2010-09-01 00:04 [commit] Werner Almesberger: modules/usd-card.fpd: renamed "depth" to "push" and increased from 1.3 mm to http://qi-hw.com/p/kicad-libs/e6fd40b 2010-09-01 00:04 thanks !! 2010-09-01 00:40 rafa: just try openwrt with setfont2 un-fuzzy-4x8-font.pnm, you'll understand :) 2010-09-01 00:41 rafa: also, i didn't quite understand what about the colors 2010-09-01 00:43 xiangfu: hi! 2010-09-01 00:43 kyak: Hi 2010-09-01 00:43 do you detect any problems with the latest openwrt-xburst image? for example, mkfs.ext2 is not working (it hangs) 2010-09-01 00:44 i have the same problem with my self-built image and with default latest one 2010-09-01 00:45 however, there was no such problem in 2010-06-15 image 2010-09-01 00:45 i believe, it could happen after switching to uClibc-0.9.32 2010-09-01 00:46 (i don't know what else software is affected) 2010-09-01 00:46 kyak: yes. maybe. I just checked  mkfs.ext2/3/4 all doesn't work. 2010-09-01 00:46 kyak: even "--help" 2010-09-01 00:46 yes 2010-09-01 00:47 kyak: ok. I will write down it in my TODO list. ahh maybe I send one email to openwrt upstream first. see if other people have the same problem. 2010-09-01 00:47 i tryed now building from scratch (i.e. not only all possible make clean's, but also manually removed staging_dir which is left for some reason). i believe some old libraries could be mixed in 2010-09-01 00:48 but i'll be able to flash only in the evening.. 2010-09-01 00:48 xiangfu: thanks! 2010-09-01 00:48 kyak: yes. I also try to do that in my PC. "make distclean" :) 2010-09-01 00:49 make distclean proved to be nto so clean , too 2010-09-01 00:49 build_dir is not removed 2010-09-01 00:49 staging_dir is kept, too 2010-09-01 00:49 and there's a lot of old stuff in staging_dir/host 2010-09-01 00:50 openwrt_xburst dir was 1.5 Gb after make clean's. Then, after i manually removed old things, it was arounf 950 Mb 2010-09-01 00:51 also, make distclean removes dl/ dir, which is nonsense 2010-09-01 00:51 the most harmless dir 2010-09-01 00:52 kyak: yes. I already mv the dl to ..  , but I didn't notice it not remove the staging_dir and build_dir. thanks for the info. 2010-09-01 00:52 no problem! 2010-09-01 00:52 then I will "make distclean" then remove those two folders . compile again. if it's fix the mkfs.ext2/3/4 problem. 2010-09-01 00:53 great! 2010-09-01 00:53 we need release a new release :) 2010-09-01 00:53 if this works 2010-09-01 00:53 yes. :) it's need 3 ~ 4 hours compile in my pc. 2010-09-01 00:54 i left it running for the night, not sure how much it took :) would be great to have some time output by make, too 2010-09-01 00:55 or just "time make" 2010-09-01 01:02 kyak: Hi "make distclean" remove the "stage_dir" and "build_dir" (I just finish the "make distclean") 2010-09-01 01:03 kyak: after "make distclean" and "mv dl/ .." . it's only 144M. 2010-09-01 01:06 xiangfu: hmm 2010-09-01 01:06 let me re-check 2010-09-01 01:06 kyak: now I am "time make" 2010-09-01 01:08 xiangfu: also, distclean removes your .config :) 2010-09-01 01:08 xiangfu: you are right, build_dir and staging_dir are removed 2010-09-01 01:08 140M now 2010-09-01 01:09 the dl is 785M. so total is 900+M. 2010-09-01 01:09 i'll take the default config from data/ and time make 2010-09-01 01:09 714M    dl 2010-09-01 01:09 :) 2010-09-01 01:14 xiangfu: can we keep a second config somewhere to build a xbboot image with some rescue tools only? 2010-09-01 01:15 for example in data/qi_lb60/conf/config.xbboot ? 2010-09-01 01:16 wolfspraul: yes. good idea. :) 2010-09-01 01:16 wolfspraul: I will do that today. 2010-09-01 01:17 2010-09-01 09:15:29 (5,82 MB/s) - «BUILD_LOG» saved [74826600/74826600] 2010-09-01 01:17 wow, what was that? is qi-hw server in a neighbour house? :) 2010-09-01 01:18 hm, not really from the BUILD_LOG if mirco had the staging_dir removed.. 2010-09-01 01:18 *not really clear 2010-09-01 01:19 I am thinking about turning the latest/ symlink on the server back to 0615, not 0826, because 0826 seems to have a number of severe regressions 2010-09-01 01:19 what do you think? 2010-09-01 01:21 maybe it needs to be renamed to devel 2010-09-01 01:21 hmm.. I think we keep the latest symlink to 0826. if we found more issue in 0826. it's good for next release. 2010-09-01 01:21 and the "latest" is official 0615 2010-09-01 01:22 ok, xiangfu argues for increasing the pain level :-) 2010-09-01 01:22 well then 2010-09-01 01:22 it's also true... 2010-09-01 01:22 it's a path of a warrior :) 2010-09-01 01:30 [commit] Xiangfu Liu: [xburst] add default command line for kernel http://qi-hw.com/p/openwrt-xburst/fb4e180 2010-09-01 01:32 xiangfu: should i ^C the beautiful clean make? :) 2010-09-01 01:34 kyak: no need.  we put the command line in uboot. we can change the command when system boot. use uboot-env tools. 2010-09-01 01:35 @IF 2I FC A@DB?C H5, 2010-09-01 01:35 how do we change it? 2010-09-01 01:36 are you saying it can be changed during runtime? 2010-09-01 01:45 kyak: yes. 2010-09-01 01:45 we have add the uboot-env to 0826 release. 2010-09-01 01:46 how to use it? 2010-09-01 01:46 does it take kernel command line from some env variable during flashing? 2010-09-01 01:48 i'm intersted, because i recently played with qemu-JZ, but without any luck.. basically because our uboot detectes that U key is always pressed for some reason and uboot from qemu-JZ doesn't like our kernel command line 2010-09-01 01:51 there are two command : fw_printenv  fw_setenv 2010-09-01 01:54 where are those commands? 2010-09-01 01:57 in your nanonote: /usr/sbin/ 2010-09-01 01:58 fw_setenv is to set the uboot'env. 2010-09-01 01:58 and in uboot's ENV , there is "commandline" 2010-09-01 01:58 after set we must reboot to make it works. 2010-09-01 01:59 about the U key. it's hard code to u-boot. 2010-09-01 01:59 oh, got it now! 2010-09-01 01:59 you need edit the u-boot source to disable the detect U key. 2010-09-01 02:00 if you set the wrong comandline. we must to connect the NN serial console to fix it. 2010-09-01 02:00 press [S] when poweron. NN will goto u-boot console. then we can fix the u-boot envs. 2010-09-01 02:00 ok, i understand ! 2010-09-01 02:02 xiangfu: are you interested in playing with qemu-JZ? 2010-09-01 02:03 kyak: I just don't have enough time to play with qemu-JZ. but I'd like to if I have time 2010-09-01 02:07 ok :) 2010-09-01 02:51 kyak: maybe the GPIO reading of non-connected pins returns the wrong value in qemu-jz? 2010-09-01 02:53 am I right in assuming that non-pressed keys should read in the same way as a non-connected keyboard? 2010-09-01 03:01 [commit] Werner Almesberger: modules/stdpass.fpd, originally from gta02-core http://qi-hw.com/p/kicad-libs/7ca5701 2010-09-01 03:06 mth: have you considered what you'd have to do to dis_connect_ the keyboard ? :-) 2010-09-01 03:08 I have no idea 2010-09-01 03:09 (assuming we're talking about then ben) you'd have to split the pcb in half :) 2010-09-01 03:10 I'm not suggesting to actually disconnect it, just as a though experiment 2010-09-01 03:10 in a matrix keyboard, a non-pressed key reads as a not-connected pin afaik 2010-09-01 03:11 so if the keyboard is not emulated by qemu, all keys should read "not pressed" 2010-09-01 03:11 but the opposite seems to be the case 2010-09-01 03:11 I had the same problem when emulating the Dingoo (which uses direct GPIO keys, not a matrix keyboard) 2010-09-01 03:12 thought experiment is less scary :) well, the keyboard is integrated into the pcb. so you can't disconnect it. but if a trace was severed, then this would appear as the absence of pressed keys, yes 2010-09-01 03:12 as far as i recall, there are no external pull-ups either 2010-09-01 03:13 each GPIO can be configured to pull up or not 2010-09-01 03:13 [commit] Xiangfu Liu: [xburst] build zImage, but not copy to bin/ http://qi-hw.com/p/openwrt-xburst/9f94731 2010-09-01 03:13 [commit] Xiangfu Liu: [xburst] add config and README file for build xbboot image http://qi-hw.com/p/openwrt-xburst/1501b07 2010-09-01 03:13 it's possible qemu ignores that setting when reading the GPIO pin 2010-09-01 04:06 wpwrak: still there? 2010-09-01 05:57 Hello there! Will usb ethernet card work on nanonote? 2010-09-01 06:02 demonoid, nanonote has no usb host, so no devices connected to it's usb port will work 2010-09-01 06:03 I see. mini-USB 2.0 differs from USB ,right? 2010-09-01 06:06 it's not the issue of mini-USB vs big-USB, it's usb host vs usb slave mode 2010-09-01 06:07 http://gallery.slashhome.se/main.php?g2_itemId=2905 2010-09-01 06:10 Hello Ornotermes, what's that? 2010-09-01 06:10 demonoid: a microSD breakout board, http://slashhome.se/p/uSD_breakout/ 2010-09-01 06:13 Ah...I just don't understand such geeky stuff 2010-09-01 06:15 all signals in microSD slot can used as GPIO (Generel Purpose Input/Output), so it can send an recieve almost any digital information 2010-09-01 06:16 but usb is probably a bit heavy :P 2010-09-01 06:22 @_@... Anyway, there seems to be some sort of usb slave2host converter 2010-09-01 06:45 demonoid: where did you find that? 2010-09-01 06:53 wpwrak: cheers kudos whatever for the last hack/mod :) 2010-09-01 06:53 !! 2010-09-01 06:53 ^_^ 2010-09-01 06:55 mth: it is definitely hardware-related, because qemu-JZ is for some other jz-base board, not specifically for Ben 2010-09-01 06:59 http://forum.xda-developers.com/showthread.php?t=629590 2010-09-01 07:45 Ornotermes: good morning :) just got up 2010-09-01 07:45 wpwrak: morning :) 2010-09-01 07:46 i been up a while, working on my break out board 2010-09-01 07:46 Ornotermes: the little board works quite well. some little issues i found: because it's thicker than uSD, it doesn't go in/out quite as easily. not sure if this is a problem. i may also have accidently deposited some adhesives this way, which would make the problem worse. 2010-09-01 07:47 so i noticed 2010-09-01 07:47 Ornotermes: the mechanical tolerance for board width is relatively narrow. if it's too small, it won't "click" 2010-09-01 07:48 it is possible to snd souwn rhe card a little to make it easier to put in/out 2010-09-01 07:48 sand* 2010-09-01 07:48 down the* 2010-09-01 07:49 Ornotermes: and you need about 1.4 mm between the edge of the "card" and any ben-facing edge of the outer board, or you can't push it in enough to click. but in think you're using a much bigger margin for this anyway, so not a problem there. 2010-09-01 07:50 i started out with big margins, i idea was that people other than ben users might be interested 2010-09-01 07:50 Ornotermes: (sand) yup. didn't want to dremel at 3 am, though :) (my mill sits in a sound-proofed box, that it's not so easy to trim the board along the z axis with that one) 2010-09-01 07:50 ah yes, good point 2010-09-01 07:50 i looked around for the specs of the card holders, but didn't find any that said how much travel this thing really needs 2010-09-01 07:51 kristianpaul: thanks ! :) 2010-09-01 07:52 s/that it's/ but it's/ 2010-09-01 07:52 need some caffeine ... 2010-09-01 07:52 it's getting time to test if the board actually works 2010-09-01 07:52 i had to do some creative fileing 2010-09-01 07:53 hehe :) hardware hacking in all senses of "hardware" :) 2010-09-01 07:53 a couple of pads shorted when the card was inserted 2010-09-01 07:53 for accessing the gpios, you can reuse my "bll" application. http://projects.qi-hardware.com/index.php/p/ben-blinkenlights/source/tree/master/bbl 2010-09-01 07:54 ofc :) 2010-09-01 07:54 (shorts) aii ... 2010-09-01 07:57 well ateast no magic smoke :) 2010-09-01 07:58 but i just reflashed so i have to start by copy som things from my card 2010-09-01 08:02 Ornotermes: i still need to find out if the corners need to be rounded or not. probably a good idea to do it in any case. 2010-09-01 08:07 i too think its a good idea, it makes less risky when inserting the card 2010-09-01 08:12 also, i found it more prectical to bridge ethernet with usb0 on my desktop and let ben get a dhcp adress 2010-09-01 08:13 oh yes. that's the only sane way for communication at the moment. i really wouldn't want to connect ethernet this way. maybe ieee 802.15.4 2010-09-01 08:13 (the latter is one of the reasons why i'm so interested in that idea :) 2010-09-01 08:14 oh, i meant instead the firewall method with static adress between ben and desktop 2010-09-01 08:14 i would peronally prefer bluetooth to connect my ben wireless 2010-09-01 08:15 i mean, even lego uses bluetooth 2010-09-01 08:15 (address) aah, i see. well, depends. i like my gadgets to be easily identified. but i'd have to go to dhcp too if i wanted to talk to more than one at a time. 2010-09-01 08:15 i've prefer bt, too. but then, even bt has openness problems :-( 2010-09-01 08:16 e.g., try to find even the data sheet of an affordable BT module 2010-09-01 08:16 yeah, but if there were internal usb with a hub it bould use usb bluetooth 2010-09-01 08:17 ah, digi-key goodies have reached buenos aires customs :) 2010-09-01 08:17 yes, that it could do. not sure about power efficiency, though. and you'd still want a bit of documentation for all the non-usb things the modules have too. 2010-09-01 08:19 plus, there are some special parameters also via BT that you may want to access. i don't quite remember what it was, but we ran into that problem at openmoko. couldn't determine one of those items. 2010-09-01 08:22 it could be an idea to try to get a dialog with a company that make bluetooth chips and try to make them to open them 2010-09-01 08:22 good luck :-) 2010-09-01 08:22 atleast enough to use them in open source hardware 2010-09-01 08:23 i'm quite sure this has been tried before 2010-09-01 08:24 and remember, sharism is a customer who doesn't generate revenue. 1000 units it not even worth hitting "reply". 2010-09-01 08:24 yeah, but a open enough chip could make a greate impact 2010-09-01 08:25 Ornotermes: the disadvantage of bridging is that interface has to be put down every time to build a bridge 2010-09-01 08:26 no very useful when you connect/disconnect Ben often 2010-09-01 08:26 kyak: really? i have it set up in my network configuration, just plug it in and start ben 2010-09-01 08:27 try pinging something, then connect Ben.. you should see your pings momentarily gone 2010-09-01 08:28 your bridge is built via udev, when you connect Ben, right? 2010-09-01 08:29 no, static pridge 2010-09-01 08:29 bridge* 2010-09-01 08:29 hm..ok, maybe.. anyway, i have no problems using NAT on Ben 2010-09-01 08:31 it's not a "problem" to use NAT but by bridging i have more access to ben from my network 2010-09-01 08:32 i'd use DNAT for that, if i needed 2010-09-01 09:55 morning :) 2010-09-01 10:04 xiangfu: hey thanks you too for pointing this feature about uart emulation over usb 2010-09-01 10:04 :) 2010-09-01 10:06 really awesome ! 2010-09-01 10:07 i wonder what else i can emulate over usb ;D 2010-09-01 11:12 wait 2010-09-01 11:12 did 2010-09-01 11:12 I seriously break gmenu2x? 2010-09-01 11:12 I had no problems.. and I never heard someone had problems either 2010-09-01 11:13 anyway, I would update to the new image 2010-09-01 11:13 if sdl wasn't broken 2010-09-01 11:13 bartbes: yes it's broken. I was already worried it could affect you. 2010-09-01 11:13 well I wouldn't mind messing with gmenu2x, but I compiled my version, ran it on my version of the image 2010-09-01 11:14 no problems whatsoever 2010-09-01 11:14 I'll try again though 2010-09-01 11:14 so, what is the exact problem with SDL anyway? 2010-09-01 11:16 updates toolchain 2010-09-01 11:16 nanomap, huh? 2010-09-01 11:16 sounds cool 2010-09-01 11:18 ah yeah, i'd have some questions about the nanomap ;) 2010-09-01 11:18 first, wht is it exactly? i haven't tried it yet 2010-09-01 11:18 second, what libs does it use? SDL perhaps? 2010-09-01 11:19 i'd be interested in porting it to the dingoo, if it runs on sdl and can display offline maps 2010-09-01 11:19 yeah, I have no idea either 2010-09-01 11:19 well, if it ran on sdl it wouldn't run on the nanonote atm, would it? 2010-09-01 11:19 :P 2010-09-01 11:19 oh.. 2010-09-01 11:19 i haven't been paying attention to owrt lately, was rather busy with jlime 2010-09-01 11:19 I think I know why gmenu2x might crash 2010-09-01 11:19 so we have broken sdl now? 2010-09-01 11:20 wolfspraul: do you have the gmenu2x crash too? 2010-09-01 11:21 bartbes: yes, for example starting dgclock will crash gmenu2x 2010-09-01 11:21 ehm 2010-09-01 11:22 I only touched the explorer.. 2010-09-01 11:22 I thought it was about the explorer anyway 2010-09-01 11:22 bartbes, explorer in the gmenu2x? 2010-09-01 11:22 yes 2010-09-01 11:22 from the list: 2010-09-01 11:22 >>  - the gmenu2x-explorer is unusable right now - gmenu2x is seg-faulting 2010-09-01 11:22 >> immediately - is anybody willing to take care of that? 2010-09-01 11:22 heh, that's not good 2010-09-01 11:22 because it *sounds* like the default dir doesn't exist 2010-09-01 11:22 we need gmenu2x soon for open-dingux release 2010-09-01 11:23 (though arguably it shouldn't segfault) 2010-09-01 11:23 bartbes, wasn't default set to / ? 2010-09-01 11:23 no 2010-09-01 11:23 /card 2010-09-01 11:23 it should read default from a config file 2010-09-01 11:23 instead of being hardcoded 2010-09-01 11:23 it's a compile-time define 2010-09-01 11:23 but yeah it should 2010-09-01 11:23 ah, so it's good 2010-09-01 11:24 well, I made it more flexible 2010-09-01 11:24 somebody already abstracted it 2010-09-01 11:24 but left some of the hardcoded lengths in 2010-09-01 11:24 oh no wait 2010-09-01 11:24 it's *not* a define 2010-09-01 11:24 OH GOD 2010-09-01 11:24 hmm.. wasn't the latest change in 28 days? 2010-09-01 11:25 I should really, really fix that 2010-09-01 11:25 I thought somebody committed.. 2010-09-01 11:25 he told me he was fixing it 2010-09-01 11:25 or there is yet another repo? 2010-09-01 11:25 not that I know of 2010-09-01 11:25 hence the conclusion I broke it 2010-09-01 11:25 i thought you were talking about some new changes 2010-09-01 11:29 I see 2010-09-01 11:29 removing /card actually causes the segfault 2010-09-01 11:29 time to fix it then 2010-09-01 11:31 but how.. 2010-09-01 11:34 I see where I went wrong 2010-09-01 11:36 right, so the segfault should be fixed 2010-09-01 11:36 now the define 2010-09-01 11:38 done 2010-09-01 11:39 [commit] bartbes: Fixed segfault when CARD_ROOT didn't exist & added _CARD_ROOT http://qi-hw.com/p/gmenu2x/6381674 2010-09-01 11:39 there we go 2010-09-01 11:39 -->Dinner 2010-09-01 11:39 if there's anything just hilight me 2010-09-01 11:39 :D 2010-09-01 11:39 good work man 2010-09-01 12:28 well, the fix fails 2010-09-01 12:28 great.. 2010-09-01 12:28 going to have to look into it later 2010-09-01 15:42 hehe, the CPU has a bit in the SD/MMC controller called "OUT_OF_RAGE" :-) 2010-09-01 16:12 [commit] bartbes: More fixed for when CARD_ROOT doesn't exist http://qi-hw.com/p/gmenu2x/cf9628a 2010-09-01 16:12 that bot is fast! 2010-09-01 16:13 awww 2010-09-01 16:13 I typoed! 2010-09-01 16:13 :( 2010-09-01 16:16 bartbes: better to have typed and typoed that to never have typed :) 2010-09-01 16:23 wpwrak: ohhh you wrote the doc for the initrd in Linux !! 2010-09-01 16:23 :) 2010-09-01 16:26 argh 2010-09-01 16:26 hmm, did i ? i rememeber writing initrd, but my memory is hazy on documentation. well, a README or so will have been there .. :) 2010-09-01 16:26 I hate this segfault 2010-09-01 16:29 hates slow customs. goodies from digi-key are still sitting there. been available for clearance since 1:45 am, almost 16 hours ago :-( 2010-09-01 16:32 have any of you ever seen the gmenu2x error "applet not found"? 2010-09-01 16:32 oh wait 2010-09-01 16:32 I know 2010-09-01 16:32 nvm 2010-09-01 16:33 nice: 2010-09-01 16:33 (gdb) backtrace 2010-09-01 16:33 #0  0x2acbe250 in ?? () 2010-09-01 16:33 ... 2010-09-01 16:33 (that's everything) 2010-09-01 16:34 wpwrak: yes a initrd.txt sort of REAMDE if you want call it that way :) 2010-09-01 16:36 ls 2010-09-01 16:36 oops 2010-09-01 16:38 oh.. 2010-09-01 16:38 I know why it segfaults 2010-09-01 16:38 just not why it doesn't prevent the situation from happening 2010-09-01 16:39 bartbes: are you trying last openwrt image? 2010-09-01 16:39 no 2010-09-01 16:39 I can't 2010-09-01 16:39 and won't 2010-09-01 16:39 no sdl 2010-09-01 16:39 ah ok 2010-09-01 16:39 but I *am* fixing gmenu2x 2010-09-01 16:39 yeah :'( 2010-09-01 16:39 or at least trying to 2010-09-01 16:40 :) 2010-09-01 16:40 hmm 2010-09-01 16:40 and now I know why it doesn't prevent the segfault too 2010-09-01 16:40 great 2010-09-01 16:40 if fix means re write, go ahead ! ;) 2010-09-01 16:40 hehe 2010-09-01 16:40 no rewrites ;) 2010-09-01 16:40 hehe i know 2010-09-01 16:41 now.. how did I cross-compile.. 2010-09-01 16:41 (without committing, that is) 2010-09-01 16:45 found it 2010-09-01 16:45 that's why I love the irc logs 2010-09-01 16:52 [commit] bartbes: Automatically switches CARD_ROOT to / when CARD_ROOT doesn't exist http://qi-hw.com/p/gmenu2x/5c05348 2010-09-01 17:02 why?! 2010-09-01 17:02 it *should* run the code I added 2010-09-01 17:03 wait a second.. 2010-09-01 17:03 I don't see the build dir being updated 2010-09-01 17:03 (aka redownloaded) 2010-09-01 17:03 that helps.. 2010-09-01 17:15 well *that* helps 2010-09-01 17:15 damn 2010-09-01 17:15 I hate this 2010-09-01 17:15 :( 2010-09-01 17:15 I might have fixed this ages ago! 2010-09-01 17:15 :@ 2010-09-01 17:15 anyway 2010-09-01 17:15 now let's update the Makefile 2010-09-01 17:17 [commit] bartbes: Updated gmenu2x makefile to use newest version http://qi-hw.com/p/openwrt-packages/6ce3263 2010-09-01 17:17 and now unrape my toolchain 2010-09-01 17:19 is proud of himself 2010-09-01 17:28 [commit] Bas Wijnen: working on mass storage http://qi-hw.com/p/iris/61d76aa 2010-09-01 17:41 roh: hmm, is mail.openmoko.org down ? 2010-09-01 17:50 [commit] Werner Almesberger: modules/stdpass.fpd: added experimental -M variants for manual soldering that http://qi-hw.com/p/kicad-libs/476841d 2010-09-01 18:08 [commit] Werner Almesberger: Get stdpass from kicad-libs. Added bbl.cmp to make pcbnew happy. http://qi-hw.com/p/ben-blinkenlights/87f1ae7 2010-09-01 18:48 wpwrak hm.. the whole host seems to be down 2010-09-01 18:53 wpwrak chandra is down, seems to be some scheduled rz stuff. 2010-09-01 18:54 means: it should be back soon 2010-09-01 18:55 roh: ah, good. thanks ! 2010-09-01 18:56 wpwrak: (bbl) the gpio is controlled by manipulating the memory, right? 2010-09-01 18:59 Ornotermes: the port registers are memory-mapped, yes 2010-09-01 19:00 good, i was thinking of controlling the gpio from python 2010-09-01 19:00 Ornotermes: the concept they use is interesting: you don't write values directly to the registers but you have a set and a clear register. each 1 you write to the set register gets set, each 1 you write to the clear register gets cleared. that way, there can be no conflicts with read-modify-write instructions 2010-09-01 19:00 this stuff is fantastic.  what about shipping a uSD breakout with every Ben? 2010-09-01 19:02 Ornotermes: hmm, i once tried to do these things on the ben with python and i found that the mmap didn't quite work. not sure whether this was just my installation or not, though. so if it doesn't work, you may want to strace it to see if mmap is really called with the correct arguments. 2010-09-01 19:02 Ornotermes: if i recall properly, the one i had trouble with was the address. i think it got shifted or such. 2010-09-01 19:03 unclouded: could be cool with a "hacker pack" that has several other hacking tools too (usb-uart converter and other stuff) 2010-09-01 19:05 wpwrak: it is worth giving a try, especially since the whole python thing is more beginner friendly 2010-09-01 19:07 *afk* 2010-09-01 19:07 wpwrak: nice source: http://projects.qi-hardware.com/index.php/p/ben-blinkenlights/source/tree/master/bbl/bbl.c  easy to read 2010-09-01 19:08 Ornotermes: yup. well, if it doesn't work and you can figure out why, you'll have done mankind yet another great favour :) 2010-09-01 19:09 unclouded: thanks :) needs a bit more comments on the registers, though. i did it in a bit of a hurry :) 2010-09-01 19:10 it's really useful having something short enough to read in one go that gives a demonstration of how to drive the SDIO pins.  that's what I meant to say 2010-09-01 19:12 just need to remove the comment about "command line parsing", which is confusing 2010-09-01 19:12 unclouded: ah yes, that's actually one of the ideas behind making it :) 2010-09-01 19:12 oh :) 2010-09-01 19:13 i'll do a bit of spring cleaning after dinner 2010-09-01 19:21 wpwrak btw: about the microchip enc28j60 stuff... spare yourself the pain 2010-09-01 19:21 that stuff sucks and dies if you cough a bit on the ethernet 2010-09-01 19:22 my current way of putting ip on avr's is using the serial and speaking zbus 2010-09-01 19:32 roh: heh. i have no plan whatsoever to connect that thing to ethernet :) 2010-09-01 19:32 roh: but this little connector should also be enough to put ieee 802.15.4 there :) 2010-09-01 19:40 which connector? 2010-09-01 19:41 the microsd, with a little pcb inside, and bit-banging it 2010-09-01 19:41 my point is only: the enc thingie is crap. not worth the work. falls over and deadlocks a lot in realworld ethernet segments 2010-09-01 19:41 there is nothing to bitbang which makes sense. 2010-09-01 19:41 maybe post that to the list for those who might want to try that ether thing 2010-09-01 19:41 also its not unusual nowadays to have atleast singledigit, if not even 2 digit mbits of multicast load on a ethernetsegment, even at home 2010-09-01 19:42 (bitbang) the at86rf230 doesn't speak sd/sdio/mmc and the ben's mmc engine can't speak spi. so bitbang it is 2010-09-01 19:42 makes this thing not only saturate the phy, but also deadlock. it seems it deadlocks a lot. 2010-09-01 19:42 means: it doesnt even recover from a saturated phy 2010-09-01 19:43 (mcast) ah, interesting. i don't think i have any on my net :) 2010-09-01 19:43 and its not that it wouldnt have mac-addr and multicast filters. 2010-09-01 19:43 wpwrak what do you think they use for iptv? ;) 2010-09-01 19:44 tcom-vdsl is multicast. you send group-joins and then get the traffic routed through your dsl line. 2010-09-01 19:44 no iptv here :) 2010-09-01 19:45 same here. but we got a hw-mpeg encoder on our cnc mill machine with a robot-cam 2010-09-01 19:45 generates 8-10mbit multicast 2010-09-01 19:47 in the end its much easier to use a serial and let something else handle ethernet. 2010-09-01 19:47 if one wants 'ip connectivity' and not neccessarily 'ethernet' 2010-09-01 19:48 wpwrak you would like zbus: http://www.ethersex.de/index.php/ZBus_Protocol 2010-09-01 19:49 dead simple 2010-09-01 19:50 about the at86rf230.. bitbanging doesnt sound that bad... or is the chip that fast on datarate? 2010-09-01 19:51 naw, data rate is 250 kbps. maximum spi speed (from buffer) is 8 Mbps. 2010-09-01 19:53 doesnt sound so bad. 2010-09-01 19:54 does the ingenic cpu have no spi at all? or is the limitation to 'no spi' only for the mmc engine? 2010-09-01 19:56 it has spi, bit it's blocked by the keyboard on the ben 2010-09-01 19:57 also, you can't have the spi engine on the mmc pins 2010-09-01 19:57 (unless you connect them in some way, of course) 2010-09-01 19:57 gnnnh. 2010-09-01 19:57 so for connecting boards to the uSD connector of the ben, it's bit-banging 2010-09-01 19:57 ok.. old problem.. keyboard suckage 2010-09-01 19:58 or hardware that speaks SD/SDIO/MMC :) 2010-09-01 19:59 yeah, the choice of signals for the keyboard is pretty much pessimal, if this word existed in english :) 2010-09-01 19:59 actually .. it does. just has a different meaning :) 2010-09-01 19:59 hrhr 2010-09-01 20:01 i'm back 2010-09-01 20:02 a microcontroler sdio - spi converter would be so cool :) 2010-09-01 20:04 Ornotermes: you could probably do it with a psoc3/5 and the little cpld inside 2010-09-01 20:06 i think that would be a bit over my head for now :P 2010-09-01 20:07 i use AVRs 2010-09-01 20:08 consider it an opportunity to lean something new :) 2010-09-01 20:08 leaRn 2010-09-01 20:08 alas, these chip still aren't in production. they've been available for evaluation only for more than a year now. not sure what's keeping them. 2010-09-01 20:09 so it's still possible that they won't really make it to the market. that would be a pity. 2010-09-01 20:09 there is atleast eval kits to buy 2010-09-01 20:10 btw, one very nice thing about uSD is that there's also a clock output there (the SD/MMC clock). just what i need for ieee 802.15.4 ;-) 2010-09-01 20:12 should be able to multiplex miso and mosi over a single data line (the transceiver seems to support such a thing). so even the small number of signals shouldn't be a big issue. 2010-09-01 20:14 fooood 2010-09-01 20:17 roh: does your food come when you call it ? :) 2010-09-01 20:18 nah.. was fetching it outside 2010-09-01 20:18 grr. mail still down :-( 2010-09-01 20:18 roh: (food) ah, good old hunter and gatherer style :) 2010-09-01 20:20 berlin is great. <300m and choice of food 2010-09-01 20:20 wpwrak: but i think a cheap avr might be good enough to convert SDIO to SPI or I2C 2010-09-01 20:22 Ornotermes: sdio is a bit tricky. you need to synchronize with the master clock and you have to be prepared to receive any time. 2010-09-01 20:23 what is the clock frequency? 2010-09-01 20:23 Ornotermes: but if the SD/MMC clock is slow enough and you have short response time, then why not 2010-09-01 20:23 Ornotermes: you could pick your own 2010-09-01 20:24 http://en.wikipedia.org/wiki/MultiMediaCard#Table 2010-09-01 20:24 Ornotermes: there's a "standard" minimum speed (dunno how much), but in this case, you can of course ignore that 2010-09-01 20:25 if it used max transfer(20Mbit/s) and 4bits it should be 5MHz? 2010-09-01 20:26 go down to less and a pcint on a avr at 20mhz should be enough 2010-09-01 20:26 like.. 2-3mhz 2010-09-01 20:27 can the uSD breakout be used to talk SPI directly? 2010-09-01 20:27 yes, but 20MHz @ 3.3V can be a bit problematic 2010-09-01 20:28 unclouded: with bit-banging, yes 2010-09-01 20:31 Ornotermes: are you thinking of something like a gateway between bit-banged SPI on the uSD breakout to I2C via an ATmega8 or similar? 2010-09-01 20:32 unclouded: no, i'm thinking of a SDIO to SPI (or I2C) interfece 2010-09-01 20:33 so ben can use the accelerated SDIO interface 2010-09-01 20:33 Ornotermes: oh I see: you're going to make something that appears as an SDIO device? 2010-09-01 20:34 i don't know if i'm really going to make it, but i'm thinking a bit about it 2010-09-01 20:35 sounds like a job for a really small cpld 2010-09-01 20:37 that would probably be the best yes, but on the other hand most AVRs is capable to bit bang usb :P 2010-09-01 20:40 roh: mail is still down :-( let's hope the machine didn't burst into flame or something ... 2010-09-01 20:40 me too. gismo mentioned something about 'rz move' ... i guess they carry aall machines over 2010-09-01 20:42 roh: wolfgang posted something about a major downtime (~9 hours) of the qi-hw machines in about two weeks. 2010-09-01 20:42 sep 19 to 20. 2010-09-01 20:43 yeah. gismo said also something about 'was a different date' 2010-09-01 20:46 maybe it's our turn tonight then. or the machine just decided it was a good general timeframe to die :) 2010-09-01 20:49 lets hope not 2010-09-01 20:52 [commit] Werner Almesberger: Cleaned up bbl.c and added comments. http://qi-hw.com/p/ben-blinkenlights/1f509de 2010-09-01 20:56 http://www.sdcard.org/developers/tech/sdcard/pls/ if anyone is interested in the "simplified" SD specs 2010-09-01 20:56 not much meat on the mechanical side there :-( 2010-09-01 21:04 oh well, i have to sleep 2010-09-01 21:04 good night 2010-09-01 21:29 wolfspraul: btw, how open can we discuss features of the various ingenic cpus, especially those of the jz4760 ? 2010-09-01 21:31 wolfspraul: btw2, do you know the dsv (data sheet viewer) system i use in ben-wpan ? (a successor of the one in gta02-core) 2010-09-01 21:42 wpwrak: 4760, I think we can discuss openly, why not. There is not that much to discuss anyway. 2010-09-01 21:44 great. things like the number of SDIO units or what other peripheral blocks are there. nothing fancy, i agree. just wanted to be sure they didn't explicitly ask you to keep all this secret or such. 2010-09-01 21:44 jz4760 = more of everything ;) 2010-09-01 21:44 my passion is not around such chips anyway, they are just temporary stepping stones. 2010-09-01 21:45 larsc: yeah, it's a bit of a wet dream come true :) 2010-09-01 21:45 larsc: e.g., 3 SD/MMC units. wow. 2010-09-01 21:45 well when I started to work with Ingenic my hope was that their engineering would be so bad that they couldn't do major upgrades that would destroy our GPL work. 2010-09-01 21:46 ;-))) 2010-09-01 21:46 so by now, larsc has been doing phantastic work on the Linux kernel side. Now let's see whether Ingenic holds their part of the teamwork. 2010-09-01 21:46 in other words - how easy it will be to support the 4760... 2010-09-01 21:47 I will try to explain to Ingenic that they should re-base their work on kernel.org 2010-09-01 21:47 not sure whether they get it though... 2010-09-01 21:47 about the 3 SDIOs 2010-09-01 21:48 Ingenic will first try to throw everything in, like lars said "more of everything" 2010-09-01 21:48 then on the way to actual customers, things are thrown out again 2010-09-01 21:48 oh yes, they should definitely do that. and be more open about the specs. it's really bad for open collaboration if you never quite know how much the person you're talking to knows. 2010-09-01 21:48 either because they failed on the engineering side (their people just can't get the stuff to work) 2010-09-01 21:48 well, even the 4750 has some substatial changes. So I doubt that most of the 4740 drivers will work for the 4760 2010-09-01 21:48 or because of packaging 2010-09-01 21:49 one interesting thing in the 4760 is the dual-core 2010-09-01 21:49 (now work for the 4760) that would be unusual. normally, function blocks are just copied with as few changes as possible 2010-09-01 21:49 maybe one core can be dedicated for GPS signal processing or such 2010-09-01 21:49 btw, 4755 is too 2010-09-01 21:50 my rough Ingenic roadmap looks like this: 2010-09-01 21:50 first - meeting with them to talk about upstream (2.6.36). I will ask them to release the 4740 programmer's manual under cc-by-nd. And I will ask them to re-base their work on upstream. 2010-09-01 21:50 Xiangfu does have a 4760 EVB, and we did some hacking on it already, u-boot etc. 2010-09-01 21:51 so I will ask for another 2 EVBs, bringing the total to 3 2010-09-01 21:51 sounds good so far 2010-09-01 21:51 then I can have one for Adam to study, and one for Germany, for exampe if Lars is interested 2010-09-01 21:51 also the 4720_ds, please. you really need both. 2010-09-01 21:52 yup. perfect 2010-09-01 21:52 then I need to talk with them seriously about 4760 packaging 2010-09-01 21:52 so far I have only heard about a 350-ball bga package 2010-09-01 21:52 basically that's the one where everything of the 'more of everything' will be brought out 2010-09-01 21:52 you prefer COB ? 2010-09-01 21:52 no no 2010-09-01 21:52 wait 2010-09-01 21:52 ;-) 2010-09-01 21:52 I am not sure whether this 350-ball bga package will ever really be sold, or demanded by customers 2010-09-01 21:53 ah, i see. well, it's not bigger than the others 2010-09-01 21:53 it could be that there is too much failed and random stuff in there, the down-side of the 'more of everything' strategy 2010-09-01 21:53 so last time the CEO also spoke about 2 other packages they are planning - 256-ball BGA and 128-pin qfp. 2010-09-01 21:54 these things are changing quite a bit, btw. it's super demand driven. 2010-09-01 21:54 not sure just putting more of the same is a big design risk. i'd worry much much more about new things. 2010-09-01 21:54 so I have to find out - which of the 4760 variants (if any) will actually become a high-volume product. 2010-09-01 21:55 I think one reason we had quite a lot of fun with the 4740, including say larsc (let me know if I'm wrong), is that the 4740 is found in many products 2010-09-01 21:55 fun = obscure bugs ? 2010-09-01 21:55 so the kernel work we did applies to Sungale, Hanvon, Dingoo - not just the exotic 'copyleft' Qi stuff 2010-09-01 21:55 no, interesting devices 2010-09-01 21:55 ah, fun = motivation 2010-09-01 21:55 yes 2010-09-01 21:56 also meaningful 2010-09-01 21:56 great kernel.org support for a bunch of EVBs? come on... 2010-09-01 21:56 we are wasting everybody's time 2010-09-01 21:56 bah, just focus on hack value ;-) 2010-09-01 21:56 there are too many things to hack on 2010-09-01 21:56 anyway 2010-09-01 21:56 that's why I stay away from 4750 and 4755 2010-09-01 21:56 they did not 'catch on' 2010-09-01 21:57 so some people are forced into them, but there really is no adoption wave, and never will be anymore 2010-09-01 21:57 oh, i see. 4750 would be a "plan B", feature-wise 2010-09-01 21:57 I will not touch it. 2010-09-01 21:57 rather do Milkymist NanoNote right away 2010-09-01 21:57 heh ;-) 2010-09-01 21:57 my work needs to make sense (create value) from a free software perspective 2010-09-01 21:57 the fact that the NanoNote helped to get 4740 support into kernel.org is good for a lot of projects 2010-09-01 21:58 i niche 4750 would also be a sourcing risk 2010-09-01 21:58 I cannot get lost in tons of proprietary features that are popping up here and there. life is too short :-) 2010-09-01 21:58 sourcing is OK 2010-09-01 21:58 there is only 1 source anyway, so strategically the risk is always the same. and I have a good relationship. 2010-09-01 21:58 so about 4760 2010-09-01 21:58 I need to understand which package may have the chance to catch on. 2010-09-01 21:59 then I know which features are inside (or rather come out in that particular package) 2010-09-01 21:59 (sourcing) i mean, if nobody else uses it and you're small, they may have to scrap it, no matter how much they like you 2010-09-01 21:59 if we do a Ya NanoNote, we are still free to choose - 350-ball bga, 256-ball bga, even 128-pin qfp (although I am afraid that one will have very little of anything) 2010-09-01 22:00 wolfspraul: hey 2010-09-01 22:00 you may want to consider having a 32 bit ram bus 2010-09-01 22:00 also, 4760 supports DDR 2010-09-01 22:00 wpwrak: making another wafer is cheap 2010-09-01 22:00 wolfspraul: just curios about  issues in SIE manufacturing process 2010-09-01 22:00 I think they currently still use 6 or 8'' wafers 2010-09-01 22:00 so you could be up to 4x as fast as the ben 2010-09-01 22:00 so that's maybe 1000 ICs 2010-09-01 22:01 i just read carlos mail about getting time short for hist studets start hacking the boards :) 2010-09-01 22:01 so i just feel curios to ask about :) 2010-09-01 22:01 (making another wafer) hmm, usually things that sound so easy aren't ;-) 2010-09-01 22:01 kristianpaul: no all is fine. there was a couple days delay in getting all tests to work. the components are with the SMT factory already, run should be Friday/Saturday this week. 2010-09-01 22:02 wpwrak: I will ask a bit more about it, always curious to learn these things. 2010-09-01 22:02 I want to focus on high-volume chips anyway. 2010-09-01 22:02 wolfspraul: usual delays i guess? nothign to worry or take care next time? 2010-09-01 22:02 so that's where 4760 stands, and that's why I don't know whether it's 3 SDIO, or 2, or 1 2010-09-01 22:02 yup. safe choice in all regards 2010-09-01 22:02 kristianpaul: the 'delay' Carlos is talking about is relative to the wiki schedule Adam wrote up. 2010-09-01 22:02 wolfspraul: ah ok ;) 2010-09-01 22:03 but that's not really fair. We are writing up the schedule as a first 'good guess', and share it. 2010-09-01 22:03 okay. let's hope for the best. something like 2 SDIO and >= 2 SPI would be nice, preferably >= 3 SPI. 2010-09-01 22:03 two weeks into the work you realize, oh well, you need another week 2010-09-01 22:03 so you push out the schedule 2010-09-01 22:03 you cna't have too many SPI anyway ;-) 2010-09-01 22:03 or SD with SPI mode support 2010-09-01 22:03 if this wiki schedule would have been written by wpwrak it would first be finished in 2 weeks, and 6 months later wpwrak would still be working on very valuable stuff 2010-09-01 22:04 :) 2010-09-01 22:04 so yes, there were 'delays' on the wiki schedule, but in my experience all is smooth and moving forward. 2010-09-01 22:04 the next big thing will be whether the SMT run is successful, and functioning boards come out 2010-09-01 22:04 sure delays happens 2010-09-01 22:04 we know Monday 2010-09-01 22:04 it's not a real delay IMO 2010-09-01 22:04 rather a miscalculation 2010-09-01 22:04 if Carlos would have asked me, I would have doubled all numbers from Adam anyway. 2010-09-01 22:04 wolfspraul: you should know by now that there's no point in asking me for a schedule ;-)) 2010-09-01 22:04 on my schedule, there is no delay. 2010-09-01 22:05 ah thats explain it more ! 2010-09-01 22:05 there you go 2010-09-01 22:05 :) 2010-09-01 22:05 I understand Carlos sentiment however. we are trying to make these boards asap. 2010-09-01 22:05 we have dropped the ball on everything else to get this done. Not more I can do. And neither can Carlos that's why he writes the mail :-) 2010-09-01 22:06 I know the students are waiting since August 10. 2010-09-01 22:06 and they even have paid already by now. 2010-09-01 22:06 wpwrak: kernel modules or programas? i need to write a sort of logic analizaer and draw on screen 3 signals at ~ 2Mhz rate using gpio, what you recommend based on your experience? 2010-09-01 22:07 wolfspraul: sure thats emotinal stress happens after the "cash get out the hands" 2010-09-01 22:07 kristianpaul: you mean kernel or user space ? where does the data come from ? gpios ? 2010-09-01 22:07 i just was worried about some more tecnical process or soemthing 2010-09-01 22:08 well on my side too. I hate pre-pay/pre-order. And essentially I accepted that because I sold the boards before I had them manufactured. 2010-09-01 22:08 what if there is a manufacturing problem now? very bad for me. 2010-09-01 22:08 but like I said, we are breaking all the rules just to get these boards made asap, because the students cannot wait 2010-09-01 22:08 wpwrak: yeah sorry, kerl or user space, from a SPI-like out-put from a IF chips, gpios from ben 2010-09-01 22:09 kristianpaul: sounds more like all user space than kernel of you don't mind running with privileges. that way, you safe the time it takes to get in and out of the kernel 2010-09-01 22:10 wpwrak: before the dsv, I'm curious about ben-blinkenlights 2010-09-01 22:10 kristianpaul: you'll probably need elevated cpu priority. do you write to the screen directly ? 2010-09-01 22:10 you made it yourself? do you need a new kernel driver or can you control the lights with the current kernel already? 2010-09-01 22:10 wpwrak: yeah all to make it fast 2010-09-01 22:11 also because later need do some FFT 2010-09-01 22:11 wolfspraul: yup, i made it. you just run bbl (fro bbl.c) as root 2010-09-01 22:11 of course in that step i dont need write to screen 2010-09-01 22:11 kristianpaul: do you need to respond to interrupts ? 2010-09-01 22:11 i just need to start debugging what i'm reading from the chip 2010-09-01 22:12 wpwrak: dont know yet.. 2010-09-01 22:12 wolfspraul: the idea to use the uSD port came from Ornotermes. he also made a nice 1200 dpi scan of a uSD card that helped to get this done more rapidly 2010-09-01 22:12 wpwrak: so no new drivers needed. just run user-space as root? 2010-09-01 22:12 wolfspraul: yup 2010-09-01 22:13 wolfspraul: bbl just reprograms the gpios. kicks out sd/mmc ;-) 2010-09-01 22:14 i think wpwrak relly like user space :=) 2010-09-01 22:14 (fortunately, sd/mmc doesn't mind. i didn't actually check if this was safe. but if it wouldn't be, i'd know by now :) 2010-09-01 22:14 hey i forgot ask you about reading data using uSD 2010-09-01 22:14 now that wpwrak is making good examples to use it on the openwrt 2010-09-01 22:15 may be i can use uSD instead of gpio.. 2010-09-01 22:15 does your discovery of the sdio possibilities mean that ben-wpan will now be attached to the microSD-connector, rather than on the side of the LCM? 2010-09-01 22:15 kristianpaul: you can tell a good kernel developer by how well he/she is at avoiding to do things in the kernel ;-) 2010-09-01 22:16 kristianpaul: you should be able to read back things without a problem 2010-09-01 22:16 wolfspraul: well, the story is a bit more complicated. for the ya/ben II, i still think the lcm is the place to go 2010-09-01 22:16 wolfspraul: but i can prototype on uSD. and the prototype could become an upgrade for the existing ben. 2010-09-01 22:17 yes I would love to manufacture crazy microSD cards 2010-09-01 22:17 :-) 2010-09-01 22:17 also, you know my opinion on proprietary microcontrollers 2010-09-01 22:17 not that I want to stop anybody 2010-09-01 22:17 but every new proprietary tool or compiler we introduce slows us down 2010-09-01 22:18 I'm talking about that CapSense stuff 2010-09-01 22:18 wpwrak: i'm just really worried that the monolic kernel we use all days, may slowdown acess and fast response talking about GPIO 2010-09-01 22:18 I agree the 8051+cpld is already a bit better. 2010-09-01 22:18 wolfspraul: so this saves me having to worry about tricky cabling for a while. the tricky cabling is particularly bad because i don't even have the right kind of cable (fpc with customized shape), so it'll be an uphill battle all the way, with many annoying failure modes that just distract from what i want to do 2010-09-01 22:18 yes yes 2010-09-01 22:18 wolfspraul: good point about microcontrollers 2010-09-01 22:18 totally agree 2010-09-01 22:19 wolfspraul: so add micro fpgas (kidding about micro) is better for you? 2010-09-01 22:19 so with any (new or old) proprietary ICs, at least I want to understand how they help on the path to more freedom. if I have such a path in mind, the proprietary IC becomes acceptable. 2010-09-01 22:19 or just re-ue and avbuse from ingenic xbust as much as we can ;)? 2010-09-01 22:19 yes sure, if that means some of the work we do in the micro-fpga can be carried forward into our own ICs later 2010-09-01 22:20 wolfspraul: that was just to answer carlo's question. i don't quite see capacitative sending in the main keyboard. and for the slider/softbuttons next to the screen, i think having to add an MCU just for that would have a poor cost/benefit ratio 2010-09-01 22:20 if anything I want to make my own ICs 2010-09-01 22:20 we need a good (small) starting point for that anyway 2010-09-01 22:20 wolfspraul: so i'm all for not having MCUs unless we really really have to. and of course, the more open and better supported with tools and the like, the better 2010-09-01 22:21 we are on the same page 2010-09-01 22:21 wpwrak: i hope you order from digikey some cpld's too ? :) 2010-09-01 22:21 and if it's a fpga/cpld programmed in verilog/vhdl, and we can later carry this forward into a bigger fpga, or into our own asic, then that's the way to go 2010-09-01 22:21 wolfspraul: in fact, i'll probably kick the MCU from IDBG and switch to FT232R. the latter now seems competitive and we don't really need the ability to run our own code in IDBG on the ben. 2010-09-01 22:21 wolfspraul: (in case that changes, i can always bring the MCU back, of course :) 2010-09-01 22:22 FT232R ! 2010-09-01 22:22 kristianpaul: no cpld/fpgas for me for now. first, lekernel has to write his Open/Free fpga synthesizer ;-) 2010-09-01 22:22 wpwrak: wut? 2010-09-01 22:23 well. 2010-09-01 22:23 a PHd take time i guess 2010-09-01 22:23 kristianpaul: i'm not particularly fond of those ft232 chips, but in this case, it just looks like the better solution. slightly fewer external components, no programming required, similar price, same size. 2010-09-01 22:24 wpwrak: i dont understand you said not to fpga/cpld but still using propietary Micro-controllers, so whats diferences? 2010-09-01 22:24 kristianpaul: ft232r is a black box for me. i don't care about what's going on inside. it does serial and bit-banging for me. that's enough. 2010-09-01 22:24 wpwrak: indeed 2010-09-01 22:25 same with micro-controllers? 2010-09-01 22:25 kristianpaul: microcontrollers, i have to program. so they're not a black box 2010-09-01 22:25 even when you know you can implement one SoC with the already made MM ? 2010-09-01 22:26 wpwrak: you program a blackbox.. 2010-09-01 22:26 wpwrak: sorry I have a stupid question for you 2010-09-01 22:26 the manufacture just give you the manual wich tell what is inside the box 2010-09-01 22:26 you buy it 2010-09-01 22:26 use it 2010-09-01 22:26 kristianpaul: fpgas are tricky. i would generally want them for the ability to hack them. but yes, if i just needed it to load mm and be happy with it, fpga would be okay. 2010-09-01 22:26 and belive it works as he said 2010-09-01 22:27 kristianpaul: i also don't mind that my usrp has a proprietary fpga. again, just a black box for me. 2010-09-01 22:27 wolfspraul: fire away :) 2010-09-01 22:27 someone wants to manufacture a wifi design around xburst with me. he says he can give me the entire design, and I can open it up if I want to. 2010-09-01 22:27 sure fpga uses propietary sinteshys sofware but at least you can model some digial hardware, is that bad?? 2010-09-01 22:27 wow 2010-09-01 22:27 :O 2010-09-01 22:27 so now, he writes "we have a power amplifier that brings RF power to antenna to 23-25dbm, I guess you don't need that" 2010-09-01 22:28 wolfspraul: what are the semantics of "entire" ? :) 2010-09-01 22:28 oh, it's still a closed wi-fi chip, nothing special, not on our level here :-) 2010-09-01 22:28 I am just trying to understand what he means with "we have a power amplifier ... to 23-25dbm" 2010-09-01 22:28 and why I would or would not need that 2010-09-01 22:29 most transceiver chips have a relatively low power output 2010-09-01 22:29 (that is, transceicers with built-in power amplifier) 2010-09-01 22:29 so sometimes you add another one to boost the signal more 2010-09-01 22:30 ah so it's stronger than usual? 2010-09-01 22:30 this one would be 300-320 mW 2010-09-01 22:30 err, 200-320 mW 2010-09-01 22:30 yes. 100 mW would be normal for wifi 2010-09-01 22:30 thats more a my wrt54gl, isnt? 2010-09-01 22:30 oh 2010-09-01 22:30 (if i got all the regulations right ;) 2010-09-01 22:31 so with that power amplifier it would not be cerfitiable as a regular wi-fi retail ap by the fcc? 2010-09-01 22:31 s/cerfitiable/certifiable/ 2010-09-01 22:31 :-) 2010-09-01 22:32 kristianpaul: according to mine, it can go up to 251 mW 2010-09-01 22:32 wpwrak: wrt54gl? 2010-09-01 22:32 how hard is it to remove such a power amplifier from a design? 2010-09-01 22:32 wolfspraul: dunno. it may be a case of "as long as you promise not to use it where you're not allowed to" ... 2010-09-01 22:33 wolfspraul: 100 mW isn't strictly the limit of the regulations. you can go up to 1 W, but with additional constraints 2010-09-01 22:33 1 W :O 2010-09-01 22:33 wolfspraul: at least in FCC land. EU is another story again. 2010-09-01 22:34 kristianpaul: yup 2010-09-01 22:35 wolfspraul: i read most of the regs with the question "could we be in conflict with this one ?", so i'm not so sure on the things that i can exclude a priori, e.g., xxl power amps 2010-09-01 22:36 for reference, my wrt54gl is set to 42 mW output by default 2010-09-01 22:36 ahh i knew it ! 2010-09-01 22:37 (remove the PA) should be easy. you're likely to find more documentation on the design without extra PA than the one with it. 2010-09-01 22:37 but that up to 251 mW is a openwrt thing or firmware that mye not obey to you? 2010-09-01 22:38 except if your transceiver needs a PA either way. they you would have to find a different PA 2010-09-01 22:38 kristianpaul: dunno. i just see that it offers 251 mW and i assume its not lying to me :) 2010-09-01 22:38 k 2010-09-01 22:39 well .. let's see what happens when i set it ... 2010-09-01 22:39 ah you never tried that before? ! 2010-09-01 22:40 nope. i just knew that tx power is usually something you can change, yet i rarely do 2010-09-01 22:40 most of the time, the ap's tx isn't the problem anyway 2010-09-01 22:41 ok I think I got it, roughly. 2010-09-01 22:41 well if he pays me enough money and I can open up some nice (if high-level) design files in the process, I'll manufacture the stuff for him :-) 2010-09-01 22:42 we won't crack open the wifi thing like this ever, anyway 2010-09-01 22:42 unfortunately 2010-09-01 22:42 :( 2010-09-01 22:43 wpwrak: did you digi-key shipment make it through customs? 2010-09-01 22:43 wolfspraul: still sitting there. they're slow today :-( 2010-09-01 22:44 wolfspraul: never had problems with things from digi-key, though. so unless they don't like the antennas for some reason, i don't expect any this time either. 2010-09-01 22:46 wolfspraul: they once dropped a shipment from openmoko off the regular process and made me clear customs myself (trip to the airport, day lost going from one office to the next, etc.) because they thought this could be something to crack stolen phones. 2010-09-01 22:46 wolfspraul: little do they know :) 2010-09-01 22:46 wpwrak: 'dropped off the regular process' - was this with fedex? 2010-09-01 22:47 wolfspraul: not sure if it was fedex or dhl. it wsn't ems. (ems goes elsewhere) 2010-09-01 22:47 wpwrak: how do you argue those itemes were not to crack phones? 2010-09-01 22:49 kristianpaul: i just tell them what they are for. and i don't explain to them that i wouldn't need all this if i really wanted to set up a phone cracking shop, and that, yes, i could explain them in detail how to do it. well, almost :-) 2010-09-01 22:50 in reality, the complexity of all this has long surpassed what governments can actually regulate 2010-09-01 22:51 but in a move of massive inertia, they are able to hold up the perception of regulation for a bit 2010-09-01 22:51 heh 2010-09-01 22:51 wpwrak: I was just asking because fedex claims to have their own customs brokers, so with fedex theoretically it should never 'drop out of the regular process' 2010-09-01 22:52 the broker will always drive the process, maybe unless you really smuggle some drugs or weapons or fissile material etc... 2010-09-01 22:52 wolfspraul: yeah, when talking to those customs guys, it's always tricky to assess where exactly between sticks and stones and the transistor they are :) 2010-09-01 22:52 well they have no chance 2010-09-01 22:52 it's a show 2010-09-01 22:52 but the show must go on 2010-09-01 22:52 ah no, fedex, dhl, they all have their brokers. they do the simple process. 2010-09-01 22:53 dsv (data sheet viewer) - no I didn't know yet. tell me more... 2010-09-01 22:54 what is that? where can I see it? 2010-09-01 22:54 if it drops out of this, you're on your own. they don't even have a broker they can recommend you. (although that may be because of some arrangements/regulations, not because they wouldn't want to help you) 2010-09-01 22:54 (dsv) check out ben-wpan 2010-09-01 22:54 then  make dsv 2010-09-01 22:54 wait until it's done. then 2010-09-01 22:54 scripts/dsv antenna 2010-09-01 22:54 scripts/dsv at86rf230 2010-09-01 22:54 scripts/dsv mcu 2010-09-01 22:54 etc. 2010-09-01 22:54 or just scripts/dsv 2010-09-01 22:55 note the time between experiencing the desire of seeing a data sheet and the moment it appears :) 2010-09-01 22:56 http://paste.debian.net/87209 how do i solve this? 2010-09-01 22:56 wolfspraul: do you need my public key? where i send it? 2010-09-01 22:57 wpwrak: yeah it's running "make dsv" now. It must be doing something _really good_, given how long it takes... 2010-09-01 22:59 *gargh* 2010-09-01 22:59 openssl broke my lighty 2010-09-01 23:00 http://redmine.lighttpd.net/projects/lighttpd/repository/revisions/2716 2010-09-01 23:00 wolfspraul: it's just downloading stuff 2010-09-01 23:00 kristianpaul: there are two ways to access the repositories on the projects server 2010-09-01 23:00 just in case some of you run into the same problem 2010-09-01 23:00 git@ and git:// 2010-09-01 23:00 roh: awhttpd for the win :) 2010-09-01 23:01 the git@ requires you to have write access to that repository (be an admin or member, and have a ssh pubkey uploaded for your account) 2010-09-01 23:01 but in the meantime, if you just want to get the sources, try git clone git://projects.qi-hardware.com/ben-blinkenlights.git 2010-09-01 23:01 i think that message you be fixed here http://projects.qi-hardware.com/index.php/p/ben-blinkenlights/source/tree/master/ 2010-09-01 23:02 it tries to be smart and sometimes tell you to use git@, sometimes git:// 2010-09-01 23:02 :-) 2010-09-01 23:02 instead of explaining the whole situation 2010-09-01 23:02 it could show both :) 2010-09-01 23:02 if you log out from your account, it will give you the git:// URL for sure 2010-09-01 23:02 yes I agree 2010-09-01 23:03 i wasnt logged frist time i try 2010-09-01 23:03 okay i uploaded my public key ! 2010-09-01 23:03 ok can you try the command I just gave you? the git:// url. what do you get? 2010-09-01 23:04 ah sure that works 2010-09-01 23:04 okay, maybe that also works as long as you only read. you should not be able to commit unless your account is an admin or member of the projects (depending on projects settings, but that's the default) 2010-09-01 23:04 now the git@ too :) 2010-09-01 23:04 wpwrak: ok yes, it is really downloading some PDFs 2010-09-01 23:05 some sites are a bit on the slow side :( 2010-09-01 23:05 btw, quite a few websites have fine print against 'deep links' 2010-09-01 23:05 wolfspraul doesnt support ssl at all 2010-09-01 23:06 they pull it out when their legal dept has you on the radar 2010-09-01 23:06 "pull it out" = change the path ? or make you not link ? 2010-09-01 23:06 (try to) 2010-09-01 23:06 it's not really enforced until there is a bigger reason to go after you 2010-09-01 23:07 wpwrak: for example, http://www.mips.com/terms-of-use/ 2010-09-01 23:08 "In addition, you may not "deep-link" or use robots, spiders or other automatic devices, programs, algorithms or methodologies to monitor, access, copy or acquire any Materials or portions of this Site." 2010-09-01 23:08 so if your script goes to such a site, it violates the terms of use of that site 2010-09-01 23:09 sorry, just fyi (you do this to me with 'regulations' all the time :-)) 2010-09-01 23:09 gn8 2010-09-01 23:09 there doesn't seem to be a safe way to access such sites then 2010-09-01 23:09 even written instructions in plain english could be read by a script, etc. 2010-09-01 23:09 you have to describe the steps to go to the document you mean in prose ("go to the homepage. click on ..., then click on ..., then ...") 2010-09-01 23:10 no need to argue with me, I just quote the terms of use :-) 2010-09-01 23:10 it's my regulation 1:1 with you... 2010-09-01 23:10 we are even now 2010-09-01 23:10 ;-) 2010-09-01 23:10 they will pull this out if they have you on the radar for something 2010-09-01 23:10 of course they will. they have lawyers sitting around who need to justify their jobs. 2010-09-01 23:11 so the solution would be something that actually parses step by step instructions 2010-09-01 23:11 I guess it's still a "program, algorithm or methodology"? 2010-09-01 23:11 I don't know, it's stupid. 2010-09-01 23:11 let's ignore this shit 2010-09-01 23:11 :-) 2010-09-01 23:11 so would be any description a human being can be reasonably expected to repeat accurately :) 2010-09-01 23:12 hey: and this is logged! 2010-09-01 23:12 yeah, ignoring it seems to be the best approach :) 2010-09-01 23:12 if they don't like it, one can always send it to some warez site and link that one ;-) 2010-09-01 23:13 [commit] Juan José Díaz Vecchio: Aded 2 phototransistors; removed accelerometer, i2c http://qi-hw.com/p/sie-ceimtun/710236c 2010-09-01 23:14 wpwrak: look here, under "supporting documentation" 2010-09-01 23:14 http://opencores.org/project,plasma 2010-09-01 23:14 the "deep link" stuff usually comes up when you try to pretent the content is yours. that's not the goal here at all 2010-09-01 23:14 they can pull it out whenever they like to. 2010-09-01 23:15 if you go down to 'supporting documentation', you have the legally accepted way to link to the PDF on the mips.com site 2010-09-01 23:15 right before "Bit/little endian" - do you see it? 2010-09-01 23:16 yup, i see it. hmm, isn't the full title, combined with google, as "deep link" too ? :) 2010-09-01 23:18 wolfspraul: the URL "dsv" uses are in the file BOOKSHELF. so would you rather not have them on projects ? 2010-09-01 23:18 no no 2010-09-01 23:18 relax 2010-09-01 23:18 :-) 2010-09-01 23:18 just checking :) 2010-09-01 23:18 I just bring this up because it's a fact, same as the regulations you are pulling up. 2010-09-01 23:18 I don't know how many tech sites have deep links rules in their terms of use now. 2010-09-01 23:19 but I could imagine that it grows, for sure. the lawyers are writing and writing and writing... 2010-09-01 23:19 and they have the law on their side 2010-09-01 23:19 so if they would ever understand what this script does, they will write you a nice letter to stop violating their terms of use. 2010-09-01 23:19 if you are lucky, the letter does not come with an invoice. 2010-09-01 23:20 yes. i know of the issue in general. but also all the work-arounds don't quite convince. 2010-09-01 23:20 the best is to get rid of proprietary IP altogether 2010-09-01 23:20 kick all those comanies out 2010-09-01 23:20 they have too much money, for the wrong things, obviously 2010-09-01 23:20 find a cure for lawyers ;-) 2010-09-01 23:21 in this case it starts with proprietary IP that they need to 'protect' 2010-09-01 23:21 the deep link nonsense is just one small little tool in the 'protection' arsenal 2010-09-01 23:21 but if you think they won't use it, for whatever reason, well, wrong 2010-09-01 23:21 they will 2010-09-01 23:22 i wonder how alldatasheets do it. after all, they even have copies. now that should enrage the nazguls. 2010-09-01 23:22 the opencores plasma homepage is so clean because they got some help from MIPS :-) 2010-09-01 23:22 lovely 2010-09-01 23:22 one of these days i'll write that english -> url parser :) 2010-09-01 23:23 I don't know how alldatasheets does it 2010-09-01 23:23 maybe somewhat undergroundish? 2010-09-01 23:23 is there a legal entity behind it? where are they located? where are their servers? etc. etc. 2010-09-01 23:23 it's a big world :-) 2010-09-01 23:23 and what is their legal strategy? react quickly to take-down notices? 2010-09-01 23:23 there are many strategies 2010-09-01 23:23 obviously the servers are still there, so whatever their strategy is, it seems to work in the real world 2010-09-01 23:24 maybe it's time for making a little satellite with a few TB of storage :) 2010-09-01 23:24 "serve *that* task-down notice" :) 2010-09-01 23:24 there you go 2010-09-01 23:24 s/task/take/ 2010-09-01 23:24 good idea 2010-09-01 23:24 the freedom satellite! 2010-09-01 23:24 I love it 2010-09-01 23:24 PCBs made in werner's diy lab 2010-09-01 23:25 ;-)) 2010-09-01 23:27 anyway, what dsv tries to solve is the generally much worse IP violations caused by the need to send around copies of data sheets 2010-09-01 23:27 need or perceived need in this case 2010-09-01 23:27 we launch it, and it hosts datasheets, ha ha! 2010-09-01 23:27 :-) 2010-09-01 23:27 but unfortunately I read that both the Chinese and Americans are working on the ability to take down satellites 2010-09-01 23:27 even 'dock' on one for 'servicing' 2010-09-01 23:28 we can add some active meteorite protection :) 2010-09-01 23:28 oh 2010-09-01 23:28 you are trying to circumvent non-redistribution clauses? 2010-09-01 23:28 nice :-) 2010-09-01 23:28 plus make it all a bit more efficient 2010-09-01 23:37 wpwrak: ok dsv finished and it all works. so it's just a way to download files into a cache and view them. 2010-09-01 23:38 yup 2010-09-01 23:38 it's nice, a way to document urls 2010-09-01 23:39 unfortunately like I said, with many of those datasheets there may be some nasty terms of use associated with those urls and pdfs 2010-09-01 23:39 the idea is that each project user/participant can get the document set this way efficiently 2010-09-01 23:39 yes. i guess we really need that english to url parser :) 2010-09-01 23:39 yes, unless there could be a central repository somewhere of greater stability/longevity than these vendor urls, it's good 2010-09-01 23:40 and such a central repository is too hard to setup right now, so it's the best compromise 2010-09-01 23:40 of course, there's also the question of whether "dsv" falls under the things they can prohibit 2010-09-01 23:40 e.g., the MIPS exclusion would also include your Web browser 2010-09-01 23:41 the central repository would also be a pretty clear violation of traditional copyright law 2010-09-01 23:41 ah sure, I try to focus on realistic threats only. we just need to keep it in mind, those policies and people to enforce them are out there. 2010-09-01 23:42 and we are not planning to steal any of their wonderful intellectual property anyway, so it's definitely different from a company that wants to sell cloned ICs, for example. 2010-09-01 23:43 but unlike the schematics diff scripts, dsv is of no use on the server, I think 2010-09-01 23:43 (running on the server) 2010-09-01 23:43 yup. we're not the enemy. of course, that doesn't have to prevent their lawyers from becoming ours 2010-09-01 23:44 at least when you view BOOKSHELF in the tree (on projects), the URLs just appear as plain text. so we're not worse than heise.de :) 2010-09-01 23:44 no, dsv is the contrary of being something for the server :) 2010-09-01 23:45 but if you can get companies to make things available at least at the level at which dsv can use them, that would be a win 2010-09-01 23:48 http://www.atmel.com/general/legal_information.asp 2010-09-01 23:48 strange. if I read this correctly, they only give you a license for non-commercial use. 2010-09-01 23:49 but I cannot imagine that they mean no professional (paid) party can download and use their documentation 2010-09-01 23:49 so maybe they mean the use of the documents per se must be non-commercial 2010-09-01 23:49 so I cannot act as a print-shop and print-out the PDFs for a fee 2010-09-01 23:49 probably... who knows... 2010-09-01 23:50 I mean I can download the PDF, go down to the street here and have a nice booklet sometime this afternoon still :-) 2010-09-01 23:51 they seem to equate "use" with "reproduction" 2010-09-01 23:52 "termination of this license" ... "Atmel may terminate this agreement immediately without notice, for any reason. Upon such termination, you must destroy all Materials." 2010-09-01 23:52 so, if they dojn't even notify you, how do you know ? :) 2010-09-01 23:53 well basically the rule is: whatever Atmel wants, will happen. 2010-09-01 23:54 nothing new then :) 2010-09-01 23:55 btw, regarding potential IP problems, the antenna is probably the riskiest bit. after all it's a TI design we're using with an Atmel product. the antenna document doesn't mention any restrictions, so it's anyone's guess if this passes as "fair use" or not 2010-09-01 23:55 I understand, but let's just move forward. 2010-09-01 23:56 the most interesting part is that you document the process, test results, etc. 2010-09-01 23:56 that gives us the knowledge, hopefully, one day if someone comes and bullies us out of a particular thing, to come up with a replacement fast 2010-09-01 23:57 worst case, we can always fall back to a chip antenna 2010-09-01 23:58 for example I have the feeling the setup of your apartment, with diy pcb etching, usrp, puts you in a position to out-innovate quite a few of the dinosaurs 2010-09-01 23:58 incidentally, there are a few cheap ones in my digi-key order :) 2010-09-01 23:59 so let's see what you find, what you think works, and then we proceed and I'm fairly optimistic we will not wake up the sleeping dogs too early 2010-09-01 23:59 trial and error cycles can indeed be short :) 2010-09-01 23:59 yes :-)