**** BEGIN LOGGING AT Thu Nov 03 02:59:58 2005 Nov 03 03:01:58 do13[offline]: ping Nov 03 03:04:30 Several patches in rmk's system or on LKML are going to conflict with ones I have in various stages of submission :-/ Nov 03 03:06:34 RP: time to get in first then :-D Nov 03 03:07:49 someone built x86/uclibc with OE? Nov 03 03:08:03 ages ago Nov 03 03:09:17 I cannot get gcc-cross-initial 3.4.2 Nov 03 03:09:20 3.4.4 Nov 03 03:09:36 hrw|work: you are plaing with an X86 build ? Nov 03 03:10:11 alan|laptop: I want to build rescue distro for own usage Nov 03 03:11:16 ok. I will try to build one for educationnal purpose : tes Nov 03 03:11:52 text processing + Net in schools on light hardware config. Nov 03 03:20:53 Hi. To get glibc to link on sh4 I needed to create a oetmp/work/glibc-2.3.5+cvs20050627-r0/libc/configparms with "no-z-defs=yes". Any suggestions on where to do this within oe? Nov 03 03:21:10 v8jlene: site/sh4-linux? Nov 03 03:24:04 hrw|work: So I should see what effect that has on the autoconf stuff and then try and put that entr in site/sh4-linux? Ok... gives me another option to look at. Thanks. Nov 03 03:24:59 v8jlene: site/* are config.cache files - keep all results of autoconf tests (iirc) Nov 03 03:30:43 XorA: Thx for /proc/sysrq-trigger info Nov 03 03:31:56 * XorA fails to alsa to work on git3 :-( Nov 03 03:31:56 chmod 222 /proc/sysrq-trigger in initscript make my machine quite maintanable Nov 03 03:32:03 hrw|work: cool Nov 03 03:32:45 atleast rebootable Nov 03 03:34:08 hrw|work: As far as I can tell configparms doesn't effect the output of configure. It's changes options that would normally go in config.make? Nov 03 03:34:40 v8jlene: dont know - I use already tested stuff usually Nov 03 03:38:26 hrw|work: Thanks. I don't see anyone else trying to build for sh4 so I think I'm on my own here ;) Nov 03 03:39:07 hrw|work: uclibc svn or 0.9.28 ? Nov 03 03:41:09 ade|desk: 0.9.28 Nov 03 03:41:27 ade|desk: gcc-cross-initial 3.4.4 fails on unable to find crt1.o Nov 03 03:41:40 ade|desk: so now I'm building glibc one Nov 03 04:09:59 lunch break! Nov 03 04:10:17 almost the best time of day Nov 03 04:15:36 god damnit Nov 03 04:15:43 another spam attack on the wiki Nov 03 04:26:47 hi Nov 03 04:38:35 hey magnet Nov 03 05:17:46 hi all Nov 03 05:20:12 hey dmunicio Nov 03 05:25:05 In conf/distro/openzaurus-3.5.4.conf, the line: DISTRO_VERSION = "3.5.4" is commented. Is this normal? Nov 03 05:25:49 Actually, that doesn't even matter - I don't seem to even be having this file read in at all Nov 03 05:26:52 I presume that setting DISTRO="openzaurus" (with no version) doesn't work? Nov 03 05:27:18 from looking at the check which is performed at the bottom of conf/distro/openzaurus.conf Nov 03 05:28:33 lardman|work: correct - versioned distro is much more usable Nov 03 05:28:47 Its probably not set so images in the wild don't get mistaken Nov 03 05:29:18 It's just that I don't seem to be getting a preferred version of something set in conf/distro/preferred-gpe-versions-2.7.inc Nov 03 05:29:29 so am trying to work back and work out why not Nov 03 05:29:57 lardman|work: Perhaps something else overrides it? Nov 03 05:30:22 RP: yes, I'll have a look through Nov 03 05:32:55 hmm, I pulled from koen's backup, it appears that it hadn't sync'ed the update (that's my guess) when I pulled, sorry for the noise Nov 03 05:42:49 he< Nov 03 05:44:58 j zecke Nov 03 05:52:03 What's the current OE version? I just pulled and updated and get: monotone: already up to date at 4741ce472626483ef14f4a9e8706f74f696beea4 Nov 03 05:53:26 hey zecke Nov 03 05:53:41 lardman|work: which branch? Nov 03 05:53:46 .dev Nov 03 05:53:56 no idea Nov 03 05:54:10 13:54 hrw@home:org.openembedded$ monotone heads Nov 03 05:54:10 monotone: branch 'org.openembedded.dev' is currently merged: Nov 03 05:54:10 4741ce472626483ef14f4a9e8706f74f696beea4 rpurdie@openembedded.org 2005-11-02T17:06:52 Nov 03 05:54:13 ah, so perhaps the change to gpe-conf hasn't made it into that branch? Nov 03 05:54:14 I don't work on .dev nowadays Nov 03 05:54:49 Shouldn't changes to the branch come back into dev? Nov 03 05:55:15 lardman|work: rather changes to branch should become from dev Nov 03 05:55:25 hrw|work: yeah Nov 03 05:56:08 hrw|work: you have .dev? Could go grep conf/distro/preferred-gpe-versions-2.7.inc for 'gpe-conf' please? Nov 03 05:56:14 s/go/you Nov 03 05:57:03 ok Nov 03 05:57:18 preferred-gpe-versions-2.7.inc:#PREFERRED_VERSION_gpe-nmf ?= "0.21" Nov 03 05:57:25 ops Nov 03 05:57:29 :) Nov 03 05:57:33 preferred-gpe-versions-2.7.inc:PREFERRED_VERSION_gpe-conf ?= "0.1.25" Nov 03 05:57:48 phew, not just me then Nov 03 05:58:28 so what's the fix for this issue then? the preferred version has been updated in the release branch, but not in .dev Nov 03 06:00:11 use the branch :) Nov 03 06:00:18 forever.... Nov 03 06:00:19 ? Nov 03 06:00:39 till the release or something Nov 03 06:00:43 the two will need to be merged at some point after the actual release Nov 03 06:00:58 no, they don't Nov 03 06:01:13 no, they don't, but that would be foolish Nov 03 06:01:23 when it's released we'll delete stuff from .dev Nov 03 06:01:28 replication of work, etc. Nov 03 06:04:55 is it only the conf files which are branched? why do I have a .bb for gpe-conf_0.1.26? Nov 03 06:12:45 what's the branch name? fam081oz354? Nov 03 06:12:52 oz354fam083 Nov 03 06:13:08 okay, and I can pull that from the usual places? Nov 03 06:15:29 yes Nov 03 06:15:45 cool, I've started it going Nov 03 06:15:47 thanks Nov 03 06:16:10 my dev images may as well benefit from the changes which have been made to the release brance Nov 03 06:19:14 hi Nov 03 06:30:02 koen: You should be submitting general changes to both branches IMO... Nov 03 06:30:22 Or only backporting changes to the release branch Nov 03 06:30:40 (unless they're only hacks for the release branch) Nov 03 06:31:15 yeah, I forgot to push the change to preferred versions Nov 03 07:02:19 can anyone look at this bug for me? http://bugs.openembedded.org/show_bug.cgi?id=435 Nov 03 07:03:38 oris_wolfbane: well Nov 03 07:03:48 oris_wolfbane: I did not bother to look into it yesterday Nov 03 07:03:57 oris_wolfbane: as you specify no versions at all... Nov 03 07:04:18 zecke, its uptodate from svn and monotone Nov 03 07:05:17 BBFILES wrong? Nov 03 07:17:52 oris_wolfbane: well did you install a old bitbake Nov 03 07:18:01 oris_wolfbane: and are now running one from svn? Nov 03 07:18:02 Ciao all Nov 03 07:18:46 zecke, i deleted the entired drectory and then got the new one from svn Nov 03 07:18:49 hi Pigi Nov 03 07:19:04 hi hrw|work ! Nov 03 07:19:09 oris_wolfbane: well Nov 03 07:19:44 find / -name 'data.py*' Nov 03 07:19:48 find / -name 'data.py*' | wc -l Nov 03 07:24:50 any idea when openembedded.org will be back AFAIK ? Nov 03 07:25:36 Pigi: when kergoth changes the dns Nov 03 07:25:41 Pigi: try oe.handhelds.org Nov 03 07:25:43 heh Nov 03 07:30:08 zecke, 38 Nov 03 07:33:37 Which version is compiled when I only give the Name of the package but no version number? Nov 03 07:34:20 Cobelius: the one with the highest preference Nov 03 07:34:54 Cobelius: with is the highest version number, unless some .conf specifies that a certain version has to be used Nov 03 07:35:08 s/with/which/ Nov 03 07:35:38 suppose I have foo_1.0.bb and foo_1.1.bb Nov 03 07:36:16 bitbake will build 1.1 by default unless there is a PREFERRED_VERSION_foo = "1.0" somewhere, of the 1.1 .bb has DEFAULT_PREFERENCE = "-1" inside Nov 03 07:36:30 s/of/or/ Nov 03 07:37:18 assuming both files are inside BBFILES Nov 03 07:37:42 okay thank you Nov 03 07:39:30 I suspect there are other ways to manipulate it, but these are the ones I know Nov 03 07:42:46 oris_wolfbane: 38? so much Nov 03 07:42:49 oris_wolfbane: which paths? Nov 03 07:47:40 zecke, just ran it again, now its 43, http://pastebin.ca/27488 Nov 03 07:48:21 kidding me? Nov 03 07:48:40 so you do not even have bitbake installed? Nov 03 07:50:12 zecke, i followed the instructions on the wiki to set the PATH Nov 03 07:50:34 oris_wolfbane: what has this to do with having bitbake installed? Nov 03 07:50:55 Running BitBake without installing it Nov 03 07:50:55 This is the preferred method of running bitbake. Simply set the PATH variable so that the tools are accessible (eg: export PATH=/stuff/bitbake/bin:$PATH). You will need to do this every time you start a terminal for development. (You can set this in your login profile, or perhaps use a script to set the necessary variables for using BitBake.) Nov 03 07:51:10 so i need to install it? Nov 03 07:51:21 oris_wolfbane: well, if you don't even find data.py you do not have any traces of bitbake Nov 03 07:51:27 oris_wolfbane: where did I say that? Nov 03 07:51:53 oris_wolfbane: start bitbake with python -v /path/to/your/bitbake whatever-you-wanted to create Nov 03 07:52:15 hrw|work: isn't it nice when people say 'this is so cool, I want to have it' and we can respond with 'it's been in OE for months' Nov 03 07:53:31 koen: its nice Nov 03 07:54:06 let's wait for the pdaX morons to start flaming in that thread Nov 03 07:55:31 koen: URI? Nov 03 07:56:22 http://www.oesf.org/forums/index.php?showtopic=15881&hl= Nov 03 07:57:07 oris_wolfbane: any updates? Nov 03 07:57:26 just ran that, http://pastebin.ca/27490 Nov 03 07:58:04 import bb.data # precompiled from /usr/local/lib/python2.4/site-packages/bb/data.pyc Nov 03 07:58:14 well you at least lied twice ;) Nov 03 07:58:42 zecke, how so? Nov 03 07:58:45 well your find should have spotted /usr/local/lib/py.../data.py Nov 03 07:58:55 and second you already installed a version of bitbake Nov 03 07:59:30 then how come the find didnt spot it? i could have installed bitbake long ago and forgotten about it Nov 03 07:59:32 do bitbake --version for me please Nov 03 07:59:55 this whole thing is a good reason why do put 'versions' into a bug report Nov 03 08:00:09 you think you run one version, but you simply don't ;) Nov 03 08:00:41 and it should not sound as harsh as it may sound Nov 03 08:00:58 you can either set the PYTHONPATH (or was it PYTHON_PATH) Nov 03 08:01:06 or remove the old bitbake installation Nov 03 08:01:22 BitBake Build Tool Core version 1.3.2, bitbake version 1.3.2 Nov 03 08:01:43 zecke, how do i remove the old one? Nov 03 08:02:28 sudo rm -rf '/' where '/' could be /usr/local/lib/python2.4/site-packages/bb Nov 03 08:02:48 heh Nov 03 08:02:58 trying to punish lazy copy/pasters? Nov 03 08:04:07 koen: I didn't want to write lambda Nov 03 08:04:24 lambada! Nov 03 08:04:34 (\x -> (x+1)) Nov 03 08:04:47 (\x -> (x+1))10 Nov 03 08:04:54 gosh I love functional programming Nov 03 08:05:08 zecke, ive uninstalled as per your instructions and still no luck Nov 03 08:05:09 funcprog is pretty fun to do Nov 03 08:05:15 *without assignments* Nov 03 08:06:28 oris_wolfbane: well rerun the python -v thingie Nov 03 08:06:35 oris_wolfbane: and look from where it pulls data.py Nov 03 08:12:16 zecke, thanks alot, thats sorted it Nov 03 08:12:29 zecke, you have a paypal donate page? Nov 03 08:17:00 oris_wolfbane: not any more Nov 03 08:17:21 oris_wolfbane: I do this here free of charge, without even expecting to get anything out of it Nov 03 08:17:28 oris_wolfbane: I'm glad it works Nov 03 08:17:35 zecke, so am i Nov 03 08:17:48 zecke, thanks again Nov 03 08:18:10 oris_wolfbane: get back to us, with your next error Nov 03 08:18:17 lol Nov 03 08:18:21 oris_wolfbane: you might want to close the bug report if you want to Nov 03 08:18:29 s/your next/the next/ Nov 03 08:18:30 anyway, bitbake-mono is much faster Nov 03 08:18:53 koen: wasn't mono -v python bitbake already faster? Nov 03 08:19:41 no, C# is much faster Nov 03 08:19:53 Forth rule anyway Nov 03 08:36:02 hi Nov 03 08:36:41 hi Bernardo Nov 03 08:37:05 RP: I've been looking at that pdf you gave me the link Nov 03 08:38:15 (and found out I'm more a ignorant that I thought...) Nov 03 08:38:17 :) Nov 03 08:39:10 on the akita, it is using 4 bits sdio interrupt mode, right? Nov 03 08:39:34 Bernardo: welcome Nov 03 08:44:13 Bernardo: 1 bit sd mode ;-) Nov 03 08:44:27 We don't do 4 bit mode or sdio yet Nov 03 08:46:25 RP: ok... Time to re-read the doc... :) I was thinking the timeout and lost interrupt might be because of the extra 2 cycles needed for 4bit multi-block Nov 03 08:47:22 Bernardo: I don't claim to fully understand SD myself and have not read much of the doucment. I think the key point was there can be two interpretations of the timeout and we only use ones of them Nov 03 08:51:22 hi drw Nov 03 08:51:33 hrw|work: morning Nov 03 08:51:51 dark morning even Nov 03 08:59:31 somebody destroyed the Manual in the Wiki Nov 03 08:59:37 can somebody redo it? Nov 03 09:00:02 bleh Nov 03 09:00:04 | spectrum_cs.c:588: warning: implicit declaration of function `handle_to_dev' Nov 03 09:00:05 | spectrum_cs.c:588: error: invalid lvalue in unary `&' Nov 03 09:00:12 Cobelius: you can?! Nov 03 09:02:02 I should be ok now Nov 03 09:04:26 thank you Nov 03 09:16:07 anyone can tell me how can I set to set locale in oe...? I don't want all the locale packages compiled Nov 03 09:18:01 RP: ok, I'll keep studying it... Nov 03 09:20:05 RP: I've sent you the latest wm97xx touch screen driver for do13 to try on the tosa. Hopefully it will be ok. Nov 03 09:21:07 incinerator: going to EdLug tonight ? Nov 03 09:21:33 indeed Nov 03 09:21:42 heho, who are you? Nov 03 09:21:53 Liam Nov 03 09:21:58 well, coool Nov 03 09:22:08 didn't know there was an edlugger involved in the project.... Nov 03 09:22:27 I'll probably be at the Auld Hoose, as well Nov 03 09:22:38 I'm involved with the audio and touch screen side of things Nov 03 09:22:50 I'll also be at the Auld Hoose Nov 03 09:24:01 cool Nov 03 09:24:30 I haven't had the time to try out the oz rc yet, but I'll probably do that at the weekend.... Nov 03 09:24:42 if it's not good enough, I'll install the new cacko 1.23 ;-) Nov 03 09:32:42 03koen 07org.oe.oz354fam083 * rbc40dc55... 10/: familiar 0.8.3: prefer orinoco 0.13e since 0.15 doesn't compile Nov 03 09:33:13 later all Nov 03 09:41:30 koen: tried 0.15 for 2.4 kernels? Nov 03 09:42:05 will OE run on a 64bit athlon? Nov 03 09:42:43 hrw|work: yes, that have the error I pasted 40 minutes ago Nov 03 09:43:50 Cobelius: in 64 bit mode the compiled gcc could segfault Nov 03 10:20:35 What is the variable PREFFERED_PROVIDERS in local.conf doing? Nov 03 10:21:47 if multiple packages provide the same thing (e.g hotplug-ng and linux-hotplug both provide 'hotplug') you can select which one you want to use Nov 03 10:22:46 okay, thank you Nov 03 10:24:51 who added udev 071? it stars from PR="r5" instead of r0 Nov 03 10:24:54 starts Nov 03 10:26:10 hrw|work: http://cia.navi.cx/stats/project/openembedded/.message/6255755 Nov 03 10:26:29 rp.. Nov 03 10:37:11 RP: ping Nov 03 10:38:09 lrg: ping Nov 03 10:40:21 lrg, RP: turn on c7x0, connect via ssh over usbnet, run alsamixer, bump playback to 100 - noise from speaker even without doing something on c7x0 Nov 03 10:43:03 lrg, RP: if I turn on 'Output Mixer Line' noise will dissapear Nov 03 11:01:45 cu Nov 03 11:15:50 night Nov 03 11:28:08 hey guys, ive tryed to compile both nano and opie-image, and i get an error both times about noe being able to login cvs because anoncvs:anoncvs doesnt exsist, any ideas? Nov 03 11:28:19 s/noe/not Nov 03 11:35:40 Should bitbake have PN as an override when working out PREFERRED_VERSION? Nov 03 11:36:18 http://www.rpsys.net/openzaurus/temp/bitbake_preferred_pn_override-r0.patch ... Nov 03 11:41:22 03koen 07org.oe.oz354fam083 * r4cca22b3... 10/: bitbake.conf + other _cvs files: change cvs.hh.org to anoncvs.hh.org Nov 03 11:43:03 isn't pn always in overrides? Nov 03 11:43:20 * koen knows nothing about python or bb internals Nov 03 11:45:34 koen try setting CVSDATE_somepackage and then PREFERRED_VERSION_somepackage="cvs-${CVSDATE Nov 03 11:45:59 }" Nov 03 11:47:02 I know PREFERRED_VERSION_somepackage="cvs-${CVSDATE}" doesn't work Nov 03 11:47:15 this is why and the above patch fixes that Nov 03 11:47:29 ah, I see Nov 03 11:47:32 good job! Nov 03 11:47:45 I'll bugzilla it Nov 03 11:56:59 ~lart me Nov 03 11:56:59 * ibot gives koen a good seeing to Nov 03 12:01:19 03koen 07org.oe.dev * r9864f624... 10/: libmimedir: add 0.3.1 release and DP down the cvs one Nov 03 12:01:36 If a student says 'I always link to /usr/lib/libbluetooth.so, so why should I change the Makefile' one more time, I'll force him to eat the ipaq Nov 03 12:02:14 koen: Is there a webcam handy? :) Nov 03 12:02:32 :) Nov 03 12:03:31 * koen gives his mini-ipaq-cluster another try Nov 03 12:04:24 ~lart nfs Nov 03 12:04:25 * ibot chops nfs in half with a free Solaris 7 CD Nov 03 12:11:19 would ipkg benefit from using sqlite to store its metadata rather than plain files? Nov 03 12:13:59 I think so Nov 03 12:15:03 it would lose the hackable-text-file feature, though Nov 03 12:15:08 seems like it would be a lot quicker for most things, since it wouldn't have to parse everything all the time Nov 03 12:15:43 seems better to have decent performance and let the hackers use a little sql, or write a front-end to tweak it by hand Nov 03 12:15:59 indeed Nov 03 12:16:11 You would still need to walk the files looking for changes Nov 03 12:16:25 unless you added a seperate 'cache' mechanism Nov 03 12:16:40 which files? Nov 03 12:16:49 the .bbs Nov 03 12:16:53 on the device you only have config files, lists and the status Nov 03 12:17:07 * CosmicPenguin somehow read OE instead of ipkg Nov 03 12:17:07 duh Nov 03 12:17:11 heh Nov 03 12:17:32 I either need more caffine or less crack Nov 03 12:17:37 probably both Nov 03 12:18:47 more of both! Nov 03 12:19:31 CosmicPenguin, installing FC4 x86_64 on an athlon machine we got from newegg Nov 03 12:19:41 excellent Nov 03 12:19:47 except for the FC4 part... :) Nov 03 12:20:06 good evening ! Nov 03 12:20:12 heh Nov 03 12:20:17 I know how to use it :) Nov 03 12:20:30 FC4 on x86_64 work good ;) Nov 03 12:20:47 hopefully oe works on it too :) Nov 03 12:20:49 FC working good? Nov 03 12:21:21 for me yes ... Nov 03 12:21:44 maybe cause i know the beast since long time ;) Nov 03 12:23:46 btw Hi koen !!! Nov 03 12:23:53 hey gremlin[it] Nov 03 12:24:45 at work i'm working on a communication program to field devices (PLC S7 from Siemens) Nov 03 12:25:20 at now using Simatic library but is really a bad library .. soon all with GPL software (libnodave) ... :) Nov 03 12:25:31 so i can run it also on an arm board ;) Nov 03 12:32:42 which kernel version does the ipaq-pxa270 build in org.openembedded.oz354fam083 use? Nov 03 12:36:02 hey Nov 03 12:36:10 hi zecke :) Nov 03 12:36:33 ~hail zecke Nov 03 12:36:35 * ibot bows down to zecke and chants, "I'M NOT WORTHY!!" Nov 03 12:37:06 thanks Nov 03 12:37:26 koen: I figure both GPE and Opie got a 'problem' due Maemo Nov 03 12:37:55 everybody is going 'wow maemo is cool'? Nov 03 12:38:14 koen not me Nov 03 12:38:14 right, If Nokia doesn't blow it Nov 03 12:38:16 and forgetting gpe+opie? Nov 03 12:38:32 it could be the Linux Mobile Platform Nov 03 12:38:57 hi zecke Nov 03 12:39:24 hildon + input methods is cool Nov 03 12:40:01 koen: well that is possible with Qtopia too ;) Nov 03 12:40:02 the browser, newsreader, mail app, video player and audio player suck Nov 03 12:40:22 koen: I was impressed when I entered the HEX WPA key Nov 03 12:40:26 koen: and the keyboard changed Nov 03 12:40:26 what rest ? Nov 03 12:40:29 the browser is opera Nov 03 12:40:47 woglinde: and it sucks big mammal ass Nov 03 12:41:19 koen hm, it is as stable as konqueror on the simpad Nov 03 12:41:48 it crashes everytime the sun shines? Nov 03 12:45:13 zecke hi Nov 03 12:45:14 zecke hi Nov 03 12:45:14 zecke hi Nov 03 12:45:18 *g* Nov 03 12:47:20 hi Nov 03 12:47:43 RP, mickeyl: 2.6.14-git3-r2 works ok with kernel-module-hostap Nov 03 12:51:56 when will the oe website return? Nov 03 12:52:35 Crofton: oe.handhelds.org Nov 03 12:53:25 thanks Nov 03 12:53:32 will openembedded.org return? Nov 03 12:56:26 will Nov 03 12:56:32 when kergoth updates the dns Nov 03 12:58:09 * Crofton searches for a trout Nov 03 12:58:42 open("/usr/local/projects/openzaurus/build/opie/lib/v5l/fast-mult/half/libcrypt.so.1", O_RDONLY) = -1 ENOENT (No such file or directory) Nov 03 12:58:46 nice try on zaurus ;) Nov 03 13:07:18 Ciao all Nov 03 13:08:31 ciao Pigi Nov 03 13:08:43 mattina Pigi Nov 03 13:09:44 hi hrw. It's not morning here. At least "buonasera" :) Nov 03 13:09:52 ~ugt Nov 03 13:09:56 ugt is probably Universial Greeting Time. Created in #mipslinux, it is a rule that states that whenever somebody enters an IRC channel it is always morning, and it is always late when the person leaves. The local time of any other people in the channel, including the greeter, is irrelevant. Nov 03 13:09:57 ~ugt Nov 03 13:10:01 heh Nov 03 13:10:30 hrw, apologize then. Always something to learn :) Nov 03 13:10:35 np Nov 03 13:10:56 sharp-compat-libs suxx - dont support freetype Nov 03 13:20:59 hi hrw|tv, Pigi Nov 03 13:21:13 hi RP ! Nov 03 13:21:18 hrw|tv: Its good to know the in kernel hostap works. We should probably switch to that... Nov 03 13:21:39 Pigi: I'm afraid I've filed a bug against ipkg... Nov 03 13:21:50 argh... let me see Nov 03 13:22:44 Pigi: do you think storing ipkg metadata in an sqllite db would be a good idea? seems like it might make it go faster Nov 03 13:24:52 mreimer, I'm not really sure about that, as the biggest problem in actual ipkg is that he needs to read way too much data Nov 03 13:25:19 and I'm not sure that an sql engine would be more performant than reading from a flat file Nov 03 13:25:57 I was thinking that ipkg might have to read less data because it wouldn't have to read the whole flat file to get what it wants Nov 03 13:26:04 but you know how it works :-) Nov 03 13:26:59 the problem is a bit different. Ipkg ATM needs to read the whole data because he need to construct all the structure to check for dependencies and so on Nov 03 13:27:48 you can make it a bit faster by removing Source: and Maintainer: field from the lists Nov 03 13:27:49 I'm not sure about how much pointer lists are used by ipkg but, believe me, they are tooooooo much Nov 03 13:28:40 that's what I asked some times ago to oe builder/maintainer, they are unusefull information for ipkg itself Nov 03 13:28:58 how about recording the dependencies in the db? then you could just do "SELECT * from dependencies where pkgname = 'foo'" Nov 03 13:29:00 and are really big piece of file that needs a lot of realloc Nov 03 13:29:50 that could ( and probably would ) break the compatibility with dpkg, and also, how would you put that info's in db in a "ipkg update" for example ? Nov 03 13:30:20 yeah, gets messy. it might end up being as slow as monotone Nov 03 13:30:36 and what about reconstructing the db ?, and what if you choose to save some space by storing the db in a temporary space ( as actually is ) ? Nov 03 13:31:16 every rebbot you need to reconstruct the initial db, and also you need to keeb in sync that db with " status" file and /usr/lib/ipkg/info/* files Nov 03 13:31:24 s/keeb/keep Nov 03 13:31:46 you wouldn't need the status file Nov 03 13:32:15 you would, believe me. Is where ipkg stores information about installed, configured or unconfigured packages. Nov 03 13:32:33 Status file is ATM the most important information ipkg has about the whole installation Nov 03 13:32:35 all that would get stored in the db Nov 03 13:33:19 I know. I used to be an sql programmer years ago, but in our case I don't think we will get good things by using a db. Nov 03 13:33:26 but who knows :) Nov 03 13:33:42 I defer to your assessment, since you know how ipkg works :-) Nov 03 13:33:56 heh Nov 03 13:34:22 I'm not sure that after more than one year of fixing bug I do really know how does ipkg works ;) Nov 03 13:34:29 :-) Nov 03 13:35:45 At least you have a very healthy fear of it Nov 03 13:35:53 heh Nov 03 13:53:06 Pigi: What is needed is to eliminate the need for downloading, storing and parsing the meta-data on the device...a true client/server system where the server figures out all of the dependency/arch stuff and just feeds the device a series of packages to install (in dep order of course) Nov 03 13:53:49 (something I have been thinking about for about a month now) Nov 03 13:54:50 well, actually I haven't seen nothing like this yet. It could be an option but then every one maintaing a sel feed should "upgrade to that sistem". Nov 03 13:55:18 Btw yes it could be an option, but also writing a new package manager could be an option :) Nov 03 13:55:48 Pigi: could be :) Nov 03 13:56:30 ( this is something I'm thinking since when I started fixing bugs on ipkg ). Nov 03 13:58:10 new package manager...new bugs :) at least you would be more familiar with the internals Nov 03 13:58:24 that's the idea. Nov 03 13:58:49 I'm still learning a bit on how ipkg/dpkg works, the I could choose to make the bigstep Nov 03 14:00:09 any package manager is simple in concept, it's just all of those little details which are a pain and make it messy Nov 03 14:03:42 what I do really need Nov 03 14:04:15 atm is someone that help me create a new set of bb files for a gpe image that has less option then the original :) Nov 03 14:30:44 zecke: busy? Nov 03 14:36:37 hi pb_ Nov 03 14:36:48 hi pigi Nov 03 14:38:01 pb_: Does http://bugs.openembedded.org/show_bug.cgi?id=436 look like the correct thing to do? Nov 03 14:41:07 need a reboot. bbl Nov 03 14:45:13 re Nov 03 14:45:22 wb ... reboot velocissimo Nov 03 14:45:29 ... so fast reboot ;) Nov 03 14:45:54 heh I was playng with udev, while installing a new image and lost all my /dev/pty* :) Nov 03 14:46:09 RP: hey Nov 03 14:46:15 RP: busy and tired ;) Nov 03 14:46:30 hi zecke Nov 03 14:46:52 zecke: Ok. Was just wondering about the patch in the bug I mentioned above but it can wait Nov 03 14:47:16 RP: ah, well I clicked on the link from the mail... (bugzilla mail) Nov 03 14:48:43 the big bad update_data ;) Nov 03 14:49:11 and del localdata... ;) I have never used new, why del Nov 03 14:49:31 well that sounded negative where it should not sound that way Nov 03 14:49:36