#qi-hardware IRC log for Friday, 2011-02-18

wpwraktuxbrain_away: alright :)00:02
rjeffriestuxbrain_away they will run 1,000 all at same time. after you oredr the first 500 they'll sell the second 500 at a lower price.00:11
rjeffriesyou may be better off ordering the 1,000 have they quoted prices at quantity 1,000 ?00:11
rjeffriesI am almost (not quite) ready to get a US quote, whuch muht be interesting. I would assume my qupute and yours should be similar price plus or minus 5 percent00:13
Action: Jay7 need keyboard manufacturer :)00:15
Jay7I definitely need hybrid of MS NEK 4000 with IBM trackpoint instead of Zoom in middle00:15
Jay7and with better mechanic.. even for $100 per unit..00:16
Jay7or may be $200 even..00:16
panda|x201xiangfu, any news about 02-14 release?01:46
wolfspra1lpanda|x201: xiangfu ran into a bug with the sound module due to kernel upgrade01:47
panda|x201wolfspra1l, ah, long time no see, wolfgang01:47
wolfspra1lreally? i'm here all the time01:48
zrafaJay7: have you gotten kernel+kexecboot(initramfs) on kernel partition in nn?01:50
Jay7zrafa: not yet, coming soon :)01:51
zrafaJay7: what is the status?01:51
Jay7this is next item in todo list :)01:51
Jay7I should merge linux-jlime-nanonote into linux-kexecboot recipe in OE01:53
Jay7to be able to build kexecboot kernel + initramfs01:53
Action: Jay7 -> sleep02:01
xiangfupanda|x201: wolfspra1l just fixed.  needs enable some option in kernel_menuconfig02:09
wolfspra1lah that was easy :-)02:16
zrafawolfspra1l: wpwrak http://www.nokiaplanm.com/02:21
zrafacheck below after the spec , there are pics of the new product02:21
zrafaif you need more, here the index of the alternatives plans of nokia : http://nokiaplans.com/02:23
wolfspra1lzrafa: planm? I read about planb the other day, the engineer's revenge02:24
wolfspra1lthat one for sure has no chance for success, let's see planm now...02:24
wpwrakplanm looks good ;-))))02:25
zrafawpwrak: wolfspra1l :) .. plana would be stable I guess02:26
wolfspra1lthe text is lame, but the pic pretty cool!02:33
wolfspra1lthat's how the first Qi smartphone will look like02:34
wolfspra1lwell those plans are fun at leat02:36
wpwrakwith all the unhappiness, i wonder how long until people act on it. tunesia, egypt, nokia ? :)02:47
steve|mOpenmokia.. hehe02:49
wpwraksteve|m: the domains still seem to be available (-:C02:52
wpwrakhmm, seems that i have a sourcing crisis at my hands :-( adhesive tape is running low ... let's see if i can make it through the night with what's left ...02:56
steve|muh oh, I smell bad supply chain management.. just-in-time sourcing :P02:57
adamw_Is there anyone know where's OM wikireader's schematic?02:58
wpwrakyeah. i actually noticed something like 2-3 days ago, but keep on forgetting to buy new one02:58
adamw_or it have never been publicly at all?03:00
wpwrakadamw_: i don't remember seeing it published03:01
adamw_wpwrak, me either. I just want to know & find it.03:02
wpwrakadamw_: btw, why on earth ? can't you design something better over lunch ? :)03:27
kristianpauldammit, i dont know what happen i just get home and my home router/swich was off... by 8 hrs.03:32
adamw_wpwrak, man! forget it.03:34
rejonhi guys03:39
kristianpaulhey rejon03:41
wpwrakrjeffries: i think you underestimate the extra value you pay for when producing things at home (as opposed to the far east) ;-)04:27
wpwrakrjeffries: but i'm kinda curious where this will lead. may learn a thing or two :)04:30
rjeffrieswpwrak not sure why you think this UBB fab would go offshore. I expect in these small qty it will be local. However, if specs are solid and venodr quality is known to my middleman, then risk of China fab seems ok.04:36
rjeffrieswha made you think it will go to China?04:37
wpwrakrjeffries: no, what i mean is that you can expect the pcb fab to communicate any small problem to you. they will factor some communication overhead into their process.04:39
wpwrakrjeffries: so you don't need to plan for an "open loop" process. that would be needed if you went to china (well, unless you know the place you're dealing with well and have established good communication)04:40
rjeffriesthe guy I plan to go through (as my interface to fab) has done multiple fabs for a good friend, those are far more complex than UBB04:41
rjeffriesalso we worked with him when I was at the company I worked at for 8 years. he is a known quantity amd reliable04:42
wpwrakrjeffries: yeah, almost everything is more complex than UBB ;-) just be careful about the cutting of the board. that's the hairy bit.04:42
rjeffriesyup that worries me. I wonder if the physical dimension spec is stated as clearly as it must be04:43
rjeffriesIcertainly understand that standard v-scorinbg is a non starter04:43
rjeffries[askshimself, why the hell did I even THINK about doing this???]04:44
wpwrakyou have about 100-200 um of tolerance before things get really bad. but yes, i don't see a way for v-scoring to meet this without trouble. a mill can do it, but needs some mechanical support. preferably better than the double-sided tape on FR2 i use ;-)04:45
wpwraki think a laser cutter would be ideal. but i don't know how common they are at pcb fabs. in any case, they have *something* that can cut boards precisely. the thinness of the board should already eliminate places with medieval technology04:47
wpwrakif the pcb fab is attentive and uses a cnc mill, they should complain about the concave corners, because a mill can't make such corners :) but it can approximate them in several ways.04:50
wpwrakif they have a laser cutter with a fine enough beam, they won't care, because it'll all be within the tolerances04:50
wpwrakalso, if they just do it with a reasonably fine "standard mill", it'll be alright (i use a 35 mil bit, which is sufficient. if they use something larger, then the board may need some manual trimming to lock nicely. it will insert/remove no matter how big the bit is, because there are only convex corners involved there, and you can mill those "perfectly")04:53
kyakxiangfu: hi!05:53
kyakxiangfu: i wasn't able to build pango.. so no config_full -\05:54
kyakxiangfu: however, i don't experience any sound issues though i've been using for a while now...05:55
xiangfukyak: the option CONFIG_I2C not enable in kernel config. by enable CONFIG_I2C, sound works fine.05:56
xiangfuI have test all the apps in gmenu2x. all works fine.05:56
kyakok, that's great news05:56
tuxbrain_awayrjeffries:(fab qty) nop, they will run a 528 run,not 1000, but on second batch a lot of initial cost will not have to be payed again so first batch is expensive than all the rest.05:57
kyak# CONFIG_I2C is not set05:57
kyakhm.. why i don't see ths issues?05:57
xiangfukyak: the 'dmesg' not report any error?05:58
kyaki reflashed couple of days ago.. maybe it's a really new bug05:58
kyakxiangfu: i didn't have a look in dmesg.. just the sound is working05:58
xiangfukyak: ok. anyway. let'us start with those images :) : http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-02182011-0648/05:59
xiangfuI will download and do fast test again . then send email to list.05:59
xiangfuthis time is real release ;-)05:59
kyakxiangfu: ok. going to try that at home05:59
larsckyak: when you build your sound modules did you had add openwrt packages selected?05:59
kyakwill update to 512 Mb rootfs :)05:59
kyaklarsc: which modules exactly?06:00
larscwell the ones used for sound06:00
kyaki'm not sure, i never thought about it. What should i look for in config?06:01
larscIf I2C was enabled during build, it is also a runtime requirement for the asoc-core.06:02
larscIf it was not it is not06:02
larscso if you build your sound modules while the i2c modules were selected, you have to load the i2c modules before you can load the sound modules06:02
kyakwell, since i don't have i2c enabled, i don't have to load it..06:04
kyaksound modules just load06:04
kyaklarsc: there was a question about using monolythic kernel on malining lists. .What do you think about it?06:05
xiangfuthen why the default build about sound module have i2c enable?06:05
larscI think we should just add the spi and i2c modules as a dependcy to sound-soc-core, that should avoid any further problems06:07
larsckyak: when?06:07
kyakxiangfu: maybe your full build has something like i2c-tools enabled, that makes the i2c avaialble in your build environment, but then there is no i2c on target due to lack of CONFIG_I2C?06:07
wolfspra1linteresting: http://www.h-online.com/open/news/item/Microsoft-bans-free-software-from-Windows-Phone-Marketplace-1191524.html06:08
kyaklarsc: one sec, it was an em-ail from xiangfu06:08
wolfspra1lso gpl software seems to be excluded from iphone/ipad, as well as wp7.06:08
wpwrakwolfspra1l: we can just lean back and let them do the marketing for us ;-)06:10
kyaklarsc: http://lists.en.qi-hardware.com/pipermail/discussion/2011-January/006836.html06:12
larsckyak: i know i had very good reasons to them into modules, but i actually can't remeber what these reasons were.06:14
kyakhehe :)06:14
kyakmaybe we could just make it into the kernel and the reasons why?06:14
kyak*find the reasons06:14
kyakxiangfu: btw. my wife beat the supertux, and there is a nasty segfault during credits on the last (26th) level :)06:16
kyakmaybe it's intended, like a victory prize :)06:16
qi-bot[commit] kyak: w3m: fix build with config.full_system http://qi-hw.com/p/openwrt-packages/880c97106:32
wpwrakatben: now 54 vias. atusb: 64 vias. grmbl.07:08
wolfspra1lsseago: reading the backlog - you are a new Ben owner - welcome! happy to see you here.07:11
xiangfukyak: oh. I even not pass the level 1 :)07:30
kyakxiangfu: :)07:31
xiangfukyak: the reflash_ben.sh progress show not correct when flash bootloader: http://pastebin.com/UX8TEHu807:32
xiangfuthe bootloader always show 1/2 . in fact it's finished.07:33
xiangfukyak: thanks for fix w3m. :)07:33
kyakxiangfu: guess it could be due to those u-boot errors you mentioned yesterday.. i'll have a look.. basically these #07:34
kyaknumber - Done:  1/2 - are taken from usbboot output07:35
kyakit reports 2 blocks to be flashed, but only flashes one?07:35
xiangfuno. it's indeed flashes two.07:36
kyakxiangfu: ok. can you upload the log somewhere?07:38
xiangfukyak: http://downloads.qi-hardware.com/people/xiangfu/tmp/log.txt07:40
kyakxiangfu: ok. i'll have a look07:41
kyakxiangfu: btw, i made the latest NanoMap from git compile. But the amount of overrides in Makefile is enourmous07:41
xiangfukyak: thanks for fix w3m , will try to build it today evening07:41
kyakthis needs to be sorted our with mirko somehow.. i don't know why i always have to override variables when porting qt4 apps07:42
kyakxiangfu: np!07:42
qi-bot[commit] kyak: NanoMap: update to the latest git http://qi-hw.com/p/openwrt-packages/f3588d807:42
kyakmirko: xiangfu -^^07:43
kyakand this has to be done in all *.pro files, as you can see07:44
kyakxiangfu: reflash_ben.sh. Got the problem:07:59
kyakif [[ "$ILINE" =~ Checking\ [[:digit:]]+\ bytes\.\.\.\ Comparing\ [[:digit:]]+\ bytes\ -\ SUCCESS$ || "$ILINE" =~ Checking\ [[:digit:]]+\ bytes\.\.\.\ no\ check\!\ End\ at\ Page\:\ [[:digit:]]+ ]]; then07:59
kyakthis is the condition to increment the number of Done blocks08:00
kyakhowever, in your case, you had Checking 524288 bytes... Comparing 524288 bytes - FAIL at off 2048, wrote 0x0, read 0xff08:00
kyaktherefor, only 1/2 blocks got flashed with SUCCESS08:00
kyakxiangfu: do you think the number of Done blocks should also increase when the block has FAILed to write?08:01
kyakthis can be done easily, but maybe it is better to leave it like it is, so that people can see how many blocks FAILed?08:02
xiangfuyes. let's leave it. we are not hiding problem. we need prompt problem. and document it well :)08:03
kyakgood :)08:04
kyakyou've also detected an undocumented feature of reflash_ben.sh ;)08:04
xiangfukyak: the 'freedroid' only works with '-q'08:25
xiangfu'emacs' can not start in gmenu2x. but works find in terminal.08:28
kyakxiangfu: freedroid - right, we discussed that already.. must have swap enabled for the sound to work08:36
kyakthis is the reason you decided to remove it from gmenu2x08:36
kyak(maybe re-add it with params=-q?)08:36
xiangfuyes. I would like to just make it work by adding "params=-q"08:37
kyakit's a cool game actually08:37
kyaka little bit too heavy for Ben, but ok08:38
qi-bot[commit] kyak: NanoMap: install monav libs http://qi-hw.com/p/openwrt-packages/193f1d308:41
kyakxiangfu: if i'm not mistaken, nielsle is the developer of NanoMap?08:42
wolfspraulkyak: yes, correct [niels]08:42
kyaki found some undocumented features in NanoMap, including this monav support.. Would be great to have it explained a little bit :)08:43
kyakif you press "L" you can add various layers.. include time and battery08:43
wolfspraulI'm a super big fan of NanoMap, but I'm not sure how much the (limited) Ben is still Niels focus. some of his latest stuff looks like he is developing on a notebook or so.08:43
kyaklooks very good, but not documented08:43
kyakmaybe we can poke niels sometime :)08:44
wolfspraulfor me it needs an index, so I can search for street names, subway lines, etc.08:44
wolfspraulthen there needs to be a way to quickly load a map of some city onto the device08:44
xiangfukyak: http://downloads.qi-hardware.com/software/images/NanoNote/Ben/testing/2011-02-16/08:44
xiangfufinally I got a version. two known issue. 1. emacs not start in gmenu2x, 2. freedroid needs '-q'08:45
kyakxiangfu: ok. Will give it a try as soon as i can08:45
kyakwolfspraul: btw, i used NanoMap on my laptop to pre-download necessary data (iamge tiles). I then uploaded it to Ben08:46
kyaki found it faster/more convenient08:46
larsc"Play as Tux in a quest to save the world from the murderous rebel bots who know no mercy" - haha, interesting :)08:50
xiangfuhttp://downloads.qi-hardware.com/software/packages/NanoNote/Ben/latest/ and http://downloads.qi-hardware.com/software/mirror-openwrt-sources/ also update.08:50
kyakxiangfu: good. so opkg update && opkg install package_name should work?08:51
xiangfukyak: sure.08:51
xiangfuof cause ;-)08:51
kyakxiangfu: were there any packages that failed to build? like, php?08:52
kyakno, php is there :)08:54
xiangfukyak: all the package in config.full_system (but w3m, plplot) are compile fine.08:54
xiangfuthere are some package not compile. next time I will record them to a file.08:55
kyaki still don't understand how you managed to build pango.. might wanna have a look at logs08:55
xiangfuwhen run "make" without V=99. there are a list of package which not compile08:55
kyakxiangfu: oh, maybe we can copy w3m to packages once it's built :)08:56
kyakso people can install it even if it's not in rootfs08:57
kyakah, that would require updating of Packages.gz... nevermind08:57
xiangfuyes. it needs update Package.gz.08:58
kyakactually, it would be a one step forward to "rolling releases" if we do that09:05
kyakif we always keep the packages/NanoNote/09:06
kyakBen/latest up to date09:06
kyakthen people can just update continously09:06
xiangfuyes. there are two issue for now.09:17
xiangfu1. those file that overwrite openwrt files  openwrt-xburst.git/files/etc/*09:17
xiangfu2. linux can not reflash the bootloader.09:18
xiangfufor 1). when can create a package, that include those file. then add small script file at /etc/uci-defaults for fix that.09:18
xiangfu2) I still not find a solution. don't have time to write a software ecc in linux.09:19
kyakflashing the bootloader can be done manually in a usual manner once in a while09:23
kyakas i see it, uboot and kernel are pretty stable, right?09:23
xiangfufor now it's pretty stable09:26
tuxbrain_awaywpwrak: have you contact avr-dude developers to include the patch? Did you mind If I fill a ticket in his tracking system to the patch be included in main avrdude, maybe this way we can create a ipk without forking the actual avrdude09:29
jow_laptopxiangfu: you can throw w3m into bin/..../packages and run "make package/index"09:29
jow_laptopwhen you approach rolling releases you should be sure to flag essential packages as hold09:30
jow_laptopso that opkg does not upgrade them, e.g. when there's a new kernel with new kmods09:30
xiangfujow_laptop: thanks for the info.09:30
kyakjow_laptop: nice hint! what's the flag name, how to do this?09:32
jow_laptopsee for example package/kernel/Makefile09:32
kyakok, i see09:33
jow_laptopanother interesting flag is "essential", opkg will refuse to uninstall such packages unless forced to09:33
jow_laptopI added that to libc and others09:33
kyakare there more flags?09:33
jow_laptopyes there are but none that make sense within the build environment09:34
jow_laptopthe kernel partition size on the current NN images is fixed, right?09:36
kyakyeah.. hasn't changed09:37
jow_laptopand its an mtd partition without filesystem?09:37
kyakno filesystem there as far as i udnerstand09:38
jow_laptopin this case you could even implement kernel upgrades, by doing an mtd write from within a preinstall script09:38
kyakxiangfu: -^^09:39
jow_laptopand as far as I understood the NN can boot without any kmods?09:39
jow_laptopthis would make it more robust in case opkg failed to upgrade them09:39
kyakhm yes, i think only sound is build as kmods09:39
xiangfuin 2011-02-16, there is one small command. 'mtd.nn'  , for flash kernel. "mtd.nn flash kernel openwrt-xburst-qi_lb60-uImage.bin"09:40
kyakjow_laptop: btw, do you have a workaround for that pango compilation issue (bug 8813) before a real fix is backported?09:40
jow_laptopkyak: you can try replacing PKG_FIXUP:=libtool with PKG_FIXUP:=autoreconf09:49
Action: xiangfu needs go. see you 09:52
wolfsprauljow_laptop: btw - thanks a lot for your suggestions, they are very appreciated!09:52
xiangfuthanks jow_laptop again. ;-)09:55
kyakjow_laptop: i did it last time when you suggested, but it didn't help -\10:05
jow_laptopkyak: I see.10:07
jow_laptopkyak: whats broken with autoreconf? missing macros?10:07
jow_laptopoh I see10:09
jow_laptopits one of those braindead projects that ship libtool in their aclocal.m410:09
kyakjow_laptop: and it's pretty strange that it is working fine for xiangfu!10:09
kyakbraindead :)10:09
jow_laptopI bet xiangfu has a suitable libtool on his host system10:10
kyaki.e. matching versions?10:10
jow_laptopso the issue is hidden10:10
jow_laptoplet me check something10:10
tuxbrain_awayhi package mantainers , I need your help10:11
jow_laptopkyak: ah yes, with PKG_FIXUP:=autoreconf do also specify  PKG_REMOVE_FILES:=aclocal.m410:12
jow_laptopkyak: see if that changes anything10:12
jow_laptopbbl... smoking break10:12
tuxbrain_awayI'm reviewing what is needed to make Nanonote and AVR/Arduino companion, first step UBB is done, now in the software part we need avr-gcc, avr-lib ...stuff to compile and  avrdude to flash10:13
kyakjow_laptop: trying it now. Thanks! (i always wondered where all these nice things are documented, like PKG_REMOVE_FILES or PKG_FLAGS)?10:14
tuxbrain_awayI have was investigating and the avr toolchain seems "not too hard" to achive, there is a script from AVRfreaks with steeps needed to build the toolchain from sources (includind needed patches)10:14
kyaktuxbrain_away: it's a bit tricky. Your host is mipsel, your build is x86, and your target is avr :) This is going to be crazy!@10:16
tuxbrain_awaylink to the Avr freaks script http://www.avrfreaks.net/index.php?name=PNphpBB2&file=viewtopic&t=4263110:17
tuxbrain_awaydo you thing thrasforming this script (download source, aply patch, build) on a openwrt Makfile would be complicated?10:18
tuxbrain_awayI have not right now an uptodate building openwrt building system , an no idea on how to manage the buildhost, but looking at the script seems quite straight forward...10:21
jow_laptopkyak: include/autotools.mk10:21
jow_laptopkyak: however autofail is such an incoherent mess that you often only understand the workarounds after poking around in it a few times10:22
jow_laptopafaik there is no doc on the internet that says you have to remove aclocal.m4 if it bundles a libtool which you don't want10:23
jow_laptopin theory you can to aclocal --force or something but that will fail if someone embedded libtool.m4 in acinclude.m410:23
Jay7can someone port kexecboot to openwrt?10:26
jow_laptopkyak: oh btw, did you try without any PKG_FIXUP at all? with a bit of luck the shipped libtool is sane enough10:27
jow_laptopand does not try to relink /usr/lib10:27
jow_laptopJay7: https://dev.openwrt.org/browser/trunk/package/kexec-tools10:30
Jay7kexec-tools is required but isn't enough :)10:32
Jay7kexecboot itself should be ported10:32
jow_laptopit seems to include kexec something boot10:32
Jay7I mean kexecboot.org10:33
jow_laptopI see10:34
jow_laptopyou can open a feature request ticket at dev.openwrt.org/newticket10:34
jow_laptopso it is not forgotten10:34
jow_laptopprovide a pointer to the package and a short description what it is needed for exactly10:35
jow_laptopbbl... meeting10:35
Jay7ok, thanks for info10:37
tuxbrain_awayjow_laptop: do you think obtaining  the avr packages can be done also opening a tiquet to openwrt?10:45
wpwraktuxbrain_away: (avrdude) no, i didn't try to push to upstream yet. the patch isn't finished yet - i also want to add the clock provision logic for UART and such.12:20
tuxbrain_awayok, I don't know what you are saying but seems important :)12:21
tuxbrain_awayso meanwhile you pach can't be summited upstream maybe we can do a avrdude_8_10 ipk on qi repos?12:26
wpwrakjow_laptop: (autofail) i see we have another member for the autocrap fan club ? ;-)12:36
wpwraktuxbrain_away: (what's missing) the idea is that the UART board will run from the host (ben) clock. so i need to change the CLKSEL (sp?) fuses to the external clock.12:38
wpwraktuxbrain_away: after this, the chip will only work if there's an external clock. afaik, this includes programming.12:39
wpwraktuxbrain_away: so the nanonote driver in avrdude will have to set up the clock output, plus provide a means for selecting the clock speed, etc. i haven't implemented this yet.12:40
wpwraktuxbrain_away: i think it'll be nice to send the whole bundle of avrdude patches in one go, that's why i'm waiting until the thing is done too.12:41
wpwraktuxbrain_away: if you're in a hurry to make packages available, you can either make a temporary package or just compile the binaries and make them, plus avrdude.conf available somewhere12:42
wpwraktuxbrain_away: well, or wait for me to do it (within about a week), because i'll need that too for the recipients of ben-wpan samples :)12:43
tuxbrain_awaynot real hurry, just personal anxiousity  :) a week is far more than ok :)12:44
wpwraktuxbrain_away: (do it) i mean to upload the precompiled binaries. the easy way ;-)12:44
wpwraklemme check if i have all the avrdude/toolchain build information in git already ...12:44
wpwrakthe avrdude stuff is there, with README (haven't tried cross-building for openwrt yet, just for jlime): http://projects.qi-hardware.com/index.php/p/ben-blinkenlights/source/tree/master/uart/avrdude12:49
wpwraknow the toolchain ...12:50
wpwrakthat one's not in git yet ... it's basically the script you found, but simplified (e.g., don't build C++ :)12:53
qi-bot[commit] Werner Almesberger: atben.brd: layout cleanup; removed de facto contacts between ground zones http://qi-hw.com/p/ben-wpan/f7f8eb212:56
qi-bot[commit] Werner Almesberger: atusb/fw3/README: build instructions for firmware and prerequisites http://qi-hw.com/p/ben-wpan/c3ac7d912:56
wpwraknow let's see how long until it shows up on projects ...12:57
wpwrakinteresting delay :)12:57
wpwraktuxbrain_away: (ubb group sales) you should have put a higher margin ;-))13:00
tuxbrain_awaycommon don't screw poor nanonoters :)13:00
wpwraktuxbrain_away: i mentioned to ron that i'd be surprised if you could eat more than one meal on what you'll make from all the UBB stuff :)13:01
wpwrakavr-gcc and friends build instructions: http://projects.qi-hardware.com/index.php/p/ben-wpan/source/tree/master/atusb/fw3/README13:02
wpwrakhmm, ben-openwrt kernel is still at ? :-( that way, the bit-bang library will have difficulties13:10
wpwrakbtw, has anyone else but xiangfu done a full ben-openwrt build recently ? his 30 hours seem oddly long. does it really take so much time ?13:14
rohwpwrak: maybe his machine is oddly slow13:14
wpwrakroh: that's what i wonder about, yes13:15
wolfspraulwpwrak: the machine is a single-core athlon13:15
wolfspraulit was at 8-9 hours before, so indeed there seems to be some slowing down13:16
wpwrakwolfspraul: hmm, sounds old-ish and slow-ish then13:16
rohget him some fresh i7 box13:16
wolfspraulor maybe that's the time to build all 2200+ packages?13:16
wpwrakmaybe he's running low on memory13:16
wpwrakroh: yeah, someone who does major builds on such a regular basis should have decent hardware13:17
wolfspraulI gladly welcome the news of hardware sponsors. wonderful.13:17
wpwrakwolfspraul: hehe ;-)13:17
wolfspraulit may be the time for all 2200 packages13:18
wolfspraulwhy don't we find that our first? ;-)13:18
wolfspraulbefore hitting the 'let's buy bigger gear' button13:19
wpwrakseems that even if it's "just" the 2200 packages, that would still need the bigger gear ?13:20
wolfspraulwpwrak: to bring is from 30 hours to? 20? and those 10 hours improve what?13:24
wolfspraulI am not against faster machines in general. there is always a sweet spot and every year the next faster generation drops 30% or more in price.13:24
wolfspraulbut I am far past the years where I thought building a 'icore7' or whatever it is machine right now will solve my problems for good13:25
wolfspraulI just pick a cheap one, and upgrade regularly.13:25
wolfspraulmeanwhile, if someone wants to donate something high-end (today), go ahead13:25
wpwrakwolfspraul: naw, i would expect a much more substantial speedup. e.g., when i built openwrt ~half a year ago, i don't think it took more than about 2 hours.13:25
wolfspraulagain - we are building 2200 packages probably13:26
wolfspraulyou can build a minimal rootfs in < 1hr13:26
wolfspraulwhat do you mean with 'build openwrt'?13:26
wpwrakwolfspraul: (cutting edge hw) yeah, avoid that. but there's nice yesteryear's hw that's usually not very expensive13:26
wolfspraulexactly. and we have that.13:26
wolfspraulI will upgrade when the next generation is replaced by the uber-next generation.13:27
wpwrakhehe :)13:27
wolfspraulwpwrak: you built the full_system 6 months ago in 2 hours?13:29
wolfspraulthat's possible, it has grown a lot since then.13:30
wolfspraulon the buildhost, it was 8-9 hours last time I checked, but of course keeps growing as we add more and more packages.13:30
wolfsprauland now xiangfu is building all packages I believe, that may explain the jump to 30 hours.13:30
wolfspraulyour machine is probably faster than the buildhost, so maybe back then the buildhost was at 4-5 hours, and your machine at 213:30
wolfspraulin a few years it will be back down to 10 minutes, unless by then it has grown back up to 20 hours :-)13:31
wolfspraulit's funny how people adapt to their environments. give them more power, and they will find a (very reasonably explained) way to use it.13:32
wolfspraulif machines could do 20 Ghz today, it would still take a few hours to build a decent openwrt rootfs :-)13:32
wolfspraulsimply because all other openwrt developers would have those 20ghz machines as well and everything would slowly 'bloat' to make full use of all available power...13:33
wolfspraulas you can see I'm pretty relaxed about all this. it'll come down again, one way or the other.13:33
wolfspraulI fire up my builds on buildhost at night, and in the morning it's there.13:34
LunohoDisn't it possible to store prebuild packages? and then genereate a rootfs by installing those packages?13:34
wolfspraulyes, we are working on that too13:34
wolfspraulso I think it's great that xiangfu is now always building all 2200+ packages13:35
wolfspraulmaybe we can create a huge ImageBuilder from which one could then create rootfses without compilation much faster13:35
wolfspraulneed to try13:35
wolfspraulimho the important thing is that multiple options and entry paths are open, to serve different peoples needs or preferences/experience13:36
wolfspraulsome like to start with a bare minimal rootfs, some like to compile from source, some like to install packages, etc.13:37
wolfspraulsome like to start with a full distro like Jlime or Debian in the first place13:37
zrafawolfspraul: if the whole build of the full_system is just needed from time to time then it is not a very big problem the 30h13:38
zrafawolfspraul: but if somebody wants to try something new and he needs to wait that time to try then maybe he would like a faster machine :P13:38
wolfsprauloh sure.13:41
wolfspraulbut who needs to build 2200 packages from source to try something new?13:41
wolfspraulI wouldn't know what that 'new thing' could be13:41
wolfspraulat the very least one could only build stuff that can even fit onto a Ben NanoNote at the same time :-)13:42
wolfspraullet's gather some more hard data first, for example whether the 30h is for all packages, or just the ones that fit into the 512mb rootfs13:42
wolfspraulI think people are reading too much into the 30h - interesting though. the power of large numbers :-)13:43
wolfspraulyou just say 'something is 100h', and then a big discussion starts13:43
wolfspraulthat much?13:43
wolfspraulwow, we should...13:43
wolfspraulcalling it a day, n8 everybody13:44
wpwrak(30h just sounds _wrong_ ;-)13:45
wpwrakphew. still undefeated, even though wolfspraul's been catching up :)13:47
tuxbrain_awayhey I'm top ten!13:48
wpwrakrjeffries has been quite silent lately. i take this as a good sign. maybe he's busy playing with his "new" ben :)13:48
wpwrakhopefully kicking himself that he didn't get one sooner ;-)13:49
zrafawpwrak: since when is that top10?14:14
rjeffriesOMG I need yo write More, listen LEXX14:15
wpwrakzrafa: (top10) you mean since when do we have this feature in qi-bot ? kyak added it a few days ago14:16
zrafawpwrak: no.. when top10 starst to collect data14:18
zrafafor example.. I know that I am not the top10 if it is collecting data since two months ago14:18
kyakzrafa: should be around 2011-02-0914:20
kyakjow_laptop: hm.. i tried removing PKG_FIXUP at all, tried setting PKG_FIXUP:=autoreconf and PKG_REMOVE_FILES:=aclocal.m4, adding include $(INCLUDE_DIR)/autotools.mk.. none of these helped14:21
tuxbrain_awayzrafa:I will also annoy you and jlime/OE crowd, I want avr toolchain in Jlime !14:21
tuxbrain_awayis not yet in OE so a recipt is needed14:22
zrafatuxbrain_away: you mean an avr toolchain which runs on mipsel?14:26
rohzrafa: why not?14:32
rohtuxbrain_away: i annoyed the openwrt guys before.. i think most stuff is already there14:32
zrafaroh: no sure, is not is more comfortable to cross compile on PC'14:33
zrafais not it*14:33
rohzrafa: linux is linux. why not run the avr-gcc on a mipsel?14:33
zrafaroh: yeah.. and why no run openoffice on nn?.. it is linux14:34
rohi mean.. your code is usually not very big. even the arduino stuff works (well.. not the java-ui but all the rest)14:34
zrafaroh: I am asking if it is not more comfortable to encourge the people to cross compile things on PC? I am okey with toolchains for every archs14:34
rohopenoffice needs hundreds of mbytes of ram and disk. avr gcc doesnt.14:34
rohcomfort is very in the eye of the observer. if i got only a nn around id hate to need a pc only to flash an avr or compile code for it.14:35
zrafaI am sure that every nn owner has a pc14:35
zrafaI have not seen buyers with shipping address from chile desert yet14:36
zrafann buyes14:36
wpwrakzrafa: actually, chances are that if someone there buys one, they'd be very well-equipped. after all, that's where a huge space observatory complex is located :)14:37
zrafaand again. I am okey with toolchains for every arch. I am asking if working with toolchains is not more comfortable on PC than nn, and encourge people to use PC for that14:37
rohif i got a pc around, i dont need a nn for anything.14:39
rohsorry.. but thats how it is.14:39
wpwrakroh: i use the ben to talk to my circuits - that's something my pc can't do easily :)14:42
rohwpwrak: i use a stk500 for that14:43
zrafaroh: (that is how it is) that is your pov, no for everyone. no sense14:43
wpwrakroh: beeeh, non-open hardware ! heretic !14:44
rohhow is it not open? you get a schematic with it14:45
zrafaon the tuxbrain_away case it would be useful I guess. He works with nn and arduino or things like that.. so having all the stuff in nn is something cool surely14:45
wpwrakah, with schem. darn. well, how about ... burn the evil capitalist exploiters and hang their minions ! there must be *something*  :)14:48
MalkiotHello there, I'm having some problems getting my Radeon HD 6950 recognised. Win7 only recognises the "standard VGA" any suggestions?15:03
LunohoDyes, install linux15:03
Malkiotappreciated, but no thx ^^15:03
LunohoDis radeon hd 6950 a copyleft hardware?15:04
Malkiotexcuse my ignorance but what is "copyleft hardware"? It's a prahpics card manufactured by AMD15:05
LunohoDMalkiot: why don't you call your local microsoft support hotline?15:06
LunohoDor amd, whatever15:06
Malkiot...never got good support from big companies before... they usually manage to nuke everything selse instead15:07
mirkokyak: pong15:12
M_RojasMalkiot: ¿?15:14
Malkiotyes rojas?15:15
M_RojasMalkiot: I think that you are in the wrong place asking that ...15:15
Malkiotah ok... know any place where I can ask?15:15
Malkiotsoz about that...15:15
M_Rojasno, I don't ... maybe ... search in google ... ;-)15:16
MalkiotI did and found this, bye ^^15:17
M_Rojasbye, good luck!15:17
M_Rojasthe room it clean ... :-D15:17
Action: panda|x201 start updating to 02-1815:25
zrafaM_Rojas: we should check why google bring M$ people with hardware problem . I have seen several different guys getting in to ask about some graphic card which does not work16:09
zrafaand many of them use win16:09
zrafaNo idea how they find this channel like a channel for support of graphics cards + win16:09
rjeffrieswpwrak you have mail16:23
wpwrakrjeffries: sounds good. now let's see what you get :)16:26
rjeffrieswpwrak I've asked Doug what he would charge to put together bid package16:29
rjeffriesbut if he is too busy maybe I do it16:29
rjeffriesthinking out loud, what if we buy panels of UBB is there a reasonable way to cut the UBB parts16:30
rjeffriesI assume "no" bit worth noodling. i actually think that is the hardest part, no?16:31
wpwrakrjeffries: the fab should deliver them pre-cut or nearly pre-cut (i.e, easy to remove from the board)16:57
rjeffrieswpwrak I figured as much17:08
rjeffriestuxbrain tuxbrain_away what color will your UBB pcbs be?17:09
rjeffriesif I go forward (and who knows?) the two versions should look different, green vs blue or red or whatever17:10
kyakmirko: gnop17:11
mirkokyak: you highlighted me before :)17:26
mirkokyak: was afaik qt related...17:26
kyakmirko: yep :)17:32
kyakmirko: please have a look here http://projects.qi-hardware.com/index.php/p/openwrt-packages/source/commit/f3588d8deea2f5885dc9523ce1b44e8696e9aad1/17:33
kyaki updated NanoMap to the latest git version, but it took hell of a lot overrides17:33
kyakmaybe it's time we fix this problem :)17:33
kyakusually there is only one *.pro file, but in this case there are 5 of them17:34
kyakand all need these overrides17:34
mirkokyak: this wasn't needed before - why now?17:52
kyakmirko: dunno. As far as i remember, it was always needed, since i ported a first qt4 app17:53
mirkokyak: well, the versions of NanoMap i maintained didn't need them at all17:54
kyakyeah, i know17:54
kyakthen you see that message in NanoMap git:17:54
kyakfirst try to fix building with OpenWRT toolchain that does not include QtCor17:54
kyake and QtGui in the include path17:54
kyakso there is a problem17:54
mirkokyak: maybe it just uses wrong include-paths?17:55
mirkoit should include this way: "#include <QtGui/foo.h>"17:55
kyakno, it uses the right paths17:55
kyakyou should build without these overrides and have a look at Makefiles17:55
kyakthey are not correct at all17:55
kyake.g. the compiler is missing17:56
kyakthe paths are wrong, too17:56
mirkoi'll take a look17:56
kyakok, thanks!17:56
xMffbtw, since the IB was discussed recently, here's a proof of concept how a web image builder could work like: http://imagebuilder.augsburg.freifunk.net/cgi-bin/newkit/18:31
kyakhehe, pretty awesome18:36
xMffits written in bash :) code is here https://www.wgaugsburg.de/svn/kamikaze/contrib/imagebuilder/trunk/18:38
kyakxMff: btw.. i wasn't able to build pango, no matter how i did it18:38
bartbesso ehm, I guess all packages have been updated in the new testing release?19:31
kyakthat's right19:36
bartbesI guess I should go flash then19:37
kyaki just did it19:37
bartbescool, new nlove in image then19:37
bartbesor at least in repo19:37
bartbesjust in time, by coincidence :P19:38
kyakyup, you got it in time :)19:38
bartbesthen there's no reason left not to update :P19:38
bartbes"present files are identical to the ones on the server - do not download them again" <-- uh oh20:03
bartbesthat's what reflash_ben.sh said20:06
bartbesI'd hate it if I am now flashing the old image again..20:06
kyakfirst of all, make sure you've got the latest version of reflash_ben.sh. Then, set VERSION="testing/2011-02-16"20:07
kyakso it would download http://downloads.qi-hardware.com/software/images/NanoNote/Ben/testing/2011-02-1620:07
kyakfrom here20:07
bartbesstill that message20:08
bartbesit's -d, not VERSION20:09
kyakhm, i'm running reflash_ben.sh with no arguments20:10
kyaksure, -d testing/2011-02-16 should still do it20:14
kyakwpwrak: are you here?21:18
kyakwpwrak: trying to build qpkg for Jlime here21:25
kyakcloned your git, running make jlime21:25
kyak(got the jlime-mipsel-toolchain)21:26
kyakit says that stdlib.h is not found. Ok, i add -I/path/to/jlime-mipsel-toolchain/usr/include to CFLAGS21:27
wpwrakhmm, gcc should do this automatically21:27
kyakthen there is a strange error usr/include/bits/types.h:129:3: error: #error21:27
kyakwpwrak: yup, both "make" and "make openwrt" run fine21:27
kyaki don't know, maybe the toolchain is broken21:28
kyakor it has some paths hardcoded21:28
wpwrakfor jlime, you could  . /usr/local/jlime-2010.1/mipsel/environment-setup21:28
wpwrakthat sets up a lot of things21:28
kyakhmm ok21:29
wpwrakif you didn't install is in /usr/local/jlime-2010.1, that would explain problems21:29
kyakfirst of all, there is no environment-setup21:30
kyakgot the toolchain from here: http://jlime.com/downloads/development/crosscompilers/jlime-mipsel-toolchain.tar.gz21:30
kyakis it correct?21:30
wpwrakbtw, qpkg isn't a replacement for opkg yet. for now, it can just tell you what packages should be installed. so you still need opkg to do the actual work. (and to put them in the right order)21:30
kyakoh! that's not good21:30
kyakbecause i ran into troubles running opkg install links-x1121:30
kyakout of memory (though i have swap)21:31
kristianpaulkyak: http://en.qi-hardware.com/wiki/JlimeToolchain ?21:31
kyakso i was hoping that qpkg would help it21:31
wpwraki'm using the one made for qi-hw, http://downloads.qi-hardware.com/jlime/toolchains/jlime-2010.1-2010.1-mipsel-linux-toolchain.tar.bz221:31
kyakkristianpaul: hm, ok.. a little bit confusing. Two sites, diferent toolchains :)21:31
kristianpaulkyak: r u aware of jlime-pkg ?21:32
kristianpaulOh !21:32
wpwrakjlime-pkg would be the most immediate solution21:32
kyakrunning it now21:32
wpwrakit's slow but memory-efficient21:32
kyakmaybe i don't need the toolchain after all21:32
kristianpaulno no ! qpkg !!21:33
wpwrakthe idea for qpkg is basically to do what jlime-pkg does, just faster (and with more complete coverage of dependencies - jlime-pkg doesn't handle all the possible cases)21:33
wpwrakbut it's not quite finished yet. priorities ... :)21:34
kyakok, i'm already running links-x11 :) good!21:34
kristianpaulin jlime?21:34
kyaki was going to explore jlime repo a bit21:35
kyakwhat does it offer, and what's actually usable/usefull21:35
kristianpauli like tangogps :-)21:35
wpwrakkristianpaul: i guess we won't hear from him for a while then :)21:36
kristianpaulwpwrak: :-)21:36
wpwrakwhat was it ... ~15k packages ? or ~20k ? even at one package per minute, and only 25% applications (and not libs or such) that would be 60-80 hours. well, that's about one week ;-)21:38
kristianpaulkyak: http://fz.hobby-site.org/hp660lx/nn/listado_de_archivos.txt.bz221:44
kristianpaulzrafa: what happened with that file???21:44
kristianpaulnp i have it21:45
kyakwpwrak: i have you a feeling that you strongly overestimating the number of packages. There are 20 (twenty) centerim-* packages - each package for different locale, devel, debug, doc. DOes that count?21:47
kristianpaulkyak: http://kristianpaul.org/~paul/tmp/list_of_file.txt.tar.gz21:47
wpwrakkyak: alright, maybe you'll be done in half a week :)21:48
kristianpaulkyak: thats a list of all files from jlime repo, i think may help you21:48
kristianpaulkyak: files per package21:48
kyakhm.. ok, thnkas. Though i'm not sure how it would help :)21:49
kristianpaulyou can grep for keywords at least21:49
kristianpaulor wel..21:49
kyakthe list of packages is enough, luckily opkg list does the job21:49
kyakwithout running out of memory21:49
kyakwpwrak: don't tempt me into counting the *real* number of packages in OE. I.e. excluding devel, debug, docs, locales, broken/not applicable for Ben etc. and comparing it with openwrt :)21:51
Action: wpwrak wonders how long a malicious idea, one having taken root, takes to worm its way through the barriers of social conditioning21:54
zrafakristianpaul: you have it? great ;)21:57
kyakhm... irssi-0.8.13?21:58
kyakand no irssi with ssl support?...21:59
zrafakyak: you should download the jlime toolchain from qi21:59
zrafakyak: that is the proper one21:59
kyakyup, got it21:59
Action: kyak installs gtkperf22:00
zrafakyak: btw, bc and top runs on jlime22:01
kyakhm, where is gtkperf?22:01
zrafawhat is that?22:02
kyakgtk performance tool22:02
zrafais that something gtk2?22:02
kyakyeah, i think so22:02
kyakdon't see gtk2 at all22:03
zrafayou do not need then gtkperf: we can tell you the output ;) slow slow.. wait... slow.. a gtk2 button appears.. wait.. slow..22:03
zrafakyak: which jlime version you use? the one on qi server?22:04
kyakit's pretty strange, cause it's working not so bad in openwrt (like in stardict)22:04
kyakno, the one from jlime.org22:05
kyakit seems that gtk2 is not ported to OE. Could it be?22:05
zrafawell, abiword works on jlime. But I do not see it comfortable. ANd iirc stardict takes a while before it runs22:05
kyakmuffinman, yes22:05
kyaki'm wondering, why are packages so old?22:06
kyaki'm looking at OE git22:06
zrafakyak: you should try the one at qi I would say. THe current jlime at jlime.com (no jlime.org) has a repo broken. Some packages are okey. others no22:06
kyakoh, ok22:06
zrafakyak: the repo has one year old. Which packages are so old? which features you miss?22:07
zrafakyak: and gkt2 is at OE since it was born I guess22:07
kyakzrafa: irssi is 0.8.13 and no ssl support22:08
kyakcenterim is 4.22.522:08
kyaki'm mostly looking at packages i'm intersted in22:08
zrafakyak: you can fill a but at bugtracker of jlime.com and the packages you want eventually will appear like magic on repo22:09
kyakok, i was just expecting that the repo is more or less frequently updated22:11
zrafayeah, once a year surely22:11
zrafaso far, in one year, just few guys asked for newer packages. In fact < 10. And at least two of them were interested to see packages form jlime to port to openwrt22:12
zrafaso the repo updates frequency is okey I would say22:12
zrafakyak: maybe if you are one of them you could try some angstron distribution, no for nn, for any other platform with a lot more users and devs. There you will find surely OE repos very often updated with a lot more of applications and a lot more saner.22:14
kyakzrafa: i don't really understand. http://cgit.openembedded.org/cgit.cgi/openembedded/tree/recipes whos repo is that?22:16
kyakhow does it work between jlime/OE and angstron/OE?22:17
kyakzrafa: is jlime-pkg avaialble in bootstrap image?22:23
--- Sat Feb 19 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!