**** BEGIN LOGGING AT Thu Dec 07 02:59:57 2006 Dec 07 13:33:20 now thats very useful Dec 07 13:33:26 use a fpga to emulate a cpu and run stuff on it Dec 07 13:33:51 very economic Dec 07 13:34:10 the LION softcore would be neat for that Dec 07 13:34:24 well there are cores for that Dec 07 13:34:26 put it takes a pretty huge fpga to get ~100 MIPS out of it Dec 07 13:34:28 i just dont see the point in a phone Dec 07 13:35:05 throw in a faster main cpu if anything Dec 07 13:35:09 SC244X Dec 07 13:36:06 s3c24a0 even :) Dec 07 13:36:45 maybe TI dual core omaps Dec 07 13:37:06 and possible even a mobile graphics accelerator Dec 07 13:37:11 I'd rather have an i.xm31 Dec 07 13:37:24 preferably the new version Dec 07 13:38:30 yeah Dec 07 13:38:34 vga hardware mpeg4 encoder Dec 07 13:39:02 pcmcia and ata interface Dec 07 13:39:05 2 sd Dec 07 13:39:41 the people on the ml want http://axon.nl/?page=product_info&lbar=products&cat_id=3 Dec 07 13:39:50 arm cpu + fpga Dec 07 13:40:02 and 18 of those in a 19" case Dec 07 13:40:26 ehehe maybe they can make one into a backpack Dec 07 13:41:21 that freescale chip looks extremely tasty ;) Dec 07 13:42:14 buz: yes, a GPU from ARM/Falanx :) but that wouldnt be a main stream "phone" anymore, but more a PMP/gaming device. Dec 07 13:42:46 or use the dsm-02 from 3dlabs Dec 07 13:42:48 the high end pocketpc devices all come with GPU Dec 07 13:42:53 dual core with 24 FPUs Dec 07 13:42:55 hey mickeyl Dec 07 13:43:00 wb mickeyl Dec 07 13:43:06 you guys planning Neov8 ? Dec 07 13:43:10 :D Dec 07 13:44:48 a pcmcia slot addon would be ultra neat Dec 07 13:45:20 buz: cf is not enough? Dec 07 13:45:26 cf would be plenty Dec 07 13:45:42 afaik there are wifi + flash combo cf cards Dec 07 13:45:43 * koen stabs minimo for install -s in the makefile Dec 07 13:46:12 hell 2 sd slots would be plenty Dec 07 13:46:20 but i'd prefer cf Dec 07 13:46:27 buz: they are but not work as two functions Dec 07 13:46:29 thats about as open as it comes Dec 07 13:46:41 they dont work as both at the same time? Dec 07 13:46:44 dooh Dec 07 13:47:28 Eye-Fi is tackling the problem from a different perspective. They?re building wifi directly into the flash storage. Their first product will be a 1GB SD card with built in Wifi. For about the same price as a 1 GB flash card sells for today - $100. Dec 07 13:47:47 heh Dec 07 13:47:49 $100 Dec 07 13:47:58 1gb sd has $17 last I checked Dec 07 13:48:07 s/has/was/ Dec 07 13:48:07 koen meant: 1gb sd was $17 last I checked Dec 07 13:48:09 well thats probably an old blg posting Dec 07 13:48:19 still i'd pay 100$ for that Dec 07 13:48:30 post is from april Dec 07 13:48:59 how in hell that would ever work in a digicam i dont understand though Dec 07 13:49:59 Eye-Film is still in development. We will announce details of our beta trial in Q3 200 Dec 07 13:50:02 it's vaporware Dec 07 13:51:55 oh beta is actually happening right now Dec 07 13:52:37 v2 needs good old SD :) Dec 07 14:32:35 the Neo name isn't the final market name, is it? Dec 07 14:32:44 just a code name, right? Dec 07 14:39:59 It's the name, and the only names allowed in the contact book is Mr Anderson and Mr Smith. ;) Dec 07 14:40:25 CM: but Neo1973 is not banane... Dec 07 14:47:05 Neo1973 is the final market name Dec 07 14:47:08 the code name was GTA01 Dec 07 14:52:10 I have nothing against the name, but 1979 would have fit me better Dec 07 14:52:50 or some less OMG!!! matrix!!! name would have been better Dec 07 15:00:40 <[g2]> GTA01 LOL Dec 07 15:00:52 <[g2]> GTM01 :0 Dec 07 15:01:40 <[g2]> mickeyl can you speak about OE meta-data deployment for the Neo at all ? Dec 07 15:01:48 sure Dec 07 15:01:51 just ask Dec 07 15:02:07 <[g2]> what's the plan regarding the meta-data ? Dec 07 15:02:17 <[g2]> are you going to ship a cut down version ? Dec 07 15:02:28 <[g2]> are you going to partition sections ? Dec 07 15:02:38 <[g2]> tools/kernel/apps Dec 07 15:03:03 <[g2]> I'm guessing you've been very busy and haven't had a lot of time to think about it Dec 07 15:03:28 * [g2] has a cut-down version of 40 .bbs that builds the arm xscale toolchain Dec 07 15:03:28 that's right. general plan: upload everything into mainstream Dec 07 15:03:40 i like the greenphone vmware sdk image Dec 07 15:03:45 conceptually, at least Dec 07 15:04:10 so we might want to upload something that guys can download to reproduce the build of the software version that ships in the images Dec 07 15:04:29 sections will be partitioned Dec 07 15:04:37 openmoko/foo Dec 07 15:04:38 SDK and OE metadata are two different things Dec 07 15:05:31 I'm not sure whether cutting down is necessary Dec 07 15:05:58 just make sure bitbake 1.8.x is out before january Dec 07 15:06:04 heh Dec 07 15:06:09 as if i have influence about that Dec 07 15:06:15 (other than the ncurses IPC stuff) Dec 07 15:06:30 sure, just send zecke some dutch snacks :) Dec 07 15:06:44 drat Dec 07 15:07:22 mickeyl: I'll be passing frankfurm AM real close tomorrow, but I won't have time to stop by :( Dec 07 15:07:22 bummer Dec 07 15:09:39 <[g2]> well given that about 40 bb's builds a full cross toolchain and is less the 1% of the recipes it's a lot simpler to manage Dec 07 15:10:25 [g2]: using a cutdown version just asks for unnecessary divergence between OE and OM Dec 07 15:11:01 <[g2]> koen clearly NOT. How much of OE meta-data is "upstream" Dec 07 15:11:17 <[g2]> OE does a great job of managing the differences Dec 07 15:11:40 <[g2]> and OE will most likely always be "upstream" for OM Dec 07 15:11:50 if you give people a small tarball to play with instead of the complete repo, you have already lost Dec 07 15:12:52 <[g2]> what is lost ? Dec 07 15:13:22 you give "OE is too big and difficult" signal to potential developers Dec 07 15:13:46 How big is the whole repo? Dec 07 15:13:52 <[g2]> 128MB+ Dec 07 15:14:11 <[g2]> that's just the meta data Dec 07 15:14:11 Ok, but that's small :) Dec 07 15:14:14 and how many recipes will get written just because there is no recipe in stripped version and no one looked into full OE before Dec 07 15:14:18 150MB of text files Dec 07 15:14:45 we wrote more code than Shakespear wrote sonnets :-P Dec 07 15:15:00 hrw|work: like people did for the OZ branch Dec 07 15:15:16 "I made a recipe for lua" "already in .dev" "oh, didn't look there" Dec 07 15:15:42 we have two use cases here Dec 07 15:15:51 We're not talking gigabytes and for use as a dev environment I'd rather have too much code than too little when I need it. Dec 07 15:15:58 <[g2]> koen there are idiots and ppl that live by NIH everywhere. One can't avoid that Dec 07 15:16:19 <[g2]> the world is full of very intelligent ppl also that "do the right thing" Dec 07 15:16:22 one is what i'd call the poweruser for whom the FIC1973 is just another OE-supported device that happens to have GSM (finally) Dec 07 15:16:43 the other one is the linux-is-cool-i-can-customize-my-phone type of user Dec 07 15:16:58 who doesn't need much more than a playground for rebuilding OpenMoko default distribution Dec 07 15:17:02 + adding what he loves to Dec 07 15:17:03 <[g2]> mickeyl _exactly_... The first case is an "oe deveoper" Dec 07 15:17:13 <[g2]> the second is an "oe user" Dec 07 15:17:25 <[g2]> s/deveoper/developer/ Dec 07 15:17:39 i agree. I don't like a stripped-down version, but we need to come up with a plan to make both types of users comfortable Dec 07 15:17:58 mickeyl: qemu image + MACHINE=native Dec 07 15:18:00 s/users/use cases/ Dec 07 15:18:01 mickeyl meant: i agree. I don't like a stripped-down version, but we need to come up with a plan to make both types of use cases comfortable Dec 07 15:18:03 <[g2]> and users will probably out number at least 100 to 1 and probably 1000 to 1 Dec 07 15:18:44 mickeyl: and as you said, a simple vmware appliance with a sdk would be perfect for group 2 Dec 07 15:18:52 CM: *nod* Dec 07 15:19:01 <[g2]> mickeyl it doesn't necessarily have to be a stripped down repo. It could simply be a snapshoted repo and allow something like rsync'ing Dec 07 15:19:09 an sdk lacks one HUGE thing: easy packaging Dec 07 15:19:12 just look at maemo Dec 07 15:19:32 koen, this is why when I talk of an SDK-image Dec 07 15:19:34 i really mean Dec 07 15:19:38 SDK + build system Dec 07 15:19:41 = SDK + OE Dec 07 15:20:14 <[g2]> mickeyl OE should built the SDK image Dec 07 15:20:31 mickeyl: have you seen the poky sdk stuff? Dec 07 15:20:38 * [g2] uses a debootstrapped Edgy right now for that purpose Dec 07 15:20:51 mickeyl: it's a qemy-arm image with all -dev packages installed Dec 07 15:20:52 koen: no time :/ Dec 07 15:21:03 koen: something like that'd becool Dec 07 15:21:22 put OE + bitbake in there with machine=native and you get: Dec 07 15:21:27 a) easy sdk Dec 07 15:21:31 b) easy packaging Dec 07 15:22:36 where 'b)' is the most important thing for a community Dec 07 15:23:01 <[g2]> koen Ok. let's do this Dec 07 15:23:05 koen: you don't like vmware? Dec 07 15:24:04 mickeyl: not for crosscompiling Dec 07 15:24:36 How hard is it to make a vmware appliance? Thinking of koens b) Dec 07 15:24:48 and a vmware image forces people onto a new desktop Dec 07 15:25:17 I use gnome for a reason, so why force me to use KDE to compile apps for the greenphone? Dec 07 15:25:20 <[g2]> koen: Ok here's the bounty throwdown Dec 07 15:25:30 desktop is one point, but vmware is a nice support shield. think all those different environments Dec 07 15:26:24 <[g2]> $500 for an OE build that builds a chroots from within OE for x86 that includes all the tools to bitbake my 40 bb recipe toolchain for gcc 4.1.1 xscale and devio Dec 07 15:27:20 <[g2]> I've got the two files needed for the test case Dec 07 15:27:46 <[g2]> interested at all ? Dec 07 15:31:57 interesting Dec 07 15:32:06 but not before january Dec 07 15:33:00 <[g2]> hrw|work I'll either be done myself or have a good alternative long before jan. Dec 07 15:33:11 sure Dec 07 15:33:33 <[g2]> personally, the Edgy debootstrap is better and easier Dec 07 15:33:47 <[g2]> but a self-hosted OE build would be cleaner Dec 07 15:34:22 MACHINE=native bitbake task-native-sdk python slugos-native (and few others) Dec 07 15:34:51 <[g2]> and it'll easily be 6 months (probably never) that OE would catch up to the LiveCD/graphics card drivers supported by Ubuntu Dec 07 15:34:51 ok need to go home Dec 07 15:35:13 You don't need a vmware image Dec 07 15:35:23 you could just operate within a chrooted environment Dec 07 15:35:26 guy from cable has to came to give me net+voip Dec 07 15:35:29 or even a loopbacked filesystem Dec 07 15:35:32 cu later Dec 07 15:35:35 <[g2]> cheers Dec 07 15:35:38 if you really wanted to go that route. Dec 07 15:36:14 using vmware you have to copy data from your machine to the VM and back again Dec 07 15:36:30 use qemu to develop on arm directly ;) Dec 07 15:36:31 at least with a chroot you can just copy it into the chrooted environment. Dec 07 15:36:54 <[g2]> buz qemu is the second layer Dec 07 15:36:54 qemu runs native arm code? Dec 07 15:37:11 <[g2]> noidd yes qemu provides ARM system and userland Dec 07 15:37:21 <[g2]> LE Dec 07 15:37:29 ah, that is shiny Dec 07 15:37:35 that makes my life a whole lot easier. Dec 07 15:37:43 <[g2]> it's a good start if you run LE Dec 07 15:38:04 <[g2]> and openmoko runs LE Dec 07 15:40:51 [g2]: BE as well Dec 07 15:41:24 <[g2]> koen BE support wasn't working very well a couple months ago. Hopefully it's fixed Dec 07 15:41:28 [g2]: so you essentially want MACHINE=qemux86 (exists) with openembedded-essential installed Dec 07 15:44:00 I need to rebuild my dev environment Dec 07 15:44:17 <[g2]> koen I think more of an MACHINE=x86 and a simple rootfs / rootfs.tar Dec 07 15:44:43 that isn't chrootable, sadly Dec 07 15:44:55 I got a "kernel too old" message Dec 07 15:45:39 * [g2] remembers them from a long time ago, but forgets exactly what that comes from Dec 07 15:46:55 could make it chrootable Dec 07 15:47:17 I guess we could do all of the above as long as we automated the builds of the various dev environments Dec 07 15:47:32 <[g2]> noidd right you need a task-bootstrap functionality included Dec 07 15:47:42 with machine-qemux86 you get a qemu-able thing (probably vmware-able) image Dec 07 15:47:57 <[g2]> koen that's the point Dec 07 15:48:34 <[g2]> if ppl are smart enough the a be able to run qemu then they can already qemu a Ubuntu iso Dec 07 15:48:39 is qemu capable of network connectivity? Dec 07 15:49:08 <[g2]> the second point, is that running natively removes all the emulation overhead Dec 07 15:50:13 <[g2]> emulation will certianly come into play at somepoint (especially for ARM) but for the native environment a chroot has many advantages Dec 07 15:50:52 cos if it is I'm happy to put online a cluster of 10-20 emu systems for people to use Dec 07 15:50:57 like the skiffclusters Dec 07 15:51:33 noidd: qemu supports networking, so nfsroot works Dec 07 15:52:21 * [g2] runs a quick test on the gigE network Dec 07 15:52:33 even with arm emulation? Dec 07 15:52:42 man, there is an interesting project hiding in there Dec 07 15:53:56 noidd: o-hand even builds qemu-arm images with distcc present Dec 07 15:54:08 so you can make your own arm cluster with x86 hardware Dec 07 15:54:34 that's what I was considering doing and making available Dec 07 15:54:51 ie, actually provide the development hosts. Dec 07 15:55:05 I have the hardware, bandwidth and Xen hypervisor available to do it. Dec 07 15:56:51 going to grab some lunch Dec 07 15:56:53 back in a while Dec 07 15:57:45 <[g2]> Ok nfs is a little slow at only 19.4MBs Dec 07 16:00:03 with jumboframes? Dec 07 16:01:43 <[g2]> koen nope Dec 07 16:01:51 <[g2]> koen that's without any tuning Dec 07 16:02:11 <[g2]> I was getting near 30MBs via scps Dec 07 16:02:47 <[g2]> it may be related to the volume as the scp returned about the same performance on a 1GB file Dec 07 16:02:51 about the same here Dec 07 16:03:07 from macbook pro -> server Dec 07 16:03:17 samba tops at ~24MB/s Dec 07 16:03:28 anyway Dec 07 16:03:31 * koen -> food Dec 07 16:03:39 <[g2]> cheers mate! Dec 07 16:03:50 <[g2]> enjoy! Dec 07 16:04:08 <[g2]> and thanks for the lively discussion Dec 07 16:53:06 ick bolsh Dec 07 16:54:35 Sorry - slip of the "new channels get focus" Dec 07 17:43:53 * aevin just hopes Neo1984 will come with a telescreen and speakwrite. Dec 07 18:20:28 24MB/s with samba? around here I top out with 3MB/s Dec 07 18:20:33 scp does 2,5times of that Dec 07 18:20:40 in 100mbit Dec 07 18:20:54 i figured something was wrong ;) Dec 07 18:27:46 <[g2]> buz we're on gigE Dec 07 18:28:27 <[g2]> you can get 11MBs with 100Mbs Ethernet if your processor/disk/network is fast enough Dec 07 21:10:23 [g2]: ping? Dec 07 21:10:36 <[g2]> hrw pong Dec 07 21:11:15 [g2]: x86 rootfs with stuff for running bitbake still wanted? Dec 07 21:11:54 <[g2]> hrw are you interested in the bounty ? Dec 07 21:12:15 sounds good Dec 07 21:12:39 define deadline time Dec 07 21:13:30 <[g2]> does '06 seem reasonable ? Dec 07 21:13:50 '06 as 20060101? Dec 07 21:14:12 <[g2]> no by 20061231 somewhere in the world Dec 07 21:14:16 20070101 I meant Dec 07 21:14:23 <[g2]> yeah 20070101 Dec 07 21:14:33 sounds doable Dec 07 21:15:39 * [g2] should write up some information and post it on my web site Dec 07 21:16:00 thats an option too Dec 07 21:16:44 invoice needed? Dec 07 21:16:55 <[g2]> nope. Dec 07 21:17:37 <[g2]> it's a bounty, not a request for services/consulting Dec 07 21:18:17 <[g2]> I want to be able to check out a monoton repo, or have a reasonable small patch set Dec 07 21:18:17 sure - wanted to know Dec 07 21:18:41 <[g2]> and bb foo-x86-rootfs Dec 07 21:18:49 "bitbake openembedded-build-native-image"? Dec 07 21:18:58 "bitbake openembedded-build-image" rather Dec 07 21:19:39 <[g2]> it could be a package or it could be a distro Dec 07 21:20:04 I use angstrom-2007.1 as base distro Dec 07 21:20:08 and it support x86 Dec 07 21:21:15 <[g2]> I think angstrom-2007.1 as a base distro would be fine Dec 07 21:21:23 <[g2]> I'll need to build it for x86 Dec 07 21:21:37 and it will be. i686 Dec 07 21:22:05 <[g2]> I'm pretty sure I can live with that Dec 07 21:22:28 <[g2]> i686 supports AMD and Intel Duo core right ? Dec 07 21:22:32 yes Dec 07 21:22:50 <[g2]> Ok... that's fine with me too Dec 07 21:24:00 cool Dec 07 21:24:44 <[g2]> paypal payment acceptable ? Dec 07 21:25:03 paypal or moneybookers Dec 07 21:25:40 preferred moneybookers - never used paypal account yet (they support Poland since october) Dec 07 21:26:13 <[g2]> I've never used moneybookers Dec 07 21:26:22 ok Dec 07 21:26:24 <[g2]> but I've go a PP acct Dec 07 21:26:29 <[g2]> s/go/got/ Dec 07 21:26:29 [g2] meant: but I've got a PP acct Dec 07 21:28:51 you guys want an independent judge of completion? Dec 07 21:29:30 ok from me Dec 07 21:30:40 (i.e. I would propose that when the task is finished, an independent third party types "bitbake openembedded-build-image", then chroots into it and builds something like SlugOS or Angstrom. Dec 07 21:31:39 rwhitby: and build 'g2-recipe-pack' Dec 07 21:32:12 07 16:52 < [g2]> $500 for an OE build that builds a chroots from within OE for x86 that includes all the tools to bitbake my 40 bb recipe toolchain for gcc 4.1.1 xscale and devio Dec 07 21:32:23 that was first version ;D Dec 07 21:32:35 yup - in fact [g2] should put on the web page the exact commands he expects to type for the acceptance test :-) Dec 07 21:32:46 yes Dec 07 21:33:01 ($500 is big bucks, so you want to have these things in writing at the start :-) Dec 07 21:33:18 exactly Dec 07 21:34:08 <[g2]> Ok I'll try to update my web site tonight and kick-off the near global bounty Dec 07 21:34:42 ok Dec 07 21:35:09 <[g2]> rwhitby has an advantage as it's daytime for him Dec 07 21:35:22 23:00 here Dec 07 21:35:24 <[g2]> he'll probably still be up Dec 07 21:36:30 [g2]: is this going to be an open bounty, or is hrw signed up exclusively now ? (I'm not interested in claiming it, but hrw would probably want to know that before he begins work) Dec 07 21:37:04 or is it a race and the first person to meet the acceptance criteria gets the cash? Dec 07 21:38:14 <[g2]> rwhitby it's an open bounty with a deadline by end-of-year 2006 Dec 07 21:45:08 I go sleep - cu Dec 07 21:45:51 <[g2]> hrw you know my main website right ? Dec 07 21:46:03 remind me Dec 07 21:46:15 <[g2]> http://www.giantshoulderinc.com Dec 07 21:46:28 myopengps.com :) Dec 07 21:46:37 cu Dec 08 02:14:16 Does anyone know if the neo1973 radio is Class A, B, C? **** ENDING LOGGING AT Fri Dec 08 02:59:59 2006