do_ps_fetch and add it before do_fetch Oct 20 16:30:09 Elive_user60: No, I'm afraid not Oct 20 16:31:20 ok, thanks Oct 20 16:31:22 bother, forgot to enable usb host in osk kernel ... Oct 20 16:31:29 rp: hi Oct 20 16:31:29 RP: with bitbake-mt we could just write out the stamps, and bb will pick those up, right? Oct 20 16:32:20 koen: It will but it less efficient than actually directly telling bitbake Oct 20 16:32:48 koen: I'm not sure we can directly tell bitbake anyway so it probably will come to that Oct 20 16:33:03 The stamps handling in bitbake needs work... Oct 20 16:35:10 rp: eh. up for fielding a question probably not relevant to 2.6 kernels? :( Oct 20 16:35:30 RP: granting that the kernel has moved on since, do you have any idea how one wo Oct 20 16:35:32 uld make use of bvdd in SDL? (assuming one's SDL has bvdd support) Oct 20 16:36:09 raduga: I have no idea how SDL works... Oct 20 16:36:22 ok, thanks :) Oct 20 16:36:28 * chouimat is back. Oct 20 16:37:05 i'm unsure if the SDL libs will automagically do any acceleration they can (if bvdd is present) or if there are special calls you need to make, to do that. Oct 20 16:37:29 the sdl in doesn't have bvdd support Oct 20 16:37:34 (what ever that means) Oct 20 16:37:52 koen: understood Oct 20 16:38:18 raduga: basically 'bvdd' is a zaurus/kernel2.4 technology Oct 20 16:38:40 koen: RP's answer was clear enough in convincing me that I probably am wasting my time pursuing this. Oct 20 16:38:48 (which is good enough) Oct 20 16:39:56 raduga: If you want to do anything, look at the 2.6 overlay support Oct 20 16:40:10 raduga: teaching sdl how to use overlays would be a more future proof thing Oct 20 16:40:27 RP: I will, when I'm back on a 2.6 kernel machine Oct 20 16:40:32 (probably not far in the future) Oct 20 16:40:41 I highly suspect sdl already knows about overlays and colorspaces Oct 20 16:40:47 koen: granted Oct 20 16:40:56 a+ Oct 20 16:47:13 RP: in most use cases for SDL (games) you want faster blits, so a qvga fb + sram would help most Oct 20 16:47:53 I could test the sram patch after my new phone arrives :) Oct 20 16:48:01 koen: right, that makes sense. We should update the kernel to do that automatically I guess Oct 20 16:48:10 a phone with camera, touchscreen or other non-phone features Oct 20 16:48:16 s/with/without/ Oct 20 16:48:50 koen: which phone? Oct 20 16:48:57 zecke: nokia 6021 Oct 20 16:49:06 zecke: I got tired of this motorola shit Oct 20 16:49:26 and I had to renew my subscribtion so the nokia was basically free Oct 20 16:49:28 koen: and it runs linux? Oct 20 16:49:42 zecke: I believe nokia ported linux to it :) Oct 20 16:50:00 koen: finally they are getting rid of Symbian :) Oct 20 16:52:01 damn HP printing driver messed up OSX Oct 20 16:56:55 python started on the osk Oct 20 16:57:07 firmware loaded into usrp from osk Oct 20 16:57:14 Crofton: congrats Oct 20 16:57:19 yay Oct 20 16:57:30 how do I get usb modules to load at boot time? Oct 20 16:58:04 bother, weather radar is down, now I need to look out the window to see if it is raing Oct 20 16:58:15 chech /etc/ for a subdir called 'modutils' or 'modules' Oct 20 16:58:55 then do "echo 'module_I_want_loaded' > /etc/modules/some-name; update-modules" Oct 20 16:59:03 repeat for every module Oct 20 16:59:28 hmm Oct 20 16:59:33 or you could add module_autoload_foo = "foo_blah" to recipes in OE Oct 20 16:59:43 that is what I want :) Oct 20 16:59:52 to the image file? Oct 20 17:00:41 ehm, I don't know where autoload statements are valid.... Oct 20 17:01:06 looks like kernel files Oct 20 17:01:27 later :) Oct 20 17:01:40 how does the ide stuff load? Oct 20 17:12:10 03cbrake 07org.oe.dev * r83c4f209... 10/ (4 files in 4 dirs): Oct 20 17:12:10 initscripts: Add custom checkroot.sh for logicpd-pxa270 Oct 20 17:12:10 init script needs to be looking for /dev/root instead of rootfs to determine Oct 20 17:12:10 if the rootfs has been mounted rw yet. This may be a setup issue in the Oct 20 17:12:10 kernel build, but until we can figure that out, this is the workaround. Oct 20 17:12:23 arrrghhh I need omnipy now ...... Oct 20 17:19:37 RP? Oct 20 17:56:29 Anyone who knows how i can solve this? : http://pastebin.ca/212105 Oct 20 17:58:37 goxboxlive: hi, seeming, where should have been -L in the Makefile, only the "L" is left, and make tries to find that command. Oct 20 17:58:48 s/seeming/seemingly Oct 20 17:59:34 ok, do you have any idea what that command might be? Oct 20 17:59:55 Is this a OE bug or is it in Opie Oct 20 17:59:59 $CC? Oct 20 18:38:17 03crofton 07org.oe.dev * r83d7c4d7... 10/ (1 packages/usrp/usrp_0.12.bb): usrp_0.12.bb : update list of include files that need installing for 0.12. Oct 20 18:38:37 Has anybody done a X driver for the PXA270 video rather than the kernel framebuffer driver? Oct 20 18:38:53 the reason I ask, is for the Xv support. Oct 20 18:38:54 RP: ping Oct 20 18:39:33 wsuetholz: there are some overlay patches floating around, but I don't know if they're for X or just the framebuffer or ... Oct 20 18:39:42 wsuetholz: no, but o-hand.com is working on iwmmxt accell for kdrive Oct 20 18:39:59 most of the patches I've seen are for the kernel. Oct 20 18:40:21 hey koen! Oct 20 18:40:34 hey mreimer Oct 20 18:40:51 mreimer: you should try Xw100 on you hx4700 Oct 20 18:41:02 (which is the default now in .dev) Oct 20 18:41:22 koen: I'm eager too. maybe today Oct 20 18:41:31 koen: I just figure out how to get hx4700 to suspend/resume a lot faster Oct 20 18:41:53 mreimer: by not booting linux at all? Oct 20 18:41:56 :p Oct 20 18:42:21 :-) Oct 20 18:42:23 if I wanted to make my own custom borzoi config, what would I need to change? Oct 20 18:42:24 no Oct 20 18:42:32 koen (and RP): It turns out that w100fb uses memcpy_{from|to}io() when it saves/restores video memory, which is completely non-optimized on ARM, copying individual bytes at a time Oct 20 18:42:44 where was OEDEM at? Berlin or...? Oct 20 18:42:50 likewise: berlin Oct 20 18:43:00 koen: can you recommend a nice hotel? Oct 20 18:43:10 likewise: hotel majestic Oct 20 18:43:19 koen: or anyone who can? Would like to spend some holidays there Oct 20 18:43:20 koen, RP: changing that to memcpy() drops the save/restore video memory from 211/64 to 40/19 jiffies, respectively Oct 20 18:43:49 koen: tnx! Oct 20 18:44:23 03crofton 07org.oe.dev * r97021a57... 10/ (3 files in 3 dirs): linux-omap1_2.6.*+git : Ad support for the USB host interface. Oct 20 18:44:33 03crofton 07org.oe.dev * r5766b94f... 10/ (11 files in 2 dirs): Rename OSSIE packages. Oct 20 18:44:50 likewise: it's ~150 meters from the 'konstanzer strasse' subway station (U7 line) Oct 20 18:45:14 * likewise is noting down Oct 20 18:45:15 oesources.org is takign forever to timeout .... Oct 20 18:45:30 likewise: If you go by ICE train get off at the spandau station, the U7 line starts there Oct 20 18:45:44 man I have wxWidgets Oct 20 18:45:48 since the hauptbahnhoff doesn't have any connections to the subway Oct 20 18:45:57 ~lart wxWidgets Oct 20 18:45:57 * ibot pushes the wall down onto wxWidgets whilst whistling innocently Oct 20 18:46:09 mreimer: sounds like a nice improvement Oct 20 18:46:30 koen: makes a big difference. it was really bothersome before Oct 20 18:46:40 koen: tnx, noted Oct 20 18:46:47 * mreimer builds Xw100 Oct 20 18:47:54 03mickeyl 07org.oe.dev * rcfb2fd13... 10/ (3 files in 3 dirs): add matchbox-config-gtk - from Poky Oct 20 18:48:01 03mickeyl 07org.oe.dev * r7e945e0f... 10/ (1 conf/distro/generic.conf): generic.conf: update to glibc 2.5 Oct 20 18:48:11 03mickeyl 07org.oe.dev * re9cc9bd9... 10/ (5 files in 2 dirs): dates: sync with O-Hand Poky Oct 20 18:48:20 03mickeyl 07org.oe.dev * r1f09cd3a... 10/ (1 packages/web/web_svn.bb): web svn sync with O-Hand Poky Oct 20 18:50:59 Crofton: consider making all the usb stuff (like usb-storage) modular so people don't have to compile a kernel when wanting to use an usb card reader :) Oct 20 18:52:11 mickey|bbl: don't forget to check if sane-srcdates is still correct after you updates Oct 20 18:52:33 koen: 2.6.17 runs on h2200 now, if you're interested Oct 20 18:52:47 mreimer: ah, cool Oct 20 18:53:05 mreimer: did you recent clock work fix cpufreq? Oct 20 18:53:32 koen: no, not yet. I haven't forgotten it; still on my list Oct 20 18:53:47 koen: looks like you've been busy with angstrom. I'll have to try it one of these days. How's it working? sounds like you're not fighting libc anymore Oct 20 18:54:07 mreimer: if you login as root it's pretty neat Oct 20 18:54:17 koen: fixing cpufreq will require some careful thinking (e.g. how to go about switching frequencies with I/O is in progress?) Oct 20 18:54:17 normal users encounter dbus timeouts Oct 20 19:03:43 mreimer: http://www.openembedded.org/bonsai/query/description/h2200/ Oct 20 19:04:18 is oesources.org down? Oct 20 19:04:28 Laibsch: it doesn't exist anymore Oct 20 19:04:47 koen: cool. (I noticed your last h2200 commit referring to cpufreq) Oct 20 19:04:56 * Laibsch checks flex bb Oct 20 19:10:10 koen: Hi! So, something in the vein of what we talked couple of days ago - hh.org CVS cleanup. I've just committed my "common buttonmap" refactor, which makes all 2.6 devices (at least, pure PDAs) have same codes for appbuttons. Oct 20 19:10:41 koen: Of course, this will propogate into GPE and OPIE configs/machine files. Oct 20 19:10:59 Laibsch, oesources is certainly being annoying right now .... Oct 20 19:11:06 koen, I'd like to start commit patched towards that in OE.dev, if you don't mind. Oct 20 19:11:14 patches Oct 20 19:11:26 koen: Has it been removed from OE meta data already? Oct 20 19:11:36 Was this scheduled? Oct 20 19:11:43 koen: tehy will be unused for couple of week more, but I'd like to have stuff in place to switch quickly then. Oct 20 19:11:46 Crofton: oesources doesn't exist anymore Oct 20 19:12:03 psokolovsky: ehm, aren't those patches dependant on kernel-version Oct 20 19:12:18 well oe is still trying to find it .. Oct 20 19:12:36 check your local.conf for CVS_TARBALL_STASH Oct 20 19:12:41 koen, yes. Oct 20 19:12:42 hi psokolovsky Oct 20 19:12:48 Hi! Oct 20 19:12:58 psokolovsky: thanks for integrating the button keycodes! Oct 20 19:12:58 psokolovsky: so it should do something like 'uname -a' during runtime, right? Oct 20 19:13:35 koen, no, we should just deprecate old crap ;-) Oct 20 19:13:51 psokolovsky: thanks for suggesting that :) Oct 20 19:13:53 koen: patch is already in 2.6.16 branch. wait for more testing for 2.6.17 Oct 20 19:13:53 thanks Oct 20 19:14:06 ~lart myself Oct 20 19:14:06 * ibot brandishes Excalibur! "With this sword, I vanquish thee, myself!" and lops off myself's head Oct 20 19:14:20 So, it should be in 2.6.16-hh8 & 2.6.17-hh3 Oct 20 19:14:24 mreimer: did you have a look at the genirq changes yet? Oct 20 19:14:41 koen: no, but it's probably going to mean a bit of work for the ASIC drivers Oct 20 19:14:51 koen: it's a lot of work to track mainline Oct 20 19:15:57 time got get h2200 into mainline Oct 20 19:17:02 koen: yeah, that would be nice, but a *lot* of time to get it cleaned up acceptably for them Oct 20 19:20:32 but you'd get updates to stuff like genirq basically for free Oct 20 19:21:11 koen: yep, that would be nice Oct 20 19:21:54 koen: do you know the status of this iwmmx acelerated kdrive? Have they released any code yet? Oct 20 19:22:12 koen: I would enjoy the work getting it into mainline, but can't afford the time. Maybe 40 hours+ of work? Oct 20 19:23:26 Elive_user60: you'd have to ask the o-hand dudes Oct 20 19:23:49 sounds like it Oct 20 19:23:57 koen: do you know what they're accelerating with iwmmxt? alpha blending and things like that? Oct 20 19:24:12 What is the correct value for CVS_TARBALL_STASH, then? Oct 20 19:24:21 mreimer: they ported the fbmmx layer to iwmmxt Oct 20 19:24:26 koen: ah Oct 20 19:25:51 mreimer: but arm can't use immediate values and requires aligned data, so it might not be *that* faster Oct 20 19:25:58 bummer Oct 20 19:26:13 cairo is getting some good love Oct 20 19:26:34 after the cairo people ignored all profiling reports Oct 20 19:26:43 yep Oct 20 19:26:55 have you seen their latest commit messages? they come with performance graphs :-) Oct 20 19:27:01 nobody will escape the blame Oct 20 19:27:26 'we want profiled" "here, profiled", "no we want now" "there, as well" "ehm, we're running out of excuses to ignore you" Oct 20 19:28:32 I reckon koens first words were 'Hello World' :) Oct 20 19:30:45 sometimes irc silence is funnier than real life post-gag silence Oct 20 19:35:30 koen: Care to fix xstroke_0.6.bb? That seems to be the only package that fetches straight from oesources.org as a quick rgrep revealed. Oct 20 19:38:11 bugs.openembedded.org is still down, too? Oct 20 19:38:31 local.conf.sample needs to be fixed, too. Oct 20 19:38:40 I'd push it, but ... ;-) Oct 20 19:40:30 Laibsch: http://www.openembedded.org/bonsai/view/rev/20587/ ? Oct 20 19:41:00 ARRGG Oct 20 19:41:12 Cool. I guess, I will change the cron from daily to every half hour ;-) Oct 20 19:42:44 koen: But I think you did forget the xstroke bb file ;-) Oct 20 19:43:03 right :) Oct 20 19:48:47 Hi johnny_a Oct 20 19:48:52 Welcome to #oe Oct 20 19:48:57 ;-) Oct 20 19:49:04 Thanks :) Oct 20 20:11:21 koen: are you there Oct 20 20:13:17 ggilbert: thank's for fixing bugzilla Oct 20 20:13:36 Power went out yesterday and people forgot to turn my machine on :\ Oct 20 20:13:40 grrr Oct 20 20:14:21 If I turn on the alsa kernel stuff in the omap kernel Oct 20 20:14:37 I need to add alsa-lib and alsa-utils, but alsa-driver? Oct 20 20:14:59 Crofton: no need for alsa-driver Oct 20 20:15:08 good :) Oct 20 20:17:58 03koen 07org.oe.dev * ra07d6e3d... 10/ (1 packages/alsa/alsa-driver_0.9.6-hh4c.bb): alsa-driver: add compatible_machine mask Oct 20 20:19:05 koen: Have you tried out your soundcard yet? Oct 20 20:19:21 bah Oct 20 20:19:24 beer time Oct 20 20:20:29 goxboxlive: I tried the soundcard on my h2200 Oct 20 20:20:48 but has it same souncard as hx4700? Oct 20 20:22:07 no, uda1380 Oct 20 20:22:42 likewise: the cse patch works,I get sound from alsa :) Oct 20 20:23:27 koen: the compiler patch I added to OE bugs? Oct 20 20:24:08 koen: cse? Oct 20 20:24:20 likewise: no, the same patch that we have had in oe for ages :) Oct 20 20:24:45 koen: hmm, ok, was it already in then?? Oct 20 20:25:11 koen: I always tend to have at least the upstream pr# in the patch name. Oct 20 20:26:54 03pfalcon 07org.oe.dev * ra2d80ffe... 10/ (1 conf/machine/htcuniversal.conf): htcuniversal: Organize machine settings per logical groups. Oct 20 20:28:25 likewise: http://www.openembedded.org/bonsai/view/rev/17736/ Oct 20 20:42:18 03pfalcon 07org.oe.dev * r8e3eb6c8... 10/ (1 conf/machine/h6300.conf): h6300: Convert to task-base, organize machine settings per logical groups. Oct 20 20:43:37 Ok, so all 2.6 pocketpc machines have been converted to task-base Oct 20 20:44:07 ah, nope, jornada56x.conf still Oct 20 20:45:55 if more than 70% of the machines are converted we can RFC the removal of task-bootstrap Oct 20 20:53:34 hmmm Oct 20 20:53:47 a quick look shows we are at 17/65 Oct 20 20:57:14 koen: Is it ok to prefer tar.bz2 over tar.gz for source downloads, if there're both provided by some project? Oct 20 20:57:27 sure Oct 20 20:57:39 why waste bandwidth and diskspace? Oct 20 20:57:54 koen, ok, will commit couple patches for that then Oct 20 20:57:55 the extra 0.1% of buildtime will go unnoticed Oct 20 20:58:25 could you rfc that to the mailing list first? Oct 20 20:58:45 the argument for bw and space are a bit moot if people are going to have both :) Oct 20 20:58:59 koen, just started to write mail and then thought it's too minor ;-) Oct 20 20:59:18 koen never underestimate the power of human stupidity :) Oct 20 21:02:03 chouimat: if only we could harnass it and use it :) Oct 20 21:04:31 koen they tried it in the usa ... and now they're stuck with Bush and the Republicans ;) Oct 20 21:05:04 not that the democrats would be a big improvement Oct 20 21:05:19 it's like choosing between 'far right' and 'extreme right' Oct 20 21:05:26 koen I agree they need a King ;) Oct 20 21:06:09 koen like the one in the Burger King ads :) Oct 20 21:07:37 Have It Your Own Way Oct 20 21:07:37 hmmm ok can't use mtn over nfs ... or maybe I do something wrong :( Oct 20 21:07:37 america needs a Koen Oct 20 21:08:16 you cna't use sqlite over nfs Oct 20 21:08:16 raduga: heh, I doubt that :) Oct 20 21:08:16 koen: a pity we can't vote for you Oct 20 21:08:16 koen k :( Oct 20 21:08:16 i suck at being a leader Oct 20 21:08:28 koen: our leaders suck at least as much Oct 20 21:08:35 chouimat: i ran into that when trying to do native dev on my netbook Oct 20 21:08:41 koen: only, they believe they are good at it Oct 20 21:10:09 koen work over samba? Oct 20 21:11:55 chouimat: didn't try that Oct 20 21:12:07 samba and I aren't good friends Oct 20 21:12:46 then I will check with sshfs :) Oct 20 21:13:18 chouimat: it works over anything that doesn't mess up locking Oct 20 21:14:40 mreimer: That's some improvement! Oct 20 21:15:02 mreimer: We used to use the optimised version but it was shot down when I merged with mainline :-( Oct 20 21:15:42 mreimer: The better patch might be to make memcpy_[from|to]io use the optimised version Oct 20 21:15:45 RP: would there be a problem with just defining memcpy_fromio() et al to be memcpy()? Oct 20 21:16:05 RP: so just change include/asm-arm/io.h Oct 20 21:16:32 mreimer: The only question - why is it like that atm? Oct 20 21:16:48 RP: yeah, I wonder why rmk even has arch/arm/kernel/io.c Oct 20 21:16:59 RP: I'll ask on l-a-k Oct 20 21:17:15 mreimer: That sounds good :) Oct 20 21:17:57 mreimer: Note its marked as needing optimisation Oct 20 21:18:18 mreimer: That suggests if we submit a patch to use the optimised xscale case, it would be accepted Oct 20 21:18:42 RP: yeah, I saw that. I compared various other platforms' io.c and saw that some of them had "optimized" versions written in C but still the "this needs optimising" comment Oct 20 21:20:27 03pfalcon 07org.oe.dev * r36451eb2... 10/ (1 packages/glibmm/glibmm_2.8.4.bb): Oct 20 21:20:27 glibmm 2.8.4: There's not much sense to have DEFAULT_PREFERENCE = "-1" for Oct 20 21:20:27 single recipe. Also, it works. Oct 20 21:20:30 RP: koen says o-hand is working on an iwmmxt-accelerated xserver? I suppose to speed up Composite? how's that coming along? Oct 20 21:21:49 mreimer: The problem is that the mmx instructions require 8 byte alignment of the data which is hard to arrange in the xserver :-( Oct 20 21:21:53 :-( Oct 20 21:22:02 but we have code which does speed up some cases Oct 20 21:22:12 RP: is IPP any use? Oct 20 21:22:22 IPP? Oct 20 21:22:50 Intel Performance Primitives or whatever they call it--that library of high performance xscale code for all kinds of idct, jpg etc. uses Oct 20 21:23:25 it's not GPL Oct 20 21:23:33 LGPL? Oct 20 21:23:37 is it? Oct 20 21:23:44 mreimer: My quick scan of the arm code suggests we should patch the memcpy io reoutines to use memcpy directly - that's what I'd propose to LAKML Oct 20 21:24:18 mreimer: I've never used it. It could be useful, depending on what we have that would use it Oct 20 21:24:30 likewise: nico claims IPP is not optimized properly :) Oct 20 21:24:30 RP: do you mean io.c, or io.h where the memcpy_fromio() is defined? Oct 20 21:25:01 RP: if you're interested IPP is at http://www.intel.com/cd/software/products/asmo-na/eng/perflib/ipp/302910.htm Oct 20 21:25:42 mreimer: I'd probably try for io.h. It seems the accelerated memcpy code is only on armv though Oct 20 21:26:23 RP: really, armv only? that's surprising Oct 20 21:27:06 mreimer: Its in mm-armv.c Oct 20 21:27:43 mreimer: Actually, ignore me :-/ Oct 20 21:27:50 arch/arm/lib/memcpy.S Oct 20 21:28:54 hard to figure out how that works, with the include at the bottom... Oct 20 21:29:08 mreimer: right, and that's always built Oct 20 21:29:24 mreimer: It just uses some common copy code... Oct 20 21:29:43 RP: so really we should dump io.c and patch io.h to use memcpy(), right? Oct 20 21:30:16 koen shfs works great ;) Oct 20 21:30:29 chouimat: cool Oct 20 21:30:54 chouimat: that means I should install fuse on my boards :) Oct 20 21:31:00 hi Oct 20 21:31:18 koen I just used shfs.sf.net (kernel module) Oct 20 21:31:56 mreimer: I'd say yes, unless there is some io space access issue I can't think of atm Oct 20 21:32:10 mreimer: Best to ask on LAKML Oct 20 21:32:15 RP: will do Oct 20 21:32:47 mreimer: Is IPP publically available? Oct 20 21:33:02 Friday night ... pub and pool time :) Oct 20 21:33:44 RP: I don't know Oct 20 21:34:12 mreimer: I think you have to pay for it Oct 20 21:34:30 :-( Oct 20 21:34:42 hrw|gone: how did your build from scratch go? Oct 20 21:36:10 * koen hugs bonsai Oct 20 21:38:14 RP: the license is at http://www.intel.com/cd/software/products/asmo-na/eng/219689.htm Oct 20 21:38:23 RP: it has a noncommercial use license Oct 20 21:40:05 mreimer: Not really suitable for OE then :-( Oct 20 21:40:11 oh yeah Oct 20 21:40:27 RP: libraries are free redistributable. the license is $199 per developer/machine Oct 20 21:40:34 that's not too bad Oct 20 21:41:13 RP: free IPP for linux download (non-commercial): http://www.intel.com/cd/software/products/asmo-na/eng/perflib/ipp/219858.htm Oct 20 21:41:51 the people at the signals& systems department love IPP Oct 20 21:41:59 it does wonders on x86 Oct 20 21:42:15 it would be nice to have something like that for libjpeg, liboil, etc. Oct 20 21:42:19 on arm that is Oct 20 21:42:35 mreimer: I'm afraid I'm dubious about what we'd be allowed/not allowed to do with it Oct 20 21:42:36 zeenix is/was working on iwmmxt for liboil Oct 20 21:42:42 great Oct 20 21:42:55 mreimer: If one of us could redistribute it for all, that would be interesting though Oct 20 21:42:57 koen: does zeenix hang out here? I have a bug to report Oct 20 21:43:21 ah, liboil. IIRC, the first thing that does on arm is segfault :-/ Oct 20 21:44:08 mreimer: see http://advogato.org/person/zeenix/diary.html?start=93 Oct 20 21:44:23 RP: yeah, that's what I thought Oct 20 21:44:25 thanks koen Oct 20 21:45:02 I should finish installing scratchbox so I can update the angstrom sdk Oct 20 21:46:03 RP: license FAQ: http://www.intel.com/cd/software/products/asmo-na/eng/219693.htm Oct 20 21:46:41 RP: .so files are redistributable Oct 20 21:47:03 RP: no royalties Oct 20 21:47:43 mreimer: If we find them useful, I'll happily propose 'OE' buys a copy and builds against commond distros ;-) Oct 20 21:47:58 RP: that would be excellent! Oct 20 21:50:20 mreimer: Prove them useful and I might even talk OH into doing it. One thing that worries me is that two developers with 8 test/build machines needs 8 copies :-/ Oct 20 21:50:45 RP: yeah, that's too bad Oct 20 21:50:49 hrw|gone: XorA|gon1 Thanks for the help w/ gnome-vfs-bus, the CVSDATE compiles fine in .oz Oct 20 21:50:53 mreimer: I do have more than one build machine although I guess I'd designate one... Oct 20 21:52:13 mreimer: I wonder if the headers are redistributable ? Oct 20 21:53:06 RP: "Redistributables" are the files listed in the "redist.txt" file that is included in the Materials. If there is no redist.txt file included in the Materials, then there are no Redistributables. Oct 20 21:53:24 hrw|gone: I'll compile a test image an commit the changes if everything appears to work ok Oct 20 21:53:44 mreimer: Right, I read that. Are the headers in there though? Oct 20 21:53:55 RP: I haven't downloaded the evaluation copy to check Oct 20 21:54:21 mreimer: The evalutaion copy has a different redist.txt :-/ Oct 20 21:55:10 RP: I wonder if the non-commercial use license could still apply? if you're not selling the libraries Oct 20 21:55:50 RP: best might be to just talk to them and get their blessing for OSS use Oct 20 21:57:28 do those require a special compiler as well? Oct 20 21:58:24 koen: "iwmmxt_le-gcc (GCC) 3.3 (MontaVista), glibc 2.3.2 and binutils 2.14 20030612" and "arm-linux-gcc (GCC) 3.3.2, glibc 2.3.2 and binutils 2.14.90.0.7 20031029" Oct 20 21:58:37 that's for pxa27x Oct 20 21:58:38 erk Oct 20 21:58:41 mvista Oct 20 21:58:56 but gcc 3.3.2 is ok, right? Oct 20 21:59:02 probably newer versions would be fine Oct 20 22:01:14 'mvista' means "patched to hell" Oct 20 22:01:59 At least its not Intel's compiler Oct 20 22:02:37 I took that to mean that it would work with Monte Vista's compiler *or* arm-linux-gcc Oct 20 22:02:50 so we'd probably be fine with what's already in oe Oct 20 22:22:05 03rpurdie 07org.oe.dev * r6a84a1b9... 10/ (1 classes/image_ipk.bbclass): image_ipkg.bbclass: Fix image task dependencies after packaging updates Oct 20 22:30:39 'night all Oct 20 22:31:26 night Oct 20 22:34:16 g'night RP Oct 20 23:26:45 hey koen, just found all the tasks/*. very nice. big help constructing a distro. Oct 20 23:36:10 anybody working on powerpc embedded? Oct 20 23:39:50 RP: you still around? Oct 20 23:40:14 RP: I downloaded the free version of IPP and see that it contains .so files but no source for the libraries Oct 20 23:40:59 RP: the libraries depend on libc.so.6, GLIBC_2.1.3, and "private flags = 602: [APCS-32] [VFP float format] [software FP] [has entry point]" Oct 20 23:41:01 mreimer, not sure if PR is around, but I am, and I also had interest in IPP Oct 20 23:41:22 hi psokolovsky. maybe IPP = free speed Oct 20 23:41:28 mreimer, but exactly, lost it completely when got complete idea about binary-only and licensing Oct 20 23:41:43 mreimer, with lotsa strings attached ;-\ Oct 20 23:42:03 the licensing doesn't seem that bad, since the libraries are redistributable, and free for non-commercial use Oct 20 23:42:17 but yeah, source would be better Oct 20 23:42:18 mreimer, actually, I once asked RP if there're kernel-level means to profile code reliable, e.g. with interrrupts off Oct 20 23:42:40 psokolovsky: did you get an answer? Oct 20 23:42:56 psokolovsky: xscale has some performance counters, but I don't know if the kernel has any code to read them Oct 20 23:43:20 mreimer, there're some great ideas around, like dynamic code generation and genetic algorithm code optimization. Oct 20 23:43:37 mreimer, I was pondeing idea to play with that for ARM in some very distant future ;-) Oct 20 23:43:41 psokolovsky: sounds clever! Oct 20 23:44:04 mreimer, yep, that would solve issues once and for long time ;-) Oct 20 23:45:16 mreimer, I don't known if you heard about, but there was a project on SF.net which was using dyngen for 3d code. It did wonders and they were bought of by some firm, and project shutdown, code mostly removef ;-) Oct 20 23:45:22 removed ;-) Oct 20 23:45:29 no, I hadn't heard Oct 20 23:45:37 is their code still available somewhere? Oct 20 23:46:04 mreimer, so well, looking at that direction may be an another idea to get a good code performance Oct 20 23:46:33 mreimer, yes, sure, as it was GPL, the only thing can do is stop support for it and their own distribution Oct 20 23:47:24 good Oct 20 23:49:11 mreimer, about optmized memcpy() - once I had a look at freebsd/openbsd, and they had really hairy memcpy() code, but they wrote it as superoptimized one. so, another place to look just in case Oct 20 23:49:55 psokolovsky_: thanks. take a look at arch/arm/kernel/io.c -- lots of opportunity to optimize there Oct 20 23:49:58 mreimer, finally, I don't understand RP's concerns with 8-byte alignments with iwmmx - well, x86 has the same. I mean, it can work /wo alignment, but performance sucks, so everyone just aligns it. Oct 20 23:50:33 mreimer, all in all, there're lots of leads to optmize perfomance, it seems, if only there was time to follow them ;-) Oct 20 23:50:50 yes, time... Oct 21 02:08:16 I'd love to see benchmarks with/without IPP. Oct 21 02:08:16 Especially under OpenCV **** ENDING LOGGING AT Sat Oct 21 02:59:57 2006