#qi-hardware IRC log for Monday, 2011-02-21

rohwpwrak: seen this? http://ubertooth.sourceforge.net/hardware/one/00:49
wpwraknice. proprietary wireless, isn't it ?00:52
rohwpwrak: ? well.. he is building something which is bt compatible i guess00:57
wpwrakwhy do you think it's BT compatible ?00:58
rjeffriesany hints on how to set date and time on dgClock on Ben?00:59
wolfspraulroh: did you see the case kristianpaul made in Colombia?00:59
rohyes. nice one01:00
Action: kristianpaul and finally with buttons01:00
wolfspraulindeed. that is testimony to the great work you did!01:00
wolfspraulstandard file formats, high quality files, well documented, etc.01:01
rjeffrieswpweak well, he says it is used to snigg Bluetooth01:01
rohmaybe somebody should talk to michael ossmann01:01
rohi mean.. he also uses kicad and such.01:01
rjeffriess/wpweak/wpwrak  what an  BAD typo01:01
rjeffriesroh agree he has his heart in right place01:02
wolfspraulkristianpaul: how did you do the buttons?01:02
rohi think one needs to show how many examples of small innovative foss hardware projects there are01:02
kristianpaulwolfspraul: a cilender from the dimensions roh told me, i use my floss 3d printer.. some glue and a small sheet of plastic01:04
wpwrakrjeffries: ah, sniff BT. yup, that sounds good. funny TI don't mention that compatibility.01:04
rjeffrieswpwrak yeah maybe not somthing they wish to advertise. I assume it is si,ilar to promiscious mode with wifi or ethernet01:06
rohrjeffries: nah. to paranoid thinking. bt usually means 'less possibilities, higher integration, also sell some fw and a cpu'01:06
rohsee the csr based stuff.01:06
rohbasically all usb bt 'chips' you can find are csr based.01:07
rjeffriesneed a bit of help: how does one set date time in dgClock01:07
rjeffriesroh ok, but for an end user having BT so one can hook up keyboard etc would be nice01:08
rjeffriesBY chips are so cheap now01:09
rohsure they are cheap. but not open. basically its a 'hardmac blackbox'01:09
rjeffriesroh I understand you01:10
rohthats what the sniffer project is about. doing all the stuff happening in the blackbox in sw on that arm cpu (its a arm in the csr devices too afaik)01:10
rjeffrieswe will soon have wpwrak's 802.15.4 stuff and 6LoWPAN but if somebody interfaces BT it will be useful option01:11
rohbasically a reversal of the miniaturisation which happend before (putting frontend, rf mod/demod and arm cpu in one chip)01:11
rohi dont think his goal is building a complete bt dongle fw01:11
rjeffriesroh so what is not to like about Ubertooth? sound sgreat to me01:11
rohits more a development tool for the lower levels.01:12
rohdifferent goal. but yes. i assume the hw could be used as such given somebody writes the firmware01:12
rjeffriesI know that. yeah ok womeone else may however01:12
rjeffrieswolfspraul have you used dgClock?01:13
rohthats the difficult part on opensource hw. one needs to be careful when differenciating what 'is possible in theory' and what 'is the goal and people work on'01:13
rjeffriesroh you are preaching to the choid. I get it. but one can dream for gosh sakes01:14
rohsure ;)01:14
rjeffrieswhen I use alt enter to exit GMU ot works, but starts dgClock i thought I;d end up in a shell01:16
wolfspraulI'm not using dgclock other than checking whether it starts, and it does.01:21
kristianpaulyou can type date on the shell >-)01:21
kristianpauli remenber it closed with control + q. may be you typed something else rjeffries ¡01:22
zrafalarsc: xiangfu : which is the size of kernel partition on nn nand? I would like to know how many bytes we have for a kernel+initramfs01:40
xiangfuzrafa: 4M01:41
zrafaxiangfu: ah.. interesting :).. thanks01:41
xiangfubootloader 4M, kenrel 4M,01:41
wolfspraulzrafa: next time any of the hard-coded partitions get rearranged, we will increase this for sure01:44
wolfspraulit was a bad decision to only make it 4 M01:44
wolfspraulwith a 2 GB NAND there is no point in not making it 16 MB or so01:44
wolfspraulor 32 even01:44
wolfspraulbut we will wait until we have a lot of thinking/pros/cons before we make any more change to the hard-coded partitions01:44
wolfspraulincluding things like ubifs attach time, multiple rootfses in one partition (jlime/openwrt dual booting), etc.01:45
wolfspraulso for now it stays at 4M :-)01:45
zrafawolfspraul: yep, 16MB would be nice . BUt well, I want to try a minimal ramdisk for new ideas, so let me see if I get something into 4M01:47
wolfspraulxiangfu: did you see what David wrote up about increasing ubifs attach time? http://en.qi-hardware.com/wiki/UbiAttachTime01:47
xiangfusorry. not yet.01:47
wolfspraulmaybe supporting sub-page reads can speed up mounting...01:47
xiangfuok I will read that.01:48
wolfspraul'decreasing ubifs attach time' I meant01:48
wolfspraulor 'increasing' the speed :-)01:48
wolfspraulwe can also try to increase the erase block size01:49
wolfspraulneed to think about the pros and cons of that01:49
wolfspraulbut if the sub-page reading theory is right, that looks like a very clean way to make things faster without any downside01:49
rjeffrieshow does one get to the shell? alt enter escapes from the launcher but dumps me intp dgClock from which ctrl-Q does nor escape01:51
rjeffriessorry to bore you guys but as a new Ben user I simply wish to get to a shell, any shell. is there something I can do during the boot sequence maybe?01:54
rjeffriesat what point wolfspraul01:54
wolfspraulafter booting into gmenu2x01:54
wolfspraulyou can go back to the gui with ctrl-alt-f501:54
rjeffriesok will try next thx01:54
wolfspraulin gmenu2x, there should be icons for terminals too01:55
rjeffrieson mine that does not see  to work sorry01:55
wolfspraulyou see the icons, you navigate to them and press <enter>, but the terminal won't come up?01:55
wolfspraulin some old gmenu2x you had to press 'x' to start something, I vaguely remember...01:56
rjeffriesall I see is applications and settings at top of launcher then two more rows one with dgclock,gmu, nanomap01:56
rjeffriesthe bottow row is stardict and explorer01:57
wolfspraulsounds like an old rootfs01:57
rjeffrieslet me get to shell now01:57
wolfsprauldo this: ctrl-alt-f2, <enter>, cat /etc/VERSION01:57
rjeffriesnow in shell thx01:58
rjeffriescat /etc/shell01:58
wolfspraulwhat does cat /etc/VERSION tell you?01:59
rjeffriesok that's cool. when the image now being tested goes gold I'll reflash02:03
rjeffriesyhanks all02:03
Action: xiangfu found one picture. http://3.bp.blogspot.com/_3pwU7-9IKfs/SuFlfOJb4xI/AAAAAAAAAmU/5roOipiyFeM/s1600-h/SDC10881.JPG02:06
xiangfuRichard Stallman, Ben NanoNote, Xiangfu :)02:07
wolfspraulha, cool!02:07
wpwrakheh ;-)02:07
wolfspraulcan we upload this into the wiki?02:07
rjeffrieswhen/where was that photo taken?02:08
rjeffriesnow we know what xiangful looks like02:09
xiangfuI can contact the author, then upload to wiki.02:10
wolfspraulxiangfu: sure, good idea02:10
kristianpaulnice pic xiangfu02:10
xiangfuI think this author.02:11
kristianpaulxiangfu: what stallman said about the nanonote?02:12
rjeffriesthanks xf can I call you "xf?" xiangfu //smile//02:12
xiangfuhmm.. he said re-product the hardware cost money.02:13
xiangfunot like re-compile software.02:13
kristianpaulThanks for all this wolfspraul ;-)02:15
kristianpaulAnd also all people buying nanonotes and mm1's :-)02:18
kristianpaulxiangfu: nice try :-)02:18
kristianpaulxiangfu: btw you have a mm1 to isnt?02:18
zrafafirst try kernel+initramfs 7.5M02:19
wolfspraul4M will be very hard02:19
zrafalet me remove stuff and do the second try02:19
zrafawolfspraul: yeah :)02:19
wolfspraullike I said we made a mistake with the 4M02:19
wolfspraulin hindsight vision is always 20/20...02:20
kristianpaulstuff = ?.. no kernel :p02:20
wolfspraulbut... we make mistakes, then we fix them. for sure. next time around 4M will be increased to give more flexibility. the 'lost' few M are totally irrelevant in comparison.02:20
zrafawolfspraul: yep, you are right. 4M seems very hard. But well, let me see..02:21
wolfspraulkristianpaul: no xiangfu doesn't have m1 yet02:22
wolfspraulbut very soon, we are working on getting the pieces together. importing to China is a hassle too nowadays.02:22
wolfspraulbut yes, the plan is definitely that xiangfu starts hacking on m1, which I am very much looking forward to.02:22
wolfspraulI am so happy about the fantastic progress xiangfu has made over the years I work with him.02:23
wolfspraulfirst time we met his English was so bad that he took some time at home to write down what he wanted to tell me in English, on a piece of paper.02:23
wolfspraulnot that my Chinese would be any better...02:23
wolfspraulsince then it feels we must have learnt like the amount of a full library :-)02:24
kristianpaulwpwrak: 10cms worked well, at least for the logic analizer :-)02:24
kristianpaulwolfspraul: (Chinese) and you actually have stardict on the nanonote... ;-)02:25
wolfspraulyes, but! the chinese dictionary was removed in the last image, I think. and it has no pinyin, so it doesn't help me.02:25
wolfsprauland no way to look up characters either.02:25
wolfspraulin terms of practicaly usefulness those apps are still by far lacking any normal electronic dictionary you can buy for 20 USD.02:26
kristianpaulyeap :/02:26
kristianpaulOther night i was thinking i a single feature nanonote, ie translation tool :-)02:27
wolfspraulwe get there, the pieces are coming together. I am very patient. Most important for me is maintenance overhead, upstream everything.02:27
xiangfuzrafa: what packages include in your  kernel+initramfs. mine is 2.6M02:27
xiangfuzrafa: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.xbboot-02152011-0100/02:27
wolfspraulmakes no sense that we focus our tiny resources on one app now, then a) everything else falls behind b) most likely we will make lots of ugly/quick hacks that will then come down again as we update the system02:27
kristianpaulupstream by sure02:27
rjeffriesmy son today asked about a single application graphing calculator version. could be successful02:28
wolfspraulso as much as I like to have fewer perfectly working apps, instead of many bad ones, the maintenance/upstream priority still trumps that02:28
wolfspraulwe need both02:28
kristianpaulThre is one in Jlime and last image i think so02:28
rjeffrieskristianpaul how useable is that app02:28
zrafaxiangfu: I do not know. I just took the bootstrap image and I am checking its content02:29
kristianpaulrjeffries: ahm sorry i regret02:29
kristianpaul"graphing calculator"..02:29
xiangfuzrafa: I am using openwrt build the image.02:29
kristianpaulxiangfu: your blog i okay?, long time since it dint reponse, well at least on my side...02:30
zrafaxiangfu: ah.. your link has a list of packages. Nice. Is your ram fs working?.. I mean, you get a shell, you can run some script, etc?02:30
wolfspraulI think his blog is on a Chinese server.02:30
zrafaxiangfu: I was thinking to use myrootfs from werner02:30
kristianpaulhmm :-/02:30
kristianpaulweird, with tor it works..02:31
wolfspraulthe problem is if he moves it to a foreign server, it will become slower/harder to access from within China, and infact depending on where he hosts it, say wordpress.com, it may become totally inaccessible without tunneling.02:31
xiangfuzrafa: yes. it works fine.02:31
wolfspraulif he moves it to a CHinese server, there are all sorts of difficulties about slowness, expensive, etc. I think one time Xiangfu told me to get your own IP on a Chinese server alone costs 50 USD / month or so.02:32
xiangfuzrafa: there only mtd stuff in that image.02:32
wolfspraulor even more.02:32
xiangfuzrafa: once we have such image "zImage" , we can using 'xbboot' to test image. no need to reflash.02:32
kristianpaulwolfspraul: so the idea still  fewer apps, good!, i was getting scare with all the  news apps coming and problems, and overhead... sure, contributions are good, but in owrt tree ;-)02:32
wolfspraulit's easy to agree to a statement like "fewer perfect apps are better than many bad ones"02:33
wolfspraulthe reality is that there are many conflicting priorities02:33
zrafaxiangfu: great. Thanks to let me know. Let me see if I get something similar. About xbboot: that is better yet. I was thinking how to test it without reflash all the time and eventually breaks uboot partition or similar :)02:33
xiangfukristianpaul: my blog is using DDNS, I don't have a static IP :(,02:33
wolfspraul100% upstreaming, making contributors happy that help bring apps they want/need to the device, testing overhead02:34
xiangfuzrafa: sudo xbboot -u 0x80600000 /PATH/TO/your_zImage.02:34
xiangfuxburst-tools, include 'usbboot' and 'xbboot'02:34
xiangfuzrafa: I think xbboot can handle 6M image.02:35
kristianpaulxiangfu: same here, but yours seems take a while to get updated in records, i' using freedns.afraid.org just in case you need a backup :-)02:35
zrafaxiangfu: I have xburst-tools, so let me check if xbboot is already there02:35
xiangfukristianpaul: thanks.02:35
zrafaxiangfu: yep, it is there. cool02:35
zrafaxiangfu: OE has a minimal-with-mtd recipe target for initramfs. I just found that. No idea how useful is that. Let me check all the options (myrootfs, your list of packages and build the rootfs, some OE recipes). Thanks a lot for the links and tips02:38
kristianpaulwolfspraul: (conflicting priorities)  i'm not sure but with the nanonote i think lots of command-line apps already in openwrt with seems less from care now can find some loving, i mean maintenance, but sure thats not a load for just qi02:38
wolfspraulkristianpaul: hey nice! http://en.qi-hardware.com/wiki/File:Sigemm1.jpg02:44
steve|mmilky-GPS? ;)02:45
kristianpaulwolfspraul: i have to do it, or i never will goint to start the migration :-)02:46
wolfspraulsteve|m: I remember one thing about your sanding... Which sandpaper did you use?02:46
kristianpaulwolfspraul: btw scope get VERY handy when doing this !!02:47
steve|mwolfspraul: uh, yeah.. documenting this is still on my todo list ;) I used corning 120, and 400 for polishing it before scanning02:48
wolfspraulkristianpaul: well, that picture looks like you got it all wired up already, now just move software over.02:48
steve|mtoday I was busy with documenting another project: http://www.youtube.com/watch?v=HZdV2iKakqA / http://wiki.steve-m.de/doku.php/epaper_display02:48
wolfspraulsteve|m: if it's still on your todo list even better, but in case it stays there forever (like 80% of my todo list items do), we just get the most important tidbits communicated upfront ;-)02:48
kristianpaulwolfspraul: yeah, working on it02:49
wolfspraulso you first desolder all components02:49
wolfspraulthen you start with corning 120 paper02:49
wolfspraulthen once you see the structures of the next layer, you switch to corning 40002:49
steve|mwolfspraul: right, I desoldered the components on both sides using hot-air..02:49
wolfspraulhow much pressure do you apply with the sandpaper?02:49
wolfspraulfrom one layer to the next, do you remember roughly how long you used the 120 paper, and how long the 400 one?02:50
wolfspraulis it like 'a few minutes' with 120, then another 'few minutes' with 400?02:50
wolfspraulor longer?02:50
kristianpaulwolfspraul: i have to choose between bugfix memory flash or ethernet bugs in order to still getting decent raw samples (8Mb ~) from the mm1, so, well, step by step or i'll get crazy :p02:51
steve|mwolfspraul: (pressure) this highly depends.. I used 120 really most of the time, and from one layer to another I would say something like 25 minutes, excluding the scanning02:51
wolfspraulyes, I can imagine. bugs everywhere. but that's why need to move to one common platform asap.02:51
kristianpaulsure :-)02:51
steve|m400 is only used like 30-50 seconds :)02:52
wolfspraulwell I think that's pretty clear already - documented :-)02:52
wolfspraulany pitfalls?02:52
wolfspraulhow much pressure do you apply with the 120 paper, roughly?02:52
wpwrakkristianpaul: any luck with analyzing the data you already collected ?02:52
kristianpaulwpwrak: i'll do my self02:53
kristianpaulwpwrak: well i still waiiting reply from the other guy02:53
kristianpaulso no luck yet :-/02:53
wpwrakkristianpaul: (diy) heh. if you want it done well, you have to do it yourself, it seems ...02:53
wolfspraulsteve|m: oh - nice post about the f3 e-paper! Too bad it's not a blog so I can't add it to the Qi planet.02:54
steve|mpitfalls.. not damaging the board on the borders is hard.. and you should always just remove the copper of one layer until you see the coating of the next layer, otherwise it's hard to tell the copper of 2 layers apart and you may remove too much02:54
wpwrakkristianpaul: pity, though. he ought to know this stuff very well.02:54
wolfspraulsteve|m: with 'borders', you mean aroud the edge of the board?02:55
wolfspraulthe middle is easier/better?02:55
kristianpaulwpwrak: 2 bits ADC sample are not common for him, the first he askme was if i have a usrp ;-)02:55
kristianpaulwpwrak: I'll send him data again as 32 bits signed intenger, so i think i will speedup all procesing that can be done on a usual computer02:57
steve|mwolfspraul: yes, exactly. the best idea to prevent this would be having some other material with the exact same thickness around the pcb, so you're not damaging the edges (which will become round otherwise)02:57
kristianpaulI'm close, i just need solve some segfaults :p02:57
wpwrakkristianpaul: i would think that converting from 2 bits to whatever he normally use should be quite trivial ...02:58
wolfspraulsteve|m: got it - thanks!02:59
kristianpaulwpwrak: also i feel a bit bad, but is not top priority, he is using matlab for analizing the data..02:59
wpwraksteve|m: run a laser rotating around the z axis over it ?02:59
wolfspraulreally nice f3 post btw, really. you are not easily stopped :-)02:59
kristianpaulwpwrak: converting sure thats the idea02:59
steve|mwolfspraul: and which worked quite will is using a few rows of gaffer-tape as an underlayment, since the non-adhesive site is quite resistant to the sandpaper03:00
steve|ms/which/what s/will/well03:00
zrafaxiangfu: he.. now 2.8M.. ;)03:00
kristianpaulzrafa: oh wow !03:01
wolfspraulyou use the gaffer-tape to protect areas that are already sanded deep enough?03:01
zrafakristianpaul: xiangfu's image is smaller and has more tools than mine I thin :)03:01
steve|mwolfspraul: no, I'm just putting the tape on the wodden desk, and the pcb on top.. since the tape doesn't outwear the sandpaper, like e.g. wood would03:02
steve|mand it's quite easy to dispose all the copper dust when renewing it from time to time :)03:02
steve|mmaybe teflon would work good as well ;)03:03
kristianpaulzrafa: Thats openwrt lightness man ! :-)03:04
kristianpaulhe i wonder running OE on my router >-)03:05
wolfsprauloh I see. the tape is protecting your wood desk.03:06
kristianpaulsteve|m: how many layers of tape you need and the end?03:07
steve|mwolfspraul: yes, and it's preventing scratches on the other side of the pcb as well, since it's somewhat soft03:07
steve|mkristianpaul: (layers of tape).. only one, but something like 4 vertical rows next to each other with ~15cm length03:08
kristianpaulsteve|m: hmm, i tought you used the tape to remove copper smothly, i mean adding tape later then etching and so on03:09
wpwrakzrafa: some time ago, i had an 1.44 MB floppy with kernel, rootfs, and the X server :)03:11
wpwrakzrafa: must have been around 1993 :-)03:11
zrafawpwrak: haha.. can you check where is that floppy?03:11
kristianpaulwpwrak: slackware flopy sets?03:11
steve|mkristianpaul: I thought of etching as well - that would be something to evaluate03:11
Action: kristianpaul remenbers when was at school03:12
steve|mwpwrak: nice.. I ran Fli4l for many years off a floppy, but without X obviously :P03:12
steve|m(okay, booting linux off a floppy without X is nothing special..)03:13
kristianpaulwell at that time X needed a separte floppy i think03:13
kristianpaulfirst boot floppy, then rootfs, ... he, some years later i finally understood how all that worked03:14
steve|mwpwrak: good old LILO ;)03:18
wpwrakkristianpaul: naw, self-constructed. just a custom-configured kernel, shell, and the X server. i made it to test the laptop i was about to buy in the shop.03:19
steve|mwow, lilo is at version 23.1 already.. hehe03:20
wpwraksteve|m: don't remember if this floppy had LILO. but yes, it would be likely :)03:21
rjeffriesfrom the shell on Ben what is an easy way to change font bigger for my not-so-young eyes?03:59
rjeffriesjust used nano editor write a one line document saved it and then cat result. one small step04:02
wolfspraulyou need to use setfont/setfont2 commands to change the font, I am hoping one day we can just iterate over all available console fonts with a hotkey, but right now you need to type a command (not sure which one exactly)04:06
rjeffrieswow Ben just burped big time. was in shell, doing nothing, a bunch of stuff flew by04:07
rjeffriesout of memory kill process 454 hotplug2 score 157 or a child04:07
rjeffriesoh well04:07
wpwrakrjeffries: did you    echo 1  >/proc/sys/vm/overcommit_memory    ?04:14
wpwrakrjeffries: without this, the kernel will use an extremely conservative memory allocation strategy, which basically means that something like half of your memory will not be used04:16
rjeffriesi will do that now04:19
wpwrakthe latest openwrt-for-ben should set this by default04:20
rjeffriesok with help from a magnifier I got it done04:24
rjeffriesdoes anyone have a short list of a few programs tuned for this screen size? for example, top is not easy to read ;)\04:25
rjeffriesanyway yes Ben works04:26
tuxbrain_awaywpwrak: I was thinking in use the ICSP header of Arduino board with the UBB, but I'm little confuse, there is only information on how to program the Atmega chip using the that header but not how to read/send data once the program is running, but I see for examble that the Arduino SD module can use both SPI pins or that ICSP header,09:10
tuxbrain_awayso I guess is also posible to read/send data trhough it , using the same SPI libraries?09:10
tuxbrain_awayIf I use the ICSP the pins, will pins 8-13 will be also unusable? Should I ask on arduino forum instead of here?09:10
wolfspraulha, this is great!09:16
wolfspraulmirko: congratulations!09:16
wolfspra1lI added the hackaday.com article to http://en.qi-hardware.com/wiki/What_others_are_saying10:02
wolfspra1lmaybe the title of that page is a little cheesy, if someone has better ideas let me know10:02
tuxbrain_awaywpwrak: ok this http://www.arduino.cc/cgi-bin/yabb2/YaBB.pl?num=1287338644 and this10:32
tuxbrain_awayhttp://www.arduino.cc/cgi-bin/yabb2/YaBB.pl?num=1288255553 had clarified my doubts. ICSP = SPI pins, use same libraries, and if you can't use booth pins at time due are the same, so pins 13,12,11 get's annuled while using ICSP10:32
kyakmirko: i already adapted all qt4 ports to work with TARGET_INCDIRS+=$(STAGING_DIR)/usr/include in qmake.mk :) So if you could review this change and commit it, this would be great!12:25
kyakand "adapted" means that i've thrown away exporting of variables via Build/Prepare as it's not needed anymore12:29
kyakxiangfu: hi! wanted to mention that i reflashed the image couple of days ago, and the bootloader was flashed 2/2 (i.e. i had no errors).12:29
kyakmaybe it's because i have older version of xboot-tools12:30
kyakor whatever it's called12:30
xiangfukyak: hmm... , flash 2011-02-18-testing u-boot. then reflash a old u-boot. then reflash 2011-02-18-testing u-boot again. the error are gone. I have tested three time.12:31
xiangfukyak: don't have time look into it.12:31
kyakok then. Let's say that you've experienced a sporadic error back then :)12:32
wpwraktuxbrain_away: (while using ICSP) well, you usually don't do anything else while flashing. you can still have other stuff connected to the pins, as long as this other stuff doesn't interfere with the signals. e.g., if you keep nSS high, a slave's MISO should be Z, so it wouldn't mess with ICSP's MISO.14:29
wpwraktuxbrain_away: (assuming the AVR is the SPI master)14:32
rjeffriesis this required every time Ben is powerd on?14:49
rjeffriesrjeffries: did you    echo 1  >/proc/sys/vm/overcommit_memory    ?14:49
wolfsprauldefinitely not 'required', it's an optimization14:52
wolfspraulwhich is included by default in the latest testing image14:53
wolfspraulthen we find out over time whether there are any unforeseen negative side-effects, and whether it will stay in as default or not14:53
wpwrakrjeffries: this change is lost when you (re)boot, yes14:54
wpwrakwolfspraul: the own downsides i know of is that, a) if you run close to the memory limits or exceed them, the system gets very slow because it spends a lot of time trying to free memory, and b) it gets a bit less predictable what processes get killed once you run out of memory for good14:56
wpwraks/own/only/  # coffee not kicking in yet14:57
wpwrakxiangfu: btw, after last week's discussion about the slowness of the build process, i tried to see how long it would take on my pc (an elderly quad-core (Q6600) with a lot of RAM). alas, the build didn't get very far, so i'll have to look into this again at a later point in time.15:03
rjeffrieswell the Ben choked when the only thing I had done was open nano editor then close it15:03
wpwrakxiangfu: one thing i noticed was that you listed  jikes  among the dependencies15:03
rjeffriesmaybe there is some other issue i dunno15:03
xiangfuwpwrak: thanks.15:04
wpwrakxiangfu: this seems to be some very old package. e.g., in ubuntu they removed it several releases ago and i found comments stating it was obsolete. what distribution are you using ?15:04
xiangfuwpwrak: I am using ubuntu 10.04, the buildhost is Debian 5.015:05
wpwrakrjeffries: just a nano shouldn't be a problem. well, unless you're editing a really big file :-)15:05
rjeffriesso it urns out that SoB (son of Ben) having more RAM might be A Good THing15:05
wpwrakrjeffries: i haven't read many disagreeing voices in the last year or so :)15:05
rjeffrieswpwrak I had EXITED nano. nothing was going on, so maybe this particular Ben is ill I dunno15:05
xiangfuwpwrak: I don't know what 'jikes' for.  I can not remember add it to dependencies.15:06
rjeffrieswpwrak it is also not at all clear that 64MB would be the right bump. Ben has ok CPU power, is RAM starved15:06
wpwrakxiangfu: hmm ... and you have jikes installed on ubuntu ? i think 10.4 doesn't have it. don't know about debian. it seems that sun java 6 or so may be a replacement for jikes (at lest that's what i found in some comments)15:07
wpwrakrjeffries: you mean you'd want more the 64 MB ?15:07
rjeffriesnot sure more thasn 64MB will be prsctical on Ben due to cost15:08
xiangfuwpwrak: no 'jiles' in ubuntu 10.04.15:08
wpwrakxiangfu: i found the jikes dependency here: http://en.qi-hardware.com/wiki/Building_Software_Image#Prerequisites15:09
xiangfuwpwrak: the build host system have 'jikes' : jikes - Fast Java compiler adhering to language and VM specifications15:09
wpwrakso maybe the dependency doesn't exist anymore. even better ;-)15:09
rjeffriesI have an exciting new Ben accessory: a magnifing glass15:09
wpwrakrjeffries: cost could be an issue, yes. not for the memory size per se but for the industry moving to DDR, so the old SDR stays at a higher price point15:10
wpwrakwolfspraul: a 4740 with a DDR interface would be very cool ;-)15:11
wpwrakrjeffries: would you like a walking stick with this ? ;-)15:11
wpwrakxiangfu: you run full builds on both systems, your ubuntu pc and the debian buildhost, correct ?15:12
rjeffriesI need to send a photo. Maybe tuxbrain will put logo on this handsome magnifier and sell it in his shop15:13
xiangfuwpwrak: recently not. only build minimal in my ubuntu pc.15:13
kristianpaulxiangfu: hi15:14
xiangfukristianpaul: hi15:14
kristianpaulxiangfu: i got a minimal build, whats the way to get single packages compiled?15:14
wpwrakrjeffries: after the "nanowar" edition, he can then make the "forefather" version ;-)15:14
kristianpaulI mean i want re-use the minimal build and just add the packages i want later15:15
xiangfukristianpaul: make package/PACKAGE_NAME/compile15:15
kristianpaulthanks xiangfu15:15
wpwrakxiangfu: okay. so i'll keep you posted when i stumble into anything. thanks !15:15
xiangfukristianpaul: or edit the ".config", change the package from "not set" to "=y"15:16
kristianpaulwpwrak: quadcore?? ha15:16
Action: kristianpaul is happy with his on core and 900Mb ram :-)15:16
xiangfuwpwrak: ok thanks. maybe the package depends on 'jikes' already removed in openwrt :)15:16
kristianpaulwpwrak: in the other side you could be the new buildhost :-)15:17
xiangfukristianpaul: output more info when compile package: make package/plplot/{clean,compile} V=9915:17
kristianpaulxiangfu: Just by curiosirty, how usefull is the upstream image provided from openwrt for the nanonote?15:22
xiangfukristianpaul: it's a like minimal rootfs15:25
kristianpaulBut may allow me to build the qi feed? (gmenu2x, startdic,, etc...) ?15:26
wpwrakkristianpaul: (buildhost) naw, not with my slow link :)15:26
xiangfukristianpaul: yes. just add those feeds to 'feeds.conf'15:26
kristianpaulwpwrak: (slow link) fair point in you favor15:26
kristianpaulhmm now dingux seems to be suffering from manuals obscurity from ingenic :-?15:28
kristianpaullarsc: seems they need the programmer's manual of the JZ4755 SoC15:34
kristianpaulI read from rss but now i cant find the blog entry..15:34
wolfspraulkristianpaul: already solved.15:35
kristianpaulwolfspraul: :-)15:36
kristianpaulxiangfu: i tought mac address in then nanonote was already asigned to the chip...hmm15:38
larsckristianpaul: wich rss? planet.qi-hardware.com?15:39
kristianpaulthat can explain some issues..15:39
kristianpaullarsc: yes, or dingux.com i think15:39
jow_laptopwpwrak: jikes dependency is introduced by sablevm-sdk package in openwrt15:53
jow_laptopI never understood who parted that and for what it is/was useful since this seems to have existed long before any gui efforts in openwrt15:54
wpwrakjow_laptop: does this sound as if the bottom line would be that sablevm may not be so useful to keep around ?15:56
jow_laptopah, its a small java vm15:56
jow_laptopsomehow I always thought about some game engine15:56
jow_laptopscumvm or how it was called15:56
wpwrakdoes anything depend on sablevm ?15:56
kristianpaulkill it! :')15:57
Action: wpwrak pats the headman's axe 15:58
kyakxiangfu: (Ethernet_over_USB#Static_Mac_Address) how i can do this when building an image?15:58
jow_laptopxiangfu: regarding the question where to hook the alsactl restore... module loading happens in /etc/init.d/boot (first initscript). You could do the restore in an /etc/init.d/alsa with START=1116:04
xiangfujow_laptop: oh. yes. thanks for the info16:04
xiangfukyak: the kernel command is in u-boot.  there are two method.16:05
xiangfu1. change u-boot code.16:05
xiangfu2. we have a package name 'nanonote-files' which have a 'mtd.nn' and this package will run "mtd.nn fw_setenv_default" when you first boot your nanonote. so change the content of mtd.nn in nanonote-files package.16:06
kristianpaulkyak: that will be very usefull indeed16:07
jow_laptopxiangfu: did you ever try to use /etc/config/network to manage the usb0 ?16:08
xiangfujow_laptop: only config the ip address , never try config mac address.16:08
jow_laptoptheoretically "option macaddr ..." should work16:08
xiangfujow_laptop: ok. will try that and report. thanks again .16:09
xiangfujow_laptop: there are two mac address of usb0. g_ether.host_mac  g_ether.dev_mac. can we setup those two mac address in /etc/config/network?16:11
mthin Debian I added "hwaddress ether 01:23:45:67:89:AB" in the right interface section in /etc/network/interfaces16:11
mthI think buildroot uses the same network config tools as Debian16:12
qi-bot[commit] Xiangfu Liu: add -q to freedroid http://qi-hw.com/p/gmenu2x/819000416:13
qi-bot[commit] Xiangfu Liu: add hnb icon, add snownews icon http://qi-hw.com/p/gmenu2x/30cd9fd16:13
kyakxiangfu: ok, thanks16:17
qi-bot[commit] Xiangfu Liu: screen, add icon http://qi-hw.com/p/gmenu2x/e9c58f516:21
qi-bot[commit] Xiangfu Liu: byobu, add /var/run/utmp http://qi-hw.com/p/openwrt-packages/a25d55516:24
xiangfukyak: since the u-boot < 1M. we put the u-boot env at offset 1M and 1.5M(backup place). just FYI,16:32
xiangfukyak: my plan is keep bootloader <= 2M. so maybe some day we can using 6M kernel partition :)16:32
kyakor maybe, using three kernels :)16:36
qi-bot[commit] Xiangfu Liu: gmenu2x, update http://qi-hw.com/p/openwrt-packages/4ba1f5016:37
qi-bot[commit] Xiangfu Liu: Revert "byobu, add /var/run/utmp" http://qi-hw.com/p/openwrt-packages/d7b53ed16:44
xiangfukyak: tri-boot. :)17:00
xiangfukyak: need goto sleep. see you17:01
wpwrakrjeffries: RFQ looks good. let's see what crawls out of it :)17:27
rjeffrieswpwrak control your enthusiasm. ;)17:30
rjeffriesor is that "curb your enthusiasm?) (TV show)17:31
rjeffriesthis looks cool. http://dangerousprototypes.com/2011/02/21/qp-state-machine-framework-for-arduino/17:51
rjeffrieskristianpaul your site is not reachable this morning, thanks to Google I can look at cached copy18:23
kristianpaulrjeffries: should be now, yday my isp had troubles18:24
kristianpaulsorry :-)18:24
rjeffriesmaybe your dydnamic dns is hosed?18:25
rjeffriesor need a refresh??18:26
kristianpaulrjeffries: it still not reachable right now?18:26
rjeffrieswill try again now18:27
rjeffriesOops! Google Chrome could not connect to www.kristianpaul.org //after a long delay//18:28
wpwraktuxbrain_away: weren't you interested in a sneak preview ? http://downloads.qi-hardware.com/people/werner/wpan/tmp/atben-110219-compare.jpg   http://downloads.qi-hardware.com/people/werner/wpan/tmp/atben-110219-compare-inserted.jpg18:47
kristianpaulwpwrak: nice !18:50
wpwrakdon't know its transmission characteristics yet, though. ubuntu upgrade broke usrp2_fft :-(19:00
tuxbrain_awaywpwrak: awesome dude!!!19:15
tuxbrain_awaysmaller=less pain :)19:16
tuxbrain_awayThe more I see it the more I think a recover it with plastic can fit19:17
tuxbrain_awayif posible in bright black plastic as NN is19:18
tuxbrain_awayor at least black19:18
tuxbrain_awaywhatever I want one of it, can we do a exchange? 5 UBB by one atben? :P19:20
tuxbrain_awaykristianpaul: don't laugh man, next target is your gps thing :P when a 8:10 version?19:22
Action: tuxbrain_away thinks than negotiate with pcbs with a man that piss tin and create smd with snaps of his finger is a noway deal19:26
kristianpaultuxbrain_away: 8:10 not yet, actually thats the last thing i'll do :-)19:28
tuxbrain_awaykristianpaul: well at least is in the todo list :) , just to focus a gps thing is more usefull in a portable device like NN more than a plugged thing like MM1, I suppose is for the FPGA part you have started with MM1, isn't it?19:30
rohhuh? what does one need an fpga for to recieve gps?19:31
tuxbrain_awayroh:kristianpaul is going beyond to use a gps module, is creating a full free gps stack, catching directly the airwaves and creating the electronic and logic needed to transform it something usable by a computer, ( kristianpaul correctme if I'm wrong)19:34
tuxbrain_awayand I have read well he is using FPGA to do that signal treatment19:35
tuxbrain_awaynext NN wet fantasy: atben+ kristianpaul gps +openmoko navigation expansion all in one 8:10 :)19:36
rohtuxbrain_away: i know. still doesnt need a fpga19:37
tuxbrain_awayok then I'm wrong :P19:37
rohfpga are good for asic development. doesnt need to be done here. whatfor? there are gps demods. and cpus. thats how ALL gps modules are built.19:37
wpwraktuxbrain_away: (exchange) naw, sorry. making the four sets for adam (2), roh (1), and richard (1) is already bad enough. this stuff takes forever, and you get slower with larger volumes :-(19:37
rohregardless if you look at ublox, sirf or broadcom, they all do gps with a hw demod and afaik even all use a arm cpu (mips would also work)19:38
tuxbrain_awaywpwrak: well whatever count with that 5 production UBB :)19:38
rohso.. please explain why we need an expensive, power-hungry, physical gigantous fpga for gps ;)19:39
wpwraktuxbrain_away: (snaps of his finger) hmm, making a set takes 1-2 days ... depending on the rate of mistakes (which goes up rapidly with larger volumes)19:39
rohwpwrak: i can feel your pain ;)19:39
wpwrakroh: thanks :)19:39
kristianpaultuxbrain_away: (catching directly) ergg not so, i use a GPS L1 Fronted for that19:40
rohkristianpaul: which is totally ok. thats what _all_ gps modules do. demodulating yourself wouldnt gain anything19:43
kristianpaulroh: tuxbrain_away fpga can speed up some tracking process,imaging doing fft on the nanonote :'/19:43
rohkristianpaul: still a waste of power.19:43
rohfpga are bad cpu.. stupid to use them mobile19:43
kristianpaulroh: (demodulating yourself wouldnt gain anything) wronng, actually thats a key point for indoor aplications19:43
rohkristianpaul: nah. not even ublox does that.19:44
kristianpaulroh: how do you know? (ublox)19:44
rohkristianpaul: do it like gps. digitize i and q and do math on it.19:44
rohno full sdr, but only the last step of the demod19:44
rohor gsm. (where we use the bitclock and sync on it (while the burst)19:45
rohsdr is always stupid coming from a 'saving space, power, cost' pov19:45
kristianpaulroh: ( fpga are bad cpu) sure, the idea is implement specific things not a cpu maybe basebadn processor19:46
rohkristianpaul: ublox is a soc vendor. fabless. they do 'the math' and the integration of done designs. afaik they used a maxim chip for L1 frontend before integrating it even more19:46
kristianpaulroh: sure, fft is not the only way to correlate and track signals19:47
rohkristianpaul: i still dont get whatfor? the openmoko gta01 had a 'l1 frontend' on a serial and needed to do the math on the main arm cpu in side the binary blob.19:47
rohand it worked (somehow)19:47
kristianpaulroh: (gta01) do you have the specs of the 'l1 frontend' ?19:48
rohi didnt like the binary part. but i dont see why we need more than the small arm7 which the complete rest of the industry uses for their 50 or so korrelators.19:48
rohkristianpaul: not much. we dont have any serial protocol description or so.19:48
kristianpaulroh: (do it like gps. digitize i and q and do math on it) thats the idea19:49
kristianpaulwhy you understood?19:49
rohkristianpaul: it was that hammerhead thingie19:49
rohsomebody even started reversing the serial bitstream http://wiki.openmoko.org/wiki/Hammerhead/Protocol19:50
rohthe company who did the chip and the binary was named 'global locate'19:51
rohwhich was bought up by broadcom while openmoko was working on the gta0119:51
rohwe switched for ublox for the gta02 (freerunner)19:52
kristianpauli dont get this,  that dint seems to be _just_ a 'l1 frontend'19:52
kristianpaulwich ublox?19:52
rohublox was antares4 or so ... its is a different solution (no binary on the main arm cpu, real nmea on serial)19:52
kristianpaulso, what's the fun on reading a nmea?19:53
kristianpaulwich is already posible, but dint help copylegft..19:53
rohublox basically brings its own arm cpu for running their binary. but in the end all gps are the same. its a L1 demod which spits out a serial bitstream consisting of adc values from I and Q.19:53
rohfor openmoko it was imporant to get a working product done. that wasnt possible with hammerhead and broadcom. broadcom wasnt even able to recompile their binary with a eabi compiler19:54
rohalso licensing was madness.19:54
rohthus we went to ublox which solved the problem by blackboxing the binary in its own arm cpu (which is ok by my levels of foss neediness)19:55
rohas long as those blackboxes have open specified interfaces i can work with them. i dont accept them running in the same domain as my free code. so binaries in kernel or userspace i need => i dont buy the product. from my pov i cannot buy hw from nvidia or (too fresh) ati hw. (and i dont. i use a 3 year old thinkpad which is completely foss supported)19:57
rohgllin was like the master example why foss and classic bullshit bingo business dont mix well. it was like a roadblock you couldnt drive around.19:57
kristianpaul_webroh: i'm not saying that mostly gps receiver dint work from correlating i/q signals20:02
kristianpaul_webwich point us back to the fat, that a dedicated hardware is needed for some hard part of the process20:02
rohkristianpaul_web: dedicated hw for doing what? working the i/q samples?20:03
rohof for doing the L1 demod up to i/q samples?20:03
kristianpaul_websure, you can do that part with the nanonote (not couting bandwich issues) but i dont see it tracking at least 4 channels to get a fixed point meawhile you are in the  bus for example20:04
rohi agreee that you need the L1 demod like all others.20:04
wpwrakroh: (nvidia and ati) i have both in my pc. yet still no closed drivers ;-)20:04
rohand the I/q samples cannot be that much data. it was a serial without hw handshake and afaik some low data rate liek 38400 or 5760020:04
rohthey do the gps correlation stuff on a slow arm7 cpus. cannot be magic20:05
kristianpaul_webyes, thats why i pointed tuxbrain_away that some dedicated hardware is required to catch gps signals and get a decent i/q data20:05
kristianpaul_webfor what i now, (with the  fronted i work) is not 57600, (as is not nema ;-))20:06
rohsure. we agree on that. sdr is plain stupid (and expensive) for that. especially since there are proper dedicated filters for the gps stuff20:06
wpwraki wonder how hard the correlation it. may just be a bunch of relatively straightforward state machines. alas, i couldn't find any sort (~4-5 bits) examples of gold codes20:06
rohkristianpaul_web: nmea would work on 4800 baud20:06
kristianpaul_websampling is about 2MSPS for at 4 bits per sample20:06
rohwpwrak: afaik the L1 baseband signal is very few bits 4bit per sample or so20:06
roh4 bit I and 4bit Q per sample or so20:07
rohits not as simple as a state-machine i believe20:08
kristianpaul_webrjeffries: seems other browout at  home or ISP troubles athome :-(20:08
kristianpaul_webfor one channel should not be so much trouble..20:08
rohafaik recent gps solutions do have like 50 parallel correlators20:09
wpwrakroh: you would do the correlation in the analog domain ?20:09
wpwrakroh: well, digitized-analog i mean20:09
rohwpwrak: no. its done inside of arm cpus in sw. afaik20:09
kristianpaul_webroh: are you aware of the freq used to spread gps/gnss signal?20:10
rohkristianpaul_web: i dunno all details (quite widely spattered) but afaik there is only a quite small band to recieve anyhow20:10
rohfor L120:10
roh1.572 ghz or so (from my head)20:11
rohthats also why these small patch antennas are manually tuned20:11
kristianpaul_webwpwrak: namuru use some acumulators in hardware20:11
roh(the 'scratches' on the corners of patch antennas'20:11
kristianpaul_webwpwrak: the akos book mentioned it a bit but  not in depth20:12
wpwrakkristianpaul_web: so how is it done ? reconstruct a bitstream and then try to match one of the gold codes ?20:13
kristianpaul_webwpwrak: actually gps-sdr (softwre project) used some precomputed fft tables20:14
rohneed to run.. bbl20:15
wpwrakso the correlation is done with FFT ? interesting20:15
kristianpaul_webyeah, but i was told also gps-sdr (software project) is not the best example to folllow in this topic20:16
kristianpaul_webwpwrak: namuru i think  uses some algorithm for serial correlation, i mean some series of multiplication and comparisson, after it found something20:17
kristianpaul_webwpwrak: of course this is a FPGA, so you can multiply it  by 3 (for the prediction/tracking part), then 12 channels20:19
rohwhy so few?20:19
kristianpaul_webwell... see in gpredict how many satellites you can get in a skyview (just GPS )20:20
rohsirf also uses arm. seems they got bought by CSR20:21
kristianpaul_webroh: you should be aware (just in case), the chip could havea correlator part and the CPU part (also know as baseband processor)20:22
rohsure. but that would be new to me.20:22
kristianpaul_webbut i may be wrong, i dont remenber all details for all comercial GPS-Receiver, and more for those are documented other just not..20:23
rohi currently need to assume its done in sw when one manuf does it that way (globallocate/brcm) since doing it in hw would mean more cost for other manufs.20:23
rohso. no. i dont know if sirf and or ublox do use hw for that. i just assume they dont since i know its possible.20:24
kristianpaul_webthats the thing !20:26
kristianpaul_webso close-sourced gps are just that...20:26
kristianpaul_webgps receiver*20:26
kristianpaul_websee later i need phone my ISP :-/20:27
qi-bot[commit] David Kühling: gnuplot: explicitely disable all unneeded features to fix build problems http://qi-hw.com/p/openwrt-packages/520487b21:18
qi-bot[commit] David Kühling: config.full_system: fix name of gnuplot package (was renamed a long time ago) http://qi-hw.com/p/openwrt-xburst/c25b5fb21:19
qi-bot[commit] David Kühling: gmenu2x: make console the foreground console before launching applications http://qi-hw.com/p/gmenu2x/530092122:56
--- Tue Feb 22 201100:00

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